YUV conversions. Initial import.

yuv.mod/common.bmx

@@ -0,0 +1,69 @@
+'
+' Copyright (c) 2024 Bruce A Henderson
+' All rights reserved.
+'
+' Redistribution and use in source and binary forms, with or without
+' modification, are permitted provided that the following conditions are met:
+'     * Redistributions of source code must retain the above copyright
+'       notice, this list of conditions and the following disclaimer.
+'     * Redistributions in binary form must reproduce the above copyright
+'       notice, this list of conditions and the following disclaimer in the
+'       documentation and/or other materials provided with the distribution.
+'     * Neither the name of the author nor the
+'       names of its contributors may be used to endorse or promote products
+'       derived from this software without specific prior written permission.
+'
+' THIS SOFTWARE IS PROVIDED BY THE AUTHOR "AS IS" AND ANY
+' EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+' WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+' DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+' DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+' (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+' LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+' ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+' (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+' SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+'
+SuperStrict
+
+Import "source.bmx"
+
+Extern
+
+	Function I420ToARGB:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+	Function I420ToABGR:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+
+	Function J420ToARGB:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+	Function J420ToABGR:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+
+	Function H420ToARGB:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+	Function H420ToABGR:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+
+	Function U420ToARGB:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+	Function U420ToABGR:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+
+	Function I422ToARGB:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+	Function I422ToABGR:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+
+	Function J422ToARGB:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+	Function J422ToABGR:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+
+	Function H422ToARGB:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+	Function H422ToABGR:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+
+	Function U422ToARGB:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+	Function U422ToABGR:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+
+	Function I444ToARGB:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+	Function I444ToABGR:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+
+	Function J444ToARGB:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+	Function J444ToABGR:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+
+	Function H444ToARGB:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+	Function H444ToABGR:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+
+	Function U444ToARGB:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+	Function U444ToABGR:Int(srcY:Byte Ptr, strideY:Int, srcU:Byte Ptr, strideU:Int, srcV:Byte Ptr, strideV:Int, dst:Byte Ptr, dstStride:Int, width:Int, height:Int)
+
+End Extern

yuv.mod/examples/example_01.bmx

@@ -0,0 +1,27 @@
+SuperStrict
+
+Framework SDL.SDLRenderMax2D
+Import Image.YUV
+
+
+Local yuv:Byte[] = LoadByteArray("code_build_play.yuv")
+Local width:Int = 1024
+Local height:Int = 768
+
+
+Local pix:TPixmap = CreatePixmap(width, height, PF_RGBA)
+
+Local yuvPtr:Byte Ptr = yuv
+ConvertI420ToABGR(yuvPtr, width, yuvPtr + width * height, width / 2, yuvPtr + width * height * 5 / 4, width / 2, pix)
+
+Local img:TImage = LoadImage(pix)
+
+Graphics 1024, 768, 0
+
+While Not KeyHit(KEY_ESCAPE)
+	Cls
+
+	DrawImage img, 0, 0
+
+	Flip
+Wend

yuv.mod/libyuv/AUTHORS

@@ -0,0 +1,7 @@
+# Names should be added to this file like so:
+# Name or Organization <email address>
+
+Google Inc.
+
+Ho Cheung <[email protected]>
+Ivan Pavlotskiy <[email protected]>

yuv.mod/libyuv/Android.bp

@@ -0,0 +1,200 @@
+package {
+    default_applicable_licenses: ["external_libyuv_license"],
+}
+
+// Added automatically by a large-scale-change
+//
+// large-scale-change included anything that looked like it might be a license
+// text as a license_text. e.g. LICENSE, NOTICE, COPYING etc.
+//
+// Please consider removing redundant or irrelevant files from 'license_text:'.
+// See: http://go/android-license-faq
+license {
+    name: "external_libyuv_license",
+    visibility: [":__subpackages__"],
+    license_kinds: [
+        "SPDX-license-identifier-BSD",
+    ],
+    license_text: [
+        "LICENSE",
+        "PATENTS",
+    ],
+}
+
+cc_library {
+    name: "libyuv",
+    vendor_available: true,
+    product_available: true,
+    host_supported: true,
+    vndk: {
+        enabled: true,
+    },
+
+    srcs: [
+        "source/compare.cc",
+        "source/compare_common.cc",
+        "source/compare_gcc.cc",
+        "source/compare_msa.cc",
+        "source/compare_neon.cc",
+        "source/compare_neon64.cc",
+        "source/convert.cc",
+        "source/convert_argb.cc",
+        "source/convert_from.cc",
+        "source/convert_from_argb.cc",
+        "source/convert_jpeg.cc",
+        "source/convert_to_argb.cc",
+        "source/convert_to_i420.cc",
+        "source/cpu_id.cc",
+        "source/mjpeg_decoder.cc",
+        "source/mjpeg_validate.cc",
+        "source/planar_functions.cc",
+        "source/rotate.cc",
+        "source/rotate_any.cc",
+        "source/rotate_argb.cc",
+        "source/rotate_common.cc",
+        "source/rotate_gcc.cc",
+        "source/rotate_msa.cc",
+        "source/rotate_neon.cc",
+        "source/rotate_neon64.cc",
+        "source/row_any.cc",
+        "source/row_common.cc",
+        "source/row_gcc.cc",
+        "source/row_msa.cc",
+        "source/row_neon.cc",
+        "source/row_neon64.cc",
+        "source/row_rvv.cc",
+        "source/scale.cc",
+        "source/scale_any.cc",
+        "source/scale_argb.cc",
+        "source/scale_common.cc",
+        "source/scale_gcc.cc",
+        "source/scale_msa.cc",
+        "source/scale_neon.cc",
+        "source/scale_neon64.cc",
+        "source/scale_rgb.cc",
+        "source/scale_rvv.cc",
+        "source/scale_uv.cc",
+        "source/video_common.cc",
+    ],
+
+    cflags: [
+        "-Wall",
+        "-Werror",
+        "-Wno-unused-parameter",
+        "-fexceptions",
+        "-DHAVE_JPEG",
+        "-DLIBYUV_UNLIMITED_DATA",
+    ],
+
+    arch: {
+        arm: {
+            cflags: ["-mfpu=neon"],
+        },
+    },
+
+    shared_libs: ["libjpeg"],
+
+    export_include_dirs: ["include"],
+
+    apex_available: [
+        "//apex_available:platform",
+        "com.android.media.swcodec",
+    ],
+    min_sdk_version: "29",
+}
+
+// compatibilty static library until all uses of libyuv_static are replaced
+// with libyuv (b/37646797)
+cc_library_static {
+    name: "libyuv_static",
+    vendor_available: true,
+    whole_static_libs: ["libyuv"],
+    apex_available: [
+        "//apex_available:platform",
+        "com.android.media.swcodec",
+    ],
+    min_sdk_version: "29",
+}
+
+cc_test {
+    name: "libyuv_unittest",
+    static_libs: ["libyuv"],
+    shared_libs: ["libjpeg"],
+    cflags: ["-Wall", "-Werror"],
+    srcs: [
+        "unit_test/basictypes_test.cc",
+        "unit_test/color_test.cc",
+        "unit_test/compare_test.cc",
+        "unit_test/convert_test.cc",
+        "unit_test/cpu_test.cc",
+        "unit_test/cpu_thread_test.cc",
+        "unit_test/math_test.cc",
+        "unit_test/planar_test.cc",
+        "unit_test/rotate_argb_test.cc",
+        "unit_test/rotate_test.cc",
+        "unit_test/scale_argb_test.cc",
+        "unit_test/scale_plane_test.cc",
+        "unit_test/scale_rgb_test.cc",
+        "unit_test/scale_test.cc",
+        "unit_test/scale_uv_test.cc",
+        "unit_test/unit_test.cc",
+        "unit_test/video_common_test.cc",
+    ],
+}
+
+cc_test {
+    name: "compare",
+    gtest: false,
+    srcs: [
+        "util/compare.cc",
+    ],
+    static_libs: ["libyuv"],
+}
+
+cc_test {
+    name: "cpuid",
+    gtest: false,
+    srcs: [
+        "util/cpuid.c",
+    ],
+    static_libs: ["libyuv"],
+}
+
+cc_test {
+    name: "i444tonv12_eg",
+    gtest: false,
+    srcs: [
+        "util/i444tonv12_eg.cc",
+    ],
+    static_libs: ["libyuv"],
+}
+
+cc_test {
+    name: "psnr",
+    gtest: false,
+    srcs: [
+        "util/psnr_main.cc",
+        "util/psnr.cc",
+        "util/ssim.cc",
+    ],
+    static_libs: ["libyuv"],
+}
+
+cc_test {
+    name: "yuvconstants",
+    gtest: false,
+    srcs: [
+        "util/yuvconstants.c",
+    ],
+    static_libs: ["libyuv"],
+}
+
+cc_test {
+    name: "yuvconvert",
+    gtest: false,
+    srcs: [
+        "util/yuvconvert.cc",
+    ],
+    static_libs: ["libyuv"],
+    shared_libs: ["libjpeg"],
+}

yuv.mod/libyuv/Android.mk

@@ -0,0 +1,110 @@
+# This is the Android makefile for libyuv for NDK.
+LOCAL_PATH:= $(call my-dir)
+
+include $(CLEAR_VARS)
+
+LOCAL_CPP_EXTENSION := .cc
+
+LOCAL_SRC_FILES := \
+    source/compare.cc           \
+    source/compare_common.cc    \
+    source/compare_gcc.cc       \
+    source/compare_msa.cc       \
+    source/compare_neon.cc      \
+    source/compare_neon64.cc    \
+    source/compare_win.cc       \
+    source/convert.cc           \
+    source/convert_argb.cc      \
+    source/convert_from.cc      \
+    source/convert_from_argb.cc \
+    source/convert_to_argb.cc   \
+    source/convert_to_i420.cc   \
+    source/cpu_id.cc            \
+    source/planar_functions.cc  \
+    source/rotate.cc            \
+    source/rotate_any.cc        \
+    source/rotate_argb.cc       \
+    source/rotate_common.cc     \
+    source/rotate_gcc.cc        \
+    source/rotate_msa.cc        \
+    source/rotate_neon.cc       \
+    source/rotate_neon64.cc     \
+    source/rotate_win.cc        \
+    source/row_any.cc           \
+    source/row_common.cc        \
+    source/row_gcc.cc           \
+    source/row_msa.cc           \
+    source/row_neon.cc          \
+    source/row_neon64.cc        \
+    source/row_win.cc           \
+    source/scale.cc             \
+    source/scale_any.cc         \
+    source/scale_argb.cc        \
+    source/scale_common.cc      \
+    source/scale_gcc.cc         \
+    source/scale_msa.cc         \
+    source/scale_neon.cc        \
+    source/scale_neon64.cc      \
+    source/scale_rgb.cc         \
+    source/scale_uv.cc          \
+    source/scale_win.cc         \
+    source/video_common.cc
+
+common_CFLAGS := -Wall -fexceptions
+ifneq ($(LIBYUV_DISABLE_JPEG), "yes")
+LOCAL_SRC_FILES += \
+    source/convert_jpeg.cc      \
+    source/mjpeg_decoder.cc     \
+    source/mjpeg_validate.cc
+common_CFLAGS += -DHAVE_JPEG
+LOCAL_SHARED_LIBRARIES := libjpeg
+endif
+
+LOCAL_CFLAGS += $(common_CFLAGS)
+LOCAL_EXPORT_C_INCLUDES := $(LOCAL_PATH)/include
+LOCAL_C_INCLUDES += $(LOCAL_PATH)/include
+LOCAL_EXPORT_C_INCLUDE_DIRS := $(LOCAL_PATH)/include
+
+LOCAL_MODULE := libyuv_static
+LOCAL_MODULE_TAGS := optional
+
+include $(BUILD_STATIC_LIBRARY)
+
+include $(CLEAR_VARS)
+
+LOCAL_WHOLE_STATIC_LIBRARIES := libyuv_static
+LOCAL_MODULE := libyuv
+ifneq ($(LIBYUV_DISABLE_JPEG), "yes")
+LOCAL_SHARED_LIBRARIES := libjpeg
+endif
+
+include $(BUILD_SHARED_LIBRARY)
+
+include $(CLEAR_VARS)
+LOCAL_STATIC_LIBRARIES := libyuv_static
+LOCAL_SHARED_LIBRARIES := libjpeg
+LOCAL_MODULE_TAGS := tests
+LOCAL_CPP_EXTENSION := .cc
+LOCAL_C_INCLUDES += $(LOCAL_PATH)/include
+LOCAL_SRC_FILES := \
+    unit_test/basictypes_test.cc    \
+    unit_test/color_test.cc         \
+    unit_test/compare_test.cc       \
+    unit_test/convert_argb_test.cc  \
+    unit_test/convert_test.cc       \
+    unit_test/cpu_test.cc           \
+    unit_test/cpu_thread_test.cc    \
+    unit_test/math_test.cc          \
+    unit_test/planar_test.cc        \
+    unit_test/rotate_argb_test.cc   \
+    unit_test/rotate_test.cc        \
+    unit_test/scale_argb_test.cc    \
+    unit_test/scale_plane_test.cc   \
+    unit_test/scale_rgb_test.cc     \
+    unit_test/scale_test.cc         \
+    unit_test/scale_uv_test.cc      \
+    unit_test/unit_test.cc          \
+    unit_test/video_common_test.cc
+
+LOCAL_MODULE := libyuv_unittest
+include $(BUILD_NATIVE_TEST)

yuv.mod/libyuv/BUILD.gn

@@ -0,0 +1,486 @@
+# Copyright 2014 The LibYuv Project Authors. All rights reserved.
+#
+# Use of this source code is governed by a BSD-style license
+# that can be found in the LICENSE file in the root of the source
+# tree. An additional intellectual property rights grant can be found
+# in the file PATENTS. All contributing project authors may
+# be found in the AUTHORS file in the root of the source tree.
+
+import("//build/config/features.gni")
+import("//testing/test.gni")
+import("libyuv.gni")
+
+declare_args() {
+  # Set to false to disable building with absl flags.
+  libyuv_use_absl_flags = true
+
+  # When building a shared library using a target in WebRTC or
+  # Chromium projects that depends on libyuv, setting this flag
+  # to true makes libyuv symbols visible inside that library.
+  libyuv_symbols_visible = false
+}
+
+config("libyuv_config") {
+  include_dirs = [ "include" ]
+  if (is_android) {
+    if (target_cpu == "arm" || target_cpu == "x86" || target_cpu == "mipsel") {
+      ldflags = [ "-Wl,--dynamic-linker,/system/bin/linker" ]
+    } else {
+      ldflags = [ "-Wl,--dynamic-linker,/system/bin/linker64" ]
+    }
+  }
+
+  # Define CHROMIUM to tell cpu_id to avoid sandbox unsafe system calls.
+  defines = [ "CHROMIUM" ]
+  if (!libyuv_use_neon) {
+    defines += [ "LIBYUV_DISABLE_NEON" ]
+  }
+  if (!libyuv_use_sve) {
+    defines += [ "LIBYUV_DISABLE_SVE" ]
+  }
+  if (!libyuv_use_sme) {
+    defines += [ "LIBYUV_DISABLE_SME" ]
+  }
+  if (libyuv_disable_rvv) {
+    defines += [ "LIBYUV_DISABLE_RVV" ]
+  }
+  if (!libyuv_use_lsx) {
+    defines += [ "LIBYUV_DISABLE_LSX" ]
+  }
+  if (!libyuv_use_lasx) {
+    defines += [ "LIBYUV_DISABLE_LASX" ]
+  }
+}
+
+# This target is built when no specific target is specified on the command line.
+group("default") {
+  testonly = true
+  deps = [ ":libyuv" ]
+  if (libyuv_include_tests) {
+    deps += [
+      ":compare",
+      ":cpuid",
+      ":i444tonv12_eg",
+      ":libyuv_unittest",
+      ":psnr",
+      ":yuvconstants",
+      ":yuvconvert",
+    ]
+  }
+}
+
+group("libyuv") {
+  all_dependent_configs = [ ":libyuv_config" ]
+  deps = []
+
+  if (is_win && target_cpu == "x64") {
+    # Compile with clang in order to get inline assembly
+    public_deps = [ ":libyuv_internal(//build/toolchain/win:win_clang_x64)" ]
+  } else {
+    public_deps = [ ":libyuv_internal" ]
+  }
+
+  if (libyuv_use_neon) {
+    deps += [ ":libyuv_neon" ]
+  }
+
+  if (libyuv_use_sve) {
+    deps += [ ":libyuv_sve" ]
+  }
+
+  if (libyuv_use_sme) {
+    deps += [ ":libyuv_sme" ]
+  }
+
+  if (libyuv_use_msa) {
+    deps += [ ":libyuv_msa" ]
+  }
+
+  if (libyuv_use_lsx) {
+    deps += [ ":libyuv_lsx" ]
+  }
+
+  if (libyuv_use_lasx) {
+    deps += [ ":libyuv_lasx" ]
+  }
+
+  if (!is_ios && !libyuv_disable_jpeg) {
+    # Make sure that clients of libyuv link with libjpeg. This can't go in
+    # libyuv_internal because in Windows x64 builds that will generate a clang
+    # build of libjpeg, and we don't want two copies.
+    deps += [ "//third_party:jpeg" ]
+  }
+}
+
+static_library("libyuv_internal") {
+  visibility = [ ":*" ]
+
+  sources = [
+    # Headers
+    "include/libyuv.h",
+    "include/libyuv/basic_types.h",
+    "include/libyuv/compare.h",
+    "include/libyuv/compare_row.h",
+    "include/libyuv/convert.h",
+    "include/libyuv/convert_argb.h",
+    "include/libyuv/convert_from.h",
+    "include/libyuv/convert_from_argb.h",
+    "include/libyuv/cpu_id.h",
+    "include/libyuv/loongson_intrinsics.h",
+    "include/libyuv/macros_msa.h",
+    "include/libyuv/mjpeg_decoder.h",
+    "include/libyuv/planar_functions.h",
+    "include/libyuv/rotate.h",
+    "include/libyuv/rotate_argb.h",
+    "include/libyuv/rotate_row.h",
+    "include/libyuv/row.h",
+    "include/libyuv/scale.h",
+    "include/libyuv/scale_argb.h",
+    "include/libyuv/scale_rgb.h",
+    "include/libyuv/scale_row.h",
+    "include/libyuv/scale_uv.h",
+    "include/libyuv/version.h",
+    "include/libyuv/video_common.h",
+
+    # Source Files
+    "source/compare.cc",
+    "source/compare_common.cc",
+    "source/compare_gcc.cc",
+    "source/compare_win.cc",
+    "source/convert.cc",
+    "source/convert_argb.cc",
+    "source/convert_from.cc",
+    "source/convert_from_argb.cc",
+    "source/convert_jpeg.cc",
+    "source/convert_to_argb.cc",
+    "source/convert_to_i420.cc",
+    "source/cpu_id.cc",
+    "source/mjpeg_decoder.cc",
+    "source/mjpeg_validate.cc",
+    "source/planar_functions.cc",
+    "source/rotate.cc",
+    "source/rotate_any.cc",
+    "source/rotate_argb.cc",
+    "source/rotate_common.cc",
+    "source/rotate_gcc.cc",
+    "source/rotate_win.cc",
+    "source/row_any.cc",
+    "source/row_common.cc",
+    "source/row_gcc.cc",
+    "source/row_rvv.cc",
+    "source/row_win.cc",
+    "source/scale.cc",
+    "source/scale_any.cc",
+    "source/scale_argb.cc",
+    "source/scale_common.cc",
+    "source/scale_gcc.cc",
+    "source/scale_rgb.cc",
+    "source/scale_rvv.cc",
+    "source/scale_uv.cc",
+    "source/scale_win.cc",
+    "source/video_common.cc",
+  ]
+
+  configs += [ ":libyuv_config" ]
+  defines = []
+  deps = []
+
+  if (libyuv_symbols_visible) {
+    configs -= [ "//build/config/gcc:symbol_visibility_hidden" ]
+    configs += [ "//build/config/gcc:symbol_visibility_default" ]
+  }
+
+  if ((!is_ios || use_blink) && !libyuv_disable_jpeg) {
+    defines += [ "HAVE_JPEG" ]
+
+    # Needed to pull in libjpeg headers. Can't add //third_party:jpeg to deps
+    # because in Windows x64 build it will get compiled with clang.
+    deps += [ "//third_party:jpeg_includes" ]
+  }
+
+  # Always enable optimization for Release and NaCl builds (to workaround
+  # crbug.com/538243).
+  if (!is_debug || is_nacl) {
+    configs -= [ "//build/config/compiler:default_optimization" ]
+
+    # Enable optimize for speed (-O2) over size (-Os).
+    configs += [ "//build/config/compiler:optimize_max" ]
+  }
+
+  # To enable AVX2 or other cpu optimization, pass flag here
+  if (!is_win) {
+    cflags = [
+      # "-mpopcnt",
+      # "-mavx2",
+      # "-mfma",
+      "-ffp-contract=fast",  # Enable fma vectorization for NEON.
+    ]
+  }
+}
+
+if (libyuv_use_neon) {
+  static_library("libyuv_neon") {
+    sources = [
+      # ARM Source Files
+      "source/compare_neon.cc",
+      "source/compare_neon64.cc",
+      "source/rotate_neon.cc",
+      "source/rotate_neon64.cc",
+      "source/row_neon.cc",
+      "source/row_neon64.cc",
+      "source/scale_neon.cc",
+      "source/scale_neon64.cc",
+    ]
+
+    deps = [ ":libyuv_internal" ]
+
+    public_configs = [ ":libyuv_config" ]
+
+    if (current_cpu == "arm64") {
+      cflags = [ "-march=armv8-a+dotprod+i8mm" ]
+    } else {
+      configs -= [ "//build/config/compiler:compiler_arm_fpu" ]
+      cflags = [ "-mfpu=neon" ]
+    }
+  }
+}
+
+if (libyuv_use_sve) {
+  static_library("libyuv_sve") {
+    sources = [ "source/row_sve.cc" ]
+
+    deps = [ ":libyuv_internal" ]
+
+    public_configs = [ ":libyuv_config" ]
+
+    # SVE2 is an Armv9-A feature.
+    cflags = [ "-march=armv9-a+sve2" ]
+  }
+}
+
+if (libyuv_use_sme) {
+  static_library("libyuv_sme") {
+    sources = [ "source/rotate_sme.cc" ]
+
+    deps = [ ":libyuv_internal" ]
+
+    public_configs = [ ":libyuv_config" ]
+
+    # SME is an Armv9-A feature.
+    cflags = [ "-march=armv9-a+sme" ]
+  }
+}
+
+if (libyuv_use_msa) {
+  static_library("libyuv_msa") {
+    sources = [
+      # MSA Source Files
+      "source/compare_msa.cc",
+      "source/rotate_msa.cc",
+      "source/row_msa.cc",
+      "source/scale_msa.cc",
+    ]
+
+    deps = [ ":libyuv_internal" ]
+
+    public_configs = [ ":libyuv_config" ]
+  }
+}
+
+if (libyuv_use_lsx) {
+  static_library("libyuv_lsx") {
+    sources = [
+      # LSX Source Files
+      "source/rotate_lsx.cc",
+      "source/row_lsx.cc",
+      "source/scale_lsx.cc",
+    ]
+
+    cflags_cc = [
+      "-mlsx",
+      "-Wno-c++11-narrowing",
+    ]
+
+    deps = [ ":libyuv_internal" ]
+
+    public_configs = [ ":libyuv_config" ]
+  }
+}
+
+if (libyuv_use_lasx) {
+  static_library("libyuv_lasx") {
+    sources = [
+      # LASX Source Files
+      "source/row_lasx.cc",
+    ]
+
+    cflags_cc = [
+      "-mlasx",
+      "-Wno-c++11-narrowing",
+    ]
+
+    deps = [ ":libyuv_internal" ]
+
+    public_configs = [ ":libyuv_config" ]
+  }
+}
+
+if (libyuv_include_tests) {
+  config("libyuv_unittest_warnings_config") {
+    if (!is_win) {
+      cflags = [
+        # TODO(fbarchard): Fix sign and unused variable warnings.
+        "-Wno-sign-compare",
+        "-Wno-unused-variable",
+      ]
+    }
+    if (is_win) {
+      cflags = [
+        "/wd4245",  # signed/unsigned mismatch
+        "/wd4189",  # local variable is initialized but not referenced
+      ]
+    }
+  }
+  config("libyuv_unittest_config") {
+    defines = [ "GTEST_RELATIVE_PATH" ]
+  }
+
+  test("libyuv_unittest") {
+    testonly = true
+
+    sources = [
+      "unit_test/basictypes_test.cc",
+      "unit_test/color_test.cc",
+      "unit_test/compare_test.cc",
+      "unit_test/convert_argb_test.cc",
+      "unit_test/convert_test.cc",
+      "unit_test/cpu_test.cc",
+      "unit_test/cpu_thread_test.cc",
+      "unit_test/math_test.cc",
+      "unit_test/planar_test.cc",
+      "unit_test/rotate_argb_test.cc",
+      "unit_test/rotate_test.cc",
+      "unit_test/scale_argb_test.cc",
+      "unit_test/scale_plane_test.cc",
+      "unit_test/scale_rgb_test.cc",
+      "unit_test/scale_test.cc",
+      "unit_test/scale_uv_test.cc",
+      "unit_test/unit_test.cc",
+      "unit_test/unit_test.h",
+      "unit_test/video_common_test.cc",
+    ]
+
+    deps = [
+      ":libyuv",
+      "//testing/gtest",
+    ]
+
+    defines = []
+    if (libyuv_use_absl_flags) {
+      defines += [ "LIBYUV_USE_ABSL_FLAGS" ]
+      deps += [
+        "//third_party/abseil-cpp/absl/flags:flag",
+        "//third_party/abseil-cpp/absl/flags:parse",
+      ]
+    }
+
+    configs += [ ":libyuv_unittest_warnings_config" ]
+
+    public_deps = [ "//testing/gtest" ]
+    public_configs = [ ":libyuv_unittest_config" ]
+
+    if (is_linux || is_chromeos) {
+      cflags = [ "-fexceptions" ]
+    }
+    if (is_ios) {
+      configs -= [ "//build/config/compiler:default_symbols" ]
+      configs += [ "//build/config/compiler:symbols" ]
+      cflags = [ "-Wno-sometimes-uninitialized" ]
+    }
+    if (!is_ios && !libyuv_disable_jpeg) {
+      defines += [ "HAVE_JPEG" ]
+    }
+    if (is_android) {
+      deps += [ "//testing/android/native_test:native_test_native_code" ]
+    }
+
+    # TODO(YangZhang): These lines can be removed when high accuracy
+    # YUV to RGB to Neon is ported.
+    if ((target_cpu == "armv7" || target_cpu == "armv7s" ||
+         (target_cpu == "arm" && arm_version >= 7) || target_cpu == "arm64") &&
+        (arm_use_neon || arm_optionally_use_neon)) {
+      defines += [ "LIBYUV_NEON" ]
+    }
+
+    defines += [
+      # Enable the following 3 macros to turn off assembly for specified CPU.
+      # "LIBYUV_DISABLE_X86",
+      # "LIBYUV_DISABLE_NEON",
+      # Enable the following macro to build libyuv as a shared library (dll).
+      # "LIBYUV_USING_SHARED_LIBRARY"
+    ]
+  }
+
+  executable("compare") {
+    sources = [
+      # sources
+      "util/compare.cc",
+    ]
+    deps = [ ":libyuv" ]
+    if (is_linux || is_chromeos) {
+      cflags = [ "-fexceptions" ]
+    }
+  }
+
+  executable("yuvconvert") {
+    sources = [
+      # sources
+      "util/yuvconvert.cc",
+    ]
+    deps = [ ":libyuv" ]
+    if (is_linux || is_chromeos) {
+      cflags = [ "-fexceptions" ]
+    }
+  }
+
+  executable("yuvconstants") {
+    sources = [
+      # sources
+      "util/yuvconstants.c",
+    ]
+    deps = [ ":libyuv" ]
+    if (is_linux || is_chromeos) {
+      cflags = [ "-fexceptions" ]
+    }
+  }
+
+  executable("psnr") {
+    sources = [
+      # sources
+      "util/psnr.cc",
+      "util/psnr_main.cc",
+      "util/ssim.cc",
+    ]
+    deps = [ ":libyuv" ]
+
+    if (!is_ios && !libyuv_disable_jpeg) {
+      defines = [ "HAVE_JPEG" ]
+    }
+  }
+
+  executable("i444tonv12_eg") {
+    sources = [
+      # sources
+      "util/i444tonv12_eg.cc",
+    ]
+    deps = [ ":libyuv" ]
+  }
+
+  executable("cpuid") {
+    sources = [
+      # sources
+      "util/cpuid.c",
+    ]
+    deps = [ ":libyuv" ]
+  }
+}

yuv.mod/libyuv/CM_linux_packages.cmake

@@ -0,0 +1,69 @@
+# determine the version number from the #define in libyuv/version.h
+EXECUTE_PROCESS (
+	COMMAND grep --perl-regex --only-matching "(?<=LIBYUV_VERSION )[0-9]+" include/libyuv/version.h
+	WORKING_DIRECTORY ${PROJECT_SOURCE_DIR}
+	OUTPUT_VARIABLE YUV_VERSION_NUMBER
+	OUTPUT_STRIP_TRAILING_WHITESPACE )
+SET ( YUV_VER_MAJOR 0 )
+SET ( YUV_VER_MINOR 0 )
+SET ( YUV_VER_PATCH ${YUV_VERSION_NUMBER} )
+SET ( YUV_VERSION ${YUV_VER_MAJOR}.${YUV_VER_MINOR}.${YUV_VER_PATCH} )
+MESSAGE ( VERBOSE "Building ver.: ${YUV_VERSION}" )
+
+# is this a 32-bit or 64-bit build?
+IF ( CMAKE_SIZEOF_VOID_P EQUAL 8 )
+	SET ( YUV_BIT_SIZE 64 )
+ELSEIF ( CMAKE_SIZEOF_VOID_P EQUAL 4 )
+	SET ( YUV_BIT_SIZE 32 )
+ELSE ()
+	MESSAGE ( FATAL_ERROR "CMAKE_SIZEOF_VOID_P=${CMAKE_SIZEOF_VOID_P}" )
+ENDIF ()
+
+# detect if this is a ARM build
+STRING (FIND "${CMAKE_CXX_COMPILER}" "arm-linux-gnueabihf-g++" pos)
+IF ( ${pos} EQUAL -1 )
+	SET ( YUV_CROSS_COMPILE_FOR_ARM7 FALSE )
+ELSE ()
+	MESSAGE ( "Cross compiling for ARM7" )
+	SET ( YUV_CROSS_COMPILE_FOR_ARM7 TRUE )
+ENDIF ()
+STRING (FIND "${CMAKE_SYSTEM_PROCESSOR}" "arm" pos)
+IF ( ${pos} EQUAL -1 )
+	SET ( YUV_COMPILE_FOR_ARM7 FALSE )
+ELSE ()
+	MESSAGE ( "Compiling for ARM" )
+	SET ( YUV_COMPILE_FOR_ARM7 TRUE )
+ENDIF ()
+
+# setup the sytem name, such as "x86-32", "amd-64", and "arm-32
+IF ( ${YUV_CROSS_COMPILE_FOR_ARM7} OR ${YUV_COMPILE_FOR_ARM7} )
+	SET ( YUV_SYSTEM_NAME "armhf-${YUV_BIT_SIZE}" )
+ELSE ()
+	IF ( YUV_BIT_SIZE EQUAL 32 )
+		SET ( YUV_SYSTEM_NAME "x86-${YUV_BIT_SIZE}" )
+	ELSE ()
+		SET ( YUV_SYSTEM_NAME "amd-${YUV_BIT_SIZE}" )
+	ENDIF ()
+ENDIF ()
+MESSAGE ( VERBOSE "Packaging for: ${YUV_SYSTEM_NAME}" )
+
+# define all the variables needed by CPack to create .deb and .rpm packages
+SET ( CPACK_PACKAGE_VENDOR					"Frank Barchard" )
+SET ( CPACK_PACKAGE_CONTACT					"[email protected]" )
+SET ( CPACK_PACKAGE_VERSION					${YUV_VERSION} )
+SET ( CPACK_PACKAGE_VERSION_MAJOR			${YUV_VER_MAJOR} )
+SET ( CPACK_PACKAGE_VERSION_MINOR			${YUV_VER_MINOR} )
+SET ( CPACK_PACKAGE_VERSION_PATCH			${YUV_VER_PATCH} )
+SET ( CPACK_RESOURCE_FILE_LICENSE			${PROJECT_SOURCE_DIR}/LICENSE )
+SET ( CPACK_SYSTEM_NAME						"linux-${YUV_SYSTEM_NAME}" )
+SET ( CPACK_PACKAGE_NAME					"libyuv" )
+SET ( CPACK_PACKAGE_DESCRIPTION_SUMMARY		"YUV library" )
+SET ( CPACK_PACKAGE_DESCRIPTION				"YUV library and YUV conversion tool" )
+SET ( CPACK_DEBIAN_PACKAGE_SECTION			"other" )
+SET ( CPACK_DEBIAN_PACKAGE_PRIORITY			"optional" )
+SET ( CPACK_DEBIAN_PACKAGE_MAINTAINER		"Frank Barchard <[email protected]>" )
+SET ( CPACK_GENERATOR						"DEB;RPM" )
+
+# create the .deb and .rpm files (you'll need build-essential and rpm tools)
+INCLUDE( CPack )
+

yuv.mod/libyuv/CMakeLists.txt

@@ -0,0 +1,221 @@
+# CMakeLists for libyuv
+# Originally created for "roxlu build system" to compile libyuv on windows
+# Run with -DTEST=ON to build unit tests
+
+include(CheckCSourceCompiles)
+
+project ( YUV C CXX )	# "C" is required even for C++ projects
+cmake_minimum_required( VERSION 2.8.12 )
+option( UNIT_TEST "Built unit tests" OFF )
+
+set ( ly_base_dir	${PROJECT_SOURCE_DIR} )
+set ( ly_src_dir	${ly_base_dir}/source )
+set ( ly_inc_dir	${ly_base_dir}/include )
+set ( ly_tst_dir	${ly_base_dir}/unit_test )
+set ( ly_lib_name	yuv )
+set ( ly_lib_static	${ly_lib_name} )
+set ( ly_lib_shared	${ly_lib_name}_shared )
+
+# We cannot use GLOB here since we want to be able to separate out files that
+# need particular flags to enable architecture extensions like AArch64's SVE.
+# TODO: More of these files could be separated out for different architectures.
+set ( ly_common_source_files
+  ${ly_src_dir}/compare.cc
+  ${ly_src_dir}/compare_common.cc
+  ${ly_src_dir}/compare_gcc.cc
+  ${ly_src_dir}/compare_msa.cc
+  ${ly_src_dir}/compare_win.cc
+  ${ly_src_dir}/convert_argb.cc
+  ${ly_src_dir}/convert.cc
+  ${ly_src_dir}/convert_from_argb.cc
+  ${ly_src_dir}/convert_from.cc
+  ${ly_src_dir}/convert_jpeg.cc
+  ${ly_src_dir}/convert_to_argb.cc
+  ${ly_src_dir}/convert_to_i420.cc
+  ${ly_src_dir}/cpu_id.cc
+  ${ly_src_dir}/mjpeg_decoder.cc
+  ${ly_src_dir}/mjpeg_validate.cc
+  ${ly_src_dir}/planar_functions.cc
+  ${ly_src_dir}/rotate_any.cc
+  ${ly_src_dir}/rotate_argb.cc
+  ${ly_src_dir}/rotate.cc
+  ${ly_src_dir}/rotate_common.cc
+  ${ly_src_dir}/rotate_gcc.cc
+  ${ly_src_dir}/rotate_lsx.cc
+  ${ly_src_dir}/rotate_msa.cc
+  ${ly_src_dir}/rotate_win.cc
+  ${ly_src_dir}/row_any.cc
+  ${ly_src_dir}/row_common.cc
+  ${ly_src_dir}/row_gcc.cc
+  ${ly_src_dir}/row_lasx.cc
+  ${ly_src_dir}/row_lsx.cc
+  ${ly_src_dir}/row_msa.cc
+  ${ly_src_dir}/row_rvv.cc
+  ${ly_src_dir}/row_win.cc
+  ${ly_src_dir}/scale_any.cc
+  ${ly_src_dir}/scale_argb.cc
+  ${ly_src_dir}/scale.cc
+  ${ly_src_dir}/scale_common.cc
+  ${ly_src_dir}/scale_gcc.cc
+  ${ly_src_dir}/scale_lsx.cc
+  ${ly_src_dir}/scale_msa.cc
+  ${ly_src_dir}/scale_rgb.cc
+  ${ly_src_dir}/scale_rvv.cc
+  ${ly_src_dir}/scale_uv.cc
+  ${ly_src_dir}/scale_win.cc
+  ${ly_src_dir}/video_common.cc)
+
+file ( GLOB_RECURSE	ly_unittest_sources ${ly_tst_dir}/*.cc )
+list ( SORT			ly_unittest_sources )
+
+include_directories( BEFORE ${ly_inc_dir} )
+
+if(MSVC)
+  add_definitions ( -D_CRT_SECURE_NO_WARNINGS )
+endif()
+
+# Need to set PIC to allow creating shared libraries from object file libraries.
+set(CMAKE_POSITION_INDEPENDENT_CODE ON)
+
+# Build the set of objects that do not need to be compiled with flags to enable
+# particular architecture features.
+add_library( ${ly_lib_name}_common_objects OBJECT ${ly_common_source_files} )
+set(ly_lib_parts $<TARGET_OBJECTS:${ly_lib_name}_common_objects>)
+
+if(NOT MSVC)
+  string(TOLOWER "${CMAKE_SYSTEM_PROCESSOR}" arch_lowercase)
+
+  if(arch_lowercase MATCHES "^arm" AND NOT arch_lowercase STREQUAL "arm64")
+    # Enable Arm Neon kernels.
+    add_definitions(-DLIBYUV_NEON=1)
+    add_library(${ly_lib_name}_neon OBJECT
+      ${ly_src_dir}/compare_neon.cc
+      ${ly_src_dir}/rotate_neon.cc
+      ${ly_src_dir}/row_neon.cc
+      ${ly_src_dir}/scale_neon.cc)
+    target_compile_options(${ly_lib_name}_neon PRIVATE -mfpu=neon)
+    list(APPEND ly_lib_parts $<TARGET_OBJECTS:${ly_lib_name}_neon>)
+  endif()
+
+  if(arch_lowercase STREQUAL "aarch64" OR arch_lowercase STREQUAL "arm64")
+    # Enable AArch64 Neon dot-product and i8mm kernels.
+    add_library(${ly_lib_name}_neon64 OBJECT
+      ${ly_src_dir}/compare_neon64.cc
+      ${ly_src_dir}/rotate_neon64.cc
+      ${ly_src_dir}/row_neon64.cc
+      ${ly_src_dir}/scale_neon64.cc)
+    target_compile_options(${ly_lib_name}_neon64 PRIVATE -march=armv8-a+dotprod+i8mm)
+    list(APPEND ly_lib_parts $<TARGET_OBJECTS:${ly_lib_name}_neon64>)
+
+    # Enable AArch64 SVE kernels.
+    add_library(${ly_lib_name}_sve OBJECT
+      ${ly_src_dir}/row_sve.cc)
+    target_compile_options(${ly_lib_name}_sve PRIVATE -march=armv9-a+sve2)
+    list(APPEND ly_lib_parts $<TARGET_OBJECTS:${ly_lib_name}_sve>)
+
+    set(OLD_CMAKE_REQURED_FLAGS ${CMAKE_REQUIRED_FLAGS})
+    set(OLD_CMAKE_TRY_COMPILE_TARGET_TYPE ${CMAKE_TRY_COMPILE_TARGET_TYPE})
+    set(CMAKE_REQUIRED_FLAGS "${CMAKE_REQUIRED_FLAGS} -march=armv9-a+sme")
+    set(CMAKE_TRY_COMPILE_TARGET_TYPE STATIC_LIBRARY)
+    # Check whether the compiler can compile SME functions; this fails
+    # with Clang for Windows.
+    check_c_source_compiles("
+__arm_locally_streaming void func(void) { }
+int main(void) { return 0; }
+    " CAN_COMPILE_SME)
+    set(CMAKE_REQUIRED_FLAGS ${OLD_CMAKE_REQURED_FLAGS})
+    set(CMAKE_TRY_COMPILE_TARGET_TYPE ${OLD_CMAKE_TRY_COMPILE_TARGET_TYPE})
+
+    if (CAN_COMPILE_SME)
+      # Enable AArch64 SME kernels.
+      add_library(${ly_lib_name}_sme OBJECT
+        ${ly_src_dir}/rotate_sme.cc)
+      target_compile_options(${ly_lib_name}_sme PRIVATE -march=armv9-a+sme)
+      list(APPEND ly_lib_parts $<TARGET_OBJECTS:${ly_lib_name}_sme>)
+    else()
+      add_definitions(-DLIBYUV_DISABLE_SME)
+    endif()
+  endif()
+endif()
+
+# this creates the static library (.a)
+add_library( ${ly_lib_static} STATIC ${ly_lib_parts})
+
+# this creates the shared library (.so)
+add_library( ${ly_lib_shared} SHARED ${ly_lib_parts})
+set_target_properties( ${ly_lib_shared} PROPERTIES OUTPUT_NAME "${ly_lib_name}" )
+set_target_properties( ${ly_lib_shared} PROPERTIES PREFIX "lib" )
+if(WIN32)
+  set_target_properties( ${ly_lib_shared} PROPERTIES IMPORT_PREFIX "lib" )
+endif()
+
+# this creates the cpuid tool
+add_executable      ( cpuid ${ly_base_dir}/util/cpuid.c )
+target_link_libraries  ( cpuid ${ly_lib_static} )
+
+# this creates the conversion tool
+add_executable			( yuvconvert ${ly_base_dir}/util/yuvconvert.cc )
+target_link_libraries	( yuvconvert ${ly_lib_static} )
+
+# this creates the yuvconstants tool
+add_executable      ( yuvconstants ${ly_base_dir}/util/yuvconstants.c )
+target_link_libraries  ( yuvconstants ${ly_lib_static} )
+
+find_package ( JPEG )
+if (JPEG_FOUND)
+  include_directories( ${JPEG_INCLUDE_DIR} )
+  target_link_libraries( ${ly_lib_shared} ${JPEG_LIBRARY} )
+  add_definitions( -DHAVE_JPEG )
+endif()
+
+if(UNIT_TEST)
+  find_library(GTEST_LIBRARY gtest)
+  if(GTEST_LIBRARY STREQUAL "GTEST_LIBRARY-NOTFOUND")
+    set(GTEST_SRC_DIR_DEFAULT /usr/src/gtest)
+    if (CMAKE_CROSSCOMPILING)
+      set(GTEST_SRC_DIR_DEFAULT ${CMAKE_SOURCE_DIR}/third_party/googletest/src/googletest)
+    endif()
+    set(GTEST_SRC_DIR ${GTEST_SRC_DIR_DEFAULT} CACHE STRING "Location of gtest sources")
+    if(EXISTS ${GTEST_SRC_DIR}/src/gtest-all.cc)
+      message(STATUS "building gtest from sources in ${GTEST_SRC_DIR}")
+      set(gtest_sources ${GTEST_SRC_DIR}/src/gtest-all.cc)
+      add_library(gtest STATIC ${gtest_sources})
+      include_directories(${GTEST_SRC_DIR})
+      include_directories(${GTEST_SRC_DIR}/include)
+      set(GTEST_LIBRARY gtest)
+    else()
+      message(FATAL_ERROR "UNIT_TEST is set but unable to find gtest library")
+    endif()
+  endif()
+
+  add_executable(libyuv_unittest ${ly_unittest_sources})
+  target_link_libraries(libyuv_unittest ${ly_lib_name} ${GTEST_LIBRARY})
+  find_library(PTHREAD_LIBRARY pthread)
+  if(NOT PTHREAD_LIBRARY STREQUAL "PTHREAD_LIBRARY-NOTFOUND")
+    target_link_libraries(libyuv_unittest pthread)
+  endif()
+  if (JPEG_FOUND)
+    target_link_libraries(libyuv_unittest ${JPEG_LIBRARY})
+  endif()
+
+  if(NACL AND NACL_LIBC STREQUAL "newlib")
+    target_link_libraries(libyuv_unittest glibc-compat)
+  endif()
+
+  find_library(GFLAGS_LIBRARY gflags)
+  if(NOT GFLAGS_LIBRARY STREQUAL "GFLAGS_LIBRARY-NOTFOUND")
+    target_link_libraries(libyuv_unittest gflags)
+    add_definitions(-DLIBYUV_USE_GFLAGS)
+  endif()
+endif()
+
+
+# install the conversion tool, .so, .a, and all the header files
+install ( PROGRAMS ${CMAKE_BINARY_DIR}/yuvconvert			DESTINATION bin )
+install ( TARGETS ${ly_lib_static}						DESTINATION lib )
+install ( TARGETS ${ly_lib_shared} LIBRARY				DESTINATION lib RUNTIME DESTINATION bin )
+install ( DIRECTORY ${PROJECT_SOURCE_DIR}/include/		DESTINATION include )
+
+# create the .deb and .rpm packages using cpack
+include ( CM_linux_packages.cmake )
+

yuv.mod/libyuv/DEPS

@@ -0,0 +1,2566 @@
+gclient_gn_args_file = 'src/build/config/gclient_args.gni'
+gclient_gn_args = [
+  'generate_location_tags',
+]
+
+vars = {
+  'chromium_git': 'https://chromium.googlesource.com',
+  'chromium_revision': 'af3d01376bec75a68f90160bfd38057d60510a2b',
+  'gn_version': 'git_revision:fae280eabe5d31accc53100137459ece19a7a295',
+  # ninja CIPD package version.
+  # https://chrome-infra-packages.appspot.com/p/infra/3pp/tools/ninja
+  'ninja_version': 'version:[email protected]',
+  # reclient CIPD package version
+  'reclient_version': 're_client_version:0.110.0.43ec6b1-gomaip',
+
+  # Keep the Chromium default of generating location tags.
+  'generate_location_tags': True,
+
+  # By default, download the fuchsia sdk from the public sdk directory.
+  'fuchsia_sdk_cipd_prefix': 'fuchsia/sdk/core/',
+  'fuchsia_version': 'version:15.20230909.2.1',
+  # By default, download the fuchsia images from the fuchsia GCS bucket.
+  'fuchsia_images_bucket': 'fuchsia',
+  'checkout_fuchsia': False,
+  # Since the images are hundreds of MB, default to only downloading the image
+  # most commonly useful for developers. Bots and developers that need to use
+  # other images can override this with additional images.
+  'checkout_fuchsia_boot_images': "terminal.qemu-x64,terminal.x64",
+  'checkout_fuchsia_product_bundles': '"{checkout_fuchsia_boot_images}" != ""',
+}
+
+deps = {
+  'src/build':
+    Var('chromium_git') + '/chromium/src/build' + '@' + '5885d3c24833ad72845a52a1b913a2b8bc651b56',
+  'src/buildtools':
+    Var('chromium_git') + '/chromium/src/buildtools' + '@' + '79ab87fa54614258c4c95891e873223371194525',
+  'src/testing':
+    Var('chromium_git') + '/chromium/src/testing' + '@' + '51e9a02297057cc0e917763a51e16680b7d16fb6',
+  'src/third_party':
+    Var('chromium_git') + '/chromium/src/third_party' + '@' + '2dc4b18abd1003ce7b1eda509dc96f12d49a9667',
+
+  'src/buildtools/linux64': {
+    'packages': [
+      {
+        'package': 'gn/gn/linux-${{arch}}',
+        'version': Var('gn_version'),
+      }
+    ],
+    'dep_type': 'cipd',
+    'condition': 'host_os == "linux"',
+  },
+
+  'src/buildtools/mac': {
+    'packages': [
+      {
+        'package': 'gn/gn/mac-${{arch}}',
+        'version': Var('gn_version'),
+      }
+    ],
+    'dep_type': 'cipd',
+    'condition': 'host_os == "mac"',
+  },
+
+  'src/buildtools/win': {
+    'packages': [
+      {
+        'package': 'gn/gn/windows-amd64',
+        'version': Var('gn_version'),
+      }
+    ],
+    'dep_type': 'cipd',
+    'condition': 'host_os == "win"',
+  },
+
+  'src/buildtools/reclient': {
+    'packages': [
+      {
+        'package': 'infra/rbe/client/${{platform}}',
+        'version': Var('reclient_version'),
+      }
+    ],
+    'dep_type': 'cipd',
+  },
+
+  'src/third_party/catapult':
+    Var('chromium_git') + '/catapult.git' + '@' + 'fa05d995e152efdae488a2aeba397cd609fdbc9d',
+  'src/third_party/clang-format/script':
+      Var('chromium_git') + '/external/github.com/llvm/llvm-project/clang/tools/clang-format.git' + '@' + 'f97059df7f8b205064625cdb5f97b56668a125ef',
+  'src/third_party/colorama/src':
+    Var('chromium_git') + '/external/colorama.git' + '@' + '3de9f013df4b470069d03d250224062e8cf15c49',
+  'src/third_party/cpu_features/src': {
+    'url': Var('chromium_git') + '/external/github.com/google/cpu_features.git' + '@' + '936b9ab5515dead115606559502e3864958f7f6e',
+    'condition': 'checkout_android',
+  },
+  'src/third_party/depot_tools':
+    Var('chromium_git') + '/chromium/tools/depot_tools.git' + '@' + 'd3e43dd4319ba169c0aaf44547eecf861f2fe5da',
+  'src/third_party/freetype/src':
+    Var('chromium_git') + '/chromium/src/third_party/freetype2.git' + '@' + '9e3c5d7e183c1a8d5ed8868d7d28ef18d3ec9ec8',
+  'third_party/fuchsia-gn-sdk': {
+    'url': Var('chromium_git') + '/chromium/src/third_party/fuchsia-gn-sdk.git' + '@' + '0d6902558d92fe3d49ba9a8f638ddea829be595b',
+    'condition': 'checkout_fuchsia',
+  },
+  'src/third_party/googletest/src':
+    Var('chromium_git') + '/external/github.com/google/googletest.git' + '@' + 'af29db7ec28d6df1c7f0f745186884091e602e07',
+  'src/third_party/harfbuzz-ng/src':
+    Var('chromium_git') + '/external/github.com/harfbuzz/harfbuzz.git' + '@' + 'db700b5670d9475cc8ed4880cc9447b232c5e432',
+  'src/third_party/libc++/src':
+    Var('chromium_git') + '/external/github.com/llvm/llvm-project/libcxx.git' + '@' + '84fb809dd6dae36d556dc0bb702c6cc2ce9d4b80',
+  'src/third_party/libc++abi/src':
+    Var('chromium_git') + '/external/github.com/llvm/llvm-project/libcxxabi.git' + '@' + '8d21803b9076b16d46c32e2f10da191ee758520c',
+  'src/third_party/libunwind/src':
+    Var('chromium_git') + '/external/github.com/llvm/llvm-project/libunwind.git' + '@' + 'f1c687e0aaf0d70b9a53a150e9be5cb63af9215f',
+  'src/third_party/libjpeg_turbo':
+    Var('chromium_git') + '/chromium/deps/libjpeg_turbo.git' + '@' + '30bdb85e302ecfc52593636b2f44af438e05e784',
+  'src/third_party/nasm':
+    Var('chromium_git') + '/chromium/deps/nasm.git' + '@' + '7fc833e889d1afda72c06220e5bed8fb43b2e5ce',
+  'src/tools':
+    Var('chromium_git') + '/chromium/src/tools' + '@' + 'a76c0dbb64c603a0d45e0c6dfae3a351b6e1adf1',
+
+  # libyuv-only dependencies (not present in Chromium).
+  'src/third_party/gtest-parallel':
+    Var('chromium_git') + '/external/webrtc/deps/third_party/gtest-parallel' + '@' + '1dad0e9f6d82ff994130b529d7d814b40eb32b0e',
+
+  'src/third_party/lss': {
+    'url': Var('chromium_git') + '/linux-syscall-support.git' + '@' + 'ce877209e11aa69dcfffbd53ef90ea1d07136521',
+    'condition': 'checkout_android or checkout_linux',
+  },
+
+  # Android deps:
+  'src/third_party/accessibility_test_framework': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/accessibility-test-framework',
+              'version': 'b5ec1e56e58e56bc1a0c77d43111c37f9b512c8a',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/kotlin_stdlib': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/kotlin_stdlib',
+              'version': 'Z1gsqhL967kFQecxKrRwXHbl-vwQjpv0l7PMUZ0EVO8C',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/kotlinc/current': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/kotlinc',
+              'version': 'Rr02Gf2EkaeSs3EhSUHhPqDHSd1AzimrM6cRYUJCPjQC',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/boringssl/src':
+    'https://boringssl.googlesource.com/boringssl.git' + '@' + '20a06474c0b4a16779311bfe98ba69dc2402101d',
+  'src/base': {
+    'url': Var('chromium_git') + '/chromium/src/base' + '@' + 'd407b7061bce341bb6e11b539ea86c46c949ac4c',
+    'condition': 'checkout_android',
+  },
+  'src/third_party/bazel': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/bazel',
+              'version': 'VjMsf48QUWw8n7XtJP2AuSjIGmbQeYdWdwyxVvIRLmAC',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/bouncycastle': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/bouncycastle',
+              'version': 'c078e87552ba26e776566fdaf0f22cd8712743d0',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/android_toolchain': {
+      'packages': [
+            {
+                'package': 'chromium/third_party/android_toolchain/android_toolchain',
+                'version': 'R_8suM8m0oHbZ1awdxGXvKEFpAOETscbfZxkkMthyk8C',
+            },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/androidx': {
+    'packages': [
+      {
+          'package': 'chromium/third_party/androidx',
+          'version': 'y7rF_rx56mD3FGhMiqnlbQ6HOqHJ95xUFNX1m-_a988C',
+      },
+    ],
+    'condition': 'checkout_android',
+    'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_support_test_runner': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_support_test_runner',
+              'version': '96d4bf848cd210fdcbca6bcc8c1b4b39cbd93141',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/android_sdk/public': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_sdk/public/build-tools/34.0.0',
+              'version': 'YK9Rzw3fDzMHVzatNN6VlyoD_81amLZpN1AbmkdOd6AC',
+          },
+          {
+              'package': 'chromium/third_party/android_sdk/public/emulator',
+              'version': '9lGp8nTUCRRWGMnI_96HcKfzjnxEJKUcfvfwmA3wXNkC',
+          },
+          {
+              'package': 'chromium/third_party/android_sdk/public/patcher',
+              'version': 'I6FNMhrXlpB-E1lOhMlvld7xt9lBVNOO83KIluXDyA0C',
+          },
+          {
+              'package': 'chromium/third_party/android_sdk/public/platform-tools',
+              'version': 'HWVsGs2HCKgSVv41FsOcsfJbNcB0UFiNrF6Tc4yRArYC',
+          },
+          {
+              'package': 'chromium/third_party/android_sdk/public/platforms/android-34',
+              'version': 'u-bhWbTME6u-DjypTgr3ZikCyeAeU6txkR9ET6Uudc8C',
+          },
+   {
+              'package': 'chromium/third_party/android_sdk/public/platforms/android-tiramisuprivacysandbox',
+              'version': 'YWMYkzyxGBgVsty0GhXL1oxbY0pGXQIgFc0Rh7ZMRPYC',
+          },
+          {
+              'package': 'chromium/third_party/android_sdk/public/sources/android-31',
+              'version': '_a_BcnANjPYw5mSKlNHa7GFY8yc1kdqj2rmQgac7yUcC',
+          },
+          {
+              'package': 'chromium/third_party/android_sdk/public/cmdline-tools',
+              'version': 'EWnL2r7oV5GtE9Ef7GyohyFam42wtMtEKYU4dCb3U1YC',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/tools/clang/dsymutil': {
+    'packages': [
+      {
+        'package': 'chromium/llvm-build-tools/dsymutil',
+        'version': 'OWlhXkmj18li3yhJk59Kmjbc5KdgLh56TwCd1qBdzlIC',
+      }
+    ],
+    'condition': 'checkout_mac',
+    'dep_type': 'cipd',
+  },
+  'src/third_party/android_build_tools/aapt2': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_build_tools/aapt2',
+              'version': 'STY0BXlZxsEhudnlXQFed-B5UpwehcoM0sYqor6qRqsC',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/byte_buddy': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/byte_buddy',
+              'version': 'c9b53316603fc2d997c899c7ca1707f809b918cd',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/byte_buddy/android_sdk_build_tools_25_0_2': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_sdk/public/build-tools',
+              'version': 'kwIs2vdfTm93yEP8LG5aSnchN4BVEdVxbqQtF4XpPdkC',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/ced/src': {
+    'url': Var('chromium_git') + '/external/github.com/google/compact_enc_det.git' + '@' + 'ba412eaaacd3186085babcd901679a48863c7dd5',
+    'condition': 'checkout_android',
+  },
+  'src/third_party/errorprone/lib': {
+      'url': Var('chromium_git') + '/chromium/third_party/errorprone.git' + '@' + '980d49e839aa4984015efed34b0134d4b2c9b6d7',
+      'condition': 'checkout_android',
+  },
+  'src/third_party/findbugs': {
+    'url': Var('chromium_git') + '/chromium/deps/findbugs.git' + '@' + '4275d9ac8610db6b1bc9a5e887f97e41b33fac67',
+    'condition': 'checkout_android',
+  },
+  'src/third_party/gson': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/gson',
+              'version': '681931c9778045903a0ed59856ce2dd8dd7bf7ca',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/guava': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/guava',
+              'version': 'a6fba501f3a0de88b9be1daa2052632de5b96a46',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/hamcrest': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/hamcrest',
+              'version': '37eccfc658fe79695d6abb6dd497463c4372032f',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/icu': {
+    'url': Var('chromium_git') + '/chromium/deps/icu.git' + '@' + 'e8c3bc9ea97d4423ad0515e5f1c064f486dae8b1',
+  },
+  'src/third_party/icu4j': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/icu4j',
+              'version': 'e87e5bed2b4935913ee26a3ebd0b723ee2344354',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/intellij': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/intellij',
+              'version': '77c2721b024b36ee073402c08e6d8428c0295336',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/jdk': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/jdk',
+              'version': 'GCFtf5t6M4HlrHj6NXedHbpHp2xjgognF8ptNci4478C',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/jsr-305/src': {
+    'url': Var('chromium_git') + '/external/jsr-305.git' + '@' + '642c508235471f7220af6d5df2d3210e3bfc0919',
+    'condition': 'checkout_android',
+  },
+  'src/third_party/junit/src': {
+    'url': Var('chromium_git') + '/external/junit.git' + '@' + '05fe2a64f59127c02135be22f416e91260d6ede6',
+    'condition': 'checkout_android',
+  },
+  'src/third_party/libunwindstack': {
+      'url': Var('chromium_git') + '/chromium/src/third_party/libunwindstack.git' + '@' + '4dbfa0e8c844c8e243b297bc185e54a99ff94f9e',
+      'condition': 'checkout_android',
+  },
+  'src/third_party/ninja': {
+    'packages': [
+      {
+        'package': 'infra/3pp/tools/ninja/${{platform}}',
+        'version': Var('ninja_version'),
+      }
+    ],
+    'dep_type': 'cipd',
+  },
+  'src/third_party/mockito/src': {
+    'url': Var('chromium_git') + '/external/mockito/mockito.git' + '@' + '7c3641bcef717ffa7d765f2c86b847d0aab1aac9',
+    'condition': 'checkout_android',
+  },
+  'src/third_party/objenesis': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/objenesis',
+              'version': 'tknDblENYi8IaJYyD6tUahUyHYZlzJ_Y74_QZSz4DpIC',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/ow2_asm': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/ow2_asm',
+              'version': 'NNAhdJzMdnutUVqfSJm5v0tVazA9l3Dd6CRwH6N4Q5kC',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/r8': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/r8',
+              'version': 'O1BBWiBTIeNUcraX8STMtQXVaCleu6SJJjWCcnfhPLkC',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  # This duplication is intentional, so we avoid updating the r8.jar used by
+  # dexing unless necessary, since each update invalidates all incremental
+  # dexing and unnecessarily slows down all bots.
+  'src/third_party/r8/d8': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/r8',
+              'version': 'vw5kLlW3-suSlCKSO9OQpFWpR8oDnvQ8k1RgKNUapQYC',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/proguard': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/proguard',
+              'version': 'Fd91BJFVlmiO6c46YMTsdy7n2f5Sk2hVVGlzPLvqZPsC',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/requests/src': {
+    'url': Var('chromium_git') + '/external/github.com/kennethreitz/requests.git' + '@' + 'c7e0fc087ceeadb8b4c84a0953a422c474093d6d',
+    'condition': 'checkout_android',
+  },
+  'src/third_party/robolectric': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/robolectric',
+              'version': 'hzetqh1qFI32FOgQroZvGcGdomrgVBJ6WKRnl1KFw6EC',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/sqlite4java': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/sqlite4java',
+              'version': 'LofjKH9dgXIAJhRYCPQlMFywSwxYimrfDeBmaHc-Z5EC',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/turbine': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/turbine',
+              'version': '2I2Nz480QsuCxpQ1lMfbigX8l5HAhX3_ykWU4TKRGo4C',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+  'src/third_party/ub-uiautomator/lib': {
+    'url': Var('chromium_git') + '/chromium/third_party/ub-uiautomator.git' + '@' + '00270549ce3161ae72ceb24712618ea28b4f9434',
+    'condition': 'checkout_android',
+  },
+
+  # iOS deps:
+  'src/ios': {
+    'url': Var('chromium_git') + '/chromium/src/ios' + '@' + 'ddd58e86cf4ebdc0db60a5d0f3c323de49bb295c',
+    'condition': 'checkout_ios'
+  },
+
+  # Everything coming after this is automatically updated by the auto-roller.
+  # === ANDROID_DEPS Generated Code Start ===
+  # Generated by //third_party/android_deps/fetch_all.py
+  'src/third_party/android_deps/libs/android_arch_core_common': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/android_arch_core_common',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/android_arch_core_runtime': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/android_arch_core_runtime',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/android_arch_lifecycle_common': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/android_arch_lifecycle_common',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/android_arch_lifecycle_common_java8': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/android_arch_lifecycle_common_java8',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/android_arch_lifecycle_livedata': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/android_arch_lifecycle_livedata',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/android_arch_lifecycle_livedata_core': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/android_arch_lifecycle_livedata_core',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/android_arch_lifecycle_runtime': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/android_arch_lifecycle_runtime',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/android_arch_lifecycle_viewmodel': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/android_arch_lifecycle_viewmodel',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_animated_vector_drawable': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_animated_vector_drawable',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_appcompat_v7': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_appcompat_v7',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_asynclayoutinflater': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_asynclayoutinflater',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_cardview_v7': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_cardview_v7',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_collections': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_collections',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_coordinatorlayout': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_coordinatorlayout',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_cursoradapter': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_cursoradapter',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_customview': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_customview',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_design': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_design',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_documentfile': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_documentfile',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_drawerlayout': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_drawerlayout',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_interpolator': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_interpolator',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_loader': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_loader',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_localbroadcastmanager': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_localbroadcastmanager',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_multidex': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_multidex',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_print': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_print',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_recyclerview_v7': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_recyclerview_v7',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_slidingpanelayout': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_slidingpanelayout',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_support_annotations': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_support_annotations',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_support_compat': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_support_compat',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_support_core_ui': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_support_core_ui',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_support_core_utils': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_support_core_utils',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_support_fragment': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_support_fragment',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_support_media_compat': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_support_media_compat',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_support_v4': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_support_v4',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_support_vector_drawable': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_support_vector_drawable',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_swiperefreshlayout': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_swiperefreshlayout',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_transition': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_transition',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_versionedparcelable': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_versionedparcelable',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_support_viewpager': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_support_viewpager',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_tools_common': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_tools_common',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_tools_layoutlib_layoutlib_api': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_tools_layoutlib_layoutlib_api',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_android_tools_sdk_common': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_android_tools_sdk_common',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_github_ben_manes_caffeine_caffeine': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_github_ben_manes_caffeine_caffeine',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_github_kevinstern_software_and_algorithms': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_github_kevinstern_software_and_algorithms',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_annotations': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_annotations',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_apps_common_testing_accessibility_framework_accessibility_test_framework': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_apps_common_testing_accessibility_framework_accessibility_test_framework',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_datatransport_transport_api': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_datatransport_transport_api',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_auth': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_auth',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_auth_api_phone': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_auth_api_phone',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_auth_base': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_auth_base',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_base': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_base',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_basement': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_basement',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_cast': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_cast',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_cast_framework': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_cast_framework',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_clearcut': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_clearcut',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_cloud_messaging': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_cloud_messaging',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_flags': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_flags',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_gcm': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_gcm',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_iid': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_iid',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_instantapps': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_instantapps',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_location': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_location',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_phenotype': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_phenotype',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_places_placereport': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_places_placereport',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_stats': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_stats',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_tasks': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_tasks',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_vision': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_vision',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_gms_play_services_vision_common': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_gms_play_services_vision_common',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_material_material': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_material_material',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_play_core_common': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_play_core_common',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_android_play_feature_delivery': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_android_play_feature_delivery',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_auto_auto_common': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_auto_auto_common',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_auto_service_auto_service': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_auto_service_auto_service',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_auto_service_auto_service_annotations': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_auto_service_auto_service_annotations',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_auto_value_auto_value_annotations': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_auto_value_auto_value_annotations',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_code_findbugs_jsr305': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_code_findbugs_jsr305',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_code_gson_gson': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_code_gson_gson',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_dagger_dagger': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_dagger_dagger',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_dagger_dagger_compiler': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_dagger_dagger_compiler',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_dagger_dagger_producers': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_dagger_dagger_producers',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_dagger_dagger_spi': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_dagger_dagger_spi',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_errorprone_error_prone_annotation': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_errorprone_error_prone_annotation',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_errorprone_error_prone_annotations': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_errorprone_error_prone_annotations',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_errorprone_error_prone_check_api': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_errorprone_error_prone_check_api',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_errorprone_error_prone_core': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_errorprone_error_prone_core',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_errorprone_error_prone_type_annotations': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_errorprone_error_prone_type_annotations',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_errorprone_javac': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_errorprone_javac',
+              'version': 'version:2@9+181-r4173-1.cr1',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_errorprone_javac_shaded': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_errorprone_javac_shaded',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_firebase_firebase_annotations': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_firebase_firebase_annotations',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_firebase_firebase_common': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_firebase_firebase_common',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_firebase_firebase_components': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_firebase_firebase_components',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_firebase_firebase_encoders': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_firebase_firebase_encoders',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_firebase_firebase_encoders_json': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_firebase_firebase_encoders_json',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_firebase_firebase_iid': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_firebase_firebase_iid',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_firebase_firebase_iid_interop': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_firebase_firebase_iid_interop',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_firebase_firebase_installations': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_firebase_firebase_installations',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_firebase_firebase_installations_interop': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_firebase_firebase_installations_interop',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_firebase_firebase_measurement_connector': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_firebase_firebase_measurement_connector',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_firebase_firebase_messaging': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_firebase_firebase_messaging',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_googlejavaformat_google_java_format': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_googlejavaformat_google_java_format',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_guava_failureaccess': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_guava_failureaccess',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_guava_guava': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_guava_guava',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_guava_guava_android': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_guava_guava_android',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_guava_listenablefuture': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_guava_listenablefuture',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_j2objc_j2objc_annotations': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_j2objc_j2objc_annotations',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_protobuf_protobuf_java': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_protobuf_protobuf_java',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_google_protobuf_protobuf_javalite': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_google_protobuf_protobuf_javalite',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_googlecode_java_diff_utils_diffutils': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_googlecode_java_diff_utils_diffutils',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_squareup_javapoet': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_squareup_javapoet',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_squareup_javawriter': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_squareup_javawriter',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_squareup_okio_okio_jvm': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_squareup_okio_okio_jvm',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/com_squareup_wire_wire_runtime_jvm': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/com_squareup_wire_wire_runtime_jvm',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/io_github_java_diff_utils_java_diff_utils': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/io_github_java_diff_utils_java_diff_utils',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/io_grpc_grpc_api': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/io_grpc_grpc_api',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/io_grpc_grpc_binder': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/io_grpc_grpc_binder',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/io_grpc_grpc_context': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/io_grpc_grpc_context',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/io_grpc_grpc_core': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/io_grpc_grpc_core',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/io_grpc_grpc_protobuf_lite': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/io_grpc_grpc_protobuf_lite',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/io_grpc_grpc_stub': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/io_grpc_grpc_stub',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/io_perfmark_perfmark_api': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/io_perfmark_perfmark_api',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/javax_annotation_javax_annotation_api': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/javax_annotation_javax_annotation_api',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/javax_annotation_jsr250_api': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/javax_annotation_jsr250_api',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/javax_inject_javax_inject': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/javax_inject_javax_inject',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/net_bytebuddy_byte_buddy': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/net_bytebuddy_byte_buddy',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/net_bytebuddy_byte_buddy_agent': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/net_bytebuddy_byte_buddy_agent',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/net_ltgt_gradle_incap_incap': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/net_ltgt_gradle_incap_incap',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_bouncycastle_bcprov_jdk18on': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_bouncycastle_bcprov_jdk18on',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_ccil_cowan_tagsoup_tagsoup': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_ccil_cowan_tagsoup_tagsoup',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_checkerframework_checker_compat_qual': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_checkerframework_checker_compat_qual',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_checkerframework_checker_qual': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_checkerframework_checker_qual',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_checkerframework_checker_util': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_checkerframework_checker_util',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_checkerframework_dataflow_errorprone': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_checkerframework_dataflow_errorprone',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_codehaus_mojo_animal_sniffer_annotations': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_codehaus_mojo_animal_sniffer_annotations',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_conscrypt_conscrypt_openjdk_uber': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_conscrypt_conscrypt_openjdk_uber',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_eclipse_jgit_org_eclipse_jgit': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_eclipse_jgit_org_eclipse_jgit',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_hamcrest_hamcrest': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_hamcrest_hamcrest',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_jetbrains_kotlin_kotlin_stdlib_jdk7': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_jetbrains_kotlin_kotlin_stdlib_jdk7',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_jetbrains_kotlin_kotlin_stdlib_jdk8': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_jetbrains_kotlin_kotlin_stdlib_jdk8',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_jetbrains_kotlinx_kotlinx_coroutines_android': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_jetbrains_kotlinx_kotlinx_coroutines_android',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_jetbrains_kotlinx_kotlinx_coroutines_core_jvm': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_jetbrains_kotlinx_kotlinx_coroutines_core_jvm',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_jetbrains_kotlinx_kotlinx_coroutines_guava': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_jetbrains_kotlinx_kotlinx_coroutines_guava',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_jetbrains_kotlinx_kotlinx_metadata_jvm': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_jetbrains_kotlinx_kotlinx_metadata_jvm',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_jsoup_jsoup': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_jsoup_jsoup',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_mockito_mockito_android': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_mockito_mockito_android',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_mockito_mockito_core': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_mockito_mockito_core',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_mockito_mockito_subclass': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_mockito_mockito_subclass',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_objenesis_objenesis': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_objenesis_objenesis',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_ow2_asm_asm': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_ow2_asm_asm',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_ow2_asm_asm_analysis': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_ow2_asm_asm_analysis',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_ow2_asm_asm_commons': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_ow2_asm_asm_commons',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_ow2_asm_asm_tree': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_ow2_asm_asm_tree',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_ow2_asm_asm_util': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_ow2_asm_asm_util',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_pcollections_pcollections': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_pcollections_pcollections',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_annotations': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_annotations',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_junit': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_junit',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_nativeruntime': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_nativeruntime',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_nativeruntime_dist_compat': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_nativeruntime_dist_compat',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_pluginapi': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_pluginapi',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_plugins_maven_dependency_resolver': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_plugins_maven_dependency_resolver',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_resources': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_resources',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_robolectric': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_robolectric',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_sandbox': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_sandbox',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_shadowapi': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_shadowapi',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_shadows_framework': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_shadows_framework',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_shadows_playservices': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_shadows_playservices',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_utils': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_utils',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  'src/third_party/android_deps/libs/org_robolectric_utils_reflector': {
+      'packages': [
+          {
+              'package': 'chromium/third_party/android_deps/libs/org_robolectric_utils_reflector',
+              'version': 'version:[email protected]',
+          },
+      ],
+      'condition': 'checkout_android',
+      'dep_type': 'cipd',
+  },
+
+  # === ANDROID_DEPS Generated Code End ===
+}
+
+hooks = [
+  {
+    # This clobbers when necessary (based on get_landmines.py). It should be
+    # an early hook but it will need to be run after syncing Chromium and
+    # setting up the links, so the script actually exists.
+    'name': 'landmines',
+    'pattern': '.',
+    'action': [
+        'python3',
+        'src/build/landmines.py',
+        '--landmine-scripts',
+        'src/tools_libyuv/get_landmines.py',
+        '--src-dir',
+        'src',
+    ],
+  },
+  # Downloads the current stable linux sysroot to build/linux/ if needed.
+  {
+    'name': 'sysroot_arm',
+    'pattern': '.',
+    'condition': 'checkout_linux and checkout_arm',
+    'action': ['python3', 'src/build/linux/sysroot_scripts/install-sysroot.py',
+               '--arch=arm'],
+  },
+  {
+    'name': 'sysroot_arm64',
+    'pattern': '.',
+    'condition': 'checkout_linux and checkout_arm64',
+    'action': ['python3', 'src/build/linux/sysroot_scripts/install-sysroot.py',
+               '--arch=arm64'],
+  },
+  {
+    'name': 'sysroot_x86',
+    'pattern': '.',
+    'condition': 'checkout_linux and (checkout_x86 or checkout_x64)',
+    'action': ['python3', 'src/build/linux/sysroot_scripts/install-sysroot.py',
+               '--arch=x86'],
+  },
+  {
+    'name': 'sysroot_mips',
+    'pattern': '.',
+    'condition': 'checkout_linux and checkout_mips',
+    'action': ['python3', 'src/build/linux/sysroot_scripts/install-sysroot.py',
+               '--arch=mips'],
+  },
+  {
+    'name': 'sysroot_x64',
+    'pattern': '.',
+    'condition': 'checkout_linux and checkout_x64',
+    'action': ['python3', 'src/build/linux/sysroot_scripts/install-sysroot.py',
+               '--arch=x64'],
+  },
+  {
+    # Update the Windows toolchain if necessary.
+    'name': 'win_toolchain',
+    'pattern': '.',
+    'action': ['python3', 'src/build/vs_toolchain.py', 'update'],
+  },
+  {
+    # Update the Mac toolchain if necessary.
+    'name': 'mac_toolchain',
+    'pattern': '.',
+    'action': ['python3', 'src/build/mac_toolchain.py'],
+    'condition': 'checkout_mac',
+  },
+  {
+    'name': 'msan_chained_origins_focal',
+    'pattern': '.',
+    'condition': 'checkout_instrumented_libraries',
+    'action': [ 'python3',
+                'src/third_party/depot_tools/download_from_google_storage.py',
+                '--no_resume',
+                '--no_auth',
+                '--bucket', 'chromium-instrumented-libraries',
+                '-s', 'src/third_party/instrumented_libraries/binaries/msan-chained-origins-focal.tgz.sha1',
+              ],
+  },
+  {
+    'name': 'msan_no_origins_focal',
+    'pattern': '.',
+    'condition': 'checkout_instrumented_libraries',
+    'action': [ 'python3',
+                'src/third_party/depot_tools/download_from_google_storage.py',
+                '--no_resume',
+                '--no_auth',
+                '--bucket', 'chromium-instrumented-libraries',
+                '-s', 'src/third_party/instrumented_libraries/binaries/msan-no-origins-focal.tgz.sha1',
+              ],
+  },
+  {
+    'name': 'msan_chained_origins_focal',
+    'pattern': '.',
+    'condition': 'checkout_instrumented_libraries',
+    'action': [ 'python3',
+                'src/third_party/depot_tools/download_from_google_storage.py',
+                '--no_resume',
+                '--no_auth',
+                '--bucket', 'chromium-instrumented-libraries',
+                '-s', 'src/third_party/instrumented_libraries/binaries/msan-chained-origins-focal.tgz.sha1',
+              ],
+  },
+  {
+    'name': 'msan_no_origins_focal',
+    'pattern': '.',
+    'condition': 'checkout_instrumented_libraries',
+    'action': [ 'python3',
+                'src/third_party/depot_tools/download_from_google_storage.py',
+                '--no_resume',
+                '--no_auth',
+                '--bucket', 'chromium-instrumented-libraries',
+                '-s', 'src/third_party/instrumented_libraries/binaries/msan-no-origins-focal.tgz.sha1',
+              ],
+  },
+  {
+    'name': 'Download Fuchsia SDK from GCS',
+    'pattern': '.',
+    'condition': 'checkout_fuchsia',
+    'action': [
+      'python3',
+      'src/build/fuchsia/update_sdk.py',
+      '--cipd-prefix={fuchsia_sdk_cipd_prefix}',
+      '--version={fuchsia_version}',
+    ],
+  },
+  {
+    'name': 'Download Fuchsia system images',
+    'pattern': '.',
+    'condition': 'checkout_fuchsia and checkout_fuchsia_product_bundles',
+    'action': [
+      'python3',
+      'src/build/fuchsia/update_product_bundles.py',
+      '{checkout_fuchsia_boot_images}',
+    ],
+  },
+  {
+    # Pull clang if needed or requested via GYP_DEFINES.
+    # Note: On Win, this should run after win_toolchain, as it may use it.
+    'name': 'clang',
+    'pattern': '.',
+    'action': ['python3', 'src/tools/clang/scripts/update.py'],
+  },
+  {
+    # Update LASTCHANGE.
+    'name': 'lastchange',
+    'pattern': '.',
+    'action': ['python3', 'src/build/util/lastchange.py',
+               '-o', 'src/build/util/LASTCHANGE'],
+  },
+  # Pull clang-format binaries using checked-in hashes.
+  {
+    'name': 'clang_format_win',
+    'pattern': '.',
+    'condition': 'host_os == "win"',
+    'action': [ 'python3',
+                'src/third_party/depot_tools/download_from_google_storage.py',
+                '--no_resume',
+                '--platform=win32',
+                '--no_auth',
+                '--bucket', 'chromium-clang-format',
+                '-s', 'src/buildtools/win/clang-format.exe.sha1',
+    ],
+  },
+  {
+    'name': 'clang_format_mac_x64',
+    'pattern': '.',
+    'condition': 'host_os == "mac" and host_cpu == "x64"',
+    'action': [ 'python3',
+                'src/third_party/depot_tools/download_from_google_storage.py',
+                '--no_resume',
+                '--platform=darwin',
+                '--no_auth',
+                '--bucket', 'chromium-clang-format',
+                '-s', 'src/buildtools/mac/clang-format.x64.sha1',
+                '-o', 'src/buildtools/mac/clang-format',
+    ],
+  },
+  {
+    'name': 'clang_format_mac_arm64',
+    'pattern': '.',
+    'condition': 'host_os == "mac" and host_cpu == "arm64"',
+    'action': [ 'python3',
+                'src/third_party/depot_tools/download_from_google_storage.py',
+                '--no_resume',
+                '--no_auth',
+                '--bucket', 'chromium-clang-format',
+                '-s', 'src/buildtools/mac/clang-format.arm64.sha1',
+                '-o', 'src/buildtools/mac/clang-format',
+     ],
+  },
+  {
+    'name': 'clang_format_linux',
+    'pattern': '.',
+    'condition': 'host_os == "linux"',
+    'action': [ 'python3',
+                'src/third_party/depot_tools/download_from_google_storage.py',
+                '--no_resume',
+                '--platform=linux*',
+                '--no_auth',
+                '--bucket', 'chromium-clang-format',
+                '-s', 'src/buildtools/linux64/clang-format.sha1',
+    ],
+  },
+  # Pull luci-go binaries (isolate, swarming) using checked-in hashes.
+  {
+    'name': 'luci-go_win',
+    'pattern': '.',
+    'action': [ 'download_from_google_storage',
+                '--no_resume',
+                '--platform=win32',
+                '--no_auth',
+                '--bucket', 'chromium-luci',
+                '-d', 'src/tools/luci-go/win64',
+    ],
+  },
+  {
+    'name': 'luci-go_mac',
+    'pattern': '.',
+    'action': [ 'download_from_google_storage',
+                '--no_resume',
+                '--platform=darwin',
+                '--no_auth',
+                '--bucket', 'chromium-luci',
+                '-d', 'src/tools/luci-go/mac64',
+    ],
+  },
+  {
+    'name': 'luci-go_linux',
+    'pattern': '.',
+    'action': [ 'download_from_google_storage',
+                '--no_resume',
+                '--platform=linux*',
+                '--no_auth',
+                '--bucket', 'chromium-luci',
+                '-d', 'src/tools/luci-go/linux64',
+    ],
+  },
+  {
+    'name': 'Generate component metadata for tests',
+    'pattern': '.',
+    'action': [
+      'vpython3',
+      'src/testing/generate_location_tags.py',
+      '--out',
+      'src/testing/location_tags.json',
+    ],
+  },
+]
+
+recursedeps = []

yuv.mod/libyuv/DIR_METADATA

@@ -0,0 +1,3 @@
+monorail {
+  component: "Internals>Images>Codecs"
+}

yuv.mod/libyuv/LICENSE

@@ -0,0 +1,29 @@
+Copyright 2011 The LibYuv Project Authors. All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+
+  * Redistributions of source code must retain the above copyright
+    notice, this list of conditions and the following disclaimer.
+
+  * Redistributions in binary form must reproduce the above copyright
+    notice, this list of conditions and the following disclaimer in
+    the documentation and/or other materials provided with the
+    distribution.
+
+  * Neither the name of Google nor the names of its contributors may
+    be used to endorse or promote products derived from this software
+    without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

yuv.mod/libyuv/OWNERS

@@ -0,0 +1,11 @@
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
+
+per-file *[email protected],[email protected]
+per-file .gitignore=*
+per-file AUTHORS=*
+per-file DEPS=*
+per-file [email protected],[email protected]

yuv.mod/libyuv/PATENTS

@@ -0,0 +1,24 @@
+Additional IP Rights Grant (Patents)
+
+"This implementation" means the copyrightable works distributed by
+Google as part of the LibYuv code package.
+
+Google hereby grants to you a perpetual, worldwide, non-exclusive,
+no-charge, irrevocable (except as stated in this section) patent
+license to make, have made, use, offer to sell, sell, import,
+transfer, and otherwise run, modify and propagate the contents of this
+implementation of the LibYuv code package, where such license applies
+only to those patent claims, both currently owned by Google and
+acquired in the future, licensable by Google that are necessarily
+infringed by this implementation of the LibYuv code package. This
+grant does not include claims that would be infringed only as a
+consequence of further modification of this implementation. If you or
+your agent or exclusive licensee institute or order or agree to the
+institution of patent litigation against any entity (including a
+cross-claim or counterclaim in a lawsuit) alleging that this
+implementation of the LibYuv code package or any code incorporated
+within this implementation of the LibYuv code package constitutes
+direct or contributory patent infringement, or inducement of patent
+infringement, then any patent rights granted to you under this License
+for this implementation of the LibYuv code package shall terminate as
+of the date such litigation is filed.

yuv.mod/libyuv/PRESUBMIT.py

@@ -0,0 +1,52 @@
+# Copyright 2017 The LibYuv Project Authors. All rights reserved.
+#
+# Use of this source code is governed by a BSD-style license
+# that can be found in the LICENSE file in the root of the source
+# tree. An additional intellectual property rights grant can be found
+# in the file PATENTS. All contributing project authors may
+# be found in the AUTHORS file in the root of the source tree.
+
+# Runs PRESUBMIT.py in py3 mode by git cl presubmit.
+USE_PYTHON3 = True
+
+def _CommonChecks(input_api, output_api):
+  """Checks common to both upload and commit."""
+  results = []
+  results.extend(input_api.canned_checks.RunPylint(input_api, output_api,
+      files_to_skip=(r'^base[\\\/].*\.py$',
+                     r'^build[\\\/].*\.py$',
+                     r'^buildtools[\\\/].*\.py$',
+                     r'^ios[\\\/].*\.py$',
+                     r'^out.*[\\\/].*\.py$',
+                     r'^testing[\\\/].*\.py$',
+                     r'^third_party[\\\/].*\.py$',
+                     r'^tools[\\\/].*\.py$',
+                     # TODO(kjellander): should arguably be checked.
+                     r'^tools_libyuv[\\\/]valgrind[\\\/].*\.py$',
+                     r'^xcodebuild.*[\\\/].*\.py$',),
+      disabled_warnings=['F0401',  # Failed to import x
+                         'E0611',  # No package y in x
+                         'W0232',  # Class has no __init__ method
+                        ],
+      pylintrc='pylintrc',
+      version='2.7'))
+  return results
+
+
+def CheckChangeOnUpload(input_api, output_api):
+  results = []
+  results.extend(_CommonChecks(input_api, output_api))
+  results.extend(
+      input_api.canned_checks.CheckGNFormatted(input_api, output_api))
+  return results
+
+
+def CheckChangeOnCommit(input_api, output_api):
+  results = []
+  results.extend(_CommonChecks(input_api, output_api))
+  results.extend(input_api.canned_checks.CheckOwners(input_api, output_api))
+  results.extend(input_api.canned_checks.CheckChangeWasUploaded(
+      input_api, output_api))
+  results.extend(input_api.canned_checks.CheckChangeHasDescription(
+      input_api, output_api))
+  return results

yuv.mod/libyuv/README.chromium

@@ -0,0 +1,10 @@
+Name: libyuv
+URL: https://chromium.googlesource.com/libyuv/libyuv/
+Version: 1895
+License: BSD
+License File: LICENSE
+Shipped: yes
+
+Description:
+libyuv is an open source project that includes YUV conversion and scaling functionality.
+

yuv.mod/libyuv/README.md

@@ -0,0 +1,19 @@
+**libyuv** is an open source project that includes YUV scaling and conversion functionality.
+
+* Scale YUV to prepare content for compression, with point, bilinear or box filter.
+* Convert to YUV from webcam formats for compression.
+* Convert to RGB formats for rendering/effects.
+* Rotate by 90/180/270 degrees to adjust for mobile devices in portrait mode.
+* Optimized for SSSE3/AVX2 on x86/x64.
+* Optimized for Neon on Arm.
+* Optimized for MSA on Mips.
+* Optimized for RVV on RISC-V.
+
+### Development
+
+See [Getting started][1] for instructions on how to get started developing.
+
+You can also browse the [docs directory][2] for more documentation.
+
+[1]: ./docs/getting_started.md
+[2]: ./docs/

yuv.mod/libyuv/build_overrides/build.gni

@@ -0,0 +1,63 @@
+# Copyright 2016 The LibYuv Project Authors. All rights reserved.
+#
+# Use of this source code is governed by a BSD-style license
+# that can be found in the LICENSE file in the root of the source
+# tree. An additional intellectual property rights grant can be found
+# in the file PATENTS. All contributing project authors may
+# be found in the AUTHORS file in the root of the source tree.
+
+# Variable that can be used to support multiple build scenarios, like having
+# Chromium specific targets in a client project's GN file etc.
+build_with_chromium = false
+
+# Some non-Chromium builds don't support building java targets.
+enable_java_templates = true
+
+# Enables assertions on safety checks in libc++.
+enable_safe_libcxx = true
+
+# Allow using custom suppressions files (currently not used by libyuv).
+asan_suppressions_file = "//build/sanitizers/asan_suppressions.cc"
+lsan_suppressions_file = "//build/sanitizers/lsan_suppressions.cc"
+tsan_suppressions_file = "//build/sanitizers/tsan_suppressions.cc"
+
+msan_blacklist_path =
+    rebase_path("//tools_libyuv/msan/blacklist.txt", root_build_dir)
+ubsan_blacklist_path =
+    rebase_path("//tools_libyuv/ubsan/blacklist.txt", root_build_dir)
+ubsan_vptr_blacklist_path =
+    rebase_path("//tools_libyuv/ubsan/vptr_blacklist.txt", root_build_dir)
+
+# For Chromium, Android 32-bit non-component, non-clang builds hit a 4GiB size
+# limit, making them requiring symbol_level=2. WebRTC doesn't hit that problem
+# so we just ignore that assert. See https://crbug.com/648948 for more info.
+ignore_elf32_limitations = true
+
+# Use bundled hermetic Xcode installation maintained by Chromium,
+# except for local iOS builds where it is unsupported.
+if (host_os == "mac") {
+  _result = exec_script("//build/mac/should_use_hermetic_xcode.py",
+                        [ target_os ],
+                        "value")
+  assert(_result != 2,
+         "Do not allow building targets with the default" +
+             "hermetic toolchain if the minimum OS version is not met.")
+  use_system_xcode = _result == 0
+}
+
+declare_args() {
+  # Tracing support requires //third_party/perfetto.
+  enable_base_tracing = false
+  use_perfetto_client_library = false
+
+  # Limits the defined //third_party/android_deps targets to only "buildCompile"
+  # and "buildCompileNoDeps" targets. This is useful for third-party
+  # repositories which do not use JUnit tests. For instance,
+  # limit_android_deps == true removes "gn gen" requirement for
+  # //third_party/robolectric .
+  limit_android_deps = false
+
+  # Allows googletest to pretty-print various absl types.
+  # Defined here rather than in gtest.gni to match chromium.
+  gtest_enable_absl_printers = true
+}

yuv.mod/libyuv/build_overrides/gtest.gni

@@ -0,0 +1,19 @@
+# Copyright (c) 2016 The LibYuv project authors. All Rights Reserved.
+#
+# Use of this source code is governed by a BSD-style license
+# that can be found in the LICENSE file in the root of the source
+# tree. An additional intellectual property rights grant can be found
+# in the file PATENTS.  All contributing project authors may
+# be found in the AUTHORS file in the root of the source tree.
+
+# Include support for registering main function in multi-process tests.
+gtest_include_multiprocess = true
+
+# Include support for platform-specific operations across unit tests.
+gtest_include_platform_test = true
+
+# Exclude support for testing Objective C code on OS X and iOS.
+gtest_include_objc_support = true
+
+# Exclude support for flushing coverage files on iOS.
+gtest_include_ios_coverage = true

yuv.mod/libyuv/build_overrides/partition_alloc.gni

@@ -0,0 +1,17 @@
+# Copyright 2022 The LibYuv Project Authors. All rights reserved.
+#
+# Use of this source code is governed by a BSD-style license
+# that can be found in the LICENSE file in the root of the source
+# tree. An additional intellectual property rights grant can be found
+# in the file PATENTS. All contributing project authors may
+# be found in the AUTHORS file in the root of the source tree.
+
+# Use default values for PartitionAlloc as standalone library from
+# base/allocator/partition_allocator/build_overrides/partition_alloc.gni
+use_partition_alloc_as_malloc_default = false
+use_allocator_shim_default = false
+enable_backup_ref_ptr_support_default = false
+enable_mte_checked_ptr_support_default = false
+put_ref_count_in_previous_slot_default = false
+enable_backup_ref_ptr_slow_checks_default = false
+enable_dangling_raw_ptr_checks_default = false

yuv.mod/libyuv/codereview.settings

@@ -0,0 +1,5 @@
+# This file is used by `git cl` to get repository specific information.
+CODE_REVIEW_SERVER: codereview.chromium.org
+GERRIT_HOST: True
+PROJECT: libyuv
+VIEW_VC: https://chromium.googlesource.com/libyuv/libyuv/+/

yuv.mod/libyuv/docs/deprecated_builds.md

@@ -0,0 +1,441 @@
+# Deprecated Builds
+
+Older documentation on build configs which are no longer supported.
+
+## Pre-requisites
+
+You'll need to have depot tools installed: https://www.chromium.org/developers/how-tos/install-depot-tools
+Refer to chromium instructions for each platform for other prerequisites.
+
+## Getting the Code
+
+Create a working directory, enter it, and run:
+
+    gclient config https://chromium.googlesource.com/libyuv/libyuv
+    gclient sync
+
+
+Then you'll get a .gclient file like:
+
+    solutions = [
+      { "name"        : "libyuv",
+        "url"         : "https://chromium.googlesource.com/libyuv/libyuv",
+        "deps_file"   : "DEPS",
+        "managed"     : True,
+        "custom_deps" : {
+        },
+        "safesync_url": "",
+      },
+    ];
+
+
+For iOS add `;target_os=['ios'];` to your OSX .gclient and run `GYP_DEFINES="OS=ios" gclient sync.`
+
+Browse the Git reprository: https://chromium.googlesource.com/libyuv/libyuv/+/master
+
+### Android
+For Android add `;target_os=['android'];` to your Linux .gclient
+
+
+    solutions = [
+      { "name"        : "libyuv",
+        "url"         : "https://chromium.googlesource.com/libyuv/libyuv",
+        "deps_file"   : "DEPS",
+        "managed"     : True,
+        "custom_deps" : {
+        },
+        "safesync_url": "",
+      },
+    ];
+    target_os = ["android", "unix"];
+
+Then run:
+
+    export GYP_DEFINES="OS=android"
+    gclient sync
+
+Caveat: Theres an error with Google Play services updates.  If you get the error "Your version of the Google Play services library is not up to date", run the following:
+
+    cd chromium/src
+    ./build/android/play_services/update.py download
+    cd ../..
+
+For Windows the gclient sync must be done from an Administrator command prompt.
+
+The sync will generate native build files for your environment using gyp (Windows: Visual Studio, OSX: XCode, Linux: make). This generation can also be forced manually: `gclient runhooks`
+
+To get just the source (not buildable):
+
+    git clone https://chromium.googlesource.com/libyuv/libyuv
+
+
+## Building the Library and Unittests
+
+### Windows
+
+    set GYP_DEFINES=target_arch=ia32
+    call python gyp_libyuv -fninja -G msvs_version=2013
+    ninja -j7 -C out\Release
+    ninja -j7 -C out\Debug
+
+    set GYP_DEFINES=target_arch=x64
+    call python gyp_libyuv -fninja -G msvs_version=2013
+    ninja -C out\Debug_x64
+    ninja -C out\Release_x64
+
+#### Building with clangcl
+    set GYP_DEFINES=clang=1 target_arch=ia32
+    call python tools\clang\scripts\update.py
+    call python gyp_libyuv -fninja libyuv_test.gyp
+    ninja -C out\Debug
+    ninja -C out\Release
+
+### OSX
+
+Clang 64 bit shown. Remove `clang=1` for GCC and change x64 to ia32 for 32 bit.
+
+    GYP_DEFINES="clang=1 target_arch=x64" ./gyp_libyuv
+    ninja -j7 -C out/Debug
+    ninja -j7 -C out/Release
+
+    GYP_DEFINES="clang=1 target_arch=ia32" ./gyp_libyuv
+    ninja -j7 -C out/Debug
+    ninja -j7 -C out/Release
+
+### iOS
+http://www.chromium.org/developers/how-tos/build-instructions-ios
+
+Add to .gclient last line: `target_os=['ios'];`
+
+armv7
+
+    GYP_DEFINES="OS=ios target_arch=armv7 target_subarch=arm32" GYP_CROSSCOMPILE=1 GYP_GENERATOR_FLAGS="output_dir=out_ios" ./gyp_libyuv
+    ninja -j7 -C out_ios/Debug-iphoneos libyuv_unittest
+    ninja -j7 -C out_ios/Release-iphoneos libyuv_unittest
+
+arm64
+
+    GYP_DEFINES="OS=ios target_arch=arm64 target_subarch=arm64" GYP_CROSSCOMPILE=1 GYP_GENERATOR_FLAGS="output_dir=out_ios" ./gyp_libyuv
+    ninja -j7 -C out_ios/Debug-iphoneos libyuv_unittest
+    ninja -j7 -C out_ios/Release-iphoneos libyuv_unittest
+
+both armv7 and arm64 (fat)
+
+    GYP_DEFINES="OS=ios target_arch=armv7 target_subarch=both" GYP_CROSSCOMPILE=1 GYP_GENERATOR_FLAGS="output_dir=out_ios" ./gyp_libyuv
+    ninja -j7 -C out_ios/Debug-iphoneos libyuv_unittest
+    ninja -j7 -C out_ios/Release-iphoneos libyuv_unittest
+
+simulator
+
+    GYP_DEFINES="OS=ios target_arch=ia32 target_subarch=arm32" GYP_CROSSCOMPILE=1 GYP_GENERATOR_FLAGS="output_dir=out_sim" ./gyp_libyuv
+    ninja -j7 -C out_sim/Debug-iphonesimulator libyuv_unittest
+    ninja -j7 -C out_sim/Release-iphonesimulator libyuv_unittest
+
+### Android
+https://code.google.com/p/chromium/wiki/AndroidBuildInstructions
+
+Add to .gclient last line: `target_os=['android'];`
+
+armv7
+
+    GYP_DEFINES="OS=android" GYP_CROSSCOMPILE=1 ./gyp_libyuv
+    ninja -j7 -C out/Debug yuv_unittest_apk
+    ninja -j7 -C out/Release yuv_unittest_apk
+
+arm64
+
+    GYP_DEFINES="OS=android target_arch=arm64 target_subarch=arm64" GYP_CROSSCOMPILE=1 ./gyp_libyuv
+    ninja -j7 -C out/Debug yuv_unittest_apk
+    ninja -j7 -C out/Release yuv_unittest_apk
+
+ia32
+
+    GYP_DEFINES="OS=android target_arch=ia32" GYP_CROSSCOMPILE=1 ./gyp_libyuv
+    ninja -j7 -C out/Debug yuv_unittest_apk
+    ninja -j7 -C out/Release yuv_unittest_apk
+
+    GYP_DEFINES="OS=android target_arch=ia32 android_full_debug=1" GYP_CROSSCOMPILE=1 ./gyp_libyuv
+    ninja -j7 -C out/Debug yuv_unittest_apk
+
+mipsel
+
+    GYP_DEFINES="OS=android target_arch=mipsel" GYP_CROSSCOMPILE=1 ./gyp_libyuv
+    ninja -j7 -C out/Debug yuv_unittest_apk
+    ninja -j7 -C out/Release yuv_unittest_apk
+
+arm32 disassembly:
+
+    llvm-objdump -d out/Release/obj/source/libyuv.row_neon.o
+
+arm64 disassembly:
+
+    llvm-objdump -d out/Release/obj/source/libyuv.row_neon64.o
+
+Running tests:
+
+    build/android/test_runner.py gtest -s libyuv_unittest -t 7200 --verbose --release --gtest_filter=*
+
+Running test as benchmark:
+
+    build/android/test_runner.py gtest -s libyuv_unittest -t 7200 --verbose --release --gtest_filter=* -a "--libyuv_width=1280 --libyuv_height=720 --libyuv_repeat=999 --libyuv_flags=-1"
+
+Running test with C code:
+
+    build/android/test_runner.py gtest -s libyuv_unittest -t 7200 --verbose --release --gtest_filter=* -a "--libyuv_width=1280 --libyuv_height=720 --libyuv_repeat=999 --libyuv_flags=1 --libyuv_cpu_info=1"
+
+#### Building with GN
+
+    gn gen out/Release "--args=is_debug=false target_cpu=\"x86\""
+    gn gen out/Debug "--args=is_debug=true target_cpu=\"x86\""
+    ninja -C out/Release
+    ninja -C out/Debug
+
+### Building Offical with GN
+
+    gn gen out/Official "--args=is_debug=false is_official_build=true is_chrome_branded=true"
+    ninja -C out/Official
+
+#### Building mips with GN
+
+mipsel
+    gn gen out/Default "--args=is_debug=false target_cpu=\"mipsel\" target_os = \"android\" mips_arch_variant = \"r6\" mips_use_msa = true is_component_build = true is_clang = false"
+    ninja -C out/Default
+
+mips64el
+    gn gen out/Default "--args=is_debug=false target_cpu=\"mips64el\" target_os = \"android\" mips_arch_variant = \"r6\" mips_use_msa = true is_component_build = true is_clang = false"
+    ninja -C out/Default
+
+### Linux
+
+    GYP_DEFINES="target_arch=x64" ./gyp_libyuv
+    ninja -j7 -C out/Debug
+    ninja -j7 -C out/Release
+
+    GYP_DEFINES="target_arch=ia32" ./gyp_libyuv
+    ninja -j7 -C out/Debug
+    ninja -j7 -C out/Release
+
+#### CentOS
+
+On CentOS 32 bit the following work around allows a sync:
+
+    export GYP_DEFINES="host_arch=ia32"
+    gclient sync
+
+### Windows Shared Library
+
+Modify libyuv.gyp from 'static_library' to 'shared_library', and add 'LIBYUV_BUILDING_SHARED_LIBRARY' to 'defines'.
+
+    gclient runhooks
+
+After this command follow the building the library instructions above.
+
+If you get a compile error for atlthunk.lib on Windows, read http://www.chromium.org/developers/how-tos/build-instructions-windows
+
+
+### Build targets
+
+    ninja -C out/Debug libyuv
+    ninja -C out/Debug libyuv_unittest
+    ninja -C out/Debug compare
+    ninja -C out/Debug yuvconvert
+    ninja -C out/Debug yuvconstants
+    ninja -C out/Debug psnr
+    ninja -C out/Debug cpuid
+
+
+## Building the Library with make
+
+### Linux
+
+    make -j7 V=1 -f linux.mk
+    make -j7 V=1 -f linux.mk clean
+    make -j7 V=1 -f linux.mk CXX=clang++
+
+## Building the Library with cmake
+
+Install cmake: http://www.cmake.org/
+
+Default debug build:
+
+    mkdir out
+    cd out
+    cmake ..
+    cmake --build .
+
+Release build/install
+
+    mkdir out
+    cd out
+    cmake -DCMAKE_INSTALL_PREFIX="/usr/lib" -DCMAKE_BUILD_TYPE="Release" ..
+    cmake --build . --config Release
+    sudo cmake --build . --target install --config Release
+
+### Windows 8 Phone
+
+Pre-requisite:
+
+* Install Visual Studio 2012 and Arm to your environment.<br>
+
+Then:
+
+    call "c:\Program Files (x86)\Microsoft Visual Studio 11.0\VC\bin\x86_arm\vcvarsx86_arm.bat"
+
+or with Visual Studio 2013:
+
+    call "c:\Program Files (x86)\Microsoft Visual Studio 12.0\VC\bin\x86_arm\vcvarsx86_arm.bat"
+    nmake /f winarm.mk clean
+    nmake /f winarm.mk
+
+### Windows Shared Library
+
+Modify libyuv.gyp from 'static_library' to 'shared_library', and add 'LIBYUV_BUILDING_SHARED_LIBRARY' to 'defines'. Then run this.
+
+    gclient runhooks
+
+After this command follow the building the library instructions above.
+
+If you get a compile error for atlthunk.lib on Windows, read http://www.chromium.org/developers/how-tos/build-instructions-windows
+
+### 64 bit Windows
+
+    set GYP_DEFINES=target_arch=x64
+    gclient runhooks V=1
+
+### ARM Linux
+
+    export GYP_DEFINES="target_arch=arm"
+    export CROSSTOOL=`<path>`/arm-none-linux-gnueabi
+    export CXX=$CROSSTOOL-g++
+    export CC=$CROSSTOOL-gcc
+    export AR=$CROSSTOOL-ar
+    export AS=$CROSSTOOL-as
+    export RANLIB=$CROSSTOOL-ranlib
+    gclient runhooks
+
+## Running Unittests
+
+### Windows
+
+    out\Release\libyuv_unittest.exe --gtest_catch_exceptions=0 --gtest_filter="*"
+
+### OSX
+
+    out/Release/libyuv_unittest --gtest_filter="*"
+
+### Linux
+
+    out/Release/libyuv_unittest --gtest_filter="*"
+
+Replace --gtest_filter="*" with specific unittest to run.  May include wildcards. e.g.
+
+    out/Release/libyuv_unittest --gtest_filter=libyuvTest.I420ToARGB_Opt
+
+## CPU Emulator tools
+
+### Intel SDE (Software Development Emulator)
+
+Pre-requisite: Install IntelSDE for Windows: http://software.intel.com/en-us/articles/intel-software-development-emulator
+
+Then run:
+
+    c:\intelsde\sde -hsw -- out\release\libyuv_unittest.exe --gtest_filter=*
+
+
+## Memory tools
+
+### Running Dr Memory memcheck for Windows
+
+Pre-requisite: Install Dr Memory for Windows and add it to your path: http://www.drmemory.org/docs/page_install_windows.html
+
+    set GYP_DEFINES=build_for_tool=drmemory target_arch=ia32
+    call python gyp_libyuv -fninja -G msvs_version=2013
+    ninja -C out\Debug
+    drmemory out\Debug\libyuv_unittest.exe --gtest_catch_exceptions=0 --gtest_filter=*
+
+### Running UBSan
+
+See Chromium instructions for sanitizers: https://www.chromium.org/developers/testing/undefinedbehaviorsanitizer
+
+Sanitizers available: TSan, MSan, ASan, UBSan, LSan
+
+    GYP_DEFINES='ubsan=1' gclient runhooks
+    ninja -C out/Release
+
+### Running Valgrind memcheck
+
+Memory errors and race conditions can be found by running tests under special memory tools. [Valgrind] [1] is an instrumentation framework for building dynamic analysis tools. Various tests and profilers are built upon it to find memory handling errors and memory leaks, for instance.
+
+[1]: http://valgrind.org
+
+    solutions = [
+      { "name"        : "libyuv",
+        "url"         : "https://chromium.googlesource.com/libyuv/libyuv",
+        "deps_file"   : "DEPS",
+        "managed"     : True,
+        "custom_deps" : {
+           "libyuv/chromium/src/third_party/valgrind": "https://chromium.googlesource.com/chromium/deps/valgrind/binaries",
+        },
+        "safesync_url": "",
+      },
+    ]
+
+Then run:
+
+    GYP_DEFINES="clang=0 target_arch=x64 build_for_tool=memcheck" python gyp_libyuv
+    ninja -C out/Debug
+    valgrind out/Debug/libyuv_unittest
+
+
+For more information, see http://www.chromium.org/developers/how-tos/using-valgrind
+
+### Running Thread Sanitizer (TSan)
+
+    GYP_DEFINES="clang=0 target_arch=x64 build_for_tool=tsan" python gyp_libyuv
+    ninja -C out/Debug
+    valgrind out/Debug/libyuv_unittest
+
+For more info, see http://www.chromium.org/developers/how-tos/using-valgrind/threadsanitizer
+
+### Running Address Sanitizer (ASan)
+
+    GYP_DEFINES="clang=0 target_arch=x64 build_for_tool=asan" python gyp_libyuv
+    ninja -C out/Debug
+    valgrind out/Debug/libyuv_unittest
+
+For more info, see http://dev.chromium.org/developers/testing/addresssanitizer
+
+## Benchmarking
+
+The unittests can be used to benchmark.
+
+### Windows
+
+    set LIBYUV_WIDTH=1280
+    set LIBYUV_HEIGHT=720
+    set LIBYUV_REPEAT=999
+    set LIBYUV_FLAGS=-1
+    out\Release\libyuv_unittest.exe --gtest_filter=*I420ToARGB_Opt
+
+
+### Linux and Mac
+
+    LIBYUV_WIDTH=1280 LIBYUV_HEIGHT=720 LIBYUV_REPEAT=1000 out/Release/libyuv_unittest --gtest_filter=*I420ToARGB_Opt
+
+    libyuvTest.I420ToARGB_Opt (547 ms)
+
+Indicates 0.547 ms/frame for 1280 x 720.
+
+## Making a change
+
+    gclient sync
+    git checkout -b mycl -t origin/master
+    git pull
+    <edit files>
+    git add -u
+    git commit -m "my change"
+    git cl lint
+    git cl try
+    git cl upload -r [email protected] -s
+    <once approved..>
+    git cl land

yuv.mod/libyuv/docs/environment_variables.md

@@ -0,0 +1,57 @@
+# Introduction
+
+For test purposes, environment variables can be set to control libyuv behavior.  These should only be used for testing, to narrow down bugs or to test performance.
+
+# CPU
+
+By default the cpu is detected and the most advanced form of SIMD is used.  But you can disable instruction sets selectively, or completely, falling back on C code.  Set the variable to 1 to disable the specified instruction set.
+
+## All CPUs
+    LIBYUV_DISABLE_ASM
+
+## Intel CPUs
+    LIBYUV_DISABLE_X86
+    LIBYUV_DISABLE_SSE2
+    LIBYUV_DISABLE_SSSE3
+    LIBYUV_DISABLE_SSE41
+    LIBYUV_DISABLE_SSE42
+    LIBYUV_DISABLE_AVX
+    LIBYUV_DISABLE_AVX2
+    LIBYUV_DISABLE_ERMS
+    LIBYUV_DISABLE_FMA3
+    LIBYUV_DISABLE_F16C
+    LIBYUV_DISABLE_AVX512BW
+    LIBYUV_DISABLE_AVX512VL
+    LIBYUV_DISABLE_AVX512VNNI
+    LIBYUV_DISABLE_AVX512VBMI
+    LIBYUV_DISABLE_AVX512VBMI2
+    LIBYUV_DISABLE_AVX512VBITALG
+    LIBYUV_DISABLE_AVX10
+    LIBYUV_DISABLE_AVXVNNI
+    LIBYUV_DISABLE_AVXVNNIINT8
+    LIBYUV_DISABLE_AMXINT8
+
+## ARM CPUs
+
+    LIBYUV_DISABLE_NEON
+
+## MIPS CPUs
+    LIBYUV_DISABLE_MSA
+
+## LOONGARCH CPUs
+    LIBYUV_DISABLE_LSX
+    LIBYUV_DISABLE_LASX
+
+## RISCV CPUs
+    LIBYUV_DISABLE_RVV
+
+# Test Width/Height/Repeat
+
+The unittests default to a small image (128x72) to run fast.  This can be set by environment variable to test a specific resolutions.
+You can also repeat the test a specified number of iterations, allowing benchmarking and profiling.
+
+    set LIBYUV_WIDTH=1280
+    set LIBYUV_HEIGHT=720
+    set LIBYUV_REPEAT=999
+    set LIBYUV_FLAGS=-1
+    set LIBYUV_CPU_INFO=-1

yuv.mod/libyuv/docs/feature_detection.md

@@ -0,0 +1,108 @@
+# Introduction
+
+Several routines in libyuv have multiple implementations specialized for a
+variety of CPU architecture extensions. Libyuv will automatically detect and
+use the latest architecture extension present on a machine for which a kernel
+implementation is available.
+
+# Feature detection on AArch64
+
+## Architecture extensions of interest
+
+The Arm 64-bit A-class architecture has a number of vector extensions which can
+be used to accelerate libyuv kernels.
+
+### Neon extensions
+
+Neon is available and mandatory in AArch64 from the base Armv8.0-A
+architecture. Neon can be used even if later extensions like the Scalable
+Vector Extension (SVE) are also present. The exception to this is if the CPU is
+currently operating in streaming mode as introduced by the Scalable Matrix
+Extension, described later.
+
+There are also a couple of architecture extensions present for Neon that we can
+take advantage of in libyuv:
+
+* The Neon DotProd extension is architecturally available from Armv8.1-A and
+  becomes mandatory from Armv8.4-A. This extension provides instructions to
+  perform a pairwise widening multiply of groups of four bytes from two source
+  vectors, taking the sum of the four widened multiply results within each
+  group to give a 32-bit result, accumulating into a destination vector.
+
+* The Neon I8MM extension extends the DotProd extension with support for
+  mixed-sign DotProds. The I8MM extension is architecturally available from
+  Armv8.1-A and becomes mandatory from Armv8.6-A. It does not strictly depend
+  on the DotProd extension being implemented, however at time of writing there
+  is no known micro-architecture implementation where I8MM is implemented
+  without the DotProd extension also being implemented.
+
+### The Scalable Vector Extension (SVE)
+
+The two Scalable Vector extensions (SVE and SVE2) provides equivalent
+functionality to most existing Neon instructions but with the ability to
+efficiently operate on vector registers with a run-time-determined vector
+length.
+
+The original version of SVE is architecturally available from Armv8.2-A and is
+primarily targeted at HPC applications. This focus means it does not include
+most of the DSP-style operations that are necessary for most libyuv
+color-conversion kernels, though it can still be used for many scaling or
+rotation kernels.
+
+SVE does not strictly depend on either of the Neon DotProd or I8MM extensions
+being implemented. The only micro-architecture at time of writing where SVE is
+implemented without these two extensions both also being implemented is the
+Fujitsu A64FX, which is not a CPU of interest for libyuv.
+
+SVE2 extends the base SVE extension with the remaining instructions from Neon,
+porting these instructions to operate on scalable vectors. SVE2 is
+architecturally available from Armv9.0-A. If SVE2 is implemented then SVE must
+also be implemented. Since Armv9.0-A is based on Armv8.5-A this implies that
+the Neon DotProd extension is also implemented. Interestingly this means that
+the I8MM extension is not mandatory since it only becomes mandatory from
+Armv8.6-A or Armv9.1-A, however there is no micro-architecture at time of
+writing where SVE2 is implemented without all previously-mentioned features
+also being implemented.
+
+### The Scalable Matrix Extension (SME)
+
+The Scalable Matrix Extension (SME) is an optional feature introduced from
+Armv9.2-A. SME exists alongside SVE and introduces new execution modes for
+applications performing extended periods of data processing. In particular SME
+introduces a few new components of interest:
+
+* Access to a scalable two-dimensional ZA tile register and new instructions to
+  interact with rows and columns of the ZA tiles. This can be useful for data
+  transformations like transposes.
+
+* A streaming SVE (SSVE) mode, during which the SVE vector length matches the
+  ZA tile register width. In typical systems where the ZA tile register width
+  is longer than the core SVE vector length, SSVE processing allows for faster
+  data processing, even if the ZA tile register is unused.  While the CPU is
+  executing in streaming mode, Neon instructions are unavailable.
+
+* When both SSVE and the ZA tile registers are enabled there are additional
+  outer-product instructions accumulating into a whole ZA tile, suitable for
+  accelerating matrix arithmetic. This is likely less useful in libyuv.
+
+## Linux and Android
+
+On AArch64 running under Linux and Android, features are detected by inspecting
+the CPU auxiliary vector via `getauxval(AT_HWCAP)` and `getauxval(AT_HWCAP2)`,
+inspecting the returned bitmask.
+
+## Windows
+
+On Windows we detect features using the `IsProcessorFeaturePresent` interface
+and passing an enum parameter for the feature we want to check. More
+information on this can be found here:
+
+    https://learn.microsoft.com/en-us/windows/win32/api/processthreadsapi/nf-processthreadsapi-isprocessorfeaturepresent#parameters
+
+## Apple Silicon
+
+On Apple Silicon we detect features using the `sysctlbyname` interface and
+passing a string representing the feature we want to detect. More information
+on this can be found here:
+
+    https://developer.apple.com/documentation/kernel/1387446-sysctlbyname/determining_instruction_set_characteristics

yuv.mod/libyuv/docs/filtering.md

@@ -0,0 +1,196 @@
+# Introduction
+
+This document discusses the current state of filtering in libyuv. An emphasis on maximum performance while avoiding memory exceptions, and minimal amount of code/complexity.  See future work at end.
+
+# LibYuv Filter Subsampling
+
+There are 2 challenges with subsampling
+
+1. centering of samples, which involves clamping on edges
+2. clipping a source region
+
+Centering depends on scale factor and filter mode.
+
+# Down Sampling
+
+If scaling down, the stepping rate is always src_width / dst_width.
+
+    dx = src_width / dst_width;
+
+e.g. If scaling from 1280x720 to 640x360, the step thru the source will be 2.0, stepping over 2 pixels of source for each pixel of destination.
+
+Centering, depends on filter mode.
+
+*Point* downsampling takes the middle pixel.
+
+    x = dx >> 1;
+
+For odd scale factors (e.g. 3x down) this is exactly the middle.  For even scale factors, this rounds up and takes the pixel to the right of center.  e.g. scale of 4x down will take pixel 2.
+
+**Bilinear** filter, uses the 2x2 pixels in the middle.
+
+    x = dx / 2 - 0.5;
+
+For odd scale factors (e.g. 3x down) this is exactly the middle, and point sampling is used.
+For even scale factors, this evenly filters the middle 2x2 pixels.  e.g. 4x down will filter pixels 1,2 at 50% in both directions.
+
+**Box** filter averages the entire box so sampling starts at 0.
+
+    x = 0;
+
+For a scale factor of 2x down, this is equivalent to bilinear.
+
+# Up Sampling
+
+**Point** upsampling use stepping rate of src_width / dst_width and a starting coordinate of 0.
+
+    x = 0;
+    dx = src_width / dst_width;
+
+e.g. If scaling from 640x360 to 1280x720 the step thru the source will be 0.0, stepping half a pixel of source for each pixel of destination. Each pixel is replicated by the scale factor.
+
+**Bilinear** filter stretches such that the first pixel of source maps to the first pixel of destination, and the last pixel of source maps to the last pixel of destination.
+
+    x = 0;
+    dx = (src_width - 1) / (dst_width - 1);
+
+This method is not technically correct, and will likely change in the future.
+
+* It is inconsistent with the bilinear down sampler.  The same method could be used for down sampling, and then it would be more reversible, but that would prevent specialized 2x down sampling.
+* Although centered, the image is slightly magnified.
+* The filtering was changed in early 2013 - previously it used:
+
+        x = 0;
+        dx = (src_width - 1) / (dst_width - 1);
+
+Which is the correct scale factor, but shifted the image left, and extruded the last pixel.  The reason for the change was to remove the extruding code from the low level row functions, allowing 3 functions to sshare the same row functions - ARGBScale, I420Scale, and ARGBInterpolate.  Then the one function was ported to many cpu variations: SSE2, SSSE3, AVX2, Neon and 'Any' version for any number of pixels and alignment.  The function is also specialized for 0,25,50,75%.
+
+The above goes still has the potential to read the last pixel 100% and last pixel + 1 0%, which may cause a memory exception.  So the left pixel goes to a fraction less than the last pixel, but filters in the minimum amount of it, and the maximum of the last pixel.
+
+    dx = FixedDiv((src_width << 16) - 0x00010001, (dst << 16) - 0x00010000);
+
+**Box** filter for upsampling switches over to Bilinear.
+
+# Scale snippet:
+
+    #define CENTERSTART(dx, s) (dx < 0) ? -((-dx >> 1) + s) : ((dx >> 1) + s)
+    #define FIXEDDIV1(src, dst) FixedDiv((src << 16) - 0x00010001, \
+                                         (dst << 16) - 0x00010000);
+
+    // Compute slope values for stepping.
+    void ScaleSlope(int src_width, int src_height,
+                    int dst_width, int dst_height,
+                    FilterMode filtering,
+                    int* x, int* y, int* dx, int* dy) {
+      assert(x != NULL);
+      assert(y != NULL);
+      assert(dx != NULL);
+      assert(dy != NULL);
+      assert(src_width != 0);
+      assert(src_height != 0);
+      assert(dst_width > 0);
+      assert(dst_height > 0);
+      if (filtering == kFilterBox) {
+        // Scale step for point sampling duplicates all pixels equally.
+        *dx = FixedDiv(Abs(src_width), dst_width);
+        *dy = FixedDiv(src_height, dst_height);
+        *x = 0;
+        *y = 0;
+      } else if (filtering == kFilterBilinear) {
+        // Scale step for bilinear sampling renders last pixel once for upsample.
+        if (dst_width <= Abs(src_width)) {
+          *dx = FixedDiv(Abs(src_width), dst_width);
+          *x = CENTERSTART(*dx, -32768);
+        } else if (dst_width > 1) {
+          *dx = FIXEDDIV1(Abs(src_width), dst_width);
+          *x = 0;
+        }
+        if (dst_height <= src_height) {
+          *dy = FixedDiv(src_height,  dst_height);
+          *y = CENTERSTART(*dy, -32768);  // 32768 = -0.5 to center bilinear.
+        } else if (dst_height > 1) {
+          *dy = FIXEDDIV1(src_height, dst_height);
+          *y = 0;
+        }
+      } else if (filtering == kFilterLinear) {
+        // Scale step for bilinear sampling renders last pixel once for upsample.
+        if (dst_width <= Abs(src_width)) {
+          *dx = FixedDiv(Abs(src_width), dst_width);
+          *x = CENTERSTART(*dx, -32768);
+        } else if (dst_width > 1) {
+          *dx = FIXEDDIV1(Abs(src_width), dst_width);
+          *x = 0;
+        }
+        *dy = FixedDiv(src_height, dst_height);
+        *y = *dy >> 1;
+      } else {
+        // Scale step for point sampling duplicates all pixels equally.
+        *dx = FixedDiv(Abs(src_width), dst_width);
+        *dy = FixedDiv(src_height, dst_height);
+        *x = CENTERSTART(*dx, 0);
+        *y = CENTERSTART(*dy, 0);
+      }
+      // Negative src_width means horizontally mirror.
+      if (src_width < 0) {
+        *x += (dst_width - 1) * *dx;
+        *dx = -*dx;
+        src_width = -src_width;
+      }
+    }
+
+# Future Work
+
+Point sampling should ideally be the same as bilinear, but pixel by pixel, round to nearest neighbor.  But as is, it is reversible and exactly matches ffmpeg at all scale factors, both up and down.  The scale factor is
+
+    dx = src_width / dst_width;
+
+The step value is centered for down sample:
+
+    x = dx / 2;
+
+Or starts at 0 for upsample.
+
+    x = 0;
+
+Bilinear filtering is currently correct for down sampling, but not for upsampling.
+Upsampling is stretching the first and last pixel of source, to the first and last pixel of destination.
+
+    dx = (src_width - 1) / (dst_width - 1);<br>
+    x = 0;
+
+It should be stretching such that the first pixel is centered in the middle of the scale factor, to match the pixel that would be sampled for down sampling by the same amount.  And same on last pixel.
+
+    dx = src_width / dst_width;<br>
+    x = dx / 2 - 0.5;
+
+This would start at -0.5 and go to last pixel + 0.5, sampling 50% from last pixel + 1.
+Then clamping would be needed.  On GPUs there are numerous ways to clamp.
+
+1. Clamp the coordinate to the edge of the texture, duplicating the first and last pixel.
+2. Blend with a constant color, such as transparent black.  Typically best for fonts.
+3. Mirror the UV coordinate, which is similar to clamping.  Good for continuous tone images.
+4. Wrap the coordinate, for texture tiling.
+5. Allow the coordinate to index beyond the image, which may be the correct data if sampling a subimage.
+6. Extrapolate the edge based on the previous pixel.  pixel -0.5 is computed from slope of pixel 0 and 1.
+
+Some of these are computational, even for a GPU, which is one reason textures are sometimes limited to power of 2 sizes.
+We do care about the clipping case, where allowing coordinates to become negative and index pixels before the image is the correct data.  But normally for simple scaling, we want to clamp to the edge pixel.  For example, if bilinear scaling from 3x3 to 30x30, we’d essentially want 10 pixels of each of the original 3 pixels.  But we want the original pixels to land in the middle of each 10 pixels, at offsets 5, 15 and 25.  There would be filtering between 5 and 15 between the original pixels 0 and 1.  And filtering between 15 and 25 from original pixels 1 and 2.  The first 5 pixels are clamped to pixel 0 and the last 5 pixels are clamped to pixel 2.
+The easiest way to implement this is copy the original 3 pixels to a buffer, and duplicate the first and last pixels.  0,1,2 becomes 0, 0,1,2, 2.  Then implement a filtering without clamping.  We call this source extruding.  Its only necessary on up sampling, since down sampler will always have valid surrounding pixels.
+Extruding is practical when the image is already copied to a temporary buffer.   It could be done to the original image, as long as the original memory is restored, but valgrind and/or memory protection would disallow this, so it requires a memcpy to a temporary buffer, which may hurt performance.  The memcpy has a performance advantage, from a cache point of view, that can actually make this technique faster, depending on hardware characteristics.
+Vertical extrusion can be done with a memcpy of the first/last row, or clamping a pointer.
+
+
+The other way to implement clamping is handle the edges with a memset.  e.g. Read first source pixel and memset the first 5 pixels.  Filter pixels 0,1,2 to 5 to 25.  Read last pixel and memset the last 5 pixels.  Blur is implemented with this method like this, which has 3 loops per row - left, middle and right.
+
+Box filter is only used for 2x down sample or more.  Its based on integer sized boxes.  Technically it should be filtered edges, but thats substantially slower (roughly 100x), and at that point you may as well do a cubic filter which is more correct.
+
+Box filter currently sums rows into a row buffer.  It does this with
+
+Mirroring will use the same slope as normal, but with a negative.
+The starting coordinate needs to consider the scale factor and filter.  e.g. box filter of 30x30 to 3x3 with mirroring would use -10 for step, but x = 20.  width (30) - dx.
+
+Step needs to be accurate, so it uses an integer divide.  This is as much as 5% of the profile.  An approximated divide is substantially faster, but the inaccuracy causes stepping beyond the original image boundaries.  3 general solutions:
+
+1. copy image to buffer with padding.  allows for small errors in stepping.
+2. hash the divide, so common values are quickly found.
+3. change api so caller provides the slope.

yuv.mod/libyuv/docs/formats.md

@@ -0,0 +1,208 @@
+# Introduction
+
+Formats (FOURCC) supported by libyuv are detailed here.
+
+# Core Formats
+
+There are 2 core formats supported by libyuv - I420 and ARGB.
+  All YUV formats can be converted to/from I420.
+  All RGB formats can be converted to/from ARGB.
+
+Filtering functions such as scaling and planar functions work on I420 and/or ARGB.
+
+# OSX Core Media Pixel Formats
+
+This is how OSX formats map to libyuv
+
+    enum {
+      kCMPixelFormat_32ARGB          = 32,      FOURCC_BGRA
+      kCMPixelFormat_32BGRA          = 'BGRA',  FOURCC_ARGB
+      kCMPixelFormat_24RGB           = 24,      FOURCC_RAW
+      kCMPixelFormat_16BE555         = 16,      Not supported.
+      kCMPixelFormat_16BE565         = 'B565',  Not supported.
+      kCMPixelFormat_16LE555         = 'L555',  FOURCC_RGBO
+      kCMPixelFormat_16LE565         = 'L565',  FOURCC_RGBP
+      kCMPixelFormat_16LE5551        = '5551',  FOURCC_RGBO
+      kCMPixelFormat_422YpCbCr8      = '2vuy',  FOURCC_UYVY
+      kCMPixelFormat_422YpCbCr8_yuvs = 'yuvs',  FOURCC_YUY2
+      kCMPixelFormat_444YpCbCr8      = 'v308',  FOURCC_I444 ?
+      kCMPixelFormat_4444YpCbCrA8    = 'v408',  Not supported.
+      kCMPixelFormat_422YpCbCr16     = 'v216',  Not supported.
+      kCMPixelFormat_422YpCbCr10     = 'v210',  FOURCC_V210 previously.  Removed now.
+      kCMPixelFormat_444YpCbCr10     = 'v410',  Not supported.
+      kCMPixelFormat_8IndexedGray_WhiteIsZero = 0x00000028,  Not supported.
+    };
+
+
+# FOURCC (Four Charactacter Code) List
+
+The following is extracted from video_common.h as a complete list of formats supported by libyuv.
+    enum FourCC {
+      // 10 Primary YUV formats: 5 planar, 2 biplanar, 2 packed.
+      FOURCC_I420 = FOURCC('I', '4', '2', '0'),
+      FOURCC_I422 = FOURCC('I', '4', '2', '2'),
+      FOURCC_I444 = FOURCC('I', '4', '4', '4'),
+      FOURCC_I400 = FOURCC('I', '4', '0', '0'),
+      FOURCC_NV21 = FOURCC('N', 'V', '2', '1'),
+      FOURCC_NV12 = FOURCC('N', 'V', '1', '2'),
+      FOURCC_YUY2 = FOURCC('Y', 'U', 'Y', '2'),
+      FOURCC_UYVY = FOURCC('U', 'Y', 'V', 'Y'),
+      FOURCC_H010 = FOURCC('H', '0', '1', '0'),  // unofficial fourcc. 10 bit lsb
+      FOURCC_U010 = FOURCC('U', '0', '1', '0'),  // bt.2020, unofficial fourcc.
+                                                 // 10 bit lsb
+
+      // 1 Secondary YUV format: row biplanar.
+      FOURCC_M420 = FOURCC('M', '4', '2', '0'),  // deprecated.
+
+      // 13 Primary RGB formats: 4 32 bpp, 2 24 bpp, 3 16 bpp, 1 10 bpc, 2 64 bpp
+      FOURCC_ARGB = FOURCC('A', 'R', 'G', 'B'),
+      FOURCC_BGRA = FOURCC('B', 'G', 'R', 'A'),
+      FOURCC_ABGR = FOURCC('A', 'B', 'G', 'R'),
+      FOURCC_AR30 = FOURCC('A', 'R', '3', '0'),  // 10 bit per channel. 2101010.
+      FOURCC_AB30 = FOURCC('A', 'B', '3', '0'),  // ABGR version of 10 bit
+      FOURCC_AR64 = FOURCC('A', 'R', '6', '4'),  // 16 bit per channel.
+      FOURCC_AB64 = FOURCC('A', 'B', '6', '4'),  // ABGR version of 16 bit
+      FOURCC_24BG = FOURCC('2', '4', 'B', 'G'),
+      FOURCC_RAW = FOURCC('r', 'a', 'w', ' '),
+      FOURCC_RGBA = FOURCC('R', 'G', 'B', 'A'),
+      FOURCC_RGBP = FOURCC('R', 'G', 'B', 'P'),  // rgb565 LE.
+      FOURCC_RGBO = FOURCC('R', 'G', 'B', 'O'),  // argb1555 LE.
+      FOURCC_R444 = FOURCC('R', '4', '4', '4'),  // argb4444 LE.
+
+      // 1 Primary Compressed YUV format.
+      FOURCC_MJPG = FOURCC('M', 'J', 'P', 'G'),
+
+      // 11 Auxiliary YUV variations: 3 with U and V planes are swapped, 1 Alias.
+      FOURCC_YV12 = FOURCC('Y', 'V', '1', '2'),
+      FOURCC_YV16 = FOURCC('Y', 'V', '1', '6'),
+      FOURCC_YV24 = FOURCC('Y', 'V', '2', '4'),
+      FOURCC_YU12 = FOURCC('Y', 'U', '1', '2'),  // Linux version of I420.
+      FOURCC_J420 = FOURCC('J', '4', '2', '0'),
+      FOURCC_J400 = FOURCC('J', '4', '0', '0'),  // unofficial fourcc
+      FOURCC_H420 = FOURCC('H', '4', '2', '0'),  // unofficial fourcc
+      FOURCC_H422 = FOURCC('H', '4', '2', '2'),  // unofficial fourcc
+      FOURCC_U420 = FOURCC('U', '4', '2', '0'),  // bt.2020, unofficial fourcc
+      FOURCC_U422 = FOURCC('U', '4', '2', '2'),  // bt.2020, unofficial fourcc
+      FOURCC_U444 = FOURCC('U', '4', '4', '4'),  // bt.2020, unofficial fourcc
+
+      // 14 Auxiliary aliases.  CanonicalFourCC() maps these to canonical fourcc.
+      FOURCC_IYUV = FOURCC('I', 'Y', 'U', 'V'),  // Alias for I420.
+      FOURCC_YU16 = FOURCC('Y', 'U', '1', '6'),  // Alias for I422.
+      FOURCC_YU24 = FOURCC('Y', 'U', '2', '4'),  // Alias for I444.
+      FOURCC_YUYV = FOURCC('Y', 'U', 'Y', 'V'),  // Alias for YUY2.
+      FOURCC_YUVS = FOURCC('y', 'u', 'v', 's'),  // Alias for YUY2 on Mac.
+      FOURCC_HDYC = FOURCC('H', 'D', 'Y', 'C'),  // Alias for UYVY.
+      FOURCC_2VUY = FOURCC('2', 'v', 'u', 'y'),  // Alias for UYVY on Mac.
+      FOURCC_JPEG = FOURCC('J', 'P', 'E', 'G'),  // Alias for MJPG.
+      FOURCC_DMB1 = FOURCC('d', 'm', 'b', '1'),  // Alias for MJPG on Mac.
+      FOURCC_BA81 = FOURCC('B', 'A', '8', '1'),  // Alias for BGGR.
+      FOURCC_RGB3 = FOURCC('R', 'G', 'B', '3'),  // Alias for RAW.
+      FOURCC_BGR3 = FOURCC('B', 'G', 'R', '3'),  // Alias for 24BG.
+      FOURCC_CM32 = FOURCC(0, 0, 0, 32),  // Alias for BGRA kCMPixelFormat_32ARGB
+      FOURCC_CM24 = FOURCC(0, 0, 0, 24),  // Alias for RAW kCMPixelFormat_24RGB
+      FOURCC_L555 = FOURCC('L', '5', '5', '5'),  // Alias for RGBO.
+      FOURCC_L565 = FOURCC('L', '5', '6', '5'),  // Alias for RGBP.
+      FOURCC_5551 = FOURCC('5', '5', '5', '1'),  // Alias for RGBO.
+
+# Planar YUV
+      The following formats contains a full size Y plane followed by 1 or 2
+        planes for UV: I420, I422, I444, I400, NV21, NV12, I400
+      The size (subsampling) of the UV varies.
+        I420, NV12 and NV21 are half width, half height
+        I422, NV16 and NV61 are half width, full height
+        I444, NV24 and NV42 are full width, full height
+        I400 and J400 have no chroma channel.
+
+# Color space
+      The YUV formats start with a letter to specify the color space. e.g. I420
+        I = BT.601 limited range
+        J = BT.601 full range     (J = JPEG that uses this)
+        H = BT.709 limited range  (H for HD)
+        F = BT.709 full range     (F for Full range)
+        U = BT.2020 limited range (U for UHD)
+        V = BT.2020 full range
+        For YUV to RGB conversions, a matrix can be passed.  See also convert_argh.h
+
+# HDR formats
+      Planar formats with 10 or 12 bits use the following fourcc:
+        I010, I012, P010, P012 are half width, half height
+        I210, I212, P210, P212 are half width, full height
+        I410, I412, P410, P412 are full width, full height
+      where
+        I is the color space (see above) and 3 planes: Y, U and V.
+        P is a biplanar format, similar to NV12 but 16 bits, with the valid bits in the high bits.  There is a Y plane and a UV plane.
+        0, 2 or 4 is the last digit of subsampling: 4:2:0, 4:2:2, or 4:4:4
+        10 or 12 is the bits per channel.  The bits are in the low bits of a 16 bit channel.
+
+# The ARGB FOURCC
+
+There are 4 ARGB layouts - ARGB, BGRA, ABGR and RGBA.  ARGB is most common by far, used for screen formats, and windows webcam drivers.
+
+The fourcc describes the order of channels in a ***register***.
+
+A fourcc provided by capturer, can be thought of string, e.g. "ARGB".
+
+On little endian machines, as an int, this would have 'A' in the lowest byte.  The FOURCC macro reverses the order:
+
+    #define FOURCC(a, b, c, d) (((uint32)(a)) | ((uint32)(b) << 8) | ((uint32)(c) << 16) | ((uint32)(d) << 24))
+
+So the "ARGB" string, read as an uint32, is
+
+    FOURCC_ARGB = FOURCC('A', 'R', 'G', 'B')
+
+If you were to read ARGB pixels as uint32's, the alpha would be in the high byte, and the blue in the lowest byte.  In memory, these are stored little endian, so 'B' is first, then 'G', 'R' and 'A' last.
+
+When calling conversion functions, the names match the FOURCC, so in this case it would be I420ToARGB().
+
+All formats can be converted to/from ARGB.
+
+Most 'planar_functions' work on ARGB (e.g. ARGBBlend).
+
+Some are channel order agnostic (e.g. ARGBScale).
+
+Some functions are symmetric (e.g. ARGBToBGRA is the same as BGRAToARGB, so its a macro).
+
+ARGBBlend expects preattenuated ARGB. The R,G,B are premultiplied by alpha.  Other functions don't care.
+
+# RGB24 and RAW
+
+There are 2 RGB layouts - RGB24 (aka 24BG) and RAW
+
+RGB24 is B,G,R in memory
+RAW is R,G,B in memory
+
+# AR30 and XR30
+
+AR30 is 2 10 10 10 ARGB stored in little endian order.
+The 2 bit alpha has 4 values.  Here are the comparable 8 bit alpha values.
+0 - 0.    00000000b = 0x00 = 0
+1 - 33%.  01010101b = 0x55 = 85
+2 - 66%.  10101010b = 0xaa = 170
+3 - 100%. 11111111b = 0xff = 255
+The 10 bit RGB values range from 0 to 1023.
+XR30 is the same as AR30 but with no alpha channel.
+
+# AB64 and AR64
+
+AB64 is similar to ABGR, with 16 bit (2 bytes) per channel. Each channel stores an unsigned short.
+In memory R is the lowest and A is the highest.
+Each channel has value ranges from 0 to 65535.
+AR64 is similar to ARGB.
+
+# NV12 and NV21
+
+NV12 is a biplanar format with a full sized Y plane followed by a single
+chroma plane with weaved U and V values.
+NV21 is the same but with weaved V and U values.
+The 12 in NV12 refers to 12 bits per pixel.  NV12 has a half width and half
+height chroma channel, and therefore is a 420 subsampling.
+NV16 is 16 bits per pixel, with half width and full height.  aka 422.
+NV24 is 24 bits per pixel with full sized chroma channel. aka 444.
+Most NV12 functions allow the destination Y pointer to be NULL.
+
+# YUY2 and UYVY
+
+YUY2 is a packed YUV format with half width, full height.
+
+YUY2 is YUYV in memory
+UYVY is UYVY in memory

yuv.mod/libyuv/docs/getting_started.md

@@ -0,0 +1,326 @@
+# Getting Started
+
+How to get and build the libyuv code.
+
+## Pre-requisites
+
+You'll need to have depot tools installed: https://www.chromium.org/developers/how-tos/install-depot-tools
+Refer to chromium instructions for each platform for other prerequisites.
+
+## Getting the Code
+
+Create a working directory, enter it, and run:
+
+    gclient config --name src https://chromium.googlesource.com/libyuv/libyuv
+    gclient sync
+
+Then you'll get a .gclient file like:
+
+    solutions = [
+      { "name"        : "src",
+        "url"         : "https://chromium.googlesource.com/libyuv/libyuv",
+        "deps_file"   : "DEPS",
+        "managed"     : True,
+        "custom_deps" : {
+        },
+        "safesync_url": "",
+      },
+    ];
+
+For iOS add `;target_os=['ios'];` to your OSX .gclient and run `gclient sync.`
+
+Browse the Git reprository: https://chromium.googlesource.com/libyuv/libyuv/+/master
+
+### Android
+For Android add `;target_os=['android'];` to your Linux .gclient
+
+    solutions = [
+      { "name"        : "src",
+        "url"         : "https://chromium.googlesource.com/libyuv/libyuv",
+        "deps_file"   : "DEPS",
+        "managed"     : True,
+        "custom_deps" : {
+        },
+        "safesync_url": "",
+      },
+    ];
+    target_os = ["android", "linux"];
+
+Then run:
+
+    gclient sync
+
+To get just the source (not buildable):
+
+    git clone https://chromium.googlesource.com/libyuv/libyuv
+
+
+## Building the Library and Unittests
+
+### Windows
+
+    call gn gen out\Release "--args=is_debug=false target_cpu=\"x64\""
+    call gn gen out\Debug "--args=is_debug=true target_cpu=\"x64\""
+    ninja -v -C out\Release
+    ninja -v -C out\Debug
+
+    call gn gen out\Release "--args=is_debug=false target_cpu=\"x86\""
+    call gn gen out\Debug "--args=is_debug=true target_cpu=\"x86\""
+    ninja -v -C out\Release
+    ninja -v -C out\Debug
+
+### macOS and Linux
+
+    gn gen out/Release "--args=is_debug=false"
+    gn gen out/Debug "--args=is_debug=true"
+    ninja -v -C out/Release
+    ninja -v -C out/Debug
+
+### Building Offical with GN
+
+    gn gen out/Official "--args=is_debug=false is_official_build=true is_chrome_branded=true"
+    ninja -C out/Official
+
+### iOS
+http://www.chromium.org/developers/how-tos/build-instructions-ios
+
+Add to .gclient last line: `target_os=['ios'];`
+
+arm64
+
+    gn gen out/Release "--args=is_debug=false target_os=\"ios\" ios_enable_code_signing=false target_cpu=\"arm64\""
+    gn gen out/Debug "--args=is_debug=true target_os=\"ios\" ios_enable_code_signing=false target_cpu=\"arm64\""
+    ninja -v -C out/Debug libyuv_unittest
+    ninja -v -C out/Release libyuv_unittest
+
+ios simulator
+
+    gn gen out/Release "--args=is_debug=false target_os=\"ios\" ios_enable_code_signing=false use_xcode_clang=true target_cpu=\"x86\""
+    gn gen out/Debug "--args=is_debug=true target_os=\"ios\" ios_enable_code_signing=false use_xcode_clang=true target_cpu=\"x86\""
+    ninja -v -C out/Debug libyuv_unittest
+    ninja -v -C out/Release libyuv_unittest
+
+ios disassembly
+
+    otool -tV ./out/Release/obj/libyuv_neon/row_neon64.o >row_neon64.txt
+
+### Android
+https://code.google.com/p/chromium/wiki/AndroidBuildInstructions
+
+Add to .gclient last line: `target_os=['android'];`
+
+arm64
+
+    gn gen out/Release "--args=is_debug=false target_os=\"android\" target_cpu=\"arm64\""
+    gn gen out/Debug "--args=is_debug=true target_os=\"android\" target_cpu=\"arm64\""
+    ninja -v -C out/Debug libyuv_unittest
+    ninja -v -C out/Release libyuv_unittest
+
+armv7
+
+    gn gen out/Release "--args=is_debug=false target_os=\"android\" target_cpu=\"arm\""
+    gn gen out/Debug "--args=is_debug=true target_os=\"android\" target_cpu=\"arm\""
+    ninja -v -C out/Debug libyuv_unittest
+    ninja -v -C out/Release libyuv_unittest
+
+ia32
+
+    gn gen out/Release "--args=is_debug=false target_os=\"android\" target_cpu=\"x86\""
+    gn gen out/Debug "--args=is_debug=true target_os=\"android\" target_cpu=\"x86\""
+    ninja -v -C out/Debug libyuv_unittest
+    ninja -v -C out/Release libyuv_unittest
+
+mips
+
+    gn gen out/Release "--args=is_debug=false target_os=\"android\" target_cpu=\"mips64el\" mips_arch_variant=\"r6\" mips_use_msa=true is_component_build=true"
+    gn gen out/Debug "--args=is_debug=true target_os=\"android\" target_cpu=\"mips64el\" mips_arch_variant=\"r6\" mips_use_msa=true is_component_build=true"
+    ninja -v -C out/Debug libyuv_unittest
+    ninja -v -C out/Release libyuv_unittest
+
+arm disassembly:
+
+    llvm-objdump -d ./out/Release/obj/libyuv/row_common.o >row_common.txt
+
+    llvm-objdump -d ./out/Release/obj/libyuv_neon/row_neon.o >row_neon.txt
+
+    llvm-objdump -d ./out/Release/obj/libyuv_neon/row_neon64.o >row_neon64.txt
+
+    Caveat: Disassembly may require optimize_max be disabled in BUILD.gn
+
+Running tests:
+
+    out/Release/bin/run_libyuv_unittest -vv --gtest_filter=*
+
+Running test as benchmark:
+
+    out/Release/bin/run_libyuv_unittest -vv --gtest_filter=* --libyuv_width=1280 --libyuv_height=720 --libyuv_repeat=999 --libyuv_flags=-1  --libyuv_cpu_info=-1
+
+Running test with C code:
+
+    out/Release/bin/run_libyuv_unittest -vv --gtest_filter=* --libyuv_width=1280 --libyuv_height=720 --libyuv_repeat=999 --libyuv_flags=1 --libyuv_cpu_info=1
+
+### Build targets
+
+    ninja -C out/Debug libyuv
+    ninja -C out/Debug libyuv_unittest
+    ninja -C out/Debug compare
+    ninja -C out/Debug yuvconvert
+    ninja -C out/Debug yuvconstants
+    ninja -C out/Debug psnr
+    ninja -C out/Debug cpuid
+
+### ARM Linux
+
+    gn gen out/Release "--args=is_debug=false target_cpu=\"arm64\""
+    gn gen out/Debug "--args=is_debug=true target_cpu=\"arm64\""
+    ninja -v -C out/Debug libyuv_unittest
+    ninja -v -C out/Release libyuv_unittest
+
+### MIPS Linux
+
+mips
+
+   gn gen out/Release "--args=is_debug=false target_os=\"linux\" target_cpu=\"mips64el\" mips_arch_variant=\"loongson3\" is_component_build=false use_sysroot=false use_gold=false"
+   gn gen out/Debug "--args=is_debug=true target_os=\"linux\" target_cpu=\"mips64el\" mips_arch_variant=\"loongson3\" is_component_build=false use_sysroot=false use_gold=false"
+   ninja -v -C out/Debug libyuv_unittest
+   ninja -v -C out/Release libyuv_unittest
+
+## Building the Library with make
+
+### Linux
+
+    make V=1 -f linux.mk
+    make V=1 -f linux.mk clean
+    make V=1 -f linux.mk CXX=clang++ CC=clang
+
+## Building the library with cmake
+
+Install cmake: http://www.cmake.org/
+
+### Default debug build:
+
+    mkdir out
+    cd out
+    cmake ..
+    cmake --build .
+
+### Release build/install
+
+    mkdir out
+    cd out
+    cmake -DCMAKE_INSTALL_PREFIX="/usr/lib" -DCMAKE_BUILD_TYPE="Release" ..
+    cmake --build . --config Release
+    sudo cmake --build . --target install --config Release
+
+### Build RPM/DEB packages
+
+    mkdir out
+    cd out
+    cmake -DCMAKE_BUILD_TYPE=Release ..
+    make -j4
+    make package
+
+## Building RISC-V target with cmake
+
+### Prerequisite: build risc-v clang toolchain and qemu
+
+If you don't have prebuilt clang and riscv64 qemu, run the script to download source and build them.
+
+    ./riscv_script/prepare_toolchain_qemu.sh
+
+After running script, clang & qemu are built in `build-toolchain-qemu/riscv-clang/` & `build-toolchain-qemu/riscv-qemu/`.
+
+### Cross-compile for RISC-V target
+    cmake -B out/Release/ -DUNIT_TEST=ON \
+          -DCMAKE_BUILD_TYPE=Release \
+          -DCMAKE_TOOLCHAIN_FILE="./riscv_script/riscv-clang.cmake" \
+          -DTOOLCHAIN_PATH={TOOLCHAIN_PATH} \
+          -DUSE_RVV=ON .
+    cmake --build out/Release/
+
+#### Customized Compiler Flags
+
+Customized compiler flags are supported by `-DRISCV_COMPILER_FLAGS="xxx"`.
+If `-DRISCV_COMPILER_FLAGS="xxx"` is manually assigned, other compile flags(e.g disable -march=xxx) will not be appended.
+
+Example:
+
+    cmake -B out/Release/ -DUNIT_TEST=ON \
+          -DCMAKE_BUILD_TYPE=Release \
+          -DCMAKE_TOOLCHAIN_FILE="./riscv_script/riscv-clang.cmake" \
+          -DRISCV_COMPILER_FLAGS="-mcpu=sifive-x280" \
+          .
+
+### Run on QEMU
+
+#### Run libyuv_unittest on QEMU
+    cd out/Release/
+    USE_RVV=ON \
+    TOOLCHAIN_PATH={TOOLCHAIN_PATH} \
+    QEMU_PREFIX_PATH={QEMU_PREFIX_PATH} \
+    ../../riscv_script/run_qemu.sh libyuv_unittest
+
+
+## Setup for Arm Cross compile
+
+See also https://www.ccoderun.ca/programming/2015-12-20_CrossCompiling/index.html
+
+    sudo apt-get install ssh dkms build-essential linux-headers-generic
+    sudo apt-get install kdevelop cmake git subversion
+    sudo apt-get install graphviz doxygen doxygen-gui
+    sudo apt-get install manpages manpages-dev manpages-posix manpages-posix-dev
+    sudo apt-get install libboost-all-dev libboost-dev libssl-dev
+    sudo apt-get install rpm terminator fish
+    sudo apt-get install g++-arm-linux-gnueabihf gcc-arm-linux-gnueabihf
+
+### Build psnr tool
+
+    cd util
+    arm-linux-gnueabihf-g++ psnr_main.cc psnr.cc ssim.cc -o psnr
+    arm-linux-gnueabihf-objdump -d psnr
+
+## Running Unittests
+
+### Windows
+
+    out\Release\libyuv_unittest.exe --gtest_catch_exceptions=0 --gtest_filter="*"
+
+### macOS and Linux
+
+    out/Release/libyuv_unittest --gtest_filter="*"
+
+Replace --gtest_filter="*" with specific unittest to run.  May include wildcards.
+    out/Release/libyuv_unittest --gtest_filter=*I420ToARGB_Opt
+
+## CPU Emulator tools
+
+### Intel SDE (Software Development Emulator)
+
+Pre-requisite: Install IntelSDE: http://software.intel.com/en-us/articles/intel-software-development-emulator
+
+Then run:
+
+    c:\intelsde\sde -hsw -- out\Release\libyuv_unittest.exe --gtest_filter=*
+
+    ~/intelsde/sde -skx -- out/Release/libyuv_unittest --gtest_filter=**I420ToARGB_Opt
+
+### Intel Architecture Code Analyzer
+
+Inset these 2 macros into assembly code to be analyzed:
+    IACA_ASM_START
+    IACA_ASM_END
+Build the code as usual, then run iaca on the object file.
+    ~/iaca-lin64/bin/iaca.sh -reduceout -arch HSW out/Release/obj/libyuv_internal/compare_gcc.o
+
+## Sanitizers
+
+    gn gen out/Release "--args=is_debug=false is_msan=true"
+    ninja -v -C out/Release
+
+Sanitizers available: asan, msan, tsan, ubsan, lsan, ubsan_vptr
+
+### Running Dr Memory memcheck for Windows
+
+Pre-requisite: Install Dr Memory for Windows and add it to your path: http://www.drmemory.org/docs/page_install_windows.html
+
+    drmemory out\Debug\libyuv_unittest.exe --gtest_catch_exceptions=0 --gtest_filter=*

yuv.mod/libyuv/docs/rotation.md

@@ -0,0 +1,107 @@
+# Introduction
+
+Rotation by multiplies of 90 degrees allows mobile devices to rotate webcams from landscape to portrait.  The higher level functions ConvertToI420 and ConvertToARGB allow rotation of any format.  Optimized functionality is supported for I420, ARGB, NV12 and NV21.
+
+# ConvertToI420
+
+    int ConvertToI420(const uint8* src_frame, size_t src_size,
+                      uint8* dst_y, int dst_stride_y,
+                      uint8* dst_u, int dst_stride_u,
+                      uint8* dst_v, int dst_stride_v,
+                      int crop_x, int crop_y,
+                      int src_width, int src_height,
+                      int crop_width, int crop_height,
+                      enum RotationMode rotation,
+                      uint32 format);
+
+This function crops, converts, and rotates.  You should think of it in that order.
+  * Crops the original image, which is src_width x src_height, to crop_width x crop_height.  At this point the image is still not rotated.
+  * Converts the cropped region to I420.  Supports inverted source for src_height negative.
+  * Rotates by 90, 180 or 270 degrees.
+The buffer the caller provides should account for rotation.  Be especially important to get stride of the destination correct.
+
+e.g.
+640 x 480 NV12 captured<br>
+Crop to 640 x 360<br>
+Rotate by 90 degrees to 360 x 640.<br>
+Caller passes stride of 360 for Y and 360 / 2 for U and V.<br>
+Caller passes crop_width of 640, crop_height of 360.<br>
+
+# ConvertToARGB
+
+    int ConvertToARGB(const uint8* src_frame, size_t src_size,
+                      uint8* dst_argb, int dst_stride_argb,
+                      int crop_x, int crop_y,
+                      int src_width, int src_height,
+                      int crop_width, int crop_height,
+                      enum RotationMode rotation,
+                      uint32 format);
+
+Same as I420, but implementation is less optimized - reads columns and writes rows, 16 bytes at a time.
+
+# I420Rotate
+
+    int I420Rotate(const uint8* src_y, int src_stride_y,
+                   const uint8* src_u, int src_stride_u,
+                   const uint8* src_v, int src_stride_v,
+                   uint8* dst_y, int dst_stride_y,
+                   uint8* dst_u, int dst_stride_u,
+                   uint8* dst_v, int dst_stride_v,
+                   int src_width, int src_height, enum RotationMode mode);
+
+Destination is rotated, so pass dst_stride_y etc that consider rotation.<br>
+Rotate by 180 can be done in place, but 90 and 270 can not.
+
+Implementation (Neon/SSE2) uses 8 x 8 block transpose, so best efficiency is with sizes and pointers that are aligned to 8.
+
+Cropping can be achieved by adjusting the src_y/u/v pointers and src_width, src_height.
+
+Lower level plane functions are provided, allowing other planar formats to be rotated.  (e.g. I444)
+
+For other planar YUV formats (I444, I422, I411, I400, NV16, NV24), the planar functions are exposed and can be called directly
+
+
+    // Rotate a plane by 0, 90, 180, or 270.
+    int RotatePlane(const uint8* src, int src_stride,
+                    uint8* dst, int dst_stride,
+                    int src_width, int src_height, enum RotationMode mode);
+
+# ARGBRotate
+
+    LIBYUV_API
+    int ARGBRotate(const uint8* src_argb, int src_stride_argb,
+                   uint8* dst_argb, int dst_stride_argb,
+                   int src_width, int src_height, enum RotationMode mode);
+
+Same as I420, but implementation is less optimized - reads columns and writes rows.
+
+Rotate by 90, or any angle, can be achieved using ARGBAffine.
+
+# Mirror - Horizontal Flip
+
+Mirror functions for horizontally flipping an image, which can be useful for 'self view' of a webcam.
+
+    int I420Mirror(const uint8* src_y, int src_stride_y,
+                   const uint8* src_u, int src_stride_u,
+                   const uint8* src_v, int src_stride_v,
+                   uint8* dst_y, int dst_stride_y,
+                   uint8* dst_u, int dst_stride_u,
+                   uint8* dst_v, int dst_stride_v,
+                   int width, int height);
+    int ARGBMirror(const uint8* src_argb, int src_stride_argb,
+                   uint8* dst_argb, int dst_stride_argb,
+                   int width, int height);
+
+Mirror functionality can also be achieved with the I420Scale and ARGBScale functions by passing negative width and/or height.
+
+# Invert - Vertical Flip
+
+Inverting can be achieved with almost any libyuv function by passing a negative source height.
+
+I420Mirror and ARGBMirror can also be used to rotate by 180 degrees by passing a negative height.
+
+# Cropping - Vertical Flip
+
+When cropping from a subsampled format like NV21, the method of setting the start pointers wont work for odd crop start y on the UV plane.
+If the height after cropping will be odd, invert the source - point to the last row, negate the strides, and pass negative height, which
+will re-invert the image as the conversion outputs.

yuv.mod/libyuv/download_vs_toolchain.py

@@ -0,0 +1,29 @@
+#!/usr/bin/env vpython3
+
+# Copyright 2014 The LibYuv Project Authors. All rights reserved.
+#
+# Use of this source code is governed by a BSD-style license
+# that can be found in the LICENSE file in the root of the source
+# tree. An additional intellectual property rights grant can be found
+# in the file PATENTS. All contributing project authors may
+# be found in the AUTHORS file in the root of the source tree.
+
+# This script is used to run the vs_toolchain.py script to download the
+# Visual Studio toolchain. It's just a temporary measure while waiting for the
+# Chrome team to move find_depot_tools into src/build to get rid of these
+# workarounds (similar one in gyp_libyuv).
+
+import os
+import sys
+
+
+checkout_root = os.path.dirname(os.path.realpath(__file__))
+sys.path.insert(0, os.path.join(checkout_root, 'build'))
+sys.path.insert(0, os.path.join(checkout_root, 'tools', 'find_depot_tools'))
+
+
+import vs_toolchain  # pylint: disable=wrong-import-position
+
+
+if __name__ == '__main__':
+  sys.exit(vs_toolchain.main())

yuv.mod/libyuv/include/libyuv.h

@@ -0,0 +1,33 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_H_
+#define INCLUDE_LIBYUV_H_
+
+#include "libyuv/basic_types.h"
+#include "libyuv/compare.h"
+#include "libyuv/convert.h"
+#include "libyuv/convert_argb.h"
+#include "libyuv/convert_from.h"
+#include "libyuv/convert_from_argb.h"
+#include "libyuv/cpu_id.h"
+#include "libyuv/mjpeg_decoder.h"
+#include "libyuv/planar_functions.h"
+#include "libyuv/rotate.h"
+#include "libyuv/rotate_argb.h"
+#include "libyuv/row.h"
+#include "libyuv/scale.h"
+#include "libyuv/scale_argb.h"
+#include "libyuv/scale_row.h"
+#include "libyuv/scale_uv.h"
+#include "libyuv/version.h"
+#include "libyuv/video_common.h"
+
+#endif  // INCLUDE_LIBYUV_H_

yuv.mod/libyuv/include/libyuv/basic_types.h

@@ -0,0 +1,68 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_BASIC_TYPES_H_
+#define INCLUDE_LIBYUV_BASIC_TYPES_H_
+
+#include <stddef.h>  // For size_t and NULL
+
+#if !defined(INT_TYPES_DEFINED) && !defined(GG_LONGLONG)
+#define INT_TYPES_DEFINED
+
+#if defined(_MSC_VER) && (_MSC_VER < 1600)
+#include <sys/types.h>  // for uintptr_t on x86
+typedef unsigned __int64 uint64_t;
+typedef __int64 int64_t;
+typedef unsigned int uint32_t;
+typedef int int32_t;
+typedef unsigned short uint16_t;
+typedef short int16_t;
+typedef unsigned char uint8_t;
+typedef signed char int8_t;
+#else
+#include <stdint.h>  // for uintptr_t and C99 types
+#endif               // defined(_MSC_VER) && (_MSC_VER < 1600)
+// Types are deprecated.  Enable this macro for legacy types.
+#ifdef LIBYUV_LEGACY_TYPES
+typedef uint64_t uint64;
+typedef int64_t int64;
+typedef uint32_t uint32;
+typedef int32_t int32;
+typedef uint16_t uint16;
+typedef int16_t int16;
+typedef uint8_t uint8;
+typedef int8_t int8;
+#endif  // LIBYUV_LEGACY_TYPES
+#endif  // INT_TYPES_DEFINED
+
+#if !defined(LIBYUV_API)
+#if defined(_WIN32) || defined(__CYGWIN__)
+#if defined(LIBYUV_BUILDING_SHARED_LIBRARY)
+#define LIBYUV_API __declspec(dllexport)
+#elif defined(LIBYUV_USING_SHARED_LIBRARY)
+#define LIBYUV_API __declspec(dllimport)
+#else
+#define LIBYUV_API
+#endif  // LIBYUV_BUILDING_SHARED_LIBRARY
+#elif defined(__GNUC__) && (__GNUC__ >= 4) && !defined(__APPLE__) && \
+    (defined(LIBYUV_BUILDING_SHARED_LIBRARY) ||                      \
+     defined(LIBYUV_USING_SHARED_LIBRARY))
+#define LIBYUV_API __attribute__((visibility("default")))
+#else
+#define LIBYUV_API
+#endif  // __GNUC__
+#endif  // LIBYUV_API
+
+// TODO(fbarchard): Remove bool macros.
+#define LIBYUV_BOOL int
+#define LIBYUV_FALSE 0
+#define LIBYUV_TRUE 1
+
+#endif  // INCLUDE_LIBYUV_BASIC_TYPES_H_

yuv.mod/libyuv/include/libyuv/compare.h

@@ -0,0 +1,111 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_COMPARE_H_
+#define INCLUDE_LIBYUV_COMPARE_H_
+
+#include "libyuv/basic_types.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Compute a hash for specified memory. Seed of 5381 recommended.
+LIBYUV_API
+uint32_t HashDjb2(const uint8_t* src, uint64_t count, uint32_t seed);
+
+// Hamming Distance
+LIBYUV_API
+uint64_t ComputeHammingDistance(const uint8_t* src_a,
+                                const uint8_t* src_b,
+                                int count);
+
+// Scan an opaque argb image and return fourcc based on alpha offset.
+// Returns FOURCC_ARGB, FOURCC_BGRA, or 0 if unknown.
+LIBYUV_API
+uint32_t ARGBDetect(const uint8_t* argb,
+                    int stride_argb,
+                    int width,
+                    int height);
+
+// Sum Square Error - used to compute Mean Square Error or PSNR.
+LIBYUV_API
+uint64_t ComputeSumSquareError(const uint8_t* src_a,
+                               const uint8_t* src_b,
+                               int count);
+
+LIBYUV_API
+uint64_t ComputeSumSquareErrorPlane(const uint8_t* src_a,
+                                    int stride_a,
+                                    const uint8_t* src_b,
+                                    int stride_b,
+                                    int width,
+                                    int height);
+
+static const int kMaxPsnr = 128;
+
+LIBYUV_API
+double SumSquareErrorToPsnr(uint64_t sse, uint64_t count);
+
+LIBYUV_API
+double CalcFramePsnr(const uint8_t* src_a,
+                     int stride_a,
+                     const uint8_t* src_b,
+                     int stride_b,
+                     int width,
+                     int height);
+
+LIBYUV_API
+double I420Psnr(const uint8_t* src_y_a,
+                int stride_y_a,
+                const uint8_t* src_u_a,
+                int stride_u_a,
+                const uint8_t* src_v_a,
+                int stride_v_a,
+                const uint8_t* src_y_b,
+                int stride_y_b,
+                const uint8_t* src_u_b,
+                int stride_u_b,
+                const uint8_t* src_v_b,
+                int stride_v_b,
+                int width,
+                int height);
+
+LIBYUV_API
+double CalcFrameSsim(const uint8_t* src_a,
+                     int stride_a,
+                     const uint8_t* src_b,
+                     int stride_b,
+                     int width,
+                     int height);
+
+LIBYUV_API
+double I420Ssim(const uint8_t* src_y_a,
+                int stride_y_a,
+                const uint8_t* src_u_a,
+                int stride_u_a,
+                const uint8_t* src_v_a,
+                int stride_v_a,
+                const uint8_t* src_y_b,
+                int stride_y_b,
+                const uint8_t* src_u_b,
+                int stride_u_b,
+                const uint8_t* src_v_b,
+                int stride_v_b,
+                int width,
+                int height);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_COMPARE_H_

yuv.mod/libyuv/include/libyuv/compare_row.h

@@ -0,0 +1,148 @@
+/*
+ *  Copyright 2013 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_COMPARE_ROW_H_
+#define INCLUDE_LIBYUV_COMPARE_ROW_H_
+
+#include "libyuv/basic_types.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+#if defined(__pnacl__) || defined(__CLR_VER) ||            \
+    (defined(__native_client__) && defined(__x86_64__)) || \
+    (defined(__i386__) && !defined(__SSE__) && !defined(__clang__))
+#define LIBYUV_DISABLE_X86
+#endif
+#if defined(__native_client__)
+#define LIBYUV_DISABLE_NEON
+#endif
+// MemorySanitizer does not support assembly code yet. http://crbug.com/344505
+#if defined(__has_feature)
+#if __has_feature(memory_sanitizer) && !defined(LIBYUV_DISABLE_NEON)
+#define LIBYUV_DISABLE_NEON
+#endif
+#if __has_feature(memory_sanitizer) && !defined(LIBYUV_DISABLE_X86)
+#define LIBYUV_DISABLE_X86
+#endif
+#endif
+// Visual C 2012 required for AVX2.
+#if defined(_M_IX86) && !defined(__clang__) && defined(_MSC_VER) && \
+    _MSC_VER >= 1700
+#define VISUALC_HAS_AVX2 1
+#endif  // VisualStudio >= 2012
+
+// clang >= 3.4.0 required for AVX2.
+#if defined(__clang__) && (defined(__x86_64__) || defined(__i386__))
+#if (__clang_major__ > 3) || (__clang_major__ == 3 && (__clang_minor__ >= 4))
+#define CLANG_HAS_AVX2 1
+#endif  // clang >= 3.4
+#endif  // __clang__
+
+// The following are available for Visual C and GCC:
+#if !defined(LIBYUV_DISABLE_X86) && \
+    (defined(__x86_64__) || defined(__i386__) || defined(_M_IX86))
+#define HAS_HASHDJB2_SSE41
+#define HAS_SUMSQUAREERROR_SSE2
+#define HAS_HAMMINGDISTANCE_SSE42
+#endif
+
+// The following are available for Visual C and clangcl 32 bit:
+#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && defined(_MSC_VER) && \
+    !defined(__clang__) &&                                                   \
+    (defined(VISUALC_HAS_AVX2) || defined(CLANG_HAS_AVX2))
+#define HAS_HASHDJB2_AVX2
+#define HAS_SUMSQUAREERROR_AVX2
+#endif
+
+// The following are available for GCC and clangcl:
+#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__))
+#define HAS_HAMMINGDISTANCE_SSSE3
+#endif
+
+// The following are available for GCC and clangcl:
+#if !defined(LIBYUV_DISABLE_X86) && defined(CLANG_HAS_AVX2) && \
+    (defined(__x86_64__) || defined(__i386__))
+#define HAS_HAMMINGDISTANCE_AVX2
+#endif
+
+// The following are available for Neon:
+#if !defined(LIBYUV_DISABLE_NEON) && \
+    (defined(__ARM_NEON__) || defined(LIBYUV_NEON) || defined(__aarch64__))
+#define HAS_HAMMINGDISTANCE_NEON
+#define HAS_SUMSQUAREERROR_NEON
+#endif
+
+// The following are available for AArch64 Neon:
+#if !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__)
+#define HAS_HASHDJB2_NEON
+
+#define HAS_HAMMINGDISTANCE_NEON_DOTPROD
+#define HAS_SUMSQUAREERROR_NEON_DOTPROD
+#endif
+
+#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
+#define HAS_HAMMINGDISTANCE_MSA
+#define HAS_SUMSQUAREERROR_MSA
+#endif
+
+uint32_t HammingDistance_C(const uint8_t* src_a,
+                           const uint8_t* src_b,
+                           int count);
+uint32_t HammingDistance_SSE42(const uint8_t* src_a,
+                               const uint8_t* src_b,
+                               int count);
+uint32_t HammingDistance_SSSE3(const uint8_t* src_a,
+                               const uint8_t* src_b,
+                               int count);
+uint32_t HammingDistance_AVX2(const uint8_t* src_a,
+                              const uint8_t* src_b,
+                              int count);
+uint32_t HammingDistance_NEON(const uint8_t* src_a,
+                              const uint8_t* src_b,
+                              int count);
+uint32_t HammingDistance_NEON_DotProd(const uint8_t* src_a,
+                                      const uint8_t* src_b,
+                                      int count);
+uint32_t HammingDistance_MSA(const uint8_t* src_a,
+                             const uint8_t* src_b,
+                             int count);
+uint32_t SumSquareError_C(const uint8_t* src_a,
+                          const uint8_t* src_b,
+                          int count);
+uint32_t SumSquareError_SSE2(const uint8_t* src_a,
+                             const uint8_t* src_b,
+                             int count);
+uint32_t SumSquareError_AVX2(const uint8_t* src_a,
+                             const uint8_t* src_b,
+                             int count);
+uint32_t SumSquareError_NEON(const uint8_t* src_a,
+                             const uint8_t* src_b,
+                             int count);
+uint32_t SumSquareError_NEON_DotProd(const uint8_t* src_a,
+                                     const uint8_t* src_b,
+                                     int count);
+uint32_t SumSquareError_MSA(const uint8_t* src_a,
+                            const uint8_t* src_b,
+                            int count);
+
+uint32_t HashDjb2_C(const uint8_t* src, int count, uint32_t seed);
+uint32_t HashDjb2_SSE41(const uint8_t* src, int count, uint32_t seed);
+uint32_t HashDjb2_AVX2(const uint8_t* src, int count, uint32_t seed);
+uint32_t HashDjb2_NEON(const uint8_t* src, int count, uint32_t seed);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_COMPARE_ROW_H_

yuv.mod/libyuv/include/libyuv/convert.h

@@ -0,0 +1,1074 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_CONVERT_H_
+#define INCLUDE_LIBYUV_CONVERT_H_
+
+#include "libyuv/basic_types.h"
+
+#include "libyuv/rotate.h"  // For enum RotationMode.
+
+// TODO(fbarchard): fix WebRTC source to include following libyuv headers:
+#include "libyuv/convert_argb.h"      // For WebRTC I420ToARGB. b/620
+#include "libyuv/convert_from.h"      // For WebRTC ConvertFromI420. b/620
+#include "libyuv/planar_functions.h"  // For WebRTC I420Rect, CopyPlane. b/618
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Convert I444 to I420.
+LIBYUV_API
+int I444ToI420(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert I444 to NV12.
+LIBYUV_API
+int I444ToNV12(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert I444 to NV21.
+LIBYUV_API
+int I444ToNV21(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height);
+
+// Convert I422 to I420.
+LIBYUV_API
+int I422ToI420(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert I422 to I444.
+LIBYUV_API
+int I422ToI444(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert I422 to I210.
+LIBYUV_API
+int I422ToI210(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert MM21 to NV12.
+LIBYUV_API
+int MM21ToNV12(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert MM21 to I420.
+LIBYUV_API
+int MM21ToI420(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert MM21 to YUY2
+LIBYUV_API
+int MM21ToYUY2(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_yuy2,
+               int dst_stride_yuy2,
+               int width,
+               int height);
+
+// Convert MT2T to P010
+// Note that src_y and src_uv point to packed 10-bit values, so the Y plane will
+// be 10 / 8 times the dimensions of the image. Also for this reason,
+// src_stride_y and src_stride_uv are given in bytes.
+LIBYUV_API
+int MT2TToP010(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert I422 to NV21.
+LIBYUV_API
+int I422ToNV21(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height);
+
+// Copy I420 to I420.
+#define I420ToI420 I420Copy
+LIBYUV_API
+int I420Copy(const uint8_t* src_y,
+             int src_stride_y,
+             const uint8_t* src_u,
+             int src_stride_u,
+             const uint8_t* src_v,
+             int src_stride_v,
+             uint8_t* dst_y,
+             int dst_stride_y,
+             uint8_t* dst_u,
+             int dst_stride_u,
+             uint8_t* dst_v,
+             int dst_stride_v,
+             int width,
+             int height);
+
+// Convert I420 to I444.
+LIBYUV_API
+int I420ToI444(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Copy I010 to I010
+#define I010ToI010 I010Copy
+#define H010ToH010 I010Copy
+LIBYUV_API
+int I010Copy(const uint16_t* src_y,
+             int src_stride_y,
+             const uint16_t* src_u,
+             int src_stride_u,
+             const uint16_t* src_v,
+             int src_stride_v,
+             uint16_t* dst_y,
+             int dst_stride_y,
+             uint16_t* dst_u,
+             int dst_stride_u,
+             uint16_t* dst_v,
+             int dst_stride_v,
+             int width,
+             int height);
+
+// Convert 10 bit YUV to 8 bit
+#define H010ToH420 I010ToI420
+LIBYUV_API
+int I010ToI420(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+#define H210ToH420 I210ToI420
+LIBYUV_API
+int I210ToI420(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+#define H210ToH422 I210ToI422
+LIBYUV_API
+int I210ToI422(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+#define H410ToH420 I410ToI420
+LIBYUV_API
+int I410ToI420(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+#define H410ToH444 I410ToI444
+LIBYUV_API
+int I410ToI444(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+#define H012ToH420 I012ToI420
+LIBYUV_API
+int I012ToI420(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+#define H212ToH422 I212ToI422
+LIBYUV_API
+int I212ToI422(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+#define H212ToH420 I212ToI420
+LIBYUV_API
+int I212ToI420(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+#define H412ToH444 I412ToI444
+LIBYUV_API
+int I412ToI444(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+#define H412ToH420 I412ToI420
+LIBYUV_API
+int I412ToI420(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert 10 bit P010 to 8 bit NV12.
+// dst_y can be NULL
+LIBYUV_API
+int P010ToNV12(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+#define I412ToI012 I410ToI010
+#define H410ToH010 I410ToI010
+#define H412ToH012 I410ToI010
+LIBYUV_API
+int I410ToI010(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+#define I212ToI012 I210ToI010
+#define H210ToH010 I210ToI010
+#define H212ToH012 I210ToI010
+LIBYUV_API
+int I210ToI010(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert I010 to I410
+LIBYUV_API
+int I010ToI410(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert I012 to I412
+#define I012ToI412 I010ToI410
+
+// Convert I210 to I410
+LIBYUV_API
+int I210ToI410(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert I212 to I412
+#define I212ToI412 I210ToI410
+
+// Convert I010 to P010
+LIBYUV_API
+int I010ToP010(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert 10 bit YUV I010 to NV12
+LIBYUV_API
+int I010ToNV12(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert I210 to P210
+LIBYUV_API
+int I210ToP210(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert I012 to P012
+LIBYUV_API
+int I012ToP012(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert I212 to P212
+LIBYUV_API
+int I212ToP212(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert I400 (grey) to I420.
+LIBYUV_API
+int I400ToI420(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert I400 (grey) to NV21.
+LIBYUV_API
+int I400ToNV21(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height);
+
+#define J400ToJ420 I400ToI420
+
+// Convert NV12 to I420.
+LIBYUV_API
+int NV12ToI420(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert NV21 to I420.
+LIBYUV_API
+int NV21ToI420(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_vu,
+               int src_stride_vu,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert NV12 to NV24.
+LIBYUV_API
+int NV12ToNV24(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert NV16 to NV24.
+LIBYUV_API
+int NV16ToNV24(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert P010 to I010.
+LIBYUV_API
+int P010ToI010(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_uv,
+               int src_stride_uv,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert P012 to I012.
+LIBYUV_API
+int P012ToI012(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_uv,
+               int src_stride_uv,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert P010 to P410.
+LIBYUV_API
+int P010ToP410(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_uv,
+               int src_stride_uv,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert P012 to P412.
+#define P012ToP412 P010ToP410
+
+// Convert P016 to P416.
+#define P016ToP416 P010ToP410
+
+// Convert P210 to P410.
+LIBYUV_API
+int P210ToP410(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_uv,
+               int src_stride_uv,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert P212 to P412.
+#define P212ToP412 P210ToP410
+
+// Convert P216 to P416.
+#define P216ToP416 P210ToP410
+
+// Convert YUY2 to I420.
+LIBYUV_API
+int YUY2ToI420(const uint8_t* src_yuy2,
+               int src_stride_yuy2,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert UYVY to I420.
+LIBYUV_API
+int UYVYToI420(const uint8_t* src_uyvy,
+               int src_stride_uyvy,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert AYUV to NV12.
+LIBYUV_API
+int AYUVToNV12(const uint8_t* src_ayuv,
+               int src_stride_ayuv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert AYUV to NV21.
+LIBYUV_API
+int AYUVToNV21(const uint8_t* src_ayuv,
+               int src_stride_ayuv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height);
+
+// Convert Android420 to I420.
+LIBYUV_API
+int Android420ToI420(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     int src_pixel_stride_uv,
+                     uint8_t* dst_y,
+                     int dst_stride_y,
+                     uint8_t* dst_u,
+                     int dst_stride_u,
+                     uint8_t* dst_v,
+                     int dst_stride_v,
+                     int width,
+                     int height);
+
+// ARGB little endian (bgra in memory) to I420.
+LIBYUV_API
+int ARGBToI420(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert ARGB to I420 with Alpha
+LIBYUV_API
+int ARGBToI420Alpha(const uint8_t* src_argb,
+                    int src_stride_argb,
+                    uint8_t* dst_y,
+                    int dst_stride_y,
+                    uint8_t* dst_u,
+                    int dst_stride_u,
+                    uint8_t* dst_v,
+                    int dst_stride_v,
+                    uint8_t* dst_a,
+                    int dst_stride_a,
+                    int width,
+                    int height);
+
+// BGRA little endian (argb in memory) to I420.
+LIBYUV_API
+int BGRAToI420(const uint8_t* src_bgra,
+               int src_stride_bgra,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// ABGR little endian (rgba in memory) to I420.
+LIBYUV_API
+int ABGRToI420(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// RGBA little endian (abgr in memory) to I420.
+LIBYUV_API
+int RGBAToI420(const uint8_t* src_rgba,
+               int src_stride_rgba,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// RGB little endian (bgr in memory) to I420.
+LIBYUV_API
+int RGB24ToI420(const uint8_t* src_rgb24,
+                int src_stride_rgb24,
+                uint8_t* dst_y,
+                int dst_stride_y,
+                uint8_t* dst_u,
+                int dst_stride_u,
+                uint8_t* dst_v,
+                int dst_stride_v,
+                int width,
+                int height);
+
+// RGB little endian (bgr in memory) to J420.
+LIBYUV_API
+int RGB24ToJ420(const uint8_t* src_rgb24,
+                int src_stride_rgb24,
+                uint8_t* dst_y,
+                int dst_stride_y,
+                uint8_t* dst_u,
+                int dst_stride_u,
+                uint8_t* dst_v,
+                int dst_stride_v,
+                int width,
+                int height);
+
+// RGB big endian (rgb in memory) to I420.
+LIBYUV_API
+int RAWToI420(const uint8_t* src_raw,
+              int src_stride_raw,
+              uint8_t* dst_y,
+              int dst_stride_y,
+              uint8_t* dst_u,
+              int dst_stride_u,
+              uint8_t* dst_v,
+              int dst_stride_v,
+              int width,
+              int height);
+
+// RGB big endian (rgb in memory) to J420.
+LIBYUV_API
+int RAWToJ420(const uint8_t* src_raw,
+              int src_stride_raw,
+              uint8_t* dst_y,
+              int dst_stride_y,
+              uint8_t* dst_u,
+              int dst_stride_u,
+              uint8_t* dst_v,
+              int dst_stride_v,
+              int width,
+              int height);
+
+// RGB16 (RGBP fourcc) little endian to I420.
+LIBYUV_API
+int RGB565ToI420(const uint8_t* src_rgb565,
+                 int src_stride_rgb565,
+                 uint8_t* dst_y,
+                 int dst_stride_y,
+                 uint8_t* dst_u,
+                 int dst_stride_u,
+                 uint8_t* dst_v,
+                 int dst_stride_v,
+                 int width,
+                 int height);
+
+// RGB15 (RGBO fourcc) little endian to I420.
+LIBYUV_API
+int ARGB1555ToI420(const uint8_t* src_argb1555,
+                   int src_stride_argb1555,
+                   uint8_t* dst_y,
+                   int dst_stride_y,
+                   uint8_t* dst_u,
+                   int dst_stride_u,
+                   uint8_t* dst_v,
+                   int dst_stride_v,
+                   int width,
+                   int height);
+
+// RGB12 (R444 fourcc) little endian to I420.
+LIBYUV_API
+int ARGB4444ToI420(const uint8_t* src_argb4444,
+                   int src_stride_argb4444,
+                   uint8_t* dst_y,
+                   int dst_stride_y,
+                   uint8_t* dst_u,
+                   int dst_stride_u,
+                   uint8_t* dst_v,
+                   int dst_stride_v,
+                   int width,
+                   int height);
+
+// RGB little endian (bgr in memory) to J400.
+LIBYUV_API
+int RGB24ToJ400(const uint8_t* src_rgb24,
+                int src_stride_rgb24,
+                uint8_t* dst_yj,
+                int dst_stride_yj,
+                int width,
+                int height);
+
+// RGB big endian (rgb in memory) to J400.
+LIBYUV_API
+int RAWToJ400(const uint8_t* src_raw,
+              int src_stride_raw,
+              uint8_t* dst_yj,
+              int dst_stride_yj,
+              int width,
+              int height);
+
+// src_width/height provided by capture.
+// dst_width/height for clipping determine final size.
+LIBYUV_API
+int MJPGToI420(const uint8_t* sample,
+               size_t sample_size,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int src_width,
+               int src_height,
+               int dst_width,
+               int dst_height);
+
+// JPEG to NV21
+LIBYUV_API
+int MJPGToNV21(const uint8_t* sample,
+               size_t sample_size,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int src_width,
+               int src_height,
+               int dst_width,
+               int dst_height);
+
+// JPEG to NV12
+LIBYUV_API
+int MJPGToNV12(const uint8_t* sample,
+               size_t sample_size,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int src_width,
+               int src_height,
+               int dst_width,
+               int dst_height);
+
+// Query size of MJPG in pixels.
+LIBYUV_API
+int MJPGSize(const uint8_t* sample,
+             size_t sample_size,
+             int* width,
+             int* height);
+
+// Convert camera sample to I420 with cropping, rotation and vertical flip.
+// "src_size" is needed to parse MJPG.
+// "dst_stride_y" number of bytes in a row of the dst_y plane.
+//   Normally this would be the same as dst_width, with recommended alignment
+//   to 16 bytes for better efficiency.
+//   If rotation of 90 or 270 is used, stride is affected. The caller should
+//   allocate the I420 buffer according to rotation.
+// "dst_stride_u" number of bytes in a row of the dst_u plane.
+//   Normally this would be the same as (dst_width + 1) / 2, with
+//   recommended alignment to 16 bytes for better efficiency.
+//   If rotation of 90 or 270 is used, stride is affected.
+// "crop_x" and "crop_y" are starting position for cropping.
+//   To center, crop_x = (src_width - dst_width) / 2
+//              crop_y = (src_height - dst_height) / 2
+// "src_width" / "src_height" is size of src_frame in pixels.
+//   "src_height" can be negative indicating a vertically flipped image source.
+// "crop_width" / "crop_height" is the size to crop the src to.
+//    Must be less than or equal to src_width/src_height
+//    Cropping parameters are pre-rotation.
+// "rotation" can be 0, 90, 180 or 270.
+// "fourcc" is a fourcc. ie 'I420', 'YUY2'
+// Returns 0 for successful; -1 for invalid parameter. Non-zero for failure.
+LIBYUV_API
+int ConvertToI420(const uint8_t* sample,
+                  size_t sample_size,
+                  uint8_t* dst_y,
+                  int dst_stride_y,
+                  uint8_t* dst_u,
+                  int dst_stride_u,
+                  uint8_t* dst_v,
+                  int dst_stride_v,
+                  int crop_x,
+                  int crop_y,
+                  int src_width,
+                  int src_height,
+                  int crop_width,
+                  int crop_height,
+                  enum RotationMode rotation,
+                  uint32_t fourcc);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_CONVERT_H_

yuv.mod/libyuv/include/libyuv/convert_argb.h

@@ -0,0 +1,2335 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_CONVERT_ARGB_H_
+#define INCLUDE_LIBYUV_CONVERT_ARGB_H_
+
+#include "libyuv/basic_types.h"
+
+#include "libyuv/rotate.h"  // For enum RotationMode.
+#include "libyuv/scale.h"   // For enum FilterMode.
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Conversion matrix for YUV to RGB
+LIBYUV_API extern const struct YuvConstants kYuvI601Constants;   // BT.601
+LIBYUV_API extern const struct YuvConstants kYuvJPEGConstants;   // BT.601 full
+LIBYUV_API extern const struct YuvConstants kYuvH709Constants;   // BT.709
+LIBYUV_API extern const struct YuvConstants kYuvF709Constants;   // BT.709 full
+LIBYUV_API extern const struct YuvConstants kYuv2020Constants;   // BT.2020
+LIBYUV_API extern const struct YuvConstants kYuvV2020Constants;  // BT.2020 full
+
+// Conversion matrix for YVU to BGR
+LIBYUV_API extern const struct YuvConstants kYvuI601Constants;   // BT.601
+LIBYUV_API extern const struct YuvConstants kYvuJPEGConstants;   // BT.601 full
+LIBYUV_API extern const struct YuvConstants kYvuH709Constants;   // BT.709
+LIBYUV_API extern const struct YuvConstants kYvuF709Constants;   // BT.709 full
+LIBYUV_API extern const struct YuvConstants kYvu2020Constants;   // BT.2020
+LIBYUV_API extern const struct YuvConstants kYvuV2020Constants;  // BT.2020 full
+
+// Macros for end swapped destination Matrix conversions.
+// Swap UV and pass mirrored kYvuJPEGConstants matrix.
+// TODO(fbarchard): Add macro for each Matrix function.
+#define kYuvI601ConstantsVU kYvuI601Constants
+#define kYuvJPEGConstantsVU kYvuJPEGConstants
+#define kYuvH709ConstantsVU kYvuH709Constants
+#define kYuvF709ConstantsVU kYvuF709Constants
+#define kYuv2020ConstantsVU kYvu2020Constants
+#define kYuvV2020ConstantsVU kYvuV2020Constants
+
+#define NV12ToABGRMatrix(a, b, c, d, e, f, g, h, i) \
+  NV21ToARGBMatrix(a, b, c, d, e, f, g##VU, h, i)
+#define NV21ToABGRMatrix(a, b, c, d, e, f, g, h, i) \
+  NV12ToARGBMatrix(a, b, c, d, e, f, g##VU, h, i)
+#define NV12ToRAWMatrix(a, b, c, d, e, f, g, h, i) \
+  NV21ToRGB24Matrix(a, b, c, d, e, f, g##VU, h, i)
+#define NV21ToRAWMatrix(a, b, c, d, e, f, g, h, i) \
+  NV12ToRGB24Matrix(a, b, c, d, e, f, g##VU, h, i)
+#define I010ToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k) \
+  I010ToARGBMatrix(a, b, e, f, c, d, g, h, i##VU, j, k)
+#define I210ToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k) \
+  I210ToARGBMatrix(a, b, e, f, c, d, g, h, i##VU, j, k)
+#define I410ToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k) \
+  I410ToARGBMatrix(a, b, e, f, c, d, g, h, i##VU, j, k)
+#define I010ToAB30Matrix(a, b, c, d, e, f, g, h, i, j, k) \
+  I010ToAR30Matrix(a, b, e, f, c, d, g, h, i##VU, j, k)
+#define I210ToAB30Matrix(a, b, c, d, e, f, g, h, i, j, k) \
+  I210ToAR30Matrix(a, b, e, f, c, d, g, h, i##VU, j, k)
+#define I410ToAB30Matrix(a, b, c, d, e, f, g, h, i, j, k) \
+  I410ToAR30Matrix(a, b, e, f, c, d, g, h, i##VU, j, k)
+#define I012ToAB30Matrix(a, b, c, d, e, f, g, h, i, j, k) \
+  I012ToAR30Matrix(a, b, e, f, c, d, g, h, i##VU, j, k)
+#define I420AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k, l, m, n) \
+  I420AlphaToARGBMatrix(a, b, e, f, c, d, g, h, i, j, k##VU, l, m, n)
+#define I422AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k, l, m, n) \
+  I422AlphaToARGBMatrix(a, b, e, f, c, d, g, h, i, j, k##VU, l, m, n)
+#define I444AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k, l, m, n) \
+  I444AlphaToARGBMatrix(a, b, e, f, c, d, g, h, i, j, k##VU, l, m, n)
+#define I010AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k, l, m, n) \
+  I010AlphaToARGBMatrix(a, b, e, f, c, d, g, h, i, j, k##VU, l, m, n)
+#define I210AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k, l, m, n) \
+  I210AlphaToARGBMatrix(a, b, e, f, c, d, g, h, i, j, k##VU, l, m, n)
+#define I410AlphaToABGRMatrix(a, b, c, d, e, f, g, h, i, j, k, l, m, n) \
+  I410AlphaToARGBMatrix(a, b, e, f, c, d, g, h, i, j, k##VU, l, m, n)
+
+// Alias.
+#define ARGBToARGB ARGBCopy
+
+// Copy ARGB to ARGB.
+LIBYUV_API
+int ARGBCopy(const uint8_t* src_argb,
+             int src_stride_argb,
+             uint8_t* dst_argb,
+             int dst_stride_argb,
+             int width,
+             int height);
+
+// Convert I420 to ARGB.
+LIBYUV_API
+int I420ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert I420 to ABGR.
+LIBYUV_API
+int I420ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert J420 to ARGB.
+LIBYUV_API
+int J420ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert J420 to ABGR.
+LIBYUV_API
+int J420ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert H420 to ARGB.
+LIBYUV_API
+int H420ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert H420 to ABGR.
+LIBYUV_API
+int H420ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert U420 to ARGB.
+LIBYUV_API
+int U420ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert U420 to ABGR.
+LIBYUV_API
+int U420ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert I422 to ARGB.
+LIBYUV_API
+int I422ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert I422 to ABGR.
+LIBYUV_API
+int I422ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert J422 to ARGB.
+LIBYUV_API
+int J422ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert J422 to ABGR.
+LIBYUV_API
+int J422ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert H422 to ARGB.
+LIBYUV_API
+int H422ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert H422 to ABGR.
+LIBYUV_API
+int H422ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert U422 to ARGB.
+LIBYUV_API
+int U422ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert U422 to ABGR.
+LIBYUV_API
+int U422ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert I444 to ARGB.
+LIBYUV_API
+int I444ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert I444 to ABGR.
+LIBYUV_API
+int I444ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert J444 to ARGB.
+LIBYUV_API
+int J444ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert J444 to ABGR.
+LIBYUV_API
+int J444ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert H444 to ARGB.
+LIBYUV_API
+int H444ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert H444 to ABGR.
+LIBYUV_API
+int H444ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert U444 to ARGB.
+LIBYUV_API
+int U444ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert U444 to ABGR.
+LIBYUV_API
+int U444ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert I444 to RGB24.
+LIBYUV_API
+int I444ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_u,
+                int src_stride_u,
+                const uint8_t* src_v,
+                int src_stride_v,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height);
+
+// Convert I444 to RAW.
+LIBYUV_API
+int I444ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height);
+
+// Convert I010 to ARGB.
+LIBYUV_API
+int I010ToARGB(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert I010 to ABGR.
+LIBYUV_API
+int I010ToABGR(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert H010 to ARGB.
+LIBYUV_API
+int H010ToARGB(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert H010 to ABGR.
+LIBYUV_API
+int H010ToABGR(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert U010 to ARGB.
+LIBYUV_API
+int U010ToARGB(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert U010 to ABGR.
+LIBYUV_API
+int U010ToABGR(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert I210 to ARGB.
+LIBYUV_API
+int I210ToARGB(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert I210 to ABGR.
+LIBYUV_API
+int I210ToABGR(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert H210 to ARGB.
+LIBYUV_API
+int H210ToARGB(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert H210 to ABGR.
+LIBYUV_API
+int H210ToABGR(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert U210 to ARGB.
+LIBYUV_API
+int U210ToARGB(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert U210 to ABGR.
+LIBYUV_API
+int U210ToABGR(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert I420 with Alpha to preattenuated ARGB.
+LIBYUV_API
+int I420AlphaToARGB(const uint8_t* src_y,
+                    int src_stride_y,
+                    const uint8_t* src_u,
+                    int src_stride_u,
+                    const uint8_t* src_v,
+                    int src_stride_v,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    int width,
+                    int height,
+                    int attenuate);
+
+// Convert I420 with Alpha to preattenuated ABGR.
+LIBYUV_API
+int I420AlphaToABGR(const uint8_t* src_y,
+                    int src_stride_y,
+                    const uint8_t* src_u,
+                    int src_stride_u,
+                    const uint8_t* src_v,
+                    int src_stride_v,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_abgr,
+                    int dst_stride_abgr,
+                    int width,
+                    int height,
+                    int attenuate);
+
+// Convert I422 with Alpha to preattenuated ARGB.
+LIBYUV_API
+int I422AlphaToARGB(const uint8_t* src_y,
+                    int src_stride_y,
+                    const uint8_t* src_u,
+                    int src_stride_u,
+                    const uint8_t* src_v,
+                    int src_stride_v,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    int width,
+                    int height,
+                    int attenuate);
+
+// Convert I422 with Alpha to preattenuated ABGR.
+LIBYUV_API
+int I422AlphaToABGR(const uint8_t* src_y,
+                    int src_stride_y,
+                    const uint8_t* src_u,
+                    int src_stride_u,
+                    const uint8_t* src_v,
+                    int src_stride_v,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_abgr,
+                    int dst_stride_abgr,
+                    int width,
+                    int height,
+                    int attenuate);
+
+// Convert I444 with Alpha to preattenuated ARGB.
+LIBYUV_API
+int I444AlphaToARGB(const uint8_t* src_y,
+                    int src_stride_y,
+                    const uint8_t* src_u,
+                    int src_stride_u,
+                    const uint8_t* src_v,
+                    int src_stride_v,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    int width,
+                    int height,
+                    int attenuate);
+
+// Convert I444 with Alpha to preattenuated ABGR.
+LIBYUV_API
+int I444AlphaToABGR(const uint8_t* src_y,
+                    int src_stride_y,
+                    const uint8_t* src_u,
+                    int src_stride_u,
+                    const uint8_t* src_v,
+                    int src_stride_v,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_abgr,
+                    int dst_stride_abgr,
+                    int width,
+                    int height,
+                    int attenuate);
+
+// Convert I400 (grey) to ARGB.  Reverse of ARGBToI400.
+LIBYUV_API
+int I400ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert J400 (jpeg grey) to ARGB.
+LIBYUV_API
+int J400ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Alias.
+#define YToARGB I400ToARGB
+
+// Convert NV12 to ARGB.
+LIBYUV_API
+int NV12ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert NV21 to ARGB.
+LIBYUV_API
+int NV21ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_vu,
+               int src_stride_vu,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert NV12 to ABGR.
+LIBYUV_API
+int NV12ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert NV21 to ABGR.
+LIBYUV_API
+int NV21ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_vu,
+               int src_stride_vu,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert NV12 to RGB24.
+LIBYUV_API
+int NV12ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_uv,
+                int src_stride_uv,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height);
+
+// Convert NV21 to RGB24.
+LIBYUV_API
+int NV21ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_vu,
+                int src_stride_vu,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height);
+
+// Convert NV21 to YUV24.
+LIBYUV_API
+int NV21ToYUV24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_vu,
+                int src_stride_vu,
+                uint8_t* dst_yuv24,
+                int dst_stride_yuv24,
+                int width,
+                int height);
+
+// Convert NV12 to RAW.
+LIBYUV_API
+int NV12ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_uv,
+              int src_stride_uv,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height);
+
+// Convert NV21 to RAW.
+LIBYUV_API
+int NV21ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_vu,
+              int src_stride_vu,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height);
+
+// Convert YUY2 to ARGB.
+LIBYUV_API
+int YUY2ToARGB(const uint8_t* src_yuy2,
+               int src_stride_yuy2,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert UYVY to ARGB.
+LIBYUV_API
+int UYVYToARGB(const uint8_t* src_uyvy,
+               int src_stride_uyvy,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert I010 to AR30.
+LIBYUV_API
+int I010ToAR30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height);
+
+// Convert H010 to AR30.
+LIBYUV_API
+int H010ToAR30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height);
+
+// Convert I010 to AB30.
+LIBYUV_API
+int I010ToAB30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height);
+
+// Convert H010 to AB30.
+LIBYUV_API
+int H010ToAB30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height);
+
+// Convert U010 to AR30.
+LIBYUV_API
+int U010ToAR30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height);
+
+// Convert U010 to AB30.
+LIBYUV_API
+int U010ToAB30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height);
+
+// Convert I210 to AR30.
+LIBYUV_API
+int I210ToAR30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height);
+
+// Convert I210 to AB30.
+LIBYUV_API
+int I210ToAB30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height);
+
+// Convert H210 to AR30.
+LIBYUV_API
+int H210ToAR30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height);
+
+// Convert H210 to AB30.
+LIBYUV_API
+int H210ToAB30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height);
+
+// Convert U210 to AR30.
+LIBYUV_API
+int U210ToAR30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height);
+
+// Convert U210 to AB30.
+LIBYUV_API
+int U210ToAB30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height);
+
+// BGRA little endian (argb in memory) to ARGB.
+LIBYUV_API
+int BGRAToARGB(const uint8_t* src_bgra,
+               int src_stride_bgra,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// ABGR little endian (rgba in memory) to ARGB.
+LIBYUV_API
+int ABGRToARGB(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// RGBA little endian (abgr in memory) to ARGB.
+LIBYUV_API
+int RGBAToARGB(const uint8_t* src_rgba,
+               int src_stride_rgba,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Deprecated function name.
+#define BG24ToARGB RGB24ToARGB
+
+// RGB little endian (bgr in memory) to ARGB.
+LIBYUV_API
+int RGB24ToARGB(const uint8_t* src_rgb24,
+                int src_stride_rgb24,
+                uint8_t* dst_argb,
+                int dst_stride_argb,
+                int width,
+                int height);
+
+// RGB big endian (rgb in memory) to ARGB.
+LIBYUV_API
+int RAWToARGB(const uint8_t* src_raw,
+              int src_stride_raw,
+              uint8_t* dst_argb,
+              int dst_stride_argb,
+              int width,
+              int height);
+
+// RGB big endian (rgb in memory) to RGBA.
+LIBYUV_API
+int RAWToRGBA(const uint8_t* src_raw,
+              int src_stride_raw,
+              uint8_t* dst_rgba,
+              int dst_stride_rgba,
+              int width,
+              int height);
+
+// RGB16 (RGBP fourcc) little endian to ARGB.
+LIBYUV_API
+int RGB565ToARGB(const uint8_t* src_rgb565,
+                 int src_stride_rgb565,
+                 uint8_t* dst_argb,
+                 int dst_stride_argb,
+                 int width,
+                 int height);
+
+// RGB15 (RGBO fourcc) little endian to ARGB.
+LIBYUV_API
+int ARGB1555ToARGB(const uint8_t* src_argb1555,
+                   int src_stride_argb1555,
+                   uint8_t* dst_argb,
+                   int dst_stride_argb,
+                   int width,
+                   int height);
+
+// RGB12 (R444 fourcc) little endian to ARGB.
+LIBYUV_API
+int ARGB4444ToARGB(const uint8_t* src_argb4444,
+                   int src_stride_argb4444,
+                   uint8_t* dst_argb,
+                   int dst_stride_argb,
+                   int width,
+                   int height);
+
+// Aliases
+#define AB30ToARGB AR30ToABGR
+#define AB30ToABGR AR30ToARGB
+#define AB30ToAR30 AR30ToAB30
+
+// Convert AR30 To ARGB.
+LIBYUV_API
+int AR30ToARGB(const uint8_t* src_ar30,
+               int src_stride_ar30,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert AR30 To ABGR.
+LIBYUV_API
+int AR30ToABGR(const uint8_t* src_ar30,
+               int src_stride_ar30,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert AR30 To AB30.
+LIBYUV_API
+int AR30ToAB30(const uint8_t* src_ar30,
+               int src_stride_ar30,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height);
+
+// Convert AR64 to ARGB.
+LIBYUV_API
+int AR64ToARGB(const uint16_t* src_ar64,
+               int src_stride_ar64,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert AB64 to ABGR.
+#define AB64ToABGR AR64ToARGB
+
+// Convert AB64 to ARGB.
+LIBYUV_API
+int AB64ToARGB(const uint16_t* src_ab64,
+               int src_stride_ab64,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert AR64 to ABGR.
+#define AR64ToABGR AB64ToARGB
+
+// Convert AR64 To AB64.
+LIBYUV_API
+int AR64ToAB64(const uint16_t* src_ar64,
+               int src_stride_ar64,
+               uint16_t* dst_ab64,
+               int dst_stride_ab64,
+               int width,
+               int height);
+
+// Convert AB64 To AR64.
+#define AB64ToAR64 AR64ToAB64
+
+// src_width/height provided by capture
+// dst_width/height for clipping determine final size.
+LIBYUV_API
+int MJPGToARGB(const uint8_t* sample,
+               size_t sample_size,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int src_width,
+               int src_height,
+               int dst_width,
+               int dst_height);
+
+// Convert Android420 to ARGB.
+LIBYUV_API
+int Android420ToARGB(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     int src_pixel_stride_uv,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     int width,
+                     int height);
+
+// Convert Android420 to ABGR.
+LIBYUV_API
+int Android420ToABGR(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     int src_pixel_stride_uv,
+                     uint8_t* dst_abgr,
+                     int dst_stride_abgr,
+                     int width,
+                     int height);
+
+// Convert NV12 to RGB565.
+LIBYUV_API
+int NV12ToRGB565(const uint8_t* src_y,
+                 int src_stride_y,
+                 const uint8_t* src_uv,
+                 int src_stride_uv,
+                 uint8_t* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height);
+
+// Convert I422 to BGRA.
+LIBYUV_API
+int I422ToBGRA(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_bgra,
+               int dst_stride_bgra,
+               int width,
+               int height);
+
+// Convert I422 to ABGR.
+LIBYUV_API
+int I422ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert I422 to RGBA.
+LIBYUV_API
+int I422ToRGBA(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_rgba,
+               int dst_stride_rgba,
+               int width,
+               int height);
+
+LIBYUV_API
+int I420ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+LIBYUV_API
+int I420ToBGRA(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_bgra,
+               int dst_stride_bgra,
+               int width,
+               int height);
+
+LIBYUV_API
+int I420ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+LIBYUV_API
+int I420ToRGBA(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_rgba,
+               int dst_stride_rgba,
+               int width,
+               int height);
+
+LIBYUV_API
+int I420ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_u,
+                int src_stride_u,
+                const uint8_t* src_v,
+                int src_stride_v,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height);
+
+LIBYUV_API
+int I420ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height);
+
+LIBYUV_API
+int H420ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_u,
+                int src_stride_u,
+                const uint8_t* src_v,
+                int src_stride_v,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height);
+
+LIBYUV_API
+int H420ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height);
+
+LIBYUV_API
+int J420ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_u,
+                int src_stride_u,
+                const uint8_t* src_v,
+                int src_stride_v,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height);
+
+LIBYUV_API
+int J420ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height);
+
+// Convert I422 to RGB24.
+LIBYUV_API
+int I422ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_u,
+                int src_stride_u,
+                const uint8_t* src_v,
+                int src_stride_v,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height);
+
+// Convert I422 to RAW.
+LIBYUV_API
+int I422ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height);
+
+LIBYUV_API
+int I420ToRGB565(const uint8_t* src_y,
+                 int src_stride_y,
+                 const uint8_t* src_u,
+                 int src_stride_u,
+                 const uint8_t* src_v,
+                 int src_stride_v,
+                 uint8_t* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height);
+
+LIBYUV_API
+int J420ToRGB565(const uint8_t* src_y,
+                 int src_stride_y,
+                 const uint8_t* src_u,
+                 int src_stride_u,
+                 const uint8_t* src_v,
+                 int src_stride_v,
+                 uint8_t* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height);
+
+LIBYUV_API
+int H420ToRGB565(const uint8_t* src_y,
+                 int src_stride_y,
+                 const uint8_t* src_u,
+                 int src_stride_u,
+                 const uint8_t* src_v,
+                 int src_stride_v,
+                 uint8_t* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height);
+
+LIBYUV_API
+int I422ToRGB565(const uint8_t* src_y,
+                 int src_stride_y,
+                 const uint8_t* src_u,
+                 int src_stride_u,
+                 const uint8_t* src_v,
+                 int src_stride_v,
+                 uint8_t* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height);
+
+// Convert I420 To RGB565 with 4x4 dither matrix (16 bytes).
+// Values in dither matrix from 0 to 7 recommended.
+// The order of the dither matrix is first byte is upper left.
+
+LIBYUV_API
+int I420ToRGB565Dither(const uint8_t* src_y,
+                       int src_stride_y,
+                       const uint8_t* src_u,
+                       int src_stride_u,
+                       const uint8_t* src_v,
+                       int src_stride_v,
+                       uint8_t* dst_rgb565,
+                       int dst_stride_rgb565,
+                       const uint8_t* dither4x4,
+                       int width,
+                       int height);
+
+LIBYUV_API
+int I420ToARGB1555(const uint8_t* src_y,
+                   int src_stride_y,
+                   const uint8_t* src_u,
+                   int src_stride_u,
+                   const uint8_t* src_v,
+                   int src_stride_v,
+                   uint8_t* dst_argb1555,
+                   int dst_stride_argb1555,
+                   int width,
+                   int height);
+
+LIBYUV_API
+int I420ToARGB4444(const uint8_t* src_y,
+                   int src_stride_y,
+                   const uint8_t* src_u,
+                   int src_stride_u,
+                   const uint8_t* src_v,
+                   int src_stride_v,
+                   uint8_t* dst_argb4444,
+                   int dst_stride_argb4444,
+                   int width,
+                   int height);
+
+// Convert I420 to AR30.
+LIBYUV_API
+int I420ToAR30(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height);
+
+// Convert I420 to AB30.
+LIBYUV_API
+int I420ToAB30(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height);
+
+// Convert H420 to AR30.
+LIBYUV_API
+int H420ToAR30(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height);
+
+// Convert H420 to AB30.
+LIBYUV_API
+int H420ToAB30(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height);
+
+// Convert I420 to ARGB with matrix.
+LIBYUV_API
+int I420ToARGBMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert I422 to ARGB with matrix.
+LIBYUV_API
+int I422ToARGBMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert I444 to ARGB with matrix.
+LIBYUV_API
+int I444ToARGBMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert I444 to RGB24 with matrix.
+LIBYUV_API
+int I444ToRGB24Matrix(const uint8_t* src_y,
+                      int src_stride_y,
+                      const uint8_t* src_u,
+                      int src_stride_u,
+                      const uint8_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_rgb24,
+                      int dst_stride_rgb24,
+                      const struct YuvConstants* yuvconstants,
+                      int width,
+                      int height);
+
+// Convert 10 bit 420 YUV to ARGB with matrix.
+LIBYUV_API
+int I010ToAR30Matrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert 10 bit 420 YUV to ARGB with matrix.
+LIBYUV_API
+int I210ToAR30Matrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert 10 bit 444 YUV to ARGB with matrix.
+LIBYUV_API
+int I410ToAR30Matrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert 10 bit YUV to ARGB with matrix.
+LIBYUV_API
+int I010ToARGBMatrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// multiply 12 bit yuv into high bits to allow any number of bits.
+LIBYUV_API
+int I012ToAR30Matrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert 12 bit YUV to ARGB with matrix.
+LIBYUV_API
+int I012ToARGBMatrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert 10 bit 422 YUV to ARGB with matrix.
+LIBYUV_API
+int I210ToARGBMatrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert 10 bit 444 YUV to ARGB with matrix.
+LIBYUV_API
+int I410ToARGBMatrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert P010 to ARGB with matrix.
+LIBYUV_API
+int P010ToARGBMatrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_uv,
+                     int src_stride_uv,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert P210 to ARGB with matrix.
+LIBYUV_API
+int P210ToARGBMatrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_uv,
+                     int src_stride_uv,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert P010 to AR30 with matrix.
+LIBYUV_API
+int P010ToAR30Matrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_uv,
+                     int src_stride_uv,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert P210 to AR30 with matrix.
+LIBYUV_API
+int P210ToAR30Matrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_uv,
+                     int src_stride_uv,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// P012 and P010 use most significant bits so the conversion is the same.
+// Convert P012 to ARGB with matrix.
+#define P012ToARGBMatrix P010ToARGBMatrix
+// Convert P012 to AR30 with matrix.
+#define P012ToAR30Matrix P010ToAR30Matrix
+// Convert P212 to ARGB with matrix.
+#define P212ToARGBMatrix P210ToARGBMatrix
+// Convert P212 to AR30 with matrix.
+#define P212ToAR30Matrix P210ToAR30Matrix
+
+// Convert P016 to ARGB with matrix.
+#define P016ToARGBMatrix P010ToARGBMatrix
+// Convert P016 to AR30 with matrix.
+#define P016ToAR30Matrix P010ToAR30Matrix
+// Convert P216 to ARGB with matrix.
+#define P216ToARGBMatrix P210ToARGBMatrix
+// Convert P216 to AR30 with matrix.
+#define P216ToAR30Matrix P210ToAR30Matrix
+
+// Convert I420 with Alpha to preattenuated ARGB with matrix.
+LIBYUV_API
+int I420AlphaToARGBMatrix(const uint8_t* src_y,
+                          int src_stride_y,
+                          const uint8_t* src_u,
+                          int src_stride_u,
+                          const uint8_t* src_v,
+                          int src_stride_v,
+                          const uint8_t* src_a,
+                          int src_stride_a,
+                          uint8_t* dst_argb,
+                          int dst_stride_argb,
+                          const struct YuvConstants* yuvconstants,
+                          int width,
+                          int height,
+                          int attenuate);
+
+// Convert I422 with Alpha to preattenuated ARGB with matrix.
+LIBYUV_API
+int I422AlphaToARGBMatrix(const uint8_t* src_y,
+                          int src_stride_y,
+                          const uint8_t* src_u,
+                          int src_stride_u,
+                          const uint8_t* src_v,
+                          int src_stride_v,
+                          const uint8_t* src_a,
+                          int src_stride_a,
+                          uint8_t* dst_argb,
+                          int dst_stride_argb,
+                          const struct YuvConstants* yuvconstants,
+                          int width,
+                          int height,
+                          int attenuate);
+
+// Convert I444 with Alpha to preattenuated ARGB with matrix.
+LIBYUV_API
+int I444AlphaToARGBMatrix(const uint8_t* src_y,
+                          int src_stride_y,
+                          const uint8_t* src_u,
+                          int src_stride_u,
+                          const uint8_t* src_v,
+                          int src_stride_v,
+                          const uint8_t* src_a,
+                          int src_stride_a,
+                          uint8_t* dst_argb,
+                          int dst_stride_argb,
+                          const struct YuvConstants* yuvconstants,
+                          int width,
+                          int height,
+                          int attenuate);
+
+// Convert I010 with Alpha to preattenuated ARGB with matrix.
+LIBYUV_API
+int I010AlphaToARGBMatrix(const uint16_t* src_y,
+                          int src_stride_y,
+                          const uint16_t* src_u,
+                          int src_stride_u,
+                          const uint16_t* src_v,
+                          int src_stride_v,
+                          const uint16_t* src_a,
+                          int src_stride_a,
+                          uint8_t* dst_argb,
+                          int dst_stride_argb,
+                          const struct YuvConstants* yuvconstants,
+                          int width,
+                          int height,
+                          int attenuate);
+
+// Convert I210 with Alpha to preattenuated ARGB with matrix.
+LIBYUV_API
+int I210AlphaToARGBMatrix(const uint16_t* src_y,
+                          int src_stride_y,
+                          const uint16_t* src_u,
+                          int src_stride_u,
+                          const uint16_t* src_v,
+                          int src_stride_v,
+                          const uint16_t* src_a,
+                          int src_stride_a,
+                          uint8_t* dst_argb,
+                          int dst_stride_argb,
+                          const struct YuvConstants* yuvconstants,
+                          int width,
+                          int height,
+                          int attenuate);
+
+// Convert I410 with Alpha to preattenuated ARGB with matrix.
+LIBYUV_API
+int I410AlphaToARGBMatrix(const uint16_t* src_y,
+                          int src_stride_y,
+                          const uint16_t* src_u,
+                          int src_stride_u,
+                          const uint16_t* src_v,
+                          int src_stride_v,
+                          const uint16_t* src_a,
+                          int src_stride_a,
+                          uint8_t* dst_argb,
+                          int dst_stride_argb,
+                          const struct YuvConstants* yuvconstants,
+                          int width,
+                          int height,
+                          int attenuate);
+
+// Convert NV12 to ARGB with matrix.
+LIBYUV_API
+int NV12ToARGBMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_uv,
+                     int src_stride_uv,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert NV21 to ARGB with matrix.
+LIBYUV_API
+int NV21ToARGBMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_vu,
+                     int src_stride_vu,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert NV12 to RGB565 with matrix.
+LIBYUV_API
+int NV12ToRGB565Matrix(const uint8_t* src_y,
+                       int src_stride_y,
+                       const uint8_t* src_uv,
+                       int src_stride_uv,
+                       uint8_t* dst_rgb565,
+                       int dst_stride_rgb565,
+                       const struct YuvConstants* yuvconstants,
+                       int width,
+                       int height);
+
+// Convert NV12 to RGB24 with matrix.
+LIBYUV_API
+int NV12ToRGB24Matrix(const uint8_t* src_y,
+                      int src_stride_y,
+                      const uint8_t* src_uv,
+                      int src_stride_uv,
+                      uint8_t* dst_rgb24,
+                      int dst_stride_rgb24,
+                      const struct YuvConstants* yuvconstants,
+                      int width,
+                      int height);
+
+// Convert NV21 to RGB24 with matrix.
+LIBYUV_API
+int NV21ToRGB24Matrix(const uint8_t* src_y,
+                      int src_stride_y,
+                      const uint8_t* src_vu,
+                      int src_stride_vu,
+                      uint8_t* dst_rgb24,
+                      int dst_stride_rgb24,
+                      const struct YuvConstants* yuvconstants,
+                      int width,
+                      int height);
+
+// Convert YUY2 to ARGB with matrix.
+LIBYUV_API
+int YUY2ToARGBMatrix(const uint8_t* src_yuy2,
+                     int src_stride_yuy2,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert UYVY to ARGB with matrix.
+LIBYUV_API
+int UYVYToARGBMatrix(const uint8_t* src_uyvy,
+                     int src_stride_uyvy,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert Android420 to ARGB with matrix.
+LIBYUV_API
+int Android420ToARGBMatrix(const uint8_t* src_y,
+                           int src_stride_y,
+                           const uint8_t* src_u,
+                           int src_stride_u,
+                           const uint8_t* src_v,
+                           int src_stride_v,
+                           int src_pixel_stride_uv,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height);
+
+// Convert I422 to RGBA with matrix.
+LIBYUV_API
+int I422ToRGBAMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_rgba,
+                     int dst_stride_rgba,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert I420 to RGBA with matrix.
+LIBYUV_API
+int I420ToRGBAMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_rgba,
+                     int dst_stride_rgba,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert I420 to RGB24 with matrix.
+LIBYUV_API
+int I420ToRGB24Matrix(const uint8_t* src_y,
+                      int src_stride_y,
+                      const uint8_t* src_u,
+                      int src_stride_u,
+                      const uint8_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_rgb24,
+                      int dst_stride_rgb24,
+                      const struct YuvConstants* yuvconstants,
+                      int width,
+                      int height);
+
+// Convert I422 to RGB24 with matrix.
+LIBYUV_API
+int I422ToRGB24Matrix(const uint8_t* src_y,
+                      int src_stride_y,
+                      const uint8_t* src_u,
+                      int src_stride_u,
+                      const uint8_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_rgb24,
+                      int dst_stride_rgb24,
+                      const struct YuvConstants* yuvconstants,
+                      int width,
+                      int height);
+
+// Convert I420 to RGB565 with specified color matrix.
+LIBYUV_API
+int I420ToRGB565Matrix(const uint8_t* src_y,
+                       int src_stride_y,
+                       const uint8_t* src_u,
+                       int src_stride_u,
+                       const uint8_t* src_v,
+                       int src_stride_v,
+                       uint8_t* dst_rgb565,
+                       int dst_stride_rgb565,
+                       const struct YuvConstants* yuvconstants,
+                       int width,
+                       int height);
+
+// Convert I422 to RGB565 with specified color matrix.
+LIBYUV_API
+int I422ToRGB565Matrix(const uint8_t* src_y,
+                       int src_stride_y,
+                       const uint8_t* src_u,
+                       int src_stride_u,
+                       const uint8_t* src_v,
+                       int src_stride_v,
+                       uint8_t* dst_rgb565,
+                       int dst_stride_rgb565,
+                       const struct YuvConstants* yuvconstants,
+                       int width,
+                       int height);
+
+// Convert I420 to AR30 with matrix.
+LIBYUV_API
+int I420ToAR30Matrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert I400 (grey) to ARGB.  Reverse of ARGBToI400.
+LIBYUV_API
+int I400ToARGBMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height);
+
+// Convert I420 to ARGB with matrix and UV filter mode.
+LIBYUV_API
+int I420ToARGBMatrixFilter(const uint8_t* src_y,
+                           int src_stride_y,
+                           const uint8_t* src_u,
+                           int src_stride_u,
+                           const uint8_t* src_v,
+                           int src_stride_v,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter);
+
+// Convert I422 to ARGB with matrix and UV filter mode.
+LIBYUV_API
+int I422ToARGBMatrixFilter(const uint8_t* src_y,
+                           int src_stride_y,
+                           const uint8_t* src_u,
+                           int src_stride_u,
+                           const uint8_t* src_v,
+                           int src_stride_v,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter);
+
+// Convert I422 to RGB24 with matrix and UV filter mode.
+LIBYUV_API
+int I422ToRGB24MatrixFilter(const uint8_t* src_y,
+                            int src_stride_y,
+                            const uint8_t* src_u,
+                            int src_stride_u,
+                            const uint8_t* src_v,
+                            int src_stride_v,
+                            uint8_t* dst_rgb24,
+                            int dst_stride_rgb24,
+                            const struct YuvConstants* yuvconstants,
+                            int width,
+                            int height,
+                            enum FilterMode filter);
+
+// Convert I420 to RGB24 with matrix and UV filter mode.
+LIBYUV_API
+int I420ToRGB24MatrixFilter(const uint8_t* src_y,
+                            int src_stride_y,
+                            const uint8_t* src_u,
+                            int src_stride_u,
+                            const uint8_t* src_v,
+                            int src_stride_v,
+                            uint8_t* dst_rgb24,
+                            int dst_stride_rgb24,
+                            const struct YuvConstants* yuvconstants,
+                            int width,
+                            int height,
+                            enum FilterMode filter);
+
+// Convert I010 to AR30 with matrix and UV filter mode.
+LIBYUV_API
+int I010ToAR30MatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_u,
+                           int src_stride_u,
+                           const uint16_t* src_v,
+                           int src_stride_v,
+                           uint8_t* dst_ar30,
+                           int dst_stride_ar30,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter);
+
+// Convert I210 to AR30 with matrix and UV filter mode.
+LIBYUV_API
+int I210ToAR30MatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_u,
+                           int src_stride_u,
+                           const uint16_t* src_v,
+                           int src_stride_v,
+                           uint8_t* dst_ar30,
+                           int dst_stride_ar30,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter);
+
+// Convert I010 to ARGB with matrix and UV filter mode.
+LIBYUV_API
+int I010ToARGBMatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_u,
+                           int src_stride_u,
+                           const uint16_t* src_v,
+                           int src_stride_v,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter);
+
+// Convert I210 to ARGB with matrix and UV filter mode.
+LIBYUV_API
+int I210ToARGBMatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_u,
+                           int src_stride_u,
+                           const uint16_t* src_v,
+                           int src_stride_v,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter);
+
+// Convert I420 with Alpha to attenuated ARGB with matrix and UV filter mode.
+LIBYUV_API
+int I420AlphaToARGBMatrixFilter(const uint8_t* src_y,
+                                int src_stride_y,
+                                const uint8_t* src_u,
+                                int src_stride_u,
+                                const uint8_t* src_v,
+                                int src_stride_v,
+                                const uint8_t* src_a,
+                                int src_stride_a,
+                                uint8_t* dst_argb,
+                                int dst_stride_argb,
+                                const struct YuvConstants* yuvconstants,
+                                int width,
+                                int height,
+                                int attenuate,
+                                enum FilterMode filter);
+
+// Convert I422 with Alpha to attenuated ARGB with matrix and UV filter mode.
+LIBYUV_API
+int I422AlphaToARGBMatrixFilter(const uint8_t* src_y,
+                                int src_stride_y,
+                                const uint8_t* src_u,
+                                int src_stride_u,
+                                const uint8_t* src_v,
+                                int src_stride_v,
+                                const uint8_t* src_a,
+                                int src_stride_a,
+                                uint8_t* dst_argb,
+                                int dst_stride_argb,
+                                const struct YuvConstants* yuvconstants,
+                                int width,
+                                int height,
+                                int attenuate,
+                                enum FilterMode filter);
+
+// Convert I010 with Alpha to attenuated ARGB with matrix and UV filter mode.
+LIBYUV_API
+int I010AlphaToARGBMatrixFilter(const uint16_t* src_y,
+                                int src_stride_y,
+                                const uint16_t* src_u,
+                                int src_stride_u,
+                                const uint16_t* src_v,
+                                int src_stride_v,
+                                const uint16_t* src_a,
+                                int src_stride_a,
+                                uint8_t* dst_argb,
+                                int dst_stride_argb,
+                                const struct YuvConstants* yuvconstants,
+                                int width,
+                                int height,
+                                int attenuate,
+                                enum FilterMode filter);
+
+// Convert I210 with Alpha to attenuated ARGB with matrix and UV filter mode.
+LIBYUV_API
+int I210AlphaToARGBMatrixFilter(const uint16_t* src_y,
+                                int src_stride_y,
+                                const uint16_t* src_u,
+                                int src_stride_u,
+                                const uint16_t* src_v,
+                                int src_stride_v,
+                                const uint16_t* src_a,
+                                int src_stride_a,
+                                uint8_t* dst_argb,
+                                int dst_stride_argb,
+                                const struct YuvConstants* yuvconstants,
+                                int width,
+                                int height,
+                                int attenuate,
+                                enum FilterMode filter);
+
+// Convert P010 to ARGB with matrix and UV filter mode.
+LIBYUV_API
+int P010ToARGBMatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_uv,
+                           int src_stride_uv,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter);
+
+// Convert P210 to ARGB with matrix and UV filter mode.
+LIBYUV_API
+int P210ToARGBMatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_uv,
+                           int src_stride_uv,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter);
+
+// Convert P010 to AR30 with matrix and UV filter mode.
+LIBYUV_API
+int P010ToAR30MatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_uv,
+                           int src_stride_uv,
+                           uint8_t* dst_ar30,
+                           int dst_stride_ar30,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter);
+
+// Convert P210 to AR30 with matrix and UV filter mode.
+LIBYUV_API
+int P210ToAR30MatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_uv,
+                           int src_stride_uv,
+                           uint8_t* dst_ar30,
+                           int dst_stride_ar30,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter);
+
+// Convert camera sample to ARGB with cropping, rotation and vertical flip.
+// "sample_size" is needed to parse MJPG.
+// "dst_stride_argb" number of bytes in a row of the dst_argb plane.
+//   Normally this would be the same as dst_width, with recommended alignment
+//   to 16 bytes for better efficiency.
+//   If rotation of 90 or 270 is used, stride is affected. The caller should
+//   allocate the I420 buffer according to rotation.
+// "dst_stride_u" number of bytes in a row of the dst_u plane.
+//   Normally this would be the same as (dst_width + 1) / 2, with
+//   recommended alignment to 16 bytes for better efficiency.
+//   If rotation of 90 or 270 is used, stride is affected.
+// "crop_x" and "crop_y" are starting position for cropping.
+//   To center, crop_x = (src_width - dst_width) / 2
+//              crop_y = (src_height - dst_height) / 2
+// "src_width" / "src_height" is size of src_frame in pixels.
+//   "src_height" can be negative indicating a vertically flipped image source.
+// "crop_width" / "crop_height" is the size to crop the src to.
+//    Must be less than or equal to src_width/src_height
+//    Cropping parameters are pre-rotation.
+// "rotation" can be 0, 90, 180 or 270.
+// "fourcc" is a fourcc. ie 'I420', 'YUY2'
+// Returns 0 for successful; -1 for invalid parameter. Non-zero for failure.
+LIBYUV_API
+int ConvertToARGB(const uint8_t* sample,
+                  size_t sample_size,
+                  uint8_t* dst_argb,
+                  int dst_stride_argb,
+                  int crop_x,
+                  int crop_y,
+                  int src_width,
+                  int src_height,
+                  int crop_width,
+                  int crop_height,
+                  enum RotationMode rotation,
+                  uint32_t fourcc);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_CONVERT_ARGB_H_

yuv.mod/libyuv/include/libyuv/convert_from.h

@@ -0,0 +1,203 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_CONVERT_FROM_H_
+#define INCLUDE_LIBYUV_CONVERT_FROM_H_
+
+#include "libyuv/basic_types.h"
+#include "libyuv/rotate.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// See Also convert.h for conversions from formats to I420.
+
+// Convert 8 bit YUV to 10 bit.
+#define H420ToH010 I420ToI010
+LIBYUV_API
+int I420ToI010(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert 8 bit YUV to 12 bit.
+#define H420ToH012 I420ToI012
+LIBYUV_API
+int I420ToI012(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+LIBYUV_API
+int I420ToI422(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+LIBYUV_API
+int I420ToI444(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Copy to I400. Source can be I420, I422, I444, I400, NV12 or NV21.
+LIBYUV_API
+int I400Copy(const uint8_t* src_y,
+             int src_stride_y,
+             uint8_t* dst_y,
+             int dst_stride_y,
+             int width,
+             int height);
+
+LIBYUV_API
+int I420ToNV12(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+LIBYUV_API
+int I420ToNV21(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height);
+
+LIBYUV_API
+int I420ToYUY2(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_yuy2,
+               int dst_stride_yuy2,
+               int width,
+               int height);
+
+LIBYUV_API
+int I420ToUYVY(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_uyvy,
+               int dst_stride_uyvy,
+               int width,
+               int height);
+
+// The following are from convert_argb.h
+// DEPRECATED: The prototypes will be removed in future.  Use convert_argb.h
+
+// Convert I420 to ARGB.
+LIBYUV_API
+int I420ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Convert I420 to ABGR.
+LIBYUV_API
+int I420ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert I420 to specified format.
+// "dst_sample_stride" is bytes in a row for the destination. Pass 0 if the
+//    buffer has contiguous rows. Can be negative. A multiple of 16 is optimal.
+LIBYUV_API
+int ConvertFromI420(const uint8_t* y,
+                    int y_stride,
+                    const uint8_t* u,
+                    int u_stride,
+                    const uint8_t* v,
+                    int v_stride,
+                    uint8_t* dst_sample,
+                    int dst_sample_stride,
+                    int width,
+                    int height,
+                    uint32_t fourcc);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_CONVERT_FROM_H_

yuv.mod/libyuv/include/libyuv/convert_from_argb.h

@@ -0,0 +1,381 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_CONVERT_FROM_ARGB_H_
+#define INCLUDE_LIBYUV_CONVERT_FROM_ARGB_H_
+
+#include "libyuv/basic_types.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Copy ARGB to ARGB.
+#define ARGBToARGB ARGBCopy
+LIBYUV_API
+int ARGBCopy(const uint8_t* src_argb,
+             int src_stride_argb,
+             uint8_t* dst_argb,
+             int dst_stride_argb,
+             int width,
+             int height);
+
+// Convert ARGB To BGRA.
+LIBYUV_API
+int ARGBToBGRA(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_bgra,
+               int dst_stride_bgra,
+               int width,
+               int height);
+
+// Convert ARGB To ABGR.
+LIBYUV_API
+int ARGBToABGR(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height);
+
+// Convert ARGB To RGBA.
+LIBYUV_API
+int ARGBToRGBA(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_rgba,
+               int dst_stride_rgba,
+               int width,
+               int height);
+
+// Aliases
+#define ARGBToAB30 ABGRToAR30
+#define ABGRToAB30 ARGBToAR30
+
+// Convert ABGR To AR30.
+LIBYUV_API
+int ABGRToAR30(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height);
+
+// Convert ARGB To AR30.
+LIBYUV_API
+int ARGBToAR30(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height);
+
+// Aliases
+#define ABGRToRGB24 ARGBToRAW
+#define ABGRToRAW ARGBToRGB24
+
+// Convert ARGB To RGB24.
+LIBYUV_API
+int ARGBToRGB24(const uint8_t* src_argb,
+                int src_stride_argb,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height);
+
+// Convert ARGB To RAW.
+LIBYUV_API
+int ARGBToRAW(const uint8_t* src_argb,
+              int src_stride_argb,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height);
+
+// Convert ARGB To RGB565.
+LIBYUV_API
+int ARGBToRGB565(const uint8_t* src_argb,
+                 int src_stride_argb,
+                 uint8_t* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height);
+
+// Convert ARGB To RGB565 with 4x4 dither matrix (16 bytes).
+// Values in dither matrix from 0 to 7 recommended.
+// The order of the dither matrix is first byte is upper left.
+// TODO(fbarchard): Consider pointer to 2d array for dither4x4.
+// const uint8_t(*dither)[4][4];
+LIBYUV_API
+int ARGBToRGB565Dither(const uint8_t* src_argb,
+                       int src_stride_argb,
+                       uint8_t* dst_rgb565,
+                       int dst_stride_rgb565,
+                       const uint8_t* dither4x4,
+                       int width,
+                       int height);
+
+// Convert ARGB To ARGB1555.
+LIBYUV_API
+int ARGBToARGB1555(const uint8_t* src_argb,
+                   int src_stride_argb,
+                   uint8_t* dst_argb1555,
+                   int dst_stride_argb1555,
+                   int width,
+                   int height);
+
+// Convert ARGB To ARGB4444.
+LIBYUV_API
+int ARGBToARGB4444(const uint8_t* src_argb,
+                   int src_stride_argb,
+                   uint8_t* dst_argb4444,
+                   int dst_stride_argb4444,
+                   int width,
+                   int height);
+
+// Convert ARGB To I444.
+LIBYUV_API
+int ARGBToI444(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert ARGB to AR64.
+LIBYUV_API
+int ARGBToAR64(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint16_t* dst_ar64,
+               int dst_stride_ar64,
+               int width,
+               int height);
+
+// Convert ABGR to AB64.
+#define ABGRToAB64 ARGBToAR64
+
+// Convert ARGB to AB64.
+LIBYUV_API
+int ARGBToAB64(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint16_t* dst_ab64,
+               int dst_stride_ab64,
+               int width,
+               int height);
+
+// Convert ABGR to AR64.
+#define ABGRToAR64 ARGBToAB64
+
+// Convert ARGB To I422.
+LIBYUV_API
+int ARGBToI422(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert ARGB To I420. (also in convert.h)
+LIBYUV_API
+int ARGBToI420(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert ARGB to J420. (JPeg full range I420).
+LIBYUV_API
+int ARGBToJ420(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               uint8_t* dst_uj,
+               int dst_stride_uj,
+               uint8_t* dst_vj,
+               int dst_stride_vj,
+               int width,
+               int height);
+
+// Convert ARGB to J422.
+LIBYUV_API
+int ARGBToJ422(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               uint8_t* dst_uj,
+               int dst_stride_uj,
+               uint8_t* dst_vj,
+               int dst_stride_vj,
+               int width,
+               int height);
+
+// Convert ARGB to J400. (JPeg full range).
+LIBYUV_API
+int ARGBToJ400(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               int width,
+               int height);
+
+// Convert ABGR to J420. (JPeg full range I420).
+LIBYUV_API
+int ABGRToJ420(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               uint8_t* dst_uj,
+               int dst_stride_uj,
+               uint8_t* dst_vj,
+               int dst_stride_vj,
+               int width,
+               int height);
+
+// Convert ABGR to J422.
+LIBYUV_API
+int ABGRToJ422(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               uint8_t* dst_uj,
+               int dst_stride_uj,
+               uint8_t* dst_vj,
+               int dst_stride_vj,
+               int width,
+               int height);
+
+// Convert ABGR to J400. (JPeg full range).
+LIBYUV_API
+int ABGRToJ400(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               int width,
+               int height);
+
+// Convert RGBA to J400. (JPeg full range).
+LIBYUV_API
+int RGBAToJ400(const uint8_t* src_rgba,
+               int src_stride_rgba,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               int width,
+               int height);
+
+// Convert ARGB to I400.
+LIBYUV_API
+int ARGBToI400(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               int width,
+               int height);
+
+// Convert ARGB to G. (Reverse of J400toARGB, which replicates G back to ARGB)
+LIBYUV_API
+int ARGBToG(const uint8_t* src_argb,
+            int src_stride_argb,
+            uint8_t* dst_g,
+            int dst_stride_g,
+            int width,
+            int height);
+
+// Convert ARGB To NV12.
+LIBYUV_API
+int ARGBToNV12(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert ARGB To NV21.
+LIBYUV_API
+int ARGBToNV21(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height);
+
+// Convert ABGR To NV12.
+LIBYUV_API
+int ABGRToNV12(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert ABGR To NV21.
+LIBYUV_API
+int ABGRToNV21(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height);
+
+// Convert ARGB To YUY2.
+LIBYUV_API
+int ARGBToYUY2(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_yuy2,
+               int dst_stride_yuy2,
+               int width,
+               int height);
+
+// Convert ARGB To UYVY.
+LIBYUV_API
+int ARGBToUYVY(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_uyvy,
+               int dst_stride_uyvy,
+               int width,
+               int height);
+
+// RAW to JNV21 full range NV21
+LIBYUV_API
+int RAWToJNV21(const uint8_t* src_raw,
+               int src_stride_raw,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_CONVERT_FROM_ARGB_H_

yuv.mod/libyuv/include/libyuv/cpu_id.h

@@ -0,0 +1,150 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_CPU_ID_H_
+#define INCLUDE_LIBYUV_CPU_ID_H_
+
+#include "libyuv/basic_types.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Internal flag to indicate cpuid requires initialization.
+static const int kCpuInitialized = 0x1;
+
+// These flags are only valid on Arm processors.
+static const int kCpuHasARM = 0x2;
+static const int kCpuHasNEON = 0x4;
+// Leave a gap to avoid setting kCpuHasX86.
+static const int kCpuHasNeonDotProd = 0x10;
+static const int kCpuHasNeonI8MM = 0x20;
+static const int kCpuHasSVE = 0x40;
+static const int kCpuHasSVE2 = 0x80;
+static const int kCpuHasSME = 0x100;
+
+// These flags are only valid on x86 processors.
+static const int kCpuHasX86 = 0x8;
+static const int kCpuHasSSE2 = 0x10;
+static const int kCpuHasSSSE3 = 0x20;
+static const int kCpuHasSSE41 = 0x40;
+static const int kCpuHasSSE42 = 0x80;
+static const int kCpuHasAVX = 0x100;
+static const int kCpuHasAVX2 = 0x200;
+static const int kCpuHasERMS = 0x400;
+static const int kCpuHasFMA3 = 0x800;
+static const int kCpuHasF16C = 0x1000;
+static const int kCpuHasAVX512BW = 0x2000;
+static const int kCpuHasAVX512VL = 0x4000;
+static const int kCpuHasAVX512VNNI = 0x8000;
+static const int kCpuHasAVX512VBMI = 0x10000;
+static const int kCpuHasAVX512VBMI2 = 0x20000;
+static const int kCpuHasAVX512VBITALG = 0x40000;
+static const int kCpuHasAVX10 = 0x80000;
+static const int kCpuHasAVXVNNI = 0x100000;
+static const int kCpuHasAVXVNNIINT8 = 0x200000;
+static const int kCpuHasAMXINT8 = 0x400000;
+
+// These flags are only valid on MIPS processors.
+static const int kCpuHasMIPS = 0x800000;
+static const int kCpuHasMSA = 0x1000000;
+
+// These flags are only valid on LOONGARCH processors.
+static const int kCpuHasLOONGARCH = 0x2000000;
+static const int kCpuHasLSX = 0x4000000;
+static const int kCpuHasLASX = 0x8000000;
+
+// These flags are only valid on RISCV processors.
+static const int kCpuHasRISCV = 0x10000000;
+static const int kCpuHasRVV = 0x20000000;
+static const int kCpuHasRVVZVFH = 0x40000000;
+
+// Optional init function. TestCpuFlag does an auto-init.
+// Returns cpu_info flags.
+LIBYUV_API
+int InitCpuFlags(void);
+
+// Detect CPU has SSE2 etc.
+// Test_flag parameter should be one of kCpuHas constants above.
+// Returns non-zero if instruction set is detected
+static __inline int TestCpuFlag(int test_flag) {
+  LIBYUV_API extern int cpu_info_;
+#ifdef __ATOMIC_RELAXED
+  int cpu_info = __atomic_load_n(&cpu_info_, __ATOMIC_RELAXED);
+#else
+  int cpu_info = cpu_info_;
+#endif
+  return (!cpu_info ? InitCpuFlags() : cpu_info) & test_flag;
+}
+
+// Internal function for parsing /proc/cpuinfo.
+LIBYUV_API
+int ArmCpuCaps(const char* cpuinfo_name);
+LIBYUV_API
+int MipsCpuCaps(const char* cpuinfo_name);
+LIBYUV_API
+int RiscvCpuCaps(const char* cpuinfo_name);
+
+#ifdef __linux__
+// On Linux, parse AArch64 features from getauxval(AT_HWCAP{,2}).
+LIBYUV_API
+int AArch64CpuCaps(unsigned long hwcap, unsigned long hwcap2);
+#else
+LIBYUV_API
+int AArch64CpuCaps();
+#endif
+
+// For testing, allow CPU flags to be disabled.
+// ie MaskCpuFlags(~kCpuHasSSSE3) to disable SSSE3.
+// MaskCpuFlags(-1) to enable all cpu specific optimizations.
+// MaskCpuFlags(1) to disable all cpu specific optimizations.
+// MaskCpuFlags(0) to reset state so next call will auto init.
+// Returns cpu_info flags.
+LIBYUV_API
+int MaskCpuFlags(int enable_flags);
+
+// Sets the CPU flags to |cpu_flags|, bypassing the detection code. |cpu_flags|
+// should be a valid combination of the kCpuHas constants above and include
+// kCpuInitialized. Use this method when running in a sandboxed process where
+// the detection code might fail (as it might access /proc/cpuinfo). In such
+// cases the cpu_info can be obtained from a non sandboxed process by calling
+// InitCpuFlags() and passed to the sandboxed process (via command line
+// parameters, IPC...) which can then call this method to initialize the CPU
+// flags.
+// Notes:
+// - when specifying 0 for |cpu_flags|, the auto initialization is enabled
+//   again.
+// - enabling CPU features that are not supported by the CPU will result in
+//   undefined behavior.
+// TODO(fbarchard): consider writing a helper function that translates from
+// other library CPU info to libyuv CPU info and add a .md doc that explains
+// CPU detection.
+static __inline void SetCpuFlags(int cpu_flags) {
+  LIBYUV_API extern int cpu_info_;
+#ifdef __ATOMIC_RELAXED
+  __atomic_store_n(&cpu_info_, cpu_flags, __ATOMIC_RELAXED);
+#else
+  cpu_info_ = cpu_flags;
+#endif
+}
+
+// Low level cpuid for X86. Returns zeros on other CPUs.
+// eax is the info type that you want.
+// ecx is typically the cpu number, and should normally be zero.
+LIBYUV_API
+void CpuId(int info_eax, int info_ecx, int* cpu_info);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_CPU_ID_H_

yuv.mod/libyuv/include/libyuv/loongson_intrinsics.h

@@ -0,0 +1,1949 @@
+/*
+ *  Copyright 2022 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_LOONGSON_INTRINSICS_H
+#define INCLUDE_LIBYUV_LOONGSON_INTRINSICS_H
+
+/*
+ * Copyright (c) 2022 Loongson Technology Corporation Limited
+ * All rights reserved.
+ * Contributed by Shiyou Yin <[email protected]>
+ *                Xiwei Gu   <[email protected]>
+ *                Lu Wang    <[email protected]>
+ *
+ * This file is a header file for loongarch builtin extension.
+ *
+ */
+
+#ifndef LOONGSON_INTRINSICS_H
+#define LOONGSON_INTRINSICS_H
+
+/**
+ * MAJOR version: Macro usage changes.
+ * MINOR version: Add new functions, or bug fixes.
+ * MICRO version: Comment changes or implementation changes.
+ */
+#define LSOM_VERSION_MAJOR 1
+#define LSOM_VERSION_MINOR 1
+#define LSOM_VERSION_MICRO 0
+
+#define DUP2_ARG1(_INS, _IN0, _IN1, _OUT0, _OUT1) \
+  {                                               \
+    _OUT0 = _INS(_IN0);                           \
+    _OUT1 = _INS(_IN1);                           \
+  }
+
+#define DUP2_ARG2(_INS, _IN0, _IN1, _IN2, _IN3, _OUT0, _OUT1) \
+  {                                                           \
+    _OUT0 = _INS(_IN0, _IN1);                                 \
+    _OUT1 = _INS(_IN2, _IN3);                                 \
+  }
+
+#define DUP2_ARG3(_INS, _IN0, _IN1, _IN2, _IN3, _IN4, _IN5, _OUT0, _OUT1) \
+  {                                                                       \
+    _OUT0 = _INS(_IN0, _IN1, _IN2);                                       \
+    _OUT1 = _INS(_IN3, _IN4, _IN5);                                       \
+  }
+
+#define DUP4_ARG1(_INS, _IN0, _IN1, _IN2, _IN3, _OUT0, _OUT1, _OUT2, _OUT3) \
+  {                                                                         \
+    DUP2_ARG1(_INS, _IN0, _IN1, _OUT0, _OUT1);                              \
+    DUP2_ARG1(_INS, _IN2, _IN3, _OUT2, _OUT3);                              \
+  }
+
+#define DUP4_ARG2(_INS, _IN0, _IN1, _IN2, _IN3, _IN4, _IN5, _IN6, _IN7, _OUT0, \
+                  _OUT1, _OUT2, _OUT3)                                         \
+  {                                                                            \
+    DUP2_ARG2(_INS, _IN0, _IN1, _IN2, _IN3, _OUT0, _OUT1);                     \
+    DUP2_ARG2(_INS, _IN4, _IN5, _IN6, _IN7, _OUT2, _OUT3);                     \
+  }
+
+#define DUP4_ARG3(_INS, _IN0, _IN1, _IN2, _IN3, _IN4, _IN5, _IN6, _IN7, _IN8, \
+                  _IN9, _IN10, _IN11, _OUT0, _OUT1, _OUT2, _OUT3)             \
+  {                                                                           \
+    DUP2_ARG3(_INS, _IN0, _IN1, _IN2, _IN3, _IN4, _IN5, _OUT0, _OUT1);        \
+    DUP2_ARG3(_INS, _IN6, _IN7, _IN8, _IN9, _IN10, _IN11, _OUT2, _OUT3);      \
+  }
+
+#ifdef __loongarch_sx
+#include <lsxintrin.h>
+/*
+ * =============================================================================
+ * Description : Dot product & addition of byte vector elements
+ * Arguments   : Inputs  - in_c, in_h, in_l
+ *               Outputs - out
+ *               Return Type - halfword
+ * Details     : Signed byte elements from in_h are multiplied by
+ *               signed byte elements from in_l, and then added adjacent to
+ *               each other to get results with the twice size of input.
+ *               Then the results plus to signed half-word elements from in_c.
+ * Example     : out = __lsx_vdp2add_h_b(in_c, in_h, in_l)
+ *        in_c : 1,2,3,4, 1,2,3,4
+ *        in_h : 1,2,3,4, 5,6,7,8, 1,2,3,4, 5,6,7,8
+ *        in_l : 8,7,6,5, 4,3,2,1, 8,7,6,5, 4,3,2,1
+ *         out : 23,40,41,26, 23,40,41,26
+ * =============================================================================
+ */
+static inline __m128i __lsx_vdp2add_h_b(__m128i in_c,
+                                        __m128i in_h,
+                                        __m128i in_l) {
+  __m128i out;
+
+  out = __lsx_vmaddwev_h_b(in_c, in_h, in_l);
+  out = __lsx_vmaddwod_h_b(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product & addition of byte vector elements
+ * Arguments   : Inputs  - in_c, in_h, in_l
+ *               Outputs - out
+ *               Return Type - halfword
+ * Details     : Unsigned byte elements from in_h are multiplied by
+ *               unsigned byte elements from in_l, and then added adjacent to
+ *               each other to get results with the twice size of input.
+ *               The results plus to signed half-word elements from in_c.
+ * Example     : out = __lsx_vdp2add_h_bu(in_c, in_h, in_l)
+ *        in_c : 1,2,3,4, 1,2,3,4
+ *        in_h : 1,2,3,4, 5,6,7,8, 1,2,3,4, 5,6,7,8
+ *        in_l : 8,7,6,5, 4,3,2,1, 8,7,6,5, 4,3,2,1
+ *         out : 23,40,41,26, 23,40,41,26
+ * =============================================================================
+ */
+static inline __m128i __lsx_vdp2add_h_bu(__m128i in_c,
+                                         __m128i in_h,
+                                         __m128i in_l) {
+  __m128i out;
+
+  out = __lsx_vmaddwev_h_bu(in_c, in_h, in_l);
+  out = __lsx_vmaddwod_h_bu(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product & addition of byte vector elements
+ * Arguments   : Inputs  - in_c, in_h, in_l
+ *               Outputs - out
+ *               Return Type - halfword
+ * Details     : Unsigned byte elements from in_h are multiplied by
+ *               signed byte elements from in_l, and then added adjacent to
+ *               each other to get results with the twice size of input.
+ *               The results plus to signed half-word elements from in_c.
+ * Example     : out = __lsx_vdp2add_h_bu_b(in_c, in_h, in_l)
+ *        in_c : 1,1,1,1, 1,1,1,1
+ *        in_h : 1,2,3,4, 5,6,7,8, 1,2,3,4, 5,6,7,8
+ *        in_l : -1,-2,-3,-4, -5,-6,-7,-8, 1,2,3,4, 5,6,7,8
+ *         out : -4,-24,-60,-112, 6,26,62,114
+ * =============================================================================
+ */
+static inline __m128i __lsx_vdp2add_h_bu_b(__m128i in_c,
+                                           __m128i in_h,
+                                           __m128i in_l) {
+  __m128i out;
+
+  out = __lsx_vmaddwev_h_bu_b(in_c, in_h, in_l);
+  out = __lsx_vmaddwod_h_bu_b(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product & addition of half-word vector elements
+ * Arguments   : Inputs  - in_c, in_h, in_l
+ *               Outputs - out
+ *               Return Type - __m128i
+ * Details     : Signed half-word elements from in_h are multiplied by
+ *               signed half-word elements from in_l, and then added adjacent to
+ *               each other to get results with the twice size of input.
+ *               Then the results plus to signed word elements from in_c.
+ * Example     : out = __lsx_vdp2add_h_b(in_c, in_h, in_l)
+ *        in_c : 1,2,3,4
+ *        in_h : 1,2,3,4, 5,6,7,8
+ *        in_l : 8,7,6,5, 4,3,2,1
+ *         out : 23,40,41,26
+ * =============================================================================
+ */
+static inline __m128i __lsx_vdp2add_w_h(__m128i in_c,
+                                        __m128i in_h,
+                                        __m128i in_l) {
+  __m128i out;
+
+  out = __lsx_vmaddwev_w_h(in_c, in_h, in_l);
+  out = __lsx_vmaddwod_w_h(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product of byte vector elements
+ * Arguments   : Inputs  - in_h, in_l
+ *               Outputs - out
+ *               Return Type - halfword
+ * Details     : Signed byte elements from in_h are multiplied by
+ *               signed byte elements from in_l, and then added adjacent to
+ *               each other to get results with the twice size of input.
+ * Example     : out = __lsx_vdp2_h_b(in_h, in_l)
+ *        in_h : 1,2,3,4, 5,6,7,8, 1,2,3,4, 5,6,7,8
+ *        in_l : 8,7,6,5, 4,3,2,1, 8,7,6,5, 4,3,2,1
+ *         out : 22,38,38,22, 22,38,38,22
+ * =============================================================================
+ */
+static inline __m128i __lsx_vdp2_h_b(__m128i in_h, __m128i in_l) {
+  __m128i out;
+
+  out = __lsx_vmulwev_h_b(in_h, in_l);
+  out = __lsx_vmaddwod_h_b(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product of byte vector elements
+ * Arguments   : Inputs  - in_h, in_l
+ *               Outputs - out
+ *               Return Type - halfword
+ * Details     : Unsigned byte elements from in_h are multiplied by
+ *               unsigned byte elements from in_l, and then added adjacent to
+ *               each other to get results with the twice size of input.
+ * Example     : out = __lsx_vdp2_h_bu(in_h, in_l)
+ *        in_h : 1,2,3,4, 5,6,7,8, 1,2,3,4, 5,6,7,8
+ *        in_l : 8,7,6,5, 4,3,2,1, 8,7,6,5, 4,3,2,1
+ *         out : 22,38,38,22, 22,38,38,22
+ * =============================================================================
+ */
+static inline __m128i __lsx_vdp2_h_bu(__m128i in_h, __m128i in_l) {
+  __m128i out;
+
+  out = __lsx_vmulwev_h_bu(in_h, in_l);
+  out = __lsx_vmaddwod_h_bu(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product of byte vector elements
+ * Arguments   : Inputs  - in_h, in_l
+ *               Outputs - out
+ *               Return Type - halfword
+ * Details     : Unsigned byte elements from in_h are multiplied by
+ *               signed byte elements from in_l, and then added adjacent to
+ *               each other to get results with the twice size of input.
+ * Example     : out = __lsx_vdp2_h_bu_b(in_h, in_l)
+ *        in_h : 1,2,3,4, 5,6,7,8, 1,2,3,4, 5,6,7,8
+ *        in_l : 8,7,6,5, 4,3,2,1, 8,7,6,5, 4,3,2,-1
+ *         out : 22,38,38,22, 22,38,38,6
+ * =============================================================================
+ */
+static inline __m128i __lsx_vdp2_h_bu_b(__m128i in_h, __m128i in_l) {
+  __m128i out;
+
+  out = __lsx_vmulwev_h_bu_b(in_h, in_l);
+  out = __lsx_vmaddwod_h_bu_b(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product of byte vector elements
+ * Arguments   : Inputs  - in_h, in_l
+ *               Outputs - out
+ *               Return Type - halfword
+ * Details     : Signed byte elements from in_h are multiplied by
+ *               signed byte elements from in_l, and then added adjacent to
+ *               each other to get results with the twice size of input.
+ * Example     : out = __lsx_vdp2_w_h(in_h, in_l)
+ *        in_h : 1,2,3,4, 5,6,7,8
+ *        in_l : 8,7,6,5, 4,3,2,1
+ *         out : 22,38,38,22
+ * =============================================================================
+ */
+static inline __m128i __lsx_vdp2_w_h(__m128i in_h, __m128i in_l) {
+  __m128i out;
+
+  out = __lsx_vmulwev_w_h(in_h, in_l);
+  out = __lsx_vmaddwod_w_h(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Clip all halfword elements of input vector between min & max
+ *               out = ((_in) < (min)) ? (min) : (((_in) > (max)) ? (max) :
+ *               (_in))
+ * Arguments   : Inputs  - _in  (input vector)
+ *                       - min  (min threshold)
+ *                       - max  (max threshold)
+ *               Outputs - out  (output vector with clipped elements)
+ *               Return Type - signed halfword
+ * Example     : out = __lsx_vclip_h(_in)
+ *         _in : -8,2,280,249, -8,255,280,249
+ *         min : 1,1,1,1, 1,1,1,1
+ *         max : 9,9,9,9, 9,9,9,9
+ *         out : 1,2,9,9, 1,9,9,9
+ * =============================================================================
+ */
+static inline __m128i __lsx_vclip_h(__m128i _in, __m128i min, __m128i max) {
+  __m128i out;
+
+  out = __lsx_vmax_h(min, _in);
+  out = __lsx_vmin_h(max, out);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Set each element of vector between 0 and 255
+ * Arguments   : Inputs  - _in
+ *               Outputs - out
+ *               Return Type - halfword
+ * Details     : Signed byte elements from _in are clamped between 0 and 255.
+ * Example     : out = __lsx_vclip255_h(_in)
+ *         _in : -8,255,280,249, -8,255,280,249
+ *         out : 0,255,255,249, 0,255,255,249
+ * =============================================================================
+ */
+static inline __m128i __lsx_vclip255_h(__m128i _in) {
+  __m128i out;
+
+  out = __lsx_vmaxi_h(_in, 0);
+  out = __lsx_vsat_hu(out, 7);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Set each element of vector between 0 and 255
+ * Arguments   : Inputs  - _in
+ *               Outputs - out
+ *               Return Type - word
+ * Details     : Signed byte elements from _in are clamped between 0 and 255.
+ * Example     : out = __lsx_vclip255_w(_in)
+ *         _in : -8,255,280,249
+ *         out : 0,255,255,249
+ * =============================================================================
+ */
+static inline __m128i __lsx_vclip255_w(__m128i _in) {
+  __m128i out;
+
+  out = __lsx_vmaxi_w(_in, 0);
+  out = __lsx_vsat_wu(out, 7);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Swap two variables
+ * Arguments   : Inputs  - _in0, _in1
+ *               Outputs - _in0, _in1 (in-place)
+ * Details     : Swapping of two input variables using xor
+ * Example     : LSX_SWAP(_in0, _in1)
+ *        _in0 : 1,2,3,4
+ *        _in1 : 5,6,7,8
+ *   _in0(out) : 5,6,7,8
+ *   _in1(out) : 1,2,3,4
+ * =============================================================================
+ */
+#define LSX_SWAP(_in0, _in1)         \
+  {                                  \
+    _in0 = __lsx_vxor_v(_in0, _in1); \
+    _in1 = __lsx_vxor_v(_in0, _in1); \
+    _in0 = __lsx_vxor_v(_in0, _in1); \
+  }
+
+/*
+ * =============================================================================
+ * Description : Transpose 4x4 block with word elements in vectors
+ * Arguments   : Inputs  - in0, in1, in2, in3
+ *               Outputs - out0, out1, out2, out3
+ * Details     :
+ * Example     :
+ *               1, 2, 3, 4            1, 5, 9,13
+ *               5, 6, 7, 8    to      2, 6,10,14
+ *               9,10,11,12  =====>    3, 7,11,15
+ *              13,14,15,16            4, 8,12,16
+ * =============================================================================
+ */
+#define LSX_TRANSPOSE4x4_W(_in0, _in1, _in2, _in3, _out0, _out1, _out2, _out3) \
+  {                                                                            \
+    __m128i _t0, _t1, _t2, _t3;                                                \
+                                                                               \
+    _t0 = __lsx_vilvl_w(_in1, _in0);                                           \
+    _t1 = __lsx_vilvh_w(_in1, _in0);                                           \
+    _t2 = __lsx_vilvl_w(_in3, _in2);                                           \
+    _t3 = __lsx_vilvh_w(_in3, _in2);                                           \
+    _out0 = __lsx_vilvl_d(_t2, _t0);                                           \
+    _out1 = __lsx_vilvh_d(_t2, _t0);                                           \
+    _out2 = __lsx_vilvl_d(_t3, _t1);                                           \
+    _out3 = __lsx_vilvh_d(_t3, _t1);                                           \
+  }
+
+/*
+ * =============================================================================
+ * Description : Transpose 8x8 block with byte elements in vectors
+ * Arguments   : Inputs  - _in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7
+ *               Outputs - _out0, _out1, _out2, _out3, _out4, _out5, _out6,
+ *               _out7
+ * Details     : The rows of the matrix become columns, and the columns
+ *               become rows.
+ * Example     : LSX_TRANSPOSE8x8_B
+ *        _in0 : 00,01,02,03,04,05,06,07, 00,00,00,00,00,00,00,00
+ *        _in1 : 10,11,12,13,14,15,16,17, 00,00,00,00,00,00,00,00
+ *        _in2 : 20,21,22,23,24,25,26,27, 00,00,00,00,00,00,00,00
+ *        _in3 : 30,31,32,33,34,35,36,37, 00,00,00,00,00,00,00,00
+ *        _in4 : 40,41,42,43,44,45,46,47, 00,00,00,00,00,00,00,00
+ *        _in5 : 50,51,52,53,54,55,56,57, 00,00,00,00,00,00,00,00
+ *        _in6 : 60,61,62,63,64,65,66,67, 00,00,00,00,00,00,00,00
+ *        _in7 : 70,71,72,73,74,75,76,77, 00,00,00,00,00,00,00,00
+ *
+ *      _ out0 : 00,10,20,30,40,50,60,70, 00,00,00,00,00,00,00,00
+ *      _ out1 : 01,11,21,31,41,51,61,71, 00,00,00,00,00,00,00,00
+ *      _ out2 : 02,12,22,32,42,52,62,72, 00,00,00,00,00,00,00,00
+ *      _ out3 : 03,13,23,33,43,53,63,73, 00,00,00,00,00,00,00,00
+ *      _ out4 : 04,14,24,34,44,54,64,74, 00,00,00,00,00,00,00,00
+ *      _ out5 : 05,15,25,35,45,55,65,75, 00,00,00,00,00,00,00,00
+ *      _ out6 : 06,16,26,36,46,56,66,76, 00,00,00,00,00,00,00,00
+ *      _ out7 : 07,17,27,37,47,57,67,77, 00,00,00,00,00,00,00,00
+ * =============================================================================
+ */
+#define LSX_TRANSPOSE8x8_B(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                           _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                           _out7)                                           \
+  {                                                                         \
+    __m128i zero = {0};                                                     \
+    __m128i shuf8 = {0x0F0E0D0C0B0A0908, 0x1716151413121110};               \
+    __m128i _t0, _t1, _t2, _t3, _t4, _t5, _t6, _t7;                         \
+                                                                            \
+    _t0 = __lsx_vilvl_b(_in2, _in0);                                        \
+    _t1 = __lsx_vilvl_b(_in3, _in1);                                        \
+    _t2 = __lsx_vilvl_b(_in6, _in4);                                        \
+    _t3 = __lsx_vilvl_b(_in7, _in5);                                        \
+    _t4 = __lsx_vilvl_b(_t1, _t0);                                          \
+    _t5 = __lsx_vilvh_b(_t1, _t0);                                          \
+    _t6 = __lsx_vilvl_b(_t3, _t2);                                          \
+    _t7 = __lsx_vilvh_b(_t3, _t2);                                          \
+    _out0 = __lsx_vilvl_w(_t6, _t4);                                        \
+    _out2 = __lsx_vilvh_w(_t6, _t4);                                        \
+    _out4 = __lsx_vilvl_w(_t7, _t5);                                        \
+    _out6 = __lsx_vilvh_w(_t7, _t5);                                        \
+    _out1 = __lsx_vshuf_b(zero, _out0, shuf8);                              \
+    _out3 = __lsx_vshuf_b(zero, _out2, shuf8);                              \
+    _out5 = __lsx_vshuf_b(zero, _out4, shuf8);                              \
+    _out7 = __lsx_vshuf_b(zero, _out6, shuf8);                              \
+  }
+
+/*
+ * =============================================================================
+ * Description : Transpose 8x8 block with half-word elements in vectors
+ * Arguments   : Inputs  - in0, in1, in2, in3, in4, in5, in6, in7
+ *               Outputs - out0, out1, out2, out3, out4, out5, out6, out7
+ * Details     :
+ * Example     :
+ *              00,01,02,03,04,05,06,07           00,10,20,30,40,50,60,70
+ *              10,11,12,13,14,15,16,17           01,11,21,31,41,51,61,71
+ *              20,21,22,23,24,25,26,27           02,12,22,32,42,52,62,72
+ *              30,31,32,33,34,35,36,37    to     03,13,23,33,43,53,63,73
+ *              40,41,42,43,44,45,46,47  ======>  04,14,24,34,44,54,64,74
+ *              50,51,52,53,54,55,56,57           05,15,25,35,45,55,65,75
+ *              60,61,62,63,64,65,66,67           06,16,26,36,46,56,66,76
+ *              70,71,72,73,74,75,76,77           07,17,27,37,47,57,67,77
+ * =============================================================================
+ */
+#define LSX_TRANSPOSE8x8_H(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                           _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                           _out7)                                           \
+  {                                                                         \
+    __m128i _s0, _s1, _t0, _t1, _t2, _t3, _t4, _t5, _t6, _t7;               \
+                                                                            \
+    _s0 = __lsx_vilvl_h(_in6, _in4);                                        \
+    _s1 = __lsx_vilvl_h(_in7, _in5);                                        \
+    _t0 = __lsx_vilvl_h(_s1, _s0);                                          \
+    _t1 = __lsx_vilvh_h(_s1, _s0);                                          \
+    _s0 = __lsx_vilvh_h(_in6, _in4);                                        \
+    _s1 = __lsx_vilvh_h(_in7, _in5);                                        \
+    _t2 = __lsx_vilvl_h(_s1, _s0);                                          \
+    _t3 = __lsx_vilvh_h(_s1, _s0);                                          \
+    _s0 = __lsx_vilvl_h(_in2, _in0);                                        \
+    _s1 = __lsx_vilvl_h(_in3, _in1);                                        \
+    _t4 = __lsx_vilvl_h(_s1, _s0);                                          \
+    _t5 = __lsx_vilvh_h(_s1, _s0);                                          \
+    _s0 = __lsx_vilvh_h(_in2, _in0);                                        \
+    _s1 = __lsx_vilvh_h(_in3, _in1);                                        \
+    _t6 = __lsx_vilvl_h(_s1, _s0);                                          \
+    _t7 = __lsx_vilvh_h(_s1, _s0);                                          \
+                                                                            \
+    _out0 = __lsx_vpickev_d(_t0, _t4);                                      \
+    _out2 = __lsx_vpickev_d(_t1, _t5);                                      \
+    _out4 = __lsx_vpickev_d(_t2, _t6);                                      \
+    _out6 = __lsx_vpickev_d(_t3, _t7);                                      \
+    _out1 = __lsx_vpickod_d(_t0, _t4);                                      \
+    _out3 = __lsx_vpickod_d(_t1, _t5);                                      \
+    _out5 = __lsx_vpickod_d(_t2, _t6);                                      \
+    _out7 = __lsx_vpickod_d(_t3, _t7);                                      \
+  }
+
+/*
+ * =============================================================================
+ * Description : Transpose input 8x4 byte block into 4x8
+ * Arguments   : Inputs  - _in0, _in1, _in2, _in3      (input 8x4 byte block)
+ *               Outputs - _out0, _out1, _out2, _out3  (output 4x8 byte block)
+ *               Return Type - as per RTYPE
+ * Details     : The rows of the matrix become columns, and the columns become
+ *               rows.
+ * Example     : LSX_TRANSPOSE8x4_B
+ *        _in0 : 00,01,02,03,00,00,00,00, 00,00,00,00,00,00,00,00
+ *        _in1 : 10,11,12,13,00,00,00,00, 00,00,00,00,00,00,00,00
+ *        _in2 : 20,21,22,23,00,00,00,00, 00,00,00,00,00,00,00,00
+ *        _in3 : 30,31,32,33,00,00,00,00, 00,00,00,00,00,00,00,00
+ *        _in4 : 40,41,42,43,00,00,00,00, 00,00,00,00,00,00,00,00
+ *        _in5 : 50,51,52,53,00,00,00,00, 00,00,00,00,00,00,00,00
+ *        _in6 : 60,61,62,63,00,00,00,00, 00,00,00,00,00,00,00,00
+ *        _in7 : 70,71,72,73,00,00,00,00, 00,00,00,00,00,00,00,00
+ *
+ *       _out0 : 00,10,20,30,40,50,60,70, 00,00,00,00,00,00,00,00
+ *       _out1 : 01,11,21,31,41,51,61,71, 00,00,00,00,00,00,00,00
+ *       _out2 : 02,12,22,32,42,52,62,72, 00,00,00,00,00,00,00,00
+ *       _out3 : 03,13,23,33,43,53,63,73, 00,00,00,00,00,00,00,00
+ * =============================================================================
+ */
+#define LSX_TRANSPOSE8x4_B(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7, \
+                           _out0, _out1, _out2, _out3)                     \
+  {                                                                        \
+    __m128i _tmp0_m, _tmp1_m, _tmp2_m, _tmp3_m;                            \
+                                                                           \
+    _tmp0_m = __lsx_vpackev_w(_in4, _in0);                                 \
+    _tmp1_m = __lsx_vpackev_w(_in5, _in1);                                 \
+    _tmp2_m = __lsx_vilvl_b(_tmp1_m, _tmp0_m);                             \
+    _tmp0_m = __lsx_vpackev_w(_in6, _in2);                                 \
+    _tmp1_m = __lsx_vpackev_w(_in7, _in3);                                 \
+                                                                           \
+    _tmp3_m = __lsx_vilvl_b(_tmp1_m, _tmp0_m);                             \
+    _tmp0_m = __lsx_vilvl_h(_tmp3_m, _tmp2_m);                             \
+    _tmp1_m = __lsx_vilvh_h(_tmp3_m, _tmp2_m);                             \
+                                                                           \
+    _out0 = __lsx_vilvl_w(_tmp1_m, _tmp0_m);                               \
+    _out2 = __lsx_vilvh_w(_tmp1_m, _tmp0_m);                               \
+    _out1 = __lsx_vilvh_d(_out2, _out0);                                   \
+    _out3 = __lsx_vilvh_d(_out0, _out2);                                   \
+  }
+
+/*
+ * =============================================================================
+ * Description : Transpose 16x8 block with byte elements in vectors
+ * Arguments   : Inputs  - in0, in1, in2, in3, in4, in5, in6, in7, in8
+ *                         in9, in10, in11, in12, in13, in14, in15
+ *               Outputs - out0, out1, out2, out3, out4, out5, out6, out7
+ * Details     :
+ * Example     :
+ *              000,001,002,003,004,005,006,007
+ *              008,009,010,011,012,013,014,015
+ *              016,017,018,019,020,021,022,023
+ *              024,025,026,027,028,029,030,031
+ *              032,033,034,035,036,037,038,039
+ *              040,041,042,043,044,045,046,047        000,008,...,112,120
+ *              048,049,050,051,052,053,054,055        001,009,...,113,121
+ *              056,057,058,059,060,061,062,063   to   002,010,...,114,122
+ *              064,068,066,067,068,069,070,071 =====> 003,011,...,115,123
+ *              072,073,074,075,076,077,078,079        004,012,...,116,124
+ *              080,081,082,083,084,085,086,087        005,013,...,117,125
+ *              088,089,090,091,092,093,094,095        006,014,...,118,126
+ *              096,097,098,099,100,101,102,103        007,015,...,119,127
+ *              104,105,106,107,108,109,110,111
+ *              112,113,114,115,116,117,118,119
+ *              120,121,122,123,124,125,126,127
+ * =============================================================================
+ */
+#define LSX_TRANSPOSE16x8_B(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                            _in8, _in9, _in10, _in11, _in12, _in13, _in14,   \
+                            _in15, _out0, _out1, _out2, _out3, _out4, _out5, \
+                            _out6, _out7)                                    \
+  {                                                                          \
+    __m128i _tmp0, _tmp1, _tmp2, _tmp3, _tmp4, _tmp5, _tmp6, _tmp7;          \
+    __m128i _t0, _t1, _t2, _t3, _t4, _t5, _t6, _t7;                          \
+    DUP4_ARG2(__lsx_vilvl_b, _in2, _in0, _in3, _in1, _in6, _in4, _in7, _in5, \
+              _tmp0, _tmp1, _tmp2, _tmp3);                                   \
+    DUP4_ARG2(__lsx_vilvl_b, _in10, _in8, _in11, _in9, _in14, _in12, _in15,  \
+              _in13, _tmp4, _tmp5, _tmp6, _tmp7);                            \
+    DUP2_ARG2(__lsx_vilvl_b, _tmp1, _tmp0, _tmp3, _tmp2, _t0, _t2);          \
+    DUP2_ARG2(__lsx_vilvh_b, _tmp1, _tmp0, _tmp3, _tmp2, _t1, _t3);          \
+    DUP2_ARG2(__lsx_vilvl_b, _tmp5, _tmp4, _tmp7, _tmp6, _t4, _t6);          \
+    DUP2_ARG2(__lsx_vilvh_b, _tmp5, _tmp4, _tmp7, _tmp6, _t5, _t7);          \
+    DUP2_ARG2(__lsx_vilvl_w, _t2, _t0, _t3, _t1, _tmp0, _tmp4);              \
+    DUP2_ARG2(__lsx_vilvh_w, _t2, _t0, _t3, _t1, _tmp2, _tmp6);              \
+    DUP2_ARG2(__lsx_vilvl_w, _t6, _t4, _t7, _t5, _tmp1, _tmp5);              \
+    DUP2_ARG2(__lsx_vilvh_w, _t6, _t4, _t7, _t5, _tmp3, _tmp7);              \
+    DUP2_ARG2(__lsx_vilvl_d, _tmp1, _tmp0, _tmp3, _tmp2, _out0, _out2);      \
+    DUP2_ARG2(__lsx_vilvh_d, _tmp1, _tmp0, _tmp3, _tmp2, _out1, _out3);      \
+    DUP2_ARG2(__lsx_vilvl_d, _tmp5, _tmp4, _tmp7, _tmp6, _out4, _out6);      \
+    DUP2_ARG2(__lsx_vilvh_d, _tmp5, _tmp4, _tmp7, _tmp6, _out5, _out7);      \
+  }
+
+/*
+ * =============================================================================
+ * Description : Butterfly of 4 input vectors
+ * Arguments   : Inputs  - in0, in1, in2, in3
+ *               Outputs - out0, out1, out2, out3
+ * Details     : Butterfly operation
+ * Example     :
+ *               out0 = in0 + in3;
+ *               out1 = in1 + in2;
+ *               out2 = in1 - in2;
+ *               out3 = in0 - in3;
+ * =============================================================================
+ */
+#define LSX_BUTTERFLY_4_B(_in0, _in1, _in2, _in3, _out0, _out1, _out2, _out3) \
+  {                                                                           \
+    _out0 = __lsx_vadd_b(_in0, _in3);                                         \
+    _out1 = __lsx_vadd_b(_in1, _in2);                                         \
+    _out2 = __lsx_vsub_b(_in1, _in2);                                         \
+    _out3 = __lsx_vsub_b(_in0, _in3);                                         \
+  }
+#define LSX_BUTTERFLY_4_H(_in0, _in1, _in2, _in3, _out0, _out1, _out2, _out3) \
+  {                                                                           \
+    _out0 = __lsx_vadd_h(_in0, _in3);                                         \
+    _out1 = __lsx_vadd_h(_in1, _in2);                                         \
+    _out2 = __lsx_vsub_h(_in1, _in2);                                         \
+    _out3 = __lsx_vsub_h(_in0, _in3);                                         \
+  }
+#define LSX_BUTTERFLY_4_W(_in0, _in1, _in2, _in3, _out0, _out1, _out2, _out3) \
+  {                                                                           \
+    _out0 = __lsx_vadd_w(_in0, _in3);                                         \
+    _out1 = __lsx_vadd_w(_in1, _in2);                                         \
+    _out2 = __lsx_vsub_w(_in1, _in2);                                         \
+    _out3 = __lsx_vsub_w(_in0, _in3);                                         \
+  }
+#define LSX_BUTTERFLY_4_D(_in0, _in1, _in2, _in3, _out0, _out1, _out2, _out3) \
+  {                                                                           \
+    _out0 = __lsx_vadd_d(_in0, _in3);                                         \
+    _out1 = __lsx_vadd_d(_in1, _in2);                                         \
+    _out2 = __lsx_vsub_d(_in1, _in2);                                         \
+    _out3 = __lsx_vsub_d(_in0, _in3);                                         \
+  }
+
+/*
+ * =============================================================================
+ * Description : Butterfly of 8 input vectors
+ * Arguments   : Inputs  - _in0, _in1, _in2, _in3, ~
+ *               Outputs - _out0, _out1, _out2, _out3, ~
+ * Details     : Butterfly operation
+ * Example     :
+ *              _out0 = _in0 + _in7;
+ *              _out1 = _in1 + _in6;
+ *              _out2 = _in2 + _in5;
+ *              _out3 = _in3 + _in4;
+ *              _out4 = _in3 - _in4;
+ *              _out5 = _in2 - _in5;
+ *              _out6 = _in1 - _in6;
+ *              _out7 = _in0 - _in7;
+ * =============================================================================
+ */
+#define LSX_BUTTERFLY_8_B(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                          _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                          _out7)                                           \
+  {                                                                        \
+    _out0 = __lsx_vadd_b(_in0, _in7);                                      \
+    _out1 = __lsx_vadd_b(_in1, _in6);                                      \
+    _out2 = __lsx_vadd_b(_in2, _in5);                                      \
+    _out3 = __lsx_vadd_b(_in3, _in4);                                      \
+    _out4 = __lsx_vsub_b(_in3, _in4);                                      \
+    _out5 = __lsx_vsub_b(_in2, _in5);                                      \
+    _out6 = __lsx_vsub_b(_in1, _in6);                                      \
+    _out7 = __lsx_vsub_b(_in0, _in7);                                      \
+  }
+
+#define LSX_BUTTERFLY_8_H(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                          _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                          _out7)                                           \
+  {                                                                        \
+    _out0 = __lsx_vadd_h(_in0, _in7);                                      \
+    _out1 = __lsx_vadd_h(_in1, _in6);                                      \
+    _out2 = __lsx_vadd_h(_in2, _in5);                                      \
+    _out3 = __lsx_vadd_h(_in3, _in4);                                      \
+    _out4 = __lsx_vsub_h(_in3, _in4);                                      \
+    _out5 = __lsx_vsub_h(_in2, _in5);                                      \
+    _out6 = __lsx_vsub_h(_in1, _in6);                                      \
+    _out7 = __lsx_vsub_h(_in0, _in7);                                      \
+  }
+
+#define LSX_BUTTERFLY_8_W(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                          _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                          _out7)                                           \
+  {                                                                        \
+    _out0 = __lsx_vadd_w(_in0, _in7);                                      \
+    _out1 = __lsx_vadd_w(_in1, _in6);                                      \
+    _out2 = __lsx_vadd_w(_in2, _in5);                                      \
+    _out3 = __lsx_vadd_w(_in3, _in4);                                      \
+    _out4 = __lsx_vsub_w(_in3, _in4);                                      \
+    _out5 = __lsx_vsub_w(_in2, _in5);                                      \
+    _out6 = __lsx_vsub_w(_in1, _in6);                                      \
+    _out7 = __lsx_vsub_w(_in0, _in7);                                      \
+  }
+
+#define LSX_BUTTERFLY_8_D(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                          _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                          _out7)                                           \
+  {                                                                        \
+    _out0 = __lsx_vadd_d(_in0, _in7);                                      \
+    _out1 = __lsx_vadd_d(_in1, _in6);                                      \
+    _out2 = __lsx_vadd_d(_in2, _in5);                                      \
+    _out3 = __lsx_vadd_d(_in3, _in4);                                      \
+    _out4 = __lsx_vsub_d(_in3, _in4);                                      \
+    _out5 = __lsx_vsub_d(_in2, _in5);                                      \
+    _out6 = __lsx_vsub_d(_in1, _in6);                                      \
+    _out7 = __lsx_vsub_d(_in0, _in7);                                      \
+  }
+
+#endif  // LSX
+
+#ifdef __loongarch_asx
+#include <lasxintrin.h>
+/*
+ * =============================================================================
+ * Description : Dot product of byte vector elements
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ *               Return Type - signed halfword
+ * Details     : Unsigned byte elements from in_h are multiplied with
+ *               unsigned byte elements from in_l producing a result
+ *               twice the size of input i.e. signed halfword.
+ *               Then this multiplied results of adjacent odd-even elements
+ *               are added to the out vector
+ * Example     : See out = __lasx_xvdp2_w_h(in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2_h_bu(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmulwev_h_bu(in_h, in_l);
+  out = __lasx_xvmaddwod_h_bu(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product of byte vector elements
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ *               Return Type - signed halfword
+ * Details     : Signed byte elements from in_h are multiplied with
+ *               signed byte elements from in_l producing a result
+ *               twice the size of input i.e. signed halfword.
+ *               Then this multiplication results of adjacent odd-even elements
+ *               are added to the out vector
+ * Example     : See out = __lasx_xvdp2_w_h(in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2_h_b(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmulwev_h_b(in_h, in_l);
+  out = __lasx_xvmaddwod_h_b(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product of halfword vector elements
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ *               Return Type - signed word
+ * Details     : Signed halfword elements from in_h are multiplied with
+ *               signed halfword elements from in_l producing a result
+ *               twice the size of input i.e. signed word.
+ *               Then this multiplied results of adjacent odd-even elements
+ *               are added to the out vector.
+ * Example     : out = __lasx_xvdp2_w_h(in_h, in_l)
+ *        in_h : 1,2,3,4, 5,6,7,8, 1,2,3,4, 5,6,7,8
+ *        in_l : 8,7,6,5, 4,3,2,1, 8,7,6,5, 4,3,2,1
+ *         out : 22,38,38,22, 22,38,38,22
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2_w_h(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmulwev_w_h(in_h, in_l);
+  out = __lasx_xvmaddwod_w_h(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product of word vector elements
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ *               Return Type - signed double
+ * Details     : Signed word elements from in_h are multiplied with
+ *               signed word elements from in_l producing a result
+ *               twice the size of input i.e. signed double-word.
+ *               Then this multiplied results of adjacent odd-even elements
+ *               are added to the out vector.
+ * Example     : See out = __lasx_xvdp2_w_h(in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2_d_w(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmulwev_d_w(in_h, in_l);
+  out = __lasx_xvmaddwod_d_w(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product of halfword vector elements
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ *               Return Type - signed word
+ * Details     : Unsigned halfword elements from in_h are multiplied with
+ *               signed halfword elements from in_l producing a result
+ *               twice the size of input i.e. unsigned word.
+ *               Multiplication result of adjacent odd-even elements
+ *               are added to the out vector
+ * Example     : See out = __lasx_xvdp2_w_h(in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2_w_hu_h(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmulwev_w_hu_h(in_h, in_l);
+  out = __lasx_xvmaddwod_w_hu_h(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product & addition of byte vector elements
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ *               Return Type - halfword
+ * Details     : Signed byte elements from in_h are multiplied with
+ *               signed byte elements from in_l producing a result
+ *               twice the size of input i.e. signed halfword.
+ *               Then this multiplied results of adjacent odd-even elements
+ *               are added to the in_c vector.
+ * Example     : See out = __lasx_xvdp2add_w_h(in_c, in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2add_h_b(__m256i in_c,
+                                          __m256i in_h,
+                                          __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmaddwev_h_b(in_c, in_h, in_l);
+  out = __lasx_xvmaddwod_h_b(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product & addition of byte vector elements
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ *               Return Type - halfword
+ * Details     : Unsigned byte elements from in_h are multiplied with
+ *               unsigned byte elements from in_l producing a result
+ *               twice the size of input i.e. signed halfword.
+ *               Then this multiplied results of adjacent odd-even elements
+ *               are added to the in_c vector.
+ * Example     : See out = __lasx_xvdp2add_w_h(in_c, in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2add_h_bu(__m256i in_c,
+                                           __m256i in_h,
+                                           __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmaddwev_h_bu(in_c, in_h, in_l);
+  out = __lasx_xvmaddwod_h_bu(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product & addition of byte vector elements
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ *               Return Type - halfword
+ * Details     : Unsigned byte elements from in_h are multiplied with
+ *               signed byte elements from in_l producing a result
+ *               twice the size of input i.e. signed halfword.
+ *               Then this multiplied results of adjacent odd-even elements
+ *               are added to the in_c vector.
+ * Example     : See out = __lasx_xvdp2add_w_h(in_c, in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2add_h_bu_b(__m256i in_c,
+                                             __m256i in_h,
+                                             __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmaddwev_h_bu_b(in_c, in_h, in_l);
+  out = __lasx_xvmaddwod_h_bu_b(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product of halfword vector elements
+ * Arguments   : Inputs - in_c, in_h, in_l
+ *               Output - out
+ *               Return Type - per RTYPE
+ * Details     : Signed halfword elements from in_h are multiplied with
+ *               signed halfword elements from in_l producing a result
+ *               twice the size of input i.e. signed word.
+ *               Multiplication result of adjacent odd-even elements
+ *               are added to the in_c vector.
+ * Example     : out = __lasx_xvdp2add_w_h(in_c, in_h, in_l)
+ *        in_c : 1,2,3,4, 1,2,3,4
+ *        in_h : 1,2,3,4, 5,6,7,8, 1,2,3,4, 5,6,7,8,
+ *        in_l : 8,7,6,5, 4,3,2,1, 8,7,6,5, 4,3,2,1,
+ *         out : 23,40,41,26, 23,40,41,26
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2add_w_h(__m256i in_c,
+                                          __m256i in_h,
+                                          __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmaddwev_w_h(in_c, in_h, in_l);
+  out = __lasx_xvmaddwod_w_h(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product of halfword vector elements
+ * Arguments   : Inputs - in_c, in_h, in_l
+ *               Output - out
+ *               Return Type - signed word
+ * Details     : Unsigned halfword elements from in_h are multiplied with
+ *               unsigned halfword elements from in_l producing a result
+ *               twice the size of input i.e. signed word.
+ *               Multiplication result of adjacent odd-even elements
+ *               are added to the in_c vector.
+ * Example     : See out = __lasx_xvdp2add_w_h(in_c, in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2add_w_hu(__m256i in_c,
+                                           __m256i in_h,
+                                           __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmaddwev_w_hu(in_c, in_h, in_l);
+  out = __lasx_xvmaddwod_w_hu(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product of halfword vector elements
+ * Arguments   : Inputs - in_c, in_h, in_l
+ *               Output - out
+ *               Return Type - signed word
+ * Details     : Unsigned halfword elements from in_h are multiplied with
+ *               signed halfword elements from in_l producing a result
+ *               twice the size of input i.e. signed word.
+ *               Multiplication result of adjacent odd-even elements
+ *               are added to the in_c vector
+ * Example     : See out = __lasx_xvdp2add_w_h(in_c, in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2add_w_hu_h(__m256i in_c,
+                                             __m256i in_h,
+                                             __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmaddwev_w_hu_h(in_c, in_h, in_l);
+  out = __lasx_xvmaddwod_w_hu_h(out, in_h, in_l);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Vector Unsigned Dot Product and Subtract
+ * Arguments   : Inputs - in_c, in_h, in_l
+ *               Output - out
+ *               Return Type - signed halfword
+ * Details     : Unsigned byte elements from in_h are multiplied with
+ *               unsigned byte elements from in_l producing a result
+ *               twice the size of input i.e. signed halfword.
+ *               Multiplication result of adjacent odd-even elements
+ *               are added together and subtracted from double width elements
+ *               in_c vector.
+ * Example     : See out = __lasx_xvdp2sub_w_h(in_c, in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2sub_h_bu(__m256i in_c,
+                                           __m256i in_h,
+                                           __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmulwev_h_bu(in_h, in_l);
+  out = __lasx_xvmaddwod_h_bu(out, in_h, in_l);
+  out = __lasx_xvsub_h(in_c, out);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Vector Signed Dot Product and Subtract
+ * Arguments   : Inputs - in_c, in_h, in_l
+ *               Output - out
+ *               Return Type - signed word
+ * Details     : Signed halfword elements from in_h are multiplied with
+ *               Signed halfword elements from in_l producing a result
+ *               twice the size of input i.e. signed word.
+ *               Multiplication result of adjacent odd-even elements
+ *               are added together and subtracted from double width elements
+ *               in_c vector.
+ * Example     : out = __lasx_xvdp2sub_w_h(in_c, in_h, in_l)
+ *        in_c : 0,0,0,0, 0,0,0,0
+ *        in_h : 3,1,3,0, 0,0,0,1, 0,0,1,1, 0,0,0,1
+ *        in_l : 2,1,1,0, 1,0,0,0, 0,0,1,0, 1,0,0,1
+ *         out : -7,-3,0,0, 0,-1,0,-1
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp2sub_w_h(__m256i in_c,
+                                          __m256i in_h,
+                                          __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmulwev_w_h(in_h, in_l);
+  out = __lasx_xvmaddwod_w_h(out, in_h, in_l);
+  out = __lasx_xvsub_w(in_c, out);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Dot product of halfword vector elements
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ *               Return Type - signed word
+ * Details     : Signed halfword elements from in_h are multiplied with
+ *               signed halfword elements from in_l producing a result
+ *               four times the size of input i.e. signed doubleword.
+ *               Then this multiplication results of four adjacent elements
+ *               are added together and stored to the out vector.
+ * Example     : out = __lasx_xvdp4_d_h(in_h, in_l)
+ *        in_h :  3,1,3,0, 0,0,0,1, 0,0,1,-1, 0,0,0,1
+ *        in_l : -2,1,1,0, 1,0,0,0, 0,0,1, 0, 1,0,0,1
+ *         out : -2,0,1,1
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvdp4_d_h(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvmulwev_w_h(in_h, in_l);
+  out = __lasx_xvmaddwod_w_h(out, in_h, in_l);
+  out = __lasx_xvhaddw_d_w(out, out);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : The high half of the vector elements are expanded and
+ *               added after being doubled.
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ * Details     : The in_h vector and the in_l vector are added after the
+ *               higher half of the two-fold sign extension (signed byte
+ *               to signed halfword) and stored to the out vector.
+ * Example     : See out = __lasx_xvaddwh_w_h(in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvaddwh_h_b(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvilvh_b(in_h, in_l);
+  out = __lasx_xvhaddw_h_b(out, out);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : The high half of the vector elements are expanded and
+ *               added after being doubled.
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ * Details     : The in_h vector and the in_l vector are added after the
+ *               higher half of the two-fold sign extension (signed halfword
+ *               to signed word) and stored to the out vector.
+ * Example     : out = __lasx_xvaddwh_w_h(in_h, in_l)
+ *        in_h : 3, 0,3,0, 0,0,0,-1, 0,0,1,-1, 0,0,0,1
+ *        in_l : 2,-1,1,2, 1,0,0, 0, 1,0,1, 0, 1,0,0,1
+ *         out : 1,0,0,-1, 1,0,0, 2
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvaddwh_w_h(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvilvh_h(in_h, in_l);
+  out = __lasx_xvhaddw_w_h(out, out);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : The low half of the vector elements are expanded and
+ *               added after being doubled.
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ * Details     : The in_h vector and the in_l vector are added after the
+ *               lower half of the two-fold sign extension (signed byte
+ *               to signed halfword) and stored to the out vector.
+ * Example     : See out = __lasx_xvaddwl_w_h(in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvaddwl_h_b(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvilvl_b(in_h, in_l);
+  out = __lasx_xvhaddw_h_b(out, out);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : The low half of the vector elements are expanded and
+ *               added after being doubled.
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ * Details     : The in_h vector and the in_l vector are added after the
+ *               lower half of the two-fold sign extension (signed halfword
+ *               to signed word) and stored to the out vector.
+ * Example     : out = __lasx_xvaddwl_w_h(in_h, in_l)
+ *        in_h : 3, 0,3,0, 0,0,0,-1, 0,0,1,-1, 0,0,0,1
+ *        in_l : 2,-1,1,2, 1,0,0, 0, 1,0,1, 0, 1,0,0,1
+ *         out : 5,-1,4,2, 1,0,2,-1
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvaddwl_w_h(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvilvl_h(in_h, in_l);
+  out = __lasx_xvhaddw_w_h(out, out);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : The low half of the vector elements are expanded and
+ *               added after being doubled.
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ * Details     : The out vector and the out vector are added after the
+ *               lower half of the two-fold zero extension (unsigned byte
+ *               to unsigned halfword) and stored to the out vector.
+ * Example     : See out = __lasx_xvaddwl_w_h(in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvaddwl_h_bu(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvilvl_b(in_h, in_l);
+  out = __lasx_xvhaddw_hu_bu(out, out);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : The low half of the vector elements are expanded and
+ *               added after being doubled.
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ * Details     : The in_l vector after double zero extension (unsigned byte to
+ *               signed halfword)，added to the in_h vector.
+ * Example     : See out = __lasx_xvaddw_w_w_h(in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvaddw_h_h_bu(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvsllwil_hu_bu(in_l, 0);
+  out = __lasx_xvadd_h(in_h, out);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : The low half of the vector elements are expanded and
+ *               added after being doubled.
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ * Details     : The in_l vector after double sign extension (signed halfword to
+ *               signed word), added to the in_h vector.
+ * Example     : out = __lasx_xvaddw_w_w_h(in_h, in_l)
+ *        in_h : 0, 1,0,0, -1,0,0,1,
+ *        in_l : 2,-1,1,2,  1,0,0,0, 0,0,1,0, 1,0,0,1,
+ *         out : 2, 0,1,2, -1,0,1,1,
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvaddw_w_w_h(__m256i in_h, __m256i in_l) {
+  __m256i out;
+
+  out = __lasx_xvsllwil_w_h(in_l, 0);
+  out = __lasx_xvadd_w(in_h, out);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Multiplication and addition calculation after expansion
+ *               of the lower half of the vector.
+ * Arguments   : Inputs - in_c, in_h, in_l
+ *               Output - out
+ * Details     : The in_h vector and the in_l vector are multiplied after
+ *               the lower half of the two-fold sign extension (signed halfword
+ *               to signed word), and the result is added to the vector in_c,
+ *               then stored to the out vector.
+ * Example     : out = __lasx_xvmaddwl_w_h(in_c, in_h, in_l)
+ *        in_c : 1,2,3,4, 5,6,7,8
+ *        in_h : 1,2,3,4, 1,2,3,4, 5,6,7,8, 5,6,7,8
+ *        in_l : 200, 300, 400, 500,  2000, 3000, 4000, 5000,
+ *              -200,-300,-400,-500, -2000,-3000,-4000,-5000
+ *         out : 201, 602,1203,2004, -995, -1794,-2793,-3992
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvmaddwl_w_h(__m256i in_c,
+                                          __m256i in_h,
+                                          __m256i in_l) {
+  __m256i tmp0, tmp1, out;
+
+  tmp0 = __lasx_xvsllwil_w_h(in_h, 0);
+  tmp1 = __lasx_xvsllwil_w_h(in_l, 0);
+  tmp0 = __lasx_xvmul_w(tmp0, tmp1);
+  out = __lasx_xvadd_w(tmp0, in_c);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Multiplication and addition calculation after expansion
+ *               of the higher half of the vector.
+ * Arguments   : Inputs - in_c, in_h, in_l
+ *               Output - out
+ * Details     : The in_h vector and the in_l vector are multiplied after
+ *               the higher half of the two-fold sign extension (signed
+ *               halfword to signed word), and the result is added to
+ *               the vector in_c, then stored to the out vector.
+ * Example     : See out = __lasx_xvmaddwl_w_h(in_c, in_h, in_l)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvmaddwh_w_h(__m256i in_c,
+                                          __m256i in_h,
+                                          __m256i in_l) {
+  __m256i tmp0, tmp1, out;
+
+  tmp0 = __lasx_xvilvh_h(in_h, in_h);
+  tmp1 = __lasx_xvilvh_h(in_l, in_l);
+  tmp0 = __lasx_xvmulwev_w_h(tmp0, tmp1);
+  out = __lasx_xvadd_w(tmp0, in_c);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Multiplication calculation after expansion of the lower
+ *               half of the vector.
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ * Details     : The in_h vector and the in_l vector are multiplied after
+ *               the lower half of the two-fold sign extension (signed
+ *               halfword to signed word), then stored to the out vector.
+ * Example     : out = __lasx_xvmulwl_w_h(in_h, in_l)
+ *        in_h : 3,-1,3,0, 0,0,0,-1, 0,0,1,-1, 0,0,0,1
+ *        in_l : 2,-1,1,2, 1,0,0, 0, 0,0,1, 0, 1,0,0,1
+ *         out : 6,1,3,0, 0,0,1,0
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvmulwl_w_h(__m256i in_h, __m256i in_l) {
+  __m256i tmp0, tmp1, out;
+
+  tmp0 = __lasx_xvsllwil_w_h(in_h, 0);
+  tmp1 = __lasx_xvsllwil_w_h(in_l, 0);
+  out = __lasx_xvmul_w(tmp0, tmp1);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Multiplication calculation after expansion of the lower
+ *               half of the vector.
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ * Details     : The in_h vector and the in_l vector are multiplied after
+ *               the lower half of the two-fold sign extension (signed
+ *               halfword to signed word), then stored to the out vector.
+ * Example     : out = __lasx_xvmulwh_w_h(in_h, in_l)
+ *        in_h : 3,-1,3,0, 0,0,0,-1, 0,0,1,-1, 0,0,0,1
+ *        in_l : 2,-1,1,2, 1,0,0, 0, 0,0,1, 0, 1,0,0,1
+ *         out : 0,0,0,0, 0,0,0,1
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvmulwh_w_h(__m256i in_h, __m256i in_l) {
+  __m256i tmp0, tmp1, out;
+
+  tmp0 = __lasx_xvilvh_h(in_h, in_h);
+  tmp1 = __lasx_xvilvh_h(in_l, in_l);
+  out = __lasx_xvmulwev_w_h(tmp0, tmp1);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : The low half of the vector elements are added to the high half
+ *               after being doubled, then saturated.
+ * Arguments   : Inputs - in_h, in_l
+ *               Output - out
+ * Details     : The in_h vector adds the in_l vector after the lower half of
+ *               the two-fold zero extension (unsigned byte to unsigned
+ *               halfword) and then saturated. The results are stored to the out
+ *               vector.
+ * Example     : out = __lasx_xvsaddw_hu_hu_bu(in_h, in_l)
+ *        in_h : 2,65532,1,2, 1,0,0,0, 0,0,1,0, 1,0,0,1
+ *        in_l : 3,6,3,0, 0,0,0,1, 0,0,1,1, 0,0,0,1, 3,18,3,0, 0,0,0,1, 0,0,1,1,
+ *               0,0,0,1
+ *        out  : 5,65535,4,2, 1,0,0,1, 3,18,4,0, 1,0,0,2,
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvsaddw_hu_hu_bu(__m256i in_h, __m256i in_l) {
+  __m256i tmp1, out;
+  __m256i zero = {0};
+
+  tmp1 = __lasx_xvilvl_b(zero, in_l);
+  out = __lasx_xvsadd_hu(in_h, tmp1);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Clip all halfword elements of input vector between min & max
+ *               out = ((in) < (min)) ? (min) : (((in) > (max)) ? (max) : (in))
+ * Arguments   : Inputs  - in    (input vector)
+ *                       - min   (min threshold)
+ *                       - max   (max threshold)
+ *               Outputs - in    (output vector with clipped elements)
+ *               Return Type - signed halfword
+ * Example     : out = __lasx_xvclip_h(in, min, max)
+ *          in : -8,2,280,249, -8,255,280,249, 4,4,4,4, 5,5,5,5
+ *         min : 1,1,1,1, 1,1,1,1, 1,1,1,1, 1,1,1,1
+ *         max : 9,9,9,9, 9,9,9,9, 9,9,9,9, 9,9,9,9
+ *         out : 1,2,9,9, 1,9,9,9, 4,4,4,4, 5,5,5,5
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvclip_h(__m256i in, __m256i min, __m256i max) {
+  __m256i out;
+
+  out = __lasx_xvmax_h(min, in);
+  out = __lasx_xvmin_h(max, out);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Clip all signed halfword elements of input vector
+ *               between 0 & 255
+ * Arguments   : Inputs  - in   (input vector)
+ *               Outputs - out  (output vector with clipped elements)
+ *               Return Type - signed halfword
+ * Example     : See out = __lasx_xvclip255_w(in)
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvclip255_h(__m256i in) {
+  __m256i out;
+
+  out = __lasx_xvmaxi_h(in, 0);
+  out = __lasx_xvsat_hu(out, 7);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Clip all signed word elements of input vector
+ *               between 0 & 255
+ * Arguments   : Inputs - in   (input vector)
+ *               Output - out  (output vector with clipped elements)
+ *               Return Type - signed word
+ * Example     : out = __lasx_xvclip255_w(in)
+ *          in : -8,255,280,249, -8,255,280,249
+ *         out :  0,255,255,249,  0,255,255,249
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvclip255_w(__m256i in) {
+  __m256i out;
+
+  out = __lasx_xvmaxi_w(in, 0);
+  out = __lasx_xvsat_wu(out, 7);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Indexed halfword element values are replicated to all
+ *               elements in output vector. If 'idx < 8' use xvsplati_l_*,
+ *               if 'idx >= 8' use xvsplati_h_*.
+ * Arguments   : Inputs - in, idx
+ *               Output - out
+ * Details     : Idx element value from in vector is replicated to all
+ *               elements in out vector.
+ *               Valid index range for halfword operation is 0-7
+ * Example     : out = __lasx_xvsplati_l_h(in, idx)
+ *          in : 20,10,11,12, 13,14,15,16, 0,0,2,0, 0,0,0,0
+ *         idx : 0x02
+ *         out : 11,11,11,11, 11,11,11,11, 11,11,11,11, 11,11,11,11
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvsplati_l_h(__m256i in, int idx) {
+  __m256i out;
+
+  out = __lasx_xvpermi_q(in, in, 0x02);
+  out = __lasx_xvreplve_h(out, idx);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Indexed halfword element values are replicated to all
+ *               elements in output vector. If 'idx < 8' use xvsplati_l_*,
+ *               if 'idx >= 8' use xvsplati_h_*.
+ * Arguments   : Inputs - in, idx
+ *               Output - out
+ * Details     : Idx element value from in vector is replicated to all
+ *               elements in out vector.
+ *               Valid index range for halfword operation is 0-7
+ * Example     : out = __lasx_xvsplati_h_h(in, idx)
+ *          in : 20,10,11,12, 13,14,15,16, 0,2,0,0, 0,0,0,0
+ *         idx : 0x09
+ *         out : 2,2,2,2, 2,2,2,2, 2,2,2,2, 2,2,2,2
+ * =============================================================================
+ */
+static inline __m256i __lasx_xvsplati_h_h(__m256i in, int idx) {
+  __m256i out;
+
+  out = __lasx_xvpermi_q(in, in, 0x13);
+  out = __lasx_xvreplve_h(out, idx);
+  return out;
+}
+
+/*
+ * =============================================================================
+ * Description : Transpose 4x4 block with double-word elements in vectors
+ * Arguments   : Inputs  - _in0, _in1, _in2, _in3
+ *               Outputs - _out0, _out1, _out2, _out3
+ * Example     : LASX_TRANSPOSE4x4_D
+ *        _in0 : 1,2,3,4
+ *        _in1 : 1,2,3,4
+ *        _in2 : 1,2,3,4
+ *        _in3 : 1,2,3,4
+ *
+ *       _out0 : 1,1,1,1
+ *       _out1 : 2,2,2,2
+ *       _out2 : 3,3,3,3
+ *       _out3 : 4,4,4,4
+ * =============================================================================
+ */
+#define LASX_TRANSPOSE4x4_D(_in0, _in1, _in2, _in3, _out0, _out1, _out2, \
+                            _out3)                                       \
+  {                                                                      \
+    __m256i _tmp0, _tmp1, _tmp2, _tmp3;                                  \
+    _tmp0 = __lasx_xvilvl_d(_in1, _in0);                                 \
+    _tmp1 = __lasx_xvilvh_d(_in1, _in0);                                 \
+    _tmp2 = __lasx_xvilvl_d(_in3, _in2);                                 \
+    _tmp3 = __lasx_xvilvh_d(_in3, _in2);                                 \
+    _out0 = __lasx_xvpermi_q(_tmp2, _tmp0, 0x20);                        \
+    _out2 = __lasx_xvpermi_q(_tmp2, _tmp0, 0x31);                        \
+    _out1 = __lasx_xvpermi_q(_tmp3, _tmp1, 0x20);                        \
+    _out3 = __lasx_xvpermi_q(_tmp3, _tmp1, 0x31);                        \
+  }
+
+/*
+ * =============================================================================
+ * Description : Transpose 8x8 block with word elements in vectors
+ * Arguments   : Inputs  - _in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7
+ *               Outputs - _out0, _out1, _out2, _out3, _out4, _out5, _out6,
+ *               _out7
+ * Example     : LASX_TRANSPOSE8x8_W
+ *        _in0 : 1,2,3,4,5,6,7,8
+ *        _in1 : 2,2,3,4,5,6,7,8
+ *        _in2 : 3,2,3,4,5,6,7,8
+ *        _in3 : 4,2,3,4,5,6,7,8
+ *        _in4 : 5,2,3,4,5,6,7,8
+ *        _in5 : 6,2,3,4,5,6,7,8
+ *        _in6 : 7,2,3,4,5,6,7,8
+ *        _in7 : 8,2,3,4,5,6,7,8
+ *
+ *       _out0 : 1,2,3,4,5,6,7,8
+ *       _out1 : 2,2,2,2,2,2,2,2
+ *       _out2 : 3,3,3,3,3,3,3,3
+ *       _out3 : 4,4,4,4,4,4,4,4
+ *       _out4 : 5,5,5,5,5,5,5,5
+ *       _out5 : 6,6,6,6,6,6,6,6
+ *       _out6 : 7,7,7,7,7,7,7,7
+ *       _out7 : 8,8,8,8,8,8,8,8
+ * =============================================================================
+ */
+#define LASX_TRANSPOSE8x8_W(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                            _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                            _out7)                                           \
+  {                                                                          \
+    __m256i _s0_m, _s1_m;                                                    \
+    __m256i _tmp0_m, _tmp1_m, _tmp2_m, _tmp3_m;                              \
+    __m256i _tmp4_m, _tmp5_m, _tmp6_m, _tmp7_m;                              \
+                                                                             \
+    _s0_m = __lasx_xvilvl_w(_in2, _in0);                                     \
+    _s1_m = __lasx_xvilvl_w(_in3, _in1);                                     \
+    _tmp0_m = __lasx_xvilvl_w(_s1_m, _s0_m);                                 \
+    _tmp1_m = __lasx_xvilvh_w(_s1_m, _s0_m);                                 \
+    _s0_m = __lasx_xvilvh_w(_in2, _in0);                                     \
+    _s1_m = __lasx_xvilvh_w(_in3, _in1);                                     \
+    _tmp2_m = __lasx_xvilvl_w(_s1_m, _s0_m);                                 \
+    _tmp3_m = __lasx_xvilvh_w(_s1_m, _s0_m);                                 \
+    _s0_m = __lasx_xvilvl_w(_in6, _in4);                                     \
+    _s1_m = __lasx_xvilvl_w(_in7, _in5);                                     \
+    _tmp4_m = __lasx_xvilvl_w(_s1_m, _s0_m);                                 \
+    _tmp5_m = __lasx_xvilvh_w(_s1_m, _s0_m);                                 \
+    _s0_m = __lasx_xvilvh_w(_in6, _in4);                                     \
+    _s1_m = __lasx_xvilvh_w(_in7, _in5);                                     \
+    _tmp6_m = __lasx_xvilvl_w(_s1_m, _s0_m);                                 \
+    _tmp7_m = __lasx_xvilvh_w(_s1_m, _s0_m);                                 \
+    _out0 = __lasx_xvpermi_q(_tmp4_m, _tmp0_m, 0x20);                        \
+    _out1 = __lasx_xvpermi_q(_tmp5_m, _tmp1_m, 0x20);                        \
+    _out2 = __lasx_xvpermi_q(_tmp6_m, _tmp2_m, 0x20);                        \
+    _out3 = __lasx_xvpermi_q(_tmp7_m, _tmp3_m, 0x20);                        \
+    _out4 = __lasx_xvpermi_q(_tmp4_m, _tmp0_m, 0x31);                        \
+    _out5 = __lasx_xvpermi_q(_tmp5_m, _tmp1_m, 0x31);                        \
+    _out6 = __lasx_xvpermi_q(_tmp6_m, _tmp2_m, 0x31);                        \
+    _out7 = __lasx_xvpermi_q(_tmp7_m, _tmp3_m, 0x31);                        \
+  }
+
+/*
+ * =============================================================================
+ * Description : Transpose input 16x8 byte block
+ * Arguments   : Inputs  - _in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,
+ *                         _in8, _in9, _in10, _in11, _in12, _in13, _in14, _in15
+ *                         (input 16x8 byte block)
+ *               Outputs - _out0, _out1, _out2, _out3, _out4, _out5, _out6,
+ *                         _out7 (output 8x16 byte block)
+ * Details     : The rows of the matrix become columns, and the columns become
+ *               rows.
+ * Example     : See LASX_TRANSPOSE16x8_H
+ * =============================================================================
+ */
+#define LASX_TRANSPOSE16x8_B(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                             _in8, _in9, _in10, _in11, _in12, _in13, _in14,   \
+                             _in15, _out0, _out1, _out2, _out3, _out4, _out5, \
+                             _out6, _out7)                                    \
+  {                                                                           \
+    __m256i _tmp0_m, _tmp1_m, _tmp2_m, _tmp3_m;                               \
+    __m256i _tmp4_m, _tmp5_m, _tmp6_m, _tmp7_m;                               \
+                                                                              \
+    _tmp0_m = __lasx_xvilvl_b(_in2, _in0);                                    \
+    _tmp1_m = __lasx_xvilvl_b(_in3, _in1);                                    \
+    _tmp2_m = __lasx_xvilvl_b(_in6, _in4);                                    \
+    _tmp3_m = __lasx_xvilvl_b(_in7, _in5);                                    \
+    _tmp4_m = __lasx_xvilvl_b(_in10, _in8);                                   \
+    _tmp5_m = __lasx_xvilvl_b(_in11, _in9);                                   \
+    _tmp6_m = __lasx_xvilvl_b(_in14, _in12);                                  \
+    _tmp7_m = __lasx_xvilvl_b(_in15, _in13);                                  \
+    _out0 = __lasx_xvilvl_b(_tmp1_m, _tmp0_m);                                \
+    _out1 = __lasx_xvilvh_b(_tmp1_m, _tmp0_m);                                \
+    _out2 = __lasx_xvilvl_b(_tmp3_m, _tmp2_m);                                \
+    _out3 = __lasx_xvilvh_b(_tmp3_m, _tmp2_m);                                \
+    _out4 = __lasx_xvilvl_b(_tmp5_m, _tmp4_m);                                \
+    _out5 = __lasx_xvilvh_b(_tmp5_m, _tmp4_m);                                \
+    _out6 = __lasx_xvilvl_b(_tmp7_m, _tmp6_m);                                \
+    _out7 = __lasx_xvilvh_b(_tmp7_m, _tmp6_m);                                \
+    _tmp0_m = __lasx_xvilvl_w(_out2, _out0);                                  \
+    _tmp2_m = __lasx_xvilvh_w(_out2, _out0);                                  \
+    _tmp4_m = __lasx_xvilvl_w(_out3, _out1);                                  \
+    _tmp6_m = __lasx_xvilvh_w(_out3, _out1);                                  \
+    _tmp1_m = __lasx_xvilvl_w(_out6, _out4);                                  \
+    _tmp3_m = __lasx_xvilvh_w(_out6, _out4);                                  \
+    _tmp5_m = __lasx_xvilvl_w(_out7, _out5);                                  \
+    _tmp7_m = __lasx_xvilvh_w(_out7, _out5);                                  \
+    _out0 = __lasx_xvilvl_d(_tmp1_m, _tmp0_m);                                \
+    _out1 = __lasx_xvilvh_d(_tmp1_m, _tmp0_m);                                \
+    _out2 = __lasx_xvilvl_d(_tmp3_m, _tmp2_m);                                \
+    _out3 = __lasx_xvilvh_d(_tmp3_m, _tmp2_m);                                \
+    _out4 = __lasx_xvilvl_d(_tmp5_m, _tmp4_m);                                \
+    _out5 = __lasx_xvilvh_d(_tmp5_m, _tmp4_m);                                \
+    _out6 = __lasx_xvilvl_d(_tmp7_m, _tmp6_m);                                \
+    _out7 = __lasx_xvilvh_d(_tmp7_m, _tmp6_m);                                \
+  }
+
+/*
+ * =============================================================================
+ * Description : Transpose input 16x8 byte block
+ * Arguments   : Inputs  - _in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,
+ *                         _in8, _in9, _in10, _in11, _in12, _in13, _in14, _in15
+ *                         (input 16x8 byte block)
+ *               Outputs - _out0, _out1, _out2, _out3, _out4, _out5, _out6,
+ *                         _out7 (output 8x16 byte block)
+ * Details     : The rows of the matrix become columns, and the columns become
+ *               rows.
+ * Example     : LASX_TRANSPOSE16x8_H
+ *        _in0 : 1,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *        _in1 : 2,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *        _in2 : 3,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *        _in3 : 4,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *        _in4 : 5,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *        _in5 : 6,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *        _in6 : 7,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *        _in7 : 8,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *        _in8 : 9,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *        _in9 : 1,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *       _in10 : 0,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *       _in11 : 2,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *       _in12 : 3,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *       _in13 : 7,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *       _in14 : 5,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *       _in15 : 6,2,3,4,5,6,7,8,0,0,0,0,0,0,0,0
+ *
+ *       _out0 : 1,2,3,4,5,6,7,8,9,1,0,2,3,7,5,6
+ *       _out1 : 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2
+ *       _out2 : 3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3
+ *       _out3 : 4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4
+ *       _out4 : 5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5
+ *       _out5 : 6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6
+ *       _out6 : 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7
+ *       _out7 : 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8
+ * =============================================================================
+ */
+#define LASX_TRANSPOSE16x8_H(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                             _in8, _in9, _in10, _in11, _in12, _in13, _in14,   \
+                             _in15, _out0, _out1, _out2, _out3, _out4, _out5, \
+                             _out6, _out7)                                    \
+  {                                                                           \
+    __m256i _tmp0_m, _tmp1_m, _tmp2_m, _tmp3_m;                               \
+    __m256i _tmp4_m, _tmp5_m, _tmp6_m, _tmp7_m;                               \
+    __m256i _t0, _t1, _t2, _t3, _t4, _t5, _t6, _t7;                           \
+                                                                              \
+    _tmp0_m = __lasx_xvilvl_h(_in2, _in0);                                    \
+    _tmp1_m = __lasx_xvilvl_h(_in3, _in1);                                    \
+    _tmp2_m = __lasx_xvilvl_h(_in6, _in4);                                    \
+    _tmp3_m = __lasx_xvilvl_h(_in7, _in5);                                    \
+    _tmp4_m = __lasx_xvilvl_h(_in10, _in8);                                   \
+    _tmp5_m = __lasx_xvilvl_h(_in11, _in9);                                   \
+    _tmp6_m = __lasx_xvilvl_h(_in14, _in12);                                  \
+    _tmp7_m = __lasx_xvilvl_h(_in15, _in13);                                  \
+    _t0 = __lasx_xvilvl_h(_tmp1_m, _tmp0_m);                                  \
+    _t1 = __lasx_xvilvh_h(_tmp1_m, _tmp0_m);                                  \
+    _t2 = __lasx_xvilvl_h(_tmp3_m, _tmp2_m);                                  \
+    _t3 = __lasx_xvilvh_h(_tmp3_m, _tmp2_m);                                  \
+    _t4 = __lasx_xvilvl_h(_tmp5_m, _tmp4_m);                                  \
+    _t5 = __lasx_xvilvh_h(_tmp5_m, _tmp4_m);                                  \
+    _t6 = __lasx_xvilvl_h(_tmp7_m, _tmp6_m);                                  \
+    _t7 = __lasx_xvilvh_h(_tmp7_m, _tmp6_m);                                  \
+    _tmp0_m = __lasx_xvilvl_d(_t2, _t0);                                      \
+    _tmp2_m = __lasx_xvilvh_d(_t2, _t0);                                      \
+    _tmp4_m = __lasx_xvilvl_d(_t3, _t1);                                      \
+    _tmp6_m = __lasx_xvilvh_d(_t3, _t1);                                      \
+    _tmp1_m = __lasx_xvilvl_d(_t6, _t4);                                      \
+    _tmp3_m = __lasx_xvilvh_d(_t6, _t4);                                      \
+    _tmp5_m = __lasx_xvilvl_d(_t7, _t5);                                      \
+    _tmp7_m = __lasx_xvilvh_d(_t7, _t5);                                      \
+    _out0 = __lasx_xvpermi_q(_tmp1_m, _tmp0_m, 0x20);                         \
+    _out1 = __lasx_xvpermi_q(_tmp3_m, _tmp2_m, 0x20);                         \
+    _out2 = __lasx_xvpermi_q(_tmp5_m, _tmp4_m, 0x20);                         \
+    _out3 = __lasx_xvpermi_q(_tmp7_m, _tmp6_m, 0x20);                         \
+                                                                              \
+    _tmp0_m = __lasx_xvilvh_h(_in2, _in0);                                    \
+    _tmp1_m = __lasx_xvilvh_h(_in3, _in1);                                    \
+    _tmp2_m = __lasx_xvilvh_h(_in6, _in4);                                    \
+    _tmp3_m = __lasx_xvilvh_h(_in7, _in5);                                    \
+    _tmp4_m = __lasx_xvilvh_h(_in10, _in8);                                   \
+    _tmp5_m = __lasx_xvilvh_h(_in11, _in9);                                   \
+    _tmp6_m = __lasx_xvilvh_h(_in14, _in12);                                  \
+    _tmp7_m = __lasx_xvilvh_h(_in15, _in13);                                  \
+    _t0 = __lasx_xvilvl_h(_tmp1_m, _tmp0_m);                                  \
+    _t1 = __lasx_xvilvh_h(_tmp1_m, _tmp0_m);                                  \
+    _t2 = __lasx_xvilvl_h(_tmp3_m, _tmp2_m);                                  \
+    _t3 = __lasx_xvilvh_h(_tmp3_m, _tmp2_m);                                  \
+    _t4 = __lasx_xvilvl_h(_tmp5_m, _tmp4_m);                                  \
+    _t5 = __lasx_xvilvh_h(_tmp5_m, _tmp4_m);                                  \
+    _t6 = __lasx_xvilvl_h(_tmp7_m, _tmp6_m);                                  \
+    _t7 = __lasx_xvilvh_h(_tmp7_m, _tmp6_m);                                  \
+    _tmp0_m = __lasx_xvilvl_d(_t2, _t0);                                      \
+    _tmp2_m = __lasx_xvilvh_d(_t2, _t0);                                      \
+    _tmp4_m = __lasx_xvilvl_d(_t3, _t1);                                      \
+    _tmp6_m = __lasx_xvilvh_d(_t3, _t1);                                      \
+    _tmp1_m = __lasx_xvilvl_d(_t6, _t4);                                      \
+    _tmp3_m = __lasx_xvilvh_d(_t6, _t4);                                      \
+    _tmp5_m = __lasx_xvilvl_d(_t7, _t5);                                      \
+    _tmp7_m = __lasx_xvilvh_d(_t7, _t5);                                      \
+    _out4 = __lasx_xvpermi_q(_tmp1_m, _tmp0_m, 0x20);                         \
+    _out5 = __lasx_xvpermi_q(_tmp3_m, _tmp2_m, 0x20);                         \
+    _out6 = __lasx_xvpermi_q(_tmp5_m, _tmp4_m, 0x20);                         \
+    _out7 = __lasx_xvpermi_q(_tmp7_m, _tmp6_m, 0x20);                         \
+  }
+
+/*
+ * =============================================================================
+ * Description : Transpose 4x4 block with halfword elements in vectors
+ * Arguments   : Inputs  - _in0, _in1, _in2, _in3
+ *               Outputs - _out0, _out1, _out2, _out3
+ *               Return Type - signed halfword
+ * Details     : The rows of the matrix become columns, and the columns become
+ *               rows.
+ * Example     : See LASX_TRANSPOSE8x8_H
+ * =============================================================================
+ */
+#define LASX_TRANSPOSE4x4_H(_in0, _in1, _in2, _in3, _out0, _out1, _out2, \
+                            _out3)                                       \
+  {                                                                      \
+    __m256i _s0_m, _s1_m;                                                \
+                                                                         \
+    _s0_m = __lasx_xvilvl_h(_in1, _in0);                                 \
+    _s1_m = __lasx_xvilvl_h(_in3, _in2);                                 \
+    _out0 = __lasx_xvilvl_w(_s1_m, _s0_m);                               \
+    _out2 = __lasx_xvilvh_w(_s1_m, _s0_m);                               \
+    _out1 = __lasx_xvilvh_d(_out0, _out0);                               \
+    _out3 = __lasx_xvilvh_d(_out2, _out2);                               \
+  }
+
+/*
+ * =============================================================================
+ * Description : Transpose input 8x8 byte block
+ * Arguments   : Inputs  - _in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7
+ *                         (input 8x8 byte block)
+ *               Outputs - _out0, _out1, _out2, _out3, _out4, _out5, _out6,
+ *                         _out7 (output 8x8 byte block)
+ * Example     : See LASX_TRANSPOSE8x8_H
+ * =============================================================================
+ */
+#define LASX_TRANSPOSE8x8_B(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                            _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                            _out7)                                           \
+  {                                                                          \
+    __m256i _tmp0_m, _tmp1_m, _tmp2_m, _tmp3_m;                              \
+    __m256i _tmp4_m, _tmp5_m, _tmp6_m, _tmp7_m;                              \
+    _tmp0_m = __lasx_xvilvl_b(_in2, _in0);                                   \
+    _tmp1_m = __lasx_xvilvl_b(_in3, _in1);                                   \
+    _tmp2_m = __lasx_xvilvl_b(_in6, _in4);                                   \
+    _tmp3_m = __lasx_xvilvl_b(_in7, _in5);                                   \
+    _tmp4_m = __lasx_xvilvl_b(_tmp1_m, _tmp0_m);                             \
+    _tmp5_m = __lasx_xvilvh_b(_tmp1_m, _tmp0_m);                             \
+    _tmp6_m = __lasx_xvilvl_b(_tmp3_m, _tmp2_m);                             \
+    _tmp7_m = __lasx_xvilvh_b(_tmp3_m, _tmp2_m);                             \
+    _out0 = __lasx_xvilvl_w(_tmp6_m, _tmp4_m);                               \
+    _out2 = __lasx_xvilvh_w(_tmp6_m, _tmp4_m);                               \
+    _out4 = __lasx_xvilvl_w(_tmp7_m, _tmp5_m);                               \
+    _out6 = __lasx_xvilvh_w(_tmp7_m, _tmp5_m);                               \
+    _out1 = __lasx_xvbsrl_v(_out0, 8);                                       \
+    _out3 = __lasx_xvbsrl_v(_out2, 8);                                       \
+    _out5 = __lasx_xvbsrl_v(_out4, 8);                                       \
+    _out7 = __lasx_xvbsrl_v(_out6, 8);                                       \
+  }
+
+/*
+ * =============================================================================
+ * Description : Transpose 8x8 block with halfword elements in vectors.
+ * Arguments   : Inputs  - _in0, _in1, ~
+ *               Outputs - _out0, _out1, ~
+ * Details     : The rows of the matrix become columns, and the columns become
+ *               rows.
+ * Example     : LASX_TRANSPOSE8x8_H
+ *        _in0 : 1,2,3,4, 5,6,7,8, 1,2,3,4, 5,6,7,8
+ *        _in1 : 8,2,3,4, 5,6,7,8, 8,2,3,4, 5,6,7,8
+ *        _in2 : 8,2,3,4, 5,6,7,8, 8,2,3,4, 5,6,7,8
+ *        _in3 : 1,2,3,4, 5,6,7,8, 1,2,3,4, 5,6,7,8
+ *        _in4 : 9,2,3,4, 5,6,7,8, 9,2,3,4, 5,6,7,8
+ *        _in5 : 1,2,3,4, 5,6,7,8, 1,2,3,4, 5,6,7,8
+ *        _in6 : 1,2,3,4, 5,6,7,8, 1,2,3,4, 5,6,7,8
+ *        _in7 : 9,2,3,4, 5,6,7,8, 9,2,3,4, 5,6,7,8
+ *
+ *       _out0 : 1,8,8,1, 9,1,1,9, 1,8,8,1, 9,1,1,9
+ *       _out1 : 2,2,2,2, 2,2,2,2, 2,2,2,2, 2,2,2,2
+ *       _out2 : 3,3,3,3, 3,3,3,3, 3,3,3,3, 3,3,3,3
+ *       _out3 : 4,4,4,4, 4,4,4,4, 4,4,4,4, 4,4,4,4
+ *       _out4 : 5,5,5,5, 5,5,5,5, 5,5,5,5, 5,5,5,5
+ *       _out5 : 6,6,6,6, 6,6,6,6, 6,6,6,6, 6,6,6,6
+ *       _out6 : 7,7,7,7, 7,7,7,7, 7,7,7,7, 7,7,7,7
+ *       _out7 : 8,8,8,8, 8,8,8,8, 8,8,8,8, 8,8,8,8
+ * =============================================================================
+ */
+#define LASX_TRANSPOSE8x8_H(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                            _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                            _out7)                                           \
+  {                                                                          \
+    __m256i _s0_m, _s1_m;                                                    \
+    __m256i _tmp0_m, _tmp1_m, _tmp2_m, _tmp3_m;                              \
+    __m256i _tmp4_m, _tmp5_m, _tmp6_m, _tmp7_m;                              \
+                                                                             \
+    _s0_m = __lasx_xvilvl_h(_in6, _in4);                                     \
+    _s1_m = __lasx_xvilvl_h(_in7, _in5);                                     \
+    _tmp0_m = __lasx_xvilvl_h(_s1_m, _s0_m);                                 \
+    _tmp1_m = __lasx_xvilvh_h(_s1_m, _s0_m);                                 \
+    _s0_m = __lasx_xvilvh_h(_in6, _in4);                                     \
+    _s1_m = __lasx_xvilvh_h(_in7, _in5);                                     \
+    _tmp2_m = __lasx_xvilvl_h(_s1_m, _s0_m);                                 \
+    _tmp3_m = __lasx_xvilvh_h(_s1_m, _s0_m);                                 \
+                                                                             \
+    _s0_m = __lasx_xvilvl_h(_in2, _in0);                                     \
+    _s1_m = __lasx_xvilvl_h(_in3, _in1);                                     \
+    _tmp4_m = __lasx_xvilvl_h(_s1_m, _s0_m);                                 \
+    _tmp5_m = __lasx_xvilvh_h(_s1_m, _s0_m);                                 \
+    _s0_m = __lasx_xvilvh_h(_in2, _in0);                                     \
+    _s1_m = __lasx_xvilvh_h(_in3, _in1);                                     \
+    _tmp6_m = __lasx_xvilvl_h(_s1_m, _s0_m);                                 \
+    _tmp7_m = __lasx_xvilvh_h(_s1_m, _s0_m);                                 \
+                                                                             \
+    _out0 = __lasx_xvpickev_d(_tmp0_m, _tmp4_m);                             \
+    _out2 = __lasx_xvpickev_d(_tmp1_m, _tmp5_m);                             \
+    _out4 = __lasx_xvpickev_d(_tmp2_m, _tmp6_m);                             \
+    _out6 = __lasx_xvpickev_d(_tmp3_m, _tmp7_m);                             \
+    _out1 = __lasx_xvpickod_d(_tmp0_m, _tmp4_m);                             \
+    _out3 = __lasx_xvpickod_d(_tmp1_m, _tmp5_m);                             \
+    _out5 = __lasx_xvpickod_d(_tmp2_m, _tmp6_m);                             \
+    _out7 = __lasx_xvpickod_d(_tmp3_m, _tmp7_m);                             \
+  }
+
+/*
+ * =============================================================================
+ * Description : Butterfly of 4 input vectors
+ * Arguments   : Inputs  - _in0, _in1, _in2, _in3
+ *               Outputs - _out0, _out1, _out2, _out3
+ * Details     : Butterfly operation
+ * Example     : LASX_BUTTERFLY_4
+ *               _out0 = _in0 + _in3;
+ *               _out1 = _in1 + _in2;
+ *               _out2 = _in1 - _in2;
+ *               _out3 = _in0 - _in3;
+ * =============================================================================
+ */
+#define LASX_BUTTERFLY_4_B(_in0, _in1, _in2, _in3, _out0, _out1, _out2, _out3) \
+  {                                                                            \
+    _out0 = __lasx_xvadd_b(_in0, _in3);                                        \
+    _out1 = __lasx_xvadd_b(_in1, _in2);                                        \
+    _out2 = __lasx_xvsub_b(_in1, _in2);                                        \
+    _out3 = __lasx_xvsub_b(_in0, _in3);                                        \
+  }
+#define LASX_BUTTERFLY_4_H(_in0, _in1, _in2, _in3, _out0, _out1, _out2, _out3) \
+  {                                                                            \
+    _out0 = __lasx_xvadd_h(_in0, _in3);                                        \
+    _out1 = __lasx_xvadd_h(_in1, _in2);                                        \
+    _out2 = __lasx_xvsub_h(_in1, _in2);                                        \
+    _out3 = __lasx_xvsub_h(_in0, _in3);                                        \
+  }
+#define LASX_BUTTERFLY_4_W(_in0, _in1, _in2, _in3, _out0, _out1, _out2, _out3) \
+  {                                                                            \
+    _out0 = __lasx_xvadd_w(_in0, _in3);                                        \
+    _out1 = __lasx_xvadd_w(_in1, _in2);                                        \
+    _out2 = __lasx_xvsub_w(_in1, _in2);                                        \
+    _out3 = __lasx_xvsub_w(_in0, _in3);                                        \
+  }
+#define LASX_BUTTERFLY_4_D(_in0, _in1, _in2, _in3, _out0, _out1, _out2, _out3) \
+  {                                                                            \
+    _out0 = __lasx_xvadd_d(_in0, _in3);                                        \
+    _out1 = __lasx_xvadd_d(_in1, _in2);                                        \
+    _out2 = __lasx_xvsub_d(_in1, _in2);                                        \
+    _out3 = __lasx_xvsub_d(_in0, _in3);                                        \
+  }
+
+/*
+ * =============================================================================
+ * Description : Butterfly of 8 input vectors
+ * Arguments   : Inputs  - _in0, _in1, _in2, _in3, ~
+ *               Outputs - _out0, _out1, _out2, _out3, ~
+ * Details     : Butterfly operation
+ * Example     : LASX_BUTTERFLY_8
+ *               _out0 = _in0 + _in7;
+ *               _out1 = _in1 + _in6;
+ *               _out2 = _in2 + _in5;
+ *               _out3 = _in3 + _in4;
+ *               _out4 = _in3 - _in4;
+ *               _out5 = _in2 - _in5;
+ *               _out6 = _in1 - _in6;
+ *               _out7 = _in0 - _in7;
+ * =============================================================================
+ */
+#define LASX_BUTTERFLY_8_B(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                           _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                           _out7)                                           \
+  {                                                                         \
+    _out0 = __lasx_xvadd_b(_in0, _in7);                                     \
+    _out1 = __lasx_xvadd_b(_in1, _in6);                                     \
+    _out2 = __lasx_xvadd_b(_in2, _in5);                                     \
+    _out3 = __lasx_xvadd_b(_in3, _in4);                                     \
+    _out4 = __lasx_xvsub_b(_in3, _in4);                                     \
+    _out5 = __lasx_xvsub_b(_in2, _in5);                                     \
+    _out6 = __lasx_xvsub_b(_in1, _in6);                                     \
+    _out7 = __lasx_xvsub_b(_in0, _in7);                                     \
+  }
+
+#define LASX_BUTTERFLY_8_H(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                           _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                           _out7)                                           \
+  {                                                                         \
+    _out0 = __lasx_xvadd_h(_in0, _in7);                                     \
+    _out1 = __lasx_xvadd_h(_in1, _in6);                                     \
+    _out2 = __lasx_xvadd_h(_in2, _in5);                                     \
+    _out3 = __lasx_xvadd_h(_in3, _in4);                                     \
+    _out4 = __lasx_xvsub_h(_in3, _in4);                                     \
+    _out5 = __lasx_xvsub_h(_in2, _in5);                                     \
+    _out6 = __lasx_xvsub_h(_in1, _in6);                                     \
+    _out7 = __lasx_xvsub_h(_in0, _in7);                                     \
+  }
+
+#define LASX_BUTTERFLY_8_W(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                           _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                           _out7)                                           \
+  {                                                                         \
+    _out0 = __lasx_xvadd_w(_in0, _in7);                                     \
+    _out1 = __lasx_xvadd_w(_in1, _in6);                                     \
+    _out2 = __lasx_xvadd_w(_in2, _in5);                                     \
+    _out3 = __lasx_xvadd_w(_in3, _in4);                                     \
+    _out4 = __lasx_xvsub_w(_in3, _in4);                                     \
+    _out5 = __lasx_xvsub_w(_in2, _in5);                                     \
+    _out6 = __lasx_xvsub_w(_in1, _in6);                                     \
+    _out7 = __lasx_xvsub_w(_in0, _in7);                                     \
+  }
+
+#define LASX_BUTTERFLY_8_D(_in0, _in1, _in2, _in3, _in4, _in5, _in6, _in7,  \
+                           _out0, _out1, _out2, _out3, _out4, _out5, _out6, \
+                           _out7)                                           \
+  {                                                                         \
+    _out0 = __lasx_xvadd_d(_in0, _in7);                                     \
+    _out1 = __lasx_xvadd_d(_in1, _in6);                                     \
+    _out2 = __lasx_xvadd_d(_in2, _in5);                                     \
+    _out3 = __lasx_xvadd_d(_in3, _in4);                                     \
+    _out4 = __lasx_xvsub_d(_in3, _in4);                                     \
+    _out5 = __lasx_xvsub_d(_in2, _in5);                                     \
+    _out6 = __lasx_xvsub_d(_in1, _in6);                                     \
+    _out7 = __lasx_xvsub_d(_in0, _in7);                                     \
+  }
+
+#endif  // LASX
+
+/*
+ * =============================================================================
+ * Description : Print out elements in vector.
+ * Arguments   : Inputs  - RTYPE, _element_num, _in0, _enter
+ *               Outputs -
+ * Details     : Print out '_element_num' elements in 'RTYPE' vector '_in0', if
+ *               '_enter' is TRUE, prefix "\nVP:" will be added first.
+ * Example     : VECT_PRINT(v4i32,4,in0,1); // in0: 1,2,3,4
+ *               VP:1,2,3,4,
+ * =============================================================================
+ */
+#define VECT_PRINT(RTYPE, element_num, in0, enter) \
+  {                                                \
+    RTYPE _tmp0 = (RTYPE)in0;                      \
+    int _i = 0;                                    \
+    if (enter)                                     \
+      printf("\nVP:");                             \
+    for (_i = 0; _i < element_num; _i++)           \
+      printf("%d,", _tmp0[_i]);                    \
+  }
+
+#endif /* LOONGSON_INTRINSICS_H */
+#endif /* INCLUDE_LIBYUV_LOONGSON_INTRINSICS_H */

yuv.mod/libyuv/include/libyuv/macros_msa.h

@@ -0,0 +1,244 @@
+/*
+ *  Copyright 2016 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_MACROS_MSA_H_
+#define INCLUDE_LIBYUV_MACROS_MSA_H_
+
+#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
+#include <msa.h>
+#include <stdint.h>
+
+#if (__mips_isa_rev >= 6)
+#define LW(psrc)                                       \
+  ({                                                   \
+    const uint8_t* psrc_lw_m = (const uint8_t*)(psrc); \
+    uint32_t val_m;                                    \
+    asm("lw  %[val_m],  %[psrc_lw_m]  \n"              \
+        : [val_m] "=r"(val_m)                          \
+        : [psrc_lw_m] "m"(*psrc_lw_m));                \
+    val_m;                                             \
+  })
+
+#if (__mips == 64)
+#define LD(psrc)                                       \
+  ({                                                   \
+    const uint8_t* psrc_ld_m = (const uint8_t*)(psrc); \
+    uint64_t val_m = 0;                                \
+    asm("ld  %[val_m],  %[psrc_ld_m]  \n"              \
+        : [val_m] "=r"(val_m)                          \
+        : [psrc_ld_m] "m"(*psrc_ld_m));                \
+    val_m;                                             \
+  })
+#else  // !(__mips == 64)
+#define LD(psrc)                                                         \
+  ({                                                                     \
+    const uint8_t* psrc_ld_m = (const uint8_t*)(psrc);                   \
+    uint32_t val0_m, val1_m;                                             \
+    uint64_t val_m = 0;                                                  \
+    val0_m = LW(psrc_ld_m);                                              \
+    val1_m = LW(psrc_ld_m + 4);                                          \
+    val_m = (uint64_t)(val1_m);                             /* NOLINT */ \
+    val_m = (uint64_t)((val_m << 32) & 0xFFFFFFFF00000000); /* NOLINT */ \
+    val_m = (uint64_t)(val_m | (uint64_t)val0_m);           /* NOLINT */ \
+    val_m;                                                               \
+  })
+#endif  // (__mips == 64)
+
+#define SW(val, pdst)                                   \
+  ({                                                    \
+    uint8_t* pdst_sw_m = (uint8_t*)(pdst); /* NOLINT */ \
+    uint32_t val_m = (val);                             \
+    asm("sw  %[val_m],  %[pdst_sw_m]  \n"               \
+        : [pdst_sw_m] "=m"(*pdst_sw_m)                  \
+        : [val_m] "r"(val_m));                          \
+  })
+
+#if (__mips == 64)
+#define SD(val, pdst)                                   \
+  ({                                                    \
+    uint8_t* pdst_sd_m = (uint8_t*)(pdst); /* NOLINT */ \
+    uint64_t val_m = (val);                             \
+    asm("sd  %[val_m],  %[pdst_sd_m]  \n"               \
+        : [pdst_sd_m] "=m"(*pdst_sd_m)                  \
+        : [val_m] "r"(val_m));                          \
+  })
+#else  // !(__mips == 64)
+#define SD(val, pdst)                                        \
+  ({                                                         \
+    uint8_t* pdst_sd_m = (uint8_t*)(pdst); /* NOLINT */      \
+    uint32_t val0_m, val1_m;                                 \
+    val0_m = (uint32_t)((val)&0x00000000FFFFFFFF);           \
+    val1_m = (uint32_t)(((val) >> 32) & 0x00000000FFFFFFFF); \
+    SW(val0_m, pdst_sd_m);                                   \
+    SW(val1_m, pdst_sd_m + 4);                               \
+  })
+#endif  // !(__mips == 64)
+#else   // !(__mips_isa_rev >= 6)
+#define LW(psrc)                                \
+  ({                                            \
+    uint8_t* psrc_lw_m = (uint8_t*)(psrc);      \
+    uint32_t val_lw_m;                          \
+                                                \
+    asm("lwr %[val_lw_m], 0(%[psrc_lw_m]) \n\t" \
+        "lwl %[val_lw_m], 3(%[psrc_lw_m]) \n\t" \
+                                                \
+        : [val_lw_m] "=&r"(val_lw_m)            \
+        : [psrc_lw_m] "r"(psrc_lw_m));          \
+                                                \
+    val_lw_m;                                   \
+  })
+
+#if (__mips == 64)
+#define LD(psrc)                                \
+  ({                                            \
+    uint8_t* psrc_ld_m = (uint8_t*)(psrc);      \
+    uint64_t val_ld_m = 0;                      \
+                                                \
+    asm("ldr %[val_ld_m], 0(%[psrc_ld_m]) \n\t" \
+        "ldl %[val_ld_m], 7(%[psrc_ld_m]) \n\t" \
+                                                \
+        : [val_ld_m] "=&r"(val_ld_m)            \
+        : [psrc_ld_m] "r"(psrc_ld_m));          \
+                                                \
+    val_ld_m;                                   \
+  })
+#else  // !(__mips == 64)
+#define LD(psrc)                                                         \
+  ({                                                                     \
+    const uint8_t* psrc_ld_m = (const uint8_t*)(psrc);                   \
+    uint32_t val0_m, val1_m;                                             \
+    uint64_t val_m = 0;                                                  \
+    val0_m = LW(psrc_ld_m);                                              \
+    val1_m = LW(psrc_ld_m + 4);                                          \
+    val_m = (uint64_t)(val1_m);                             /* NOLINT */ \
+    val_m = (uint64_t)((val_m << 32) & 0xFFFFFFFF00000000); /* NOLINT */ \
+    val_m = (uint64_t)(val_m | (uint64_t)val0_m);           /* NOLINT */ \
+    val_m;                                                               \
+  })
+#endif  // (__mips == 64)
+
+#define SW(val, pdst)                                   \
+  ({                                                    \
+    uint8_t* pdst_sw_m = (uint8_t*)(pdst); /* NOLINT */ \
+    uint32_t val_m = (val);                             \
+    asm("usw  %[val_m],  %[pdst_sw_m]  \n"              \
+        : [pdst_sw_m] "=m"(*pdst_sw_m)                  \
+        : [val_m] "r"(val_m));                          \
+  })
+
+#define SD(val, pdst)                                        \
+  ({                                                         \
+    uint8_t* pdst_sd_m = (uint8_t*)(pdst); /* NOLINT */      \
+    uint32_t val0_m, val1_m;                                 \
+    val0_m = (uint32_t)((val)&0x00000000FFFFFFFF);           \
+    val1_m = (uint32_t)(((val) >> 32) & 0x00000000FFFFFFFF); \
+    SW(val0_m, pdst_sd_m);                                   \
+    SW(val1_m, pdst_sd_m + 4);                               \
+  })
+#endif  // (__mips_isa_rev >= 6)
+
+// TODO(fbarchard): Consider removing __VAR_ARGS versions.
+#define LD_B(RTYPE, psrc) *((RTYPE*)(psrc)) /* NOLINT */
+#define LD_UB(...) LD_B(const v16u8, __VA_ARGS__)
+
+#define LD_H(RTYPE, psrc) *((RTYPE*)(psrc)) /* NOLINT */
+#define LD_UH(...) LD_H(const v8u16, __VA_ARGS__)
+
+#define ST_B(RTYPE, in, pdst) *((RTYPE*)(pdst)) = (in) /* NOLINT */
+#define ST_UB(...) ST_B(v16u8, __VA_ARGS__)
+
+#define ST_H(RTYPE, in, pdst) *((RTYPE*)(pdst)) = (in) /* NOLINT */
+#define ST_UH(...) ST_H(v8u16, __VA_ARGS__)
+
+/* Description : Load two vectors with 16 'byte' sized elements
+   Arguments   : Inputs  - psrc, stride
+                 Outputs - out0, out1
+                 Return Type - as per RTYPE
+   Details     : Load 16 byte elements in 'out0' from (psrc)
+                 Load 16 byte elements in 'out1' from (psrc + stride)
+*/
+#define LD_B2(RTYPE, psrc, stride, out0, out1) \
+  {                                            \
+    out0 = LD_B(RTYPE, (psrc));                \
+    out1 = LD_B(RTYPE, (psrc) + stride);       \
+  }
+#define LD_UB2(...) LD_B2(const v16u8, __VA_ARGS__)
+
+#define LD_B4(RTYPE, psrc, stride, out0, out1, out2, out3) \
+  {                                                        \
+    LD_B2(RTYPE, (psrc), stride, out0, out1);              \
+    LD_B2(RTYPE, (psrc) + 2 * stride, stride, out2, out3); \
+  }
+#define LD_UB4(...) LD_B4(const v16u8, __VA_ARGS__)
+
+/* Description : Store two vectors with stride each having 16 'byte' sized
+                 elements
+   Arguments   : Inputs - in0, in1, pdst, stride
+   Details     : Store 16 byte elements from 'in0' to (pdst)
+                 Store 16 byte elements from 'in1' to (pdst + stride)
+*/
+#define ST_B2(RTYPE, in0, in1, pdst, stride) \
+  {                                          \
+    ST_B(RTYPE, in0, (pdst));                \
+    ST_B(RTYPE, in1, (pdst) + stride);       \
+  }
+#define ST_UB2(...) ST_B2(v16u8, __VA_ARGS__)
+
+#define ST_B4(RTYPE, in0, in1, in2, in3, pdst, stride)   \
+  {                                                      \
+    ST_B2(RTYPE, in0, in1, (pdst), stride);              \
+    ST_B2(RTYPE, in2, in3, (pdst) + 2 * stride, stride); \
+  }
+#define ST_UB4(...) ST_B4(v16u8, __VA_ARGS__)
+
+/* Description : Store vectors of 8 halfword elements with stride
+   Arguments   : Inputs - in0, in1, pdst, stride
+   Details     : Store 8 halfword elements from 'in0' to (pdst)
+                 Store 8 halfword elements from 'in1' to (pdst + stride)
+*/
+#define ST_H2(RTYPE, in0, in1, pdst, stride) \
+  {                                          \
+    ST_H(RTYPE, in0, (pdst));                \
+    ST_H(RTYPE, in1, (pdst) + stride);       \
+  }
+#define ST_UH2(...) ST_H2(v8u16, __VA_ARGS__)
+
+// TODO(fbarchard): Consider using __msa_vshf_b and __msa_ilvr_b directly.
+/* Description : Shuffle byte vector elements as per mask vector
+   Arguments   : Inputs  - in0, in1, in2, in3, mask0, mask1
+                 Outputs - out0, out1
+                 Return Type - as per RTYPE
+   Details     : Byte elements from 'in0' & 'in1' are copied selectively to
+                 'out0' as per control vector 'mask0'
+*/
+#define VSHF_B2(RTYPE, in0, in1, in2, in3, mask0, mask1, out0, out1)  \
+  {                                                                   \
+    out0 = (RTYPE)__msa_vshf_b((v16i8)mask0, (v16i8)in1, (v16i8)in0); \
+    out1 = (RTYPE)__msa_vshf_b((v16i8)mask1, (v16i8)in3, (v16i8)in2); \
+  }
+#define VSHF_B2_UB(...) VSHF_B2(v16u8, __VA_ARGS__)
+
+/* Description : Interleave both left and right half of input vectors
+   Arguments   : Inputs  - in0, in1
+                 Outputs - out0, out1
+                 Return Type - as per RTYPE
+   Details     : Right half of byte elements from 'in0' and 'in1' are
+                 interleaved and written to 'out0'
+*/
+#define ILVRL_B2(RTYPE, in0, in1, out0, out1)           \
+  {                                                     \
+    out0 = (RTYPE)__msa_ilvr_b((v16i8)in0, (v16i8)in1); \
+    out1 = (RTYPE)__msa_ilvl_b((v16i8)in0, (v16i8)in1); \
+  }
+#define ILVRL_B2_UB(...) ILVRL_B2(v16u8, __VA_ARGS__)
+
+#endif /* !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa) */
+
+#endif  // INCLUDE_LIBYUV_MACROS_MSA_H_

yuv.mod/libyuv/include/libyuv/mjpeg_decoder.h

@@ -0,0 +1,195 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_MJPEG_DECODER_H_
+#define INCLUDE_LIBYUV_MJPEG_DECODER_H_
+
+#include "libyuv/basic_types.h"
+
+#ifdef __cplusplus
+// NOTE: For a simplified public API use convert.h MJPGToI420().
+
+struct jpeg_common_struct;
+struct jpeg_decompress_struct;
+struct jpeg_source_mgr;
+
+namespace libyuv {
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+LIBYUV_BOOL ValidateJpeg(const uint8_t* sample, size_t sample_size);
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+
+static const uint32_t kUnknownDataSize = 0xFFFFFFFF;
+
+enum JpegSubsamplingType {
+  kJpegYuv420,
+  kJpegYuv422,
+  kJpegYuv444,
+  kJpegYuv400,
+  kJpegUnknown
+};
+
+struct Buffer {
+  const uint8_t* data;
+  int len;
+};
+
+struct BufferVector {
+  Buffer* buffers;
+  int len;
+  int pos;
+};
+
+struct SetJmpErrorMgr;
+
+// MJPEG ("Motion JPEG") is a pseudo-standard video codec where the frames are
+// simply independent JPEG images with a fixed huffman table (which is omitted).
+// It is rarely used in video transmission, but is common as a camera capture
+// format, especially in Logitech devices. This class implements a decoder for
+// MJPEG frames.
+//
+// See http://tools.ietf.org/html/rfc2435
+class LIBYUV_API MJpegDecoder {
+ public:
+  typedef void (*CallbackFunction)(void* opaque,
+                                   const uint8_t* const* data,
+                                   const int* strides,
+                                   int rows);
+
+  static const int kColorSpaceUnknown;
+  static const int kColorSpaceGrayscale;
+  static const int kColorSpaceRgb;
+  static const int kColorSpaceYCbCr;
+  static const int kColorSpaceCMYK;
+  static const int kColorSpaceYCCK;
+
+  MJpegDecoder();
+  ~MJpegDecoder();
+
+  // Loads a new frame, reads its headers, and determines the uncompressed
+  // image format.
+  // Returns LIBYUV_TRUE if image looks valid and format is supported.
+  // If return value is LIBYUV_TRUE, then the values for all the following
+  // getters are populated.
+  // src_len is the size of the compressed mjpeg frame in bytes.
+  LIBYUV_BOOL LoadFrame(const uint8_t* src, size_t src_len);
+
+  // Returns width of the last loaded frame in pixels.
+  int GetWidth();
+
+  // Returns height of the last loaded frame in pixels.
+  int GetHeight();
+
+  // Returns format of the last loaded frame. The return value is one of the
+  // kColorSpace* constants.
+  int GetColorSpace();
+
+  // Number of color components in the color space.
+  int GetNumComponents();
+
+  // Sample factors of the n-th component.
+  int GetHorizSampFactor(int component);
+
+  int GetVertSampFactor(int component);
+
+  int GetHorizSubSampFactor(int component);
+
+  int GetVertSubSampFactor(int component);
+
+  // Public for testability.
+  int GetImageScanlinesPerImcuRow();
+
+  // Public for testability.
+  int GetComponentScanlinesPerImcuRow(int component);
+
+  // Width of a component in bytes.
+  int GetComponentWidth(int component);
+
+  // Height of a component.
+  int GetComponentHeight(int component);
+
+  // Width of a component in bytes with padding for DCTSIZE. Public for testing.
+  int GetComponentStride(int component);
+
+  // Size of a component in bytes.
+  int GetComponentSize(int component);
+
+  // Call this after LoadFrame() if you decide you don't want to decode it
+  // after all.
+  LIBYUV_BOOL UnloadFrame();
+
+  // Decodes the entire image into a one-buffer-per-color-component format.
+  // dst_width must match exactly. dst_height must be <= to image height; if
+  // less, the image is cropped. "planes" must have size equal to at least
+  // GetNumComponents() and they must point to non-overlapping buffers of size
+  // at least GetComponentSize(i). The pointers in planes are incremented
+  // to point to after the end of the written data.
+  // TODO(fbarchard): Add dst_x, dst_y to allow specific rect to be decoded.
+  LIBYUV_BOOL DecodeToBuffers(uint8_t** planes, int dst_width, int dst_height);
+
+  // Decodes the entire image and passes the data via repeated calls to a
+  // callback function. Each call will get the data for a whole number of
+  // image scanlines.
+  // TODO(fbarchard): Add dst_x, dst_y to allow specific rect to be decoded.
+  LIBYUV_BOOL DecodeToCallback(CallbackFunction fn,
+                               void* opaque,
+                               int dst_width,
+                               int dst_height);
+
+  // The helper function which recognizes the jpeg sub-sampling type.
+  static JpegSubsamplingType JpegSubsamplingTypeHelper(
+      int* subsample_x,
+      int* subsample_y,
+      int number_of_components);
+
+ private:
+  void AllocOutputBuffers(int num_outbufs);
+  void DestroyOutputBuffers();
+
+  LIBYUV_BOOL StartDecode();
+  LIBYUV_BOOL FinishDecode();
+
+  void SetScanlinePointers(uint8_t** data);
+  LIBYUV_BOOL DecodeImcuRow();
+
+  int GetComponentScanlinePadding(int component);
+
+  // A buffer holding the input data for a frame.
+  Buffer buf_;
+  BufferVector buf_vec_;
+
+  jpeg_decompress_struct* decompress_struct_;
+  jpeg_source_mgr* source_mgr_;
+  SetJmpErrorMgr* error_mgr_;
+
+  // LIBYUV_TRUE iff at least one component has scanline padding. (i.e.,
+  // GetComponentScanlinePadding() != 0.)
+  LIBYUV_BOOL has_scanline_padding_;
+
+  // Temporaries used to point to scanline outputs.
+  int num_outbufs_;  // Outermost size of all arrays below.
+  uint8_t*** scanlines_;
+  int* scanlines_sizes_;
+  // Temporary buffer used for decoding when we can't decode directly to the
+  // output buffers. Large enough for just one iMCU row.
+  uint8_t** databuf_;
+  int* databuf_strides_;
+};
+
+}  // namespace libyuv
+
+#endif  //  __cplusplus
+#endif  // INCLUDE_LIBYUV_MJPEG_DECODER_H_

yuv.mod/libyuv/include/libyuv/planar_functions.h

@@ -0,0 +1,1158 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_PLANAR_FUNCTIONS_H_
+#define INCLUDE_LIBYUV_PLANAR_FUNCTIONS_H_
+
+#include "libyuv/basic_types.h"
+
+// TODO(fbarchard): Remove the following headers includes.
+#include "libyuv/convert.h"
+#include "libyuv/convert_argb.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// TODO(fbarchard): Move cpu macros to row.h
+#if defined(__pnacl__) || defined(__CLR_VER) ||            \
+    (defined(__native_client__) && defined(__x86_64__)) || \
+    (defined(__i386__) && !defined(__SSE__) && !defined(__clang__))
+#define LIBYUV_DISABLE_X86
+#endif
+// MemorySanitizer does not support assembly code yet. http://crbug.com/344505
+#if defined(__has_feature)
+#if __has_feature(memory_sanitizer) && !defined(LIBYUV_DISABLE_NEON)
+#define LIBYUV_DISABLE_NEON
+#endif
+#if __has_feature(memory_sanitizer) && !defined(LIBYUV_DISABLE_X86)
+#define LIBYUV_DISABLE_X86
+#endif
+#endif
+// The following are available on all x86 platforms:
+#if !defined(LIBYUV_DISABLE_X86) && \
+    (defined(_M_IX86) || defined(__x86_64__) || defined(__i386__))
+#define HAS_ARGBAFFINEROW_SSE2
+#endif
+
+// Copy a plane of data.
+LIBYUV_API
+void CopyPlane(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               int width,
+               int height);
+
+LIBYUV_API
+void CopyPlane_16(const uint16_t* src_y,
+                  int src_stride_y,
+                  uint16_t* dst_y,
+                  int dst_stride_y,
+                  int width,
+                  int height);
+
+LIBYUV_API
+void Convert16To8Plane(const uint16_t* src_y,
+                       int src_stride_y,
+                       uint8_t* dst_y,
+                       int dst_stride_y,
+                       int scale,  // 16384 for 10 bits
+                       int width,
+                       int height);
+
+LIBYUV_API
+void Convert8To16Plane(const uint8_t* src_y,
+                       int src_stride_y,
+                       uint16_t* dst_y,
+                       int dst_stride_y,
+                       int scale,  // 1024 for 10 bits
+                       int width,
+                       int height);
+
+// Set a plane of data to a 32 bit value.
+LIBYUV_API
+void SetPlane(uint8_t* dst_y,
+              int dst_stride_y,
+              int width,
+              int height,
+              uint32_t value);
+
+// Convert a plane of tiles of 16 x H to linear.
+LIBYUV_API
+int DetilePlane(const uint8_t* src_y,
+                int src_stride_y,
+                uint8_t* dst_y,
+                int dst_stride_y,
+                int width,
+                int height,
+                int tile_height);
+
+// Convert a plane of 16 bit tiles of 16 x H to linear.
+LIBYUV_API
+int DetilePlane_16(const uint16_t* src_y,
+                   int src_stride_y,
+                   uint16_t* dst_y,
+                   int dst_stride_y,
+                   int width,
+                   int height,
+                   int tile_height);
+
+// Convert a UV plane of tiles of 16 x H into linear U and V planes.
+LIBYUV_API
+void DetileSplitUVPlane(const uint8_t* src_uv,
+                        int src_stride_uv,
+                        uint8_t* dst_u,
+                        int dst_stride_u,
+                        uint8_t* dst_v,
+                        int dst_stride_v,
+                        int width,
+                        int height,
+                        int tile_height);
+
+// Convert a Y and UV plane of tiles into interlaced YUY2.
+LIBYUV_API
+void DetileToYUY2(const uint8_t* src_y,
+                  int src_stride_y,
+                  const uint8_t* src_uv,
+                  int src_stride_uv,
+                  uint8_t* dst_yuy2,
+                  int dst_stride_yuy2,
+                  int width,
+                  int height,
+                  int tile_height);
+
+// Split interleaved UV plane into separate U and V planes.
+LIBYUV_API
+void SplitUVPlane(const uint8_t* src_uv,
+                  int src_stride_uv,
+                  uint8_t* dst_u,
+                  int dst_stride_u,
+                  uint8_t* dst_v,
+                  int dst_stride_v,
+                  int width,
+                  int height);
+
+// Merge separate U and V planes into one interleaved UV plane.
+LIBYUV_API
+void MergeUVPlane(const uint8_t* src_u,
+                  int src_stride_u,
+                  const uint8_t* src_v,
+                  int src_stride_v,
+                  uint8_t* dst_uv,
+                  int dst_stride_uv,
+                  int width,
+                  int height);
+
+// Split interleaved msb UV plane into separate lsb U and V planes.
+LIBYUV_API
+void SplitUVPlane_16(const uint16_t* src_uv,
+                     int src_stride_uv,
+                     uint16_t* dst_u,
+                     int dst_stride_u,
+                     uint16_t* dst_v,
+                     int dst_stride_v,
+                     int width,
+                     int height,
+                     int depth);
+
+// Merge separate lsb U and V planes into one interleaved msb UV plane.
+LIBYUV_API
+void MergeUVPlane_16(const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint16_t* dst_uv,
+                     int dst_stride_uv,
+                     int width,
+                     int height,
+                     int depth);
+
+// Convert lsb plane to msb plane
+LIBYUV_API
+void ConvertToMSBPlane_16(const uint16_t* src_y,
+                          int src_stride_y,
+                          uint16_t* dst_y,
+                          int dst_stride_y,
+                          int width,
+                          int height,
+                          int depth);
+
+// Convert msb plane to lsb plane
+LIBYUV_API
+void ConvertToLSBPlane_16(const uint16_t* src_y,
+                          int src_stride_y,
+                          uint16_t* dst_y,
+                          int dst_stride_y,
+                          int width,
+                          int height,
+                          int depth);
+
+// Scale U and V to half width and height and merge into interleaved UV plane.
+// width and height are source size, allowing odd sizes.
+// Use for converting I444 or I422 to NV12.
+LIBYUV_API
+void HalfMergeUVPlane(const uint8_t* src_u,
+                      int src_stride_u,
+                      const uint8_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_uv,
+                      int dst_stride_uv,
+                      int width,
+                      int height);
+
+// Swap U and V channels in interleaved UV plane.
+LIBYUV_API
+void SwapUVPlane(const uint8_t* src_uv,
+                 int src_stride_uv,
+                 uint8_t* dst_vu,
+                 int dst_stride_vu,
+                 int width,
+                 int height);
+
+// Split interleaved RGB plane into separate R, G and B planes.
+LIBYUV_API
+void SplitRGBPlane(const uint8_t* src_rgb,
+                   int src_stride_rgb,
+                   uint8_t* dst_r,
+                   int dst_stride_r,
+                   uint8_t* dst_g,
+                   int dst_stride_g,
+                   uint8_t* dst_b,
+                   int dst_stride_b,
+                   int width,
+                   int height);
+
+// Merge separate R, G and B planes into one interleaved RGB plane.
+LIBYUV_API
+void MergeRGBPlane(const uint8_t* src_r,
+                   int src_stride_r,
+                   const uint8_t* src_g,
+                   int src_stride_g,
+                   const uint8_t* src_b,
+                   int src_stride_b,
+                   uint8_t* dst_rgb,
+                   int dst_stride_rgb,
+                   int width,
+                   int height);
+
+// Split interleaved ARGB plane into separate R, G, B and A planes.
+// dst_a can be NULL to discard alpha plane.
+LIBYUV_API
+void SplitARGBPlane(const uint8_t* src_argb,
+                    int src_stride_argb,
+                    uint8_t* dst_r,
+                    int dst_stride_r,
+                    uint8_t* dst_g,
+                    int dst_stride_g,
+                    uint8_t* dst_b,
+                    int dst_stride_b,
+                    uint8_t* dst_a,
+                    int dst_stride_a,
+                    int width,
+                    int height);
+
+// Merge separate R, G, B and A planes into one interleaved ARGB plane.
+// src_a can be NULL to fill opaque value to alpha.
+LIBYUV_API
+void MergeARGBPlane(const uint8_t* src_r,
+                    int src_stride_r,
+                    const uint8_t* src_g,
+                    int src_stride_g,
+                    const uint8_t* src_b,
+                    int src_stride_b,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    int width,
+                    int height);
+
+// Merge separate 'depth' bit R, G and B planes stored in lsb
+// into one interleaved XR30 plane.
+// depth should in range [10, 16]
+LIBYUV_API
+void MergeXR30Plane(const uint16_t* src_r,
+                    int src_stride_r,
+                    const uint16_t* src_g,
+                    int src_stride_g,
+                    const uint16_t* src_b,
+                    int src_stride_b,
+                    uint8_t* dst_ar30,
+                    int dst_stride_ar30,
+                    int width,
+                    int height,
+                    int depth);
+
+// Merge separate 'depth' bit R, G, B and A planes stored in lsb
+// into one interleaved AR64 plane.
+// src_a can be NULL to fill opaque value to alpha.
+// depth should in range [1, 16]
+LIBYUV_API
+void MergeAR64Plane(const uint16_t* src_r,
+                    int src_stride_r,
+                    const uint16_t* src_g,
+                    int src_stride_g,
+                    const uint16_t* src_b,
+                    int src_stride_b,
+                    const uint16_t* src_a,
+                    int src_stride_a,
+                    uint16_t* dst_ar64,
+                    int dst_stride_ar64,
+                    int width,
+                    int height,
+                    int depth);
+
+// Merge separate 'depth' bit R, G, B and A planes stored in lsb
+// into one interleaved ARGB plane.
+// src_a can be NULL to fill opaque value to alpha.
+// depth should in range [8, 16]
+LIBYUV_API
+void MergeARGB16To8Plane(const uint16_t* src_r,
+                         int src_stride_r,
+                         const uint16_t* src_g,
+                         int src_stride_g,
+                         const uint16_t* src_b,
+                         int src_stride_b,
+                         const uint16_t* src_a,
+                         int src_stride_a,
+                         uint8_t* dst_argb,
+                         int dst_stride_argb,
+                         int width,
+                         int height,
+                         int depth);
+
+// Copy I400.  Supports inverting.
+LIBYUV_API
+int I400ToI400(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               int width,
+               int height);
+
+#define J400ToJ400 I400ToI400
+
+// Copy I422 to I422.
+#define I422ToI422 I422Copy
+LIBYUV_API
+int I422Copy(const uint8_t* src_y,
+             int src_stride_y,
+             const uint8_t* src_u,
+             int src_stride_u,
+             const uint8_t* src_v,
+             int src_stride_v,
+             uint8_t* dst_y,
+             int dst_stride_y,
+             uint8_t* dst_u,
+             int dst_stride_u,
+             uint8_t* dst_v,
+             int dst_stride_v,
+             int width,
+             int height);
+
+// Copy I444 to I444.
+#define I444ToI444 I444Copy
+LIBYUV_API
+int I444Copy(const uint8_t* src_y,
+             int src_stride_y,
+             const uint8_t* src_u,
+             int src_stride_u,
+             const uint8_t* src_v,
+             int src_stride_v,
+             uint8_t* dst_y,
+             int dst_stride_y,
+             uint8_t* dst_u,
+             int dst_stride_u,
+             uint8_t* dst_v,
+             int dst_stride_v,
+             int width,
+             int height);
+
+// Copy I210 to I210.
+#define I210ToI210 I210Copy
+LIBYUV_API
+int I210Copy(const uint16_t* src_y,
+             int src_stride_y,
+             const uint16_t* src_u,
+             int src_stride_u,
+             const uint16_t* src_v,
+             int src_stride_v,
+             uint16_t* dst_y,
+             int dst_stride_y,
+             uint16_t* dst_u,
+             int dst_stride_u,
+             uint16_t* dst_v,
+             int dst_stride_v,
+             int width,
+             int height);
+
+// Copy I410 to I410.
+#define I410ToI410 I410Copy
+LIBYUV_API
+int I410Copy(const uint16_t* src_y,
+             int src_stride_y,
+             const uint16_t* src_u,
+             int src_stride_u,
+             const uint16_t* src_v,
+             int src_stride_v,
+             uint16_t* dst_y,
+             int dst_stride_y,
+             uint16_t* dst_u,
+             int dst_stride_u,
+             uint16_t* dst_v,
+             int dst_stride_v,
+             int width,
+             int height);
+
+// Copy NV12. Supports inverting.
+LIBYUV_API
+int NV12Copy(const uint8_t* src_y,
+             int src_stride_y,
+             const uint8_t* src_uv,
+             int src_stride_uv,
+             uint8_t* dst_y,
+             int dst_stride_y,
+             uint8_t* dst_uv,
+             int dst_stride_uv,
+             int width,
+             int height);
+
+// Copy NV21. Supports inverting.
+LIBYUV_API
+int NV21Copy(const uint8_t* src_y,
+             int src_stride_y,
+             const uint8_t* src_vu,
+             int src_stride_vu,
+             uint8_t* dst_y,
+             int dst_stride_y,
+             uint8_t* dst_vu,
+             int dst_stride_vu,
+             int width,
+             int height);
+
+// Convert YUY2 to I422.
+LIBYUV_API
+int YUY2ToI422(const uint8_t* src_yuy2,
+               int src_stride_yuy2,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Convert UYVY to I422.
+LIBYUV_API
+int UYVYToI422(const uint8_t* src_uyvy,
+               int src_stride_uyvy,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+LIBYUV_API
+int YUY2ToNV12(const uint8_t* src_yuy2,
+               int src_stride_yuy2,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+LIBYUV_API
+int UYVYToNV12(const uint8_t* src_uyvy,
+               int src_stride_uyvy,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Convert NV21 to NV12.
+LIBYUV_API
+int NV21ToNV12(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_vu,
+               int src_stride_vu,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+LIBYUV_API
+int YUY2ToY(const uint8_t* src_yuy2,
+            int src_stride_yuy2,
+            uint8_t* dst_y,
+            int dst_stride_y,
+            int width,
+            int height);
+
+LIBYUV_API
+int UYVYToY(const uint8_t* src_uyvy,
+            int src_stride_uyvy,
+            uint8_t* dst_y,
+            int dst_stride_y,
+            int width,
+            int height);
+
+// Convert I420 to I400. (calls CopyPlane ignoring u/v).
+LIBYUV_API
+int I420ToI400(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               int width,
+               int height);
+
+// Alias
+#define J420ToJ400 I420ToI400
+#define I420ToI420Mirror I420Mirror
+
+// I420 mirror.
+LIBYUV_API
+int I420Mirror(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height);
+
+// Alias
+#define I400ToI400Mirror I400Mirror
+
+// I400 mirror.  A single plane is mirrored horizontally.
+// Pass negative height to achieve 180 degree rotation.
+LIBYUV_API
+int I400Mirror(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               int width,
+               int height);
+
+// Alias
+#define NV12ToNV12Mirror NV12Mirror
+
+// NV12 mirror.
+LIBYUV_API
+int NV12Mirror(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height);
+
+// Alias
+#define ARGBToARGBMirror ARGBMirror
+
+// ARGB mirror.
+LIBYUV_API
+int ARGBMirror(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Alias
+#define RGB24ToRGB24Mirror RGB24Mirror
+
+// RGB24 mirror.
+LIBYUV_API
+int RGB24Mirror(const uint8_t* src_rgb24,
+                int src_stride_rgb24,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height);
+
+// Mirror a plane of data.
+LIBYUV_API
+void MirrorPlane(const uint8_t* src_y,
+                 int src_stride_y,
+                 uint8_t* dst_y,
+                 int dst_stride_y,
+                 int width,
+                 int height);
+
+// Mirror a plane of UV data.
+LIBYUV_API
+void MirrorUVPlane(const uint8_t* src_uv,
+                   int src_stride_uv,
+                   uint8_t* dst_uv,
+                   int dst_stride_uv,
+                   int width,
+                   int height);
+
+// Alias
+#define RGB24ToRAW RAWToRGB24
+
+LIBYUV_API
+int RAWToRGB24(const uint8_t* src_raw,
+               int src_stride_raw,
+               uint8_t* dst_rgb24,
+               int dst_stride_rgb24,
+               int width,
+               int height);
+
+// Draw a rectangle into I420.
+LIBYUV_API
+int I420Rect(uint8_t* dst_y,
+             int dst_stride_y,
+             uint8_t* dst_u,
+             int dst_stride_u,
+             uint8_t* dst_v,
+             int dst_stride_v,
+             int x,
+             int y,
+             int width,
+             int height,
+             int value_y,
+             int value_u,
+             int value_v);
+
+// Draw a rectangle into ARGB.
+LIBYUV_API
+int ARGBRect(uint8_t* dst_argb,
+             int dst_stride_argb,
+             int dst_x,
+             int dst_y,
+             int width,
+             int height,
+             uint32_t value);
+
+// Convert ARGB to gray scale ARGB.
+LIBYUV_API
+int ARGBGrayTo(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height);
+
+// Make a rectangle of ARGB gray scale.
+LIBYUV_API
+int ARGBGray(uint8_t* dst_argb,
+             int dst_stride_argb,
+             int dst_x,
+             int dst_y,
+             int width,
+             int height);
+
+// Make a rectangle of ARGB Sepia tone.
+LIBYUV_API
+int ARGBSepia(uint8_t* dst_argb,
+              int dst_stride_argb,
+              int dst_x,
+              int dst_y,
+              int width,
+              int height);
+
+// Apply a matrix rotation to each ARGB pixel.
+// matrix_argb is 4 signed ARGB values. -128 to 127 representing -2 to 2.
+// The first 4 coefficients apply to B, G, R, A and produce B of the output.
+// The next 4 coefficients apply to B, G, R, A and produce G of the output.
+// The next 4 coefficients apply to B, G, R, A and produce R of the output.
+// The last 4 coefficients apply to B, G, R, A and produce A of the output.
+LIBYUV_API
+int ARGBColorMatrix(const uint8_t* src_argb,
+                    int src_stride_argb,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    const int8_t* matrix_argb,
+                    int width,
+                    int height);
+
+// Deprecated. Use ARGBColorMatrix instead.
+// Apply a matrix rotation to each ARGB pixel.
+// matrix_argb is 3 signed ARGB values. -128 to 127 representing -1 to 1.
+// The first 4 coefficients apply to B, G, R, A and produce B of the output.
+// The next 4 coefficients apply to B, G, R, A and produce G of the output.
+// The last 4 coefficients apply to B, G, R, A and produce R of the output.
+LIBYUV_API
+int RGBColorMatrix(uint8_t* dst_argb,
+                   int dst_stride_argb,
+                   const int8_t* matrix_rgb,
+                   int dst_x,
+                   int dst_y,
+                   int width,
+                   int height);
+
+// Apply a color table each ARGB pixel.
+// Table contains 256 ARGB values.
+LIBYUV_API
+int ARGBColorTable(uint8_t* dst_argb,
+                   int dst_stride_argb,
+                   const uint8_t* table_argb,
+                   int dst_x,
+                   int dst_y,
+                   int width,
+                   int height);
+
+// Apply a color table each ARGB pixel but preserve destination alpha.
+// Table contains 256 ARGB values.
+LIBYUV_API
+int RGBColorTable(uint8_t* dst_argb,
+                  int dst_stride_argb,
+                  const uint8_t* table_argb,
+                  int dst_x,
+                  int dst_y,
+                  int width,
+                  int height);
+
+// Apply a luma/color table each ARGB pixel but preserve destination alpha.
+// Table contains 32768 values indexed by [Y][C] where 7 it 7 bit luma from
+// RGB (YJ style) and C is an 8 bit color component (R, G or B).
+LIBYUV_API
+int ARGBLumaColorTable(const uint8_t* src_argb,
+                       int src_stride_argb,
+                       uint8_t* dst_argb,
+                       int dst_stride_argb,
+                       const uint8_t* luma,
+                       int width,
+                       int height);
+
+// Apply a 3 term polynomial to ARGB values.
+// poly points to a 4x4 matrix.  The first row is constants.  The 2nd row is
+// coefficients for b, g, r and a.  The 3rd row is coefficients for b squared,
+// g squared, r squared and a squared.  The 4rd row is coefficients for b to
+// the 3, g to the 3, r to the 3 and a to the 3.  The values are summed and
+// result clamped to 0 to 255.
+// A polynomial approximation can be dirived using software such as 'R'.
+
+LIBYUV_API
+int ARGBPolynomial(const uint8_t* src_argb,
+                   int src_stride_argb,
+                   uint8_t* dst_argb,
+                   int dst_stride_argb,
+                   const float* poly,
+                   int width,
+                   int height);
+
+// Convert plane of 16 bit shorts to half floats.
+// Source values are multiplied by scale before storing as half float.
+//
+// Note: Unlike other libyuv functions that operate on uint16_t buffers, the
+// src_stride_y and dst_stride_y parameters of HalfFloatPlane() are in bytes,
+// not in units of uint16_t.
+LIBYUV_API
+int HalfFloatPlane(const uint16_t* src_y,
+                   int src_stride_y,
+                   uint16_t* dst_y,
+                   int dst_stride_y,
+                   float scale,
+                   int width,
+                   int height);
+
+// Convert a buffer of bytes to floats, scale the values and store as floats.
+LIBYUV_API
+int ByteToFloat(const uint8_t* src_y, float* dst_y, float scale, int width);
+
+// Quantize a rectangle of ARGB. Alpha unaffected.
+// scale is a 16 bit fractional fixed point scaler between 0 and 65535.
+// interval_size should be a value between 1 and 255.
+// interval_offset should be a value between 0 and 255.
+LIBYUV_API
+int ARGBQuantize(uint8_t* dst_argb,
+                 int dst_stride_argb,
+                 int scale,
+                 int interval_size,
+                 int interval_offset,
+                 int dst_x,
+                 int dst_y,
+                 int width,
+                 int height);
+
+// Copy ARGB to ARGB.
+LIBYUV_API
+int ARGBCopy(const uint8_t* src_argb,
+             int src_stride_argb,
+             uint8_t* dst_argb,
+             int dst_stride_argb,
+             int width,
+             int height);
+
+// Copy Alpha channel of ARGB to alpha of ARGB.
+LIBYUV_API
+int ARGBCopyAlpha(const uint8_t* src_argb,
+                  int src_stride_argb,
+                  uint8_t* dst_argb,
+                  int dst_stride_argb,
+                  int width,
+                  int height);
+
+// Extract the alpha channel from ARGB.
+LIBYUV_API
+int ARGBExtractAlpha(const uint8_t* src_argb,
+                     int src_stride_argb,
+                     uint8_t* dst_a,
+                     int dst_stride_a,
+                     int width,
+                     int height);
+
+// Copy Y channel to Alpha of ARGB.
+LIBYUV_API
+int ARGBCopyYToAlpha(const uint8_t* src_y,
+                     int src_stride_y,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     int width,
+                     int height);
+
+// Alpha Blend ARGB images and store to destination.
+// Source is pre-multiplied by alpha using ARGBAttenuate.
+// Alpha of destination is set to 255.
+LIBYUV_API
+int ARGBBlend(const uint8_t* src_argb0,
+              int src_stride_argb0,
+              const uint8_t* src_argb1,
+              int src_stride_argb1,
+              uint8_t* dst_argb,
+              int dst_stride_argb,
+              int width,
+              int height);
+
+// Alpha Blend plane and store to destination.
+// Source is not pre-multiplied by alpha.
+LIBYUV_API
+int BlendPlane(const uint8_t* src_y0,
+               int src_stride_y0,
+               const uint8_t* src_y1,
+               int src_stride_y1,
+               const uint8_t* alpha,
+               int alpha_stride,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               int width,
+               int height);
+
+// Alpha Blend YUV images and store to destination.
+// Source is not pre-multiplied by alpha.
+// Alpha is full width x height and subsampled to half size to apply to UV.
+LIBYUV_API
+int I420Blend(const uint8_t* src_y0,
+              int src_stride_y0,
+              const uint8_t* src_u0,
+              int src_stride_u0,
+              const uint8_t* src_v0,
+              int src_stride_v0,
+              const uint8_t* src_y1,
+              int src_stride_y1,
+              const uint8_t* src_u1,
+              int src_stride_u1,
+              const uint8_t* src_v1,
+              int src_stride_v1,
+              const uint8_t* alpha,
+              int alpha_stride,
+              uint8_t* dst_y,
+              int dst_stride_y,
+              uint8_t* dst_u,
+              int dst_stride_u,
+              uint8_t* dst_v,
+              int dst_stride_v,
+              int width,
+              int height);
+
+// Multiply ARGB image by ARGB image. Shifted down by 8. Saturates to 255.
+LIBYUV_API
+int ARGBMultiply(const uint8_t* src_argb0,
+                 int src_stride_argb0,
+                 const uint8_t* src_argb1,
+                 int src_stride_argb1,
+                 uint8_t* dst_argb,
+                 int dst_stride_argb,
+                 int width,
+                 int height);
+
+// Add ARGB image with ARGB image. Saturates to 255.
+LIBYUV_API
+int ARGBAdd(const uint8_t* src_argb0,
+            int src_stride_argb0,
+            const uint8_t* src_argb1,
+            int src_stride_argb1,
+            uint8_t* dst_argb,
+            int dst_stride_argb,
+            int width,
+            int height);
+
+// Subtract ARGB image (argb1) from ARGB image (argb0). Saturates to 0.
+LIBYUV_API
+int ARGBSubtract(const uint8_t* src_argb0,
+                 int src_stride_argb0,
+                 const uint8_t* src_argb1,
+                 int src_stride_argb1,
+                 uint8_t* dst_argb,
+                 int dst_stride_argb,
+                 int width,
+                 int height);
+
+// Convert I422 to YUY2.
+LIBYUV_API
+int I422ToYUY2(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_yuy2,
+               int dst_stride_yuy2,
+               int width,
+               int height);
+
+// Convert I422 to UYVY.
+LIBYUV_API
+int I422ToUYVY(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_uyvy,
+               int dst_stride_uyvy,
+               int width,
+               int height);
+
+// Convert unattentuated ARGB to preattenuated ARGB.
+LIBYUV_API
+int ARGBAttenuate(const uint8_t* src_argb,
+                  int src_stride_argb,
+                  uint8_t* dst_argb,
+                  int dst_stride_argb,
+                  int width,
+                  int height);
+
+// Convert preattentuated ARGB to unattenuated ARGB.
+LIBYUV_API
+int ARGBUnattenuate(const uint8_t* src_argb,
+                    int src_stride_argb,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    int width,
+                    int height);
+
+// Internal function - do not call directly.
+// Computes table of cumulative sum for image where the value is the sum
+// of all values above and to the left of the entry. Used by ARGBBlur.
+LIBYUV_API
+int ARGBComputeCumulativeSum(const uint8_t* src_argb,
+                             int src_stride_argb,
+                             int32_t* dst_cumsum,
+                             int dst_stride32_cumsum,
+                             int width,
+                             int height);
+
+// Blur ARGB image.
+// dst_cumsum table of width * (height + 1) * 16 bytes aligned to
+//   16 byte boundary.
+// dst_stride32_cumsum is number of ints in a row (width * 4).
+// radius is number of pixels around the center.  e.g. 1 = 3x3. 2=5x5.
+// Blur is optimized for radius of 5 (11x11) or less.
+LIBYUV_API
+int ARGBBlur(const uint8_t* src_argb,
+             int src_stride_argb,
+             uint8_t* dst_argb,
+             int dst_stride_argb,
+             int32_t* dst_cumsum,
+             int dst_stride32_cumsum,
+             int width,
+             int height,
+             int radius);
+
+// Gaussian 5x5 blur a float plane.
+// Coefficients of 1, 4, 6, 4, 1.
+// Each destination pixel is a blur of the 5x5
+// pixels from the source.
+// Source edges are clamped.
+LIBYUV_API
+int GaussPlane_F32(const float* src,
+                   int src_stride,
+                   float* dst,
+                   int dst_stride,
+                   int width,
+                   int height);
+
+// Multiply ARGB image by ARGB value.
+LIBYUV_API
+int ARGBShade(const uint8_t* src_argb,
+              int src_stride_argb,
+              uint8_t* dst_argb,
+              int dst_stride_argb,
+              int width,
+              int height,
+              uint32_t value);
+
+// Interpolate between two images using specified amount of interpolation
+// (0 to 255) and store to destination.
+// 'interpolation' is specified as 8 bit fraction where 0 means 100% src0
+// and 255 means 1% src0 and 99% src1.
+LIBYUV_API
+int InterpolatePlane(const uint8_t* src0,
+                     int src_stride0,
+                     const uint8_t* src1,
+                     int src_stride1,
+                     uint8_t* dst,
+                     int dst_stride,
+                     int width,
+                     int height,
+                     int interpolation);
+
+// Interpolate between two images using specified amount of interpolation
+// (0 to 255) and store to destination.
+// 'interpolation' is specified as 8 bit fraction where 0 means 100% src0
+// and 255 means 1% src0 and 99% src1.
+LIBYUV_API
+int InterpolatePlane_16(const uint16_t* src0,
+                        int src_stride0,  // measured in 16 bit pixels
+                        const uint16_t* src1,
+                        int src_stride1,
+                        uint16_t* dst,
+                        int dst_stride,
+                        int width,
+                        int height,
+                        int interpolation);
+
+// Interpolate between two ARGB images using specified amount of interpolation
+// Internally calls InterpolatePlane with width * 4 (bpp).
+LIBYUV_API
+int ARGBInterpolate(const uint8_t* src_argb0,
+                    int src_stride_argb0,
+                    const uint8_t* src_argb1,
+                    int src_stride_argb1,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    int width,
+                    int height,
+                    int interpolation);
+
+// Interpolate between two YUV images using specified amount of interpolation
+// Internally calls InterpolatePlane on each plane where the U and V planes
+// are half width and half height.
+LIBYUV_API
+int I420Interpolate(const uint8_t* src0_y,
+                    int src0_stride_y,
+                    const uint8_t* src0_u,
+                    int src0_stride_u,
+                    const uint8_t* src0_v,
+                    int src0_stride_v,
+                    const uint8_t* src1_y,
+                    int src1_stride_y,
+                    const uint8_t* src1_u,
+                    int src1_stride_u,
+                    const uint8_t* src1_v,
+                    int src1_stride_v,
+                    uint8_t* dst_y,
+                    int dst_stride_y,
+                    uint8_t* dst_u,
+                    int dst_stride_u,
+                    uint8_t* dst_v,
+                    int dst_stride_v,
+                    int width,
+                    int height,
+                    int interpolation);
+
+// Row function for copying pixels from a source with a slope to a row
+// of destination. Useful for scaling, rotation, mirror, texture mapping.
+LIBYUV_API
+void ARGBAffineRow_C(const uint8_t* src_argb,
+                     int src_argb_stride,
+                     uint8_t* dst_argb,
+                     const float* uv_dudv,
+                     int width);
+// TODO(fbarchard): Move ARGBAffineRow_SSE2 to row.h
+LIBYUV_API
+void ARGBAffineRow_SSE2(const uint8_t* src_argb,
+                        int src_argb_stride,
+                        uint8_t* dst_argb,
+                        const float* uv_dudv,
+                        int width);
+
+// Shuffle ARGB channel order.  e.g. BGRA to ARGB.
+// shuffler is 16 bytes.
+LIBYUV_API
+int ARGBShuffle(const uint8_t* src_bgra,
+                int src_stride_bgra,
+                uint8_t* dst_argb,
+                int dst_stride_argb,
+                const uint8_t* shuffler,
+                int width,
+                int height);
+
+// Shuffle AR64 channel order.  e.g. AR64 to AB64.
+// shuffler is 16 bytes.
+LIBYUV_API
+int AR64Shuffle(const uint16_t* src_ar64,
+                int src_stride_ar64,
+                uint16_t* dst_ar64,
+                int dst_stride_ar64,
+                const uint8_t* shuffler,
+                int width,
+                int height);
+
+// Sobel ARGB effect with planar output.
+LIBYUV_API
+int ARGBSobelToPlane(const uint8_t* src_argb,
+                     int src_stride_argb,
+                     uint8_t* dst_y,
+                     int dst_stride_y,
+                     int width,
+                     int height);
+
+// Sobel ARGB effect.
+LIBYUV_API
+int ARGBSobel(const uint8_t* src_argb,
+              int src_stride_argb,
+              uint8_t* dst_argb,
+              int dst_stride_argb,
+              int width,
+              int height);
+
+// Sobel ARGB effect w/ Sobel X, Sobel, Sobel Y in ARGB.
+LIBYUV_API
+int ARGBSobelXY(const uint8_t* src_argb,
+                int src_stride_argb,
+                uint8_t* dst_argb,
+                int dst_stride_argb,
+                int width,
+                int height);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_PLANAR_FUNCTIONS_H_

yuv.mod/libyuv/include/libyuv/rotate.h

@@ -0,0 +1,296 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_ROTATE_H_
+#define INCLUDE_LIBYUV_ROTATE_H_
+
+#include "libyuv/basic_types.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Supported rotation.
+typedef enum RotationMode {
+  kRotate0 = 0,      // No rotation.
+  kRotate90 = 90,    // Rotate 90 degrees clockwise.
+  kRotate180 = 180,  // Rotate 180 degrees.
+  kRotate270 = 270,  // Rotate 270 degrees clockwise.
+
+  // Deprecated.
+  kRotateNone = 0,
+  kRotateClockwise = 90,
+  kRotateCounterClockwise = 270,
+} RotationModeEnum;
+
+// Rotate I420 frame.
+LIBYUV_API
+int I420Rotate(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height,
+               enum RotationMode mode);
+
+// Rotate I422 frame.
+LIBYUV_API
+int I422Rotate(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height,
+               enum RotationMode mode);
+
+// Rotate I444 frame.
+LIBYUV_API
+int I444Rotate(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height,
+               enum RotationMode mode);
+
+// Rotate I010 frame.
+LIBYUV_API
+int I010Rotate(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height,
+               enum RotationMode mode);
+
+// Rotate I210 frame.
+LIBYUV_API
+int I210Rotate(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height,
+               enum RotationMode mode);
+
+// Rotate I410 frame.
+LIBYUV_API
+int I410Rotate(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height,
+               enum RotationMode mode);
+
+// Rotate NV12 input and store in I420.
+LIBYUV_API
+int NV12ToI420Rotate(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_uv,
+                     int src_stride_uv,
+                     uint8_t* dst_y,
+                     int dst_stride_y,
+                     uint8_t* dst_u,
+                     int dst_stride_u,
+                     uint8_t* dst_v,
+                     int dst_stride_v,
+                     int width,
+                     int height,
+                     enum RotationMode mode);
+
+// Convert Android420 to I420 with rotation.
+// "rotation" can be 0, 90, 180 or 270.
+LIBYUV_API
+int Android420ToI420Rotate(const uint8_t* src_y,
+                           int src_stride_y,
+                           const uint8_t* src_u,
+                           int src_stride_u,
+                           const uint8_t* src_v,
+                           int src_stride_v,
+                           int src_pixel_stride_uv,
+                           uint8_t* dst_y,
+                           int dst_stride_y,
+                           uint8_t* dst_u,
+                           int dst_stride_u,
+                           uint8_t* dst_v,
+                           int dst_stride_v,
+                           int width,
+                           int height,
+                           enum RotationMode rotation);
+
+// Rotate a plane by 0, 90, 180, or 270.
+LIBYUV_API
+int RotatePlane(const uint8_t* src,
+                int src_stride,
+                uint8_t* dst,
+                int dst_stride,
+                int width,
+                int height,
+                enum RotationMode mode);
+
+// Rotate planes by 90, 180, 270. Deprecated.
+LIBYUV_API
+void RotatePlane90(const uint8_t* src,
+                   int src_stride,
+                   uint8_t* dst,
+                   int dst_stride,
+                   int width,
+                   int height);
+
+LIBYUV_API
+void RotatePlane180(const uint8_t* src,
+                    int src_stride,
+                    uint8_t* dst,
+                    int dst_stride,
+                    int width,
+                    int height);
+
+LIBYUV_API
+void RotatePlane270(const uint8_t* src,
+                    int src_stride,
+                    uint8_t* dst,
+                    int dst_stride,
+                    int width,
+                    int height);
+
+// Rotate a plane by 0, 90, 180, or 270.
+LIBYUV_API
+int RotatePlane_16(const uint16_t* src,
+                   int src_stride,
+                   uint16_t* dst,
+                   int dst_stride,
+                   int width,
+                   int height,
+                   enum RotationMode mode);
+
+// Rotations for when U and V are interleaved.
+// These functions take one UV input pointer and
+// split the data into two buffers while
+// rotating them.
+// width and height expected to be half size for NV12.
+LIBYUV_API
+int SplitRotateUV(const uint8_t* src_uv,
+                  int src_stride_uv,
+                  uint8_t* dst_u,
+                  int dst_stride_u,
+                  uint8_t* dst_v,
+                  int dst_stride_v,
+                  int width,
+                  int height,
+                  enum RotationMode mode);
+
+LIBYUV_API
+void SplitRotateUV90(const uint8_t* src,
+                     int src_stride,
+                     uint8_t* dst_a,
+                     int dst_stride_a,
+                     uint8_t* dst_b,
+                     int dst_stride_b,
+                     int width,
+                     int height);
+
+LIBYUV_API
+void SplitRotateUV180(const uint8_t* src,
+                      int src_stride,
+                      uint8_t* dst_a,
+                      int dst_stride_a,
+                      uint8_t* dst_b,
+                      int dst_stride_b,
+                      int width,
+                      int height);
+
+LIBYUV_API
+void SplitRotateUV270(const uint8_t* src,
+                      int src_stride,
+                      uint8_t* dst_a,
+                      int dst_stride_a,
+                      uint8_t* dst_b,
+                      int dst_stride_b,
+                      int width,
+                      int height);
+
+// The 90 and 270 functions are based on transposes.
+// Doing a transpose with reversing the read/write
+// order will result in a rotation by +- 90 degrees.
+// Deprecated.
+LIBYUV_API
+void TransposePlane(const uint8_t* src,
+                    int src_stride,
+                    uint8_t* dst,
+                    int dst_stride,
+                    int width,
+                    int height);
+
+LIBYUV_API
+void SplitTransposeUV(const uint8_t* src,
+                      int src_stride,
+                      uint8_t* dst_a,
+                      int dst_stride_a,
+                      uint8_t* dst_b,
+                      int dst_stride_b,
+                      int width,
+                      int height);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_ROTATE_H_

yuv.mod/libyuv/include/libyuv/rotate_argb.h

@@ -0,0 +1,37 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_ROTATE_ARGB_H_
+#define INCLUDE_LIBYUV_ROTATE_ARGB_H_
+
+#include "libyuv/basic_types.h"
+#include "libyuv/rotate.h"  // For RotationMode.
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Rotate ARGB frame
+LIBYUV_API
+int ARGBRotate(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int src_width,
+               int src_height,
+               enum RotationMode mode);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_ROTATE_ARGB_H_

yuv.mod/libyuv/include/libyuv/rotate_row.h

@@ -0,0 +1,311 @@
+/*
+ *  Copyright 2013 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_ROTATE_ROW_H_
+#define INCLUDE_LIBYUV_ROTATE_ROW_H_
+
+#include "libyuv/basic_types.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+#if defined(__pnacl__) || defined(__CLR_VER) ||            \
+    (defined(__native_client__) && defined(__x86_64__)) || \
+    (defined(__i386__) && !defined(__SSE__) && !defined(__clang__))
+#define LIBYUV_DISABLE_X86
+#endif
+#if defined(__native_client__)
+#define LIBYUV_DISABLE_NEON
+#endif
+
+// temporary disable SME
+#if !defined(LIBYUV_DISABLE_SME)
+#define LIBYUV_DISABLE_SME
+#endif
+
+// MemorySanitizer does not support assembly code yet. http://crbug.com/344505
+#if defined(__has_feature)
+#if __has_feature(memory_sanitizer) && !defined(LIBYUV_DISABLE_NEON)
+#define LIBYUV_DISABLE_NEON
+#endif
+#if __has_feature(memory_sanitizer) && !defined(LIBYUV_DISABLE_X86)
+#define LIBYUV_DISABLE_X86
+#endif
+#endif
+// The following are available for Visual C 32 bit:
+#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && defined(_MSC_VER) && \
+    !defined(__clang__)
+#define HAS_TRANSPOSEWX8_SSSE3
+#define HAS_TRANSPOSEUVWX8_SSE2
+#endif
+
+// The following are available for GCC 32 or 64 bit:
+#if !defined(LIBYUV_DISABLE_X86) && (defined(__i386__) || defined(__x86_64__))
+#define HAS_TRANSPOSEWX8_SSSE3
+#define HAS_TRANSPOSE4X4_32_SSE2
+#define HAS_TRANSPOSE4X4_32_AVX2
+#endif
+
+// The following are available for 64 bit GCC:
+#if !defined(LIBYUV_DISABLE_X86) && defined(__x86_64__)
+#define HAS_TRANSPOSEWX8_FAST_SSSE3
+#define HAS_TRANSPOSEUVWX8_SSE2
+#endif
+
+#if !defined(LIBYUV_DISABLE_NEON) && \
+    (defined(__ARM_NEON__) || defined(LIBYUV_NEON) || defined(__aarch64__))
+#if defined(__aarch64__)
+#define HAS_TRANSPOSEWX16_NEON
+#else
+#define HAS_TRANSPOSEWX8_NEON
+#endif
+#define HAS_TRANSPOSEUVWX8_NEON
+#define HAS_TRANSPOSE4X4_32_NEON
+#endif
+
+#if !defined(LIBYUV_DISABLE_SME) && defined(__aarch64__)
+#define HAS_TRANSPOSEWXH_SME
+#define HAS_TRANSPOSEUVWXH_SME
+#endif
+
+#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
+#define HAS_TRANSPOSEWX16_MSA
+#define HAS_TRANSPOSEUVWX16_MSA
+#endif
+
+#if !defined(LIBYUV_DISABLE_LSX) && defined(__loongarch_sx)
+#define HAS_TRANSPOSEWX16_LSX
+#define HAS_TRANSPOSEUVWX16_LSX
+#endif
+
+void TransposeWxH_C(const uint8_t* src,
+                    int src_stride,
+                    uint8_t* dst,
+                    int dst_stride,
+                    int width,
+                    int height);
+
+void TransposeWx8_C(const uint8_t* src,
+                    int src_stride,
+                    uint8_t* dst,
+                    int dst_stride,
+                    int width);
+void TransposeWx16_C(const uint8_t* src,
+                     int src_stride,
+                     uint8_t* dst,
+                     int dst_stride,
+                     int width);
+void TransposeWx8_NEON(const uint8_t* src,
+                       int src_stride,
+                       uint8_t* dst,
+                       int dst_stride,
+                       int width);
+void TransposeWx16_NEON(const uint8_t* src,
+                        int src_stride,
+                        uint8_t* dst,
+                        int dst_stride,
+                        int width);
+void TransposeWxH_SME(const uint8_t* src,
+                      int src_stride,
+                      uint8_t* dst,
+                      int dst_stride,
+                      int width,
+                      int height);
+void TransposeWx8_SSSE3(const uint8_t* src,
+                        int src_stride,
+                        uint8_t* dst,
+                        int dst_stride,
+                        int width);
+void TransposeWx8_Fast_SSSE3(const uint8_t* src,
+                             int src_stride,
+                             uint8_t* dst,
+                             int dst_stride,
+                             int width);
+void TransposeWx16_MSA(const uint8_t* src,
+                       int src_stride,
+                       uint8_t* dst,
+                       int dst_stride,
+                       int width);
+void TransposeWx16_LSX(const uint8_t* src,
+                       int src_stride,
+                       uint8_t* dst,
+                       int dst_stride,
+                       int width);
+
+void TransposeWx8_Any_NEON(const uint8_t* src,
+                           int src_stride,
+                           uint8_t* dst,
+                           int dst_stride,
+                           int width);
+void TransposeWx16_Any_NEON(const uint8_t* src,
+                            int src_stride,
+                            uint8_t* dst,
+                            int dst_stride,
+                            int width);
+void TransposeWx8_Any_SSSE3(const uint8_t* src,
+                            int src_stride,
+                            uint8_t* dst,
+                            int dst_stride,
+                            int width);
+void TransposeWx8_Fast_Any_SSSE3(const uint8_t* src,
+                                 int src_stride,
+                                 uint8_t* dst,
+                                 int dst_stride,
+                                 int width);
+void TransposeWx16_Any_MSA(const uint8_t* src,
+                           int src_stride,
+                           uint8_t* dst,
+                           int dst_stride,
+                           int width);
+void TransposeWx16_Any_LSX(const uint8_t* src,
+                           int src_stride,
+                           uint8_t* dst,
+                           int dst_stride,
+                           int width);
+
+void TransposeUVWxH_C(const uint8_t* src,
+                      int src_stride,
+                      uint8_t* dst_a,
+                      int dst_stride_a,
+                      uint8_t* dst_b,
+                      int dst_stride_b,
+                      int width,
+                      int height);
+
+void TransposeUVWx8_C(const uint8_t* src,
+                      int src_stride,
+                      uint8_t* dst_a,
+                      int dst_stride_a,
+                      uint8_t* dst_b,
+                      int dst_stride_b,
+                      int width);
+void TransposeUVWx16_C(const uint8_t* src,
+                       int src_stride,
+                       uint8_t* dst_a,
+                       int dst_stride_a,
+                       uint8_t* dst_b,
+                       int dst_stride_b,
+                       int width);
+void TransposeUVWx8_SSE2(const uint8_t* src,
+                         int src_stride,
+                         uint8_t* dst_a,
+                         int dst_stride_a,
+                         uint8_t* dst_b,
+                         int dst_stride_b,
+                         int width);
+void TransposeUVWx8_NEON(const uint8_t* src,
+                         int src_stride,
+                         uint8_t* dst_a,
+                         int dst_stride_a,
+                         uint8_t* dst_b,
+                         int dst_stride_b,
+                         int width);
+void TransposeUVWxH_SME(const uint8_t* src,
+                        int src_stride,
+                        uint8_t* dst_a,
+                        int dst_stride_a,
+                        uint8_t* dst_b,
+                        int dst_stride_b,
+                        int width,
+                        int height);
+void TransposeUVWx16_MSA(const uint8_t* src,
+                         int src_stride,
+                         uint8_t* dst_a,
+                         int dst_stride_a,
+                         uint8_t* dst_b,
+                         int dst_stride_b,
+                         int width);
+void TransposeUVWx16_LSX(const uint8_t* src,
+                         int src_stride,
+                         uint8_t* dst_a,
+                         int dst_stride_a,
+                         uint8_t* dst_b,
+                         int dst_stride_b,
+                         int width);
+
+void TransposeUVWx8_Any_SSE2(const uint8_t* src,
+                             int src_stride,
+                             uint8_t* dst_a,
+                             int dst_stride_a,
+                             uint8_t* dst_b,
+                             int dst_stride_b,
+                             int width);
+void TransposeUVWx8_Any_NEON(const uint8_t* src,
+                             int src_stride,
+                             uint8_t* dst_a,
+                             int dst_stride_a,
+                             uint8_t* dst_b,
+                             int dst_stride_b,
+                             int width);
+void TransposeUVWx16_Any_MSA(const uint8_t* src,
+                             int src_stride,
+                             uint8_t* dst_a,
+                             int dst_stride_a,
+                             uint8_t* dst_b,
+                             int dst_stride_b,
+                             int width);
+void TransposeUVWx16_Any_LSX(const uint8_t* src,
+                             int src_stride,
+                             uint8_t* dst_a,
+                             int dst_stride_a,
+                             uint8_t* dst_b,
+                             int dst_stride_b,
+                             int width);
+void TransposeWxH_16_C(const uint16_t* src,
+                       int src_stride,
+                       uint16_t* dst,
+                       int dst_stride,
+                       int width,
+                       int height);
+
+void TransposeWx8_16_C(const uint16_t* src,
+                       int src_stride,
+                       uint16_t* dst,
+                       int dst_stride,
+                       int width);
+void TransposeWx1_16_C(const uint16_t* src,
+                       int src_stride,
+                       uint16_t* dst,
+                       int dst_stride,
+                       int width);
+
+// Transpose 32 bit values (ARGB)
+void Transpose4x4_32_NEON(const uint8_t* src,
+                          int src_stride,
+                          uint8_t* dst,
+                          int dst_stride,
+                          int width);
+
+void Transpose4x4_32_SSE2(const uint8_t* src,
+                          int src_stride,
+                          uint8_t* dst,
+                          int dst_stride,
+                          int width);
+
+void Transpose4x4_32_AVX2(const uint8_t* src,
+                          int src_stride,
+                          uint8_t* dst,
+                          int dst_stride,
+                          int width);
+
+void Transpose4x4_32_C(const uint8_t* src,
+                       int src_stride,
+                       uint8_t* dst,
+                       int dst_stride,
+                       int width);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_ROTATE_ROW_H_

yuv.mod/libyuv/include/libyuv/row.h

@@ -0,0 +1,6714 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_ROW_H_
+#define INCLUDE_LIBYUV_ROW_H_
+
+#include <stddef.h>  // For NULL
+#include <stdlib.h>  // For malloc
+
+#include "libyuv/basic_types.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+#if defined(__pnacl__) || defined(__CLR_VER) ||            \
+    (defined(__native_client__) && defined(__x86_64__)) || \
+    (defined(__i386__) && !defined(__SSE__) && !defined(__clang__))
+#define LIBYUV_DISABLE_X86
+#endif
+#if defined(__native_client__)
+#define LIBYUV_DISABLE_NEON
+#endif
+// MemorySanitizer does not support assembly code yet. http://crbug.com/344505
+#if defined(__has_feature)
+#if __has_feature(memory_sanitizer) && !defined(LIBYUV_DISABLE_NEON)
+#define LIBYUV_DISABLE_NEON
+#endif
+#if __has_feature(memory_sanitizer) && !defined(LIBYUV_DISABLE_X86)
+#define LIBYUV_DISABLE_X86
+#endif
+#endif
+// clang >= 3.5.0 required for Arm64.
+#if defined(__clang__) && defined(__aarch64__) && !defined(LIBYUV_DISABLE_NEON)
+#if (__clang_major__ < 3) || (__clang_major__ == 3 && (__clang_minor__ < 5))
+#define LIBYUV_DISABLE_NEON
+#endif  // clang >= 3.5
+#endif  // __clang__
+
+// GCC >= 4.7.0 required for AVX2.
+#if defined(__GNUC__) && (defined(__x86_64__) || defined(__i386__))
+#if (__GNUC__ > 4) || (__GNUC__ == 4 && (__GNUC_MINOR__ >= 7))
+#define GCC_HAS_AVX2 1
+#endif  // GNUC >= 4.7
+#endif  // __GNUC__
+
+// clang >= 3.4.0 required for AVX2.
+#if defined(__clang__) && (defined(__x86_64__) || defined(__i386__))
+#if (__clang_major__ > 3) || (__clang_major__ == 3 && (__clang_minor__ >= 4))
+#define CLANG_HAS_AVX2 1
+#endif  // clang >= 3.4
+#endif  // __clang__
+
+// clang >= 6.0.0 required for AVX512.
+#if defined(__clang__) && (defined(__x86_64__) || defined(__i386__))
+// clang in xcode follows a different versioning scheme.
+// TODO(fbarchard): fix xcode 9 ios b/789.
+#if (__clang_major__ >= 7) && !defined(__APPLE__)
+#define CLANG_HAS_AVX512 1
+#endif  // clang >= 7
+#endif  // __clang__
+
+// Visual C 2012 required for AVX2.
+#if defined(_M_IX86) && !defined(__clang__) && defined(_MSC_VER) && \
+    _MSC_VER >= 1700
+#define VISUALC_HAS_AVX2 1
+#endif  // VisualStudio >= 2012
+
+// The following are available on all x86 platforms:
+#if !defined(LIBYUV_DISABLE_X86) && \
+    (defined(_M_IX86) || defined(__x86_64__) || defined(__i386__))
+// Conversions:
+#define HAS_ABGRTOYROW_SSSE3
+#define HAS_ARGB1555TOARGBROW_SSE2
+#define HAS_ARGB4444TOARGBROW_SSE2
+#define HAS_ARGBEXTRACTALPHAROW_SSE2
+#define HAS_ARGBSETROW_X86
+#define HAS_ARGBSHUFFLEROW_SSSE3
+#define HAS_ARGBTOARGB1555ROW_SSE2
+#define HAS_ARGBTOARGB4444ROW_SSE2
+#define HAS_ARGBTORAWROW_SSSE3
+#define HAS_ARGBTORGB24ROW_SSSE3
+#define HAS_ARGBTORGB565DITHERROW_SSE2
+#define HAS_ARGBTORGB565ROW_SSE2
+#define HAS_ARGBTOYJROW_SSSE3
+#define HAS_ARGBTOYROW_SSSE3
+#define HAS_BGRATOYROW_SSSE3
+#define HAS_COPYROW_ERMS
+#define HAS_COPYROW_SSE2
+#define HAS_H422TOARGBROW_SSSE3
+#define HAS_HALFFLOATROW_SSE2
+#define HAS_I422TOARGB1555ROW_SSSE3
+#define HAS_I422TOARGB4444ROW_SSSE3
+#define HAS_I422TOARGBROW_SSSE3
+#define HAS_I422TORGB24ROW_SSSE3
+#define HAS_I422TORGB565ROW_SSSE3
+#define HAS_I422TORGBAROW_SSSE3
+#define HAS_I422TOUYVYROW_SSE2
+#define HAS_I422TOYUY2ROW_SSE2
+#define HAS_I444TOARGBROW_SSSE3
+#define HAS_I444TORGB24ROW_SSSE3
+#define HAS_INTERPOLATEROW_SSSE3
+#define HAS_J400TOARGBROW_SSE2
+#define HAS_J422TOARGBROW_SSSE3
+#define HAS_MERGEUVROW_SSE2
+#define HAS_MIRRORROW_SSSE3
+#define HAS_MIRRORSPLITUVROW_SSSE3
+#define HAS_NV12TOARGBROW_SSSE3
+#define HAS_NV12TORGB24ROW_SSSE3
+#define HAS_NV12TORGB565ROW_SSSE3
+#define HAS_NV21TOARGBROW_SSSE3
+#define HAS_NV21TORGB24ROW_SSSE3
+#define HAS_RAWTOARGBROW_SSSE3
+#define HAS_RAWTORGB24ROW_SSSE3
+#define HAS_RAWTOYJROW_SSSE3
+#define HAS_RAWTOYROW_SSSE3
+#define HAS_RGB24TOARGBROW_SSSE3
+#define HAS_RGB24TOYJROW_SSSE3
+#define HAS_RGB24TOYROW_SSSE3
+#define HAS_RGB565TOARGBROW_SSE2
+#define HAS_RGBATOYROW_SSSE3
+#define HAS_SETROW_ERMS
+#define HAS_SETROW_X86
+#define HAS_SPLITUVROW_SSE2
+#define HAS_UYVYTOARGBROW_SSSE3
+#define HAS_UYVYTOUV422ROW_SSE2
+#define HAS_UYVYTOUVROW_SSE2
+#define HAS_UYVYTOYROW_SSE2
+#define HAS_YUY2TOARGBROW_SSSE3
+#define HAS_YUY2TOUV422ROW_SSE2
+#define HAS_YUY2TOUVROW_SSE2
+#define HAS_YUY2TOYROW_SSE2
+#if !defined(LIBYUV_BIT_EXACT)
+#define HAS_ABGRTOUVROW_SSSE3
+#define HAS_ARGBTOUV444ROW_SSSE3
+#define HAS_ARGBTOUVJROW_SSSE3
+#define HAS_ARGBTOUVROW_SSSE3
+#define HAS_BGRATOUVROW_SSSE3
+#define HAS_RGBATOUVROW_SSSE3
+#endif
+
+// Effects:
+#define HAS_ARGBADDROW_SSE2
+#define HAS_ARGBAFFINEROW_SSE2
+#define HAS_ARGBBLENDROW_SSSE3
+#define HAS_ARGBCOLORMATRIXROW_SSSE3
+#define HAS_ARGBCOLORTABLEROW_X86
+#define HAS_ARGBCOPYALPHAROW_SSE2
+#define HAS_ARGBCOPYYTOALPHAROW_SSE2
+#define HAS_ARGBGRAYROW_SSSE3
+#define HAS_ARGBLUMACOLORTABLEROW_SSSE3
+#define HAS_ARGBMIRRORROW_SSE2
+#define HAS_ARGBMULTIPLYROW_SSE2
+#define HAS_ARGBPOLYNOMIALROW_SSE2
+#define HAS_ARGBQUANTIZEROW_SSE2
+#define HAS_ARGBSEPIAROW_SSSE3
+#define HAS_ARGBSHADEROW_SSE2
+#define HAS_ARGBSUBTRACTROW_SSE2
+#define HAS_BLENDPLANEROW_SSSE3
+#define HAS_COMPUTECUMULATIVESUMROW_SSE2
+#define HAS_CUMULATIVESUMTOAVERAGEROW_SSE2
+#define HAS_RGBCOLORTABLEROW_X86
+#define HAS_SOBELROW_SSE2
+#define HAS_SOBELTOPLANEROW_SSE2
+#define HAS_SOBELXROW_SSE2
+#define HAS_SOBELXYROW_SSE2
+#define HAS_SOBELYROW_SSE2
+
+// The following functions fail on gcc/clang 32 bit with fpic and framepointer.
+// caveat: clangcl uses row_win.cc which works.
+#if defined(__x86_64__) || !defined(__pic__) || defined(__clang__) || \
+    defined(_MSC_VER)
+// TODO(fbarchard): fix build error on android_full_debug=1
+// https://code.google.com/p/libyuv/issues/detail?id=517
+#define HAS_I422ALPHATOARGBROW_SSSE3
+#define HAS_I444ALPHATOARGBROW_SSSE3
+#endif
+#endif
+
+// The following are available on all x86 platforms, but
+// require VS2012, clang 3.4 or gcc 4.7.
+#if !defined(LIBYUV_DISABLE_X86) &&                          \
+    (defined(VISUALC_HAS_AVX2) || defined(CLANG_HAS_AVX2) || \
+     defined(GCC_HAS_AVX2))
+#define HAS_ARGBCOPYALPHAROW_AVX2
+#define HAS_ARGBCOPYYTOALPHAROW_AVX2
+#define HAS_ARGBEXTRACTALPHAROW_AVX2
+#define HAS_ARGBMIRRORROW_AVX2
+#define HAS_ARGBPOLYNOMIALROW_AVX2
+#define HAS_ARGBSHUFFLEROW_AVX2
+#define HAS_ARGBTORGB565DITHERROW_AVX2
+#define HAS_ARGBTOYJROW_AVX2
+#define HAS_ARGBTOYROW_AVX2
+#define HAS_COPYROW_AVX
+#define HAS_H422TOARGBROW_AVX2
+#define HAS_HALFFLOATROW_AVX2
+#define HAS_I422TOARGB1555ROW_AVX2
+#define HAS_I422TOARGB4444ROW_AVX2
+#define HAS_I422TOARGBROW_AVX2
+#define HAS_I422TORGB24ROW_AVX2
+#define HAS_I422TORGB565ROW_AVX2
+#define HAS_I422TORGBAROW_AVX2
+#define HAS_I444TOARGBROW_AVX2
+#define HAS_I444TORGB24ROW_AVX2
+#define HAS_INTERPOLATEROW_AVX2
+#define HAS_J422TOARGBROW_AVX2
+#define HAS_MERGEUVROW_AVX2
+#define HAS_MIRRORROW_AVX2
+#define HAS_NV12TOARGBROW_AVX2
+#define HAS_NV12TORGB24ROW_AVX2
+#define HAS_NV12TORGB565ROW_AVX2
+#define HAS_NV21TOARGBROW_AVX2
+#define HAS_NV21TORGB24ROW_AVX2
+#define HAS_RAWTOYJROW_AVX2
+#define HAS_RGB24TOYJROW_AVX2
+#define HAS_SPLITUVROW_AVX2
+#define HAS_UYVYTOARGBROW_AVX2
+#define HAS_UYVYTOUV422ROW_AVX2
+#define HAS_UYVYTOUVROW_AVX2
+#define HAS_UYVYTOYROW_AVX2
+#define HAS_YUY2TOARGBROW_AVX2
+#define HAS_YUY2TOUV422ROW_AVX2
+#define HAS_YUY2TOUVROW_AVX2
+#define HAS_YUY2TOYROW_AVX2
+//  #define HAS_HALFFLOATROW_F16C  // Enable to test half float cast
+#if !defined(LIBYUV_BIT_EXACT)
+#define HAS_ARGBTOUVJROW_AVX2
+#define HAS_ARGBTOUVROW_AVX2
+#endif
+
+// Effects:
+#define HAS_ARGBADDROW_AVX2
+#define HAS_ARGBMULTIPLYROW_AVX2
+#define HAS_ARGBSUBTRACTROW_AVX2
+#define HAS_BLENDPLANEROW_AVX2
+
+#if defined(__x86_64__) || !defined(__pic__) || defined(__clang__) || \
+    defined(_MSC_VER)
+// TODO(fbarchard): fix build error on android_full_debug=1
+// https://code.google.com/p/libyuv/issues/detail?id=517
+#define HAS_I422ALPHATOARGBROW_AVX2
+#define HAS_I444ALPHATOARGBROW_AVX2
+#endif
+#endif
+
+// The following are available for AVX2 Visual C 32 bit:
+// TODO(fbarchard): Port to gcc.
+#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && defined(_MSC_VER) && \
+    !defined(__clang__) && defined(VISUALC_HAS_AVX2)
+#define HAS_ARGB1555TOARGBROW_AVX2
+#define HAS_ARGB4444TOARGBROW_AVX2
+#define HAS_ARGBTOARGB1555ROW_AVX2
+#define HAS_ARGBTOARGB4444ROW_AVX2
+#define HAS_ARGBTORGB565ROW_AVX2
+#define HAS_J400TOARGBROW_AVX2
+#define HAS_RGB565TOARGBROW_AVX2
+#endif
+
+// The following are also available on x64 Visual C.
+#if !defined(LIBYUV_DISABLE_X86) && defined(_MSC_VER) && defined(_M_X64) && \
+    (!defined(__clang__) || defined(__SSSE3__))
+#define HAS_I444ALPHATOARGBROW_SSSE3
+#define HAS_I444TOARGBROW_SSSE3
+#define HAS_I422ALPHATOARGBROW_SSSE3
+#define HAS_I422TOARGBROW_SSSE3
+#endif
+
+// The following are available for gcc/clang x86 platforms:
+// TODO(fbarchard): Port to Visual C
+#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__))
+#define HAS_AB64TOARGBROW_SSSE3
+#define HAS_ABGRTOAR30ROW_SSSE3
+#define HAS_ABGRTOYJROW_SSSE3
+#define HAS_AR64TOARGBROW_SSSE3
+#define HAS_ARGBATTENUATEROW_SSSE3
+#define HAS_ARGBTOAB64ROW_SSSE3
+#define HAS_ARGBTOAR30ROW_SSSE3
+#define HAS_ARGBTOAR64ROW_SSSE3
+#define HAS_ARGBUNATTENUATEROW_SSE2
+#define HAS_CONVERT16TO8ROW_SSSE3
+#define HAS_CONVERT8TO16ROW_SSE2
+#define HAS_DETILEROW_16_SSE2
+#define HAS_DETILEROW_SSE2
+#define HAS_DETILESPLITUVROW_SSSE3
+#define HAS_DETILETOYUY2_SSE2
+#define HAS_HALFMERGEUVROW_SSSE3
+#define HAS_I210TOAR30ROW_SSSE3
+#define HAS_I210TOARGBROW_SSSE3
+#define HAS_I212TOAR30ROW_SSSE3
+#define HAS_I212TOARGBROW_SSSE3
+#define HAS_I400TOARGBROW_SSE2
+#define HAS_I410TOAR30ROW_SSSE3
+#define HAS_I410TOARGBROW_SSSE3
+#define HAS_I422TOAR30ROW_SSSE3
+#define HAS_MERGEARGBROW_SSE2
+#define HAS_MERGERGBROW_SSSE3
+#define HAS_MERGEXRGBROW_SSE2
+#define HAS_MIRRORUVROW_SSSE3
+#define HAS_NV21TOYUV24ROW_SSSE3
+#define HAS_P210TOAR30ROW_SSSE3
+#define HAS_P210TOARGBROW_SSSE3
+#define HAS_P410TOAR30ROW_SSSE3
+#define HAS_P410TOARGBROW_SSSE3
+#define HAS_RAWTORGBAROW_SSSE3
+#define HAS_RGB24MIRRORROW_SSSE3
+#define HAS_RGBATOYJROW_SSSE3
+#define HAS_SPLITARGBROW_SSE2
+#define HAS_SPLITARGBROW_SSSE3
+#define HAS_SPLITRGBROW_SSSE3
+#define HAS_SPLITXRGBROW_SSE2
+#define HAS_SPLITXRGBROW_SSSE3
+#define HAS_SWAPUVROW_SSSE3
+#define HAS_YUY2TONVUVROW_SSE2
+#if !defined(LIBYUV_BIT_EXACT)
+#define HAS_ABGRTOUVJROW_SSSE3
+#endif
+
+#if defined(__x86_64__) || !defined(__pic__)
+// TODO(fbarchard): fix build error on android_full_debug=1
+// https://code.google.com/p/libyuv/issues/detail?id=517
+#define HAS_I210ALPHATOARGBROW_SSSE3
+#define HAS_I410ALPHATOARGBROW_SSSE3
+#endif
+#endif
+
+// The following are available for AVX2 gcc/clang x86 platforms:
+// TODO(fbarchard): Port to Visual C
+#if !defined(LIBYUV_DISABLE_X86) &&               \
+    (defined(__x86_64__) || defined(__i386__)) && \
+    (defined(CLANG_HAS_AVX2) || defined(GCC_HAS_AVX2))
+#define HAS_AB64TOARGBROW_AVX2
+#define HAS_ABGRTOAR30ROW_AVX2
+#define HAS_ABGRTOYJROW_AVX2
+#define HAS_ABGRTOYROW_AVX2
+#define HAS_AR64TOARGBROW_AVX2
+#define HAS_ARGBATTENUATEROW_AVX2
+#define HAS_ARGBTOAB64ROW_AVX2
+#define HAS_ARGBTOAR30ROW_AVX2
+#define HAS_ARGBTOAR64ROW_AVX2
+#define HAS_ARGBTORAWROW_AVX2
+#define HAS_ARGBTORGB24ROW_AVX2
+#define HAS_ARGBUNATTENUATEROW_AVX2
+#define HAS_CONVERT16TO8ROW_AVX2
+#define HAS_CONVERT8TO16ROW_AVX2
+#define HAS_DETILEROW_16_AVX
+#define HAS_DIVIDEROW_16_AVX2
+#define HAS_HALFMERGEUVROW_AVX2
+#define HAS_I210TOAR30ROW_AVX2
+#define HAS_I210TOARGBROW_AVX2
+#define HAS_I212TOAR30ROW_AVX2
+#define HAS_I212TOARGBROW_AVX2
+#define HAS_I400TOARGBROW_AVX2
+#define HAS_I410TOAR30ROW_AVX2
+#define HAS_I410TOARGBROW_AVX2
+#define HAS_I422TOAR30ROW_AVX2
+#define HAS_I422TOUYVYROW_AVX2
+#define HAS_I422TOYUY2ROW_AVX2
+#define HAS_INTERPOLATEROW_16TO8_AVX2
+#define HAS_MERGEAR64ROW_AVX2
+#define HAS_MERGEARGB16TO8ROW_AVX2
+#define HAS_MERGEARGBROW_AVX2
+#define HAS_MERGEUVROW_16_AVX2
+#define HAS_MERGEXR30ROW_AVX2
+#define HAS_MERGEXR64ROW_AVX2
+#define HAS_MERGEXRGB16TO8ROW_AVX2
+#define HAS_MERGEXRGBROW_AVX2
+#define HAS_MIRRORUVROW_AVX2
+#define HAS_MULTIPLYROW_16_AVX2
+#define HAS_NV21TOYUV24ROW_AVX2
+#define HAS_P210TOAR30ROW_AVX2
+#define HAS_P210TOARGBROW_AVX2
+#define HAS_P410TOAR30ROW_AVX2
+#define HAS_P410TOARGBROW_AVX2
+#define HAS_RGBATOYJROW_AVX2
+#define HAS_SPLITARGBROW_AVX2
+#define HAS_SPLITUVROW_16_AVX2
+#define HAS_SPLITXRGBROW_AVX2
+#define HAS_SWAPUVROW_AVX2
+#define HAS_YUY2TONVUVROW_AVX2
+#if !defined(LIBYUV_BIT_EXACT)
+#define HAS_ABGRTOUVJROW_AVX2
+#define HAS_ABGRTOUVROW_AVX2
+#endif
+
+#if defined(__x86_64__) || !defined(__pic__)
+// TODO(fbarchard): fix build error on android_full_debug=1
+// https://code.google.com/p/libyuv/issues/detail?id=517
+#define HAS_I210ALPHATOARGBROW_AVX2
+#define HAS_I410ALPHATOARGBROW_AVX2
+#endif
+#endif
+
+// The following are available for AVX512 clang x86 platforms:
+// TODO(fbarchard): Port to GCC and Visual C
+// TODO(b/42280744): re-enable HAS_ARGBTORGB24ROW_AVX512VBMI.
+#if !defined(LIBYUV_DISABLE_X86) && \
+    (defined(__x86_64__) || defined(__i386__)) && defined(CLANG_HAS_AVX512)
+#define HAS_ARGBTORGB24ROW_AVX512VBMI
+#define HAS_CONVERT16TO8ROW_AVX512BW
+#define HAS_MERGEUVROW_AVX512BW
+#endif
+
+// The following are available for AVX512 clang x64 platforms:
+// TODO(fbarchard): Port to x86
+#if !defined(LIBYUV_DISABLE_X86) && defined(__x86_64__) && \
+    (defined(CLANG_HAS_AVX512))
+#define HAS_I422TOARGBROW_AVX512BW
+#endif
+
+// The following are available on Neon platforms:
+#if !defined(LIBYUV_DISABLE_NEON) && \
+    (defined(__aarch64__) || defined(__ARM_NEON__) || defined(LIBYUV_NEON))
+#define HAS_AB64TOARGBROW_NEON
+#define HAS_ABGRTOUVJROW_NEON
+#define HAS_ABGRTOUVROW_NEON
+#define HAS_ABGRTOYJROW_NEON
+#define HAS_ABGRTOYROW_NEON
+#define HAS_AR64TOARGBROW_NEON
+#define HAS_ARGB1555TOARGBROW_NEON
+#define HAS_ARGB1555TOUVROW_NEON
+#define HAS_ARGB1555TOYROW_NEON
+#define HAS_ARGB4444TOARGBROW_NEON
+#define HAS_ARGB4444TOUVROW_NEON
+#define HAS_ARGB4444TOYROW_NEON
+#define HAS_ARGBEXTRACTALPHAROW_NEON
+#define HAS_ARGBSETROW_NEON
+#define HAS_ARGBTOAB64ROW_NEON
+#define HAS_ARGBTOAR64ROW_NEON
+#define HAS_ARGBTOARGB1555ROW_NEON
+#define HAS_ARGBTOARGB4444ROW_NEON
+#define HAS_ARGBTORAWROW_NEON
+#define HAS_ARGBTORGB24ROW_NEON
+#define HAS_ARGBTORGB565DITHERROW_NEON
+#define HAS_ARGBTORGB565ROW_NEON
+#define HAS_ARGBTOUV444ROW_NEON
+#define HAS_ARGBTOUVJROW_NEON
+#define HAS_ARGBTOUVROW_NEON
+#define HAS_ARGBTOYJROW_NEON
+#define HAS_ARGBTOYROW_NEON
+#define HAS_AYUVTOUVROW_NEON
+#define HAS_AYUVTOVUROW_NEON
+#define HAS_AYUVTOYROW_NEON
+#define HAS_BGRATOUVROW_NEON
+#define HAS_BGRATOYROW_NEON
+#define HAS_BYTETOFLOATROW_NEON
+#define HAS_CONVERT16TO8ROW_NEON
+#define HAS_COPYROW_NEON
+#define HAS_DETILEROW_16_NEON
+#define HAS_DETILEROW_NEON
+#define HAS_DETILESPLITUVROW_NEON
+#define HAS_DETILETOYUY2_NEON
+#define HAS_UNPACKMT2T_NEON
+#define HAS_DIVIDEROW_16_NEON
+#define HAS_HALFFLOATROW_NEON
+#define HAS_HALFMERGEUVROW_NEON
+#define HAS_I400TOARGBROW_NEON
+#define HAS_I422ALPHATOARGBROW_NEON
+#define HAS_I422TOARGB1555ROW_NEON
+#define HAS_I422TOARGB4444ROW_NEON
+#define HAS_I422TOARGBROW_NEON
+#define HAS_I422TORGB24ROW_NEON
+#define HAS_I422TORGB565ROW_NEON
+#define HAS_I422TORGBAROW_NEON
+#define HAS_I422TOUYVYROW_NEON
+#define HAS_I422TOYUY2ROW_NEON
+#define HAS_I444ALPHATOARGBROW_NEON
+#define HAS_I444TOARGBROW_NEON
+#define HAS_I444TORGB24ROW_NEON
+#define HAS_INTERPOLATEROW_16_NEON
+#define HAS_INTERPOLATEROW_NEON
+#define HAS_J400TOARGBROW_NEON
+#define HAS_MERGEAR64ROW_NEON
+#define HAS_MERGEARGB16TO8ROW_NEON
+#define HAS_MERGEARGBROW_NEON
+#define HAS_MERGEUVROW_16_NEON
+#define HAS_MERGEUVROW_NEON
+#define HAS_MERGEXR30ROW_NEON
+#define HAS_MERGEXR64ROW_NEON
+#define HAS_MERGEXRGB16TO8ROW_NEON
+#define HAS_MERGEXRGBROW_NEON
+#define HAS_MIRRORROW_NEON
+#define HAS_MIRRORSPLITUVROW_NEON
+#define HAS_MIRRORUVROW_NEON
+#define HAS_MULTIPLYROW_16_NEON
+#define HAS_NV12TOARGBROW_NEON
+#define HAS_NV12TORGB24ROW_NEON
+#define HAS_NV12TORGB565ROW_NEON
+#define HAS_NV21TOARGBROW_NEON
+#define HAS_NV21TORGB24ROW_NEON
+#define HAS_NV21TOYUV24ROW_NEON
+#define HAS_RAWTOARGBROW_NEON
+#define HAS_RAWTORGB24ROW_NEON
+#define HAS_RAWTORGBAROW_NEON
+#define HAS_RAWTOUVJROW_NEON
+#define HAS_RAWTOUVROW_NEON
+#define HAS_RAWTOYJROW_NEON
+#define HAS_RAWTOYROW_NEON
+#define HAS_RGB24TOARGBROW_NEON
+#define HAS_RGB24TOUVJROW_NEON
+#define HAS_RGB24TOUVROW_NEON
+#define HAS_RGB24TOYJROW_NEON
+#define HAS_RGB24TOYROW_NEON
+#define HAS_RGB565TOARGBROW_NEON
+#define HAS_RGB565TOUVROW_NEON
+#define HAS_RGB565TOYROW_NEON
+#define HAS_RGBATOUVROW_NEON
+#define HAS_RGBATOYJROW_NEON
+#define HAS_RGBATOYROW_NEON
+#define HAS_SETROW_NEON
+#define HAS_SPLITARGBROW_NEON
+#define HAS_SPLITRGBROW_NEON
+#define HAS_SPLITUVROW_16_NEON
+#define HAS_SPLITUVROW_NEON
+#define HAS_SPLITXRGBROW_NEON
+#define HAS_SWAPUVROW_NEON
+#define HAS_UYVYTOARGBROW_NEON
+#define HAS_UYVYTOUV422ROW_NEON
+#define HAS_UYVYTOUVROW_NEON
+#define HAS_UYVYTOYROW_NEON
+#define HAS_YUY2TOARGBROW_NEON
+#define HAS_YUY2TONVUVROW_NEON
+#define HAS_YUY2TOUV422ROW_NEON
+#define HAS_YUY2TOUVROW_NEON
+#define HAS_YUY2TOYROW_NEON
+
+// Effects:
+#define HAS_ARGBADDROW_NEON
+#define HAS_ARGBATTENUATEROW_NEON
+#define HAS_ARGBBLENDROW_NEON
+#define HAS_ARGBCOLORMATRIXROW_NEON
+#define HAS_ARGBGRAYROW_NEON
+#define HAS_ARGBMIRRORROW_NEON
+#define HAS_ARGBMULTIPLYROW_NEON
+#define HAS_ARGBQUANTIZEROW_NEON
+#define HAS_ARGBSEPIAROW_NEON
+#define HAS_ARGBSHADEROW_NEON
+#define HAS_ARGBSHUFFLEROW_NEON
+#define HAS_ARGBSUBTRACTROW_NEON
+#define HAS_RGB24MIRRORROW_NEON
+#define HAS_SOBELROW_NEON
+#define HAS_SOBELTOPLANEROW_NEON
+#define HAS_SOBELXROW_NEON
+#define HAS_SOBELXYROW_NEON
+#define HAS_SOBELYROW_NEON
+#endif
+
+// The following are available on AArch64 platforms:
+#if !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__)
+#define HAS_ARGBTOAR30ROW_NEON
+#define HAS_ABGRTOAR30ROW_NEON
+#define HAS_I210ALPHATOARGBROW_NEON
+#define HAS_I410ALPHATOARGBROW_NEON
+#define HAS_I210TOARGBROW_NEON
+#define HAS_I410TOARGBROW_NEON
+#define HAS_I210TOAR30ROW_NEON
+#define HAS_I410TOAR30ROW_NEON
+#define HAS_I212TOARGBROW_NEON
+#define HAS_I212TOAR30ROW_NEON
+#define HAS_I422TOAR30ROW_NEON
+#define HAS_P210TOAR30ROW_NEON
+#define HAS_P210TOARGBROW_NEON
+#define HAS_P410TOAR30ROW_NEON
+#define HAS_P410TOARGBROW_NEON
+
+#define HAS_ABGRTOYJROW_NEON_DOTPROD
+#define HAS_ABGRTOYROW_NEON_DOTPROD
+#define HAS_ARGBTOYJROW_NEON_DOTPROD
+#define HAS_ARGBTOYROW_NEON_DOTPROD
+#define HAS_BGRATOYROW_NEON_DOTPROD
+#define HAS_RGBATOYJROW_NEON_DOTPROD
+#define HAS_RGBATOYROW_NEON_DOTPROD
+
+#define HAS_ARGBCOLORMATRIXROW_NEON_I8MM
+#define HAS_ARGBTOUV444ROW_NEON_I8MM
+#endif
+
+// The following are available on AArch64 SVE platforms:
+#if !defined(LIBYUV_DISABLE_SVE) && defined(__aarch64__)
+#define HAS_ABGRTOUVJROW_SVE2
+#define HAS_ABGRTOUVROW_SVE2
+#define HAS_ARGB1555TOARGBROW_SVE2
+#define HAS_ARGBTORAWROW_SVE2
+#define HAS_ARGBTORGB24ROW_SVE2
+#define HAS_ARGBTORGB565DITHERROW_SVE2
+#define HAS_ARGBTORGB565ROW_SVE2
+#define HAS_ARGBTOUVJROW_SVE2
+#define HAS_ARGBTOUVROW_SVE2
+#define HAS_AYUVTOUVROW_SVE2
+#define HAS_AYUVTOVUROW_SVE2
+#define HAS_BGRATOUVROW_SVE2
+#define HAS_I400TOARGBROW_SVE2
+#define HAS_I422ALPHATOARGBROW_SVE2
+#define HAS_I422TOARGBROW_SVE2
+#define HAS_I422TORGBAROW_SVE2
+#define HAS_I444ALPHATOARGBROW_SVE2
+#define HAS_I444TOARGBROW_SVE2
+#define HAS_NV12TOARGBROW_SVE2
+#define HAS_NV21TOARGBROW_SVE2
+#define HAS_RAWTOARGBROW_SVE2
+#define HAS_RAWTORGB24ROW_SVE2
+#define HAS_RAWTORGBAROW_SVE2
+#define HAS_RGB24TOARGBROW_SVE2
+#define HAS_RGBATOUVROW_SVE2
+#define HAS_UYVYTOARGBROW_SVE2
+#define HAS_YUY2TOARGBROW_SVE2
+#endif
+
+// The following are available on AArch64 platforms:
+#if !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__)
+#define HAS_GAUSSCOL_F32_NEON
+#define HAS_GAUSSROW_F32_NEON
+#define HAS_INTERPOLATEROW_16TO8_NEON
+#define HAS_SCALESUMSAMPLES_NEON
+#endif
+#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
+#define HAS_ABGRTOUVJROW_MSA
+#define HAS_ABGRTOUVROW_MSA
+#define HAS_ABGRTOYROW_MSA
+#define HAS_ARGB1555TOARGBROW_MSA
+#define HAS_ARGB1555TOUVROW_MSA
+#define HAS_ARGB1555TOYROW_MSA
+#define HAS_ARGB4444TOARGBROW_MSA
+#define HAS_ARGBADDROW_MSA
+#define HAS_ARGBATTENUATEROW_MSA
+#define HAS_ARGBBLENDROW_MSA
+#define HAS_ARGBCOLORMATRIXROW_MSA
+#define HAS_ARGBEXTRACTALPHAROW_MSA
+#define HAS_ARGBGRAYROW_MSA
+#define HAS_ARGBMIRRORROW_MSA
+#define HAS_ARGBMULTIPLYROW_MSA
+#define HAS_ARGBQUANTIZEROW_MSA
+#define HAS_ARGBSEPIAROW_MSA
+#define HAS_ARGBSETROW_MSA
+#define HAS_ARGBSHADEROW_MSA
+#define HAS_ARGBSHUFFLEROW_MSA
+#define HAS_ARGBSUBTRACTROW_MSA
+#define HAS_ARGBTOARGB1555ROW_MSA
+#define HAS_ARGBTOARGB4444ROW_MSA
+#define HAS_ARGBTORAWROW_MSA
+#define HAS_ARGBTORGB24ROW_MSA
+#define HAS_ARGBTORGB565DITHERROW_MSA
+#define HAS_ARGBTORGB565ROW_MSA
+#define HAS_ARGBTOUV444ROW_MSA
+#define HAS_ARGBTOUVJROW_MSA
+#define HAS_ARGBTOUVROW_MSA
+#define HAS_ARGBTOYJROW_MSA
+#define HAS_ARGBTOYROW_MSA
+#define HAS_BGRATOUVROW_MSA
+#define HAS_BGRATOYROW_MSA
+#define HAS_HALFFLOATROW_MSA
+#define HAS_I400TOARGBROW_MSA
+#define HAS_I422ALPHATOARGBROW_MSA
+#define HAS_I422TOARGB1555ROW_MSA
+#define HAS_I422TOARGB4444ROW_MSA
+#define HAS_I422TOARGBROW_MSA
+#define HAS_I422TORGB24ROW_MSA
+#define HAS_I422TORGB565ROW_MSA
+#define HAS_I422TORGBAROW_MSA
+#define HAS_I422TOUYVYROW_MSA
+#define HAS_I422TOYUY2ROW_MSA
+#define HAS_I444TOARGBROW_MSA
+#define HAS_INTERPOLATEROW_MSA
+#define HAS_J400TOARGBROW_MSA
+#define HAS_MERGEUVROW_MSA
+#define HAS_MIRRORROW_MSA
+#define HAS_MIRRORSPLITUVROW_MSA
+#define HAS_MIRRORUVROW_MSA
+#define HAS_NV12TOARGBROW_MSA
+#define HAS_NV12TORGB565ROW_MSA
+#define HAS_NV21TOARGBROW_MSA
+#define HAS_RAWTOARGBROW_MSA
+#define HAS_RAWTORGB24ROW_MSA
+#define HAS_RAWTOUVROW_MSA
+#define HAS_RAWTOYROW_MSA
+#define HAS_RGB24TOARGBROW_MSA
+#define HAS_RGB24TOUVROW_MSA
+#define HAS_RGB24TOYROW_MSA
+#define HAS_RGB565TOARGBROW_MSA
+#define HAS_RGB565TOUVROW_MSA
+#define HAS_RGB565TOYROW_MSA
+#define HAS_RGBATOUVROW_MSA
+#define HAS_RGBATOYROW_MSA
+#define HAS_SETROW_MSA
+#define HAS_SOBELROW_MSA
+#define HAS_SOBELTOPLANEROW_MSA
+#define HAS_SOBELXROW_MSA
+#define HAS_SOBELXYROW_MSA
+#define HAS_SOBELYROW_MSA
+#define HAS_SPLITUVROW_MSA
+#define HAS_UYVYTOARGBROW_MSA
+#define HAS_UYVYTOUVROW_MSA
+#define HAS_UYVYTOYROW_MSA
+#define HAS_YUY2TOARGBROW_MSA
+#define HAS_YUY2TOUV422ROW_MSA
+#define HAS_YUY2TOUVROW_MSA
+#define HAS_YUY2TOYROW_MSA
+#endif
+
+#if !defined(LIBYUV_DISABLE_LSX) && defined(__loongarch_sx)
+#define HAS_ABGRTOUVROW_LSX
+#define HAS_ABGRTOYROW_LSX
+#define HAS_ARGB1555TOARGBROW_LSX
+#define HAS_ARGB1555TOUVROW_LSX
+#define HAS_ARGB1555TOYROW_LSX
+#define HAS_ARGB4444TOARGBROW_LSX
+#define HAS_ARGBADDROW_LSX
+#define HAS_ARGBATTENUATEROW_LSX
+#define HAS_ARGBBLENDROW_LSX
+#define HAS_ARGBCOLORMATRIXROW_LSX
+#define HAS_ARGBEXTRACTALPHAROW_LSX
+#define HAS_ARGBGRAYROW_LSX
+#define HAS_ARGBSEPIAROW_LSX
+#define HAS_ARGBSHADEROW_LSX
+#define HAS_ARGBSHUFFLEROW_LSX
+#define HAS_ARGBSUBTRACTROW_LSX
+#define HAS_ARGBQUANTIZEROW_LSX
+#define HAS_ARGBSETROW_LSX
+#define HAS_ARGBTOARGB1555ROW_LSX
+#define HAS_ARGBTOARGB4444ROW_LSX
+#define HAS_ARGBTORAWROW_LSX
+#define HAS_ARGBTORGB24ROW_LSX
+#define HAS_ARGBTORGB565ROW_LSX
+#define HAS_ARGBTORGB565DITHERROW_LSX
+#define HAS_ARGBTOUVJROW_LSX
+#define HAS_ARGBTOUV444ROW_LSX
+#define HAS_ARGBTOUVROW_LSX
+#define HAS_ARGBTOYJROW_LSX
+#define HAS_ARGBMIRRORROW_LSX
+#define HAS_ARGBMULTIPLYROW_LSX
+#define HAS_BGRATOUVROW_LSX
+#define HAS_BGRATOYROW_LSX
+#define HAS_I400TOARGBROW_LSX
+#define HAS_I444TOARGBROW_LSX
+#define HAS_INTERPOLATEROW_LSX
+#define HAS_I422ALPHATOARGBROW_LSX
+#define HAS_I422TOARGB1555ROW_LSX
+#define HAS_I422TOARGB4444ROW_LSX
+#define HAS_I422TORGB24ROW_LSX
+#define HAS_I422TORGB565ROW_LSX
+#define HAS_I422TORGBAROW_LSX
+#define HAS_I422TOUYVYROW_LSX
+#define HAS_I422TOYUY2ROW_LSX
+#define HAS_J400TOARGBROW_LSX
+#define HAS_MERGEUVROW_LSX
+#define HAS_MIRRORROW_LSX
+#define HAS_MIRRORUVROW_LSX
+#define HAS_MIRRORSPLITUVROW_LSX
+#define HAS_NV12TOARGBROW_LSX
+#define HAS_NV12TORGB565ROW_LSX
+#define HAS_NV21TOARGBROW_LSX
+#define HAS_RAWTOARGBROW_LSX
+#define HAS_RAWTORGB24ROW_LSX
+#define HAS_RAWTOUVROW_LSX
+#define HAS_RAWTOYROW_LSX
+#define HAS_RGB24TOARGBROW_LSX
+#define HAS_RGB24TOUVROW_LSX
+#define HAS_RGB24TOYROW_LSX
+#define HAS_RGB565TOARGBROW_LSX
+#define HAS_RGB565TOUVROW_LSX
+#define HAS_RGB565TOYROW_LSX
+#define HAS_RGBATOUVROW_LSX
+#define HAS_RGBATOYROW_LSX
+#define HAS_SETROW_LSX
+#define HAS_SOBELROW_LSX
+#define HAS_SOBELTOPLANEROW_LSX
+#define HAS_SOBELXYROW_LSX
+#define HAS_SPLITUVROW_LSX
+#define HAS_UYVYTOARGBROW_LSX
+#define HAS_UYVYTOUV422ROW_LSX
+#define HAS_UYVYTOUVROW_LSX
+#define HAS_UYVYTOYROW_LSX
+#define HAS_YUY2TOARGBROW_LSX
+#define HAS_YUY2TOUVROW_LSX
+#define HAS_YUY2TOUV422ROW_LSX
+#define HAS_YUY2TOYROW_LSX
+#define HAS_ARGBTOYROW_LSX
+#define HAS_ABGRTOYJROW_LSX
+#define HAS_RGBATOYJROW_LSX
+#define HAS_RGB24TOYJROW_LSX
+#define HAS_RAWTOYJROW_LSX
+#endif
+
+#if !defined(LIBYUV_DISABLE_LSX) && defined(__loongarch_sx)
+#define HAS_I422TOARGBROW_LSX
+#endif
+
+#if !defined(LIBYUV_DISABLE_LASX) && defined(__loongarch_asx)
+#define HAS_ARGB1555TOARGBROW_LASX
+#define HAS_ARGB1555TOUVROW_LASX
+#define HAS_ARGB1555TOYROW_LASX
+#define HAS_ARGB4444TOARGBROW_LASX
+#define HAS_ARGBADDROW_LASX
+#define HAS_ARGBATTENUATEROW_LASX
+#define HAS_ARGBGRAYROW_LASX
+#define HAS_ARGBMIRRORROW_LASX
+#define HAS_ARGBMULTIPLYROW_LASX
+#define HAS_ARGBSEPIAROW_LASX
+#define HAS_ARGBSHADEROW_LASX
+#define HAS_ARGBSHUFFLEROW_LASX
+#define HAS_ARGBSUBTRACTROW_LASX
+#define HAS_ARGBTOARGB1555ROW_LASX
+#define HAS_ARGBTOARGB4444ROW_LASX
+#define HAS_ARGBTORAWROW_LASX
+#define HAS_ARGBTORGB24ROW_LASX
+#define HAS_ARGBTORGB565DITHERROW_LASX
+#define HAS_ARGBTORGB565ROW_LASX
+#define HAS_ARGBTOUV444ROW_LASX
+#define HAS_ARGBTOUVJROW_LASX
+#define HAS_ARGBTOUVROW_LASX
+#define HAS_ARGBTOYJROW_LASX
+#define HAS_ARGBTOYROW_LASX
+#define HAS_ABGRTOYJROW_LASX
+#define HAS_ABGRTOYROW_LASX
+#define HAS_I422ALPHATOARGBROW_LASX
+#define HAS_I422TOARGB1555ROW_LASX
+#define HAS_I422TOARGB4444ROW_LASX
+#define HAS_I422TOARGBROW_LASX
+#define HAS_I422TORGB24ROW_LASX
+#define HAS_I422TORGB565ROW_LASX
+#define HAS_I422TORGBAROW_LASX
+#define HAS_I422TOUYVYROW_LASX
+#define HAS_I422TOYUY2ROW_LASX
+#define HAS_MIRRORROW_LASX
+#define HAS_MIRRORUVROW_LASX
+#define HAS_NV12TOARGBROW_LASX
+#define HAS_NV12TORGB565ROW_LASX
+#define HAS_NV21TOARGBROW_LASX
+#define HAS_RAWTOARGBROW_LASX
+#define HAS_RAWTOUVROW_LASX
+#define HAS_RAWTOYROW_LASX
+#define HAS_RGB24TOARGBROW_LASX
+#define HAS_RGB24TOUVROW_LASX
+#define HAS_RGB24TOYROW_LASX
+#define HAS_RGB565TOARGBROW_LASX
+#define HAS_RGB565TOUVROW_LASX
+#define HAS_RGB565TOYROW_LASX
+#define HAS_UYVYTOUV422ROW_LASX
+#define HAS_UYVYTOUVROW_LASX
+#define HAS_UYVYTOYROW_LASX
+#define HAS_YUY2TOUV422ROW_LASX
+#define HAS_YUY2TOUVROW_LASX
+#define HAS_YUY2TOYROW_LASX
+#define HAS_RGBATOYROW_LASX
+#define HAS_RGBATOYJROW_LASX
+#define HAS_BGRATOYROW_LASX
+#define HAS_RGB24TOYJROW_LASX
+#define HAS_RAWTOYJROW_LASX
+#endif
+
+#if !defined(LIBYUV_DISABLE_RVV) && defined(__riscv_vector)
+#if __riscv_v_intrinsic > 11000
+// Since v0.12, TUPLE_TYPE is introduced for segment load and store.
+#define LIBYUV_RVV_HAS_TUPLE_TYPE
+// Since v0.12, VXRM(fixed-point rounding mode) is included in arguments of
+// fixed-point intrinsics.
+#define LIBYUV_RVV_HAS_VXRM_ARG
+#endif
+
+#define HAS_COPYROW_RVV
+#define HAS_AB64TOARGBROW_RVV
+#define HAS_ABGRTOYJROW_RVV
+#define HAS_ABGRTOYROW_RVV
+#define HAS_AR64TOARGBROW_RVV
+#define HAS_AR64TOAB64ROW_RVV
+#define HAS_ARGBATTENUATEROW_RVV
+#define HAS_ARGBBLENDROW_RVV
+#define HAS_ARGBCOPYYTOALPHAROW_RVV
+#define HAS_ARGBEXTRACTALPHAROW_RVV
+#define HAS_ARGBTOAB64ROW_RVV
+#define HAS_ARGBTOABGRROW_RVV
+#define HAS_ARGBTOAR64ROW_RVV
+#define HAS_ARGBTOBGRAROW_RVV
+#define HAS_ARGBTORAWROW_RVV
+#define HAS_ARGBTORGB24ROW_RVV
+#define HAS_ARGBTORGBAROW_RVV
+#define HAS_ARGBTOYJROW_RVV
+#define HAS_ARGBTOYMATRIXROW_RVV
+#define HAS_ARGBTOYROW_RVV
+#define HAS_BGRATOYROW_RVV
+#define HAS_BLENDPLANEROW_RVV
+#define HAS_I400TOARGBROW_RVV
+#define HAS_I422ALPHATOARGBROW_RVV
+#define HAS_I422TOARGBROW_RVV
+#define HAS_I422TORGB24ROW_RVV
+#define HAS_I422TORGBAROW_RVV
+#define HAS_I444ALPHATOARGBROW_RVV
+#define HAS_I444TOARGBROW_RVV
+#define HAS_I444TORGB24ROW_RVV
+#define HAS_INTERPOLATEROW_RVV
+#define HAS_J400TOARGBROW_RVV
+#define HAS_MERGEARGBROW_RVV
+#define HAS_MERGERGBROW_RVV
+#define HAS_MERGEUVROW_RVV
+#define HAS_MERGEXRGBROW_RVV
+#define HAS_NV12TOARGBROW_RVV
+#define HAS_NV12TORGB24ROW_RVV
+#define HAS_NV21TOARGBROW_RVV
+#define HAS_NV21TORGB24ROW_RVV
+#define HAS_RAWTOARGBROW_RVV
+#define HAS_RAWTORGB24ROW_RVV
+#define HAS_RAWTORGBAROW_RVV
+#define HAS_RAWTOYJROW_RVV
+#define HAS_RAWTOYROW_RVV
+#define HAS_RGB24TOARGBROW_RVV
+#define HAS_RGB24TOYJROW_RVV
+#define HAS_RGB24TOYROW_RVV
+#define HAS_RGBATOARGBROW_RVV
+#define HAS_RGBATOYJROW_RVV
+#define HAS_RGBATOYMATRIXROW_RVV
+#define HAS_RGBATOYROW_RVV
+#define HAS_RGBTOYMATRIXROW_RVV
+#define HAS_SPLITARGBROW_RVV
+#define HAS_SPLITRGBROW_RVV
+#define HAS_SPLITUVROW_RVV
+#define HAS_SPLITXRGBROW_RVV
+#endif
+
+#if defined(_MSC_VER) && !defined(__CLR_VER) && !defined(__clang__)
+#if defined(VISUALC_HAS_AVX2)
+#define SIMD_ALIGNED(var) __declspec(align(32)) var
+#else
+#define SIMD_ALIGNED(var) __declspec(align(16)) var
+#endif
+#define LIBYUV_NOINLINE __declspec(noinline)
+typedef __declspec(align(16)) int16_t vec16[8];
+typedef __declspec(align(16)) int32_t vec32[4];
+typedef __declspec(align(16)) float vecf32[4];
+typedef __declspec(align(16)) int8_t vec8[16];
+typedef __declspec(align(16)) uint16_t uvec16[8];
+typedef __declspec(align(16)) uint32_t uvec32[4];
+typedef __declspec(align(16)) uint8_t uvec8[16];
+typedef __declspec(align(32)) int16_t lvec16[16];
+typedef __declspec(align(32)) int32_t lvec32[8];
+typedef __declspec(align(32)) int8_t lvec8[32];
+typedef __declspec(align(32)) uint16_t ulvec16[16];
+typedef __declspec(align(32)) uint32_t ulvec32[8];
+typedef __declspec(align(32)) uint8_t ulvec8[32];
+#elif !defined(__pnacl__) && (defined(__GNUC__) || defined(__clang__))
+// Caveat GCC 4.2 to 4.7 have a known issue using vectors with const.
+#if defined(CLANG_HAS_AVX2) || defined(GCC_HAS_AVX2)
+#define SIMD_ALIGNED(var) var __attribute__((aligned(32)))
+#else
+#define SIMD_ALIGNED(var) var __attribute__((aligned(16)))
+#endif
+#define LIBYUV_NOINLINE __attribute__((noinline))
+typedef int16_t __attribute__((vector_size(16))) vec16;
+typedef int32_t __attribute__((vector_size(16))) vec32;
+typedef float __attribute__((vector_size(16))) vecf32;
+typedef int8_t __attribute__((vector_size(16))) vec8;
+typedef uint16_t __attribute__((vector_size(16))) uvec16;
+typedef uint32_t __attribute__((vector_size(16))) uvec32;
+typedef uint8_t __attribute__((vector_size(16))) uvec8;
+typedef int16_t __attribute__((vector_size(32))) lvec16;
+typedef int32_t __attribute__((vector_size(32))) lvec32;
+typedef int8_t __attribute__((vector_size(32))) lvec8;
+typedef uint16_t __attribute__((vector_size(32))) ulvec16;
+typedef uint32_t __attribute__((vector_size(32))) ulvec32;
+typedef uint8_t __attribute__((vector_size(32))) ulvec8;
+#else
+#define SIMD_ALIGNED(var) var
+#define LIBYUV_NOINLINE
+typedef int16_t vec16[8];
+typedef int32_t vec32[4];
+typedef float vecf32[4];
+typedef int8_t vec8[16];
+typedef uint16_t uvec16[8];
+typedef uint32_t uvec32[4];
+typedef uint8_t uvec8[16];
+typedef int16_t lvec16[16];
+typedef int32_t lvec32[8];
+typedef int8_t lvec8[32];
+typedef uint16_t ulvec16[16];
+typedef uint32_t ulvec32[8];
+typedef uint8_t ulvec8[32];
+#endif
+
+#if defined(__aarch64__) || defined(__arm__) || defined(__riscv)
+// This struct is for ARM and RISC-V color conversion.
+struct YuvConstants {
+  uvec8 kUVCoeff;
+  vec16 kRGBCoeffBias;
+};
+#else
+// This struct is for Intel color conversion.
+struct YuvConstants {
+  uint8_t kUVToB[32];
+  uint8_t kUVToG[32];
+  uint8_t kUVToR[32];
+  int16_t kYToRgb[16];
+  int16_t kYBiasToRgb[16];
+};
+
+// Offsets into YuvConstants structure
+#define KUVTOB 0
+#define KUVTOG 32
+#define KUVTOR 64
+#define KYTORGB 96
+#define KYBIASTORGB 128
+
+#endif
+
+#define IS_ALIGNED(p, a) (!((uintptr_t)(p) & ((a)-1)))
+
+#define align_buffer_64(var, size)                                         \
+  void* var##_mem = malloc((size) + 63);                      /* NOLINT */ \
+  uint8_t* var = (uint8_t*)(((intptr_t)var##_mem + 63) & ~63) /* NOLINT */
+
+#define free_aligned_buffer_64(var) \
+  free(var##_mem);                  \
+  var = NULL
+
+#if defined(__APPLE__) || defined(__x86_64__) || defined(__llvm__)
+#define OMITFP
+#else
+#define OMITFP __attribute__((optimize("omit-frame-pointer")))
+#endif
+
+// NaCL macros for GCC x86 and x64.
+#if defined(__native_client__)
+#define LABELALIGN ".p2align 5\n"
+#else
+#define LABELALIGN
+#endif
+
+// Intel Code Analizer markers.  Insert IACA_START IACA_END around code to be
+// measured and then run with iaca -64 libyuv_unittest.
+// IACA_ASM_START amd IACA_ASM_END are equivalents that can be used within
+// inline assembly blocks.
+// example of iaca:
+// ~/iaca-lin64/bin/iaca.sh -64 -analysis LATENCY out/Release/libyuv_unittest
+
+#if defined(__x86_64__) || defined(__i386__)
+
+#define IACA_ASM_START  \
+  ".byte 0x0F, 0x0B\n"  \
+  " movl $111, %%ebx\n" \
+  ".byte 0x64, 0x67, 0x90\n"
+
+#define IACA_ASM_END         \
+  " movl $222, %%ebx\n"      \
+  ".byte 0x64, 0x67, 0x90\n" \
+  ".byte 0x0F, 0x0B\n"
+
+#define IACA_SSC_MARK(MARK_ID)                        \
+  __asm__ __volatile__("\n\t  movl $" #MARK_ID        \
+                       ", %%ebx"                      \
+                       "\n\t  .byte 0x64, 0x67, 0x90" \
+                       :                              \
+                       :                              \
+                       : "memory");
+
+#define IACA_UD_BYTES __asm__ __volatile__("\n\t .byte 0x0F, 0x0B");
+
+#else /* Visual C */
+#define IACA_UD_BYTES \
+  { __asm _emit 0x0F __asm _emit 0x0B }
+
+#define IACA_SSC_MARK(x) \
+  { __asm mov ebx, x __asm _emit 0x64 __asm _emit 0x67 __asm _emit 0x90 }
+
+#define IACA_VC64_START __writegsbyte(111, 111);
+#define IACA_VC64_END __writegsbyte(222, 222);
+#endif
+
+#define IACA_START     \
+  {                    \
+    IACA_UD_BYTES      \
+    IACA_SSC_MARK(111) \
+  }
+#define IACA_END       \
+  {                    \
+    IACA_SSC_MARK(222) \
+    IACA_UD_BYTES      \
+  }
+
+void I210AlphaToARGBRow_NEON(const uint16_t* src_y,
+                             const uint16_t* src_u,
+                             const uint16_t* src_v,
+                             const uint16_t* src_a,
+                             uint8_t* rgb_buf,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I410AlphaToARGBRow_NEON(const uint16_t* src_y,
+                             const uint16_t* src_u,
+                             const uint16_t* src_v,
+                             const uint16_t* src_a,
+                             uint8_t* rgb_buf,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I444ToARGBRow_NEON(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I444ToARGBRow_SVE2(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I444ToRGB24Row_NEON(const uint8_t* src_y,
+                         const uint8_t* src_u,
+                         const uint8_t* src_v,
+                         uint8_t* dst_rgb24,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I210ToARGBRow_NEON(const uint16_t* src_y,
+                        const uint16_t* src_u,
+                        const uint16_t* src_v,
+                        uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I410ToARGBRow_NEON(const uint16_t* src_y,
+                        const uint16_t* src_u,
+                        const uint16_t* src_v,
+                        uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I210ToAR30Row_NEON(const uint16_t* src_y,
+                        const uint16_t* src_u,
+                        const uint16_t* src_v,
+                        uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I410ToAR30Row_NEON(const uint16_t* src_y,
+                        const uint16_t* src_u,
+                        const uint16_t* src_v,
+                        uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I212ToARGBRow_NEON(const uint16_t* src_y,
+                        const uint16_t* src_u,
+                        const uint16_t* src_v,
+                        uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I212ToAR30Row_NEON(const uint16_t* src_y,
+                        const uint16_t* src_u,
+                        const uint16_t* src_v,
+                        uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I422ToARGBRow_NEON(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I422ToARGBRow_SVE2(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I422ToAR30Row_NEON(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I444AlphaToARGBRow_NEON(const uint8_t* src_y,
+                             const uint8_t* src_u,
+                             const uint8_t* src_v,
+                             const uint8_t* src_a,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I444AlphaToARGBRow_SVE2(const uint8_t* src_y,
+                             const uint8_t* src_u,
+                             const uint8_t* src_v,
+                             const uint8_t* src_a,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422AlphaToARGBRow_NEON(const uint8_t* src_y,
+                             const uint8_t* src_u,
+                             const uint8_t* src_v,
+                             const uint8_t* src_a,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422AlphaToARGBRow_SVE2(const uint8_t* src_y,
+                             const uint8_t* src_u,
+                             const uint8_t* src_v,
+                             const uint8_t* src_a,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToRGBARow_NEON(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_rgba,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I422ToRGBARow_SVE2(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_rgba,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I422ToRGB24Row_NEON(const uint8_t* src_y,
+                         const uint8_t* src_u,
+                         const uint8_t* src_v,
+                         uint8_t* dst_rgb24,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I422ToRGB565Row_NEON(const uint8_t* src_y,
+                          const uint8_t* src_u,
+                          const uint8_t* src_v,
+                          uint8_t* dst_rgb565,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void I422ToARGB1555Row_NEON(const uint8_t* src_y,
+                            const uint8_t* src_u,
+                            const uint8_t* src_v,
+                            uint8_t* dst_argb1555,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToARGB4444Row_NEON(const uint8_t* src_y,
+                            const uint8_t* src_u,
+                            const uint8_t* src_v,
+                            uint8_t* dst_argb4444,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void NV12ToARGBRow_NEON(const uint8_t* src_y,
+                        const uint8_t* src_uv,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void NV12ToARGBRow_SVE2(const uint8_t* src_y,
+                        const uint8_t* src_uv,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void NV12ToRGB565Row_NEON(const uint8_t* src_y,
+                          const uint8_t* src_uv,
+                          uint8_t* dst_rgb565,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void NV21ToARGBRow_NEON(const uint8_t* src_y,
+                        const uint8_t* src_vu,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void NV21ToARGBRow_SVE2(const uint8_t* src_y,
+                        const uint8_t* src_vu,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void NV12ToRGB24Row_NEON(const uint8_t* src_y,
+                         const uint8_t* src_uv,
+                         uint8_t* dst_rgb24,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void NV21ToRGB24Row_NEON(const uint8_t* src_y,
+                         const uint8_t* src_vu,
+                         uint8_t* dst_rgb24,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void NV21ToYUV24Row_NEON(const uint8_t* src_y,
+                         const uint8_t* src_vu,
+                         uint8_t* dst_yuv24,
+                         int width);
+void YUY2ToARGBRow_NEON(const uint8_t* src_yuy2,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void YUY2ToARGBRow_SVE2(const uint8_t* src_yuy2,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void UYVYToARGBRow_NEON(const uint8_t* src_uyvy,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void UYVYToARGBRow_SVE2(const uint8_t* src_uyvy,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I444ToARGBRow_RVV(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void I444AlphaToARGBRow_RVV(const uint8_t* src_y,
+                            const uint8_t* src_u,
+                            const uint8_t* src_v,
+                            const uint8_t* src_a,
+                            uint8_t* dst_argb,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I444ToRGB24Row_RVV(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_rgb24,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I422ToARGBRow_RVV(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void I422AlphaToARGBRow_RVV(const uint8_t* src_y,
+                            const uint8_t* src_u,
+                            const uint8_t* src_v,
+                            const uint8_t* src_a,
+                            uint8_t* dst_argb,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToRGBARow_RVV(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_rgba,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void I422ToRGB24Row_RVV(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_rgb24,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I444ToARGBRow_MSA(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void I444ToARGBRow_LSX(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+
+void I422ToARGBRow_MSA(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void I422ToARGBRow_LSX(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void I422ToARGBRow_LASX(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I422ToRGBARow_MSA(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void I422ToRGBARow_LSX(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void I422ToRGBARow_LASX(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I422AlphaToARGBRow_MSA(const uint8_t* src_y,
+                            const uint8_t* src_u,
+                            const uint8_t* src_v,
+                            const uint8_t* src_a,
+                            uint8_t* dst_argb,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422AlphaToARGBRow_LSX(const uint8_t* src_y,
+                            const uint8_t* src_u,
+                            const uint8_t* src_v,
+                            const uint8_t* src_a,
+                            uint8_t* dst_argb,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422AlphaToARGBRow_LASX(const uint8_t* src_y,
+                             const uint8_t* src_u,
+                             const uint8_t* src_v,
+                             const uint8_t* src_a,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToRGB24Row_MSA(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I422ToRGB24Row_LSX(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I422ToRGB24Row_LASX(const uint8_t* src_y,
+                         const uint8_t* src_u,
+                         const uint8_t* src_v,
+                         uint8_t* dst_argb,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I422ToRGB565Row_MSA(const uint8_t* src_y,
+                         const uint8_t* src_u,
+                         const uint8_t* src_v,
+                         uint8_t* dst_rgb565,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I422ToRGB565Row_LSX(const uint8_t* src_y,
+                         const uint8_t* src_u,
+                         const uint8_t* src_v,
+                         uint8_t* dst_rgb565,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I422ToRGB565Row_LASX(const uint8_t* src_y,
+                          const uint8_t* src_u,
+                          const uint8_t* src_v,
+                          uint8_t* dst_rgb565,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void I422ToARGB4444Row_MSA(const uint8_t* src_y,
+                           const uint8_t* src_u,
+                           const uint8_t* src_v,
+                           uint8_t* dst_argb4444,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I422ToARGB4444Row_LSX(const uint8_t* src_y,
+                           const uint8_t* src_u,
+                           const uint8_t* src_v,
+                           uint8_t* dst_argb4444,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I422ToARGB4444Row_LASX(const uint8_t* src_y,
+                            const uint8_t* src_u,
+                            const uint8_t* src_v,
+                            uint8_t* dst_argb4444,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToARGB1555Row_MSA(const uint8_t* src_y,
+                           const uint8_t* src_u,
+                           const uint8_t* src_v,
+                           uint8_t* dst_argb1555,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I422ToARGB1555Row_LSX(const uint8_t* src_y,
+                           const uint8_t* src_u,
+                           const uint8_t* src_v,
+                           uint8_t* dst_argb1555,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I422ToARGB1555Row_LASX(const uint8_t* src_y,
+                            const uint8_t* src_u,
+                            const uint8_t* src_v,
+                            uint8_t* dst_argb1555,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void NV12ToARGBRow_MSA(const uint8_t* src_y,
+                       const uint8_t* src_uv,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void NV12ToRGB565Row_MSA(const uint8_t* src_y,
+                         const uint8_t* src_uv,
+                         uint8_t* dst_rgb565,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void NV21ToARGBRow_MSA(const uint8_t* src_y,
+                       const uint8_t* src_vu,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void YUY2ToARGBRow_MSA(const uint8_t* src_yuy2,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void UYVYToARGBRow_MSA(const uint8_t* src_uyvy,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+
+void NV12ToARGBRow_LSX(const uint8_t* src_y,
+                       const uint8_t* src_uv,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void NV12ToARGBRow_LASX(const uint8_t* src_y,
+                        const uint8_t* src_uv,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void NV12ToRGB565Row_LSX(const uint8_t* src_y,
+                         const uint8_t* src_uv,
+                         uint8_t* dst_rgb565,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void NV12ToRGB565Row_LASX(const uint8_t* src_y,
+                          const uint8_t* src_uv,
+                          uint8_t* dst_rgb565,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void NV21ToARGBRow_LSX(const uint8_t* src_y,
+                       const uint8_t* src_vu,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void NV21ToARGBRow_LASX(const uint8_t* src_y,
+                        const uint8_t* src_vu,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void YUY2ToARGBRow_LSX(const uint8_t* src_yuy2,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void UYVYToARGBRow_LSX(const uint8_t* src_uyvy,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void NV12ToARGBRow_RVV(const uint8_t* src_y,
+                       const uint8_t* src_uv,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void NV21ToARGBRow_RVV(const uint8_t* src_y,
+                       const uint8_t* src_vu,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void NV12ToRGB24Row_RVV(const uint8_t* src_y,
+                        const uint8_t* src_uv,
+                        uint8_t* dst_rgb24,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void NV21ToRGB24Row_RVV(const uint8_t* src_y,
+                        const uint8_t* src_vu,
+                        uint8_t* dst_rgb24,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+
+void ARGBToYRow_AVX2(const uint8_t* src_argb, uint8_t* dst_y, int width);
+void ARGBToYRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYRow_AVX2(const uint8_t* src_abgr, uint8_t* dst_y, int width);
+void ABGRToYRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYRow_SSSE3(const uint8_t* src_argb, uint8_t* dst_y, int width);
+void ARGBToYJRow_SSSE3(const uint8_t* src_argb, uint8_t* dst_y, int width);
+void ARGBToYJRow_AVX2(const uint8_t* src_argb, uint8_t* dst_y, int width);
+void ARGBToYJRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYRow_SSSE3(const uint8_t* src_abgr, uint8_t* dst_y, int width);
+void ABGRToYJRow_SSSE3(const uint8_t* src_abgr, uint8_t* dst_y, int width);
+void ABGRToYJRow_AVX2(const uint8_t* src_abgr, uint8_t* dst_y, int width);
+void ABGRToYJRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYJRow_AVX2(const uint8_t* src_rgba, uint8_t* dst_y, int width);
+void RGBAToYJRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYJRow_SSSE3(const uint8_t* src_rgba, uint8_t* dst_y, int width);
+void BGRAToYRow_SSSE3(const uint8_t* src_bgra, uint8_t* dst_y, int width);
+void ABGRToYRow_SSSE3(const uint8_t* src_abgr, uint8_t* dst_y, int width);
+void RGBAToYRow_SSSE3(const uint8_t* src_rgba, uint8_t* dst_y, int width);
+void RGB24ToYRow_SSSE3(const uint8_t* src_rgb24, uint8_t* dst_y, int width);
+void RGB24ToYJRow_SSSE3(const uint8_t* src_rgb24, uint8_t* dst_yj, int width);
+void RAWToYRow_SSSE3(const uint8_t* src_raw, uint8_t* dst_y, int width);
+void RAWToYJRow_SSSE3(const uint8_t* src_raw, uint8_t* dst_yj, int width);
+void RGB24ToYJRow_AVX2(const uint8_t* src_rgb24, uint8_t* dst_yj, int width);
+void RAWToYJRow_AVX2(const uint8_t* src_raw, uint8_t* dst_yj, int width);
+void ARGBToYRow_NEON(const uint8_t* src_argb, uint8_t* dst_y, int width);
+void ARGBToYJRow_NEON(const uint8_t* src_argb, uint8_t* dst_yj, int width);
+void ABGRToYJRow_NEON(const uint8_t* src_abgr, uint8_t* dst_yj, int width);
+void RGBAToYJRow_NEON(const uint8_t* src_rgba, uint8_t* dst_yj, int width);
+void ARGBToYRow_NEON_DotProd(const uint8_t* src_argb,
+                             uint8_t* dst_y,
+                             int width);
+void ARGBToYJRow_NEON_DotProd(const uint8_t* src_argb,
+                              uint8_t* dst_yj,
+                              int width);
+void ABGRToYJRow_NEON_DotProd(const uint8_t* src_abgr,
+                              uint8_t* dst_yj,
+                              int width);
+void RGBAToYJRow_NEON_DotProd(const uint8_t* src_rgba,
+                              uint8_t* dst_yj,
+                              int width);
+void ARGBToYRow_RVV(const uint8_t* src_argb, uint8_t* dst_y, int width);
+void ARGBToYJRow_RVV(const uint8_t* src_argb, uint8_t* dst_yj, int width);
+void ABGRToYJRow_RVV(const uint8_t* src_rgba, uint8_t* dst_yj, int width);
+void RGBAToYJRow_RVV(const uint8_t* src_rgba, uint8_t* dst_yj, int width);
+void ARGBToYRow_MSA(const uint8_t* src_argb0, uint8_t* dst_y, int width);
+void ARGBToYJRow_MSA(const uint8_t* src_argb0, uint8_t* dst_y, int width);
+void ARGBToYRow_LSX(const uint8_t* src_argb0, uint8_t* dst_y, int width);
+void ARGBToYRow_LASX(const uint8_t* src_argb0, uint8_t* dst_y, int width);
+void ARGBToYJRow_LSX(const uint8_t* src_argb0, uint8_t* dst_y, int width);
+void ABGRToYJRow_LSX(const uint8_t* src_abgr, uint8_t* dst_yj, int width);
+void RGBAToYJRow_LSX(const uint8_t* src_rgba, uint8_t* dst_yj, int width);
+void ARGBToYJRow_LASX(const uint8_t* src_argb0, uint8_t* dst_y, int width);
+void ABGRToYJRow_LASX(const uint8_t* src_abgr, uint8_t* dst_yj, int width);
+void RGBAToYJRow_LASX(const uint8_t* src_rgba, uint8_t* dst_yj, int width);
+void ARGBToUV444Row_NEON(const uint8_t* src_argb,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void ARGBToUV444Row_NEON_I8MM(const uint8_t* src_argb,
+                              uint8_t* dst_u,
+                              uint8_t* dst_v,
+                              int width);
+void ARGBToUVRow_NEON(const uint8_t* src_argb,
+                      int src_stride_argb,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void ARGBToUVRow_SVE2(const uint8_t* src_argb,
+                      int src_stride_argb,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void ARGBToUV444Row_MSA(const uint8_t* src_argb,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void ARGBToUVRow_MSA(const uint8_t* src_argb,
+                     int src_stride_argb,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void ARGBToUVRow_LSX(const uint8_t* src_argb,
+                     int src_stride_argb,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void ARGBToUVRow_LASX(const uint8_t* src_argb,
+                      int src_stride_argb,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void ARGBToUV444Row_LSX(const uint8_t* src_argb,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void ARGBToUV444Row_LASX(const uint8_t* src_argb,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void ARGBToUVJRow_NEON(const uint8_t* src_argb,
+                       int src_stride_argb,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void ARGBToUVJRow_SVE2(const uint8_t* src_argb,
+                       int src_stride_argb,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void ABGRToUVJRow_NEON(const uint8_t* src_abgr,
+                       int src_stride_abgr,
+                       uint8_t* dst_uj,
+                       uint8_t* dst_vj,
+                       int width);
+void ABGRToUVJRow_SVE2(const uint8_t* src_abgr,
+                       int src_stride_abgr,
+                       uint8_t* dst_uj,
+                       uint8_t* dst_vj,
+                       int width);
+void BGRAToUVRow_NEON(const uint8_t* src_bgra,
+                      int src_stride_bgra,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void BGRAToUVRow_SVE2(const uint8_t* src_bgra,
+                      int src_stride_bgra,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void ABGRToUVRow_NEON(const uint8_t* src_abgr,
+                      int src_stride_abgr,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void ABGRToUVRow_SVE2(const uint8_t* src_abgr,
+                      int src_stride_abgr,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void RGBAToUVRow_NEON(const uint8_t* src_rgba,
+                      int src_stride_rgba,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void RGBAToUVRow_SVE2(const uint8_t* src_rgba,
+                      int src_stride_rgba,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void RGB24ToUVRow_NEON(const uint8_t* src_rgb24,
+                       int src_stride_rgb24,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void RAWToUVRow_NEON(const uint8_t* src_raw,
+                     int src_stride_raw,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void RGB24ToUVJRow_NEON(const uint8_t* src_rgb24,
+                        int src_stride_rgb24,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void RAWToUVJRow_NEON(const uint8_t* src_raw,
+                      int src_stride_raw,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void RGB565ToUVRow_NEON(const uint8_t* src_rgb565,
+                        int src_stride_rgb565,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void ARGB1555ToUVRow_NEON(const uint8_t* src_argb1555,
+                          int src_stride_argb1555,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void ARGB4444ToUVRow_NEON(const uint8_t* src_argb4444,
+                          int src_stride_argb4444,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void ARGBToUVJRow_MSA(const uint8_t* src_rgb,
+                      int src_stride_rgb,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void ABGRToUVJRow_MSA(const uint8_t* src_rgb,
+                      int src_stride_rgb,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void BGRAToUVRow_MSA(const uint8_t* src_rgb,
+                     int src_stride_rgb,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void ABGRToUVRow_MSA(const uint8_t* src_rgb,
+                     int src_stride_rgb,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void RGBAToUVRow_MSA(const uint8_t* src_rgb,
+                     int src_stride_rgb,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void RGB24ToUVRow_MSA(const uint8_t* src_rgb,
+                      int src_stride_rgb,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void RAWToUVRow_MSA(const uint8_t* src_rgb,
+                    int src_stride_rgb,
+                    uint8_t* dst_u,
+                    uint8_t* dst_v,
+                    int width);
+void RGB565ToUVRow_MSA(const uint8_t* src_rgb565,
+                       int src_stride_rgb565,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void ARGB1555ToUVRow_MSA(const uint8_t* src_argb1555,
+                         int src_stride_argb1555,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void BGRAToUVRow_LSX(const uint8_t* src_bgra,
+                     int src_stride_bgra,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void ABGRToUVRow_LSX(const uint8_t* src_abgr,
+                     int src_stride_abgr,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void RGBAToUVRow_LSX(const uint8_t* src_rgba,
+                     int src_stride_rgba,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void ARGBToUVJRow_LSX(const uint8_t* src_argb,
+                      int src_stride_argb,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void ARGBToUVJRow_LASX(const uint8_t* src_argb,
+                       int src_stride_argb,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void ARGB1555ToUVRow_LSX(const uint8_t* src_argb1555,
+                         int src_stride_argb1555,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void ARGB1555ToUVRow_LASX(const uint8_t* src_argb1555,
+                          int src_stride_argb1555,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void RGB565ToUVRow_LSX(const uint8_t* src_rgb565,
+                       int src_stride_rgb565,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void RGB565ToUVRow_LASX(const uint8_t* src_rgb565,
+                        int src_stride_rgb565,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void RGB24ToUVRow_LSX(const uint8_t* src_rgb24,
+                      int src_stride_rgb24,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void RGB24ToUVRow_LASX(const uint8_t* src_rgb24,
+                       int src_stride_rgb24,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void RAWToUVRow_LSX(const uint8_t* src_raw,
+                    int src_stride_raw,
+                    uint8_t* dst_u,
+                    uint8_t* dst_v,
+                    int width);
+void RAWToUVRow_LASX(const uint8_t* src_raw,
+                     int src_stride_raw,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void BGRAToYRow_NEON(const uint8_t* src_bgra, uint8_t* dst_y, int width);
+void ABGRToYRow_NEON(const uint8_t* src_abgr, uint8_t* dst_y, int width);
+void RGBAToYRow_NEON(const uint8_t* src_rgba, uint8_t* dst_y, int width);
+void BGRAToYRow_NEON_DotProd(const uint8_t* src_bgra,
+                             uint8_t* dst_y,
+                             int width);
+void ABGRToYRow_NEON_DotProd(const uint8_t* src_abgr,
+                             uint8_t* dst_y,
+                             int width);
+void RGBAToYRow_NEON_DotProd(const uint8_t* src_rgba,
+                             uint8_t* dst_y,
+                             int width);
+void RGB24ToYRow_NEON(const uint8_t* src_rgb24, uint8_t* dst_y, int width);
+void RGB24ToYJRow_NEON(const uint8_t* src_rgb24, uint8_t* dst_yj, int width);
+void RAWToYRow_NEON(const uint8_t* src_raw, uint8_t* dst_y, int width);
+void RAWToYJRow_NEON(const uint8_t* src_raw, uint8_t* dst_yj, int width);
+void RGB565ToYRow_NEON(const uint8_t* src_rgb565, uint8_t* dst_y, int width);
+void ARGB1555ToYRow_NEON(const uint8_t* src_argb1555,
+                         uint8_t* dst_y,
+                         int width);
+void ARGB4444ToYRow_NEON(const uint8_t* src_argb4444,
+                         uint8_t* dst_y,
+                         int width);
+void BGRAToYRow_RVV(const uint8_t* src_bgra, uint8_t* dst_y, int width);
+void ABGRToYRow_RVV(const uint8_t* src_abgr, uint8_t* dst_y, int width);
+void RGBAToYRow_RVV(const uint8_t* src_rgba, uint8_t* dst_y, int width);
+void RGB24ToYRow_RVV(const uint8_t* src_rgb24, uint8_t* dst_y, int width);
+void RGB24ToYJRow_RVV(const uint8_t* src_rgb24, uint8_t* dst_yj, int width);
+void RAWToYRow_RVV(const uint8_t* src_raw, uint8_t* dst_y, int width);
+void RAWToYJRow_RVV(const uint8_t* src_raw, uint8_t* dst_yj, int width);
+void BGRAToYRow_MSA(const uint8_t* src_argb, uint8_t* dst_y, int width);
+void ABGRToYRow_MSA(const uint8_t* src_argb, uint8_t* dst_y, int width);
+void RGBAToYRow_MSA(const uint8_t* src_argb, uint8_t* dst_y, int width);
+void RGB24ToYRow_MSA(const uint8_t* src_argb, uint8_t* dst_y, int width);
+void RAWToYRow_MSA(const uint8_t* src_argb, uint8_t* dst_y, int width);
+void RGB565ToYRow_MSA(const uint8_t* src_rgb565, uint8_t* dst_y, int width);
+void ARGB1555ToYRow_MSA(const uint8_t* src_argb1555, uint8_t* dst_y, int width);
+
+void BGRAToYRow_LSX(const uint8_t* src_bgra, uint8_t* dst_y, int width);
+void ABGRToYRow_LSX(const uint8_t* src_abgr, uint8_t* dst_y, int width);
+void RGBAToYRow_LSX(const uint8_t* src_rgba, uint8_t* dst_y, int width);
+void ARGB1555ToYRow_LSX(const uint8_t* src_argb1555, uint8_t* dst_y, int width);
+void RGB24ToYJRow_LSX(const uint8_t* src_rgb24, uint8_t* dst_yj, int width);
+void ABGRToYRow_LASX(const uint8_t* src_abgr, uint8_t* dst_y, int width);
+void ARGB1555ToYRow_LASX(const uint8_t* src_argb1555,
+                         uint8_t* dst_y,
+                         int width);
+void RGB565ToYRow_LSX(const uint8_t* src_rgb565, uint8_t* dst_y, int width);
+void RGB565ToYRow_LASX(const uint8_t* src_rgb565, uint8_t* dst_y, int width);
+void RGB24ToYRow_LSX(const uint8_t* src_rgb24, uint8_t* dst_y, int width);
+void RGB24ToYRow_LASX(const uint8_t* src_rgb24, uint8_t* dst_y, int width);
+void RAWToYRow_LSX(const uint8_t* src_raw, uint8_t* dst_y, int width);
+void RAWToYRow_LASX(const uint8_t* src_raw, uint8_t* dst_y, int width);
+void RGBAToYRow_LASX(const uint8_t* src_rgba, uint8_t* dst_y, int width);
+void BGRAToYRow_LASX(const uint8_t* src_bgra, uint8_t* dst_y, int width);
+void RGB24ToYJRow_LASX(const uint8_t* src_rgb24, uint8_t* dst_yj, int width);
+void RAWToYJRow_LSX(const uint8_t* src_raw, uint8_t* dst_yj, int width);
+void RAWToYJRow_LASX(const uint8_t* src_raw, uint8_t* dst_yj, int width);
+
+void ARGBToYRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
+void ARGBToYJRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
+void ABGRToYJRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
+void RGBAToYJRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
+void BGRAToYRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
+void ABGRToYRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
+void RGBAToYRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
+void RGB24ToYRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
+void RGB24ToYJRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
+void RAWToYRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
+void RAWToYJRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
+void RGB565ToYRow_C(const uint8_t* src_rgb565, uint8_t* dst_y, int width);
+void ARGB1555ToYRow_C(const uint8_t* src_argb1555, uint8_t* dst_y, int width);
+void ARGB4444ToYRow_C(const uint8_t* src_argb4444, uint8_t* dst_y, int width);
+void ARGBToYRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYJRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYJRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYJRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void BGRAToYRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB24ToYRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB24ToYJRow_Any_SSSE3(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void RAWToYRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToYJRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB24ToYJRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToYJRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int width);
+void ARGBToYJRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYJRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int width);
+void ABGRToYJRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYJRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int width);
+void RGBAToYJRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYJRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int width);
+void BGRAToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void BGRAToYRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int width);
+void ABGRToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int width);
+void RGBAToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int width);
+void RGB24ToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB24ToYJRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToYJRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB565ToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGB1555ToYRow_Any_NEON(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGB4444ToYRow_Any_NEON(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void BGRAToYRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYJRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB24ToYRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToYRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB565ToYRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGB1555ToYRow_Any_MSA(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+
+void BGRAToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYJRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB24ToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB565ToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYJRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYJRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB24ToYJRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToYJRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGB1555ToYRow_Any_LSX(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+
+void RGB565ToYRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB24ToYRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYJRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYJRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToYRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYJRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void BGRAToYRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB24ToYJRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToYJRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGB1555ToYRow_Any_LASX(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+
+void ARGBToUVRow_AVX2(const uint8_t* src_argb,
+                      int src_stride_argb,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void ABGRToUVRow_AVX2(const uint8_t* src_abgr,
+                      int src_stride_abgr,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void ARGBToUVJRow_AVX2(const uint8_t* src_argb,
+                       int src_stride_argb,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void ABGRToUVJRow_AVX2(const uint8_t* src_abgr,
+                       int src_stride_abgr,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void ARGBToUVRow_SSSE3(const uint8_t* src_argb,
+                       int src_stride_argb,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void ARGBToUVJRow_SSSE3(const uint8_t* src_argb,
+                        int src_stride_argb,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void ABGRToUVJRow_SSSE3(const uint8_t* src_abgr,
+                        int src_stride_abgr,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void BGRAToUVRow_SSSE3(const uint8_t* src_bgra,
+                       int src_stride_bgra,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void ABGRToUVRow_SSSE3(const uint8_t* src_abgr,
+                       int src_stride_abgr,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void RGBAToUVRow_SSSE3(const uint8_t* src_rgba,
+                       int src_stride_rgba,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void ARGBToUVRow_Any_AVX2(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void ABGRToUVRow_Any_AVX2(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void ARGBToUVJRow_Any_AVX2(const uint8_t* src_ptr,
+                           int src_stride,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void ABGRToUVJRow_Any_AVX2(const uint8_t* src_ptr,
+                           int src_stride,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void ARGBToUVRow_Any_SSSE3(const uint8_t* src_ptr,
+                           int src_stride,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void ARGBToUVJRow_Any_SSSE3(const uint8_t* src_ptr,
+                            int src_stride,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void ABGRToUVJRow_Any_SSSE3(const uint8_t* src_ptr,
+                            int src_stride,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void BGRAToUVRow_Any_SSSE3(const uint8_t* src_ptr,
+                           int src_stride,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void ABGRToUVRow_Any_SSSE3(const uint8_t* src_ptr,
+                           int src_stride,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void RGBAToUVRow_Any_SSSE3(const uint8_t* src_ptr,
+                           int src_stride,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void ARGBToUV444Row_Any_NEON(const uint8_t* src_ptr,
+                             uint8_t* dst_u,
+                             uint8_t* dst_v,
+                             int width);
+void ARGBToUV444Row_Any_NEON_I8MM(const uint8_t* src_ptr,
+                                  uint8_t* dst_u,
+                                  uint8_t* dst_v,
+                                  int width);
+void ARGBToUVRow_Any_NEON(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void ARGBToUVRow_Any_SVE2(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void ARGBToUV444Row_Any_MSA(const uint8_t* src_ptr,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void ARGBToUVRow_Any_MSA(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void ARGBToUVRow_Any_LSX(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void ARGBToUVRow_Any_LASX(const uint8_t* src_ptr,
+                          int src_stride_ptr,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void ARGBToUV444Row_Any_LSX(const uint8_t* src_ptr,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void ARGBToUV444Row_Any_LASX(const uint8_t* src_ptr,
+                             uint8_t* dst_u,
+                             uint8_t* dst_v,
+                             int width);
+void ARGBToUVJRow_Any_NEON(const uint8_t* src_ptr,
+                           int src_stride,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void ARGBToUVJRow_Any_SVE2(const uint8_t* src_ptr,
+                           int src_stride,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void ABGRToUVJRow_Any_NEON(const uint8_t* src_ptr,
+                           int src_stride,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void ABGRToUVJRow_Any_SVE2(const uint8_t* src_ptr,
+                           int src_stride,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void BGRAToUVRow_Any_NEON(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void BGRAToUVRow_Any_SVE2(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void ABGRToUVRow_Any_NEON(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void ABGRToUVRow_Any_SVE2(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void RGBAToUVRow_Any_NEON(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void RGBAToUVRow_Any_SVE2(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void RGB24ToUVRow_Any_NEON(const uint8_t* src_ptr,
+                           int src_stride,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void RAWToUVRow_Any_NEON(const uint8_t* src_ptr,
+                         int src_stride,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void RGB24ToUVJRow_Any_NEON(const uint8_t* src_ptr,
+                            int src_stride,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void RAWToUVJRow_Any_NEON(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void RGB565ToUVRow_Any_NEON(const uint8_t* src_ptr,
+                            int src_stride,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void ARGB1555ToUVRow_Any_NEON(const uint8_t* src_ptr,
+                              int src_stride,
+                              uint8_t* dst_u,
+                              uint8_t* dst_v,
+                              int width);
+void ARGB4444ToUVRow_Any_NEON(const uint8_t* src_ptr,
+                              int src_stride,
+                              uint8_t* dst_u,
+                              uint8_t* dst_v,
+                              int width);
+void ARGBToUVJRow_Any_MSA(const uint8_t* src_ptr,
+                          int src_stride_ptr,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void BGRAToUVRow_Any_MSA(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void ABGRToUVRow_Any_MSA(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void RGBAToUVRow_Any_MSA(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void RGB24ToUVRow_Any_MSA(const uint8_t* src_ptr,
+                          int src_stride_ptr,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void RAWToUVRow_Any_MSA(const uint8_t* src_ptr,
+                        int src_stride_ptr,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void RGB565ToUVRow_Any_MSA(const uint8_t* src_ptr,
+                           int src_stride_ptr,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void ARGB1555ToUVRow_Any_MSA(const uint8_t* src_ptr,
+                             int src_stride_ptr,
+                             uint8_t* dst_u,
+                             uint8_t* dst_v,
+                             int width);
+void ABGRToUVRow_Any_LSX(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void BGRAToUVRow_Any_LSX(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void RGBAToUVRow_Any_LSX(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void ARGBToUVJRow_Any_LSX(const uint8_t* src_ptr,
+                          int src_stride_ptr,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void ARGBToUVJRow_Any_LASX(const uint8_t* src_ptr,
+                           int src_stride_ptr,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void ARGB1555ToUVRow_Any_LSX(const uint8_t* src_ptr,
+                             int src_stride_ptr,
+                             uint8_t* dst_u,
+                             uint8_t* dst_v,
+                             int width);
+void ARGB1555ToUVRow_Any_LASX(const uint8_t* src_ptr,
+                              int src_stride_ptr,
+                              uint8_t* dst_u,
+                              uint8_t* dst_v,
+                              int width);
+void RGB565ToUVRow_Any_LSX(const uint8_t* src_ptr,
+                           int src_stride_ptr,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void RGB565ToUVRow_Any_LASX(const uint8_t* src_ptr,
+                            int src_stride_ptr,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void RGB24ToUVRow_Any_LSX(const uint8_t* src_ptr,
+                          int src_stride_ptr,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void RGB24ToUVRow_Any_LASX(const uint8_t* src_ptr,
+                           int src_stride_ptr,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void RAWToUVRow_Any_LSX(const uint8_t* src_ptr,
+                        int src_stride_ptr,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void RAWToUVRow_Any_LASX(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void ARGBToUVRow_C(const uint8_t* src_rgb,
+                   int src_stride_rgb,
+                   uint8_t* dst_u,
+                   uint8_t* dst_v,
+                   int width);
+void ARGBToUVJRow_C(const uint8_t* src_rgb,
+                    int src_stride_rgb,
+                    uint8_t* dst_u,
+                    uint8_t* dst_v,
+                    int width);
+void ABGRToUVJRow_C(const uint8_t* src_rgb,
+                    int src_stride_rgb,
+                    uint8_t* dst_u,
+                    uint8_t* dst_v,
+                    int width);
+void ARGBToUVRow_C(const uint8_t* src_rgb,
+                   int src_stride_rgb,
+                   uint8_t* dst_u,
+                   uint8_t* dst_v,
+                   int width);
+void BGRAToUVRow_C(const uint8_t* src_rgb,
+                   int src_stride_rgb,
+                   uint8_t* dst_u,
+                   uint8_t* dst_v,
+                   int width);
+void ABGRToUVRow_C(const uint8_t* src_rgb,
+                   int src_stride_rgb,
+                   uint8_t* dst_u,
+                   uint8_t* dst_v,
+                   int width);
+void RGBAToUVRow_C(const uint8_t* src_rgb,
+                   int src_stride_rgb,
+                   uint8_t* dst_u,
+                   uint8_t* dst_v,
+                   int width);
+void RGBAToUVJRow_C(const uint8_t* src_rgb,
+                    int src_stride_rgb,
+                    uint8_t* dst_u,
+                    uint8_t* dst_v,
+                    int width);
+void RGB24ToUVRow_C(const uint8_t* src_rgb,
+                    int src_stride_rgb,
+                    uint8_t* dst_u,
+                    uint8_t* dst_v,
+                    int width);
+void RAWToUVRow_C(const uint8_t* src_rgb,
+                  int src_stride_rgb,
+                  uint8_t* dst_u,
+                  uint8_t* dst_v,
+                  int width);
+void RGB24ToUVJRow_C(const uint8_t* src_rgb,
+                     int src_stride_rgb,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void RAWToUVJRow_C(const uint8_t* src_rgb,
+                   int src_stride_rgb,
+                   uint8_t* dst_u,
+                   uint8_t* dst_v,
+                   int width);
+void RGB565ToUVRow_C(const uint8_t* src_rgb565,
+                     int src_stride_rgb565,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void ARGB1555ToUVRow_C(const uint8_t* src_argb1555,
+                       int src_stride_argb1555,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+void ARGB4444ToUVRow_C(const uint8_t* src_argb4444,
+                       int src_stride_argb4444,
+                       uint8_t* dst_u,
+                       uint8_t* dst_v,
+                       int width);
+
+void ARGBToUV444Row_SSSE3(const uint8_t* src_argb,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void ARGBToUV444Row_Any_SSSE3(const uint8_t* src_ptr,
+                              uint8_t* dst_u,
+                              uint8_t* dst_v,
+                              int width);
+
+void ARGBToUV444Row_C(const uint8_t* src_argb,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+
+void MirrorRow_AVX2(const uint8_t* src, uint8_t* dst, int width);
+void MirrorRow_SSSE3(const uint8_t* src, uint8_t* dst, int width);
+void MirrorRow_NEON(const uint8_t* src, uint8_t* dst, int width);
+void MirrorRow_MSA(const uint8_t* src, uint8_t* dst, int width);
+void MirrorRow_LSX(const uint8_t* src, uint8_t* dst, int width);
+void MirrorRow_LASX(const uint8_t* src, uint8_t* dst, int width);
+void MirrorRow_C(const uint8_t* src, uint8_t* dst, int width);
+void MirrorRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void MirrorRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void MirrorRow_Any_SSE2(const uint8_t* src, uint8_t* dst, int width);
+void MirrorRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void MirrorRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void MirrorRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void MirrorRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void MirrorUVRow_AVX2(const uint8_t* src_uv, uint8_t* dst_uv, int width);
+void MirrorUVRow_SSSE3(const uint8_t* src_uv, uint8_t* dst_uv, int width);
+void MirrorUVRow_NEON(const uint8_t* src_uv, uint8_t* dst_uv, int width);
+void MirrorUVRow_MSA(const uint8_t* src_uv, uint8_t* dst_uv, int width);
+void MirrorUVRow_LSX(const uint8_t* src_uv, uint8_t* dst_uv, int width);
+void MirrorUVRow_LASX(const uint8_t* src_uv, uint8_t* dst_uv, int width);
+void MirrorUVRow_C(const uint8_t* src_uv, uint8_t* dst_uv, int width);
+void MirrorUVRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void MirrorUVRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void MirrorUVRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void MirrorUVRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void MirrorUVRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void MirrorUVRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+
+void MirrorSplitUVRow_SSSE3(const uint8_t* src,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void MirrorSplitUVRow_NEON(const uint8_t* src_uv,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void MirrorSplitUVRow_MSA(const uint8_t* src_uv,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void MirrorSplitUVRow_LSX(const uint8_t* src_uv,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void MirrorSplitUVRow_C(const uint8_t* src_uv,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+
+void MirrorRow_16_C(const uint16_t* src, uint16_t* dst, int width);
+
+void ARGBMirrorRow_AVX2(const uint8_t* src, uint8_t* dst, int width);
+void ARGBMirrorRow_SSE2(const uint8_t* src, uint8_t* dst, int width);
+void ARGBMirrorRow_NEON(const uint8_t* src_argb, uint8_t* dst_argb, int width);
+void ARGBMirrorRow_MSA(const uint8_t* src, uint8_t* dst, int width);
+void ARGBMirrorRow_LSX(const uint8_t* src, uint8_t* dst, int width);
+void ARGBMirrorRow_LASX(const uint8_t* src, uint8_t* dst, int width);
+void ARGBMirrorRow_C(const uint8_t* src, uint8_t* dst, int width);
+void ARGBMirrorRow_Any_AVX2(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void ARGBMirrorRow_Any_SSE2(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void ARGBMirrorRow_Any_NEON(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void ARGBMirrorRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBMirrorRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBMirrorRow_Any_LASX(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+
+void RGB24MirrorRow_SSSE3(const uint8_t* src_rgb24,
+                          uint8_t* dst_rgb24,
+                          int width);
+void RGB24MirrorRow_NEON(const uint8_t* src_rgb24,
+                         uint8_t* dst_rgb24,
+                         int width);
+void RGB24MirrorRow_C(const uint8_t* src_rgb24, uint8_t* dst_rgb24, int width);
+void RGB24MirrorRow_Any_SSSE3(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void RGB24MirrorRow_Any_NEON(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+
+void SplitUVRow_C(const uint8_t* src_uv,
+                  uint8_t* dst_u,
+                  uint8_t* dst_v,
+                  int width);
+void SplitUVRow_SSE2(const uint8_t* src_uv,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void SplitUVRow_AVX2(const uint8_t* src_uv,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void SplitUVRow_NEON(const uint8_t* src_uv,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void SplitUVRow_MSA(const uint8_t* src_uv,
+                    uint8_t* dst_u,
+                    uint8_t* dst_v,
+                    int width);
+void SplitUVRow_LSX(const uint8_t* src_uv,
+                    uint8_t* dst_u,
+                    uint8_t* dst_v,
+                    int width);
+void SplitUVRow_RVV(const uint8_t* src_uv,
+                    uint8_t* dst_u,
+                    uint8_t* dst_v,
+                    int width);
+void SplitUVRow_Any_SSE2(const uint8_t* src_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void SplitUVRow_Any_AVX2(const uint8_t* src_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void SplitUVRow_Any_NEON(const uint8_t* src_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void SplitUVRow_Any_MSA(const uint8_t* src_ptr,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void SplitUVRow_Any_LSX(const uint8_t* src_ptr,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void DetileRow_C(const uint8_t* src,
+                 ptrdiff_t src_tile_stride,
+                 uint8_t* dst,
+                 int width);
+void DetileRow_NEON(const uint8_t* src,
+                    ptrdiff_t src_tile_stride,
+                    uint8_t* dst,
+                    int width);
+void DetileRow_Any_NEON(const uint8_t* src,
+                        ptrdiff_t src_tile_stride,
+                        uint8_t* dst,
+                        int width);
+void DetileRow_SSE2(const uint8_t* src,
+                    ptrdiff_t src_tile_stride,
+                    uint8_t* dst,
+                    int width);
+void DetileRow_Any_SSE2(const uint8_t* src,
+                        ptrdiff_t src_tile_stride,
+                        uint8_t* dst,
+                        int width);
+void DetileRow_AVX(const uint8_t* src,
+                   ptrdiff_t src_tile_stride,
+                   uint8_t* dst,
+                   int width);
+void DetileRow_Any_AVX(const uint8_t* src,
+                       ptrdiff_t src_tile_stride,
+                       uint8_t* dst,
+                       int width);
+void DetileRow_16_C(const uint16_t* src,
+                    ptrdiff_t src_tile_stride,
+                    uint16_t* dst,
+                    int width);
+void DetileRow_16_NEON(const uint16_t* src,
+                       ptrdiff_t src_tile_stride,
+                       uint16_t* dst,
+                       int width);
+void DetileRow_16_Any_NEON(const uint16_t* src,
+                           ptrdiff_t src_tile_stride,
+                           uint16_t* dst,
+                           int width);
+void DetileRow_16_SSE2(const uint16_t* src,
+                       ptrdiff_t src_tile_stride,
+                       uint16_t* dst,
+                       int width);
+void DetileRow_16_Any_SSE2(const uint16_t* src,
+                           ptrdiff_t src_tile_stride,
+                           uint16_t* dst,
+                           int width);
+void DetileRow_16_AVX(const uint16_t* src,
+                      ptrdiff_t src_tile_stride,
+                      uint16_t* dst,
+                      int width);
+void DetileRow_16_Any_AVX(const uint16_t* src,
+                          ptrdiff_t src_tile_stride,
+                          uint16_t* dst,
+                          int width);
+void DetileSplitUVRow_C(const uint8_t* src_uv,
+                        ptrdiff_t src_tile_stride,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void DetileSplitUVRow_SSSE3(const uint8_t* src_uv,
+                            ptrdiff_t src_tile_stride,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void DetileSplitUVRow_Any_SSSE3(const uint8_t* src_uv,
+                                ptrdiff_t src_tile_stride,
+                                uint8_t* dst_u,
+                                uint8_t* dst_v,
+                                int width);
+void DetileSplitUVRow_NEON(const uint8_t* src_uv,
+                           ptrdiff_t src_tile_stride,
+                           uint8_t* dst_u,
+                           uint8_t* dst_v,
+                           int width);
+void DetileSplitUVRow_Any_NEON(const uint8_t* src_uv,
+                               ptrdiff_t src_tile_stride,
+                               uint8_t* dst_u,
+                               uint8_t* dst_v,
+                               int width);
+void DetileToYUY2_C(const uint8_t* src_y,
+                    ptrdiff_t src_y_tile_stride,
+                    const uint8_t* src_uv,
+                    ptrdiff_t src_uv_tile_stride,
+                    uint8_t* dst_yuy2,
+                    int width);
+void DetileToYUY2_SSE2(const uint8_t* src_y,
+                       ptrdiff_t src_y_tile_stride,
+                       const uint8_t* src_uv,
+                       ptrdiff_t src_uv_tile_stride,
+                       uint8_t* dst_yuy2,
+                       int width);
+void DetileToYUY2_Any_SSE2(const uint8_t* src_y,
+                           ptrdiff_t src_y_tile_stride,
+                           const uint8_t* src_uv,
+                           ptrdiff_t src_uv_tile_stride,
+                           uint8_t* dst_yuy2,
+                           int width);
+void DetileToYUY2_NEON(const uint8_t* src_y,
+                       ptrdiff_t src_y_tile_stride,
+                       const uint8_t* src_uv,
+                       ptrdiff_t src_uv_tile_stride,
+                       uint8_t* dst_yuy2,
+                       int width);
+void DetileToYUY2_Any_NEON(const uint8_t* src_y,
+                           ptrdiff_t src_y_tile_stride,
+                           const uint8_t* src_uv,
+                           ptrdiff_t src_uv_tile_stride,
+                           uint8_t* dst_yuy2,
+                           int width);
+void UnpackMT2T_C(const uint8_t* src, uint16_t* dst, size_t size);
+void UnpackMT2T_NEON(const uint8_t* src, uint16_t* dst, size_t size);
+void MergeUVRow_C(const uint8_t* src_u,
+                  const uint8_t* src_v,
+                  uint8_t* dst_uv,
+                  int width);
+void MergeUVRow_SSE2(const uint8_t* src_u,
+                     const uint8_t* src_v,
+                     uint8_t* dst_uv,
+                     int width);
+void MergeUVRow_AVX2(const uint8_t* src_u,
+                     const uint8_t* src_v,
+                     uint8_t* dst_uv,
+                     int width);
+void MergeUVRow_AVX512BW(const uint8_t* src_u,
+                         const uint8_t* src_v,
+                         uint8_t* dst_uv,
+                         int width);
+void MergeUVRow_NEON(const uint8_t* src_u,
+                     const uint8_t* src_v,
+                     uint8_t* dst_uv,
+                     int width);
+void MergeUVRow_MSA(const uint8_t* src_u,
+                    const uint8_t* src_v,
+                    uint8_t* dst_uv,
+                    int width);
+void MergeUVRow_LSX(const uint8_t* src_u,
+                    const uint8_t* src_v,
+                    uint8_t* dst_uv,
+                    int width);
+void MergeUVRow_RVV(const uint8_t* src_u,
+                    const uint8_t* src_v,
+                    uint8_t* dst_uv,
+                    int width);
+void MergeUVRow_Any_SSE2(const uint8_t* y_buf,
+                         const uint8_t* uv_buf,
+                         uint8_t* dst_ptr,
+                         int width);
+void MergeUVRow_Any_AVX2(const uint8_t* y_buf,
+                         const uint8_t* uv_buf,
+                         uint8_t* dst_ptr,
+                         int width);
+void MergeUVRow_Any_AVX512BW(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             int width);
+void MergeUVRow_Any_NEON(const uint8_t* y_buf,
+                         const uint8_t* uv_buf,
+                         uint8_t* dst_ptr,
+                         int width);
+void MergeUVRow_Any_MSA(const uint8_t* y_buf,
+                        const uint8_t* uv_buf,
+                        uint8_t* dst_ptr,
+                        int width);
+void MergeUVRow_Any_LSX(const uint8_t* y_buf,
+                        const uint8_t* uv_buf,
+                        uint8_t* dst_ptr,
+                        int width);
+
+void HalfMergeUVRow_C(const uint8_t* src_u,
+                      int src_stride_u,
+                      const uint8_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_uv,
+                      int width);
+
+void HalfMergeUVRow_NEON(const uint8_t* src_u,
+                         int src_stride_u,
+                         const uint8_t* src_v,
+                         int src_stride_v,
+                         uint8_t* dst_uv,
+                         int width);
+
+void HalfMergeUVRow_SSSE3(const uint8_t* src_u,
+                          int src_stride_u,
+                          const uint8_t* src_v,
+                          int src_stride_v,
+                          uint8_t* dst_uv,
+                          int width);
+
+void HalfMergeUVRow_AVX2(const uint8_t* src_u,
+                         int src_stride_u,
+                         const uint8_t* src_v,
+                         int src_stride_v,
+                         uint8_t* dst_uv,
+                         int width);
+
+void SplitRGBRow_C(const uint8_t* src_rgb,
+                   uint8_t* dst_r,
+                   uint8_t* dst_g,
+                   uint8_t* dst_b,
+                   int width);
+void SplitRGBRow_SSSE3(const uint8_t* src_rgb,
+                       uint8_t* dst_r,
+                       uint8_t* dst_g,
+                       uint8_t* dst_b,
+                       int width);
+void SplitRGBRow_NEON(const uint8_t* src_rgb,
+                      uint8_t* dst_r,
+                      uint8_t* dst_g,
+                      uint8_t* dst_b,
+                      int width);
+void SplitRGBRow_RVV(const uint8_t* src_rgb,
+                     uint8_t* dst_r,
+                     uint8_t* dst_g,
+                     uint8_t* dst_b,
+                     int width);
+void SplitRGBRow_Any_SSSE3(const uint8_t* src_ptr,
+                           uint8_t* dst_r,
+                           uint8_t* dst_g,
+                           uint8_t* dst_b,
+                           int width);
+void SplitRGBRow_Any_NEON(const uint8_t* src_ptr,
+                          uint8_t* dst_r,
+                          uint8_t* dst_g,
+                          uint8_t* dst_b,
+                          int width);
+
+void MergeRGBRow_C(const uint8_t* src_r,
+                   const uint8_t* src_g,
+                   const uint8_t* src_b,
+                   uint8_t* dst_rgb,
+                   int width);
+void MergeRGBRow_SSSE3(const uint8_t* src_r,
+                       const uint8_t* src_g,
+                       const uint8_t* src_b,
+                       uint8_t* dst_rgb,
+                       int width);
+void MergeRGBRow_NEON(const uint8_t* src_r,
+                      const uint8_t* src_g,
+                      const uint8_t* src_b,
+                      uint8_t* dst_rgb,
+                      int width);
+void MergeRGBRow_RVV(const uint8_t* src_r,
+                     const uint8_t* src_g,
+                     const uint8_t* src_b,
+                     uint8_t* dst_rgb,
+                     int width);
+void MergeRGBRow_Any_SSSE3(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           int width);
+void MergeRGBRow_Any_NEON(const uint8_t* src_r,
+                          const uint8_t* src_g,
+                          const uint8_t* src_b,
+                          uint8_t* dst_rgb,
+                          int width);
+void MergeARGBRow_C(const uint8_t* src_r,
+                    const uint8_t* src_g,
+                    const uint8_t* src_b,
+                    const uint8_t* src_a,
+                    uint8_t* dst_argb,
+                    int width);
+void MergeARGBRow_SSE2(const uint8_t* src_r,
+                       const uint8_t* src_g,
+                       const uint8_t* src_b,
+                       const uint8_t* src_a,
+                       uint8_t* dst_argb,
+                       int width);
+void MergeARGBRow_AVX2(const uint8_t* src_r,
+                       const uint8_t* src_g,
+                       const uint8_t* src_b,
+                       const uint8_t* src_a,
+                       uint8_t* dst_argb,
+                       int width);
+void MergeARGBRow_NEON(const uint8_t* src_r,
+                       const uint8_t* src_g,
+                       const uint8_t* src_b,
+                       const uint8_t* src_a,
+                       uint8_t* dst_argb,
+                       int width);
+void MergeARGBRow_RVV(const uint8_t* src_r,
+                      const uint8_t* src_g,
+                      const uint8_t* src_b,
+                      const uint8_t* src_a,
+                      uint8_t* dst_argb,
+                      int width);
+void MergeARGBRow_Any_SSE2(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           const uint8_t* a_buf,
+                           uint8_t* dst_ptr,
+                           int width);
+void MergeARGBRow_Any_AVX2(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           const uint8_t* a_buf,
+                           uint8_t* dst_ptr,
+                           int width);
+void MergeARGBRow_Any_NEON(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           const uint8_t* a_buf,
+                           uint8_t* dst_ptr,
+                           int width);
+void SplitARGBRow_C(const uint8_t* src_argb,
+                    uint8_t* dst_r,
+                    uint8_t* dst_g,
+                    uint8_t* dst_b,
+                    uint8_t* dst_a,
+                    int width);
+void SplitARGBRow_SSE2(const uint8_t* src_argb,
+                       uint8_t* dst_r,
+                       uint8_t* dst_g,
+                       uint8_t* dst_b,
+                       uint8_t* dst_a,
+                       int width);
+void SplitARGBRow_SSSE3(const uint8_t* src_argb,
+                        uint8_t* dst_r,
+                        uint8_t* dst_g,
+                        uint8_t* dst_b,
+                        uint8_t* dst_a,
+                        int width);
+void SplitARGBRow_AVX2(const uint8_t* src_argb,
+                       uint8_t* dst_r,
+                       uint8_t* dst_g,
+                       uint8_t* dst_b,
+                       uint8_t* dst_a,
+                       int width);
+void SplitARGBRow_NEON(const uint8_t* src_rgba,
+                       uint8_t* dst_r,
+                       uint8_t* dst_g,
+                       uint8_t* dst_b,
+                       uint8_t* dst_a,
+                       int width);
+void SplitARGBRow_RVV(const uint8_t* src_rgba,
+                      uint8_t* dst_r,
+                      uint8_t* dst_g,
+                      uint8_t* dst_b,
+                      uint8_t* dst_a,
+                      int width);
+void SplitARGBRow_Any_SSE2(const uint8_t* src_ptr,
+                           uint8_t* dst_r,
+                           uint8_t* dst_g,
+                           uint8_t* dst_b,
+                           uint8_t* dst_a,
+                           int width);
+void SplitARGBRow_Any_SSSE3(const uint8_t* src_ptr,
+                            uint8_t* dst_r,
+                            uint8_t* dst_g,
+                            uint8_t* dst_b,
+                            uint8_t* dst_a,
+                            int width);
+void SplitARGBRow_Any_AVX2(const uint8_t* src_ptr,
+                           uint8_t* dst_r,
+                           uint8_t* dst_g,
+                           uint8_t* dst_b,
+                           uint8_t* dst_a,
+                           int width);
+void SplitARGBRow_Any_NEON(const uint8_t* src_ptr,
+                           uint8_t* dst_r,
+                           uint8_t* dst_g,
+                           uint8_t* dst_b,
+                           uint8_t* dst_a,
+                           int width);
+void MergeXRGBRow_C(const uint8_t* src_r,
+                    const uint8_t* src_g,
+                    const uint8_t* src_b,
+                    uint8_t* dst_argb,
+                    int width);
+void MergeXRGBRow_SSE2(const uint8_t* src_r,
+                       const uint8_t* src_g,
+                       const uint8_t* src_b,
+                       uint8_t* dst_argb,
+                       int width);
+void MergeXRGBRow_AVX2(const uint8_t* src_r,
+                       const uint8_t* src_g,
+                       const uint8_t* src_b,
+                       uint8_t* dst_argb,
+                       int width);
+void MergeXRGBRow_NEON(const uint8_t* src_r,
+                       const uint8_t* src_g,
+                       const uint8_t* src_b,
+                       uint8_t* dst_argb,
+                       int width);
+void MergeXRGBRow_RVV(const uint8_t* src_r,
+                      const uint8_t* src_g,
+                      const uint8_t* src_b,
+                      uint8_t* dst_argb,
+                      int width);
+void MergeXRGBRow_Any_SSE2(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           int width);
+void MergeXRGBRow_Any_AVX2(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           int width);
+void MergeXRGBRow_Any_NEON(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           int width);
+void SplitXRGBRow_C(const uint8_t* src_argb,
+                    uint8_t* dst_r,
+                    uint8_t* dst_g,
+                    uint8_t* dst_b,
+                    int width);
+void SplitXRGBRow_SSE2(const uint8_t* src_argb,
+                       uint8_t* dst_r,
+                       uint8_t* dst_g,
+                       uint8_t* dst_b,
+                       int width);
+void SplitXRGBRow_SSSE3(const uint8_t* src_argb,
+                        uint8_t* dst_r,
+                        uint8_t* dst_g,
+                        uint8_t* dst_b,
+                        int width);
+void SplitXRGBRow_AVX2(const uint8_t* src_argb,
+                       uint8_t* dst_r,
+                       uint8_t* dst_g,
+                       uint8_t* dst_b,
+                       int width);
+void SplitXRGBRow_NEON(const uint8_t* src_rgba,
+                       uint8_t* dst_r,
+                       uint8_t* dst_g,
+                       uint8_t* dst_b,
+                       int width);
+void SplitXRGBRow_RVV(const uint8_t* src_rgba,
+                      uint8_t* dst_r,
+                      uint8_t* dst_g,
+                      uint8_t* dst_b,
+                      int width);
+void SplitXRGBRow_Any_SSE2(const uint8_t* src_ptr,
+                           uint8_t* dst_r,
+                           uint8_t* dst_g,
+                           uint8_t* dst_b,
+                           int width);
+void SplitXRGBRow_Any_SSSE3(const uint8_t* src_ptr,
+                            uint8_t* dst_r,
+                            uint8_t* dst_g,
+                            uint8_t* dst_b,
+                            int width);
+void SplitXRGBRow_Any_AVX2(const uint8_t* src_ptr,
+                           uint8_t* dst_r,
+                           uint8_t* dst_g,
+                           uint8_t* dst_b,
+                           int width);
+void SplitXRGBRow_Any_NEON(const uint8_t* src_ptr,
+                           uint8_t* dst_r,
+                           uint8_t* dst_g,
+                           uint8_t* dst_b,
+                           int width);
+
+void MergeXR30Row_C(const uint16_t* src_r,
+                    const uint16_t* src_g,
+                    const uint16_t* src_b,
+                    uint8_t* dst_ar30,
+                    int depth,
+                    int width);
+void MergeAR64Row_C(const uint16_t* src_r,
+                    const uint16_t* src_g,
+                    const uint16_t* src_b,
+                    const uint16_t* src_a,
+                    uint16_t* dst_ar64,
+                    int depth,
+                    int width);
+void MergeARGB16To8Row_C(const uint16_t* src_r,
+                         const uint16_t* src_g,
+                         const uint16_t* src_b,
+                         const uint16_t* src_a,
+                         uint8_t* dst_argb,
+                         int depth,
+                         int width);
+void MergeXR64Row_C(const uint16_t* src_r,
+                    const uint16_t* src_g,
+                    const uint16_t* src_b,
+                    uint16_t* dst_ar64,
+                    int depth,
+                    int width);
+void MergeXRGB16To8Row_C(const uint16_t* src_r,
+                         const uint16_t* src_g,
+                         const uint16_t* src_b,
+                         uint8_t* dst_argb,
+                         int depth,
+                         int width);
+void MergeXR30Row_AVX2(const uint16_t* src_r,
+                       const uint16_t* src_g,
+                       const uint16_t* src_b,
+                       uint8_t* dst_ar30,
+                       int depth,
+                       int width);
+void MergeAR64Row_AVX2(const uint16_t* src_r,
+                       const uint16_t* src_g,
+                       const uint16_t* src_b,
+                       const uint16_t* src_a,
+                       uint16_t* dst_ar64,
+                       int depth,
+                       int width);
+void MergeARGB16To8Row_AVX2(const uint16_t* src_r,
+                            const uint16_t* src_g,
+                            const uint16_t* src_b,
+                            const uint16_t* src_a,
+                            uint8_t* dst_argb,
+                            int depth,
+                            int width);
+void MergeXR64Row_AVX2(const uint16_t* src_r,
+                       const uint16_t* src_g,
+                       const uint16_t* src_b,
+                       uint16_t* dst_ar64,
+                       int depth,
+                       int width);
+void MergeXRGB16To8Row_AVX2(const uint16_t* src_r,
+                            const uint16_t* src_g,
+                            const uint16_t* src_b,
+                            uint8_t* dst_argb,
+                            int depth,
+                            int width);
+void MergeXR30Row_NEON(const uint16_t* src_r,
+                       const uint16_t* src_g,
+                       const uint16_t* src_b,
+                       uint8_t* dst_ar30,
+                       int depth,
+                       int width);
+void MergeXR30Row_10_NEON(const uint16_t* src_r,
+                          const uint16_t* src_g,
+                          const uint16_t* src_b,
+                          uint8_t* dst_ar30,
+                          int /* depth */,
+                          int width);
+void MergeAR64Row_NEON(const uint16_t* src_r,
+                       const uint16_t* src_g,
+                       const uint16_t* src_b,
+                       const uint16_t* src_a,
+                       uint16_t* dst_ar64,
+                       int depth,
+                       int width);
+void MergeARGB16To8Row_NEON(const uint16_t* src_r,
+                            const uint16_t* src_g,
+                            const uint16_t* src_b,
+                            const uint16_t* src_a,
+                            uint8_t* dst_argb,
+                            int depth,
+                            int width);
+void MergeXR64Row_NEON(const uint16_t* src_r,
+                       const uint16_t* src_g,
+                       const uint16_t* src_b,
+                       uint16_t* dst_ar64,
+                       int depth,
+                       int width);
+void MergeXRGB16To8Row_NEON(const uint16_t* src_r,
+                            const uint16_t* src_g,
+                            const uint16_t* src_b,
+                            uint8_t* dst_argb,
+                            int depth,
+                            int width);
+void MergeXR30Row_Any_AVX2(const uint16_t* r_buf,
+                           const uint16_t* g_buf,
+                           const uint16_t* b_buf,
+                           uint8_t* dst_ptr,
+                           int depth,
+                           int width);
+void MergeAR64Row_Any_AVX2(const uint16_t* r_buf,
+                           const uint16_t* g_buf,
+                           const uint16_t* b_buf,
+                           const uint16_t* a_buf,
+                           uint16_t* dst_ptr,
+                           int depth,
+                           int width);
+void MergeXR64Row_Any_AVX2(const uint16_t* r_buf,
+                           const uint16_t* g_buf,
+                           const uint16_t* b_buf,
+                           uint16_t* dst_ptr,
+                           int depth,
+                           int width);
+void MergeARGB16To8Row_Any_AVX2(const uint16_t* r_buf,
+                                const uint16_t* g_buf,
+                                const uint16_t* b_buf,
+                                const uint16_t* a_buf,
+                                uint8_t* dst_ptr,
+                                int depth,
+                                int width);
+void MergeXRGB16To8Row_Any_AVX2(const uint16_t* r_buf,
+                                const uint16_t* g_buf,
+                                const uint16_t* b_buf,
+                                uint8_t* dst_ptr,
+                                int depth,
+                                int width);
+void MergeXR30Row_Any_NEON(const uint16_t* r_buf,
+                           const uint16_t* g_buf,
+                           const uint16_t* b_buf,
+                           uint8_t* dst_ptr,
+                           int depth,
+                           int width);
+void MergeXR30Row_10_Any_NEON(const uint16_t* r_buf,
+                              const uint16_t* g_buf,
+                              const uint16_t* b_buf,
+                              uint8_t* dst_ptr,
+                              int depth,
+                              int width);
+void MergeAR64Row_Any_NEON(const uint16_t* r_buf,
+                           const uint16_t* g_buf,
+                           const uint16_t* b_buf,
+                           const uint16_t* a_buf,
+                           uint16_t* dst_ptr,
+                           int depth,
+                           int width);
+void MergeARGB16To8Row_Any_NEON(const uint16_t* r_buf,
+                                const uint16_t* g_buf,
+                                const uint16_t* b_buf,
+                                const uint16_t* a_buf,
+                                uint8_t* dst_ptr,
+                                int depth,
+                                int width);
+void MergeXR64Row_Any_NEON(const uint16_t* r_buf,
+                           const uint16_t* g_buf,
+                           const uint16_t* b_buf,
+                           uint16_t* dst_ptr,
+                           int depth,
+                           int width);
+void MergeXRGB16To8Row_Any_NEON(const uint16_t* r_buf,
+                                const uint16_t* g_buf,
+                                const uint16_t* b_buf,
+                                uint8_t* dst_ptr,
+                                int depth,
+                                int width);
+
+void MergeUVRow_16_C(const uint16_t* src_u,
+                     const uint16_t* src_v,
+                     uint16_t* dst_uv,
+                     int depth,
+                     int width);
+void MergeUVRow_16_AVX2(const uint16_t* src_u,
+                        const uint16_t* src_v,
+                        uint16_t* dst_uv,
+                        int depth,
+                        int width);
+void MergeUVRow_16_Any_AVX2(const uint16_t* src_u,
+                            const uint16_t* src_v,
+                            uint16_t* dst_uv,
+                            int depth,
+                            int width);
+void MergeUVRow_16_NEON(const uint16_t* src_u,
+                        const uint16_t* src_v,
+                        uint16_t* dst_uv,
+                        int depth,
+                        int width);
+void MergeUVRow_16_Any_NEON(const uint16_t* src_u,
+                            const uint16_t* src_v,
+                            uint16_t* dst_uv,
+                            int depth,
+                            int width);
+
+void SplitUVRow_16_C(const uint16_t* src_uv,
+                     uint16_t* dst_u,
+                     uint16_t* dst_v,
+                     int depth,
+                     int width);
+void SplitUVRow_16_AVX2(const uint16_t* src_uv,
+                        uint16_t* dst_u,
+                        uint16_t* dst_v,
+                        int depth,
+                        int width);
+void SplitUVRow_16_Any_AVX2(const uint16_t* src_uv,
+                            uint16_t* dst_u,
+                            uint16_t* dst_v,
+                            int depth,
+                            int width);
+void SplitUVRow_16_NEON(const uint16_t* src_uv,
+                        uint16_t* dst_u,
+                        uint16_t* dst_v,
+                        int depth,
+                        int width);
+void SplitUVRow_16_Any_NEON(const uint16_t* src_uv,
+                            uint16_t* dst_u,
+                            uint16_t* dst_v,
+                            int depth,
+                            int width);
+
+void MultiplyRow_16_C(const uint16_t* src_y,
+                      uint16_t* dst_y,
+                      int scale,
+                      int width);
+void MultiplyRow_16_AVX2(const uint16_t* src_y,
+                         uint16_t* dst_y,
+                         int scale,
+                         int width);
+void MultiplyRow_16_Any_AVX2(const uint16_t* src_ptr,
+                             uint16_t* dst_ptr,
+                             int scale,
+                             int width);
+void MultiplyRow_16_NEON(const uint16_t* src_y,
+                         uint16_t* dst_y,
+                         int scale,
+                         int width);
+void MultiplyRow_16_Any_NEON(const uint16_t* src_ptr,
+                             uint16_t* dst_ptr,
+                             int scale,
+                             int width);
+
+void DivideRow_16_C(const uint16_t* src_y,
+                    uint16_t* dst_y,
+                    int scale,
+                    int width);
+void DivideRow_16_AVX2(const uint16_t* src_y,
+                       uint16_t* dst_y,
+                       int scale,
+                       int width);
+void DivideRow_16_Any_AVX2(const uint16_t* src_ptr,
+                           uint16_t* dst_ptr,
+                           int scale,
+                           int width);
+void DivideRow_16_NEON(const uint16_t* src_y,
+                       uint16_t* dst_y,
+                       int scale,
+                       int width);
+void DivideRow_16_Any_NEON(const uint16_t* src_ptr,
+                           uint16_t* dst_ptr,
+                           int scale,
+                           int width);
+
+void Convert8To16Row_C(const uint8_t* src_y,
+                       uint16_t* dst_y,
+                       int scale,
+                       int width);
+void Convert8To16Row_SSE2(const uint8_t* src_y,
+                          uint16_t* dst_y,
+                          int scale,
+                          int width);
+void Convert8To16Row_AVX2(const uint8_t* src_y,
+                          uint16_t* dst_y,
+                          int scale,
+                          int width);
+void Convert8To16Row_Any_SSE2(const uint8_t* src_ptr,
+                              uint16_t* dst_ptr,
+                              int scale,
+                              int width);
+void Convert8To16Row_Any_AVX2(const uint8_t* src_ptr,
+                              uint16_t* dst_ptr,
+                              int scale,
+                              int width);
+
+void Convert16To8Row_C(const uint16_t* src_y,
+                       uint8_t* dst_y,
+                       int scale,
+                       int width);
+void Convert16To8Row_SSSE3(const uint16_t* src_y,
+                           uint8_t* dst_y,
+                           int scale,
+                           int width);
+void Convert16To8Row_AVX2(const uint16_t* src_y,
+                          uint8_t* dst_y,
+                          int scale,
+                          int width);
+void Convert16To8Row_AVX512BW(const uint16_t* src_y,
+                             uint8_t* dst_y,
+                             int scale,
+                             int width);
+void Convert16To8Row_Any_SSSE3(const uint16_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int scale,
+                               int width);
+void Convert16To8Row_Any_AVX2(const uint16_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int scale,
+                              int width);
+void Convert16To8Row_Any_AVX512BW(const uint16_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int scale,
+                                  int width);
+void Convert16To8Row_NEON(const uint16_t* src_y,
+                          uint8_t* dst_y,
+                          int scale,
+                          int width);
+void Convert16To8Row_Any_NEON(const uint16_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int scale,
+                              int width);
+
+void CopyRow_SSE2(const uint8_t* src, uint8_t* dst, int width);
+void CopyRow_AVX(const uint8_t* src, uint8_t* dst, int width);
+void CopyRow_ERMS(const uint8_t* src, uint8_t* dst, int width);
+void CopyRow_NEON(const uint8_t* src, uint8_t* dst, int width);
+void CopyRow_MIPS(const uint8_t* src, uint8_t* dst, int count);
+void CopyRow_RVV(const uint8_t* src, uint8_t* dst, int count);
+void CopyRow_C(const uint8_t* src, uint8_t* dst, int count);
+void CopyRow_Any_SSE2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void CopyRow_Any_AVX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void CopyRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+
+void CopyRow_16_C(const uint16_t* src, uint16_t* dst, int count);
+
+void ARGBCopyAlphaRow_C(const uint8_t* src, uint8_t* dst, int width);
+void ARGBCopyAlphaRow_SSE2(const uint8_t* src, uint8_t* dst, int width);
+void ARGBCopyAlphaRow_AVX2(const uint8_t* src, uint8_t* dst, int width);
+void ARGBCopyAlphaRow_Any_SSE2(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+void ARGBCopyAlphaRow_Any_AVX2(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+
+void ARGBExtractAlphaRow_C(const uint8_t* src_argb, uint8_t* dst_a, int width);
+void ARGBExtractAlphaRow_SSE2(const uint8_t* src_argb,
+                              uint8_t* dst_a,
+                              int width);
+void ARGBExtractAlphaRow_AVX2(const uint8_t* src_argb,
+                              uint8_t* dst_a,
+                              int width);
+void ARGBExtractAlphaRow_NEON(const uint8_t* src_argb,
+                              uint8_t* dst_a,
+                              int width);
+void ARGBExtractAlphaRow_MSA(const uint8_t* src_argb,
+                             uint8_t* dst_a,
+                             int width);
+void ARGBExtractAlphaRow_LSX(const uint8_t* src_argb,
+                             uint8_t* dst_a,
+                             int width);
+void ARGBExtractAlphaRow_RVV(const uint8_t* src_argb,
+                             uint8_t* dst_a,
+                             int width);
+void ARGBExtractAlphaRow_Any_SSE2(const uint8_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int width);
+void ARGBExtractAlphaRow_Any_AVX2(const uint8_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int width);
+void ARGBExtractAlphaRow_Any_NEON(const uint8_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int width);
+void ARGBExtractAlphaRow_Any_MSA(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int width);
+void ARGBExtractAlphaRow_Any_LSX(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int width);
+
+void ARGBCopyYToAlphaRow_C(const uint8_t* src, uint8_t* dst, int width);
+void ARGBCopyYToAlphaRow_SSE2(const uint8_t* src, uint8_t* dst, int width);
+void ARGBCopyYToAlphaRow_AVX2(const uint8_t* src, uint8_t* dst, int width);
+void ARGBCopyYToAlphaRow_RVV(const uint8_t* src, uint8_t* dst, int width);
+void ARGBCopyYToAlphaRow_Any_SSE2(const uint8_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int width);
+void ARGBCopyYToAlphaRow_Any_AVX2(const uint8_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int width);
+
+void SetRow_C(uint8_t* dst, uint8_t v8, int width);
+void SetRow_MSA(uint8_t* dst, uint8_t v8, int width);
+void SetRow_X86(uint8_t* dst, uint8_t v8, int width);
+void SetRow_ERMS(uint8_t* dst, uint8_t v8, int width);
+void SetRow_NEON(uint8_t* dst, uint8_t v8, int width);
+void SetRow_LSX(uint8_t* dst, uint8_t v8, int width);
+void SetRow_Any_X86(uint8_t* dst_ptr, uint8_t v32, int width);
+void SetRow_Any_NEON(uint8_t* dst_ptr, uint8_t v32, int width);
+void SetRow_Any_LSX(uint8_t* dst_ptr, uint8_t v32, int width);
+
+void ARGBSetRow_C(uint8_t* dst_argb, uint32_t v32, int width);
+void ARGBSetRow_X86(uint8_t* dst_argb, uint32_t v32, int width);
+void ARGBSetRow_NEON(uint8_t* dst, uint32_t v32, int width);
+void ARGBSetRow_Any_NEON(uint8_t* dst_ptr, uint32_t v32, int width);
+void ARGBSetRow_MSA(uint8_t* dst_argb, uint32_t v32, int width);
+void ARGBSetRow_Any_MSA(uint8_t* dst_ptr, uint32_t v32, int width);
+void ARGBSetRow_LSX(uint8_t* dst_argb, uint32_t v32, int width);
+void ARGBSetRow_Any_LSX(uint8_t* dst_ptr, uint32_t v32, int width);
+
+// ARGBShufflers for BGRAToARGB etc.
+void ARGBShuffleRow_C(const uint8_t* src_argb,
+                      uint8_t* dst_argb,
+                      const uint8_t* shuffler,
+                      int width);
+void ARGBShuffleRow_SSSE3(const uint8_t* src_argb,
+                          uint8_t* dst_argb,
+                          const uint8_t* shuffler,
+                          int width);
+void ARGBShuffleRow_AVX2(const uint8_t* src_argb,
+                         uint8_t* dst_argb,
+                         const uint8_t* shuffler,
+                         int width);
+void ARGBShuffleRow_NEON(const uint8_t* src_argb,
+                         uint8_t* dst_argb,
+                         const uint8_t* shuffler,
+                         int width);
+void ARGBShuffleRow_MSA(const uint8_t* src_argb,
+                        uint8_t* dst_argb,
+                        const uint8_t* shuffler,
+                        int width);
+void ARGBShuffleRow_LSX(const uint8_t* src_argb,
+                        uint8_t* dst_argb,
+                        const uint8_t* shuffler,
+                        int width);
+void ARGBShuffleRow_LASX(const uint8_t* src_argb,
+                         uint8_t* dst_argb,
+                         const uint8_t* shuffler,
+                         int width);
+void ARGBShuffleRow_Any_SSSE3(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              const uint8_t* param,
+                              int width);
+void ARGBShuffleRow_Any_AVX2(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             const uint8_t* param,
+                             int width);
+void ARGBShuffleRow_Any_NEON(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             const uint8_t* param,
+                             int width);
+void ARGBShuffleRow_Any_MSA(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            const uint8_t* param,
+                            int width);
+void ARGBShuffleRow_Any_LSX(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            const uint8_t* param,
+                            int width);
+void ARGBShuffleRow_Any_LASX(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             const uint8_t* param,
+                             int width);
+
+void RGB24ToARGBRow_SSSE3(const uint8_t* src_rgb24,
+                          uint8_t* dst_argb,
+                          int width);
+void RAWToARGBRow_SSSE3(const uint8_t* src_raw, uint8_t* dst_argb, int width);
+void RAWToRGBARow_SSSE3(const uint8_t* src_raw, uint8_t* dst_rgba, int width);
+void RAWToRGB24Row_SSSE3(const uint8_t* src_raw, uint8_t* dst_rgb24, int width);
+void RGB565ToARGBRow_SSE2(const uint8_t* src, uint8_t* dst, int width);
+void ARGB1555ToARGBRow_SSE2(const uint8_t* src, uint8_t* dst, int width);
+void ARGB4444ToARGBRow_SSE2(const uint8_t* src, uint8_t* dst, int width);
+void RGB565ToARGBRow_AVX2(const uint8_t* src_rgb565,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGB1555ToARGBRow_AVX2(const uint8_t* src_argb1555,
+                            uint8_t* dst_argb,
+                            int width);
+void ARGB4444ToARGBRow_AVX2(const uint8_t* src_argb4444,
+                            uint8_t* dst_argb,
+                            int width);
+
+void RGB24ToARGBRow_NEON(const uint8_t* src_rgb24,
+                         uint8_t* dst_argb,
+                         int width);
+void RGB24ToARGBRow_SVE2(const uint8_t* src_rgb24,
+                         uint8_t* dst_argb,
+                         int width);
+void RGB24ToARGBRow_MSA(const uint8_t* src_rgb24, uint8_t* dst_argb, int width);
+void RGB24ToARGBRow_LSX(const uint8_t* src_rgb24, uint8_t* dst_argb, int width);
+void RGB24ToARGBRow_LASX(const uint8_t* src_rgb24,
+                         uint8_t* dst_argb,
+                         int width);
+void RGB24ToARGBRow_RVV(const uint8_t* src_rgb24, uint8_t* dst_argb, int width);
+void RAWToARGBRow_NEON(const uint8_t* src_raw, uint8_t* dst_argb, int width);
+void RAWToARGBRow_SVE2(const uint8_t* src_raw, uint8_t* dst_argb, int width);
+void RAWToRGBARow_NEON(const uint8_t* src_raw, uint8_t* dst_rgba, int width);
+void RAWToRGBARow_SVE2(const uint8_t* src_raw, uint8_t* dst_rgba, int width);
+void RAWToARGBRow_MSA(const uint8_t* src_raw, uint8_t* dst_argb, int width);
+void RAWToARGBRow_LSX(const uint8_t* src_raw, uint8_t* dst_argb, int width);
+void RAWToARGBRow_LASX(const uint8_t* src_raw, uint8_t* dst_argb, int width);
+void RAWToARGBRow_RVV(const uint8_t* src_raw, uint8_t* dst_argb, int width);
+void RAWToRGBARow_RVV(const uint8_t* src_raw, uint8_t* dst_rgba, int width);
+void RAWToRGB24Row_NEON(const uint8_t* src_raw, uint8_t* dst_rgb24, int width);
+void RAWToRGB24Row_SVE2(const uint8_t* src_raw, uint8_t* dst_rgb24, int width);
+void RAWToRGB24Row_MSA(const uint8_t* src_raw, uint8_t* dst_rgb24, int width);
+void RAWToRGB24Row_LSX(const uint8_t* src_raw, uint8_t* dst_rgb24, int width);
+void RAWToRGB24Row_RVV(const uint8_t* src_raw, uint8_t* dst_rgb24, int width);
+void RGB565ToARGBRow_NEON(const uint8_t* src_rgb565,
+                          uint8_t* dst_argb,
+                          int width);
+void RGB565ToARGBRow_MSA(const uint8_t* src_rgb565,
+                         uint8_t* dst_argb,
+                         int width);
+void RGB565ToARGBRow_LSX(const uint8_t* src_rgb565,
+                         uint8_t* dst_argb,
+                         int width);
+void RGB565ToARGBRow_LASX(const uint8_t* src_rgb565,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGB1555ToARGBRow_NEON(const uint8_t* src_argb1555,
+                            uint8_t* dst_argb,
+                            int width);
+void ARGB1555ToARGBRow_SVE2(const uint8_t* src_argb1555,
+                            uint8_t* dst_argb,
+                            int width);
+void ARGB1555ToARGBRow_MSA(const uint8_t* src_argb1555,
+                           uint8_t* dst_argb,
+                           int width);
+void ARGB1555ToARGBRow_LSX(const uint8_t* src_argb1555,
+                           uint8_t* dst_argb,
+                           int width);
+void ARGB1555ToARGBRow_LASX(const uint8_t* src_argb1555,
+                            uint8_t* dst_argb,
+                            int width);
+void ARGB4444ToARGBRow_NEON(const uint8_t* src_argb4444,
+                            uint8_t* dst_argb,
+                            int width);
+void ARGB4444ToARGBRow_MSA(const uint8_t* src_argb4444,
+                           uint8_t* dst_argb,
+                           int width);
+void ARGB4444ToARGBRow_LSX(const uint8_t* src_argb4444,
+                           uint8_t* dst_argb,
+                           int width);
+void ARGB4444ToARGBRow_LASX(const uint8_t* src_argb4444,
+                            uint8_t* dst_argb,
+                            int width);
+void RGB24ToARGBRow_C(const uint8_t* src_rgb24, uint8_t* dst_argb, int width);
+void RAWToARGBRow_C(const uint8_t* src_raw, uint8_t* dst_argb, int width);
+void RAWToRGBARow_C(const uint8_t* src_raw, uint8_t* dst_rgba, int width);
+void RAWToRGB24Row_C(const uint8_t* src_raw, uint8_t* dst_rgb24, int width);
+void RGB565ToARGBRow_C(const uint8_t* src_rgb565, uint8_t* dst_argb, int width);
+void ARGB1555ToARGBRow_C(const uint8_t* src_argb1555,
+                         uint8_t* dst_argb,
+                         int width);
+void ARGB4444ToARGBRow_C(const uint8_t* src_argb4444,
+                         uint8_t* dst_argb,
+                         int width);
+void AR30ToARGBRow_C(const uint8_t* src_ar30, uint8_t* dst_argb, int width);
+void AR30ToABGRRow_C(const uint8_t* src_ar30, uint8_t* dst_abgr, int width);
+void ARGBToAR30Row_C(const uint8_t* src_argb, uint8_t* dst_ar30, int width);
+void AR30ToAB30Row_C(const uint8_t* src_ar30, uint8_t* dst_ab30, int width);
+
+void RGB24ToARGBRow_Any_SSSE3(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void RAWToARGBRow_Any_SSSE3(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void RAWToRGBARow_Any_SSSE3(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void RAWToRGB24Row_Any_SSSE3(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+
+void RGB565ToARGBRow_Any_SSE2(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGB1555ToARGBRow_Any_SSE2(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void ARGB4444ToARGBRow_Any_SSE2(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void RGB565ToARGBRow_Any_AVX2(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGB1555ToARGBRow_Any_AVX2(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void ARGB4444ToARGBRow_Any_AVX2(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+
+void RGB24ToARGBRow_Any_NEON(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void RGB24ToARGBRow_Any_MSA(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void RGB24ToARGBRow_Any_LSX(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void RGB24ToARGBRow_Any_LASX(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void RAWToARGBRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToRGBARow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToARGBRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToARGBRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToARGBRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToRGB24Row_Any_NEON(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void RAWToRGB24Row_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RAWToRGB24Row_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGB565ToARGBRow_Any_NEON(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void RGB565ToARGBRow_Any_MSA(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void RGB565ToARGBRow_Any_LSX(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void RGB565ToARGBRow_Any_LASX(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGB1555ToARGBRow_Any_NEON(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void ARGB1555ToARGBRow_Any_MSA(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+void ARGB4444ToARGBRow_Any_NEON(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void ARGB1555ToARGBRow_Any_LSX(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+void ARGB1555ToARGBRow_Any_LASX(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+
+void ARGB4444ToARGBRow_Any_MSA(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+void ARGB4444ToARGBRow_Any_LSX(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+void ARGB4444ToARGBRow_Any_LASX(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+
+void ARGBToRGB24Row_SSSE3(const uint8_t* src, uint8_t* dst, int width);
+void ARGBToRAWRow_SSSE3(const uint8_t* src, uint8_t* dst, int width);
+void ARGBToRGB565Row_SSE2(const uint8_t* src, uint8_t* dst, int width);
+void ARGBToARGB1555Row_SSE2(const uint8_t* src, uint8_t* dst, int width);
+void ARGBToARGB4444Row_SSE2(const uint8_t* src, uint8_t* dst, int width);
+void ABGRToAR30Row_SSSE3(const uint8_t* src, uint8_t* dst, int width);
+void ARGBToAR30Row_SSSE3(const uint8_t* src, uint8_t* dst, int width);
+
+void ARGBToRAWRow_AVX2(const uint8_t* src, uint8_t* dst, int width);
+void ARGBToRGB24Row_AVX2(const uint8_t* src, uint8_t* dst, int width);
+
+void ARGBToRGB24Row_AVX512VBMI(const uint8_t* src, uint8_t* dst, int width);
+
+void ARGBToRGB565DitherRow_C(const uint8_t* src_argb,
+                             uint8_t* dst_rgb,
+                             uint32_t dither4,
+                             int width);
+void ARGBToRGB565DitherRow_SSE2(const uint8_t* src,
+                                uint8_t* dst,
+                                uint32_t dither4,
+                                int width);
+void ARGBToRGB565DitherRow_AVX2(const uint8_t* src,
+                                uint8_t* dst,
+                                uint32_t dither4,
+                                int width);
+
+void ARGBToRGB565Row_AVX2(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToARGB1555Row_AVX2(const uint8_t* src_argb,
+                            uint8_t* dst_rgb,
+                            int width);
+void ARGBToARGB4444Row_AVX2(const uint8_t* src_argb,
+                            uint8_t* dst_rgb,
+                            int width);
+void ABGRToAR30Row_AVX2(const uint8_t* src, uint8_t* dst, int width);
+void ARGBToAR30Row_AVX2(const uint8_t* src, uint8_t* dst, int width);
+
+void ARGBToRGB24Row_NEON(const uint8_t* src_argb,
+                         uint8_t* dst_rgb24,
+                         int width);
+void ARGBToRGB24Row_SVE2(const uint8_t* src_argb,
+                         uint8_t* dst_rgb24,
+                         int width);
+void ARGBToRAWRow_NEON(const uint8_t* src_argb, uint8_t* dst_raw, int width);
+void ARGBToRAWRow_SVE2(const uint8_t* src_argb, uint8_t* dst_raw, int width);
+void ARGBToRGB565Row_NEON(const uint8_t* src_argb,
+                          uint8_t* dst_rgb565,
+                          int width);
+void ARGBToRGB565Row_SVE2(const uint8_t* src_argb,
+                          uint8_t* dst_rgb565,
+                          int width);
+void ARGBToARGB1555Row_NEON(const uint8_t* src_argb,
+                            uint8_t* dst_argb1555,
+                            int width);
+void ARGBToARGB4444Row_NEON(const uint8_t* src_argb,
+                            uint8_t* dst_argb4444,
+                            int width);
+void ARGBToRGB565DitherRow_NEON(const uint8_t* src_argb,
+                                uint8_t* dst_rgb,
+                                uint32_t dither4,
+                                int width);
+void ARGBToRGB565DitherRow_SVE2(const uint8_t* src_argb,
+                                uint8_t* dst_rgb,
+                                uint32_t dither4,
+                                int width);
+void ARGBToRGB24Row_MSA(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToRAWRow_MSA(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToRGB565Row_MSA(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToARGB1555Row_MSA(const uint8_t* src_argb,
+                           uint8_t* dst_rgb,
+                           int width);
+void ARGBToARGB4444Row_MSA(const uint8_t* src_argb,
+                           uint8_t* dst_rgb,
+                           int width);
+void ARGBToRGB565DitherRow_MSA(const uint8_t* src_argb,
+                               uint8_t* dst_rgb,
+                               uint32_t dither4,
+                               int width);
+void ARGBToRGB565DitherRow_LSX(const uint8_t* src_argb,
+                               uint8_t* dst_rgb,
+                               uint32_t dither4,
+                               int width);
+void ARGBToRGB565DitherRow_LASX(const uint8_t* src_argb,
+                                uint8_t* dst_rgb,
+                                uint32_t dither4,
+                                int width);
+
+void ARGBToRGB24Row_LSX(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToRGB24Row_LASX(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToRAWRow_LSX(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToRAWRow_LASX(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToRGB565Row_LSX(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToRGB565Row_LASX(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToARGB1555Row_LSX(const uint8_t* src_argb,
+                           uint8_t* dst_rgb,
+                           int width);
+void ARGBToARGB1555Row_LASX(const uint8_t* src_argb,
+                            uint8_t* dst_rgb,
+                            int width);
+void ARGBToARGB4444Row_LSX(const uint8_t* src_argb,
+                           uint8_t* dst_rgb,
+                           int width);
+void ARGBToARGB4444Row_LASX(const uint8_t* src_argb,
+                            uint8_t* dst_rgb,
+                            int width);
+
+void ARGBToRAWRow_RVV(const uint8_t* src_argb, uint8_t* dst_raw, int width);
+void ARGBToABGRRow_RVV(const uint8_t* src_argb, uint8_t* dst_abgr, int width);
+void ARGBToBGRARow_RVV(const uint8_t* src_argb, uint8_t* dst_rgba, int width);
+void ARGBToRGBARow_RVV(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToRGB24Row_RVV(const uint8_t* src_argb, uint8_t* dst_rgb24, int width);
+
+void ARGBToABGRRow_C(const uint8_t* src_argb, uint8_t* dst_abgr, int width);
+void ARGBToBGRARow_C(const uint8_t* src_argb, uint8_t* dst_bgra, int width);
+void ARGBToRGBARow_C(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToRGB24Row_C(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToRAWRow_C(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToRGB565Row_C(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToARGB1555Row_C(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ARGBToARGB4444Row_C(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
+void ABGRToAR30Row_C(const uint8_t* src_abgr, uint8_t* dst_ar30, int width);
+void ARGBToAR30Row_C(const uint8_t* src_argb, uint8_t* dst_ar30, int width);
+
+void ARGBToAR64Row_C(const uint8_t* src_argb, uint16_t* dst_ar64, int width);
+void ARGBToAB64Row_C(const uint8_t* src_argb, uint16_t* dst_ab64, int width);
+void AR64ToARGBRow_C(const uint16_t* src_ar64, uint8_t* dst_argb, int width);
+void AB64ToARGBRow_C(const uint16_t* src_ab64, uint8_t* dst_argb, int width);
+void AR64ToAB64Row_C(const uint16_t* src_ar64, uint16_t* dst_ab64, int width);
+void RGBAToARGBRow_C(const uint8_t* src_rgba, uint8_t* dst_argb, int width);
+void AR64ShuffleRow_C(const uint8_t* src_ar64,
+                      uint8_t* dst_ar64,
+                      const uint8_t* shuffler,
+                      int width);
+void ARGBToAR64Row_SSSE3(const uint8_t* src_argb,
+                         uint16_t* dst_ar64,
+                         int width);
+void ARGBToAB64Row_SSSE3(const uint8_t* src_argb,
+                         uint16_t* dst_ab64,
+                         int width);
+void AR64ToARGBRow_SSSE3(const uint16_t* src_ar64,
+                         uint8_t* dst_argb,
+                         int width);
+void AB64ToARGBRow_SSSE3(const uint16_t* src_ab64,
+                         uint8_t* dst_argb,
+                         int width);
+void ARGBToAR64Row_AVX2(const uint8_t* src_argb, uint16_t* dst_ar64, int width);
+void ARGBToAB64Row_AVX2(const uint8_t* src_argb, uint16_t* dst_ab64, int width);
+void AR64ToARGBRow_AVX2(const uint16_t* src_ar64, uint8_t* dst_argb, int width);
+void AB64ToARGBRow_AVX2(const uint16_t* src_ab64, uint8_t* dst_argb, int width);
+void ARGBToAR64Row_NEON(const uint8_t* src_argb, uint16_t* dst_ar64, int width);
+void ARGBToAB64Row_NEON(const uint8_t* src_argb, uint16_t* dst_ab64, int width);
+void AR64ToARGBRow_NEON(const uint16_t* src_ar64, uint8_t* dst_argb, int width);
+void AB64ToARGBRow_NEON(const uint16_t* src_ab64, uint8_t* dst_argb, int width);
+void ARGBToAR64Row_RVV(const uint8_t* src_argb, uint16_t* dst_ar64, int width);
+void ARGBToAB64Row_RVV(const uint8_t* src_argb, uint16_t* dst_ab64, int width);
+void AR64ToARGBRow_RVV(const uint16_t* src_ar64, uint8_t* dst_argb, int width);
+void AB64ToARGBRow_RVV(const uint16_t* src_ab64, uint8_t* dst_argb, int width);
+void AR64ToAB64Row_RVV(const uint16_t* src_ar64, uint16_t* dst_ab64, int width);
+void RGBAToARGBRow_RVV(const uint8_t* src_rgba, uint8_t* dst_argb, int width);
+void ARGBToAR64Row_Any_SSSE3(const uint8_t* src_ptr,
+                             uint16_t* dst_ptr,
+                             int width);
+void ARGBToAB64Row_Any_SSSE3(const uint8_t* src_ptr,
+                             uint16_t* dst_ptr,
+                             int width);
+void AR64ToARGBRow_Any_SSSE3(const uint16_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void AB64ToARGBRow_Any_SSSE3(const uint16_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGBToAR64Row_Any_AVX2(const uint8_t* src_ptr,
+                            uint16_t* dst_ptr,
+                            int width);
+void ARGBToAB64Row_Any_AVX2(const uint8_t* src_ptr,
+                            uint16_t* dst_ptr,
+                            int width);
+void AR64ToARGBRow_Any_AVX2(const uint16_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void AB64ToARGBRow_Any_AVX2(const uint16_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void ARGBToAR64Row_Any_NEON(const uint8_t* src_ptr,
+                            uint16_t* dst_ptr,
+                            int width);
+void ARGBToAB64Row_Any_NEON(const uint8_t* src_ptr,
+                            uint16_t* dst_ptr,
+                            int width);
+void AR64ToARGBRow_Any_NEON(const uint16_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void AB64ToARGBRow_Any_NEON(const uint16_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+
+void J400ToARGBRow_SSE2(const uint8_t* src_y, uint8_t* dst_argb, int width);
+void J400ToARGBRow_AVX2(const uint8_t* src_y, uint8_t* dst_argb, int width);
+void J400ToARGBRow_NEON(const uint8_t* src_y, uint8_t* dst_argb, int width);
+void J400ToARGBRow_MSA(const uint8_t* src_y, uint8_t* dst_argb, int width);
+void J400ToARGBRow_LSX(const uint8_t* src_y, uint8_t* dst_argb, int width);
+void J400ToARGBRow_RVV(const uint8_t* src_y, uint8_t* dst_argb, int width);
+void J400ToARGBRow_C(const uint8_t* src_y, uint8_t* dst_argb, int width);
+void J400ToARGBRow_Any_SSE2(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void J400ToARGBRow_Any_AVX2(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void J400ToARGBRow_Any_NEON(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void J400ToARGBRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void J400ToARGBRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+
+void I444ToARGBRow_C(const uint8_t* src_y,
+                     const uint8_t* src_u,
+                     const uint8_t* src_v,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void I444ToRGB24Row_C(const uint8_t* src_y,
+                      const uint8_t* src_u,
+                      const uint8_t* src_v,
+                      uint8_t* rgb_buf,
+                      const struct YuvConstants* yuvconstants,
+                      int width);
+void I422ToARGBRow_C(const uint8_t* src_y,
+                     const uint8_t* src_u,
+                     const uint8_t* src_v,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void I422ToAR30Row_C(const uint8_t* src_y,
+                     const uint8_t* src_u,
+                     const uint8_t* src_v,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void I210ToAR30Row_C(const uint16_t* src_y,
+                     const uint16_t* src_u,
+                     const uint16_t* src_v,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void I210ToARGBRow_C(const uint16_t* src_y,
+                     const uint16_t* src_u,
+                     const uint16_t* src_v,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void I212ToAR30Row_C(const uint16_t* src_y,
+                     const uint16_t* src_u,
+                     const uint16_t* src_v,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void I212ToARGBRow_C(const uint16_t* src_y,
+                     const uint16_t* src_u,
+                     const uint16_t* src_v,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void I410ToAR30Row_C(const uint16_t* src_y,
+                     const uint16_t* src_u,
+                     const uint16_t* src_v,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void I410ToARGBRow_C(const uint16_t* src_y,
+                     const uint16_t* src_u,
+                     const uint16_t* src_v,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void I210AlphaToARGBRow_C(const uint16_t* src_y,
+                          const uint16_t* src_u,
+                          const uint16_t* src_v,
+                          const uint16_t* src_a,
+                          uint8_t* rgb_buf,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void I410AlphaToARGBRow_C(const uint16_t* src_y,
+                          const uint16_t* src_u,
+                          const uint16_t* src_v,
+                          const uint16_t* src_a,
+                          uint8_t* rgb_buf,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void I444AlphaToARGBRow_C(const uint8_t* src_y,
+                          const uint8_t* src_u,
+                          const uint8_t* src_v,
+                          const uint8_t* src_a,
+                          uint8_t* rgb_buf,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void I422AlphaToARGBRow_C(const uint8_t* src_y,
+                          const uint8_t* src_u,
+                          const uint8_t* src_v,
+                          const uint8_t* src_a,
+                          uint8_t* rgb_buf,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void NV12ToARGBRow_C(const uint8_t* src_y,
+                     const uint8_t* src_uv,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void NV12ToRGB565Row_C(const uint8_t* src_y,
+                       const uint8_t* src_uv,
+                       uint8_t* dst_rgb565,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void NV21ToARGBRow_C(const uint8_t* src_y,
+                     const uint8_t* src_vu,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void NV12ToRGB24Row_C(const uint8_t* src_y,
+                      const uint8_t* src_uv,
+                      uint8_t* rgb_buf,
+                      const struct YuvConstants* yuvconstants,
+                      int width);
+void NV21ToRGB24Row_C(const uint8_t* src_y,
+                      const uint8_t* src_vu,
+                      uint8_t* rgb_buf,
+                      const struct YuvConstants* yuvconstants,
+                      int width);
+void NV21ToYUV24Row_C(const uint8_t* src_y,
+                      const uint8_t* src_vu,
+                      uint8_t* dst_yuv24,
+                      int width);
+void YUY2ToARGBRow_C(const uint8_t* src_yuy2,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void UYVYToARGBRow_C(const uint8_t* src_uyvy,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void P210ToARGBRow_C(const uint16_t* src_y,
+                     const uint16_t* src_uv,
+                     uint8_t* dst_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void P410ToARGBRow_C(const uint16_t* src_y,
+                     const uint16_t* src_uv,
+                     uint8_t* dst_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void P210ToAR30Row_C(const uint16_t* src_y,
+                     const uint16_t* src_uv,
+                     uint8_t* dst_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void P410ToAR30Row_C(const uint16_t* src_y,
+                     const uint16_t* src_uv,
+                     uint8_t* dst_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+
+void I422ToRGBARow_C(const uint8_t* src_y,
+                     const uint8_t* src_u,
+                     const uint8_t* src_v,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void I422ToRGB24Row_C(const uint8_t* src_y,
+                      const uint8_t* src_u,
+                      const uint8_t* src_v,
+                      uint8_t* rgb_buf,
+                      const struct YuvConstants* yuvconstants,
+                      int width);
+void I422ToARGB4444Row_C(const uint8_t* src_y,
+                         const uint8_t* src_u,
+                         const uint8_t* src_v,
+                         uint8_t* dst_argb4444,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I422ToARGB1555Row_C(const uint8_t* src_y,
+                         const uint8_t* src_u,
+                         const uint8_t* src_v,
+                         uint8_t* dst_argb1555,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I422ToRGB565Row_C(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_rgb565,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void I422ToARGBRow_AVX2(const uint8_t* y_buf,
+                        const uint8_t* u_buf,
+                        const uint8_t* v_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I422ToARGBRow_AVX512BW(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_argb,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToRGBARow_AVX2(const uint8_t* y_buf,
+                        const uint8_t* u_buf,
+                        const uint8_t* v_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I444ToARGBRow_SSSE3(const uint8_t* y_buf,
+                         const uint8_t* u_buf,
+                         const uint8_t* v_buf,
+                         uint8_t* dst_argb,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I444ToARGBRow_AVX2(const uint8_t* y_buf,
+                        const uint8_t* u_buf,
+                        const uint8_t* v_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I444ToRGB24Row_SSSE3(const uint8_t* y_buf,
+                          const uint8_t* u_buf,
+                          const uint8_t* v_buf,
+                          uint8_t* dst_rgb24,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void I444ToRGB24Row_AVX2(const uint8_t* y_buf,
+                         const uint8_t* u_buf,
+                         const uint8_t* v_buf,
+                         uint8_t* dst_rgb24,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I422ToARGBRow_SSSE3(const uint8_t* y_buf,
+                         const uint8_t* u_buf,
+                         const uint8_t* v_buf,
+                         uint8_t* dst_argb,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+
+void I422ToAR30Row_SSSE3(const uint8_t* y_buf,
+                         const uint8_t* u_buf,
+                         const uint8_t* v_buf,
+                         uint8_t* dst_ar30,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I210ToAR30Row_SSSE3(const uint16_t* y_buf,
+                         const uint16_t* u_buf,
+                         const uint16_t* v_buf,
+                         uint8_t* dst_ar30,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I210ToARGBRow_SSSE3(const uint16_t* y_buf,
+                         const uint16_t* u_buf,
+                         const uint16_t* v_buf,
+                         uint8_t* dst_argb,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I212ToAR30Row_SSSE3(const uint16_t* y_buf,
+                         const uint16_t* u_buf,
+                         const uint16_t* v_buf,
+                         uint8_t* dst_ar30,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I212ToARGBRow_SSSE3(const uint16_t* y_buf,
+                         const uint16_t* u_buf,
+                         const uint16_t* v_buf,
+                         uint8_t* dst_argb,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I410ToAR30Row_SSSE3(const uint16_t* y_buf,
+                         const uint16_t* u_buf,
+                         const uint16_t* v_buf,
+                         uint8_t* dst_ar30,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I410ToARGBRow_SSSE3(const uint16_t* y_buf,
+                         const uint16_t* u_buf,
+                         const uint16_t* v_buf,
+                         uint8_t* dst_argb,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I210AlphaToARGBRow_SSSE3(const uint16_t* y_buf,
+                              const uint16_t* u_buf,
+                              const uint16_t* v_buf,
+                              const uint16_t* a_buf,
+                              uint8_t* dst_argb,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void I410AlphaToARGBRow_SSSE3(const uint16_t* y_buf,
+                              const uint16_t* u_buf,
+                              const uint16_t* v_buf,
+                              const uint16_t* a_buf,
+                              uint8_t* dst_argb,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void I422ToAR30Row_AVX2(const uint8_t* y_buf,
+                        const uint8_t* u_buf,
+                        const uint8_t* v_buf,
+                        uint8_t* dst_ar30,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I210ToARGBRow_AVX2(const uint16_t* y_buf,
+                        const uint16_t* u_buf,
+                        const uint16_t* v_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I210ToAR30Row_AVX2(const uint16_t* y_buf,
+                        const uint16_t* u_buf,
+                        const uint16_t* v_buf,
+                        uint8_t* dst_ar30,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I212ToARGBRow_AVX2(const uint16_t* y_buf,
+                        const uint16_t* u_buf,
+                        const uint16_t* v_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I212ToAR30Row_AVX2(const uint16_t* y_buf,
+                        const uint16_t* u_buf,
+                        const uint16_t* v_buf,
+                        uint8_t* dst_ar30,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I410ToAR30Row_AVX2(const uint16_t* y_buf,
+                        const uint16_t* u_buf,
+                        const uint16_t* v_buf,
+                        uint8_t* dst_ar30,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I410ToARGBRow_AVX2(const uint16_t* y_buf,
+                        const uint16_t* u_buf,
+                        const uint16_t* v_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I210AlphaToARGBRow_AVX2(const uint16_t* y_buf,
+                             const uint16_t* u_buf,
+                             const uint16_t* v_buf,
+                             const uint16_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I410AlphaToARGBRow_AVX2(const uint16_t* y_buf,
+                             const uint16_t* u_buf,
+                             const uint16_t* v_buf,
+                             const uint16_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I444AlphaToARGBRow_SSSE3(const uint8_t* y_buf,
+                              const uint8_t* u_buf,
+                              const uint8_t* v_buf,
+                              const uint8_t* a_buf,
+                              uint8_t* dst_argb,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void I444AlphaToARGBRow_AVX2(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             const uint8_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422AlphaToARGBRow_SSSE3(const uint8_t* y_buf,
+                              const uint8_t* u_buf,
+                              const uint8_t* v_buf,
+                              const uint8_t* a_buf,
+                              uint8_t* dst_argb,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void I422AlphaToARGBRow_AVX2(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             const uint8_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void NV12ToARGBRow_SSSE3(const uint8_t* y_buf,
+                         const uint8_t* uv_buf,
+                         uint8_t* dst_argb,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void NV12ToARGBRow_AVX2(const uint8_t* y_buf,
+                        const uint8_t* uv_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void NV12ToRGB24Row_SSSE3(const uint8_t* src_y,
+                          const uint8_t* src_uv,
+                          uint8_t* dst_rgb24,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void NV21ToRGB24Row_SSSE3(const uint8_t* src_y,
+                          const uint8_t* src_vu,
+                          uint8_t* dst_rgb24,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void NV12ToRGB565Row_SSSE3(const uint8_t* src_y,
+                           const uint8_t* src_uv,
+                           uint8_t* dst_rgb565,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void NV12ToRGB24Row_AVX2(const uint8_t* src_y,
+                         const uint8_t* src_uv,
+                         uint8_t* dst_rgb24,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void NV21ToRGB24Row_AVX2(const uint8_t* src_y,
+                         const uint8_t* src_vu,
+                         uint8_t* dst_rgb24,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void NV21ToYUV24Row_SSSE3(const uint8_t* src_y,
+                          const uint8_t* src_vu,
+                          uint8_t* dst_yuv24,
+                          int width);
+void NV21ToYUV24Row_AVX2(const uint8_t* src_y,
+                         const uint8_t* src_vu,
+                         uint8_t* dst_yuv24,
+                         int width);
+void NV12ToRGB565Row_AVX2(const uint8_t* src_y,
+                          const uint8_t* src_uv,
+                          uint8_t* dst_rgb565,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void NV21ToARGBRow_SSSE3(const uint8_t* y_buf,
+                         const uint8_t* vu_buf,
+                         uint8_t* dst_argb,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void NV21ToARGBRow_AVX2(const uint8_t* y_buf,
+                        const uint8_t* vu_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void YUY2ToARGBRow_SSSE3(const uint8_t* yuy2_buf,
+                         uint8_t* dst_argb,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void UYVYToARGBRow_SSSE3(const uint8_t* uyvy_buf,
+                         uint8_t* dst_argb,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void YUY2ToARGBRow_AVX2(const uint8_t* yuy2_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void UYVYToARGBRow_AVX2(const uint8_t* uyvy_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+
+void P210ToARGBRow_SSSE3(const uint16_t* y_buf,
+                         const uint16_t* uv_buf,
+                         uint8_t* dst_argb,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void P410ToARGBRow_SSSE3(const uint16_t* y_buf,
+                         const uint16_t* uv_buf,
+                         uint8_t* dst_argb,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void P210ToAR30Row_SSSE3(const uint16_t* y_buf,
+                         const uint16_t* uv_buf,
+                         uint8_t* dst_ar30,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void P410ToAR30Row_SSSE3(const uint16_t* y_buf,
+                         const uint16_t* uv_buf,
+                         uint8_t* dst_ar30,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void P210ToARGBRow_AVX2(const uint16_t* y_buf,
+                        const uint16_t* uv_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void P410ToARGBRow_AVX2(const uint16_t* y_buf,
+                        const uint16_t* uv_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void P210ToAR30Row_AVX2(const uint16_t* y_buf,
+                        const uint16_t* uv_buf,
+                        uint8_t* dst_ar30,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void P410ToAR30Row_AVX2(const uint16_t* y_buf,
+                        const uint16_t* uv_buf,
+                        uint8_t* dst_ar30,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+
+void I422ToRGBARow_SSSE3(const uint8_t* y_buf,
+                         const uint8_t* u_buf,
+                         const uint8_t* v_buf,
+                         uint8_t* dst_rgba,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I422ToARGB4444Row_SSSE3(const uint8_t* src_y,
+                             const uint8_t* src_u,
+                             const uint8_t* src_v,
+                             uint8_t* dst_argb4444,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToARGB4444Row_AVX2(const uint8_t* src_y,
+                            const uint8_t* src_u,
+                            const uint8_t* src_v,
+                            uint8_t* dst_argb4444,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToARGB1555Row_SSSE3(const uint8_t* src_y,
+                             const uint8_t* src_u,
+                             const uint8_t* src_v,
+                             uint8_t* dst_argb1555,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToARGB1555Row_AVX2(const uint8_t* src_y,
+                            const uint8_t* src_u,
+                            const uint8_t* src_v,
+                            uint8_t* dst_argb1555,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToRGB565Row_SSSE3(const uint8_t* src_y,
+                           const uint8_t* src_u,
+                           const uint8_t* src_v,
+                           uint8_t* dst_rgb565,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I422ToRGB565Row_AVX2(const uint8_t* src_y,
+                          const uint8_t* src_u,
+                          const uint8_t* src_v,
+                          uint8_t* dst_rgb565,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void I422ToRGB24Row_SSSE3(const uint8_t* y_buf,
+                          const uint8_t* u_buf,
+                          const uint8_t* v_buf,
+                          uint8_t* dst_rgb24,
+                          const struct YuvConstants* yuvconstants,
+                          int width);
+void I422ToRGB24Row_AVX2(const uint8_t* src_y,
+                         const uint8_t* src_u,
+                         const uint8_t* src_v,
+                         uint8_t* dst_rgb24,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I422ToARGBRow_Any_AVX2(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToARGBRow_Any_AVX512BW(const uint8_t* y_buf,
+                                const uint8_t* u_buf,
+                                const uint8_t* v_buf,
+                                uint8_t* dst_ptr,
+                                const struct YuvConstants* yuvconstants,
+                                int width);
+void I422ToRGBARow_Any_AVX2(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I444ToARGBRow_Any_SSSE3(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I444ToRGB24Row_Any_SSSE3(const uint8_t* y_buf,
+                              const uint8_t* u_buf,
+                              const uint8_t* v_buf,
+                              uint8_t* dst_ptr,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void I444ToARGBRow_Any_AVX2(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I444ToRGB24Row_Any_AVX2(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToARGBRow_Any_SSSE3(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToAR30Row_Any_SSSE3(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I210ToAR30Row_Any_SSSE3(const uint16_t* y_buf,
+                             const uint16_t* u_buf,
+                             const uint16_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I210ToARGBRow_Any_SSSE3(const uint16_t* y_buf,
+                             const uint16_t* u_buf,
+                             const uint16_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I212ToAR30Row_Any_SSSE3(const uint16_t* y_buf,
+                             const uint16_t* u_buf,
+                             const uint16_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I212ToARGBRow_Any_SSSE3(const uint16_t* y_buf,
+                             const uint16_t* u_buf,
+                             const uint16_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I410ToAR30Row_Any_SSSE3(const uint16_t* y_buf,
+                             const uint16_t* u_buf,
+                             const uint16_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I410ToARGBRow_Any_SSSE3(const uint16_t* y_buf,
+                             const uint16_t* u_buf,
+                             const uint16_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I210AlphaToARGBRow_Any_SSSE3(const uint16_t* y_buf,
+                                  const uint16_t* u_buf,
+                                  const uint16_t* v_buf,
+                                  const uint16_t* a_buf,
+                                  uint8_t* dst_ptr,
+                                  const struct YuvConstants* yuvconstants,
+                                  int width);
+void I410AlphaToARGBRow_Any_SSSE3(const uint16_t* y_buf,
+                                  const uint16_t* u_buf,
+                                  const uint16_t* v_buf,
+                                  const uint16_t* a_buf,
+                                  uint8_t* dst_ptr,
+                                  const struct YuvConstants* yuvconstants,
+                                  int width);
+void I422ToAR30Row_Any_AVX2(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I210ToARGBRow_Any_AVX2(const uint16_t* y_buf,
+                            const uint16_t* u_buf,
+                            const uint16_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I210ToAR30Row_Any_AVX2(const uint16_t* y_buf,
+                            const uint16_t* u_buf,
+                            const uint16_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I212ToARGBRow_Any_AVX2(const uint16_t* y_buf,
+                            const uint16_t* u_buf,
+                            const uint16_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I212ToAR30Row_Any_AVX2(const uint16_t* y_buf,
+                            const uint16_t* u_buf,
+                            const uint16_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I410ToAR30Row_Any_AVX2(const uint16_t* y_buf,
+                            const uint16_t* u_buf,
+                            const uint16_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I410ToARGBRow_Any_AVX2(const uint16_t* y_buf,
+                            const uint16_t* u_buf,
+                            const uint16_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I210AlphaToARGBRow_Any_AVX2(const uint16_t* y_buf,
+                                 const uint16_t* u_buf,
+                                 const uint16_t* v_buf,
+                                 const uint16_t* a_buf,
+                                 uint8_t* dst_ptr,
+                                 const struct YuvConstants* yuvconstants,
+                                 int width);
+void I410AlphaToARGBRow_Any_AVX2(const uint16_t* y_buf,
+                                 const uint16_t* u_buf,
+                                 const uint16_t* v_buf,
+                                 const uint16_t* a_buf,
+                                 uint8_t* dst_ptr,
+                                 const struct YuvConstants* yuvconstants,
+                                 int width);
+void I444AlphaToARGBRow_Any_SSSE3(const uint8_t* y_buf,
+                                  const uint8_t* u_buf,
+                                  const uint8_t* v_buf,
+                                  const uint8_t* a_buf,
+                                  uint8_t* dst_ptr,
+                                  const struct YuvConstants* yuvconstants,
+                                  int width);
+void I444AlphaToARGBRow_Any_AVX2(const uint8_t* y_buf,
+                                 const uint8_t* u_buf,
+                                 const uint8_t* v_buf,
+                                 const uint8_t* a_buf,
+                                 uint8_t* dst_ptr,
+                                 const struct YuvConstants* yuvconstants,
+                                 int width);
+void I422AlphaToARGBRow_Any_SSSE3(const uint8_t* y_buf,
+                                  const uint8_t* u_buf,
+                                  const uint8_t* v_buf,
+                                  const uint8_t* a_buf,
+                                  uint8_t* dst_ptr,
+                                  const struct YuvConstants* yuvconstants,
+                                  int width);
+void I422AlphaToARGBRow_Any_AVX2(const uint8_t* y_buf,
+                                 const uint8_t* u_buf,
+                                 const uint8_t* v_buf,
+                                 const uint8_t* a_buf,
+                                 uint8_t* dst_ptr,
+                                 const struct YuvConstants* yuvconstants,
+                                 int width);
+void NV12ToARGBRow_Any_SSSE3(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void NV12ToARGBRow_Any_AVX2(const uint8_t* y_buf,
+                            const uint8_t* uv_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void NV21ToARGBRow_Any_SSSE3(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void NV21ToARGBRow_Any_AVX2(const uint8_t* y_buf,
+                            const uint8_t* uv_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void NV12ToRGB24Row_Any_SSSE3(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void NV21ToRGB24Row_Any_SSSE3(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void NV12ToRGB24Row_Any_AVX2(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void NV21ToRGB24Row_Any_AVX2(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void NV21ToYUV24Row_Any_SSSE3(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              int width);
+void NV21ToYUV24Row_Any_AVX2(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             int width);
+void NV12ToRGB565Row_Any_SSSE3(const uint8_t* y_buf,
+                               const uint8_t* uv_buf,
+                               uint8_t* dst_ptr,
+                               const struct YuvConstants* yuvconstants,
+                               int width);
+void NV12ToRGB565Row_Any_AVX2(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void YUY2ToARGBRow_Any_SSSE3(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void UYVYToARGBRow_Any_SSSE3(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void YUY2ToARGBRow_Any_AVX2(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void UYVYToARGBRow_Any_AVX2(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void P210ToARGBRow_Any_SSSE3(const uint16_t* y_buf,
+                             const uint16_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void P410ToARGBRow_Any_SSSE3(const uint16_t* y_buf,
+                             const uint16_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void P210ToAR30Row_Any_SSSE3(const uint16_t* y_buf,
+                             const uint16_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void P410ToAR30Row_Any_SSSE3(const uint16_t* y_buf,
+                             const uint16_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void P210ToARGBRow_Any_AVX2(const uint16_t* y_buf,
+                            const uint16_t* uv_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void P410ToARGBRow_Any_AVX2(const uint16_t* y_buf,
+                            const uint16_t* uv_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void P210ToAR30Row_Any_AVX2(const uint16_t* y_buf,
+                            const uint16_t* uv_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void P410ToAR30Row_Any_AVX2(const uint16_t* y_buf,
+                            const uint16_t* uv_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToRGBARow_Any_SSSE3(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToARGB4444Row_Any_SSSE3(const uint8_t* y_buf,
+                                 const uint8_t* u_buf,
+                                 const uint8_t* v_buf,
+                                 uint8_t* dst_ptr,
+                                 const struct YuvConstants* yuvconstants,
+                                 int width);
+void I422ToARGB4444Row_Any_AVX2(const uint8_t* y_buf,
+                                const uint8_t* u_buf,
+                                const uint8_t* v_buf,
+                                uint8_t* dst_ptr,
+                                const struct YuvConstants* yuvconstants,
+                                int width);
+void I422ToARGB1555Row_Any_SSSE3(const uint8_t* y_buf,
+                                 const uint8_t* u_buf,
+                                 const uint8_t* v_buf,
+                                 uint8_t* dst_ptr,
+                                 const struct YuvConstants* yuvconstants,
+                                 int width);
+void I422ToARGB1555Row_Any_AVX2(const uint8_t* y_buf,
+                                const uint8_t* u_buf,
+                                const uint8_t* v_buf,
+                                uint8_t* dst_ptr,
+                                const struct YuvConstants* yuvconstants,
+                                int width);
+void I422ToRGB565Row_Any_SSSE3(const uint8_t* y_buf,
+                               const uint8_t* u_buf,
+                               const uint8_t* v_buf,
+                               uint8_t* dst_ptr,
+                               const struct YuvConstants* yuvconstants,
+                               int width);
+void I422ToRGB565Row_Any_AVX2(const uint8_t* y_buf,
+                              const uint8_t* u_buf,
+                              const uint8_t* v_buf,
+                              uint8_t* dst_ptr,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void I422ToRGB24Row_Any_SSSE3(const uint8_t* y_buf,
+                              const uint8_t* u_buf,
+                              const uint8_t* v_buf,
+                              uint8_t* dst_ptr,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void I422ToRGB24Row_Any_AVX2(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+
+void I400ToARGBRow_C(const uint8_t* src_y,
+                     uint8_t* rgb_buf,
+                     const struct YuvConstants* yuvconstants,
+                     int width);
+void I400ToARGBRow_SSE2(const uint8_t* y_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I400ToARGBRow_AVX2(const uint8_t* y_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I400ToARGBRow_NEON(const uint8_t* src_y,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I400ToARGBRow_SVE2(const uint8_t* src_y,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void I400ToARGBRow_MSA(const uint8_t* src_y,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void I400ToARGBRow_LSX(const uint8_t* src_y,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void I400ToARGBRow_RVV(const uint8_t* src_y,
+                       uint8_t* dst_argb,
+                       const struct YuvConstants* yuvconstants,
+                       int width);
+void I400ToARGBRow_Any_SSE2(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* param,
+                            int width);
+void I400ToARGBRow_Any_AVX2(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* param,
+                            int width);
+void I400ToARGBRow_Any_NEON(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* param,
+                            int width);
+void I400ToARGBRow_Any_MSA(const uint8_t* src_ptr,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I400ToARGBRow_Any_LSX(const uint8_t* src_ptr,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+
+// ARGB preattenuated alpha blend.
+void ARGBBlendRow_SSSE3(const uint8_t* src_argb,
+                        const uint8_t* src_argb1,
+                        uint8_t* dst_argb,
+                        int width);
+void ARGBBlendRow_NEON(const uint8_t* src_argb,
+                       const uint8_t* src_argb1,
+                       uint8_t* dst_argb,
+                       int width);
+void ARGBBlendRow_MSA(const uint8_t* src_argb0,
+                      const uint8_t* src_argb1,
+                      uint8_t* dst_argb,
+                      int width);
+void ARGBBlendRow_LSX(const uint8_t* src_argb0,
+                      const uint8_t* src_argb1,
+                      uint8_t* dst_argb,
+                      int width);
+void ARGBBlendRow_RVV(const uint8_t* src_argb0,
+                      const uint8_t* src_argb1,
+                      uint8_t* dst_argb,
+                      int width);
+void ARGBBlendRow_C(const uint8_t* src_argb,
+                    const uint8_t* src_argb1,
+                    uint8_t* dst_argb,
+                    int width);
+
+// Unattenuated planar alpha blend.
+void BlendPlaneRow_SSSE3(const uint8_t* src0,
+                         const uint8_t* src1,
+                         const uint8_t* alpha,
+                         uint8_t* dst,
+                         int width);
+void BlendPlaneRow_Any_SSSE3(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             uint8_t* dst_ptr,
+                             int width);
+void BlendPlaneRow_AVX2(const uint8_t* src0,
+                        const uint8_t* src1,
+                        const uint8_t* alpha,
+                        uint8_t* dst,
+                        int width);
+void BlendPlaneRow_Any_AVX2(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            int width);
+void BlendPlaneRow_RVV(const uint8_t* src0,
+                       const uint8_t* src1,
+                       const uint8_t* alpha,
+                       uint8_t* dst,
+                       int width);
+void BlendPlaneRow_C(const uint8_t* src0,
+                     const uint8_t* src1,
+                     const uint8_t* alpha,
+                     uint8_t* dst,
+                     int width);
+
+// ARGB multiply images. Same API as Blend, but these require
+// pointer and width alignment for SSE2.
+void ARGBMultiplyRow_C(const uint8_t* src_argb,
+                       const uint8_t* src_argb1,
+                       uint8_t* dst_argb,
+                       int width);
+void ARGBMultiplyRow_SSE2(const uint8_t* src_argb,
+                          const uint8_t* src_argb1,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGBMultiplyRow_Any_SSE2(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBMultiplyRow_AVX2(const uint8_t* src_argb,
+                          const uint8_t* src_argb1,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGBMultiplyRow_Any_AVX2(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBMultiplyRow_NEON(const uint8_t* src_argb,
+                          const uint8_t* src_argb1,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGBMultiplyRow_Any_NEON(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBMultiplyRow_MSA(const uint8_t* src_argb0,
+                         const uint8_t* src_argb1,
+                         uint8_t* dst_argb,
+                         int width);
+void ARGBMultiplyRow_Any_MSA(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGBMultiplyRow_LSX(const uint8_t* src_argb0,
+                         const uint8_t* src_argb1,
+                         uint8_t* dst_argb,
+                         int width);
+void ARGBMultiplyRow_LASX(const uint8_t* src_argb0,
+                          const uint8_t* src_argb1,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGBMultiplyRow_Any_LSX(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGBMultiplyRow_Any_LASX(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              int width);
+
+// ARGB add images.
+void ARGBAddRow_C(const uint8_t* src_argb,
+                  const uint8_t* src_argb1,
+                  uint8_t* dst_argb,
+                  int width);
+void ARGBAddRow_SSE2(const uint8_t* src_argb,
+                     const uint8_t* src_argb1,
+                     uint8_t* dst_argb,
+                     int width);
+void ARGBAddRow_Any_SSE2(const uint8_t* y_buf,
+                         const uint8_t* uv_buf,
+                         uint8_t* dst_ptr,
+                         int width);
+void ARGBAddRow_AVX2(const uint8_t* src_argb,
+                     const uint8_t* src_argb1,
+                     uint8_t* dst_argb,
+                     int width);
+void ARGBAddRow_Any_AVX2(const uint8_t* y_buf,
+                         const uint8_t* uv_buf,
+                         uint8_t* dst_ptr,
+                         int width);
+void ARGBAddRow_NEON(const uint8_t* src_argb,
+                     const uint8_t* src_argb1,
+                     uint8_t* dst_argb,
+                     int width);
+void ARGBAddRow_Any_NEON(const uint8_t* y_buf,
+                         const uint8_t* uv_buf,
+                         uint8_t* dst_ptr,
+                         int width);
+void ARGBAddRow_MSA(const uint8_t* src_argb0,
+                    const uint8_t* src_argb1,
+                    uint8_t* dst_argb,
+                    int width);
+void ARGBAddRow_Any_MSA(const uint8_t* y_buf,
+                        const uint8_t* uv_buf,
+                        uint8_t* dst_ptr,
+                        int width);
+void ARGBAddRow_LSX(const uint8_t* src_argb0,
+                    const uint8_t* src_argb1,
+                    uint8_t* dst_argb,
+                    int width);
+void ARGBAddRow_LASX(const uint8_t* src_argb0,
+                     const uint8_t* src_argb1,
+                     uint8_t* dst_argb,
+                     int width);
+void ARGBAddRow_Any_LSX(const uint8_t* y_buf,
+                        const uint8_t* uv_buf,
+                        uint8_t* dst_ptr,
+                        int width);
+void ARGBAddRow_Any_LASX(const uint8_t* y_buf,
+                         const uint8_t* uv_buf,
+                         uint8_t* dst_ptr,
+                         int width);
+
+// ARGB subtract images. Same API as Blend, but these require
+// pointer and width alignment for SSE2.
+void ARGBSubtractRow_C(const uint8_t* src_argb,
+                       const uint8_t* src_argb1,
+                       uint8_t* dst_argb,
+                       int width);
+void ARGBSubtractRow_SSE2(const uint8_t* src_argb,
+                          const uint8_t* src_argb1,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGBSubtractRow_Any_SSE2(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBSubtractRow_AVX2(const uint8_t* src_argb,
+                          const uint8_t* src_argb1,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGBSubtractRow_Any_AVX2(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBSubtractRow_NEON(const uint8_t* src_argb,
+                          const uint8_t* src_argb1,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGBSubtractRow_Any_NEON(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBSubtractRow_MSA(const uint8_t* src_argb0,
+                         const uint8_t* src_argb1,
+                         uint8_t* dst_argb,
+                         int width);
+void ARGBSubtractRow_Any_MSA(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGBSubtractRow_LSX(const uint8_t* src_argb0,
+                         const uint8_t* src_argb1,
+                         uint8_t* dst_argb,
+                         int width);
+void ARGBSubtractRow_LASX(const uint8_t* src_argb0,
+                          const uint8_t* src_argb1,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGBSubtractRow_Any_LSX(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGBSubtractRow_Any_LASX(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              int width);
+
+void ARGBToRGB24Row_Any_SSSE3(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBToRAWRow_Any_SSSE3(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void ARGBToRGB565Row_Any_SSE2(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBToARGB1555Row_Any_SSE2(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void ARGBToARGB4444Row_Any_SSE2(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void ABGRToAR30Row_Any_SSSE3(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGBToAR30Row_Any_SSSE3(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGBToRAWRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToRGB24Row_Any_AVX2(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGBToRGB24Row_Any_AVX512VBMI(const uint8_t* src_ptr,
+                                   uint8_t* dst_ptr,
+                                   int width);
+void ARGBToRGB565DitherRow_Any_SSE2(const uint8_t* src_ptr,
+                                    uint8_t* dst_ptr,
+                                    const uint32_t param,
+                                    int width);
+void ARGBToRGB565DitherRow_Any_AVX2(const uint8_t* src_ptr,
+                                    uint8_t* dst_ptr,
+                                    const uint32_t param,
+                                    int width);
+
+void ARGBToRGB565Row_Any_AVX2(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBToARGB1555Row_Any_AVX2(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void ARGBToARGB4444Row_Any_AVX2(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void ABGRToAR30Row_Any_AVX2(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void ARGBToAR30Row_Any_AVX2(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+
+void ARGBToRGB24Row_Any_NEON(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGBToRAWRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToRGB565Row_Any_NEON(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBToARGB1555Row_Any_NEON(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void ARGBToARGB4444Row_Any_NEON(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void ARGBToRGB565DitherRow_Any_NEON(const uint8_t* src_ptr,
+                                    uint8_t* dst_ptr,
+                                    const uint32_t param,
+                                    int width);
+void ARGBToRGB24Row_Any_MSA(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void ARGBToRAWRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToRGB565Row_Any_MSA(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGBToARGB1555Row_Any_MSA(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+void ARGBToARGB4444Row_Any_MSA(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+void ARGBToRGB565DitherRow_Any_MSA(const uint8_t* src_ptr,
+                                   uint8_t* dst_ptr,
+                                   const uint32_t param,
+                                   int width);
+void ARGBToRGB565DitherRow_Any_LSX(const uint8_t* src_ptr,
+                                   uint8_t* dst_ptr,
+                                   const uint32_t param,
+                                   int width);
+void ARGBToRGB565DitherRow_Any_LASX(const uint8_t* src_ptr,
+                                    uint8_t* dst_ptr,
+                                    const uint32_t param,
+                                    int width);
+void ARGBToRGB24Row_Any_LSX(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void ARGBToRGB24Row_Any_LASX(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGBToRAWRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToRAWRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToRGB565Row_Any_LSX(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int width);
+void ARGBToRGB565Row_Any_LASX(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBToARGB1555Row_Any_LSX(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+void ARGBToARGB1555Row_Any_LASX(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void ARGBToARGB4444Row_Any_LSX(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+void ARGBToARGB4444Row_Any_LASX(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+
+void I444ToARGBRow_Any_NEON(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I444ToRGB24Row_Any_NEON(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToARGBRow_Any_NEON(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I210ToARGBRow_Any_NEON(const uint16_t* y_buf,
+                            const uint16_t* u_buf,
+                            const uint16_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I410ToARGBRow_Any_NEON(const uint16_t* y_buf,
+                            const uint16_t* u_buf,
+                            const uint16_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I210ToAR30Row_Any_NEON(const uint16_t* y_buf,
+                            const uint16_t* u_buf,
+                            const uint16_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I410ToAR30Row_Any_NEON(const uint16_t* y_buf,
+                            const uint16_t* u_buf,
+                            const uint16_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I212ToARGBRow_Any_NEON(const uint16_t* y_buf,
+                            const uint16_t* u_buf,
+                            const uint16_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I212ToAR30Row_Any_NEON(const uint16_t* y_buf,
+                            const uint16_t* u_buf,
+                            const uint16_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I444AlphaToARGBRow_Any_NEON(const uint8_t* y_buf,
+                                 const uint8_t* u_buf,
+                                 const uint8_t* v_buf,
+                                 const uint8_t* a_buf,
+                                 uint8_t* dst_ptr,
+                                 const struct YuvConstants* yuvconstants,
+                                 int width);
+void I422AlphaToARGBRow_Any_NEON(const uint8_t* y_buf,
+                                 const uint8_t* u_buf,
+                                 const uint8_t* v_buf,
+                                 const uint8_t* a_buf,
+                                 uint8_t* dst_ptr,
+                                 const struct YuvConstants* yuvconstants,
+                                 int width);
+void I410AlphaToARGBRow_Any_NEON(const uint16_t* y_buf,
+                                 const uint16_t* u_buf,
+                                 const uint16_t* v_buf,
+                                 const uint16_t* a_buf,
+                                 uint8_t* dst_ptr,
+                                 const struct YuvConstants* yuvconstants,
+                                 int width);
+void I210AlphaToARGBRow_Any_NEON(const uint16_t* y_buf,
+                                 const uint16_t* u_buf,
+                                 const uint16_t* v_buf,
+                                 const uint16_t* a_buf,
+                                 uint8_t* dst_ptr,
+                                 const struct YuvConstants* yuvconstants,
+                                 int width);
+void I422ToRGBARow_Any_NEON(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToRGB24Row_Any_NEON(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToARGB4444Row_Any_NEON(const uint8_t* y_buf,
+                                const uint8_t* u_buf,
+                                const uint8_t* v_buf,
+                                uint8_t* dst_ptr,
+                                const struct YuvConstants* yuvconstants,
+                                int width);
+void I422ToARGB1555Row_Any_NEON(const uint8_t* y_buf,
+                                const uint8_t* u_buf,
+                                const uint8_t* v_buf,
+                                uint8_t* dst_ptr,
+                                const struct YuvConstants* yuvconstants,
+                                int width);
+void I422ToRGB565Row_Any_NEON(const uint8_t* y_buf,
+                              const uint8_t* u_buf,
+                              const uint8_t* v_buf,
+                              uint8_t* dst_ptr,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void I422ToAR30Row_Any_NEON(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void NV12ToARGBRow_Any_NEON(const uint8_t* y_buf,
+                            const uint8_t* uv_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void NV21ToARGBRow_Any_NEON(const uint8_t* y_buf,
+                            const uint8_t* uv_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void NV12ToRGB24Row_Any_NEON(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void NV21ToRGB24Row_Any_NEON(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void NV21ToYUV24Row_Any_NEON(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             int width);
+void NV12ToRGB565Row_Any_NEON(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void YUY2ToARGBRow_Any_NEON(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void UYVYToARGBRow_Any_NEON(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void ARGBToAR30Row_NEON(const uint8_t* src, uint8_t* dst, int width);
+void ABGRToAR30Row_NEON(const uint8_t* src, uint8_t* dst, int width);
+void ABGRToAR30Row_Any_NEON(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void ARGBToAR30Row_Any_NEON(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int width);
+void P210ToARGBRow_NEON(const uint16_t* y_buf,
+                        const uint16_t* uv_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void P410ToARGBRow_NEON(const uint16_t* y_buf,
+                        const uint16_t* uv_buf,
+                        uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void P210ToAR30Row_NEON(const uint16_t* y_buf,
+                        const uint16_t* uv_buf,
+                        uint8_t* dst_ar30,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void P410ToAR30Row_NEON(const uint16_t* y_buf,
+                        const uint16_t* uv_buf,
+                        uint8_t* dst_ar30,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
+void P210ToARGBRow_Any_NEON(const uint16_t* y_buf,
+                            const uint16_t* uv_buf,
+                            uint8_t* dst_argb,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void P410ToARGBRow_Any_NEON(const uint16_t* y_buf,
+                            const uint16_t* uv_buf,
+                            uint8_t* dst_argb,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void P210ToAR30Row_Any_NEON(const uint16_t* y_buf,
+                            const uint16_t* uv_buf,
+                            uint8_t* dst_ar30,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void P410ToAR30Row_Any_NEON(const uint16_t* y_buf,
+                            const uint16_t* uv_buf,
+                            uint8_t* dst_ar30,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I444ToARGBRow_Any_MSA(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I444ToARGBRow_Any_LSX(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I422ToARGBRow_Any_MSA(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I422ToARGBRow_Any_LSX(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I422ToARGBRow_Any_LASX(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToRGBARow_Any_MSA(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I422ToRGBARow_Any_LSX(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I422ToRGBARow_Any_LASX(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422AlphaToARGBRow_Any_MSA(const uint8_t* y_buf,
+                                const uint8_t* u_buf,
+                                const uint8_t* v_buf,
+                                const uint8_t* a_buf,
+                                uint8_t* dst_ptr,
+                                const struct YuvConstants* yuvconstants,
+                                int width);
+void I422AlphaToARGBRow_Any_LSX(const uint8_t* y_buf,
+                                const uint8_t* u_buf,
+                                const uint8_t* v_buf,
+                                const uint8_t* a_buf,
+                                uint8_t* dst_ptr,
+                                const struct YuvConstants* yuvconstants,
+                                int width);
+void I422AlphaToARGBRow_Any_LASX(const uint8_t* y_buf,
+                                 const uint8_t* u_buf,
+                                 const uint8_t* v_buf,
+                                 const uint8_t* a_buf,
+                                 uint8_t* dst_ptr,
+                                 const struct YuvConstants* yuvconstants,
+                                 int width);
+void I422ToRGB24Row_Any_MSA(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToRGB24Row_Any_LSX(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToRGB24Row_Any_LASX(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToRGB565Row_Any_MSA(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToRGB565Row_Any_LSX(const uint8_t* y_buf,
+                             const uint8_t* u_buf,
+                             const uint8_t* v_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToRGB565Row_Any_LASX(const uint8_t* y_buf,
+                              const uint8_t* u_buf,
+                              const uint8_t* v_buf,
+                              uint8_t* dst_ptr,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void I422ToARGB4444Row_Any_MSA(const uint8_t* y_buf,
+                               const uint8_t* u_buf,
+                               const uint8_t* v_buf,
+                               uint8_t* dst_ptr,
+                               const struct YuvConstants* yuvconstants,
+                               int width);
+void I422ToARGB4444Row_Any_LSX(const uint8_t* y_buf,
+                               const uint8_t* u_buf,
+                               const uint8_t* v_buf,
+                               uint8_t* dst_ptr,
+                               const struct YuvConstants* yuvconstants,
+                               int width);
+void I422ToARGB4444Row_Any_LASX(const uint8_t* y_buf,
+                                const uint8_t* u_buf,
+                                const uint8_t* v_buf,
+                                uint8_t* dst_ptr,
+                                const struct YuvConstants* yuvconstants,
+                                int width);
+void I422ToARGB1555Row_Any_MSA(const uint8_t* y_buf,
+                               const uint8_t* u_buf,
+                               const uint8_t* v_buf,
+                               uint8_t* dst_ptr,
+                               const struct YuvConstants* yuvconstants,
+                               int width);
+void I422ToARGB1555Row_Any_LSX(const uint8_t* y_buf,
+                               const uint8_t* u_buf,
+                               const uint8_t* v_buf,
+                               uint8_t* dst_ptr,
+                               const struct YuvConstants* yuvconstants,
+                               int width);
+void I422ToARGB1555Row_Any_LASX(const uint8_t* y_buf,
+                                const uint8_t* u_buf,
+                                const uint8_t* v_buf,
+                                uint8_t* dst_ptr,
+                                const struct YuvConstants* yuvconstants,
+                                int width);
+void NV12ToARGBRow_Any_MSA(const uint8_t* y_buf,
+                           const uint8_t* uv_buf,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void NV12ToRGB565Row_Any_MSA(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void NV21ToARGBRow_Any_MSA(const uint8_t* y_buf,
+                           const uint8_t* uv_buf,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void YUY2ToARGBRow_Any_MSA(const uint8_t* src_ptr,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void UYVYToARGBRow_Any_MSA(const uint8_t* src_ptr,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+
+void NV12ToARGBRow_Any_LSX(const uint8_t* y_buf,
+                           const uint8_t* uv_buf,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void NV12ToARGBRow_Any_LASX(const uint8_t* y_buf,
+                            const uint8_t* uv_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void NV12ToRGB565Row_Any_LSX(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void NV12ToRGB565Row_Any_LASX(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              const struct YuvConstants* yuvconstants,
+                              int width);
+void NV21ToARGBRow_Any_LSX(const uint8_t* y_buf,
+                           const uint8_t* uv_buf,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void NV21ToARGBRow_Any_LASX(const uint8_t* y_buf,
+                            const uint8_t* uv_buf,
+                            uint8_t* dst_ptr,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void YUY2ToARGBRow_Any_LSX(const uint8_t* src_ptr,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void UYVYToARGBRow_Any_LSX(const uint8_t* src_ptr,
+                           uint8_t* dst_ptr,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+
+void YUY2ToYRow_AVX2(const uint8_t* src_yuy2, uint8_t* dst_y, int width);
+void YUY2ToUVRow_AVX2(const uint8_t* src_yuy2,
+                      int stride_yuy2,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void YUY2ToNVUVRow_AVX2(const uint8_t* src_yuy2,
+                        int stride_yuy2,
+                        uint8_t* dst_uv,
+                        int width);
+void YUY2ToUV422Row_AVX2(const uint8_t* src_yuy2,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void YUY2ToYRow_SSE2(const uint8_t* src_yuy2, uint8_t* dst_y, int width);
+void YUY2ToUVRow_SSE2(const uint8_t* src_yuy2,
+                      int stride_yuy2,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void YUY2ToNVUVRow_SSE2(const uint8_t* src_yuy2,
+                        int stride_yuy2,
+                        uint8_t* dst_uv,
+                        int width);
+void YUY2ToUV422Row_SSE2(const uint8_t* src_yuy2,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void YUY2ToYRow_NEON(const uint8_t* src_yuy2, uint8_t* dst_y, int width);
+void YUY2ToUVRow_NEON(const uint8_t* src_yuy2,
+                      int stride_yuy2,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void YUY2ToNVUVRow_NEON(const uint8_t* src_yuy2,
+                        int stride_yuy2,
+                        uint8_t* dst_uv,
+                        int width);
+void YUY2ToUV422Row_NEON(const uint8_t* src_yuy2,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void YUY2ToYRow_MSA(const uint8_t* src_yuy2, uint8_t* dst_y, int width);
+void YUY2ToYRow_LSX(const uint8_t* src_yuy2, uint8_t* dst_y, int width);
+void YUY2ToYRow_LASX(const uint8_t* src_yuy2, uint8_t* dst_y, int width);
+void YUY2ToUVRow_MSA(const uint8_t* src_yuy2,
+                     int src_stride_yuy2,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void YUY2ToUVRow_LSX(const uint8_t* src_yuy2,
+                     int src_stride_yuy2,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void YUY2ToUVRow_LASX(const uint8_t* src_yuy2,
+                      int src_stride_yuy2,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void YUY2ToUV422Row_MSA(const uint8_t* src_yuy2,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void YUY2ToUV422Row_LSX(const uint8_t* src_yuy2,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void YUY2ToUV422Row_LASX(const uint8_t* src_yuy2,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void YUY2ToYRow_C(const uint8_t* src_yuy2, uint8_t* dst_y, int width);
+void YUY2ToUVRow_C(const uint8_t* src_yuy2,
+                   int src_stride_yuy2,
+                   uint8_t* dst_u,
+                   uint8_t* dst_v,
+                   int width);
+void YUY2ToNVUVRow_C(const uint8_t* src_yuy2,
+                     int src_stride_yuy2,
+                     uint8_t* dst_uv,
+                     int width);
+void YUY2ToUV422Row_C(const uint8_t* src_yuy2,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void YUY2ToYRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void YUY2ToUVRow_Any_AVX2(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void YUY2ToNVUVRow_Any_AVX2(const uint8_t* src_yuy2,
+                            int stride_yuy2,
+                            uint8_t* dst_uv,
+                            int width);
+void YUY2ToUV422Row_Any_AVX2(const uint8_t* src_ptr,
+                             uint8_t* dst_u,
+                             uint8_t* dst_v,
+                             int width);
+void YUY2ToYRow_Any_SSE2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void YUY2ToUVRow_Any_SSE2(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void YUY2ToNVUVRow_Any_SSE2(const uint8_t* src_yuy2,
+                            int stride_yuy2,
+                            uint8_t* dst_uv,
+                            int width);
+void YUY2ToUV422Row_Any_SSE2(const uint8_t* src_ptr,
+                             uint8_t* dst_u,
+                             uint8_t* dst_v,
+                             int width);
+void YUY2ToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void YUY2ToUVRow_Any_NEON(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void YUY2ToNVUVRow_Any_NEON(const uint8_t* src_yuy2,
+                            int stride_yuy2,
+                            uint8_t* dst_uv,
+                            int width);
+void YUY2ToUV422Row_Any_NEON(const uint8_t* src_ptr,
+                             uint8_t* dst_u,
+                             uint8_t* dst_v,
+                             int width);
+void YUY2ToYRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void YUY2ToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void YUY2ToYRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void YUY2ToUVRow_Any_MSA(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void YUY2ToUVRow_Any_LSX(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void YUY2ToUVRow_Any_LASX(const uint8_t* src_ptr,
+                          int src_stride_ptr,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void YUY2ToUV422Row_Any_MSA(const uint8_t* src_ptr,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void YUY2ToUV422Row_Any_LSX(const uint8_t* src_ptr,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void YUY2ToUV422Row_Any_LASX(const uint8_t* src_ptr,
+                             uint8_t* dst_u,
+                             uint8_t* dst_v,
+                             int width);
+void UYVYToYRow_AVX2(const uint8_t* src_uyvy, uint8_t* dst_y, int width);
+void UYVYToUVRow_AVX2(const uint8_t* src_uyvy,
+                      int stride_uyvy,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void UYVYToUV422Row_AVX2(const uint8_t* src_uyvy,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void UYVYToYRow_SSE2(const uint8_t* src_uyvy, uint8_t* dst_y, int width);
+void UYVYToUVRow_SSE2(const uint8_t* src_uyvy,
+                      int stride_uyvy,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void UYVYToUV422Row_SSE2(const uint8_t* src_uyvy,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void UYVYToYRow_AVX2(const uint8_t* src_uyvy, uint8_t* dst_y, int width);
+void UYVYToUVRow_AVX2(const uint8_t* src_uyvy,
+                      int stride_uyvy,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void UYVYToUV422Row_AVX2(const uint8_t* src_uyvy,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void UYVYToYRow_NEON(const uint8_t* src_uyvy, uint8_t* dst_y, int width);
+void UYVYToUVRow_NEON(const uint8_t* src_uyvy,
+                      int stride_uyvy,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void UYVYToUV422Row_NEON(const uint8_t* src_uyvy,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void UYVYToYRow_MSA(const uint8_t* src_uyvy, uint8_t* dst_y, int width);
+void UYVYToYRow_LSX(const uint8_t* src_uyvy, uint8_t* dst_y, int width);
+void UYVYToYRow_LASX(const uint8_t* src_uyvy, uint8_t* dst_y, int width);
+void UYVYToUVRow_MSA(const uint8_t* src_uyvy,
+                     int src_stride_uyvy,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void UYVYToUVRow_LSX(const uint8_t* src_uyvy,
+                     int src_stride_uyvy,
+                     uint8_t* dst_u,
+                     uint8_t* dst_v,
+                     int width);
+void UYVYToUVRow_LASX(const uint8_t* src_uyvy,
+                      int src_stride_uyvy,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void UYVYToUV422Row_MSA(const uint8_t* src_uyvy,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void UYVYToUV422Row_LSX(const uint8_t* src_uyvy,
+                        uint8_t* dst_u,
+                        uint8_t* dst_v,
+                        int width);
+void UYVYToUV422Row_LASX(const uint8_t* src_uyvy,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+
+void UYVYToYRow_C(const uint8_t* src_uyvy, uint8_t* dst_y, int width);
+void UYVYToUVRow_C(const uint8_t* src_uyvy,
+                   int src_stride_uyvy,
+                   uint8_t* dst_u,
+                   uint8_t* dst_v,
+                   int width);
+void UYVYToUV422Row_C(const uint8_t* src_uyvy,
+                      uint8_t* dst_u,
+                      uint8_t* dst_v,
+                      int width);
+void UYVYToYRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void UYVYToUVRow_Any_AVX2(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void UYVYToUV422Row_Any_AVX2(const uint8_t* src_ptr,
+                             uint8_t* dst_u,
+                             uint8_t* dst_v,
+                             int width);
+void UYVYToYRow_Any_SSE2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void UYVYToUVRow_Any_SSE2(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void UYVYToUV422Row_Any_SSE2(const uint8_t* src_ptr,
+                             uint8_t* dst_u,
+                             uint8_t* dst_v,
+                             int width);
+void UYVYToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void UYVYToUVRow_Any_NEON(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void UYVYToUV422Row_Any_NEON(const uint8_t* src_ptr,
+                             uint8_t* dst_u,
+                             uint8_t* dst_v,
+                             int width);
+void UYVYToYRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void UYVYToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void UYVYToYRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void UYVYToUVRow_Any_MSA(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void UYVYToUVRow_Any_LSX(const uint8_t* src_ptr,
+                         int src_stride_ptr,
+                         uint8_t* dst_u,
+                         uint8_t* dst_v,
+                         int width);
+void UYVYToUVRow_Any_LASX(const uint8_t* src_ptr,
+                          int src_stride_ptr,
+                          uint8_t* dst_u,
+                          uint8_t* dst_v,
+                          int width);
+void UYVYToUV422Row_Any_MSA(const uint8_t* src_ptr,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void UYVYToUV422Row_Any_LSX(const uint8_t* src_ptr,
+                            uint8_t* dst_u,
+                            uint8_t* dst_v,
+                            int width);
+void UYVYToUV422Row_Any_LASX(const uint8_t* src_ptr,
+                             uint8_t* dst_u,
+                             uint8_t* dst_v,
+                             int width);
+void SwapUVRow_C(const uint8_t* src_uv, uint8_t* dst_vu, int width);
+void SwapUVRow_NEON(const uint8_t* src_uv, uint8_t* dst_vu, int width);
+void SwapUVRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void SwapUVRow_SSSE3(const uint8_t* src_uv, uint8_t* dst_vu, int width);
+void SwapUVRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void SwapUVRow_AVX2(const uint8_t* src_uv, uint8_t* dst_vu, int width);
+void SwapUVRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void AYUVToYRow_C(const uint8_t* src_ayuv, uint8_t* dst_y, int width);
+void AYUVToUVRow_C(const uint8_t* src_ayuv,
+                   int src_stride_ayuv,
+                   uint8_t* dst_uv,
+                   int width);
+void AYUVToVURow_C(const uint8_t* src_ayuv,
+                   int src_stride_ayuv,
+                   uint8_t* dst_vu,
+                   int width);
+void AYUVToYRow_NEON(const uint8_t* src_ayuv, uint8_t* dst_y, int width);
+void AYUVToUVRow_NEON(const uint8_t* src_ayuv,
+                      int src_stride_ayuv,
+                      uint8_t* dst_uv,
+                      int width);
+void AYUVToUVRow_SVE2(const uint8_t* src_ayuv,
+                      int src_stride_ayuv,
+                      uint8_t* dst_uv,
+                      int width);
+void AYUVToVURow_NEON(const uint8_t* src_ayuv,
+                      int src_stride_ayuv,
+                      uint8_t* dst_vu,
+                      int width);
+void AYUVToVURow_SVE2(const uint8_t* src_ayuv,
+                      int src_stride_ayuv,
+                      uint8_t* dst_vu,
+                      int width);
+void AYUVToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void AYUVToUVRow_Any_NEON(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_vu,
+                          int width);
+void AYUVToUVRow_Any_SVE2(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_vu,
+                          int width);
+void AYUVToVURow_Any_NEON(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_vu,
+                          int width);
+void AYUVToVURow_Any_SVE2(const uint8_t* src_ptr,
+                          int src_stride,
+                          uint8_t* dst_vu,
+                          int width);
+
+void I422ToYUY2Row_C(const uint8_t* src_y,
+                     const uint8_t* src_u,
+                     const uint8_t* src_v,
+                     uint8_t* dst_frame,
+                     int width);
+void I422ToUYVYRow_C(const uint8_t* src_y,
+                     const uint8_t* src_u,
+                     const uint8_t* src_v,
+                     uint8_t* dst_frame,
+                     int width);
+void I422ToYUY2Row_SSE2(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_yuy2,
+                        int width);
+void I422ToUYVYRow_SSE2(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_uyvy,
+                        int width);
+void I422ToYUY2Row_Any_SSE2(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            int width);
+void I422ToUYVYRow_Any_SSE2(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            int width);
+void I422ToYUY2Row_AVX2(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_yuy2,
+                        int width);
+void I422ToUYVYRow_AVX2(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_uyvy,
+                        int width);
+void I422ToYUY2Row_Any_AVX2(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            int width);
+void I422ToUYVYRow_Any_AVX2(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            int width);
+void I422ToYUY2Row_NEON(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_yuy2,
+                        int width);
+void I422ToUYVYRow_NEON(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_uyvy,
+                        int width);
+void I422ToYUY2Row_Any_NEON(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            int width);
+void I422ToUYVYRow_Any_NEON(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            int width);
+void I422ToYUY2Row_MSA(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_yuy2,
+                       int width);
+void I422ToYUY2Row_LSX(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_yuy2,
+                       int width);
+void I422ToYUY2Row_LASX(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_yuy2,
+                        int width);
+void I422ToUYVYRow_MSA(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_uyvy,
+                       int width);
+void I422ToUYVYRow_LSX(const uint8_t* src_y,
+                       const uint8_t* src_u,
+                       const uint8_t* src_v,
+                       uint8_t* dst_uyvy,
+                       int width);
+void I422ToUYVYRow_LASX(const uint8_t* src_y,
+                        const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        uint8_t* dst_uyvy,
+                        int width);
+void I422ToYUY2Row_Any_MSA(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           int width);
+void I422ToYUY2Row_Any_LSX(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           int width);
+void I422ToYUY2Row_Any_LASX(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            int width);
+void I422ToUYVYRow_Any_MSA(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           int width);
+void I422ToUYVYRow_Any_LSX(const uint8_t* y_buf,
+                           const uint8_t* u_buf,
+                           const uint8_t* v_buf,
+                           uint8_t* dst_ptr,
+                           int width);
+void I422ToUYVYRow_Any_LASX(const uint8_t* y_buf,
+                            const uint8_t* u_buf,
+                            const uint8_t* v_buf,
+                            uint8_t* dst_ptr,
+                            int width);
+
+// Effects related row functions.
+void ARGBAttenuateRow_C(const uint8_t* src_argb, uint8_t* dst_argb, int width);
+void ARGBAttenuateRow_SSSE3(const uint8_t* src_argb,
+                            uint8_t* dst_argb,
+                            int width);
+void ARGBAttenuateRow_AVX2(const uint8_t* src_argb,
+                           uint8_t* dst_argb,
+                           int width);
+void ARGBAttenuateRow_NEON(const uint8_t* src_argb,
+                           uint8_t* dst_argb,
+                           int width);
+void ARGBAttenuateRow_MSA(const uint8_t* src_argb,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGBAttenuateRow_LSX(const uint8_t* src_argb,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGBAttenuateRow_LASX(const uint8_t* src_argb,
+                           uint8_t* dst_argb,
+                           int width);
+void ARGBAttenuateRow_RVV(const uint8_t* src_argb,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGBAttenuateRow_Any_SSSE3(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int width);
+void ARGBAttenuateRow_Any_AVX2(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+void ARGBAttenuateRow_Any_NEON(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+void ARGBAttenuateRow_Any_MSA(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBAttenuateRow_Any_LSX(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int width);
+void ARGBAttenuateRow_Any_LASX(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int width);
+
+// Inverse table for unattenuate, shared by C and SSE2.
+extern const uint32_t fixed_invtbl8[256];
+void ARGBUnattenuateRow_C(const uint8_t* src_argb,
+                          uint8_t* dst_argb,
+                          int width);
+void ARGBUnattenuateRow_SSE2(const uint8_t* src_argb,
+                             uint8_t* dst_argb,
+                             int width);
+void ARGBUnattenuateRow_AVX2(const uint8_t* src_argb,
+                             uint8_t* dst_argb,
+                             int width);
+void ARGBUnattenuateRow_Any_SSE2(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int width);
+void ARGBUnattenuateRow_Any_AVX2(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int width);
+
+void ARGBGrayRow_C(const uint8_t* src_argb, uint8_t* dst_argb, int width);
+void ARGBGrayRow_SSSE3(const uint8_t* src_argb, uint8_t* dst_argb, int width);
+void ARGBGrayRow_NEON(const uint8_t* src_argb, uint8_t* dst_argb, int width);
+void ARGBGrayRow_MSA(const uint8_t* src_argb, uint8_t* dst_argb, int width);
+void ARGBGrayRow_LSX(const uint8_t* src_argb, uint8_t* dst_argb, int width);
+void ARGBGrayRow_LASX(const uint8_t* src_argb, uint8_t* dst_argb, int width);
+
+void ARGBSepiaRow_C(uint8_t* dst_argb, int width);
+void ARGBSepiaRow_SSSE3(uint8_t* dst_argb, int width);
+void ARGBSepiaRow_NEON(uint8_t* dst_argb, int width);
+void ARGBSepiaRow_MSA(uint8_t* dst_argb, int width);
+void ARGBSepiaRow_LSX(uint8_t* dst_argb, int width);
+void ARGBSepiaRow_LASX(uint8_t* dst_argb, int width);
+
+void ARGBColorMatrixRow_C(const uint8_t* src_argb,
+                          uint8_t* dst_argb,
+                          const int8_t* matrix_argb,
+                          int width);
+void ARGBColorMatrixRow_SSSE3(const uint8_t* src_argb,
+                              uint8_t* dst_argb,
+                              const int8_t* matrix_argb,
+                              int width);
+void ARGBColorMatrixRow_NEON(const uint8_t* src_argb,
+                             uint8_t* dst_argb,
+                             const int8_t* matrix_argb,
+                             int width);
+void ARGBColorMatrixRow_NEON_I8MM(const uint8_t* src_argb,
+                                  uint8_t* dst_argb,
+                                  const int8_t* matrix_argb,
+                                  int width);
+void ARGBColorMatrixRow_MSA(const uint8_t* src_argb,
+                            uint8_t* dst_argb,
+                            const int8_t* matrix_argb,
+                            int width);
+void ARGBColorMatrixRow_LSX(const uint8_t* src_argb,
+                            uint8_t* dst_argb,
+                            const int8_t* matrix_argb,
+                            int width);
+
+void ARGBColorTableRow_C(uint8_t* dst_argb,
+                         const uint8_t* table_argb,
+                         int width);
+void ARGBColorTableRow_X86(uint8_t* dst_argb,
+                           const uint8_t* table_argb,
+                           int width);
+
+void RGBColorTableRow_C(uint8_t* dst_argb,
+                        const uint8_t* table_argb,
+                        int width);
+void RGBColorTableRow_X86(uint8_t* dst_argb,
+                          const uint8_t* table_argb,
+                          int width);
+
+void ARGBQuantizeRow_C(uint8_t* dst_argb,
+                       int scale,
+                       int interval_size,
+                       int interval_offset,
+                       int width);
+void ARGBQuantizeRow_SSE2(uint8_t* dst_argb,
+                          int scale,
+                          int interval_size,
+                          int interval_offset,
+                          int width);
+void ARGBQuantizeRow_NEON(uint8_t* dst_argb,
+                          int scale,
+                          int interval_size,
+                          int interval_offset,
+                          int width);
+void ARGBQuantizeRow_MSA(uint8_t* dst_argb,
+                         int scale,
+                         int interval_size,
+                         int interval_offset,
+                         int width);
+void ARGBQuantizeRow_LSX(uint8_t* dst_argb,
+                         int scale,
+                         int interval_size,
+                         int interval_offset,
+                         int width);
+
+void ARGBShadeRow_C(const uint8_t* src_argb,
+                    uint8_t* dst_argb,
+                    int width,
+                    uint32_t value);
+void ARGBShadeRow_SSE2(const uint8_t* src_argb,
+                       uint8_t* dst_argb,
+                       int width,
+                       uint32_t value);
+void ARGBShadeRow_NEON(const uint8_t* src_argb,
+                       uint8_t* dst_argb,
+                       int width,
+                       uint32_t value);
+void ARGBShadeRow_MSA(const uint8_t* src_argb,
+                      uint8_t* dst_argb,
+                      int width,
+                      uint32_t value);
+void ARGBShadeRow_LSX(const uint8_t* src_argb,
+                      uint8_t* dst_argb,
+                      int width,
+                      uint32_t value);
+void ARGBShadeRow_LASX(const uint8_t* src_argb,
+                       uint8_t* dst_argb,
+                       int width,
+                       uint32_t value);
+
+// Used for blur.
+void CumulativeSumToAverageRow_SSE2(const int32_t* topleft,
+                                    const int32_t* botleft,
+                                    int width,
+                                    int area,
+                                    uint8_t* dst,
+                                    int count);
+void ComputeCumulativeSumRow_SSE2(const uint8_t* row,
+                                  int32_t* cumsum,
+                                  const int32_t* previous_cumsum,
+                                  int width);
+
+void CumulativeSumToAverageRow_C(const int32_t* tl,
+                                 const int32_t* bl,
+                                 int w,
+                                 int area,
+                                 uint8_t* dst,
+                                 int count);
+void ComputeCumulativeSumRow_C(const uint8_t* row,
+                               int32_t* cumsum,
+                               const int32_t* previous_cumsum,
+                               int width);
+
+LIBYUV_API
+void ARGBAffineRow_C(const uint8_t* src_argb,
+                     int src_argb_stride,
+                     uint8_t* dst_argb,
+                     const float* uv_dudv,
+                     int width);
+LIBYUV_API
+void ARGBAffineRow_SSE2(const uint8_t* src_argb,
+                        int src_argb_stride,
+                        uint8_t* dst_argb,
+                        const float* src_dudv,
+                        int width);
+
+// Used for I420Scale, ARGBScale, and ARGBInterpolate.
+void InterpolateRow_C(uint8_t* dst_ptr,
+                      const uint8_t* src_ptr,
+                      ptrdiff_t src_stride,
+                      int width,
+                      int source_y_fraction);
+void InterpolateRow_SSSE3(uint8_t* dst_ptr,
+                          const uint8_t* src_ptr,
+                          ptrdiff_t src_stride,
+                          int dst_width,
+                          int source_y_fraction);
+void InterpolateRow_AVX2(uint8_t* dst_ptr,
+                         const uint8_t* src_ptr,
+                         ptrdiff_t src_stride,
+                         int dst_width,
+                         int source_y_fraction);
+void InterpolateRow_NEON(uint8_t* dst_ptr,
+                         const uint8_t* src_ptr,
+                         ptrdiff_t src_stride,
+                         int dst_width,
+                         int source_y_fraction);
+void InterpolateRow_MSA(uint8_t* dst_ptr,
+                        const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        int width,
+                        int source_y_fraction);
+void InterpolateRow_LSX(uint8_t* dst_ptr,
+                        const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        int width,
+                        int source_y_fraction);
+void InterpolateRow_RVV(uint8_t* dst_ptr,
+                        const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        int width,
+                        int source_y_fraction);
+void InterpolateRow_Any_NEON(uint8_t* dst_ptr,
+                             const uint8_t* src_ptr,
+                             ptrdiff_t src_stride_ptr,
+                             int width,
+                             int source_y_fraction);
+void InterpolateRow_Any_SSSE3(uint8_t* dst_ptr,
+                              const uint8_t* src_ptr,
+                              ptrdiff_t src_stride_ptr,
+                              int width,
+                              int source_y_fraction);
+void InterpolateRow_Any_AVX2(uint8_t* dst_ptr,
+                             const uint8_t* src_ptr,
+                             ptrdiff_t src_stride_ptr,
+                             int width,
+                             int source_y_fraction);
+void InterpolateRow_Any_MSA(uint8_t* dst_ptr,
+                            const uint8_t* src_ptr,
+                            ptrdiff_t src_stride_ptr,
+                            int width,
+                            int source_y_fraction);
+void InterpolateRow_Any_LSX(uint8_t* dst_ptr,
+                            const uint8_t* src_ptr,
+                            ptrdiff_t src_stride_ptr,
+                            int width,
+                            int source_y_fraction);
+
+void InterpolateRow_16_C(uint16_t* dst_ptr,
+                         const uint16_t* src_ptr,
+                         ptrdiff_t src_stride,
+                         int width,
+                         int source_y_fraction);
+void InterpolateRow_16_NEON(uint16_t* dst_ptr,
+                            const uint16_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            int width,
+                            int source_y_fraction);
+void InterpolateRow_16_Any_NEON(uint16_t* dst_ptr,
+                                const uint16_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                int width,
+                                int source_y_fraction);
+
+void InterpolateRow_16To8_C(uint8_t* dst_ptr,
+                            const uint16_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            int scale,
+                            int width,
+                            int source_y_fraction);
+void InterpolateRow_16To8_NEON(uint8_t* dst_ptr,
+                               const uint16_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               int scale,
+                               int width,
+                               int source_y_fraction);
+void InterpolateRow_16To8_Any_NEON(uint8_t* dst_ptr,
+                                   const uint16_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   int scale,
+                                   int width,
+                                   int source_y_fraction);
+void InterpolateRow_16To8_AVX2(uint8_t* dst_ptr,
+                               const uint16_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               int scale,
+                               int width,
+                               int source_y_fraction);
+void InterpolateRow_16To8_Any_AVX2(uint8_t* dst_ptr,
+                                   const uint16_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   int scale,
+                                   int width,
+                                   int source_y_fraction);
+
+// Sobel images.
+void SobelXRow_C(const uint8_t* src_y0,
+                 const uint8_t* src_y1,
+                 const uint8_t* src_y2,
+                 uint8_t* dst_sobelx,
+                 int width);
+void SobelXRow_SSE2(const uint8_t* src_y0,
+                    const uint8_t* src_y1,
+                    const uint8_t* src_y2,
+                    uint8_t* dst_sobelx,
+                    int width);
+void SobelXRow_NEON(const uint8_t* src_y0,
+                    const uint8_t* src_y1,
+                    const uint8_t* src_y2,
+                    uint8_t* dst_sobelx,
+                    int width);
+void SobelXRow_MSA(const uint8_t* src_y0,
+                   const uint8_t* src_y1,
+                   const uint8_t* src_y2,
+                   uint8_t* dst_sobelx,
+                   int width);
+void SobelYRow_C(const uint8_t* src_y0,
+                 const uint8_t* src_y1,
+                 uint8_t* dst_sobely,
+                 int width);
+void SobelYRow_SSE2(const uint8_t* src_y0,
+                    const uint8_t* src_y1,
+                    uint8_t* dst_sobely,
+                    int width);
+void SobelYRow_NEON(const uint8_t* src_y0,
+                    const uint8_t* src_y1,
+                    uint8_t* dst_sobely,
+                    int width);
+void SobelYRow_MSA(const uint8_t* src_y0,
+                   const uint8_t* src_y1,
+                   uint8_t* dst_sobely,
+                   int width);
+void SobelRow_C(const uint8_t* src_sobelx,
+                const uint8_t* src_sobely,
+                uint8_t* dst_argb,
+                int width);
+void SobelRow_SSE2(const uint8_t* src_sobelx,
+                   const uint8_t* src_sobely,
+                   uint8_t* dst_argb,
+                   int width);
+void SobelRow_NEON(const uint8_t* src_sobelx,
+                   const uint8_t* src_sobely,
+                   uint8_t* dst_argb,
+                   int width);
+void SobelRow_MSA(const uint8_t* src_sobelx,
+                  const uint8_t* src_sobely,
+                  uint8_t* dst_argb,
+                  int width);
+void SobelRow_LSX(const uint8_t* src_sobelx,
+                  const uint8_t* src_sobely,
+                  uint8_t* dst_argb,
+                  int width);
+void SobelToPlaneRow_C(const uint8_t* src_sobelx,
+                       const uint8_t* src_sobely,
+                       uint8_t* dst_y,
+                       int width);
+void SobelToPlaneRow_SSE2(const uint8_t* src_sobelx,
+                          const uint8_t* src_sobely,
+                          uint8_t* dst_y,
+                          int width);
+void SobelToPlaneRow_NEON(const uint8_t* src_sobelx,
+                          const uint8_t* src_sobely,
+                          uint8_t* dst_y,
+                          int width);
+void SobelToPlaneRow_MSA(const uint8_t* src_sobelx,
+                         const uint8_t* src_sobely,
+                         uint8_t* dst_y,
+                         int width);
+void SobelToPlaneRow_LSX(const uint8_t* src_sobelx,
+                         const uint8_t* src_sobely,
+                         uint8_t* dst_y,
+                         int width);
+void SobelXYRow_C(const uint8_t* src_sobelx,
+                  const uint8_t* src_sobely,
+                  uint8_t* dst_argb,
+                  int width);
+void SobelXYRow_SSE2(const uint8_t* src_sobelx,
+                     const uint8_t* src_sobely,
+                     uint8_t* dst_argb,
+                     int width);
+void SobelXYRow_NEON(const uint8_t* src_sobelx,
+                     const uint8_t* src_sobely,
+                     uint8_t* dst_argb,
+                     int width);
+void SobelXYRow_MSA(const uint8_t* src_sobelx,
+                    const uint8_t* src_sobely,
+                    uint8_t* dst_argb,
+                    int width);
+void SobelXYRow_LSX(const uint8_t* src_sobelx,
+                    const uint8_t* src_sobely,
+                    uint8_t* dst_argb,
+                    int width);
+void SobelRow_Any_SSE2(const uint8_t* y_buf,
+                       const uint8_t* uv_buf,
+                       uint8_t* dst_ptr,
+                       int width);
+void SobelRow_Any_NEON(const uint8_t* y_buf,
+                       const uint8_t* uv_buf,
+                       uint8_t* dst_ptr,
+                       int width);
+void SobelRow_Any_MSA(const uint8_t* y_buf,
+                      const uint8_t* uv_buf,
+                      uint8_t* dst_ptr,
+                      int width);
+void SobelRow_Any_LSX(const uint8_t* y_buf,
+                      const uint8_t* uv_buf,
+                      uint8_t* dst_ptr,
+                      int width);
+void SobelToPlaneRow_Any_SSE2(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              int width);
+void SobelToPlaneRow_Any_NEON(const uint8_t* y_buf,
+                              const uint8_t* uv_buf,
+                              uint8_t* dst_ptr,
+                              int width);
+void SobelToPlaneRow_Any_MSA(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             int width);
+void SobelToPlaneRow_Any_LSX(const uint8_t* y_buf,
+                             const uint8_t* uv_buf,
+                             uint8_t* dst_ptr,
+                             int width);
+void SobelXYRow_Any_SSE2(const uint8_t* y_buf,
+                         const uint8_t* uv_buf,
+                         uint8_t* dst_ptr,
+                         int width);
+void SobelXYRow_Any_NEON(const uint8_t* y_buf,
+                         const uint8_t* uv_buf,
+                         uint8_t* dst_ptr,
+                         int width);
+void SobelXYRow_Any_MSA(const uint8_t* y_buf,
+                        const uint8_t* uv_buf,
+                        uint8_t* dst_ptr,
+                        int width);
+void SobelXYRow_Any_LSX(const uint8_t* y_buf,
+                        const uint8_t* uv_buf,
+                        uint8_t* dst_ptr,
+                        int width);
+
+void ARGBPolynomialRow_C(const uint8_t* src_argb,
+                         uint8_t* dst_argb,
+                         const float* poly,
+                         int width);
+void ARGBPolynomialRow_SSE2(const uint8_t* src_argb,
+                            uint8_t* dst_argb,
+                            const float* poly,
+                            int width);
+void ARGBPolynomialRow_AVX2(const uint8_t* src_argb,
+                            uint8_t* dst_argb,
+                            const float* poly,
+                            int width);
+
+// Scale and convert to half float.
+void HalfFloatRow_C(const uint16_t* src, uint16_t* dst, float scale, int width);
+void HalfFloatRow_SSE2(const uint16_t* src,
+                       uint16_t* dst,
+                       float scale,
+                       int width);
+void HalfFloatRow_Any_SSE2(const uint16_t* src_ptr,
+                           uint16_t* dst_ptr,
+                           float param,
+                           int width);
+void HalfFloatRow_AVX2(const uint16_t* src,
+                       uint16_t* dst,
+                       float scale,
+                       int width);
+void HalfFloatRow_Any_AVX2(const uint16_t* src_ptr,
+                           uint16_t* dst_ptr,
+                           float param,
+                           int width);
+void HalfFloatRow_F16C(const uint16_t* src,
+                       uint16_t* dst,
+                       float scale,
+                       int width);
+void HalfFloatRow_Any_F16C(const uint16_t* src,
+                           uint16_t* dst,
+                           float scale,
+                           int width);
+void HalfFloat1Row_F16C(const uint16_t* src,
+                        uint16_t* dst,
+                        float scale,
+                        int width);
+void HalfFloat1Row_Any_F16C(const uint16_t* src,
+                            uint16_t* dst,
+                            float scale,
+                            int width);
+void HalfFloatRow_NEON(const uint16_t* src,
+                       uint16_t* dst,
+                       float scale,
+                       int width);
+void HalfFloatRow_Any_NEON(const uint16_t* src_ptr,
+                           uint16_t* dst_ptr,
+                           float param,
+                           int width);
+void HalfFloat1Row_NEON(const uint16_t* src,
+                        uint16_t* dst,
+                        float scale,
+                        int width);
+void HalfFloat1Row_Any_NEON(const uint16_t* src_ptr,
+                            uint16_t* dst_ptr,
+                            float param,
+                            int width);
+void HalfFloatRow_MSA(const uint16_t* src,
+                      uint16_t* dst,
+                      float scale,
+                      int width);
+void HalfFloatRow_Any_MSA(const uint16_t* src_ptr,
+                          uint16_t* dst_ptr,
+                          float param,
+                          int width);
+void HalfFloatRow_LSX(const uint16_t* src,
+                      uint16_t* dst,
+                      float scale,
+                      int width);
+void HalfFloatRow_Any_LSX(const uint16_t* src_ptr,
+                          uint16_t* dst_ptr,
+                          float param,
+                          int width);
+void ByteToFloatRow_C(const uint8_t* src, float* dst, float scale, int width);
+void ByteToFloatRow_NEON(const uint8_t* src,
+                         float* dst,
+                         float scale,
+                         int width);
+void ByteToFloatRow_Any_NEON(const uint8_t* src_ptr,
+                             float* dst_ptr,
+                             float param,
+                             int width);
+// Convert FP16 Half Floats to FP32 Floats
+void ConvertFP16ToFP32Row_NEON(const uint16_t* src,  // fp16
+                               float* dst,
+                               int width);
+// Convert a column of FP16 Half Floats to a row of FP32 Floats
+void ConvertFP16ToFP32Column_NEON(const uint16_t* src,  // fp16
+                                  int src_stride,       // stride in elements
+                                  float* dst,
+                                  int width);
+// Convert FP32 Floats to FP16 Half Floats
+void ConvertFP32ToFP16Row_NEON(const float* src,
+                               uint16_t* dst,  // fp16
+                               int width);
+void ARGBLumaColorTableRow_C(const uint8_t* src_argb,
+                             uint8_t* dst_argb,
+                             int width,
+                             const uint8_t* luma,
+                             uint32_t lumacoeff);
+void ARGBLumaColorTableRow_SSSE3(const uint8_t* src_argb,
+                                 uint8_t* dst_argb,
+                                 int width,
+                                 const uint8_t* luma,
+                                 uint32_t lumacoeff);
+
+float ScaleMaxSamples_C(const float* src, float* dst, float scale, int width);
+float ScaleMaxSamples_NEON(const float* src,
+                           float* dst,
+                           float scale,
+                           int width);
+float ScaleSumSamples_C(const float* src, float* dst, float scale, int width);
+float ScaleSumSamples_NEON(const float* src,
+                           float* dst,
+                           float scale,
+                           int width);
+void ScaleSamples_C(const float* src, float* dst, float scale, int width);
+void ScaleSamples_NEON(const float* src, float* dst, float scale, int width);
+
+void GaussRow_F32_NEON(const float* src, float* dst, int width);
+void GaussRow_F32_C(const float* src, float* dst, int width);
+
+void GaussCol_F32_NEON(const float* src0,
+                       const float* src1,
+                       const float* src2,
+                       const float* src3,
+                       const float* src4,
+                       float* dst,
+                       int width);
+
+void GaussCol_F32_C(const float* src0,
+                    const float* src1,
+                    const float* src2,
+                    const float* src3,
+                    const float* src4,
+                    float* dst,
+                    int width);
+
+void GaussRow_C(const uint32_t* src, uint16_t* dst, int width);
+void GaussRow_NEON(const uint32_t* src, uint16_t* dst, int width);
+void GaussCol_C(const uint16_t* src0,
+                const uint16_t* src1,
+                const uint16_t* src2,
+                const uint16_t* src3,
+                const uint16_t* src4,
+                uint32_t* dst,
+                int width);
+void GaussCol_NEON(const uint16_t* src0,
+                   const uint16_t* src1,
+                   const uint16_t* src2,
+                   const uint16_t* src3,
+                   const uint16_t* src4,
+                   uint32_t* dst,
+                   int width);
+
+void ClampFloatToZero_SSE2(const float* src_x, float* dst_y, int width);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_ROW_H_

yuv.mod/libyuv/include/libyuv/scale.h

@@ -0,0 +1,321 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_SCALE_H_
+#define INCLUDE_LIBYUV_SCALE_H_
+
+#include "libyuv/basic_types.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Supported filtering.
+typedef enum FilterMode {
+  kFilterNone = 0,      // Point sample; Fastest.
+  kFilterLinear = 1,    // Filter horizontally only.
+  kFilterBilinear = 2,  // Faster than box, but lower quality scaling down.
+  kFilterBox = 3        // Highest quality.
+} FilterModeEnum;
+
+// Scale a YUV plane.
+// Returns 0 if successful.
+LIBYUV_API
+int ScalePlane(const uint8_t* src,
+               int src_stride,
+               int src_width,
+               int src_height,
+               uint8_t* dst,
+               int dst_stride,
+               int dst_width,
+               int dst_height,
+               enum FilterMode filtering);
+
+LIBYUV_API
+int ScalePlane_16(const uint16_t* src,
+                  int src_stride,
+                  int src_width,
+                  int src_height,
+                  uint16_t* dst,
+                  int dst_stride,
+                  int dst_width,
+                  int dst_height,
+                  enum FilterMode filtering);
+
+// Sample is expected to be in the low 12 bits.
+LIBYUV_API
+int ScalePlane_12(const uint16_t* src,
+                  int src_stride,
+                  int src_width,
+                  int src_height,
+                  uint16_t* dst,
+                  int dst_stride,
+                  int dst_width,
+                  int dst_height,
+                  enum FilterMode filtering);
+
+// Scales a YUV 4:2:0 image from the src width and height to the
+// dst width and height.
+// If filtering is kFilterNone, a simple nearest-neighbor algorithm is
+// used. This produces basic (blocky) quality at the fastest speed.
+// If filtering is kFilterBilinear, interpolation is used to produce a better
+// quality image, at the expense of speed.
+// If filtering is kFilterBox, averaging is used to produce ever better
+// quality image, at further expense of speed.
+// Returns 0 if successful.
+
+LIBYUV_API
+int I420Scale(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              int src_width,
+              int src_height,
+              uint8_t* dst_y,
+              int dst_stride_y,
+              uint8_t* dst_u,
+              int dst_stride_u,
+              uint8_t* dst_v,
+              int dst_stride_v,
+              int dst_width,
+              int dst_height,
+              enum FilterMode filtering);
+
+LIBYUV_API
+int I420Scale_16(const uint16_t* src_y,
+                 int src_stride_y,
+                 const uint16_t* src_u,
+                 int src_stride_u,
+                 const uint16_t* src_v,
+                 int src_stride_v,
+                 int src_width,
+                 int src_height,
+                 uint16_t* dst_y,
+                 int dst_stride_y,
+                 uint16_t* dst_u,
+                 int dst_stride_u,
+                 uint16_t* dst_v,
+                 int dst_stride_v,
+                 int dst_width,
+                 int dst_height,
+                 enum FilterMode filtering);
+
+LIBYUV_API
+int I420Scale_12(const uint16_t* src_y,
+                 int src_stride_y,
+                 const uint16_t* src_u,
+                 int src_stride_u,
+                 const uint16_t* src_v,
+                 int src_stride_v,
+                 int src_width,
+                 int src_height,
+                 uint16_t* dst_y,
+                 int dst_stride_y,
+                 uint16_t* dst_u,
+                 int dst_stride_u,
+                 uint16_t* dst_v,
+                 int dst_stride_v,
+                 int dst_width,
+                 int dst_height,
+                 enum FilterMode filtering);
+
+// Scales a YUV 4:4:4 image from the src width and height to the
+// dst width and height.
+// If filtering is kFilterNone, a simple nearest-neighbor algorithm is
+// used. This produces basic (blocky) quality at the fastest speed.
+// If filtering is kFilterBilinear, interpolation is used to produce a better
+// quality image, at the expense of speed.
+// If filtering is kFilterBox, averaging is used to produce ever better
+// quality image, at further expense of speed.
+// Returns 0 if successful.
+
+LIBYUV_API
+int I444Scale(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              int src_width,
+              int src_height,
+              uint8_t* dst_y,
+              int dst_stride_y,
+              uint8_t* dst_u,
+              int dst_stride_u,
+              uint8_t* dst_v,
+              int dst_stride_v,
+              int dst_width,
+              int dst_height,
+              enum FilterMode filtering);
+
+LIBYUV_API
+int I444Scale_16(const uint16_t* src_y,
+                 int src_stride_y,
+                 const uint16_t* src_u,
+                 int src_stride_u,
+                 const uint16_t* src_v,
+                 int src_stride_v,
+                 int src_width,
+                 int src_height,
+                 uint16_t* dst_y,
+                 int dst_stride_y,
+                 uint16_t* dst_u,
+                 int dst_stride_u,
+                 uint16_t* dst_v,
+                 int dst_stride_v,
+                 int dst_width,
+                 int dst_height,
+                 enum FilterMode filtering);
+
+LIBYUV_API
+int I444Scale_12(const uint16_t* src_y,
+                 int src_stride_y,
+                 const uint16_t* src_u,
+                 int src_stride_u,
+                 const uint16_t* src_v,
+                 int src_stride_v,
+                 int src_width,
+                 int src_height,
+                 uint16_t* dst_y,
+                 int dst_stride_y,
+                 uint16_t* dst_u,
+                 int dst_stride_u,
+                 uint16_t* dst_v,
+                 int dst_stride_v,
+                 int dst_width,
+                 int dst_height,
+                 enum FilterMode filtering);
+
+// Scales a YUV 4:2:2 image from the src width and height to the
+// dst width and height.
+// If filtering is kFilterNone, a simple nearest-neighbor algorithm is
+// used. This produces basic (blocky) quality at the fastest speed.
+// If filtering is kFilterBilinear, interpolation is used to produce a better
+// quality image, at the expense of speed.
+// If filtering is kFilterBox, averaging is used to produce ever better
+// quality image, at further expense of speed.
+// Returns 0 if successful.
+LIBYUV_API
+int I422Scale(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              int src_width,
+              int src_height,
+              uint8_t* dst_y,
+              int dst_stride_y,
+              uint8_t* dst_u,
+              int dst_stride_u,
+              uint8_t* dst_v,
+              int dst_stride_v,
+              int dst_width,
+              int dst_height,
+              enum FilterMode filtering);
+
+LIBYUV_API
+int I422Scale_16(const uint16_t* src_y,
+                 int src_stride_y,
+                 const uint16_t* src_u,
+                 int src_stride_u,
+                 const uint16_t* src_v,
+                 int src_stride_v,
+                 int src_width,
+                 int src_height,
+                 uint16_t* dst_y,
+                 int dst_stride_y,
+                 uint16_t* dst_u,
+                 int dst_stride_u,
+                 uint16_t* dst_v,
+                 int dst_stride_v,
+                 int dst_width,
+                 int dst_height,
+                 enum FilterMode filtering);
+
+LIBYUV_API
+int I422Scale_12(const uint16_t* src_y,
+                 int src_stride_y,
+                 const uint16_t* src_u,
+                 int src_stride_u,
+                 const uint16_t* src_v,
+                 int src_stride_v,
+                 int src_width,
+                 int src_height,
+                 uint16_t* dst_y,
+                 int dst_stride_y,
+                 uint16_t* dst_u,
+                 int dst_stride_u,
+                 uint16_t* dst_v,
+                 int dst_stride_v,
+                 int dst_width,
+                 int dst_height,
+                 enum FilterMode filtering);
+
+// Scales an NV12 image from the src width and height to the
+// dst width and height.
+// If filtering is kFilterNone, a simple nearest-neighbor algorithm is
+// used. This produces basic (blocky) quality at the fastest speed.
+// If filtering is kFilterBilinear, interpolation is used to produce a better
+// quality image, at the expense of speed.
+// kFilterBox is not supported for the UV channel and will be treated as
+// bilinear.
+// Returns 0 if successful.
+
+LIBYUV_API
+int NV12Scale(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_uv,
+              int src_stride_uv,
+              int src_width,
+              int src_height,
+              uint8_t* dst_y,
+              int dst_stride_y,
+              uint8_t* dst_uv,
+              int dst_stride_uv,
+              int dst_width,
+              int dst_height,
+              enum FilterMode filtering);
+
+#ifdef __cplusplus
+// Legacy API.  Deprecated.
+LIBYUV_API
+int Scale(const uint8_t* src_y,
+          const uint8_t* src_u,
+          const uint8_t* src_v,
+          int src_stride_y,
+          int src_stride_u,
+          int src_stride_v,
+          int src_width,
+          int src_height,
+          uint8_t* dst_y,
+          uint8_t* dst_u,
+          uint8_t* dst_v,
+          int dst_stride_y,
+          int dst_stride_u,
+          int dst_stride_v,
+          int dst_width,
+          int dst_height,
+          LIBYUV_BOOL interpolate);
+
+// For testing, allow disabling of specialized scalers.
+LIBYUV_API
+void SetUseReferenceImpl(LIBYUV_BOOL use);
+#endif  // __cplusplus
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_SCALE_H_

yuv.mod/libyuv/include/libyuv/scale_argb.h

@@ -0,0 +1,76 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_SCALE_ARGB_H_
+#define INCLUDE_LIBYUV_SCALE_ARGB_H_
+
+#include "libyuv/basic_types.h"
+#include "libyuv/scale.h"  // For FilterMode
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+LIBYUV_API
+int ARGBScale(const uint8_t* src_argb,
+              int src_stride_argb,
+              int src_width,
+              int src_height,
+              uint8_t* dst_argb,
+              int dst_stride_argb,
+              int dst_width,
+              int dst_height,
+              enum FilterMode filtering);
+
+// Clipped scale takes destination rectangle coordinates for clip values.
+LIBYUV_API
+int ARGBScaleClip(const uint8_t* src_argb,
+                  int src_stride_argb,
+                  int src_width,
+                  int src_height,
+                  uint8_t* dst_argb,
+                  int dst_stride_argb,
+                  int dst_width,
+                  int dst_height,
+                  int clip_x,
+                  int clip_y,
+                  int clip_width,
+                  int clip_height,
+                  enum FilterMode filtering);
+
+// Scale with YUV conversion to ARGB and clipping.
+LIBYUV_API
+int YUVToARGBScaleClip(const uint8_t* src_y,
+                       int src_stride_y,
+                       const uint8_t* src_u,
+                       int src_stride_u,
+                       const uint8_t* src_v,
+                       int src_stride_v,
+                       uint32_t src_fourcc,
+                       int src_width,
+                       int src_height,
+                       uint8_t* dst_argb,
+                       int dst_stride_argb,
+                       uint32_t dst_fourcc,
+                       int dst_width,
+                       int dst_height,
+                       int clip_x,
+                       int clip_y,
+                       int clip_width,
+                       int clip_height,
+                       enum FilterMode filtering);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_SCALE_ARGB_H_

yuv.mod/libyuv/include/libyuv/scale_rgb.h

@@ -0,0 +1,42 @@
+/*
+ *  Copyright 2022 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_SCALE_RGB_H_
+#define INCLUDE_LIBYUV_SCALE_RGB_H_
+
+#include "libyuv/basic_types.h"
+#include "libyuv/scale.h"  // For FilterMode
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// RGB can be RAW, RGB24 or YUV24
+// RGB scales 24 bit images by converting a row at a time to ARGB
+// and using ARGB row functions to scale, then convert to RGB.
+// TODO(fbarchard): Allow input/output formats to be specified.
+LIBYUV_API
+int RGBScale(const uint8_t* src_rgb,
+             int src_stride_rgb,
+             int src_width,
+             int src_height,
+             uint8_t* dst_rgb,
+             int dst_stride_rgb,
+             int dst_width,
+             int dst_height,
+             enum FilterMode filtering);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_SCALE_UV_H_

yuv.mod/libyuv/include/libyuv/scale_row.h

@@ -0,0 +1,1898 @@
+/*
+ *  Copyright 2013 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_SCALE_ROW_H_
+#define INCLUDE_LIBYUV_SCALE_ROW_H_
+
+#include "libyuv/basic_types.h"
+#include "libyuv/scale.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+#if defined(__pnacl__) || defined(__CLR_VER) ||            \
+    (defined(__native_client__) && defined(__x86_64__)) || \
+    (defined(__i386__) && !defined(__SSE__) && !defined(__clang__))
+#define LIBYUV_DISABLE_X86
+#endif
+#if defined(__native_client__)
+#define LIBYUV_DISABLE_NEON
+#endif
+// MemorySanitizer does not support assembly code yet. http://crbug.com/344505
+#if defined(__has_feature)
+#if __has_feature(memory_sanitizer) && !defined(LIBYUV_DISABLE_NEON)
+#define LIBYUV_DISABLE_NEON
+#endif
+#if __has_feature(memory_sanitizer) && !defined(LIBYUV_DISABLE_X86)
+#define LIBYUV_DISABLE_X86
+#endif
+#endif
+// GCC >= 4.7.0 required for AVX2.
+#if defined(__GNUC__) && (defined(__x86_64__) || defined(__i386__))
+#if (__GNUC__ > 4) || (__GNUC__ == 4 && (__GNUC_MINOR__ >= 7))
+#define GCC_HAS_AVX2 1
+#endif  // GNUC >= 4.7
+#endif  // __GNUC__
+
+// clang >= 3.4.0 required for AVX2.
+#if defined(__clang__) && (defined(__x86_64__) || defined(__i386__))
+#if (__clang_major__ > 3) || (__clang_major__ == 3 && (__clang_minor__ >= 4))
+#define CLANG_HAS_AVX2 1
+#endif  // clang >= 3.4
+#endif  // __clang__
+
+// Visual C 2012 required for AVX2.
+#if defined(_M_IX86) && !defined(__clang__) && defined(_MSC_VER) && \
+    _MSC_VER >= 1700
+#define VISUALC_HAS_AVX2 1
+#endif  // VisualStudio >= 2012
+
+// The following are available on all x86 platforms:
+#if !defined(LIBYUV_DISABLE_X86) && \
+    (defined(_M_IX86) || defined(__x86_64__) || defined(__i386__))
+#define HAS_FIXEDDIV1_X86
+#define HAS_FIXEDDIV_X86
+#define HAS_SCALEADDROW_SSE2
+#define HAS_SCALEARGBCOLS_SSE2
+#define HAS_SCALEARGBCOLSUP2_SSE2
+#define HAS_SCALEARGBFILTERCOLS_SSSE3
+#define HAS_SCALEARGBROWDOWN2_SSE2
+#define HAS_SCALEARGBROWDOWNEVEN_SSE2
+#define HAS_SCALECOLSUP2_SSE2
+#define HAS_SCALEFILTERCOLS_SSSE3
+#define HAS_SCALEROWDOWN2_SSSE3
+#define HAS_SCALEROWDOWN34_SSSE3
+#define HAS_SCALEROWDOWN38_SSSE3
+#define HAS_SCALEROWDOWN4_SSSE3
+#endif
+
+// The following are available for gcc/clang x86 platforms:
+// TODO(fbarchard): Port to Visual C
+#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__))
+#define HAS_SCALEUVROWDOWN2BOX_SSSE3
+#define HAS_SCALEROWUP2_LINEAR_SSE2
+#define HAS_SCALEROWUP2_LINEAR_SSSE3
+#define HAS_SCALEROWUP2_BILINEAR_SSE2
+#define HAS_SCALEROWUP2_BILINEAR_SSSE3
+#define HAS_SCALEROWUP2_LINEAR_12_SSSE3
+#define HAS_SCALEROWUP2_BILINEAR_12_SSSE3
+#define HAS_SCALEROWUP2_LINEAR_16_SSE2
+#define HAS_SCALEROWUP2_BILINEAR_16_SSE2
+#define HAS_SCALEUVROWUP2_LINEAR_SSSE3
+#define HAS_SCALEUVROWUP2_BILINEAR_SSSE3
+#define HAS_SCALEUVROWUP2_LINEAR_16_SSE41
+#define HAS_SCALEUVROWUP2_BILINEAR_16_SSE41
+#endif
+
+// The following are available for gcc/clang x86 platforms, but
+// require clang 3.4 or gcc 4.7.
+// TODO(fbarchard): Port to Visual C
+#if !defined(LIBYUV_DISABLE_X86) &&               \
+    (defined(__x86_64__) || defined(__i386__)) && \
+    (defined(CLANG_HAS_AVX2) || defined(GCC_HAS_AVX2))
+#define HAS_SCALEUVROWDOWN2BOX_AVX2
+#define HAS_SCALEROWUP2_LINEAR_AVX2
+#define HAS_SCALEROWUP2_BILINEAR_AVX2
+#define HAS_SCALEROWUP2_LINEAR_12_AVX2
+#define HAS_SCALEROWUP2_BILINEAR_12_AVX2
+#define HAS_SCALEROWUP2_LINEAR_16_AVX2
+#define HAS_SCALEROWUP2_BILINEAR_16_AVX2
+#define HAS_SCALEUVROWUP2_LINEAR_AVX2
+#define HAS_SCALEUVROWUP2_BILINEAR_AVX2
+#define HAS_SCALEUVROWUP2_LINEAR_16_AVX2
+#define HAS_SCALEUVROWUP2_BILINEAR_16_AVX2
+#endif
+
+// The following are available on all x86 platforms, but
+// require VS2012, clang 3.4 or gcc 4.7.
+// The code supports NaCL but requires a new compiler and validator.
+#if !defined(LIBYUV_DISABLE_X86) &&                          \
+    (defined(VISUALC_HAS_AVX2) || defined(CLANG_HAS_AVX2) || \
+     defined(GCC_HAS_AVX2))
+#define HAS_SCALEADDROW_AVX2
+#define HAS_SCALEROWDOWN2_AVX2
+#define HAS_SCALEROWDOWN4_AVX2
+#endif
+
+// The following are available on Neon platforms:
+#if !defined(LIBYUV_DISABLE_NEON) && \
+    (defined(__ARM_NEON__) || defined(LIBYUV_NEON) || defined(__aarch64__))
+#define HAS_SCALEADDROW_NEON
+#define HAS_SCALEARGBCOLS_NEON
+#define HAS_SCALEARGBFILTERCOLS_NEON
+#define HAS_SCALEARGBROWDOWN2_NEON
+#define HAS_SCALEARGBROWDOWNEVEN_NEON
+#define HAS_SCALEFILTERCOLS_NEON
+#define HAS_SCALEROWDOWN2_NEON
+#define HAS_SCALEROWDOWN34_NEON
+#define HAS_SCALEROWDOWN38_NEON
+#define HAS_SCALEROWDOWN4_NEON
+#define HAS_SCALEUVROWDOWN2_NEON
+#define HAS_SCALEUVROWDOWN2LINEAR_NEON
+#define HAS_SCALEUVROWDOWN2BOX_NEON
+#define HAS_SCALEUVROWDOWNEVEN_NEON
+#define HAS_SCALEROWUP2_LINEAR_NEON
+#define HAS_SCALEROWUP2_BILINEAR_NEON
+#define HAS_SCALEROWUP2_LINEAR_12_NEON
+#define HAS_SCALEROWUP2_BILINEAR_12_NEON
+#define HAS_SCALEROWUP2_LINEAR_16_NEON
+#define HAS_SCALEROWUP2_BILINEAR_16_NEON
+#define HAS_SCALEUVROWUP2_LINEAR_NEON
+#define HAS_SCALEUVROWUP2_BILINEAR_NEON
+#define HAS_SCALEUVROWUP2_LINEAR_16_NEON
+#define HAS_SCALEUVROWUP2_BILINEAR_16_NEON
+#endif
+
+#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
+#define HAS_SCALEADDROW_MSA
+#define HAS_SCALEARGBCOLS_MSA
+#define HAS_SCALEARGBFILTERCOLS_MSA
+#define HAS_SCALEARGBROWDOWN2_MSA
+#define HAS_SCALEARGBROWDOWNEVEN_MSA
+#define HAS_SCALEFILTERCOLS_MSA
+#define HAS_SCALEROWDOWN2_MSA
+#define HAS_SCALEROWDOWN34_MSA
+#define HAS_SCALEROWDOWN38_MSA
+#define HAS_SCALEROWDOWN4_MSA
+#endif
+
+#if !defined(LIBYUV_DISABLE_LSX) && defined(__loongarch_sx)
+#define HAS_SCALEARGBROWDOWN2_LSX
+#define HAS_SCALEARGBROWDOWNEVEN_LSX
+#define HAS_SCALEROWDOWN2_LSX
+#define HAS_SCALEROWDOWN4_LSX
+#define HAS_SCALEROWDOWN38_LSX
+#define HAS_SCALEFILTERCOLS_LSX
+#define HAS_SCALEADDROW_LSX
+#define HAS_SCALEARGBCOLS_LSX
+#define HAS_SCALEARGBFILTERCOLS_LSX
+#define HAS_SCALEROWDOWN34_LSX
+#endif
+
+#if !defined(LIBYUV_DISABLE_RVV) && defined(__riscv_vector)
+#define HAS_SCALEADDROW_RVV
+// TODO: Test ScaleARGBRowDownEven_RVV and enable it
+// #define HAS_SCALEARGBROWDOWNEVEN_RVV
+#if defined(__riscv_zve64x)
+#define HAS_SCALEUVROWDOWN4_RVV
+#endif
+#define HAS_SCALEUVROWDOWNEVEN_RVV
+#define HAS_SCALEARGBROWDOWN2_RVV
+#define HAS_SCALEARGBROWDOWN2BOX_RVV
+#define HAS_SCALEARGBROWDOWN2LINEAR_RVV
+#define HAS_SCALEARGBROWDOWNEVENBOX_RVV
+#define HAS_SCALEROWDOWN2_RVV
+#define HAS_SCALEROWDOWN2BOX_RVV
+#define HAS_SCALEROWDOWN2LINEAR_RVV
+#define HAS_SCALEROWDOWN34_0_BOX_RVV
+#define HAS_SCALEROWDOWN34_1_BOX_RVV
+#define HAS_SCALEROWDOWN34_RVV
+#define HAS_SCALEROWDOWN38_2_BOX_RVV
+#define HAS_SCALEROWDOWN38_3_BOX_RVV
+#define HAS_SCALEROWDOWN38_RVV
+#define HAS_SCALEROWDOWN4_RVV
+#define HAS_SCALEROWDOWN4BOX_RVV
+#define HAS_SCALEROWUP2_BILINEAR_RVV
+#define HAS_SCALEROWUP2_LINEAR_RVV
+#define HAS_SCALEUVROWDOWN2_RVV
+#define HAS_SCALEUVROWDOWN2BOX_RVV
+#define HAS_SCALEUVROWDOWN2LINEAR_RVV
+#define HAS_SCALEUVROWUP2_BILINEAR_RVV
+#define HAS_SCALEUVROWUP2_LINEAR_RVV
+#endif
+
+// Scale ARGB vertically with bilinear interpolation.
+void ScalePlaneVertical(int src_height,
+                        int dst_width,
+                        int dst_height,
+                        int src_stride,
+                        int dst_stride,
+                        const uint8_t* src_argb,
+                        uint8_t* dst_argb,
+                        int x,
+                        int y,
+                        int dy,
+                        int bpp,
+                        enum FilterMode filtering);
+
+void ScalePlaneVertical_16(int src_height,
+                           int dst_width,
+                           int dst_height,
+                           int src_stride,
+                           int dst_stride,
+                           const uint16_t* src_argb,
+                           uint16_t* dst_argb,
+                           int x,
+                           int y,
+                           int dy,
+                           int wpp,
+                           enum FilterMode filtering);
+
+void ScalePlaneVertical_16To8(int src_height,
+                              int dst_width,
+                              int dst_height,
+                              int src_stride,
+                              int dst_stride,
+                              const uint16_t* src_argb,
+                              uint8_t* dst_argb,
+                              int x,
+                              int y,
+                              int dy,
+                              int wpp,
+                              int scale,
+                              enum FilterMode filtering);
+
+void ScalePlaneDown2_16To8(int src_width,
+                           int src_height,
+                           int dst_width,
+                           int dst_height,
+                           int src_stride,
+                           int dst_stride,
+                           const uint16_t* src_ptr,
+                           uint8_t* dst_ptr,
+                           int scale,
+                           enum FilterMode filtering);
+
+// Simplify the filtering based on scale factors.
+enum FilterMode ScaleFilterReduce(int src_width,
+                                  int src_height,
+                                  int dst_width,
+                                  int dst_height,
+                                  enum FilterMode filtering);
+
+// Divide num by div and return as 16.16 fixed point result.
+int FixedDiv_C(int num, int div);
+int FixedDiv_X86(int num, int div);
+int FixedDiv_MIPS(int num, int div);
+// Divide num - 1 by div - 1 and return as 16.16 fixed point result.
+int FixedDiv1_C(int num, int div);
+int FixedDiv1_X86(int num, int div);
+int FixedDiv1_MIPS(int num, int div);
+#ifdef HAS_FIXEDDIV_X86
+#define FixedDiv FixedDiv_X86
+#define FixedDiv1 FixedDiv1_X86
+#elif defined HAS_FIXEDDIV_MIPS
+#define FixedDiv FixedDiv_MIPS
+#define FixedDiv1 FixedDiv1_MIPS
+#else
+#define FixedDiv FixedDiv_C
+#define FixedDiv1 FixedDiv1_C
+#endif
+
+// Compute slope values for stepping.
+void ScaleSlope(int src_width,
+                int src_height,
+                int dst_width,
+                int dst_height,
+                enum FilterMode filtering,
+                int* x,
+                int* y,
+                int* dx,
+                int* dy);
+
+void ScaleRowDown2_C(const uint8_t* src_ptr,
+                     ptrdiff_t src_stride,
+                     uint8_t* dst,
+                     int dst_width);
+void ScaleRowDown2_16_C(const uint16_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint16_t* dst,
+                        int dst_width);
+void ScaleRowDown2_16To8_C(const uint16_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst,
+                           int dst_width,
+                           int scale);
+void ScaleRowDown2_16To8_Odd_C(const uint16_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst,
+                               int dst_width,
+                               int scale);
+void ScaleRowDown2Linear_C(const uint8_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst,
+                           int dst_width);
+void ScaleRowDown2Linear_16_C(const uint16_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint16_t* dst,
+                              int dst_width);
+void ScaleRowDown2Linear_16To8_C(const uint16_t* src_ptr,
+                                 ptrdiff_t src_stride,
+                                 uint8_t* dst,
+                                 int dst_width,
+                                 int scale);
+void ScaleRowDown2Linear_16To8_Odd_C(const uint16_t* src_ptr,
+                                     ptrdiff_t src_stride,
+                                     uint8_t* dst,
+                                     int dst_width,
+                                     int scale);
+void ScaleRowDown2Box_C(const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint8_t* dst,
+                        int dst_width);
+void ScaleRowDown2Box_Odd_C(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst,
+                            int dst_width);
+void ScaleRowDown2Box_16_C(const uint16_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint16_t* dst,
+                           int dst_width);
+void ScaleRowDown2Box_16_NEON(const uint16_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint16_t* dst,
+                              int dst_width);
+void ScaleRowDown2Box_16To8_C(const uint16_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst,
+                              int dst_width,
+                              int scale);
+void ScaleRowDown2Box_16To8_Odd_C(const uint16_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst,
+                                  int dst_width,
+                                  int scale);
+void ScaleRowDown4_C(const uint8_t* src_ptr,
+                     ptrdiff_t src_stride,
+                     uint8_t* dst,
+                     int dst_width);
+void ScaleRowDown4_16_C(const uint16_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint16_t* dst,
+                        int dst_width);
+void ScaleRowDown4Box_C(const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint8_t* dst,
+                        int dst_width);
+void ScaleRowDown4Box_16_C(const uint16_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint16_t* dst,
+                           int dst_width);
+void ScaleRowDown34_C(const uint8_t* src_ptr,
+                      ptrdiff_t src_stride,
+                      uint8_t* dst,
+                      int dst_width);
+void ScaleRowDown34_16_C(const uint16_t* src_ptr,
+                         ptrdiff_t src_stride,
+                         uint16_t* dst,
+                         int dst_width);
+void ScaleRowDown34_0_Box_C(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* d,
+                            int dst_width);
+void ScaleRowDown34_0_Box_16_C(const uint16_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint16_t* d,
+                               int dst_width);
+void ScaleRowDown34_1_Box_C(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* d,
+                            int dst_width);
+void ScaleRowDown34_1_Box_16_C(const uint16_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint16_t* d,
+                               int dst_width);
+
+void ScaleRowUp2_Linear_C(const uint8_t* src_ptr,
+                          uint8_t* dst_ptr,
+                          int dst_width);
+void ScaleRowUp2_Bilinear_C(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            ptrdiff_t dst_stride,
+                            int dst_width);
+void ScaleRowUp2_Linear_16_C(const uint16_t* src_ptr,
+                             uint16_t* dst_ptr,
+                             int dst_width);
+void ScaleRowUp2_Bilinear_16_C(const uint16_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint16_t* dst_ptr,
+                               ptrdiff_t dst_stride,
+                               int dst_width);
+void ScaleRowUp2_Linear_Any_C(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowUp2_Bilinear_Any_C(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                ptrdiff_t dst_stride,
+                                int dst_width);
+void ScaleRowUp2_Linear_16_Any_C(const uint16_t* src_ptr,
+                                 uint16_t* dst_ptr,
+                                 int dst_width);
+void ScaleRowUp2_Bilinear_16_Any_C(const uint16_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint16_t* dst_ptr,
+                                   ptrdiff_t dst_stride,
+                                   int dst_width);
+
+void ScaleCols_C(uint8_t* dst_ptr,
+                 const uint8_t* src_ptr,
+                 int dst_width,
+                 int x,
+                 int dx);
+void ScaleCols_16_C(uint16_t* dst_ptr,
+                    const uint16_t* src_ptr,
+                    int dst_width,
+                    int x,
+                    int dx);
+void ScaleColsUp2_C(uint8_t* dst_ptr,
+                    const uint8_t* src_ptr,
+                    int dst_width,
+                    int,
+                    int);
+void ScaleColsUp2_16_C(uint16_t* dst_ptr,
+                       const uint16_t* src_ptr,
+                       int dst_width,
+                       int,
+                       int);
+void ScaleFilterCols_C(uint8_t* dst_ptr,
+                       const uint8_t* src_ptr,
+                       int dst_width,
+                       int x,
+                       int dx);
+void ScaleFilterCols_16_C(uint16_t* dst_ptr,
+                          const uint16_t* src_ptr,
+                          int dst_width,
+                          int x,
+                          int dx);
+void ScaleFilterCols64_C(uint8_t* dst_ptr,
+                         const uint8_t* src_ptr,
+                         int dst_width,
+                         int x32,
+                         int dx);
+void ScaleFilterCols64_16_C(uint16_t* dst_ptr,
+                            const uint16_t* src_ptr,
+                            int dst_width,
+                            int x32,
+                            int dx);
+void ScaleRowDown38_C(const uint8_t* src_ptr,
+                      ptrdiff_t src_stride,
+                      uint8_t* dst,
+                      int dst_width);
+void ScaleRowDown38_16_C(const uint16_t* src_ptr,
+                         ptrdiff_t src_stride,
+                         uint16_t* dst,
+                         int dst_width);
+void ScaleRowDown38_3_Box_C(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowDown38_3_Box_16_C(const uint16_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint16_t* dst_ptr,
+                               int dst_width);
+void ScaleRowDown38_2_Box_C(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowDown38_2_Box_16_C(const uint16_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint16_t* dst_ptr,
+                               int dst_width);
+void ScaleAddRow_C(const uint8_t* src_ptr, uint16_t* dst_ptr, int src_width);
+void ScaleAddRow_16_C(const uint16_t* src_ptr,
+                      uint32_t* dst_ptr,
+                      int src_width);
+void ScaleARGBRowDown2_C(const uint8_t* src_argb,
+                         ptrdiff_t src_stride,
+                         uint8_t* dst_argb,
+                         int dst_width);
+void ScaleARGBRowDown2Linear_C(const uint8_t* src_argb,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_argb,
+                               int dst_width);
+void ScaleARGBRowDown2Box_C(const uint8_t* src_argb,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_argb,
+                            int dst_width);
+void ScaleARGBRowDownEven_C(const uint8_t* src_argb,
+                            ptrdiff_t src_stride,
+                            int src_stepx,
+                            uint8_t* dst_argb,
+                            int dst_width);
+void ScaleARGBRowDownEvenBox_C(const uint8_t* src_argb,
+                               ptrdiff_t src_stride,
+                               int src_stepx,
+                               uint8_t* dst_argb,
+                               int dst_width);
+void ScaleARGBCols_C(uint8_t* dst_argb,
+                     const uint8_t* src_argb,
+                     int dst_width,
+                     int x,
+                     int dx);
+void ScaleARGBCols64_C(uint8_t* dst_argb,
+                       const uint8_t* src_argb,
+                       int dst_width,
+                       int x32,
+                       int dx);
+void ScaleARGBColsUp2_C(uint8_t* dst_argb,
+                        const uint8_t* src_argb,
+                        int dst_width,
+                        int,
+                        int);
+void ScaleARGBFilterCols_C(uint8_t* dst_argb,
+                           const uint8_t* src_argb,
+                           int dst_width,
+                           int x,
+                           int dx);
+void ScaleARGBFilterCols64_C(uint8_t* dst_argb,
+                             const uint8_t* src_argb,
+                             int dst_width,
+                             int x32,
+                             int dx);
+void ScaleUVRowDown2_C(const uint8_t* src_uv,
+                       ptrdiff_t src_stride,
+                       uint8_t* dst_uv,
+                       int dst_width);
+void ScaleUVRowDown2Linear_C(const uint8_t* src_uv,
+                             ptrdiff_t src_stride,
+                             uint8_t* dst_uv,
+                             int dst_width);
+void ScaleUVRowDown2Box_C(const uint8_t* src_uv,
+                          ptrdiff_t src_stride,
+                          uint8_t* dst_uv,
+                          int dst_width);
+void ScaleUVRowDownEven_C(const uint8_t* src_uv,
+                          ptrdiff_t src_stride,
+                          int src_stepx,
+                          uint8_t* dst_uv,
+                          int dst_width);
+void ScaleUVRowDownEvenBox_C(const uint8_t* src_uv,
+                             ptrdiff_t src_stride,
+                             int src_stepx,
+                             uint8_t* dst_uv,
+                             int dst_width);
+
+void ScaleUVRowUp2_Linear_C(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleUVRowUp2_Bilinear_C(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              ptrdiff_t dst_stride,
+                              int dst_width);
+void ScaleUVRowUp2_Linear_Any_C(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+void ScaleUVRowUp2_Bilinear_Any_C(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  ptrdiff_t dst_stride,
+                                  int dst_width);
+void ScaleUVRowUp2_Linear_16_C(const uint16_t* src_ptr,
+                               uint16_t* dst_ptr,
+                               int dst_width);
+void ScaleUVRowUp2_Bilinear_16_C(const uint16_t* src_ptr,
+                                 ptrdiff_t src_stride,
+                                 uint16_t* dst_ptr,
+                                 ptrdiff_t dst_stride,
+                                 int dst_width);
+void ScaleUVRowUp2_Linear_16_Any_C(const uint16_t* src_ptr,
+                                   uint16_t* dst_ptr,
+                                   int dst_width);
+void ScaleUVRowUp2_Bilinear_16_Any_C(const uint16_t* src_ptr,
+                                     ptrdiff_t src_stride,
+                                     uint16_t* dst_ptr,
+                                     ptrdiff_t dst_stride,
+                                     int dst_width);
+
+void ScaleUVCols_C(uint8_t* dst_uv,
+                   const uint8_t* src_uv,
+                   int dst_width,
+                   int x,
+                   int dx);
+void ScaleUVCols64_C(uint8_t* dst_uv,
+                     const uint8_t* src_uv,
+                     int dst_width,
+                     int x32,
+                     int dx);
+void ScaleUVColsUp2_C(uint8_t* dst_uv,
+                      const uint8_t* src_uv,
+                      int dst_width,
+                      int,
+                      int);
+void ScaleUVFilterCols_C(uint8_t* dst_uv,
+                         const uint8_t* src_uv,
+                         int dst_width,
+                         int x,
+                         int dx);
+void ScaleUVFilterCols64_C(uint8_t* dst_uv,
+                           const uint8_t* src_uv,
+                           int dst_width,
+                           int x32,
+                           int dx);
+
+// Specialized scalers for x86.
+void ScaleRowDown2_SSSE3(const uint8_t* src_ptr,
+                         ptrdiff_t src_stride,
+                         uint8_t* dst_ptr,
+                         int dst_width);
+void ScaleRowDown2Linear_SSSE3(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+void ScaleRowDown2Box_SSSE3(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowDown2_AVX2(const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint8_t* dst_ptr,
+                        int dst_width);
+void ScaleRowDown2Linear_AVX2(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowDown2Box_AVX2(const uint8_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst_ptr,
+                           int dst_width);
+void ScaleRowDown4_SSSE3(const uint8_t* src_ptr,
+                         ptrdiff_t src_stride,
+                         uint8_t* dst_ptr,
+                         int dst_width);
+void ScaleRowDown4Box_SSSE3(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowDown4_AVX2(const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint8_t* dst_ptr,
+                        int dst_width);
+void ScaleRowDown4Box_AVX2(const uint8_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst_ptr,
+                           int dst_width);
+
+void ScaleRowDown34_SSSE3(const uint8_t* src_ptr,
+                          ptrdiff_t src_stride,
+                          uint8_t* dst_ptr,
+                          int dst_width);
+void ScaleRowDown34_1_Box_SSSE3(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+void ScaleRowDown34_0_Box_SSSE3(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+void ScaleRowDown38_SSSE3(const uint8_t* src_ptr,
+                          ptrdiff_t src_stride,
+                          uint8_t* dst_ptr,
+                          int dst_width);
+void ScaleRowDown38_3_Box_SSSE3(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+void ScaleRowDown38_2_Box_SSSE3(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+
+void ScaleRowUp2_Linear_SSE2(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int dst_width);
+void ScaleRowUp2_Bilinear_SSE2(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               ptrdiff_t dst_stride,
+                               int dst_width);
+void ScaleRowUp2_Linear_12_SSSE3(const uint16_t* src_ptr,
+                                 uint16_t* dst_ptr,
+                                 int dst_width);
+void ScaleRowUp2_Bilinear_12_SSSE3(const uint16_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint16_t* dst_ptr,
+                                   ptrdiff_t dst_stride,
+                                   int dst_width);
+void ScaleRowUp2_Linear_16_SSE2(const uint16_t* src_ptr,
+                                uint16_t* dst_ptr,
+                                int dst_width);
+void ScaleRowUp2_Bilinear_16_SSE2(const uint16_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint16_t* dst_ptr,
+                                  ptrdiff_t dst_stride,
+                                  int dst_width);
+void ScaleRowUp2_Linear_SSSE3(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowUp2_Bilinear_SSSE3(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                ptrdiff_t dst_stride,
+                                int dst_width);
+void ScaleRowUp2_Linear_AVX2(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int dst_width);
+void ScaleRowUp2_Bilinear_AVX2(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               ptrdiff_t dst_stride,
+                               int dst_width);
+void ScaleRowUp2_Linear_12_AVX2(const uint16_t* src_ptr,
+                                uint16_t* dst_ptr,
+                                int dst_width);
+void ScaleRowUp2_Bilinear_12_AVX2(const uint16_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint16_t* dst_ptr,
+                                  ptrdiff_t dst_stride,
+                                  int dst_width);
+void ScaleRowUp2_Linear_16_AVX2(const uint16_t* src_ptr,
+                                uint16_t* dst_ptr,
+                                int dst_width);
+void ScaleRowUp2_Bilinear_16_AVX2(const uint16_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint16_t* dst_ptr,
+                                  ptrdiff_t dst_stride,
+                                  int dst_width);
+void ScaleRowUp2_Linear_Any_SSE2(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int dst_width);
+void ScaleRowUp2_Bilinear_Any_SSE2(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint8_t* dst_ptr,
+                                   ptrdiff_t dst_stride,
+                                   int dst_width);
+void ScaleRowUp2_Linear_12_Any_SSSE3(const uint16_t* src_ptr,
+                                     uint16_t* dst_ptr,
+                                     int dst_width);
+void ScaleRowUp2_Bilinear_12_Any_SSSE3(const uint16_t* src_ptr,
+                                       ptrdiff_t src_stride,
+                                       uint16_t* dst_ptr,
+                                       ptrdiff_t dst_stride,
+                                       int dst_width);
+void ScaleRowUp2_Linear_16_Any_SSE2(const uint16_t* src_ptr,
+                                    uint16_t* dst_ptr,
+                                    int dst_width);
+void ScaleRowUp2_Bilinear_16_Any_SSE2(const uint16_t* src_ptr,
+                                      ptrdiff_t src_stride,
+                                      uint16_t* dst_ptr,
+                                      ptrdiff_t dst_stride,
+                                      int dst_width);
+void ScaleRowUp2_Linear_Any_SSSE3(const uint8_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleRowUp2_Bilinear_Any_SSSE3(const uint8_t* src_ptr,
+                                    ptrdiff_t src_stride,
+                                    uint8_t* dst_ptr,
+                                    ptrdiff_t dst_stride,
+                                    int dst_width);
+void ScaleRowUp2_Linear_Any_AVX2(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int dst_width);
+void ScaleRowUp2_Bilinear_Any_AVX2(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint8_t* dst_ptr,
+                                   ptrdiff_t dst_stride,
+                                   int dst_width);
+void ScaleRowUp2_Linear_12_Any_AVX2(const uint16_t* src_ptr,
+                                    uint16_t* dst_ptr,
+                                    int dst_width);
+void ScaleRowUp2_Bilinear_12_Any_AVX2(const uint16_t* src_ptr,
+                                      ptrdiff_t src_stride,
+                                      uint16_t* dst_ptr,
+                                      ptrdiff_t dst_stride,
+                                      int dst_width);
+void ScaleRowUp2_Linear_16_Any_AVX2(const uint16_t* src_ptr,
+                                    uint16_t* dst_ptr,
+                                    int dst_width);
+void ScaleRowUp2_Bilinear_16_Any_AVX2(const uint16_t* src_ptr,
+                                      ptrdiff_t src_stride,
+                                      uint16_t* dst_ptr,
+                                      ptrdiff_t dst_stride,
+                                      int dst_width);
+
+void ScaleRowDown2_Any_SSSE3(const uint8_t* src_ptr,
+                             ptrdiff_t src_stride,
+                             uint8_t* dst_ptr,
+                             int dst_width);
+void ScaleRowDown2Linear_Any_SSSE3(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+void ScaleRowDown2Box_Any_SSSE3(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+void ScaleRowDown2Box_Odd_SSSE3(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+void ScaleRowDown2_Any_AVX2(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowDown2Linear_Any_AVX2(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleRowDown2Box_Any_AVX2(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+void ScaleRowDown2Box_Odd_AVX2(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+void ScaleRowDown4_Any_SSSE3(const uint8_t* src_ptr,
+                             ptrdiff_t src_stride,
+                             uint8_t* dst_ptr,
+                             int dst_width);
+void ScaleRowDown4Box_Any_SSSE3(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+void ScaleRowDown4_Any_AVX2(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowDown4Box_Any_AVX2(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+
+void ScaleRowDown34_Any_SSSE3(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowDown34_1_Box_Any_SSSE3(const uint8_t* src_ptr,
+                                    ptrdiff_t src_stride,
+                                    uint8_t* dst_ptr,
+                                    int dst_width);
+void ScaleRowDown34_0_Box_Any_SSSE3(const uint8_t* src_ptr,
+                                    ptrdiff_t src_stride,
+                                    uint8_t* dst_ptr,
+                                    int dst_width);
+void ScaleRowDown38_Any_SSSE3(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowDown38_3_Box_Any_SSSE3(const uint8_t* src_ptr,
+                                    ptrdiff_t src_stride,
+                                    uint8_t* dst_ptr,
+                                    int dst_width);
+void ScaleRowDown38_2_Box_Any_SSSE3(const uint8_t* src_ptr,
+                                    ptrdiff_t src_stride,
+                                    uint8_t* dst_ptr,
+                                    int dst_width);
+
+void ScaleAddRow_SSE2(const uint8_t* src_ptr, uint16_t* dst_ptr, int src_width);
+void ScaleAddRow_AVX2(const uint8_t* src_ptr, uint16_t* dst_ptr, int src_width);
+void ScaleAddRow_Any_SSE2(const uint8_t* src_ptr,
+                          uint16_t* dst_ptr,
+                          int src_width);
+void ScaleAddRow_Any_AVX2(const uint8_t* src_ptr,
+                          uint16_t* dst_ptr,
+                          int src_width);
+
+void ScaleFilterCols_SSSE3(uint8_t* dst_ptr,
+                           const uint8_t* src_ptr,
+                           int dst_width,
+                           int x,
+                           int dx);
+void ScaleColsUp2_SSE2(uint8_t* dst_ptr,
+                       const uint8_t* src_ptr,
+                       int dst_width,
+                       int x,
+                       int dx);
+
+// ARGB Column functions
+void ScaleARGBCols_SSE2(uint8_t* dst_argb,
+                        const uint8_t* src_argb,
+                        int dst_width,
+                        int x,
+                        int dx);
+void ScaleARGBFilterCols_SSSE3(uint8_t* dst_argb,
+                               const uint8_t* src_argb,
+                               int dst_width,
+                               int x,
+                               int dx);
+void ScaleARGBColsUp2_SSE2(uint8_t* dst_argb,
+                           const uint8_t* src_argb,
+                           int dst_width,
+                           int x,
+                           int dx);
+void ScaleARGBFilterCols_NEON(uint8_t* dst_argb,
+                              const uint8_t* src_argb,
+                              int dst_width,
+                              int x,
+                              int dx);
+void ScaleARGBCols_NEON(uint8_t* dst_argb,
+                        const uint8_t* src_argb,
+                        int dst_width,
+                        int x,
+                        int dx);
+void ScaleARGBFilterCols_Any_NEON(uint8_t* dst_ptr,
+                                  const uint8_t* src_ptr,
+                                  int dst_width,
+                                  int x,
+                                  int dx);
+void ScaleARGBCols_Any_NEON(uint8_t* dst_ptr,
+                            const uint8_t* src_ptr,
+                            int dst_width,
+                            int x,
+                            int dx);
+void ScaleARGBFilterCols_MSA(uint8_t* dst_argb,
+                             const uint8_t* src_argb,
+                             int dst_width,
+                             int x,
+                             int dx);
+void ScaleARGBCols_MSA(uint8_t* dst_argb,
+                       const uint8_t* src_argb,
+                       int dst_width,
+                       int x,
+                       int dx);
+void ScaleARGBFilterCols_Any_MSA(uint8_t* dst_ptr,
+                                 const uint8_t* src_ptr,
+                                 int dst_width,
+                                 int x,
+                                 int dx);
+void ScaleARGBCols_Any_MSA(uint8_t* dst_ptr,
+                           const uint8_t* src_ptr,
+                           int dst_width,
+                           int x,
+                           int dx);
+
+// ARGB Row functions
+void ScaleARGBRowDown2_SSE2(const uint8_t* src_argb,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_argb,
+                            int dst_width);
+void ScaleARGBRowDown2Linear_SSE2(const uint8_t* src_argb,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_argb,
+                                  int dst_width);
+void ScaleARGBRowDown2Box_SSE2(const uint8_t* src_argb,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_argb,
+                               int dst_width);
+void ScaleARGBRowDown2_NEON(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst,
+                            int dst_width);
+void ScaleARGBRowDown2Linear_NEON(const uint8_t* src_argb,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_argb,
+                                  int dst_width);
+void ScaleARGBRowDown2Box_NEON(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst,
+                               int dst_width);
+void ScaleARGBRowDown2_RVV(const uint8_t* src_argb,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst_argb,
+                           int dst_width);
+void ScaleARGBRowDown2Linear_RVV(const uint8_t* src_argb,
+                                 ptrdiff_t src_stride,
+                                 uint8_t* dst_argb,
+                                 int dst_width);
+void ScaleARGBRowDown2Box_RVV(const uint8_t* src_argb,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_argb,
+                              int dst_width);
+void ScaleARGBRowDown2_MSA(const uint8_t* src_argb,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst_argb,
+                           int dst_width);
+void ScaleARGBRowDown2Linear_MSA(const uint8_t* src_argb,
+                                 ptrdiff_t src_stride,
+                                 uint8_t* dst_argb,
+                                 int dst_width);
+void ScaleARGBRowDown2Box_MSA(const uint8_t* src_argb,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_argb,
+                              int dst_width);
+void ScaleARGBRowDown2_LSX(const uint8_t* src_argb,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst_argb,
+                           int dst_width);
+void ScaleARGBRowDown2Linear_LSX(const uint8_t* src_argb,
+                                 ptrdiff_t src_stride,
+                                 uint8_t* dst_argb,
+                                 int dst_width);
+void ScaleARGBRowDown2Box_LSX(const uint8_t* src_argb,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_argb,
+                              int dst_width);
+void ScaleARGBRowDown2_Any_SSE2(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+void ScaleARGBRowDown2Linear_Any_SSE2(const uint8_t* src_ptr,
+                                      ptrdiff_t src_stride,
+                                      uint8_t* dst_ptr,
+                                      int dst_width);
+void ScaleARGBRowDown2Box_Any_SSE2(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+void ScaleARGBRowDown2_Any_NEON(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+void ScaleARGBRowDown2Linear_Any_NEON(const uint8_t* src_ptr,
+                                      ptrdiff_t src_stride,
+                                      uint8_t* dst_ptr,
+                                      int dst_width);
+void ScaleARGBRowDown2Box_Any_NEON(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+void ScaleARGBRowDown2_Any_MSA(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+void ScaleARGBRowDown2Linear_Any_MSA(const uint8_t* src_ptr,
+                                     ptrdiff_t src_stride,
+                                     uint8_t* dst_ptr,
+                                     int dst_width);
+void ScaleARGBRowDown2Box_Any_MSA(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleARGBRowDown2_Any_LSX(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+void ScaleARGBRowDown2Linear_Any_LSX(const uint8_t* src_ptr,
+                                     ptrdiff_t src_stride,
+                                     uint8_t* dst_ptr,
+                                     int dst_width);
+void ScaleARGBRowDown2Box_Any_LSX(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleARGBRowDownEven_SSE2(const uint8_t* src_argb,
+                               ptrdiff_t src_stride,
+                               int src_stepx,
+                               uint8_t* dst_argb,
+                               int dst_width);
+void ScaleARGBRowDownEvenBox_SSE2(const uint8_t* src_argb,
+                                  ptrdiff_t src_stride,
+                                  int src_stepx,
+                                  uint8_t* dst_argb,
+                                  int dst_width);
+void ScaleARGBRowDownEven_NEON(const uint8_t* src_argb,
+                               ptrdiff_t src_stride,
+                               int src_stepx,
+                               uint8_t* dst_argb,
+                               int dst_width);
+void ScaleARGBRowDownEvenBox_NEON(const uint8_t* src_argb,
+                                  ptrdiff_t src_stride,
+                                  int src_stepx,
+                                  uint8_t* dst_argb,
+                                  int dst_width);
+void ScaleARGBRowDownEven_MSA(const uint8_t* src_argb,
+                              ptrdiff_t src_stride,
+                              int32_t src_stepx,
+                              uint8_t* dst_argb,
+                              int dst_width);
+void ScaleARGBRowDownEvenBox_MSA(const uint8_t* src_argb,
+                                 ptrdiff_t src_stride,
+                                 int src_stepx,
+                                 uint8_t* dst_argb,
+                                 int dst_width);
+void ScaleARGBRowDownEven_LSX(const uint8_t* src_argb,
+                              ptrdiff_t src_stride,
+                              int32_t src_stepx,
+                              uint8_t* dst_argb,
+                              int dst_width);
+void ScaleARGBRowDownEvenBox_LSX(const uint8_t* src_argb,
+                                 ptrdiff_t src_stride,
+                                 int src_stepx,
+                                 uint8_t* dst_argb,
+                                 int dst_width);
+void ScaleARGBRowDownEven_RVV(const uint8_t* src_argb,
+                              ptrdiff_t src_stride,
+                              int32_t src_stepx,
+                              uint8_t* dst_argb,
+                              int dst_width);
+void ScaleARGBRowDownEvenBox_RVV(const uint8_t* src_argb,
+                                 ptrdiff_t src_stride,
+                                 int src_stepx,
+                                 uint8_t* dst_argb,
+                                 int dst_width);
+void ScaleARGBRowDownEven_Any_SSE2(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   int src_stepx,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+void ScaleARGBRowDownEvenBox_Any_SSE2(const uint8_t* src_ptr,
+                                      ptrdiff_t src_stride,
+                                      int src_stepx,
+                                      uint8_t* dst_ptr,
+                                      int dst_width);
+void ScaleARGBRowDownEven_Any_NEON(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   int src_stepx,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+void ScaleARGBRowDownEvenBox_Any_NEON(const uint8_t* src_ptr,
+                                      ptrdiff_t src_stride,
+                                      int src_stepx,
+                                      uint8_t* dst_ptr,
+                                      int dst_width);
+void ScaleARGBRowDownEven_Any_MSA(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  int32_t src_stepx,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleARGBRowDownEvenBox_Any_MSA(const uint8_t* src_ptr,
+                                     ptrdiff_t src_stride,
+                                     int src_stepx,
+                                     uint8_t* dst_ptr,
+                                     int dst_width);
+void ScaleARGBRowDownEven_Any_LSX(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  int32_t src_stepx,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleARGBRowDownEvenBox_Any_LSX(const uint8_t* src_ptr,
+                                     ptrdiff_t src_stride,
+                                     int src_stepx,
+                                     uint8_t* dst_ptr,
+                                     int dst_width);
+
+// UV Row functions
+void ScaleUVRowDown2_SSSE3(const uint8_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst_uv,
+                           int dst_width);
+void ScaleUVRowDown2Linear_SSSE3(const uint8_t* src_ptr,
+                                 ptrdiff_t src_stride,
+                                 uint8_t* dst_uv,
+                                 int dst_width);
+void ScaleUVRowDown2Box_SSSE3(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_uv,
+                              int dst_width);
+void ScaleUVRowDown2Box_AVX2(const uint8_t* src_ptr,
+                             ptrdiff_t src_stride,
+                             uint8_t* dst_uv,
+                             int dst_width);
+void ScaleUVRowDown2_NEON(const uint8_t* src_ptr,
+                          ptrdiff_t src_stride,
+                          uint8_t* dst,
+                          int dst_width);
+void ScaleUVRowDown2Linear_NEON(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_uv,
+                                int dst_width);
+void ScaleUVRowDown2Box_NEON(const uint8_t* src_ptr,
+                             ptrdiff_t src_stride,
+                             uint8_t* dst,
+                             int dst_width);
+void ScaleUVRowDown2_MSA(const uint8_t* src_ptr,
+                         ptrdiff_t src_stride,
+                         uint8_t* dst_uv,
+                         int dst_width);
+void ScaleUVRowDown2Linear_MSA(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_uv,
+                               int dst_width);
+void ScaleUVRowDown2Box_MSA(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_uv,
+                            int dst_width);
+void ScaleUVRowDown2_RVV(const uint8_t* src_ptr,
+                         ptrdiff_t src_stride,
+                         uint8_t* dst_uv,
+                         int dst_width);
+void ScaleUVRowDown2Linear_RVV(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_uv,
+                               int dst_width);
+void ScaleUVRowDown2Box_RVV(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst,
+                            int dst_width);
+void ScaleUVRowDown2_Any_SSSE3(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+void ScaleUVRowDown2Linear_Any_SSSE3(const uint8_t* src_ptr,
+                                     ptrdiff_t src_stride,
+                                     uint8_t* dst_ptr,
+                                     int dst_width);
+void ScaleUVRowDown2Box_Any_SSSE3(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleUVRowDown2Box_Any_AVX2(const uint8_t* src_ptr,
+                                 ptrdiff_t src_stride,
+                                 uint8_t* dst_ptr,
+                                 int dst_width);
+void ScaleUVRowDown2_Any_NEON(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleUVRowDown2Linear_Any_NEON(const uint8_t* src_ptr,
+                                    ptrdiff_t src_stride,
+                                    uint8_t* dst_ptr,
+                                    int dst_width);
+void ScaleUVRowDown2Box_Any_NEON(const uint8_t* src_ptr,
+                                 ptrdiff_t src_stride,
+                                 uint8_t* dst_ptr,
+                                 int dst_width);
+void ScaleUVRowDown2_Any_MSA(const uint8_t* src_ptr,
+                             ptrdiff_t src_stride,
+                             uint8_t* dst_ptr,
+                             int dst_width);
+void ScaleUVRowDown2Linear_Any_MSA(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+void ScaleUVRowDown2Box_Any_MSA(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+void ScaleUVRowDownEven_SSSE3(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              int src_stepx,
+                              uint8_t* dst_uv,
+                              int dst_width);
+void ScaleUVRowDownEvenBox_SSSE3(const uint8_t* src_ptr,
+                                 ptrdiff_t src_stride,
+                                 int src_stepx,
+                                 uint8_t* dst_uv,
+                                 int dst_width);
+void ScaleUVRowDownEven_NEON(const uint8_t* src_ptr,
+                             ptrdiff_t src_stride,
+                             int src_stepx,
+                             uint8_t* dst_uv,
+                             int dst_width);
+void ScaleUVRowDownEvenBox_NEON(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                int src_stepx,
+                                uint8_t* dst_uv,
+                                int dst_width);
+void ScaleUVRowDown4_RVV(const uint8_t* src_ptr,
+                         ptrdiff_t src_stride,
+                         int32_t src_stepx,
+                         uint8_t* dst_uv,
+                         int dst_width);
+void ScaleUVRowDownEven_RVV(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            int32_t src_stepx,
+                            uint8_t* dst_uv,
+                            int dst_width);
+void ScaleUVRowDownEven_MSA(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            int32_t src_stepx,
+                            uint8_t* dst_uv,
+                            int dst_width);
+void ScaleUVRowDownEvenBox_MSA(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               int src_stepx,
+                               uint8_t* dst_uv,
+                               int dst_width);
+void ScaleUVRowDownEven_Any_SSSE3(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  int src_stepx,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleUVRowDownEvenBox_Any_SSSE3(const uint8_t* src_ptr,
+                                     ptrdiff_t src_stride,
+                                     int src_stepx,
+                                     uint8_t* dst_ptr,
+                                     int dst_width);
+void ScaleUVRowDownEven_Any_NEON(const uint8_t* src_ptr,
+                                 ptrdiff_t src_stride,
+                                 int src_stepx,
+                                 uint8_t* dst_ptr,
+                                 int dst_width);
+void ScaleUVRowDownEvenBox_Any_NEON(const uint8_t* src_ptr,
+                                    ptrdiff_t src_stride,
+                                    int src_stepx,
+                                    uint8_t* dst_ptr,
+                                    int dst_width);
+void ScaleUVRowDownEven_Any_MSA(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                int32_t src_stepx,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+void ScaleUVRowDownEvenBox_Any_MSA(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   int src_stepx,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+
+void ScaleUVRowUp2_Linear_SSSE3(const uint8_t* src_ptr,
+                                uint8_t* dst_ptr,
+                                int dst_width);
+void ScaleUVRowUp2_Bilinear_SSSE3(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  ptrdiff_t dst_stride,
+                                  int dst_width);
+void ScaleUVRowUp2_Linear_Any_SSSE3(const uint8_t* src_ptr,
+                                    uint8_t* dst_ptr,
+                                    int dst_width);
+void ScaleUVRowUp2_Bilinear_Any_SSSE3(const uint8_t* src_ptr,
+                                      ptrdiff_t src_stride,
+                                      uint8_t* dst_ptr,
+                                      ptrdiff_t dst_stride,
+                                      int dst_width);
+void ScaleUVRowUp2_Linear_AVX2(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+void ScaleUVRowUp2_Bilinear_AVX2(const uint8_t* src_ptr,
+                                 ptrdiff_t src_stride,
+                                 uint8_t* dst_ptr,
+                                 ptrdiff_t dst_stride,
+                                 int dst_width);
+void ScaleUVRowUp2_Linear_Any_AVX2(const uint8_t* src_ptr,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+void ScaleUVRowUp2_Bilinear_Any_AVX2(const uint8_t* src_ptr,
+                                     ptrdiff_t src_stride,
+                                     uint8_t* dst_ptr,
+                                     ptrdiff_t dst_stride,
+                                     int dst_width);
+void ScaleUVRowUp2_Linear_NEON(const uint8_t* src_ptr,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+void ScaleUVRowUp2_Bilinear_NEON(const uint8_t* src_ptr,
+                                 ptrdiff_t src_stride,
+                                 uint8_t* dst_ptr,
+                                 ptrdiff_t dst_stride,
+                                 int dst_width);
+void ScaleUVRowUp2_Linear_Any_NEON(const uint8_t* src_ptr,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+void ScaleUVRowUp2_Bilinear_Any_NEON(const uint8_t* src_ptr,
+                                     ptrdiff_t src_stride,
+                                     uint8_t* dst_ptr,
+                                     ptrdiff_t dst_stride,
+                                     int dst_width);
+void ScaleUVRowUp2_Linear_RVV(const uint8_t* src_ptr,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleUVRowUp2_Bilinear_RVV(const uint8_t* src_ptr,
+                                ptrdiff_t src_stride,
+                                uint8_t* dst_ptr,
+                                ptrdiff_t dst_stride,
+                                int dst_width);
+void ScaleUVRowUp2_Linear_16_SSE41(const uint16_t* src_ptr,
+                                   uint16_t* dst_ptr,
+                                   int dst_width);
+void ScaleUVRowUp2_Bilinear_16_SSE41(const uint16_t* src_ptr,
+                                     ptrdiff_t src_stride,
+                                     uint16_t* dst_ptr,
+                                     ptrdiff_t dst_stride,
+                                     int dst_width);
+void ScaleUVRowUp2_Linear_16_Any_SSE41(const uint16_t* src_ptr,
+                                       uint16_t* dst_ptr,
+                                       int dst_width);
+void ScaleUVRowUp2_Bilinear_16_Any_SSE41(const uint16_t* src_ptr,
+                                         ptrdiff_t src_stride,
+                                         uint16_t* dst_ptr,
+                                         ptrdiff_t dst_stride,
+                                         int dst_width);
+void ScaleUVRowUp2_Linear_16_AVX2(const uint16_t* src_ptr,
+                                  uint16_t* dst_ptr,
+                                  int dst_width);
+void ScaleUVRowUp2_Bilinear_16_AVX2(const uint16_t* src_ptr,
+                                    ptrdiff_t src_stride,
+                                    uint16_t* dst_ptr,
+                                    ptrdiff_t dst_stride,
+                                    int dst_width);
+void ScaleUVRowUp2_Linear_16_Any_AVX2(const uint16_t* src_ptr,
+                                      uint16_t* dst_ptr,
+                                      int dst_width);
+void ScaleUVRowUp2_Bilinear_16_Any_AVX2(const uint16_t* src_ptr,
+                                        ptrdiff_t src_stride,
+                                        uint16_t* dst_ptr,
+                                        ptrdiff_t dst_stride,
+                                        int dst_width);
+void ScaleUVRowUp2_Linear_16_NEON(const uint16_t* src_ptr,
+                                  uint16_t* dst_ptr,
+                                  int dst_width);
+void ScaleUVRowUp2_Bilinear_16_NEON(const uint16_t* src_ptr,
+                                    ptrdiff_t src_stride,
+                                    uint16_t* dst_ptr,
+                                    ptrdiff_t dst_stride,
+                                    int dst_width);
+void ScaleUVRowUp2_Linear_16_Any_NEON(const uint16_t* src_ptr,
+                                      uint16_t* dst_ptr,
+                                      int dst_width);
+void ScaleUVRowUp2_Bilinear_16_Any_NEON(const uint16_t* src_ptr,
+                                        ptrdiff_t src_stride,
+                                        uint16_t* dst_ptr,
+                                        ptrdiff_t dst_stride,
+                                        int dst_width);
+
+// ScaleRowDown2Box also used by planar functions
+// NEON downscalers with interpolation.
+
+// Note - not static due to reuse in convert for 444 to 420.
+void ScaleRowDown2_NEON(const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint8_t* dst,
+                        int dst_width);
+void ScaleRowDown2Linear_NEON(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst,
+                              int dst_width);
+void ScaleRowDown2Box_NEON(const uint8_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst,
+                           int dst_width);
+
+void ScaleRowDown4_NEON(const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint8_t* dst_ptr,
+                        int dst_width);
+void ScaleRowDown4Box_NEON(const uint8_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst_ptr,
+                           int dst_width);
+
+// Down scale from 4 to 3 pixels. Use the neon multilane read/write
+//  to load up the every 4th pixel into a 4 different registers.
+// Point samples 32 pixels to 24 pixels.
+void ScaleRowDown34_NEON(const uint8_t* src_ptr,
+                         ptrdiff_t src_stride,
+                         uint8_t* dst_ptr,
+                         int dst_width);
+void ScaleRowDown34_0_Box_NEON(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+void ScaleRowDown34_1_Box_NEON(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+
+// 32 -> 12
+void ScaleRowDown38_NEON(const uint8_t* src_ptr,
+                         ptrdiff_t src_stride,
+                         uint8_t* dst_ptr,
+                         int dst_width);
+// 32x3 -> 12x1
+void ScaleRowDown38_3_Box_NEON(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+// 32x2 -> 12x1
+void ScaleRowDown38_2_Box_NEON(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+
+void ScaleRowDown2_Any_NEON(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowDown2Linear_Any_NEON(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleRowDown2Box_Any_NEON(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+void ScaleRowDown2Box_Odd_NEON(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+void ScaleRowDown4_Any_NEON(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowDown4Box_Any_NEON(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               int dst_width);
+void ScaleRowDown34_Any_NEON(const uint8_t* src_ptr,
+                             ptrdiff_t src_stride,
+                             uint8_t* dst_ptr,
+                             int dst_width);
+void ScaleRowDown34_0_Box_Any_NEON(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+void ScaleRowDown34_1_Box_Any_NEON(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+// 32 -> 12
+void ScaleRowDown38_Any_NEON(const uint8_t* src_ptr,
+                             ptrdiff_t src_stride,
+                             uint8_t* dst_ptr,
+                             int dst_width);
+// 32x3 -> 12x1
+void ScaleRowDown38_3_Box_Any_NEON(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+// 32x2 -> 12x1
+void ScaleRowDown38_2_Box_Any_NEON(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint8_t* dst_ptr,
+                                   int dst_width);
+
+void ScaleRowUp2_Linear_NEON(const uint8_t* src_ptr,
+                             uint8_t* dst_ptr,
+                             int dst_width);
+void ScaleRowUp2_Bilinear_NEON(const uint8_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst_ptr,
+                               ptrdiff_t dst_stride,
+                               int dst_width);
+void ScaleRowUp2_Linear_12_NEON(const uint16_t* src_ptr,
+                                uint16_t* dst_ptr,
+                                int dst_width);
+void ScaleRowUp2_Bilinear_12_NEON(const uint16_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint16_t* dst_ptr,
+                                  ptrdiff_t dst_stride,
+                                  int dst_width);
+void ScaleRowUp2_Linear_16_NEON(const uint16_t* src_ptr,
+                                uint16_t* dst_ptr,
+                                int dst_width);
+void ScaleRowUp2_Bilinear_16_NEON(const uint16_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint16_t* dst_ptr,
+                                  ptrdiff_t dst_stride,
+                                  int dst_width);
+void ScaleRowUp2_Linear_Any_NEON(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int dst_width);
+void ScaleRowUp2_Bilinear_Any_NEON(const uint8_t* src_ptr,
+                                   ptrdiff_t src_stride,
+                                   uint8_t* dst_ptr,
+                                   ptrdiff_t dst_stride,
+                                   int dst_width);
+void ScaleRowUp2_Linear_12_Any_NEON(const uint16_t* src_ptr,
+                                    uint16_t* dst_ptr,
+                                    int dst_width);
+void ScaleRowUp2_Bilinear_12_Any_NEON(const uint16_t* src_ptr,
+                                      ptrdiff_t src_stride,
+                                      uint16_t* dst_ptr,
+                                      ptrdiff_t dst_stride,
+                                      int dst_width);
+void ScaleRowUp2_Linear_16_Any_NEON(const uint16_t* src_ptr,
+                                    uint16_t* dst_ptr,
+                                    int dst_width);
+void ScaleRowUp2_Bilinear_16_Any_NEON(const uint16_t* src_ptr,
+                                      ptrdiff_t src_stride,
+                                      uint16_t* dst_ptr,
+                                      ptrdiff_t dst_stride,
+                                      int dst_width);
+
+void ScaleAddRow_NEON(const uint8_t* src_ptr, uint16_t* dst_ptr, int src_width);
+void ScaleAddRow_Any_NEON(const uint8_t* src_ptr,
+                          uint16_t* dst_ptr,
+                          int src_width);
+
+void ScaleFilterCols_NEON(uint8_t* dst_ptr,
+                          const uint8_t* src_ptr,
+                          int dst_width,
+                          int x,
+                          int dx);
+
+void ScaleFilterCols_Any_NEON(uint8_t* dst_ptr,
+                              const uint8_t* src_ptr,
+                              int dst_width,
+                              int x,
+                              int dx);
+
+void ScaleRowDown2_MSA(const uint8_t* src_ptr,
+                       ptrdiff_t src_stride,
+                       uint8_t* dst,
+                       int dst_width);
+void ScaleRowDown2Linear_MSA(const uint8_t* src_ptr,
+                             ptrdiff_t src_stride,
+                             uint8_t* dst,
+                             int dst_width);
+void ScaleRowDown2Box_MSA(const uint8_t* src_ptr,
+                          ptrdiff_t src_stride,
+                          uint8_t* dst,
+                          int dst_width);
+void ScaleRowDown4_MSA(const uint8_t* src_ptr,
+                       ptrdiff_t src_stride,
+                       uint8_t* dst,
+                       int dst_width);
+void ScaleRowDown4Box_MSA(const uint8_t* src_ptr,
+                          ptrdiff_t src_stride,
+                          uint8_t* dst,
+                          int dst_width);
+void ScaleRowDown38_MSA(const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint8_t* dst,
+                        int dst_width);
+void ScaleRowDown38_2_Box_MSA(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowDown38_3_Box_MSA(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleAddRow_MSA(const uint8_t* src_ptr, uint16_t* dst_ptr, int src_width);
+void ScaleFilterCols_MSA(uint8_t* dst_ptr,
+                         const uint8_t* src_ptr,
+                         int dst_width,
+                         int x,
+                         int dx);
+void ScaleRowDown34_MSA(const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint8_t* dst,
+                        int dst_width);
+void ScaleRowDown34_0_Box_MSA(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* d,
+                              int dst_width);
+void ScaleRowDown34_1_Box_MSA(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* d,
+                              int dst_width);
+
+void ScaleRowDown2_Any_MSA(const uint8_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst_ptr,
+                           int dst_width);
+void ScaleRowDown2Linear_Any_MSA(const uint8_t* src_ptr,
+                                 ptrdiff_t src_stride,
+                                 uint8_t* dst_ptr,
+                                 int dst_width);
+void ScaleRowDown2Box_Any_MSA(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowDown4_Any_MSA(const uint8_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst_ptr,
+                           int dst_width);
+void ScaleRowDown4Box_Any_MSA(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowDown38_Any_MSA(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowDown38_2_Box_Any_MSA(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleRowDown38_3_Box_Any_MSA(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleAddRow_Any_MSA(const uint8_t* src_ptr,
+                         uint16_t* dst_ptr,
+                         int src_width);
+void ScaleFilterCols_Any_MSA(uint8_t* dst_ptr,
+                             const uint8_t* src_ptr,
+                             int dst_width,
+                             int x,
+                             int dx);
+void ScaleRowDown34_Any_MSA(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowDown34_0_Box_Any_MSA(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleRowDown34_1_Box_Any_MSA(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+
+void ScaleRowDown2_LSX(const uint8_t* src_ptr,
+                       ptrdiff_t src_stride,
+                       uint8_t* dst,
+                       int dst_width);
+void ScaleRowDown2Linear_LSX(const uint8_t* src_ptr,
+                             ptrdiff_t src_stride,
+                             uint8_t* dst,
+                             int dst_width);
+void ScaleRowDown2Box_LSX(const uint8_t* src_ptr,
+                          ptrdiff_t src_stride,
+                          uint8_t* dst,
+                          int dst_width);
+void ScaleRowDown4_LSX(const uint8_t* src_ptr,
+                       ptrdiff_t src_stride,
+                       uint8_t* dst,
+                       int dst_width);
+void ScaleRowDown4Box_LSX(const uint8_t* src_ptr,
+                          ptrdiff_t src_stride,
+                          uint8_t* dst,
+                          int dst_width);
+void ScaleRowDown38_LSX(const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint8_t* dst,
+                        int dst_width);
+void ScaleRowDown38_2_Box_LSX(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowDown38_3_Box_LSX(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleAddRow_LSX(const uint8_t* src_ptr, uint16_t* dst_ptr, int src_width);
+void ScaleFilterCols_LSX(uint8_t* dst_ptr,
+                         const uint8_t* src_ptr,
+                         int dst_width,
+                         int x,
+                         int dx);
+void ScaleARGBFilterCols_LSX(uint8_t* dst_argb,
+                             const uint8_t* src_argb,
+                             int dst_width,
+                             int x,
+                             int dx);
+void ScaleARGBCols_LSX(uint8_t* dst_argb,
+                       const uint8_t* src_argb,
+                       int dst_width,
+                       int x,
+                       int dx);
+void ScaleRowDown34_LSX(const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint8_t* dst,
+                        int dst_width);
+void ScaleRowDown34_0_Box_LSX(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* d,
+                              int dst_width);
+void ScaleRowDown34_1_Box_LSX(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* d,
+                              int dst_width);
+void ScaleRowDown2_Any_LSX(const uint8_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst_ptr,
+                           int dst_width);
+void ScaleRowDown2Linear_Any_LSX(const uint8_t* src_ptr,
+                                 ptrdiff_t src_stride,
+                                 uint8_t* dst_ptr,
+                                 int dst_width);
+void ScaleRowDown2Box_Any_LSX(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowDown4_Any_LSX(const uint8_t* src_ptr,
+                           ptrdiff_t src_stride,
+                           uint8_t* dst_ptr,
+                           int dst_width);
+void ScaleRowDown4Box_Any_LSX(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowDown38_Any_LSX(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowDown38_2_Box_Any_LSX(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleRowDown38_3_Box_Any_LSX(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleAddRow_Any_LSX(const uint8_t* src_ptr,
+                         uint16_t* dst_ptr,
+                         int src_width);
+void ScaleFilterCols_Any_LSX(uint8_t* dst_ptr,
+                             const uint8_t* src_ptr,
+                             int dst_width,
+                             int x,
+                             int dx);
+void ScaleARGBCols_Any_LSX(uint8_t* dst_ptr,
+                           const uint8_t* src_ptr,
+                           int dst_width,
+                           int x,
+                           int dx);
+void ScaleARGBFilterCols_Any_LSX(uint8_t* dst_ptr,
+                                 const uint8_t* src_ptr,
+                                 int dst_width,
+                                 int x,
+                                 int dx);
+void ScaleRowDown34_Any_LSX(const uint8_t* src_ptr,
+                            ptrdiff_t src_stride,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowDown34_0_Box_Any_LSX(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+void ScaleRowDown34_1_Box_Any_LSX(const uint8_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst_ptr,
+                                  int dst_width);
+
+void ScaleAddRow_RVV(const uint8_t* src_ptr, uint16_t* dst_ptr, int src_width);
+void ScaleRowDown2_RVV(const uint8_t* src_ptr,
+                       ptrdiff_t src_stride,
+                       uint8_t* dst,
+                       int dst_width);
+void ScaleRowDown2Linear_RVV(const uint8_t* src_ptr,
+                             ptrdiff_t src_stride,
+                             uint8_t* dst,
+                             int dst_width);
+void ScaleRowDown2Box_RVV(const uint8_t* src_ptr,
+                          ptrdiff_t src_stride,
+                          uint8_t* dst,
+                          int dst_width);
+
+void ScaleRowDown4_RVV(const uint8_t* src_ptr,
+                       ptrdiff_t src_stride,
+                       uint8_t* dst_ptr,
+                       int dst_width);
+void ScaleRowDown4Box_RVV(const uint8_t* src_ptr,
+                          ptrdiff_t src_stride,
+                          uint8_t* dst_ptr,
+                          int dst_width);
+void ScaleRowDown34_RVV(const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint8_t* dst_ptr,
+                        int dst_width);
+void ScaleRowDown34_0_Box_RVV(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowDown34_1_Box_RVV(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowDown38_RVV(const uint8_t* src_ptr,
+                        ptrdiff_t src_stride,
+                        uint8_t* dst,
+                        int dst_width);
+void ScaleRowDown38_3_Box_RVV(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+void ScaleRowDown38_2_Box_RVV(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              int dst_width);
+
+void ScaleRowUp2_Linear_RVV(const uint8_t* src_ptr,
+                            uint8_t* dst_ptr,
+                            int dst_width);
+void ScaleRowUp2_Bilinear_RVV(const uint8_t* src_ptr,
+                              ptrdiff_t src_stride,
+                              uint8_t* dst_ptr,
+                              ptrdiff_t dst_stride,
+                              int dst_width);
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_SCALE_ROW_H_

yuv.mod/libyuv/include/libyuv/scale_uv.h

@@ -0,0 +1,51 @@
+/*
+ *  Copyright 2020 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_SCALE_UV_H_
+#define INCLUDE_LIBYUV_SCALE_UV_H_
+
+#include "libyuv/basic_types.h"
+#include "libyuv/scale.h"  // For FilterMode
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+LIBYUV_API
+int UVScale(const uint8_t* src_uv,
+            int src_stride_uv,
+            int src_width,
+            int src_height,
+            uint8_t* dst_uv,
+            int dst_stride_uv,
+            int dst_width,
+            int dst_height,
+            enum FilterMode filtering);
+
+// Scale a 16 bit UV image.
+// This function is currently incomplete, it can't handle all cases.
+LIBYUV_API
+int UVScale_16(const uint16_t* src_uv,
+               int src_stride_uv,
+               int src_width,
+               int src_height,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int dst_width,
+               int dst_height,
+               enum FilterMode filtering);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_SCALE_UV_H_

yuv.mod/libyuv/include/libyuv/version.h

@@ -0,0 +1,16 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef INCLUDE_LIBYUV_VERSION_H_
+#define INCLUDE_LIBYUV_VERSION_H_
+
+#define LIBYUV_VERSION 1895
+
+#endif  // INCLUDE_LIBYUV_VERSION_H_

yuv.mod/libyuv/include/libyuv/video_common.h

@@ -0,0 +1,222 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+// Common definitions for video, including fourcc and VideoFormat.
+
+#ifndef INCLUDE_LIBYUV_VIDEO_COMMON_H_
+#define INCLUDE_LIBYUV_VIDEO_COMMON_H_
+
+#include "libyuv/basic_types.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+//////////////////////////////////////////////////////////////////////////////
+// Definition of FourCC codes
+//////////////////////////////////////////////////////////////////////////////
+
+// Convert four characters to a FourCC code.
+// Needs to be a macro otherwise the OS X compiler complains when the kFormat*
+// constants are used in a switch.
+#ifdef __cplusplus
+#define FOURCC(a, b, c, d)                                        \
+  ((static_cast<uint32_t>(a)) | (static_cast<uint32_t>(b) << 8) | \
+   (static_cast<uint32_t>(c) << 16) | /* NOLINT */                \
+   (static_cast<uint32_t>(d) << 24))  /* NOLINT */
+#else
+#define FOURCC(a, b, c, d)                                     \
+  (((uint32_t)(a)) | ((uint32_t)(b) << 8) |       /* NOLINT */ \
+   ((uint32_t)(c) << 16) | ((uint32_t)(d) << 24)) /* NOLINT */
+#endif
+
+// Some pages discussing FourCC codes:
+//   http://www.fourcc.org/yuv.php
+//   http://v4l2spec.bytesex.org/spec/book1.htm
+//   http://developer.apple.com/quicktime/icefloe/dispatch020.html
+//   http://msdn.microsoft.com/library/windows/desktop/dd206750.aspx#nv12
+//   http://people.xiph.org/~xiphmont/containers/nut/nut4cc.txt
+
+// FourCC codes grouped according to implementation efficiency.
+// Primary formats should convert in 1 efficient step.
+// Secondary formats are converted in 2 steps.
+// Auxilliary formats call primary converters.
+enum FourCC {
+  // 10 Primary YUV formats: 5 planar, 2 biplanar, 2 packed.
+  FOURCC_I420 = FOURCC('I', '4', '2', '0'),
+  FOURCC_I422 = FOURCC('I', '4', '2', '2'),
+  FOURCC_I444 = FOURCC('I', '4', '4', '4'),
+  FOURCC_I400 = FOURCC('I', '4', '0', '0'),
+  FOURCC_NV21 = FOURCC('N', 'V', '2', '1'),
+  FOURCC_NV12 = FOURCC('N', 'V', '1', '2'),
+  FOURCC_YUY2 = FOURCC('Y', 'U', 'Y', '2'),
+  FOURCC_UYVY = FOURCC('U', 'Y', 'V', 'Y'),
+  FOURCC_I010 = FOURCC('I', '0', '1', '0'),  // bt.601 10 bit 420
+  FOURCC_I210 = FOURCC('I', '2', '1', '0'),  // bt.601 10 bit 422
+
+  // 1 Secondary YUV format: row biplanar.  deprecated.
+  FOURCC_M420 = FOURCC('M', '4', '2', '0'),
+
+  // 13 Primary RGB formats: 4 32 bpp, 2 24 bpp, 3 16 bpp, 1 10 bpc 2 64 bpp
+  FOURCC_ARGB = FOURCC('A', 'R', 'G', 'B'),
+  FOURCC_BGRA = FOURCC('B', 'G', 'R', 'A'),
+  FOURCC_ABGR = FOURCC('A', 'B', 'G', 'R'),
+  FOURCC_AR30 = FOURCC('A', 'R', '3', '0'),  // 10 bit per channel. 2101010.
+  FOURCC_AB30 = FOURCC('A', 'B', '3', '0'),  // ABGR version of 10 bit
+  FOURCC_AR64 = FOURCC('A', 'R', '6', '4'),  // 16 bit per channel.
+  FOURCC_AB64 = FOURCC('A', 'B', '6', '4'),  // ABGR version of 16 bit
+  FOURCC_24BG = FOURCC('2', '4', 'B', 'G'),
+  FOURCC_RAW = FOURCC('r', 'a', 'w', ' '),
+  FOURCC_RGBA = FOURCC('R', 'G', 'B', 'A'),
+  FOURCC_RGBP = FOURCC('R', 'G', 'B', 'P'),  // rgb565 LE.
+  FOURCC_RGBO = FOURCC('R', 'G', 'B', 'O'),  // argb1555 LE.
+  FOURCC_R444 = FOURCC('R', '4', '4', '4'),  // argb4444 LE.
+
+  // 1 Primary Compressed YUV format.
+  FOURCC_MJPG = FOURCC('M', 'J', 'P', 'G'),
+
+  // 14 Auxiliary YUV variations: 3 with U and V planes are swapped, 1 Alias.
+  FOURCC_YV12 = FOURCC('Y', 'V', '1', '2'),
+  FOURCC_YV16 = FOURCC('Y', 'V', '1', '6'),
+  FOURCC_YV24 = FOURCC('Y', 'V', '2', '4'),
+  FOURCC_YU12 = FOURCC('Y', 'U', '1', '2'),  // Linux version of I420.
+  FOURCC_J420 =
+      FOURCC('J', '4', '2', '0'),  // jpeg (bt.601 full), unofficial fourcc
+  FOURCC_J422 =
+      FOURCC('J', '4', '2', '2'),  // jpeg (bt.601 full), unofficial fourcc
+  FOURCC_J444 =
+      FOURCC('J', '4', '4', '4'),  // jpeg (bt.601 full), unofficial fourcc
+  FOURCC_J400 =
+      FOURCC('J', '4', '0', '0'),  // jpeg (bt.601 full), unofficial fourcc
+  FOURCC_F420 = FOURCC('F', '4', '2', '0'),  // bt.709 full, unofficial fourcc
+  FOURCC_F422 = FOURCC('F', '4', '2', '2'),  // bt.709 full, unofficial fourcc
+  FOURCC_F444 = FOURCC('F', '4', '4', '4'),  // bt.709 full, unofficial fourcc
+  FOURCC_H420 = FOURCC('H', '4', '2', '0'),  // bt.709, unofficial fourcc
+  FOURCC_H422 = FOURCC('H', '4', '2', '2'),  // bt.709, unofficial fourcc
+  FOURCC_H444 = FOURCC('H', '4', '4', '4'),  // bt.709, unofficial fourcc
+  FOURCC_U420 = FOURCC('U', '4', '2', '0'),  // bt.2020, unofficial fourcc
+  FOURCC_U422 = FOURCC('U', '4', '2', '2'),  // bt.2020, unofficial fourcc
+  FOURCC_U444 = FOURCC('U', '4', '4', '4'),  // bt.2020, unofficial fourcc
+  FOURCC_F010 = FOURCC('F', '0', '1', '0'),  // bt.709 full range 10 bit 420
+  FOURCC_H010 = FOURCC('H', '0', '1', '0'),  // bt.709 10 bit 420
+  FOURCC_U010 = FOURCC('U', '0', '1', '0'),  // bt.2020 10 bit 420
+  FOURCC_F210 = FOURCC('F', '2', '1', '0'),  // bt.709 full range 10 bit 422
+  FOURCC_H210 = FOURCC('H', '2', '1', '0'),  // bt.709 10 bit 422
+  FOURCC_U210 = FOURCC('U', '2', '1', '0'),  // bt.2020 10 bit 422
+  FOURCC_P010 = FOURCC('P', '0', '1', '0'),
+  FOURCC_P210 = FOURCC('P', '2', '1', '0'),
+
+  // 14 Auxiliary aliases.  CanonicalFourCC() maps these to canonical fourcc.
+  FOURCC_IYUV = FOURCC('I', 'Y', 'U', 'V'),  // Alias for I420.
+  FOURCC_YU16 = FOURCC('Y', 'U', '1', '6'),  // Alias for I422.
+  FOURCC_YU24 = FOURCC('Y', 'U', '2', '4'),  // Alias for I444.
+  FOURCC_YUYV = FOURCC('Y', 'U', 'Y', 'V'),  // Alias for YUY2.
+  FOURCC_YUVS = FOURCC('y', 'u', 'v', 's'),  // Alias for YUY2 on Mac.
+  FOURCC_HDYC = FOURCC('H', 'D', 'Y', 'C'),  // Alias for UYVY.
+  FOURCC_2VUY = FOURCC('2', 'v', 'u', 'y'),  // Alias for UYVY on Mac.
+  FOURCC_JPEG = FOURCC('J', 'P', 'E', 'G'),  // Alias for MJPG.
+  FOURCC_DMB1 = FOURCC('d', 'm', 'b', '1'),  // Alias for MJPG on Mac.
+  FOURCC_BA81 = FOURCC('B', 'A', '8', '1'),  // Alias for BGGR.
+  FOURCC_RGB3 = FOURCC('R', 'G', 'B', '3'),  // Alias for RAW.
+  FOURCC_BGR3 = FOURCC('B', 'G', 'R', '3'),  // Alias for 24BG.
+  FOURCC_CM32 = FOURCC(0, 0, 0, 32),  // Alias for BGRA kCMPixelFormat_32ARGB
+  FOURCC_CM24 = FOURCC(0, 0, 0, 24),  // Alias for RAW kCMPixelFormat_24RGB
+  FOURCC_L555 = FOURCC('L', '5', '5', '5'),  // Alias for RGBO.
+  FOURCC_L565 = FOURCC('L', '5', '6', '5'),  // Alias for RGBP.
+  FOURCC_5551 = FOURCC('5', '5', '5', '1'),  // Alias for RGBO.
+
+  // deprecated formats.  Not supported, but defined for backward compatibility.
+  FOURCC_I411 = FOURCC('I', '4', '1', '1'),
+  FOURCC_Q420 = FOURCC('Q', '4', '2', '0'),
+  FOURCC_RGGB = FOURCC('R', 'G', 'G', 'B'),
+  FOURCC_BGGR = FOURCC('B', 'G', 'G', 'R'),
+  FOURCC_GRBG = FOURCC('G', 'R', 'B', 'G'),
+  FOURCC_GBRG = FOURCC('G', 'B', 'R', 'G'),
+  FOURCC_H264 = FOURCC('H', '2', '6', '4'),
+
+  // Match any fourcc.
+  FOURCC_ANY = -1,
+};
+
+enum FourCCBpp {
+  // Canonical fourcc codes used in our code.
+  FOURCC_BPP_I420 = 12,
+  FOURCC_BPP_I422 = 16,
+  FOURCC_BPP_I444 = 24,
+  FOURCC_BPP_I411 = 12,
+  FOURCC_BPP_I400 = 8,
+  FOURCC_BPP_NV21 = 12,
+  FOURCC_BPP_NV12 = 12,
+  FOURCC_BPP_YUY2 = 16,
+  FOURCC_BPP_UYVY = 16,
+  FOURCC_BPP_M420 = 12,  // deprecated
+  FOURCC_BPP_Q420 = 12,
+  FOURCC_BPP_ARGB = 32,
+  FOURCC_BPP_BGRA = 32,
+  FOURCC_BPP_ABGR = 32,
+  FOURCC_BPP_RGBA = 32,
+  FOURCC_BPP_AR30 = 32,
+  FOURCC_BPP_AB30 = 32,
+  FOURCC_BPP_AR64 = 64,
+  FOURCC_BPP_AB64 = 64,
+  FOURCC_BPP_24BG = 24,
+  FOURCC_BPP_RAW = 24,
+  FOURCC_BPP_RGBP = 16,
+  FOURCC_BPP_RGBO = 16,
+  FOURCC_BPP_R444 = 16,
+  FOURCC_BPP_RGGB = 8,
+  FOURCC_BPP_BGGR = 8,
+  FOURCC_BPP_GRBG = 8,
+  FOURCC_BPP_GBRG = 8,
+  FOURCC_BPP_YV12 = 12,
+  FOURCC_BPP_YV16 = 16,
+  FOURCC_BPP_YV24 = 24,
+  FOURCC_BPP_YU12 = 12,
+  FOURCC_BPP_J420 = 12,
+  FOURCC_BPP_J400 = 8,
+  FOURCC_BPP_H420 = 12,
+  FOURCC_BPP_H422 = 16,
+  FOURCC_BPP_I010 = 15,
+  FOURCC_BPP_I210 = 20,
+  FOURCC_BPP_H010 = 15,
+  FOURCC_BPP_H210 = 20,
+  FOURCC_BPP_P010 = 15,
+  FOURCC_BPP_P210 = 20,
+  FOURCC_BPP_MJPG = 0,  // 0 means unknown.
+  FOURCC_BPP_H264 = 0,
+  FOURCC_BPP_IYUV = 12,
+  FOURCC_BPP_YU16 = 16,
+  FOURCC_BPP_YU24 = 24,
+  FOURCC_BPP_YUYV = 16,
+  FOURCC_BPP_YUVS = 16,
+  FOURCC_BPP_HDYC = 16,
+  FOURCC_BPP_2VUY = 16,
+  FOURCC_BPP_JPEG = 1,
+  FOURCC_BPP_DMB1 = 1,
+  FOURCC_BPP_BA81 = 8,
+  FOURCC_BPP_RGB3 = 24,
+  FOURCC_BPP_BGR3 = 24,
+  FOURCC_BPP_CM32 = 32,
+  FOURCC_BPP_CM24 = 24,
+
+  // Match any fourcc.
+  FOURCC_BPP_ANY = 0,  // 0 means unknown.
+};
+
+// Converts fourcc aliases into canonical ones.
+LIBYUV_API uint32_t CanonicalFourCC(uint32_t fourcc);
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // INCLUDE_LIBYUV_VIDEO_COMMON_H_

yuv.mod/libyuv/infra/config/OWNERS

@@ -0,0 +1,3 @@
[email protected]
[email protected]
[email protected]

yuv.mod/libyuv/infra/config/PRESUBMIT.py

@@ -0,0 +1,13 @@
+# Copyright 2018 The PDFium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+USE_PYTHON3 = True
+
+
+def CheckChangeOnUpload(input_api, output_api):
+  return input_api.canned_checks.CheckChangedLUCIConfigs(input_api, output_api)
+
+
+def CheckChangeOnCommit(input_api, output_api):
+  return input_api.canned_checks.CheckChangedLUCIConfigs(input_api, output_api)

yuv.mod/libyuv/infra/config/README.md

@@ -0,0 +1,2 @@
+This folder contains libyuv project-wide configurations
+for chrome-infra services.

yuv.mod/libyuv/infra/config/codereview.settings

@@ -0,0 +1,6 @@
+# This file is used by gcl and git-cl to get repository specific information.
+CODE_REVIEW_SERVER: codereview.chromium.org
+PROJECT: libyuv
+GERRIT_HOST: True
+VIEW_VC: https://chromium.googlesource.com/libyuv/libyuv/+/
+

yuv.mod/libyuv/infra/config/commit-queue.cfg

@@ -0,0 +1,143 @@
+# Auto-generated by lucicfg.
+# Do not modify manually.
+#
+# For the schema of this file, see Config message:
+#   https://config.luci.app/schemas/projects:commit-queue.cfg
+
+cq_status_host: "chromium-cq-status.appspot.com"
+submit_options {
+  max_burst: 4
+  burst_delay {
+    seconds: 480
+  }
+}
+config_groups {
+  name: "config"
+  gerrit {
+    url: "https://chromium-review.googlesource.com"
+    projects {
+      name: "libyuv/libyuv"
+      ref_regexp: "refs/heads/infra/config"
+    }
+  }
+  verifiers {
+    gerrit_cq_ability {
+      committer_list: "project-libyuv-committers"
+      dry_run_access_list: "project-libyuv-tryjob-access"
+    }
+    tryjob {
+      builders {
+        name: "libyuv/try/presubmit"
+      }
+      retry_config {
+        single_quota: 1
+        global_quota: 2
+        failure_weight: 1
+        transient_failure_weight: 1
+        timeout_weight: 2
+      }
+    }
+  }
+}
+config_groups {
+  name: "master"
+  gerrit {
+    url: "https://chromium-review.googlesource.com"
+    projects {
+      name: "libyuv/libyuv"
+      ref_regexp: "refs/heads/main"
+      ref_regexp: "refs/heads/master"
+    }
+  }
+  verifiers {
+    gerrit_cq_ability {
+      committer_list: "project-libyuv-committers"
+      dry_run_access_list: "project-libyuv-tryjob-access"
+    }
+    tryjob {
+      builders {
+        name: "libyuv/try/android"
+        experiment_percentage: 100
+      }
+      builders {
+        name: "libyuv/try/android_arm64"
+        experiment_percentage: 100
+      }
+      builders {
+        name: "libyuv/try/android_rel"
+        experiment_percentage: 100
+      }
+      builders {
+        name: "libyuv/try/android_x64"
+      }
+      builders {
+        name: "libyuv/try/android_x86"
+      }
+      builders {
+        name: "libyuv/try/ios_arm64"
+      }
+      builders {
+        name: "libyuv/try/ios_arm64_rel"
+      }
+      builders {
+        name: "libyuv/try/linux"
+      }
+      builders {
+        name: "libyuv/try/linux_asan"
+      }
+      builders {
+        name: "libyuv/try/linux_gcc"
+        experiment_percentage: 100
+      }
+      builders {
+        name: "libyuv/try/linux_msan"
+      }
+      builders {
+        name: "libyuv/try/linux_rel"
+      }
+      builders {
+        name: "libyuv/try/linux_tsan2"
+      }
+      builders {
+        name: "libyuv/try/linux_ubsan"
+      }
+      builders {
+        name: "libyuv/try/linux_ubsan_vptr"
+      }
+      builders {
+        name: "libyuv/try/mac"
+      }
+      builders {
+        name: "libyuv/try/mac_asan"
+      }
+      builders {
+        name: "libyuv/try/mac_rel"
+      }
+      builders {
+        name: "libyuv/try/win"
+      }
+      builders {
+        name: "libyuv/try/win_clang"
+      }
+      builders {
+        name: "libyuv/try/win_clang_rel"
+      }
+      builders {
+        name: "libyuv/try/win_rel"
+      }
+      builders {
+        name: "libyuv/try/win_x64_clang_rel"
+      }
+      builders {
+        name: "libyuv/try/win_x64_rel"
+      }
+      retry_config {
+        single_quota: 1
+        global_quota: 2
+        failure_weight: 1
+        transient_failure_weight: 1
+        timeout_weight: 2
+      }
+    }
+  }
+}

yuv.mod/libyuv/infra/config/cr-buildbucket.cfg

@@ -0,0 +1,1704 @@
+# Auto-generated by lucicfg.
+# Do not modify manually.
+#
+# For the schema of this file, see BuildbucketCfg message:
+#   https://config.luci.app/schemas/projects:buildbucket.cfg
+
+buckets {
+  name: "ci"
+  acls {
+    role: WRITER
+    group: "project-libyuv-admins"
+  }
+  acls {
+    group: "all"
+  }
+  swarming {
+    builders {
+      name: "Android ARM64 Debug"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Android Debug"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Android Release"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Android Tester ARM32 Debug (Nexus 5X)"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "device_type:walleye"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Android Tester ARM32 Release (Nexus 5X)"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "device_type:walleye"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Android Tester ARM64 Debug (Nexus 5X)"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "device_type:walleye"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Android32 x86 Debug"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Android64 x64 Debug"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Linux Asan"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Linux MSan"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Linux Tsan v2"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Linux UBSan"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Linux UBSan vptr"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Linux32 Debug"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Linux32 Release"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Linux64 Debug"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Linux64 Release"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Mac Asan"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Mac-12"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Mac64 Debug"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Mac-12"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Mac64 Release"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Mac-12"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Win32 Debug"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Win32 Debug (Clang)"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Win32 Release"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Win32 Release (Clang)"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Win64 Debug"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Win64 Debug (Clang)"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Win64 Release"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "Win64 Release (Clang)"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "iOS ARM64 Debug"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Mac-12"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "iOS ARM64 Release"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Mac-12"
+      dimensions: "pool:luci.flex.ci"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-trusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "client.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "[email protected]"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+  }
+}
+buckets {
+  name: "cron"
+  acls {
+    role: WRITER
+    group: "project-libyuv-admins"
+  }
+  acls {
+    group: "all"
+  }
+  swarming {
+    builders {
+      name: "DEPS Autoroller"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Linux"
+      dimensions: "pool:luci.webrtc.cron"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "recipe": "libyuv/roll_deps"'
+        '}'
+      execution_timeout_secs: 7200
+      build_numbers: YES
+      service_account: "libyuv-ci-autoroll-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+  }
+}
+buckets {
+  name: "try"
+  acls {
+    role: WRITER
+    group: "project-libyuv-admins"
+  }
+  acls {
+    group: "all"
+  }
+  acls {
+    role: SCHEDULER
+    group: "project-libyuv-tryjob-access"
+  }
+  acls {
+    role: SCHEDULER
+    group: "service-account-cq"
+  }
+  swarming {
+    builders {
+      name: "android"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "device_type:walleye"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "android_arm64"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "device_type:walleye"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "android_rel"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "device_type:walleye"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "android_x64"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "android_x86"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "ios_arm64"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Mac-12"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "ios_arm64_rel"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Mac-12"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "linux"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "linux_asan"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "linux_gcc"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "linux_msan"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "linux_rel"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "linux_tsan2"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "linux_ubsan"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "linux_ubsan_vptr"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "mac"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Mac-12"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "mac_asan"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Mac-12"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "mac_rel"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Mac-12"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "presubmit"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Ubuntu-22.04"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "run_presubmit",'
+        '  "repo_name": "libyuv",'
+        '  "runhooks": true'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "win"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "win_clang"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "win_clang_rel"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "win_rel"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "win_x64_clang_rel"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+    builders {
+      name: "win_x64_rel"
+      swarming_host: "chromium-swarm.appspot.com"
+      swarming_tags: "vpython:native-python-wrapper"
+      dimensions: "cores:8"
+      dimensions: "cpu:x86-64"
+      dimensions: "os:Windows-10"
+      dimensions: "pool:luci.flex.try"
+      exe {
+        cipd_package: "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build"
+        cipd_version: "refs/heads/main"
+        cmd: "luciexe"
+      }
+      properties:
+        '{'
+        '  "$build/reclient": {'
+        '    "instance": "rbe-webrtc-untrusted",'
+        '    "metrics_project": "chromium-reclient-metrics"'
+        '  },'
+        '  "builder_group": "tryserver.libyuv",'
+        '  "recipe": "libyuv/libyuv"'
+        '}'
+      execution_timeout_secs: 10800
+      build_numbers: YES
+      service_account: "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+      experiments {
+        key: "luci.recipes.use_python3"
+        value: 100
+      }
+    }
+  }
+}

yuv.mod/libyuv/infra/config/luci-logdog.cfg

@@ -0,0 +1,9 @@
+# Auto-generated by lucicfg.
+# Do not modify manually.
+#
+# For the schema of this file, see ProjectConfig message:
+#   https://config.luci.app/schemas/projects:luci-logdog.cfg
+
+reader_auth_groups: "all"
+writer_auth_groups: "luci-logdog-chromium-writers"
+archive_gs_bucket: "chromium-luci-logdog"

yuv.mod/libyuv/infra/config/luci-milo.cfg

@@ -0,0 +1,246 @@
+# Auto-generated by lucicfg.
+# Do not modify manually.
+#
+# For the schema of this file, see Project message:
+#   https://config.luci.app/schemas/projects:luci-milo.cfg
+
+consoles {
+  id: "main"
+  name: "libyuv Main Console"
+  repo_url: "https://chromium.googlesource.com/libyuv/libyuv"
+  refs: "regexp:refs/heads/main"
+  manifest_name: "REVISION"
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Android ARM64 Debug"
+    category: "Android|Builder"
+    short_name: "dbg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Android Debug"
+    category: "Android|Builder"
+    short_name: "dbg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Android Release"
+    category: "Android|Builder"
+    short_name: "rel"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Android32 x86 Debug"
+    category: "Android|Builder|x86"
+    short_name: "dbg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Android64 x64 Debug"
+    category: "Android|Builder|x64"
+    short_name: "dbg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Android Tester ARM32 Debug (Nexus 5X)"
+    category: "Android|Tester|ARM 32"
+    short_name: "dbg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Android Tester ARM32 Release (Nexus 5X)"
+    category: "Android|Tester|ARM 32"
+    short_name: "rel"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Android Tester ARM64 Debug (Nexus 5X)"
+    category: "Android|Tester|ARM 64"
+    short_name: "dbg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Linux Asan"
+    category: "Linux"
+    short_name: "asan"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Linux MSan"
+    category: "Linux"
+    short_name: "msan"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Linux Tsan v2"
+    category: "Linux"
+    short_name: "tsan"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Linux UBSan"
+    category: "Linux|UBSan"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Linux UBSan vptr"
+    category: "Linux|UBSan"
+    short_name: "vptr"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Linux32 Debug"
+    category: "Linux|32"
+    short_name: "dbg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Linux32 Release"
+    category: "Linux|32"
+    short_name: "rel"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Linux64 Debug"
+    category: "Linux|64"
+    short_name: "dbg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Linux64 Release"
+    category: "Linux|64"
+    short_name: "rel"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Mac Asan"
+    category: "Mac"
+    short_name: "asan"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Mac64 Debug"
+    category: "Mac"
+    short_name: "dbg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Mac64 Release"
+    category: "Mac"
+    short_name: "rel"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Win32 Debug"
+    category: "Win|32|Debug"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Win32 Debug (Clang)"
+    category: "Win|32|Debug"
+    short_name: "clg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Win32 Release"
+    category: "Win|32|Release"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Win32 Release (Clang)"
+    category: "Win|32|Release"
+    short_name: "clg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Win64 Debug"
+    category: "Win|64|Debug"
+    short_name: "clg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Win64 Debug (Clang)"
+    category: "Win|64|Debug"
+    short_name: "clg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Win64 Release"
+    category: "Win|64|Release"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/Win64 Release (Clang)"
+    category: "Win|64|Release"
+    short_name: "clg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/iOS ARM64 Debug"
+    category: "iOS|ARM64"
+    short_name: "dbg"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.ci/iOS ARM64 Release"
+    category: "iOS|ARM64"
+    short_name: "rel"
+  }
+  include_experimental_builds: true
+}
+consoles {
+  id: "cron"
+  name: "Cron"
+  builders {
+    name: "buildbucket/luci.libyuv.cron/DEPS Autoroller"
+  }
+  builder_view_only: true
+}
+consoles {
+  id: "try"
+  name: "libyuv Try Builders"
+  builders {
+    name: "buildbucket/luci.libyuv.try/android"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/android_arm64"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/android_rel"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/android_x64"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/android_x86"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/ios_arm64"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/ios_arm64_rel"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/linux"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/linux_asan"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/linux_gcc"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/linux_msan"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/linux_rel"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/linux_tsan2"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/linux_ubsan"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/linux_ubsan_vptr"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/mac"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/mac_asan"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/mac_rel"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/win"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/win_clang"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/win_clang_rel"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/win_rel"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/win_x64_clang_rel"
+  }
+  builders {
+    name: "buildbucket/luci.libyuv.try/win_x64_rel"
+  }
+  builder_view_only: true
+}
+logo_url: "https://storage.googleapis.com/chrome-infra-public/logo/libyuv-logo.png"

yuv.mod/libyuv/infra/config/luci-scheduler.cfg

@@ -0,0 +1,385 @@
+# Auto-generated by lucicfg.
+# Do not modify manually.
+#
+# For the schema of this file, see ProjectConfig message:
+#   https://config.luci.app/schemas/projects:luci-scheduler.cfg
+
+job {
+  id: "Android ARM64 Debug"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Android ARM64 Debug"
+  }
+}
+job {
+  id: "Android Debug"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Android Debug"
+  }
+}
+job {
+  id: "Android Release"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Android Release"
+  }
+}
+job {
+  id: "Android Tester ARM32 Debug (Nexus 5X)"
+  realm: "ci"
+  acls {
+    role: TRIGGERER
+    granted_to: "[email protected]"
+  }
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Android Tester ARM32 Debug (Nexus 5X)"
+  }
+}
+job {
+  id: "Android Tester ARM32 Release (Nexus 5X)"
+  realm: "ci"
+  acls {
+    role: TRIGGERER
+    granted_to: "[email protected]"
+  }
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Android Tester ARM32 Release (Nexus 5X)"
+  }
+}
+job {
+  id: "Android Tester ARM64 Debug (Nexus 5X)"
+  realm: "ci"
+  acls {
+    role: TRIGGERER
+    granted_to: "[email protected]"
+  }
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Android Tester ARM64 Debug (Nexus 5X)"
+  }
+}
+job {
+  id: "Android32 x86 Debug"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Android32 x86 Debug"
+  }
+}
+job {
+  id: "Android64 x64 Debug"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Android64 x64 Debug"
+  }
+}
+job {
+  id: "DEPS Autoroller"
+  realm: "cron"
+  schedule: "0 14 * * *"
+  acl_sets: "cron"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "cron"
+    builder: "DEPS Autoroller"
+  }
+}
+job {
+  id: "Linux Asan"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Linux Asan"
+  }
+}
+job {
+  id: "Linux MSan"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Linux MSan"
+  }
+}
+job {
+  id: "Linux Tsan v2"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Linux Tsan v2"
+  }
+}
+job {
+  id: "Linux UBSan"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Linux UBSan"
+  }
+}
+job {
+  id: "Linux UBSan vptr"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Linux UBSan vptr"
+  }
+}
+job {
+  id: "Linux32 Debug"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Linux32 Debug"
+  }
+}
+job {
+  id: "Linux32 Release"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Linux32 Release"
+  }
+}
+job {
+  id: "Linux64 Debug"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Linux64 Debug"
+  }
+}
+job {
+  id: "Linux64 Release"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Linux64 Release"
+  }
+}
+job {
+  id: "Mac Asan"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Mac Asan"
+  }
+}
+job {
+  id: "Mac64 Debug"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Mac64 Debug"
+  }
+}
+job {
+  id: "Mac64 Release"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Mac64 Release"
+  }
+}
+job {
+  id: "Win32 Debug"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Win32 Debug"
+  }
+}
+job {
+  id: "Win32 Debug (Clang)"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Win32 Debug (Clang)"
+  }
+}
+job {
+  id: "Win32 Release"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Win32 Release"
+  }
+}
+job {
+  id: "Win32 Release (Clang)"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Win32 Release (Clang)"
+  }
+}
+job {
+  id: "Win64 Debug"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Win64 Debug"
+  }
+}
+job {
+  id: "Win64 Debug (Clang)"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Win64 Debug (Clang)"
+  }
+}
+job {
+  id: "Win64 Release"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Win64 Release"
+  }
+}
+job {
+  id: "Win64 Release (Clang)"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "Win64 Release (Clang)"
+  }
+}
+job {
+  id: "iOS ARM64 Debug"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "iOS ARM64 Debug"
+  }
+}
+job {
+  id: "iOS ARM64 Release"
+  realm: "ci"
+  acl_sets: "ci"
+  buildbucket {
+    server: "cr-buildbucket.appspot.com"
+    bucket: "ci"
+    builder: "iOS ARM64 Release"
+  }
+}
+trigger {
+  id: "master-gitiles-trigger"
+  realm: "ci"
+  acl_sets: "ci"
+  triggers: "Android ARM64 Debug"
+  triggers: "Android Debug"
+  triggers: "Android Release"
+  triggers: "Android32 x86 Debug"
+  triggers: "Android64 x64 Debug"
+  triggers: "Linux Asan"
+  triggers: "Linux MSan"
+  triggers: "Linux Tsan v2"
+  triggers: "Linux UBSan"
+  triggers: "Linux UBSan vptr"
+  triggers: "Linux32 Debug"
+  triggers: "Linux32 Release"
+  triggers: "Linux64 Debug"
+  triggers: "Linux64 Release"
+  triggers: "Mac Asan"
+  triggers: "Mac64 Debug"
+  triggers: "Mac64 Release"
+  triggers: "Win32 Debug"
+  triggers: "Win32 Debug (Clang)"
+  triggers: "Win32 Release"
+  triggers: "Win32 Release (Clang)"
+  triggers: "Win64 Debug"
+  triggers: "Win64 Debug (Clang)"
+  triggers: "Win64 Release"
+  triggers: "Win64 Release (Clang)"
+  triggers: "iOS ARM64 Debug"
+  triggers: "iOS ARM64 Release"
+  gitiles {
+    repo: "https://chromium.googlesource.com/libyuv/libyuv"
+    refs: "regexp:refs/heads/main"
+  }
+}
+acl_sets {
+  name: "ci"
+  acls {
+    role: OWNER
+    granted_to: "group:project-libyuv-admins"
+  }
+  acls {
+    granted_to: "group:all"
+  }
+}
+acl_sets {
+  name: "cron"
+  acls {
+    role: OWNER
+    granted_to: "group:project-libyuv-admins"
+  }
+  acls {
+    granted_to: "group:all"
+  }
+}

yuv.mod/libyuv/infra/config/main.star

@@ -0,0 +1,344 @@
+#!/usr/bin/env lucicfg
+# https://chromium.googlesource.com/infra/luci/luci-go/+/master/lucicfg/doc/
+
+"""LUCI project configuration for libyuv CQ and CI."""
+
+lucicfg.check_version("1.30.9")
+
+LIBYUV_GIT = "https://chromium.googlesource.com/libyuv/libyuv"
+LIBYUV_GERRIT = "https://chromium-review.googlesource.com/libyuv/libyuv"
+
+RECLIENT_CI = {
+    "instance": "rbe-webrtc-trusted",
+    "metrics_project": "chromium-reclient-metrics",
+}
+
+RECLIENT_CQ = {
+    "instance": "rbe-webrtc-untrusted",
+    "metrics_project": "chromium-reclient-metrics",
+}
+
+# Use LUCI Scheduler BBv2 names and add Scheduler realms configs.
+lucicfg.enable_experiment("crbug.com/1182002")
+
+luci.builder.defaults.experiments.set(
+    {
+        "luci.recipes.use_python3": 100,
+    },
+)
+
+lucicfg.config(
+    lint_checks = ["default"],
+    config_dir = ".",
+    tracked_files = [
+        "commit-queue.cfg",
+        "cr-buildbucket.cfg",
+        "luci-logdog.cfg",
+        "luci-milo.cfg",
+        "luci-scheduler.cfg",
+        "project.cfg",
+        "realms.cfg",
+    ],
+)
+
+# Generates project.cfg
+
+luci.project(
+    name = "libyuv",
+    buildbucket = "cr-buildbucket.appspot.com",
+    logdog = "luci-logdog.appspot.com",
+    milo = "luci-milo.appspot.com",
+    notify = "luci-notify.appspot.com",
+    scheduler = "luci-scheduler.appspot.com",
+    swarming = "chromium-swarm.appspot.com",
+    acls = [
+        acl.entry(acl.PROJECT_CONFIGS_READER, groups = ["all"]),
+        acl.entry(acl.LOGDOG_READER, groups = ["all"]),
+        acl.entry(acl.LOGDOG_WRITER, groups = ["luci-logdog-chromium-writers"]),
+        acl.entry(acl.SCHEDULER_READER, groups = ["all"]),
+        acl.entry(acl.SCHEDULER_OWNER, groups = ["project-libyuv-admins"]),
+        acl.entry(acl.BUILDBUCKET_READER, groups = ["all"]),
+        acl.entry(acl.BUILDBUCKET_OWNER, groups = ["project-libyuv-admins"]),
+    ],
+    bindings = [
+        luci.binding(
+            roles = "role/swarming.taskTriggerer",  # for LED tasks.
+            groups = "project-libyuv-admins",
+        ),
+        luci.binding(
+            roles = "role/configs.validator",
+            users = "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com",
+        ),
+    ],
+)
+
+# Generates luci-logdog.cfg
+
+luci.logdog(
+    gs_bucket = "chromium-luci-logdog",
+)
+
+# Generates luci-scheduler.cfg
+
+luci.gitiles_poller(
+    name = "master-gitiles-trigger",
+    bucket = "ci",
+    repo = LIBYUV_GIT,
+)
+
+# Generates luci-milo.cfg
+
+luci.milo(
+    logo = "https://storage.googleapis.com/chrome-infra-public/logo/libyuv-logo.png",
+)
+
+def libyuv_ci_view(name, category, short_name):
+    return luci.console_view_entry(
+        console_view = "main",
+        builder = name,
+        category = category,
+        short_name = short_name,
+    )
+
+def libyuv_try_view(name):
+    return luci.list_view_entry(
+        list_view = "try",
+        builder = name,
+    )
+
+luci.console_view(
+    name = "main",
+    title = "libyuv Main Console",
+    include_experimental_builds = True,
+    repo = LIBYUV_GIT,
+)
+
+luci.list_view(
+    name = "cron",
+    title = "Cron",
+    entries = ["DEPS Autoroller"],
+)
+
+luci.list_view(
+    name = "try",
+    title = "libyuv Try Builders",
+)
+
+# Generates commit-queue.cfg
+
+def libyuv_try_job_verifier(name, cq_group, experiment_percentage):
+    return luci.cq_tryjob_verifier(
+        builder = name,
+        cq_group = cq_group,
+        experiment_percentage = experiment_percentage,
+    )
+
+luci.cq(
+    status_host = "chromium-cq-status.appspot.com",
+    submit_max_burst = 4,
+    submit_burst_delay = 8 * time.minute,
+)
+
+luci.cq_group(
+    name = "master",
+    watch = [
+        cq.refset(
+            repo = LIBYUV_GERRIT,
+            refs = ["refs/heads/main", "refs/heads/master"],
+        ),
+    ],
+    acls = [
+        acl.entry(acl.CQ_COMMITTER, groups = ["project-libyuv-committers"]),
+        acl.entry(acl.CQ_DRY_RUNNER, groups = ["project-libyuv-tryjob-access"]),
+    ],
+    retry_config = cq.RETRY_ALL_FAILURES,
+    cancel_stale_tryjobs = True,
+)
+
+luci.cq_group(
+    name = "config",
+    watch = [
+        cq.refset(
+            repo = LIBYUV_GERRIT,
+            refs = ["refs/heads/infra/config"],
+        ),
+    ],
+    acls = [
+        acl.entry(acl.CQ_COMMITTER, groups = ["project-libyuv-committers"]),
+        acl.entry(acl.CQ_DRY_RUNNER, groups = ["project-libyuv-tryjob-access"]),
+    ],
+    retry_config = cq.RETRY_ALL_FAILURES,
+    cancel_stale_tryjobs = True,
+)
+
+# Generates cr-buildbucket.cfg
+
+luci.bucket(
+    name = "ci",
+)
+luci.bucket(
+    name = "try",
+    acls = [
+        acl.entry(acl.BUILDBUCKET_TRIGGERER, groups = [
+            "project-libyuv-tryjob-access",
+            "service-account-cq",
+        ]),
+    ],
+)
+luci.bucket(
+    name = "cron",
+)
+
+def get_os_dimensions(os):
+    if os == "android":
+        return {"device_type": "walleye"}
+    if os == "ios" or os == "mac":
+        return {"os": "Mac-12", "cpu": "x86-64"}
+    elif os == "win":
+        return {"os": "Windows-10", "cores": "8", "cpu": "x86-64"}
+    elif os == "linux":
+        return {"os": "Ubuntu-22.04", "cores": "8", "cpu": "x86-64"}
+    return {}
+
+def libyuv_ci_builder(name, dimensions, properties, triggered_by):
+    return luci.builder(
+        name = name,
+        dimensions = dimensions,
+        properties = properties,
+        bucket = "ci",
+        service_account = "[email protected]",
+        triggered_by = triggered_by,
+        swarming_tags = ["vpython:native-python-wrapper"],
+        execution_timeout = 180 * time.minute,
+        build_numbers = True,
+        executable = luci.recipe(
+            name = "libyuv/libyuv",
+            cipd_package = "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build",
+            use_python3 = True,
+        ),
+    )
+
+def libyuv_try_builder(name, dimensions, properties, recipe_name = "libyuv/libyuv"):
+    return luci.builder(
+        name = name,
+        dimensions = dimensions,
+        properties = properties,
+        bucket = "try",
+        service_account = "libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com",
+        swarming_tags = ["vpython:native-python-wrapper"],
+        execution_timeout = 180 * time.minute,
+        build_numbers = True,
+        executable = luci.recipe(
+            name = recipe_name,
+            cipd_package = "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build",
+            use_python3 = True,
+        ),
+    )
+
+def ci_builder(name, os, category, short_name = None):
+    dimensions = get_os_dimensions(os)
+    properties = {"$build/reclient": RECLIENT_CI}
+
+    dimensions["pool"] = "luci.flex.ci"
+    properties["builder_group"] = "client.libyuv"
+
+    triggered_by = ["master-gitiles-trigger" if os != "android" else "Android Debug"]
+    libyuv_ci_view(name, category, short_name)
+    return libyuv_ci_builder(name, dimensions, properties, triggered_by)
+
+def try_builder(name, os, experiment_percentage = None):
+    dimensions = get_os_dimensions(os)
+    properties = {"$build/reclient": RECLIENT_CQ}
+
+    dimensions["pool"] = "luci.flex.try"
+    properties["builder_group"] = "tryserver.libyuv"
+
+    if name == "presubmit":
+        recipe_name = "run_presubmit"
+        properties["repo_name"] = "libyuv"
+        properties["runhooks"] = True
+        libyuv_try_job_verifier(name, "config", experiment_percentage)
+        return libyuv_try_builder(name, dimensions, properties, recipe_name)
+
+    libyuv_try_job_verifier(name, "master", experiment_percentage)
+    libyuv_try_view(name)
+    return libyuv_try_builder(name, dimensions, properties)
+
+luci.builder(
+    name = "DEPS Autoroller",
+    bucket = "cron",
+    service_account = "libyuv-ci-autoroll-builder@chops-service-accounts.iam.gserviceaccount.com",
+    dimensions = {
+        "pool": "luci.webrtc.cron",
+        "os": "Linux",
+        "cpu": "x86-64",
+    },
+    swarming_tags = ["vpython:native-python-wrapper"],
+    execution_timeout = 120 * time.minute,
+    build_numbers = True,
+    schedule = "0 14 * * *",  # Every 2 hours.
+    executable = luci.recipe(
+        name = "libyuv/roll_deps",
+        cipd_package = "infra/recipe_bundles/chromium.googlesource.com/chromium/tools/build",
+        use_python3 = True,
+    ),
+)
+
+ci_builder("Android ARM64 Debug", "linux", "Android|Builder", "dbg")
+ci_builder("Android Debug", "linux", "Android|Builder", "dbg")
+ci_builder("Android Release", "linux", "Android|Builder", "rel")
+ci_builder("Android32 x86 Debug", "linux", "Android|Builder|x86", "dbg")
+ci_builder("Android64 x64 Debug", "linux", "Android|Builder|x64", "dbg")
+ci_builder("Android Tester ARM32 Debug (Nexus 5X)", "android", "Android|Tester|ARM 32", "dbg")
+ci_builder("Android Tester ARM32 Release (Nexus 5X)", "android", "Android|Tester|ARM 32", "rel")
+ci_builder("Android Tester ARM64 Debug (Nexus 5X)", "android", "Android|Tester|ARM 64", "dbg")
+ci_builder("Linux Asan", "linux", "Linux", "asan")
+ci_builder("Linux MSan", "linux", "Linux", "msan")
+ci_builder("Linux Tsan v2", "linux", "Linux", "tsan")
+ci_builder("Linux UBSan", "linux", "Linux|UBSan")
+ci_builder("Linux UBSan vptr", "linux", "Linux|UBSan", "vptr")
+ci_builder("Linux32 Debug", "linux", "Linux|32", "dbg")
+ci_builder("Linux32 Release", "linux", "Linux|32", "rel")
+ci_builder("Linux64 Debug", "linux", "Linux|64", "dbg")
+ci_builder("Linux64 Release", "linux", "Linux|64", "rel")
+ci_builder("Mac Asan", "mac", "Mac", "asan")
+ci_builder("Mac64 Debug", "mac", "Mac", "dbg")
+ci_builder("Mac64 Release", "mac", "Mac", "rel")
+ci_builder("Win32 Debug", "win", "Win|32|Debug")
+ci_builder("Win32 Debug (Clang)", "win", "Win|32|Debug", "clg")
+ci_builder("Win32 Release", "win", "Win|32|Release")
+ci_builder("Win32 Release (Clang)", "win", "Win|32|Release", "clg")
+ci_builder("Win64 Debug", "win", "Win|64|Debug", "clg")
+ci_builder("Win64 Debug (Clang)", "win", "Win|64|Debug", "clg")
+ci_builder("Win64 Release", "win", "Win|64|Release")
+ci_builder("Win64 Release (Clang)", "win", "Win|64|Release", "clg")
+ci_builder("iOS ARM64 Debug", "ios", "iOS|ARM64", "dbg")
+ci_builder("iOS ARM64 Release", "ios", "iOS|ARM64", "rel")
+
+# TODO(crbug.com/1242847): make this not experimental.
+try_builder("android", "android", experiment_percentage = 100)
+try_builder("android_arm64", "android", experiment_percentage = 100)
+try_builder("android_rel", "android", experiment_percentage = 100)
+
+try_builder("android_x64", "linux")
+try_builder("android_x86", "linux")
+try_builder("ios_arm64", "ios")
+try_builder("ios_arm64_rel", "ios")
+try_builder("linux", "linux")
+try_builder("linux_asan", "linux")
+try_builder("linux_gcc", "linux", experiment_percentage = 100)
+try_builder("linux_msan", "linux")
+try_builder("linux_rel", "linux")
+try_builder("linux_tsan2", "linux")
+try_builder("linux_ubsan", "linux")
+try_builder("linux_ubsan_vptr", "linux")
+try_builder("mac", "mac")
+try_builder("mac_asan", "mac")
+try_builder("mac_rel", "mac")
+try_builder("win", "win")
+try_builder("win_clang", "win")
+try_builder("win_clang_rel", "win")
+try_builder("win_rel", "win")
+try_builder("win_x64_clang_rel", "win")
+try_builder("win_x64_rel", "win")
+try_builder("presubmit", "linux")

yuv.mod/libyuv/infra/config/project.cfg

@@ -0,0 +1,15 @@
+# Auto-generated by lucicfg.
+# Do not modify manually.
+#
+# For the schema of this file, see ProjectCfg message:
+#   https://config.luci.app/schemas/projects:project.cfg
+
+name: "libyuv"
+access: "group:all"
+lucicfg {
+  version: "1.43.6"
+  package_dir: "."
+  config_dir: "."
+  entry_point: "main.star"
+  experiments: "crbug.com/1182002"
+}

yuv.mod/libyuv/infra/config/realms.cfg

@@ -0,0 +1,83 @@
+# Auto-generated by lucicfg.
+# Do not modify manually.
+#
+# For the schema of this file, see RealmsCfg message:
+#   https://config.luci.app/schemas/projects:realms.cfg
+
+realms {
+  name: "@root"
+  bindings {
+    role: "role/buildbucket.owner"
+    principals: "group:project-libyuv-admins"
+  }
+  bindings {
+    role: "role/buildbucket.reader"
+    principals: "group:all"
+  }
+  bindings {
+    role: "role/configs.reader"
+    principals: "group:all"
+  }
+  bindings {
+    role: "role/configs.validator"
+    principals: "user:libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+  }
+  bindings {
+    role: "role/logdog.reader"
+    principals: "group:all"
+  }
+  bindings {
+    role: "role/logdog.writer"
+    principals: "group:luci-logdog-chromium-writers"
+  }
+  bindings {
+    role: "role/scheduler.owner"
+    principals: "group:project-libyuv-admins"
+  }
+  bindings {
+    role: "role/scheduler.reader"
+    principals: "group:all"
+  }
+  bindings {
+    role: "role/swarming.taskTriggerer"
+    principals: "group:project-libyuv-admins"
+  }
+}
+realms {
+  name: "ci"
+  bindings {
+    role: "role/buildbucket.builderServiceAccount"
+    principals: "user:[email protected]"
+  }
+  bindings {
+    role: "role/scheduler.triggerer"
+    principals: "user:[email protected]"
+    conditions {
+      restrict {
+        attribute: "scheduler.job.name"
+        values: "Android Tester ARM32 Debug (Nexus 5X)"
+        values: "Android Tester ARM32 Release (Nexus 5X)"
+        values: "Android Tester ARM64 Debug (Nexus 5X)"
+      }
+    }
+  }
+}
+realms {
+  name: "cron"
+  bindings {
+    role: "role/buildbucket.builderServiceAccount"
+    principals: "user:libyuv-ci-autoroll-builder@chops-service-accounts.iam.gserviceaccount.com"
+  }
+}
+realms {
+  name: "try"
+  bindings {
+    role: "role/buildbucket.builderServiceAccount"
+    principals: "user:libyuv-try-builder@chops-service-accounts.iam.gserviceaccount.com"
+  }
+  bindings {
+    role: "role/buildbucket.triggerer"
+    principals: "group:project-libyuv-tryjob-access"
+    principals: "group:service-account-cq"
+  }
+}

yuv.mod/libyuv/libyuv.gni

@@ -0,0 +1,34 @@
+# Copyright 2016 The LibYuv Project Authors. All rights reserved.
+#
+# Use of this source code is governed by a BSD-style license
+# that can be found in the LICENSE file in the root of the source
+# tree. An additional intellectual property rights grant can be found
+# in the file PATENTS. All contributing project authors may
+# be found in the AUTHORS file in the root of the source tree.
+
+import("//build/config/arm.gni")
+import("//build/config/loongarch64.gni")
+import("//build/config/mips.gni")
+import("//build_overrides/build.gni")
+
+declare_args() {
+  libyuv_include_tests = !build_with_chromium
+  libyuv_disable_jpeg = false
+  libyuv_disable_rvv = false
+  libyuv_use_neon =
+      current_cpu == "arm64" ||
+      (current_cpu == "arm" && (arm_use_neon || arm_optionally_use_neon))
+  libyuv_use_sve = current_cpu == "arm64"
+
+  # Restrict to is_linux to work around undefined symbol linker errors on
+  # Android, Fuchsia, macOS, and compilation errors on Windows.
+  # TODO: bug 359006069 - Remove the is_linux restriction after the linker and
+  # compilation errors are fixed.
+  libyuv_use_sme = current_cpu == "arm64" && is_linux
+  libyuv_use_msa =
+      (current_cpu == "mips64el" || current_cpu == "mipsel") && mips_use_msa
+  libyuv_use_mmi =
+      (current_cpu == "mips64el" || current_cpu == "mipsel") && mips_use_mmi
+  libyuv_use_lsx = current_cpu == "loong64" && loongarch64_use_lsx
+  libyuv_use_lasx = current_cpu == "loong64" && loongarch64_use_lasx
+}

yuv.mod/libyuv/libyuv.gyp

@@ -0,0 +1,162 @@
+# Copyright 2011 The LibYuv Project Authors. All rights reserved.
+#
+# Use of this source code is governed by a BSD-style license
+# that can be found in the LICENSE file in the root of the source
+# tree. An additional intellectual property rights grant can be found
+# in the file PATENTS. All contributing project authors may
+# be found in the AUTHORS file in the root of the source tree.
+
+{
+  'includes': [
+    'libyuv.gypi',
+  ],
+  # Make sure that if we are being compiled to an xcodeproj, nothing tries to
+  # include a .pch.
+  'xcode_settings': {
+    'GCC_PREFIX_HEADER': '',
+    'GCC_PRECOMPILE_PREFIX_HEADER': 'NO',
+  },
+  'variables': {
+    'use_system_libjpeg%': 0,
+    # Can be enabled if your jpeg has GYP support.
+    'libyuv_disable_jpeg%': 1,
+    # 'chromium_code' treats libyuv as internal and increases warning level.
+    'chromium_code': 1,
+    # clang compiler default variable usable by other apps that include libyuv.
+    'clang%': 0,
+    # Link-Time Optimizations.
+    'use_lto%': 0,
+    'mips_msa%': 0,  # Default to msa off.
+    'build_neon': 0,
+    'build_msa': 0,
+    'conditions': [
+       ['(target_arch == "armv7" or target_arch == "armv7s" or \
+       (target_arch == "arm" and arm_version >= 7) or target_arch == "arm64")\
+       and (arm_neon == 1 or arm_neon_optional == 1)', {
+         'build_neon': 1,
+       }],
+       ['(target_arch == "mipsel" or target_arch == "mips64el")\
+       and (mips_msa == 1)',
+       {
+         'build_msa': 1,
+       }],
+    ],
+  },
+
+  'targets': [
+    {
+      'target_name': 'libyuv',
+      # Change type to 'shared_library' to build .so or .dll files.
+      'type': 'static_library',
+      'variables': {
+        'optimize': 'max',  # enable O2 and ltcg.
+      },
+      # Allows libyuv.a redistributable library without external dependencies.
+      'standalone_static_library': 1,
+      'conditions': [
+       # Disable -Wunused-parameter
+        ['clang == 1', {
+          'cflags': [
+            '-Wno-unused-parameter',
+         ],
+        }],
+        ['build_neon != 0', {
+          'defines': [
+            'LIBYUV_NEON',
+          ],
+          'cflags!': [
+            '-mfpu=vfp',
+            '-mfpu=vfpv3',
+            '-mfpu=vfpv3-d16',
+            # '-mthumb',  # arm32 not thumb
+          ],
+          'conditions': [
+            # Disable LTO in libyuv_neon target due to gcc 4.9 compiler bug.
+            ['clang == 0 and use_lto == 1', {
+              'cflags!': [
+                '-flto',
+                '-ffat-lto-objects',
+              ],
+            }],
+            # arm64 does not need -mfpu=neon option as neon is not optional
+            ['target_arch != "arm64"', {
+              'cflags': [
+                '-mfpu=neon',
+                # '-marm',  # arm32 not thumb
+              ],
+            }],
+          ],
+        }],
+        ['build_msa != 0', {
+          'defines': [
+            'LIBYUV_MSA',
+          ],
+        }],
+        ['OS != "ios" and libyuv_disable_jpeg != 1', {
+          'defines': [
+            'HAVE_JPEG'
+          ],
+          'conditions': [
+            # Caveat system jpeg support may not support motion jpeg
+            [ 'use_system_libjpeg == 1', {
+              'dependencies': [
+                 '<(DEPTH)/third_party/libjpeg/libjpeg.gyp:libjpeg',
+              ],
+            }, {
+              'dependencies': [
+                 '<(DEPTH)/third_party/libjpeg_turbo/libjpeg.gyp:libjpeg',
+              ],
+            }],
+            [ 'use_system_libjpeg == 1', {
+              'link_settings': {
+                'libraries': [
+                  '-ljpeg',
+                ],
+              }
+            }],
+          ],
+        }],
+      ], #conditions
+      'defines': [
+        # Enable the following 3 macros to turn off assembly for specified CPU.
+        # 'LIBYUV_DISABLE_X86',
+        # 'LIBYUV_DISABLE_NEON',
+        # 'LIBYUV_DISABLE_DSPR2',
+        # Enable the following macro to build libyuv as a shared library (dll).
+        # 'LIBYUV_USING_SHARED_LIBRARY',
+        # TODO(fbarchard): Make these into gyp defines.
+      ],
+      'include_dirs': [
+        'include',
+        '.',
+      ],
+      'direct_dependent_settings': {
+        'include_dirs': [
+          'include',
+          '.',
+        ],
+        'conditions': [
+          ['OS == "android" and target_arch == "arm64"', {
+            'ldflags': [
+              '-Wl,--dynamic-linker,/system/bin/linker64',
+            ],
+          }],
+          ['OS == "android" and target_arch != "arm64"', {
+            'ldflags': [
+              '-Wl,--dynamic-linker,/system/bin/linker',
+            ],
+          }],
+        ], #conditions
+      },
+      'sources': [
+        '<@(libyuv_sources)',
+      ],
+    },
+  ], # targets.
+}
+
+# Local Variables:
+# tab-width:2
+# indent-tabs-mode:nil
+# End:
+# vim: set expandtab tabstop=2 shiftwidth=2:

yuv.mod/libyuv/libyuv.gypi

@@ -0,0 +1,87 @@
+# Copyright 2014 The LibYuv Project Authors. All rights reserved.
+#
+# Use of this source code is governed by a BSD-style license
+# that can be found in the LICENSE file in the root of the source
+# tree. An additional intellectual property rights grant can be found
+# in the file PATENTS. All contributing project authors may
+# be found in the AUTHORS file in the root of the source tree.
+
+{
+  'variables': {
+    'libyuv_sources': [
+      # includes.
+      'include/libyuv.h',
+      'include/libyuv/basic_types.h',
+      'include/libyuv/compare.h',
+      'include/libyuv/compare_row.h',
+      'include/libyuv/convert.h',
+      'include/libyuv/convert_argb.h',
+      'include/libyuv/convert_from.h',
+      'include/libyuv/convert_from_argb.h',
+      'include/libyuv/cpu_id.h',
+      'include/libyuv/loongson_intrinsics.h',
+      'include/libyuv/macros_msa.h',
+      'include/libyuv/mjpeg_decoder.h',
+      'include/libyuv/planar_functions.h',
+      'include/libyuv/rotate.h',
+      'include/libyuv/rotate_argb.h',
+      'include/libyuv/rotate_row.h',
+      'include/libyuv/row.h',
+      'include/libyuv/scale.h',
+      'include/libyuv/scale_argb.h',
+      'include/libyuv/scale_rgb.h',
+      'include/libyuv/scale_row.h',
+      'include/libyuv/scale_uv.h',
+      'include/libyuv/version.h',
+      'include/libyuv/video_common.h',
+
+      # sources.
+      'source/compare.cc',
+      'source/compare_common.cc',
+      'source/compare_gcc.cc',
+      'source/compare_msa.cc',
+      'source/compare_neon.cc',
+      'source/compare_neon64.cc',
+      'source/compare_win.cc',
+      'source/convert.cc',
+      'source/convert_argb.cc',
+      'source/convert_from.cc',
+      'source/convert_from_argb.cc',
+      'source/convert_jpeg.cc',
+      'source/convert_to_argb.cc',
+      'source/convert_to_i420.cc',
+      'source/cpu_id.cc',
+      'source/mjpeg_decoder.cc',
+      'source/mjpeg_validate.cc',
+      'source/planar_functions.cc',
+      'source/rotate.cc',
+      'source/rotate_any.cc',
+      'source/rotate_argb.cc',
+      'source/rotate_common.cc',
+      'source/rotate_gcc.cc',
+      'source/rotate_msa.cc',
+      'source/rotate_neon.cc',
+      'source/rotate_neon64.cc',
+      'source/rotate_win.cc',
+      'source/row_any.cc',
+      'source/row_common.cc',
+      'source/row_gcc.cc',
+      'source/row_msa.cc',
+      'source/row_neon.cc',
+      'source/row_neon64.cc',
+      'source/row_win.cc',
+      'source/scale.cc',
+      'source/scale_any.cc',
+      'source/scale_argb.cc',
+      'source/scale_common.cc',
+      'source/scale_gcc.cc',
+      'source/scale_msa.cc',
+      'source/scale_neon.cc',
+      'source/scale_neon64.cc',
+      'source/scale_rgb.cc',
+      'source/scale_uv.cc',
+      'source/scale_win.cc',
+      'source/video_common.cc',
+    ],
+  }
+}

yuv.mod/libyuv/linux.mk

@@ -0,0 +1,100 @@
+# This is a generic makefile for libyuv for gcc.
+# make -f linux.mk CXX=clang++
+
+CC?=gcc
+CFLAGS?=-O2 -fomit-frame-pointer
+CFLAGS+=-Iinclude/
+
+CXX?=g++
+CXXFLAGS?=-O2 -fomit-frame-pointer
+CXXFLAGS+=-Iinclude/
+
+LOCAL_OBJ_FILES := \
+	source/compare.o           \
+	source/compare_common.o    \
+	source/compare_gcc.o       \
+	source/compare_msa.o       \
+	source/compare_neon.o      \
+	source/compare_neon64.o    \
+	source/compare_win.o       \
+	source/convert.o           \
+	source/convert_argb.o      \
+	source/convert_from.o      \
+	source/convert_from_argb.o \
+	source/convert_jpeg.o      \
+	source/convert_to_argb.o   \
+	source/convert_to_i420.o   \
+	source/cpu_id.o            \
+	source/mjpeg_decoder.o     \
+	source/mjpeg_validate.o    \
+	source/planar_functions.o  \
+	source/rotate.o            \
+	source/rotate_any.o        \
+	source/rotate_argb.o       \
+	source/rotate_common.o     \
+	source/rotate_gcc.o        \
+	source/rotate_lsx.o        \
+	source/rotate_msa.o        \
+	source/rotate_neon.o       \
+	source/rotate_neon64.o     \
+	source/rotate_win.o        \
+	source/row_any.o           \
+	source/row_common.o        \
+	source/row_gcc.o           \
+	source/row_lasx.o          \
+	source/row_lsx.o           \
+	source/row_msa.o           \
+	source/row_neon.o          \
+	source/row_neon64.o        \
+	source/row_rvv.o           \
+	source/row_win.o           \
+	source/scale.o             \
+	source/scale_any.o         \
+	source/scale_argb.o        \
+	source/scale_common.o      \
+	source/scale_gcc.o         \
+	source/scale_lsx.o         \
+	source/scale_msa.o         \
+	source/scale_neon.o        \
+	source/scale_neon64.o      \
+	source/scale_rgb.o         \
+	source/scale_rvv.o         \
+	source/scale_uv.o          \
+	source/scale_win.o         \
+	source/video_common.o
+
+.cc.o:
+	$(CXX) -c $(CXXFLAGS) $*.cc -o $*.o
+
+.c.o:
+	$(CC) -c $(CFLAGS) $*.c -o $*.o
+
+all: libyuv.a i444tonv12_eg yuvconvert yuvconstants cpuid psnr
+
+libyuv.a: $(LOCAL_OBJ_FILES)
+	$(AR) $(ARFLAGS) $@ $(LOCAL_OBJ_FILES)
+
+# A C++ test utility that uses libyuv conversion.
+yuvconvert: util/yuvconvert.cc libyuv.a
+	$(CXX) $(CXXFLAGS) -Iutil/ -o $@ util/yuvconvert.cc libyuv.a
+
+# A C test utility that generates yuvconstants for yuv to rgb.
+yuvconstants: util/yuvconstants.c libyuv.a
+	$(CXX) $(CXXFLAGS) -Iutil/ -lm -o $@ util/yuvconstants.c libyuv.a
+
+# A standalone test utility
+psnr: util/psnr.cc
+	$(CXX) $(CXXFLAGS) -Iutil/ -o $@ util/psnr.cc util/psnr_main.cc util/ssim.cc
+
+# A simple conversion example.
+i444tonv12_eg: util/i444tonv12_eg.cc libyuv.a
+	$(CXX) $(CXXFLAGS) -o $@ util/i444tonv12_eg.cc libyuv.a
+
+# A C test utility that uses libyuv conversion from C.
+# gcc 4.4 and older require -fno-exceptions to avoid link error on __gxx_personality_v0
+# CC=gcc-4.4 CXXFLAGS=-fno-exceptions CXX=g++-4.4 make -f linux.mk
+cpuid: util/cpuid.c libyuv.a
+	$(CC) $(CFLAGS) -o $@ util/cpuid.c libyuv.a
+
+clean:
+	/bin/rm -f source/*.o *.ii *.s libyuv.a i444tonv12_eg yuvconvert yuvconstants cpuid psnr

yuv.mod/libyuv/public.mk

@@ -0,0 +1,13 @@
+# This file contains all the common make variables which are useful for
+# anyone depending on this library.
+# Note that dependencies on NDK are not directly listed since NDK auto adds
+# them.
+
+LIBYUV_INCLUDES := $(LIBYUV_PATH)/include
+
+LIBYUV_C_FLAGS :=
+
+LIBYUV_CPP_FLAGS :=
+
+LIBYUV_LDLIBS :=
+LIBYUV_DEP_MODULES :=

yuv.mod/libyuv/pylintrc

@@ -0,0 +1,17 @@
+[MESSAGES CONTROL]
+
+# Disable the message, report, category or checker with the given id(s).
+# TODO(kjellander): Reduce this list to as small as possible.
+disable=I0010,I0011,bad-continuation,broad-except,duplicate-code,eval-used,exec-used,fixme,invalid-name,missing-docstring,no-init,no-member,too-few-public-methods,too-many-ancestors,too-many-arguments,too-many-branches,too-many-function-args,too-many-instance-attributes,too-many-lines,too-many-locals,too-many-public-methods,too-many-return-statements,too-many-statements
+
+
+[REPORTS]
+
+# Don't write out full reports, just messages.
+reports=no
+
+
+[FORMAT]
+
+# We use two spaces for indents, instead of the usual four spaces or tab.
+indent-string='  '

yuv.mod/libyuv/riscv_script/prepare_toolchain_qemu.sh

@@ -0,0 +1,74 @@
+#!/bin/bash
+set -ev
+
+# Download & build RISC-V Clang toolchain & QEMU emulator. 
+# RISC-V Clang is for cross compile with  the RISC-V Vector ISA.
+# RISC-V QEMU is used to run the test suite.
+#
+# Requirements: Linux host w/ working C++ compiler, git, cmake, ninja, wget, tar
+
+# NOTE: this script must be run from the top-level directory of the LIBYUV_SRC_DIR.
+
+RISCV_TRIPLE="riscv64-unknown-linux-gnu"
+RISCV_QEMU="qemu-riscv64"
+
+LIBYUV_SRC_DIR=$(pwd)
+BUILD_DIR="$LIBYUV_SRC_DIR"/build-toolchain-qemu
+INSTALL_QEMU="$BUILD_DIR"/riscv-qemu
+INSTALL_CLANG="$BUILD_DIR"/riscv-clang
+
+LLVM_VERSION="16.0.0"
+LLVM_NAME=llvm-project-"$LLVM_VERSION".src
+
+RISCV_GNU_TOOLCHAIN="$BUILD_DIR"/riscv-gnu-toolchain
+RISCV_CLANG_TOOLCHAIN="$BUILD_DIR"/"$LLVM_NAME"
+
+QEMU_NAME="qemu-7.0.0"
+
+mkdir -p "$BUILD_DIR"
+cd "$BUILD_DIR"
+
+# Download and install RISC-V GNU Toolchain (needed to build Clang)
+if [ ! -d "$RISCV_GNU_TOOLCHAIN" ]
+then
+  git clone [email protected]:riscv/riscv-gnu-toolchain.git
+  pushd "$RISCV_GNU_TOOLCHAIN"
+  git submodule update --init --recursive
+  ./configure --with-cmodel=medany --prefix="$INSTALL_CLANG"
+  ionice nice make linux -j `nproc` install
+  popd
+fi
+
+# Download Clang toolchain & build cross compiler
+if [ ! -d "$RISCV_CLANG_TOOLCHAIN" ]
+then
+  wget https://github.com/llvm/llvm-project/releases/download/llvmorg-"$LLVM_VERSION"/"$LLVM_NAME".tar.xz
+  tar xvJf "$LLVM_NAME".tar.xz
+  pushd "$RISCV_CLANG_TOOLCHAIN"
+	cmake -DCMAKE_INSTALL_PREFIX="$INSTALL_CLANG" \
+      -DCMAKE_BUILD_TYPE=Release \
+      -DLLVM_TARGETS_TO_BUILD="RISCV" \
+      -DLLVM_ENABLE_PROJECTS="clang" \
+      -DLLVM_DEFAULT_TARGET_TRIPLE="$RISCV_TRIPLE" \
+      -DLLVM_INSTALL_TOOLCHAIN_ONLY=On \
+      -DDEFAULT_SYSROOT=../sysroot \
+      -G "Ninja" "$RISCV_CLANG_TOOLCHAIN"/llvm
+	ionice nice ninja -j `nproc`
+	ionice nice ninja -j `nproc` install
+  popd
+  pushd "$INSTALL_CLANG"/bin
+  ln -sf clang "$RISCV_TRIPLE"-clang
+  ln -sf clang++ "$RISCV_TRIPLE"-clang++
+  popd
+fi
+
+# Download QEMU and build the riscv64 Linux usermode emulator
+if [ ! -d "$QEMU_NAME" ]
+then
+  wget https://download.qemu.org/"$QEMU_NAME".tar.xz
+  tar xvJf "$QEMU_NAME".tar.xz
+  pushd "$QEMU_NAME"
+  ./configure --target-list=riscv64-linux-user --prefix="$INSTALL_QEMU"
+  ionice nice make -j `nproc` install
+  popd
+fi

yuv.mod/libyuv/riscv_script/riscv-clang.cmake

@@ -0,0 +1,56 @@
+set(CMAKE_CROSSCOMPILING TRUE)
+set(CMAKE_SYSTEM_NAME "Linux")
+set(CMAKE_SYSTEM_PROCESSOR "riscv64")
+
+option(USE_RVV "Enable riscv vector or not." ON)
+option(USE_AUTO_VECTORIZER "Enable riscv auto vectorizer or not." OFF)
+
+# Avoid to use system path for cross-compile
+set(CMAKE_FIND_USE_CMAKE_SYSTEM_PATH FALSE)
+
+set(TOOLCHAIN_PATH "" CACHE STRING "The toolcahin path.")
+if(NOT TOOLCHAIN_PATH)
+  set(TOOLCHAIN_PATH ${CMAKE_SOURCE_DIR}/build-toolchain-qemu/riscv-clang)
+endif()
+
+set(TOOLCHAIN_PREFIX "riscv64-unknown-linux-gnu-" CACHE STRING "The toolcahin prefix.")
+
+# toolchain setting
+set(CMAKE_C_COMPILER "${TOOLCHAIN_PATH}/bin/${TOOLCHAIN_PREFIX}clang")
+set(CMAKE_CXX_COMPILER "${TOOLCHAIN_PATH}/bin/${TOOLCHAIN_PREFIX}clang++")
+
+# CMake will just use the host-side tools for the following tools, so we setup them here.
+set(CMAKE_C_COMPILER_AR "${TOOLCHAIN_PATH}/bin/llvm-ar")
+set(CMAKE_CXX_COMPILER_AR "${TOOLCHAIN_PATH}/bin/llvm-ar")
+set(CMAKE_C_COMPILER_RANLIB "${TOOLCHAIN_PATH}/bin/llvm-ranlib")
+set(CMAKE_CXX_COMPILER_RANLIB "${TOOLCHAIN_PATH}/bin/llvm-ranlib")
+set(CMAKE_OBJDUMP "${TOOLCHAIN_PATH}/bin/llvm-objdump")
+set(CMAKE_OBJCOPY "${TOOLCHAIN_PATH}/bin/llvm-objcopy")
+
+# compile options
+set(RISCV_COMPILER_FLAGS "" CACHE STRING "Compile flags")
+# if user provides RISCV_COMPILER_FLAGS, appeding compile flags is avoided.
+if(RISCV_COMPILER_FLAGS STREQUAL "")
+  message(STATUS "USE_RVV: ${USE_RVV}")
+  message(STATUS "USE_AUTO_VECTORIZER: ${USE_AUTO_VECTORIZER}")
+  if(USE_RVV)
+    list(APPEND RISCV_COMPILER_FLAGS "-march=rv64gcv")
+    if(NOT USE_AUTO_VECTORIZER)
+      # Disable auto-vectorizer
+      add_compile_options(-fno-vectorize -fno-slp-vectorize)
+    endif()
+  else()
+    list(APPEND RISCV_COMPILER_FLAGS "-march=rv64gc")
+  endif()
+endif()
+add_compile_options("-Wuninitialized")
+message(STATUS "RISCV_COMPILER_FLAGS: ${RISCV_COMPILER_FLAGS}")
+
+set(CMAKE_C_FLAGS             "${RISCV_COMPILER_FLAGS} ${CMAKE_C_FLAGS}")
+set(CMAKE_CXX_FLAGS           "${RISCV_COMPILER_FLAGS} ${CMAKE_CXX_FLAGS}")
+
+set(RISCV_LINKER_FLAGS "-lstdc++ -lpthread -lm -ldl")
+set(RISCV_LINKER_FLAGS_EXE)
+set(CMAKE_SHARED_LINKER_FLAGS "${RISCV_LINKER_FLAGS} ${CMAKE_SHARED_LINKER_FLAGS}")
+set(CMAKE_MODULE_LINKER_FLAGS "${RISCV_LINKER_FLAGS} ${CMAKE_MODULE_LINKER_FLAGS}")
+set(CMAKE_EXE_LINKER_FLAGS    "${RISCV_LINKER_FLAGS} ${RISCV_LINKER_FLAGS_EXE} ${CMAKE_EXE_LINKER_FLAGS}")

yuv.mod/libyuv/riscv_script/run_qemu.sh

@@ -0,0 +1,15 @@
+#!/bin/bash
+set -x
+set -e
+
+USE_RVV="${USE_RVV:-OFF}"
+TOOLCHAIN_PATH="${TOOLCHAIN_PATH:-../../build-toolchain-qemu/riscv-clang}"
+QEMU_PREFIX_PATH="${QEMU_PREFIX_PATH:-../../build-toolchain-qemu/riscv-qemu/}"
+
+if [ "${USE_RVV}" = "ON" ];then
+  QEMU_OPTION="-cpu rv64,zba=true,zbb=true,zbc=true,zbs=true,v=true,vlen=512,elen=64,vext_spec=v1.0 -L ${TOOLCHAIN_PATH}/sysroot"
+else
+  QEMU_OPTION="-cpu rv64,zba=true,zbb=true,zbc=true,zbs=true -L ${TOOLCHAIN_PATH}/sysroot"
+fi
+
+$QEMU_PREFIX_PATH/bin/qemu-riscv64 $QEMU_OPTION $@

yuv.mod/libyuv/source/compare.cc

@@ -0,0 +1,445 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/compare.h"
+
+#include <float.h>
+#include <math.h>
+#ifdef _OPENMP
+#include <omp.h>
+#endif
+
+#include "libyuv/basic_types.h"
+#include "libyuv/compare_row.h"
+#include "libyuv/cpu_id.h"
+#include "libyuv/row.h"
+#include "libyuv/video_common.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// hash seed of 5381 recommended.
+LIBYUV_API
+uint32_t HashDjb2(const uint8_t* src, uint64_t count, uint32_t seed) {
+  const int kBlockSize = 1 << 15;  // 32768;
+  int remainder;
+  uint32_t (*HashDjb2_SSE)(const uint8_t* src, int count, uint32_t seed) =
+      HashDjb2_C;
+#if defined(HAS_HASHDJB2_SSE41)
+  if (TestCpuFlag(kCpuHasSSE41)) {
+    HashDjb2_SSE = HashDjb2_SSE41;
+  }
+#endif
+#if defined(HAS_HASHDJB2_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    HashDjb2_SSE = HashDjb2_AVX2;
+  }
+#endif
+#if defined(HAS_HASHDJB2_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    HashDjb2_SSE = HashDjb2_NEON;
+  }
+#endif
+
+  while (count >= (uint64_t)kBlockSize) {
+    seed = HashDjb2_SSE(src, kBlockSize, seed);
+    src += kBlockSize;
+    count -= kBlockSize;
+  }
+  remainder = (int)count & ~15;
+  if (remainder) {
+    seed = HashDjb2_SSE(src, remainder, seed);
+    src += remainder;
+    count -= remainder;
+  }
+  remainder = (int)count & 15;
+  if (remainder) {
+    seed = HashDjb2_C(src, remainder, seed);
+  }
+  return seed;
+}
+
+static uint32_t ARGBDetectRow_C(const uint8_t* argb, int width) {
+  int x;
+  for (x = 0; x < width - 1; x += 2) {
+    if (argb[0] != 255) {  // First byte is not Alpha of 255, so not ARGB.
+      return FOURCC_BGRA;
+    }
+    if (argb[3] != 255) {  // Fourth byte is not Alpha of 255, so not BGRA.
+      return FOURCC_ARGB;
+    }
+    if (argb[4] != 255) {  // Second pixel first byte is not Alpha of 255.
+      return FOURCC_BGRA;
+    }
+    if (argb[7] != 255) {  // Second pixel fourth byte is not Alpha of 255.
+      return FOURCC_ARGB;
+    }
+    argb += 8;
+  }
+  if (width & 1) {
+    if (argb[0] != 255) {  // First byte is not Alpha of 255, so not ARGB.
+      return FOURCC_BGRA;
+    }
+    if (argb[3] != 255) {  // 4th byte is not Alpha of 255, so not BGRA.
+      return FOURCC_ARGB;
+    }
+  }
+  return 0;
+}
+
+// Scan an opaque argb image and return fourcc based on alpha offset.
+// Returns FOURCC_ARGB, FOURCC_BGRA, or 0 if unknown.
+LIBYUV_API
+uint32_t ARGBDetect(const uint8_t* argb,
+                    int stride_argb,
+                    int width,
+                    int height) {
+  uint32_t fourcc = 0;
+  int h;
+
+  // Coalesce rows.
+  if (stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    stride_argb = 0;
+  }
+  for (h = 0; h < height && fourcc == 0; ++h) {
+    fourcc = ARGBDetectRow_C(argb, width);
+    argb += stride_argb;
+  }
+  return fourcc;
+}
+
+// NEON version accumulates in 16 bit shorts which overflow at 65536 bytes.
+// So actual maximum is 1 less loop, which is 64436 - 32 bytes.
+
+LIBYUV_API
+uint64_t ComputeHammingDistance(const uint8_t* src_a,
+                                const uint8_t* src_b,
+                                int count) {
+  const int kBlockSize = 1 << 15;  // 32768;
+  const int kSimdSize = 64;
+  // SIMD for multiple of 64, and C for remainder
+  int remainder = count & (kBlockSize - 1) & ~(kSimdSize - 1);
+  uint64_t diff = 0;
+  int i;
+  uint32_t (*HammingDistance)(const uint8_t* src_a, const uint8_t* src_b,
+                              int count) = HammingDistance_C;
+#if defined(HAS_HAMMINGDISTANCE_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    HammingDistance = HammingDistance_NEON;
+  }
+#endif
+#if defined(HAS_HAMMINGDISTANCE_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    HammingDistance = HammingDistance_NEON_DotProd;
+  }
+#endif
+#if defined(HAS_HAMMINGDISTANCE_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    HammingDistance = HammingDistance_SSSE3;
+  }
+#endif
+#if defined(HAS_HAMMINGDISTANCE_SSE42)
+  if (TestCpuFlag(kCpuHasSSE42)) {
+    HammingDistance = HammingDistance_SSE42;
+  }
+#endif
+#if defined(HAS_HAMMINGDISTANCE_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    HammingDistance = HammingDistance_AVX2;
+  }
+#endif
+#if defined(HAS_HAMMINGDISTANCE_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    HammingDistance = HammingDistance_MSA;
+  }
+#endif
+
+#ifdef _OPENMP
+#pragma omp parallel for reduction(+ : diff)
+#endif
+  for (i = 0; i < (count - (kBlockSize - 1)); i += kBlockSize) {
+    diff += HammingDistance(src_a + i, src_b + i, kBlockSize);
+  }
+  src_a += count & ~(kBlockSize - 1);
+  src_b += count & ~(kBlockSize - 1);
+  if (remainder) {
+    diff += HammingDistance(src_a, src_b, remainder);
+    src_a += remainder;
+    src_b += remainder;
+  }
+  remainder = count & (kSimdSize - 1);
+  if (remainder) {
+    diff += HammingDistance_C(src_a, src_b, remainder);
+  }
+  return diff;
+}
+
+// TODO(fbarchard): Refactor into row function.
+LIBYUV_API
+uint64_t ComputeSumSquareError(const uint8_t* src_a,
+                               const uint8_t* src_b,
+                               int count) {
+  // SumSquareError returns values 0 to 65535 for each squared difference.
+  // Up to 65536 of those can be summed and remain within a uint32_t.
+  // After each block of 65536 pixels, accumulate into a uint64_t.
+  const int kBlockSize = 65536;
+  int remainder = count & (kBlockSize - 1) & ~31;
+  uint64_t sse = 0;
+  int i;
+  uint32_t (*SumSquareError)(const uint8_t* src_a, const uint8_t* src_b,
+                             int count) = SumSquareError_C;
+#if defined(HAS_SUMSQUAREERROR_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SumSquareError = SumSquareError_NEON;
+  }
+#endif
+#if defined(HAS_SUMSQUAREERROR_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    SumSquareError = SumSquareError_NEON_DotProd;
+  }
+#endif
+#if defined(HAS_SUMSQUAREERROR_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    // Note only used for multiples of 16 so count is not checked.
+    SumSquareError = SumSquareError_SSE2;
+  }
+#endif
+#if defined(HAS_SUMSQUAREERROR_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    // Note only used for multiples of 32 so count is not checked.
+    SumSquareError = SumSquareError_AVX2;
+  }
+#endif
+#if defined(HAS_SUMSQUAREERROR_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    SumSquareError = SumSquareError_MSA;
+  }
+#endif
+#ifdef _OPENMP
+#pragma omp parallel for reduction(+ : sse)
+#endif
+  for (i = 0; i < (count - (kBlockSize - 1)); i += kBlockSize) {
+    sse += SumSquareError(src_a + i, src_b + i, kBlockSize);
+  }
+  src_a += count & ~(kBlockSize - 1);
+  src_b += count & ~(kBlockSize - 1);
+  if (remainder) {
+    sse += SumSquareError(src_a, src_b, remainder);
+    src_a += remainder;
+    src_b += remainder;
+  }
+  remainder = count & 31;
+  if (remainder) {
+    sse += SumSquareError_C(src_a, src_b, remainder);
+  }
+  return sse;
+}
+
+LIBYUV_API
+uint64_t ComputeSumSquareErrorPlane(const uint8_t* src_a,
+                                    int stride_a,
+                                    const uint8_t* src_b,
+                                    int stride_b,
+                                    int width,
+                                    int height) {
+  uint64_t sse = 0;
+  int h;
+  // Coalesce rows.
+  if (stride_a == width && stride_b == width) {
+    width *= height;
+    height = 1;
+    stride_a = stride_b = 0;
+  }
+  for (h = 0; h < height; ++h) {
+    sse += ComputeSumSquareError(src_a, src_b, width);
+    src_a += stride_a;
+    src_b += stride_b;
+  }
+  return sse;
+}
+
+LIBYUV_API
+double SumSquareErrorToPsnr(uint64_t sse, uint64_t count) {
+  double psnr;
+  if (sse > 0) {
+    double mse = (double)count / (double)sse;
+    psnr = 10.0 * log10(255.0 * 255.0 * mse);
+  } else {
+    psnr = kMaxPsnr;  // Limit to prevent divide by 0
+  }
+
+  if (psnr > kMaxPsnr) {
+    psnr = kMaxPsnr;
+  }
+
+  return psnr;
+}
+
+LIBYUV_API
+double CalcFramePsnr(const uint8_t* src_a,
+                     int stride_a,
+                     const uint8_t* src_b,
+                     int stride_b,
+                     int width,
+                     int height) {
+  const uint64_t samples = (uint64_t)width * (uint64_t)height;
+  const uint64_t sse = ComputeSumSquareErrorPlane(src_a, stride_a, src_b,
+                                                  stride_b, width, height);
+  return SumSquareErrorToPsnr(sse, samples);
+}
+
+LIBYUV_API
+double I420Psnr(const uint8_t* src_y_a,
+                int stride_y_a,
+                const uint8_t* src_u_a,
+                int stride_u_a,
+                const uint8_t* src_v_a,
+                int stride_v_a,
+                const uint8_t* src_y_b,
+                int stride_y_b,
+                const uint8_t* src_u_b,
+                int stride_u_b,
+                const uint8_t* src_v_b,
+                int stride_v_b,
+                int width,
+                int height) {
+  const uint64_t sse_y = ComputeSumSquareErrorPlane(
+      src_y_a, stride_y_a, src_y_b, stride_y_b, width, height);
+  const int width_uv = (width + 1) >> 1;
+  const int height_uv = (height + 1) >> 1;
+  const uint64_t sse_u = ComputeSumSquareErrorPlane(
+      src_u_a, stride_u_a, src_u_b, stride_u_b, width_uv, height_uv);
+  const uint64_t sse_v = ComputeSumSquareErrorPlane(
+      src_v_a, stride_v_a, src_v_b, stride_v_b, width_uv, height_uv);
+  const uint64_t samples = (uint64_t)width * (uint64_t)height +
+                           2 * ((uint64_t)width_uv * (uint64_t)height_uv);
+  const uint64_t sse = sse_y + sse_u + sse_v;
+  return SumSquareErrorToPsnr(sse, samples);
+}
+
+static const int64_t cc1 = 26634;   // (64^2*(.01*255)^2
+static const int64_t cc2 = 239708;  // (64^2*(.03*255)^2
+
+static double Ssim8x8_C(const uint8_t* src_a,
+                        int stride_a,
+                        const uint8_t* src_b,
+                        int stride_b) {
+  int64_t sum_a = 0;
+  int64_t sum_b = 0;
+  int64_t sum_sq_a = 0;
+  int64_t sum_sq_b = 0;
+  int64_t sum_axb = 0;
+
+  int i;
+  for (i = 0; i < 8; ++i) {
+    int j;
+    for (j = 0; j < 8; ++j) {
+      sum_a += src_a[j];
+      sum_b += src_b[j];
+      sum_sq_a += src_a[j] * src_a[j];
+      sum_sq_b += src_b[j] * src_b[j];
+      sum_axb += src_a[j] * src_b[j];
+    }
+
+    src_a += stride_a;
+    src_b += stride_b;
+  }
+
+  {
+    const int64_t count = 64;
+    // scale the constants by number of pixels
+    const int64_t c1 = (cc1 * count * count) >> 12;
+    const int64_t c2 = (cc2 * count * count) >> 12;
+
+    const int64_t sum_a_x_sum_b = sum_a * sum_b;
+
+    const int64_t ssim_n = (2 * sum_a_x_sum_b + c1) *
+                           (2 * count * sum_axb - 2 * sum_a_x_sum_b + c2);
+
+    const int64_t sum_a_sq = sum_a * sum_a;
+    const int64_t sum_b_sq = sum_b * sum_b;
+
+    const int64_t ssim_d =
+        (sum_a_sq + sum_b_sq + c1) *
+        (count * sum_sq_a - sum_a_sq + count * sum_sq_b - sum_b_sq + c2);
+
+    if (ssim_d == 0) {
+      return DBL_MAX;
+    }
+    return (double)ssim_n / (double)ssim_d;
+  }
+}
+
+// We are using a 8x8 moving window with starting location of each 8x8 window
+// on the 4x4 pixel grid. Such arrangement allows the windows to overlap
+// block boundaries to penalize blocking artifacts.
+LIBYUV_API
+double CalcFrameSsim(const uint8_t* src_a,
+                     int stride_a,
+                     const uint8_t* src_b,
+                     int stride_b,
+                     int width,
+                     int height) {
+  int samples = 0;
+  double ssim_total = 0;
+  double (*Ssim8x8)(const uint8_t* src_a, int stride_a, const uint8_t* src_b,
+                    int stride_b) = Ssim8x8_C;
+
+  // sample point start with each 4x4 location
+  int i;
+  for (i = 0; i < height - 8; i += 4) {
+    int j;
+    for (j = 0; j < width - 8; j += 4) {
+      ssim_total += Ssim8x8(src_a + j, stride_a, src_b + j, stride_b);
+      samples++;
+    }
+
+    src_a += stride_a * 4;
+    src_b += stride_b * 4;
+  }
+
+  ssim_total /= samples;
+  return ssim_total;
+}
+
+LIBYUV_API
+double I420Ssim(const uint8_t* src_y_a,
+                int stride_y_a,
+                const uint8_t* src_u_a,
+                int stride_u_a,
+                const uint8_t* src_v_a,
+                int stride_v_a,
+                const uint8_t* src_y_b,
+                int stride_y_b,
+                const uint8_t* src_u_b,
+                int stride_u_b,
+                const uint8_t* src_v_b,
+                int stride_v_b,
+                int width,
+                int height) {
+  const double ssim_y =
+      CalcFrameSsim(src_y_a, stride_y_a, src_y_b, stride_y_b, width, height);
+  const int width_uv = (width + 1) >> 1;
+  const int height_uv = (height + 1) >> 1;
+  const double ssim_u = CalcFrameSsim(src_u_a, stride_u_a, src_u_b, stride_u_b,
+                                      width_uv, height_uv);
+  const double ssim_v = CalcFrameSsim(src_v_a, stride_v_a, src_v_b, stride_v_b,
+                                      width_uv, height_uv);
+  return ssim_y * 0.8 + 0.1 * (ssim_u + ssim_v);
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/compare_common.cc

@@ -0,0 +1,74 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/basic_types.h"
+
+#include "libyuv/compare_row.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Hakmem method for hamming distance.
+uint32_t HammingDistance_C(const uint8_t* src_a,
+                           const uint8_t* src_b,
+                           int count) {
+  uint32_t diff = 0u;
+
+  int i;
+  for (i = 0; i < count - 3; i += 4) {
+    uint32_t x = *((const uint32_t*)src_a) ^ *((const uint32_t*)src_b);
+    uint32_t u = x - ((x >> 1) & 0x55555555);
+    u = ((u >> 2) & 0x33333333) + (u & 0x33333333);
+    diff += ((((u + (u >> 4)) & 0x0f0f0f0f) * 0x01010101) >> 24);
+    src_a += 4;
+    src_b += 4;
+  }
+
+  for (; i < count; ++i) {
+    uint32_t x = *src_a ^ *src_b;
+    uint32_t u = x - ((x >> 1) & 0x55);
+    u = ((u >> 2) & 0x33) + (u & 0x33);
+    diff += (u + (u >> 4)) & 0x0f;
+    src_a += 1;
+    src_b += 1;
+  }
+
+  return diff;
+}
+
+uint32_t SumSquareError_C(const uint8_t* src_a,
+                          const uint8_t* src_b,
+                          int count) {
+  uint32_t sse = 0u;
+  int i;
+  for (i = 0; i < count; ++i) {
+    int diff = src_a[i] - src_b[i];
+    sse += (uint32_t)(diff * diff);
+  }
+  return sse;
+}
+
+// hash seed of 5381 recommended.
+// Internal C version of HashDjb2 with int sized count for efficiency.
+uint32_t HashDjb2_C(const uint8_t* src, int count, uint32_t seed) {
+  uint32_t hash = seed;
+  int i;
+  for (i = 0; i < count; ++i) {
+    hash += (hash << 5) + src[i];
+  }
+  return hash;
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/compare_gcc.cc

@@ -0,0 +1,360 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/basic_types.h"
+
+#include "libyuv/compare_row.h"
+#include "libyuv/row.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// This module is for GCC x86 and x64.
+#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__))
+
+// "memory" clobber prevents the reads from being removed
+
+#if defined(__x86_64__)
+uint32_t HammingDistance_SSE42(const uint8_t* src_a,
+                               const uint8_t* src_b,
+                               int count) {
+  uint64_t diff;
+
+      asm volatile (
+      "xor         %3,%3                         \n"
+      "xor         %%r8,%%r8                     \n"
+      "xor         %%r9,%%r9                     \n"
+      "xor         %%r10,%%r10                   \n"
+
+      // Process 32 bytes per loop.
+      LABELALIGN
+      "1:                                        \n"
+      "mov         (%0),%%rcx                    \n"
+      "mov         0x8(%0),%%rdx                 \n"
+      "xor         (%1),%%rcx                    \n"
+      "xor         0x8(%1),%%rdx                 \n"
+      "popcnt      %%rcx,%%rcx                   \n"
+      "popcnt      %%rdx,%%rdx                   \n"
+      "mov         0x10(%0),%%rsi                \n"
+      "mov         0x18(%0),%%rdi                \n"
+      "xor         0x10(%1),%%rsi                \n"
+      "xor         0x18(%1),%%rdi                \n"
+      "popcnt      %%rsi,%%rsi                   \n"
+      "popcnt      %%rdi,%%rdi                   \n"
+      "add         $0x20,%0                      \n"
+      "add         $0x20,%1                      \n"
+      "add         %%rcx,%3                      \n"
+      "add         %%rdx,%%r8                    \n"
+      "add         %%rsi,%%r9                    \n"
+      "add         %%rdi,%%r10                   \n"
+      "sub         $0x20,%2                      \n"
+      "jg          1b                            \n"
+
+      "add         %%r8, %3                      \n"
+      "add         %%r9, %3                      \n"
+      "add         %%r10, %3                     \n"
+      : "+r"(src_a),  // %0
+        "+r"(src_b),  // %1
+        "+r"(count),  // %2
+        "=&r"(diff)   // %3
+      :
+      : "cc", "memory", "rcx", "rdx", "rsi", "rdi", "r8", "r9", "r10");
+
+  return (uint32_t)(diff);
+}
+#else
+uint32_t HammingDistance_SSE42(const uint8_t* src_a,
+                               const uint8_t* src_b,
+                               int count) {
+  uint32_t diff = 0u;
+
+  asm volatile (
+      // Process 16 bytes per loop.
+      LABELALIGN
+      "1:                                        \n"
+      "mov         (%0),%%ecx                    \n"
+      "mov         0x4(%0),%%edx                 \n"
+      "xor         (%1),%%ecx                    \n"
+      "xor         0x4(%1),%%edx                 \n"
+      "popcnt      %%ecx,%%ecx                   \n"
+      "add         %%ecx,%3                      \n"
+      "popcnt      %%edx,%%edx                   \n"
+      "add         %%edx,%3                      \n"
+      "mov         0x8(%0),%%ecx                 \n"
+      "mov         0xc(%0),%%edx                 \n"
+      "xor         0x8(%1),%%ecx                 \n"
+      "xor         0xc(%1),%%edx                 \n"
+      "popcnt      %%ecx,%%ecx                   \n"
+      "add         %%ecx,%3                      \n"
+      "popcnt      %%edx,%%edx                   \n"
+      "add         %%edx,%3                      \n"
+      "add         $0x10,%0                      \n"
+      "add         $0x10,%1                      \n"
+      "sub         $0x10,%2                      \n"
+      "jg          1b                            \n"
+      : "+r"(src_a),  // %0
+        "+r"(src_b),  // %1
+        "+r"(count),  // %2
+        "+r"(diff)    // %3
+      :
+      : "cc", "memory", "ecx", "edx");
+
+  return diff;
+}
+#endif
+
+static const vec8 kNibbleMask = {15, 15, 15, 15, 15, 15, 15, 15,
+                                 15, 15, 15, 15, 15, 15, 15, 15};
+static const vec8 kBitCount = {0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};
+
+uint32_t HammingDistance_SSSE3(const uint8_t* src_a,
+                               const uint8_t* src_b,
+                               int count) {
+  uint32_t diff;
+
+  asm volatile (
+      "movdqa      %4,%%xmm2                     \n"
+      "movdqa      %5,%%xmm3                     \n"
+      "pxor        %%xmm0,%%xmm0                 \n"
+      "pxor        %%xmm1,%%xmm1                 \n"
+      "sub         %0,%1                         \n"
+
+      LABELALIGN
+      "1:                                        \n"
+      "movdqa      (%0),%%xmm4                   \n"
+      "movdqa      0x10(%0), %%xmm5              \n"
+      "pxor        (%0,%1), %%xmm4               \n"
+      "movdqa      %%xmm4,%%xmm6                 \n"
+      "pand        %%xmm2,%%xmm6                 \n"
+      "psrlw       $0x4,%%xmm4                   \n"
+      "movdqa      %%xmm3,%%xmm7                 \n"
+      "pshufb      %%xmm6,%%xmm7                 \n"
+      "pand        %%xmm2,%%xmm4                 \n"
+      "movdqa      %%xmm3,%%xmm6                 \n"
+      "pshufb      %%xmm4,%%xmm6                 \n"
+      "paddb       %%xmm7,%%xmm6                 \n"
+      "pxor        0x10(%0,%1),%%xmm5            \n"
+      "add         $0x20,%0                      \n"
+      "movdqa      %%xmm5,%%xmm4                 \n"
+      "pand        %%xmm2,%%xmm5                 \n"
+      "psrlw       $0x4,%%xmm4                   \n"
+      "movdqa      %%xmm3,%%xmm7                 \n"
+      "pshufb      %%xmm5,%%xmm7                 \n"
+      "pand        %%xmm2,%%xmm4                 \n"
+      "movdqa      %%xmm3,%%xmm5                 \n"
+      "pshufb      %%xmm4,%%xmm5                 \n"
+      "paddb       %%xmm7,%%xmm5                 \n"
+      "paddb       %%xmm5,%%xmm6                 \n"
+      "psadbw      %%xmm1,%%xmm6                 \n"
+      "paddd       %%xmm6,%%xmm0                 \n"
+      "sub         $0x20,%2                      \n"
+      "jg          1b                            \n"
+
+      "pshufd      $0xaa,%%xmm0,%%xmm1           \n"
+      "paddd       %%xmm1,%%xmm0                 \n"
+      "movd        %%xmm0, %3                    \n"
+      : "+r"(src_a),       // %0
+        "+r"(src_b),       // %1
+        "+r"(count),       // %2
+        "=r"(diff)         // %3
+      : "m"(kNibbleMask),  // %4
+        "m"(kBitCount)     // %5
+      : "cc", "memory", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6",
+        "xmm7");
+
+  return diff;
+}
+
+#ifdef HAS_HAMMINGDISTANCE_AVX2
+uint32_t HammingDistance_AVX2(const uint8_t* src_a,
+                              const uint8_t* src_b,
+                              int count) {
+  uint32_t diff;
+
+      asm volatile (
+      "vbroadcastf128 %4,%%ymm2                  \n"
+      "vbroadcastf128 %5,%%ymm3                  \n"
+      "vpxor       %%ymm0,%%ymm0,%%ymm0          \n"
+      "vpxor       %%ymm1,%%ymm1,%%ymm1          \n"
+      "sub         %0,%1                         \n"
+
+      LABELALIGN
+      "1:                                        \n"
+      "vmovdqa     (%0),%%ymm4                   \n"
+      "vmovdqa     0x20(%0), %%ymm5              \n"
+      "vpxor       (%0,%1), %%ymm4, %%ymm4       \n"
+      "vpand       %%ymm2,%%ymm4,%%ymm6          \n"
+      "vpsrlw      $0x4,%%ymm4,%%ymm4            \n"
+      "vpshufb     %%ymm6,%%ymm3,%%ymm6          \n"
+      "vpand       %%ymm2,%%ymm4,%%ymm4          \n"
+      "vpshufb     %%ymm4,%%ymm3,%%ymm4          \n"
+      "vpaddb      %%ymm4,%%ymm6,%%ymm6          \n"
+      "vpxor       0x20(%0,%1),%%ymm5,%%ymm4     \n"
+      "add         $0x40,%0                      \n"
+      "vpand       %%ymm2,%%ymm4,%%ymm5          \n"
+      "vpsrlw      $0x4,%%ymm4,%%ymm4            \n"
+      "vpshufb     %%ymm5,%%ymm3,%%ymm5          \n"
+      "vpand       %%ymm2,%%ymm4,%%ymm4          \n"
+      "vpshufb     %%ymm4,%%ymm3,%%ymm4          \n"
+      "vpaddb      %%ymm5,%%ymm4,%%ymm4          \n"
+      "vpaddb      %%ymm6,%%ymm4,%%ymm4          \n"
+      "vpsadbw     %%ymm1,%%ymm4,%%ymm4          \n"
+      "vpaddd      %%ymm0,%%ymm4,%%ymm0          \n"
+      "sub         $0x40,%2                      \n"
+      "jg          1b                            \n"
+
+      "vpermq      $0xb1,%%ymm0,%%ymm1           \n"
+      "vpaddd      %%ymm1,%%ymm0,%%ymm0          \n"
+      "vpermq      $0xaa,%%ymm0,%%ymm1           \n"
+      "vpaddd      %%ymm1,%%ymm0,%%ymm0          \n"
+      "vmovd       %%xmm0,%3                     \n"
+      "vzeroupper                                \n"
+      : "+r"(src_a),       // %0
+        "+r"(src_b),       // %1
+        "+r"(count),       // %2
+        "=r"(diff)         // %3
+      : "m"(kNibbleMask),  // %4
+        "m"(kBitCount)     // %5
+      : "cc", "memory", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6");
+
+  return diff;
+}
+#endif  // HAS_HAMMINGDISTANCE_AVX2
+
+uint32_t SumSquareError_SSE2(const uint8_t* src_a,
+                             const uint8_t* src_b,
+                             int count) {
+  uint32_t sse;
+      asm volatile (
+      "pxor        %%xmm0,%%xmm0                 \n"
+      "pxor        %%xmm5,%%xmm5                 \n"
+
+      LABELALIGN
+      "1:                                        \n"
+      "movdqu      (%0),%%xmm1                   \n"
+      "lea         0x10(%0),%0                   \n"
+      "movdqu      (%1),%%xmm2                   \n"
+      "lea         0x10(%1),%1                   \n"
+      "movdqa      %%xmm1,%%xmm3                 \n"
+      "psubusb     %%xmm2,%%xmm1                 \n"
+      "psubusb     %%xmm3,%%xmm2                 \n"
+      "por         %%xmm2,%%xmm1                 \n"
+      "movdqa      %%xmm1,%%xmm2                 \n"
+      "punpcklbw   %%xmm5,%%xmm1                 \n"
+      "punpckhbw   %%xmm5,%%xmm2                 \n"
+      "pmaddwd     %%xmm1,%%xmm1                 \n"
+      "pmaddwd     %%xmm2,%%xmm2                 \n"
+      "paddd       %%xmm1,%%xmm0                 \n"
+      "paddd       %%xmm2,%%xmm0                 \n"
+      "sub         $0x10,%2                      \n"
+      "jg          1b                            \n"
+
+      "pshufd      $0xee,%%xmm0,%%xmm1           \n"
+      "paddd       %%xmm1,%%xmm0                 \n"
+      "pshufd      $0x1,%%xmm0,%%xmm1            \n"
+      "paddd       %%xmm1,%%xmm0                 \n"
+      "movd        %%xmm0,%3                     \n"
+      : "+r"(src_a),  // %0
+        "+r"(src_b),  // %1
+        "+r"(count),  // %2
+        "=r"(sse)     // %3
+      :
+      : "cc", "memory", "xmm0", "xmm1", "xmm2", "xmm3", "xmm5");
+  return sse;
+}
+
+static const uvec32 kHash16x33 = {0x92d9e201, 0, 0, 0};  // 33 ^ 16
+static const uvec32 kHashMul0 = {
+    0x0c3525e1,  // 33 ^ 15
+    0xa3476dc1,  // 33 ^ 14
+    0x3b4039a1,  // 33 ^ 13
+    0x4f5f0981,  // 33 ^ 12
+};
+static const uvec32 kHashMul1 = {
+    0x30f35d61,  // 33 ^ 11
+    0x855cb541,  // 33 ^ 10
+    0x040a9121,  // 33 ^ 9
+    0x747c7101,  // 33 ^ 8
+};
+static const uvec32 kHashMul2 = {
+    0xec41d4e1,  // 33 ^ 7
+    0x4cfa3cc1,  // 33 ^ 6
+    0x025528a1,  // 33 ^ 5
+    0x00121881,  // 33 ^ 4
+};
+static const uvec32 kHashMul3 = {
+    0x00008c61,  // 33 ^ 3
+    0x00000441,  // 33 ^ 2
+    0x00000021,  // 33 ^ 1
+    0x00000001,  // 33 ^ 0
+};
+
+uint32_t HashDjb2_SSE41(const uint8_t* src, int count, uint32_t seed) {
+  uint32_t hash;
+      asm volatile (
+      "movd        %2,%%xmm0                     \n"
+      "pxor        %%xmm7,%%xmm7                 \n"
+      "movdqa      %4,%%xmm6                     \n"
+
+      LABELALIGN
+      "1:                                        \n"
+      "movdqu      (%0),%%xmm1                   \n"
+      "lea         0x10(%0),%0                   \n"
+      "pmulld      %%xmm6,%%xmm0                 \n"
+      "movdqa      %5,%%xmm5                     \n"
+      "movdqa      %%xmm1,%%xmm2                 \n"
+      "punpcklbw   %%xmm7,%%xmm2                 \n"
+      "movdqa      %%xmm2,%%xmm3                 \n"
+      "punpcklwd   %%xmm7,%%xmm3                 \n"
+      "pmulld      %%xmm5,%%xmm3                 \n"
+      "movdqa      %6,%%xmm5                     \n"
+      "movdqa      %%xmm2,%%xmm4                 \n"
+      "punpckhwd   %%xmm7,%%xmm4                 \n"
+      "pmulld      %%xmm5,%%xmm4                 \n"
+      "movdqa      %7,%%xmm5                     \n"
+      "punpckhbw   %%xmm7,%%xmm1                 \n"
+      "movdqa      %%xmm1,%%xmm2                 \n"
+      "punpcklwd   %%xmm7,%%xmm2                 \n"
+      "pmulld      %%xmm5,%%xmm2                 \n"
+      "movdqa      %8,%%xmm5                     \n"
+      "punpckhwd   %%xmm7,%%xmm1                 \n"
+      "pmulld      %%xmm5,%%xmm1                 \n"
+      "paddd       %%xmm4,%%xmm3                 \n"
+      "paddd       %%xmm2,%%xmm1                 \n"
+      "paddd       %%xmm3,%%xmm1                 \n"
+      "pshufd      $0xe,%%xmm1,%%xmm2            \n"
+      "paddd       %%xmm2,%%xmm1                 \n"
+      "pshufd      $0x1,%%xmm1,%%xmm2            \n"
+      "paddd       %%xmm2,%%xmm1                 \n"
+      "paddd       %%xmm1,%%xmm0                 \n"
+      "sub         $0x10,%1                      \n"
+      "jg          1b                            \n"
+      "movd        %%xmm0,%3                     \n"
+      : "+r"(src),        // %0
+        "+r"(count),      // %1
+        "+rm"(seed),      // %2
+        "=r"(hash)        // %3
+      : "m"(kHash16x33),  // %4
+        "m"(kHashMul0),   // %5
+        "m"(kHashMul1),   // %6
+        "m"(kHashMul2),   // %7
+        "m"(kHashMul3)    // %8
+      : "cc", "memory", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6",
+        "xmm7");
+  return hash;
+}
+#endif  // defined(__x86_64__) || (defined(__i386__) && !defined(__pic__)))
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/compare_msa.cc

@@ -0,0 +1,97 @@
+/*
+ *  Copyright 2017 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/basic_types.h"
+
+#include "libyuv/compare_row.h"
+#include "libyuv/row.h"
+
+// This module is for GCC MSA
+#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
+#include "libyuv/macros_msa.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+uint32_t HammingDistance_MSA(const uint8_t* src_a,
+                             const uint8_t* src_b,
+                             int count) {
+  uint32_t diff = 0u;
+  int i;
+  v16u8 src0, src1, src2, src3;
+  v2i64 vec0 = {0}, vec1 = {0};
+
+  for (i = 0; i < count; i += 32) {
+    src0 = (v16u8)__msa_ld_b((v16i8*)src_a, 0);
+    src1 = (v16u8)__msa_ld_b((v16i8*)src_a, 16);
+    src2 = (v16u8)__msa_ld_b((v16i8*)src_b, 0);
+    src3 = (v16u8)__msa_ld_b((v16i8*)src_b, 16);
+    src0 ^= src2;
+    src1 ^= src3;
+    vec0 += __msa_pcnt_d((v2i64)src0);
+    vec1 += __msa_pcnt_d((v2i64)src1);
+    src_a += 32;
+    src_b += 32;
+  }
+
+  vec0 += vec1;
+  diff = (uint32_t)__msa_copy_u_w((v4i32)vec0, 0);
+  diff += (uint32_t)__msa_copy_u_w((v4i32)vec0, 2);
+  return diff;
+}
+
+uint32_t SumSquareError_MSA(const uint8_t* src_a,
+                            const uint8_t* src_b,
+                            int count) {
+  uint32_t sse = 0u;
+  int i;
+  v16u8 src0, src1, src2, src3;
+  v8i16 vec0, vec1, vec2, vec3;
+  v4i32 reg0 = {0}, reg1 = {0}, reg2 = {0}, reg3 = {0};
+  v2i64 tmp0;
+
+  for (i = 0; i < count; i += 32) {
+    src0 = (v16u8)__msa_ld_b((v16i8*)src_a, 0);
+    src1 = (v16u8)__msa_ld_b((v16i8*)src_a, 16);
+    src2 = (v16u8)__msa_ld_b((v16i8*)src_b, 0);
+    src3 = (v16u8)__msa_ld_b((v16i8*)src_b, 16);
+    vec0 = (v8i16)__msa_ilvr_b((v16i8)src2, (v16i8)src0);
+    vec1 = (v8i16)__msa_ilvl_b((v16i8)src2, (v16i8)src0);
+    vec2 = (v8i16)__msa_ilvr_b((v16i8)src3, (v16i8)src1);
+    vec3 = (v8i16)__msa_ilvl_b((v16i8)src3, (v16i8)src1);
+    vec0 = __msa_hsub_u_h((v16u8)vec0, (v16u8)vec0);
+    vec1 = __msa_hsub_u_h((v16u8)vec1, (v16u8)vec1);
+    vec2 = __msa_hsub_u_h((v16u8)vec2, (v16u8)vec2);
+    vec3 = __msa_hsub_u_h((v16u8)vec3, (v16u8)vec3);
+    reg0 = __msa_dpadd_s_w(reg0, vec0, vec0);
+    reg1 = __msa_dpadd_s_w(reg1, vec1, vec1);
+    reg2 = __msa_dpadd_s_w(reg2, vec2, vec2);
+    reg3 = __msa_dpadd_s_w(reg3, vec3, vec3);
+    src_a += 32;
+    src_b += 32;
+  }
+
+  reg0 += reg1;
+  reg2 += reg3;
+  reg0 += reg2;
+  tmp0 = __msa_hadd_s_d(reg0, reg0);
+  sse = (uint32_t)__msa_copy_u_w((v4i32)tmp0, 0);
+  sse += (uint32_t)__msa_copy_u_w((v4i32)tmp0, 2);
+  return sse;
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)

yuv.mod/libyuv/source/compare_neon.cc

@@ -0,0 +1,96 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/basic_types.h"
+
+#include "libyuv/compare_row.h"
+#include "libyuv/row.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+#if !defined(LIBYUV_DISABLE_NEON) && defined(__ARM_NEON__) && \
+    !defined(__aarch64__)
+
+// 256 bits at a time
+// uses short accumulator which restricts count to 131 KB
+uint32_t HammingDistance_NEON(const uint8_t* src_a,
+                              const uint8_t* src_b,
+                              int count) {
+  uint32_t diff;
+
+  asm volatile (
+      "vmov.u16    q4, #0                        \n"  // accumulator
+
+      "1:                                        \n"
+      "vld1.8      {q0, q1}, [%0]!               \n"
+      "vld1.8      {q2, q3}, [%1]!               \n"
+      "veor.32     q0, q0, q2                    \n"
+      "veor.32     q1, q1, q3                    \n"
+      "vcnt.i8     q0, q0                        \n"
+      "vcnt.i8     q1, q1                        \n"
+      "subs        %2, %2, #32                   \n"
+      "vadd.u8     q0, q0, q1                    \n"  // 16 byte counts
+      "vpadal.u8   q4, q0                        \n"  // 8 shorts
+      "bgt         1b                            \n"
+
+      "vpaddl.u16  q0, q4                        \n"  // 4 ints
+      "vpadd.u32   d0, d0, d1                    \n"
+      "vpadd.u32   d0, d0, d0                    \n"
+      "vmov.32     %3, d0[0]                     \n"
+
+      : "+r"(src_a), "+r"(src_b), "+r"(count), "=r"(diff)
+      :
+      : "cc", "q0", "q1", "q2", "q3", "q4");
+  return diff;
+}
+
+uint32_t SumSquareError_NEON(const uint8_t* src_a,
+                             const uint8_t* src_b,
+                             int count) {
+  uint32_t sse;
+  asm volatile (
+      "vmov.u8     q8, #0                        \n"
+      "vmov.u8     q10, #0                       \n"
+      "vmov.u8     q9, #0                        \n"
+      "vmov.u8     q11, #0                       \n"
+
+      "1:                                        \n"
+      "vld1.8      {q0}, [%0]!                   \n"
+      "vld1.8      {q1}, [%1]!                   \n"
+      "subs        %2, %2, #16                   \n"
+      "vsubl.u8    q2, d0, d2                    \n"
+      "vsubl.u8    q3, d1, d3                    \n"
+      "vmlal.s16   q8, d4, d4                    \n"
+      "vmlal.s16   q9, d6, d6                    \n"
+      "vmlal.s16   q10, d5, d5                   \n"
+      "vmlal.s16   q11, d7, d7                   \n"
+      "bgt         1b                            \n"
+
+      "vadd.u32    q8, q8, q9                    \n"
+      "vadd.u32    q10, q10, q11                 \n"
+      "vadd.u32    q11, q8, q10                  \n"
+      "vpaddl.u32  q1, q11                       \n"
+      "vadd.u64    d0, d2, d3                    \n"
+      "vmov.32     %3, d0[0]                     \n"
+      : "+r"(src_a), "+r"(src_b), "+r"(count), "=r"(sse)
+      :
+      : "memory", "cc", "q0", "q1", "q2", "q3", "q8", "q9", "q10", "q11");
+  return sse;
+}
+
+#endif  // defined(__ARM_NEON__) && !defined(__aarch64__)
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/compare_neon64.cc

@@ -0,0 +1,223 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/basic_types.h"
+
+#include "libyuv/compare_row.h"
+#include "libyuv/row.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+#if !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__)
+
+// 256 bits at a time
+// uses short accumulator which restricts count to 131 KB
+uint32_t HammingDistance_NEON(const uint8_t* src_a,
+                              const uint8_t* src_b,
+                              int count) {
+  uint32_t diff;
+  asm volatile (
+      "movi        v4.8h, #0                     \n"
+
+      "1:                                        \n"
+      "ld1         {v0.16b, v1.16b}, [%0], #32   \n"
+      "ld1         {v2.16b, v3.16b}, [%1], #32   \n"
+      "eor         v0.16b, v0.16b, v2.16b        \n"
+      "prfm        pldl1keep, [%0, 448]          \n"  // prefetch 7 lines ahead
+      "eor         v1.16b, v1.16b, v3.16b        \n"
+      "cnt         v0.16b, v0.16b                \n"
+      "prfm        pldl1keep, [%1, 448]          \n"
+      "cnt         v1.16b, v1.16b                \n"
+      "subs        %w2, %w2, #32                 \n"
+      "add         v0.16b, v0.16b, v1.16b        \n"
+      "uadalp      v4.8h, v0.16b                 \n"
+      "b.gt        1b                            \n"
+
+      "uaddlv      s4, v4.8h                     \n"
+      "fmov        %w3, s4                       \n"
+      : "+r"(src_a), "+r"(src_b), "+r"(count), "=r"(diff)
+      :
+      : "memory", "cc", "v0", "v1", "v2", "v3", "v4");
+  return diff;
+}
+
+uint32_t SumSquareError_NEON(const uint8_t* src_a,
+                             const uint8_t* src_b,
+                             int count) {
+  uint32_t sse;
+  asm volatile (
+      "movi        v16.16b, #0                   \n"
+      "movi        v17.16b, #0                   \n"
+      "movi        v18.16b, #0                   \n"
+      "movi        v19.16b, #0                   \n"
+
+      "1:                                        \n"
+      "ld1         {v0.16b}, [%0], #16           \n"
+      "ld1         {v1.16b}, [%1], #16           \n"
+      "subs        %w2, %w2, #16                 \n"
+      "usubl       v2.8h, v0.8b, v1.8b           \n"
+      "usubl2      v3.8h, v0.16b, v1.16b         \n"
+      "prfm        pldl1keep, [%0, 448]          \n"  // prefetch 7 lines ahead
+      "smlal       v16.4s, v2.4h, v2.4h          \n"
+      "smlal       v17.4s, v3.4h, v3.4h          \n"
+      "prfm        pldl1keep, [%1, 448]          \n"
+      "smlal2      v18.4s, v2.8h, v2.8h          \n"
+      "smlal2      v19.4s, v3.8h, v3.8h          \n"
+      "b.gt        1b                            \n"
+
+      "add         v16.4s, v16.4s, v17.4s        \n"
+      "add         v18.4s, v18.4s, v19.4s        \n"
+      "add         v19.4s, v16.4s, v18.4s        \n"
+      "addv        s0, v19.4s                    \n"
+      "fmov        %w3, s0                       \n"
+      : "+r"(src_a), "+r"(src_b), "+r"(count), "=r"(sse)
+      :
+      : "memory", "cc", "v0", "v1", "v2", "v3", "v16", "v17", "v18", "v19");
+  return sse;
+}
+
+static const uvec32 kDjb2Multiplicands[] = {
+    {0x0c3525e1,   // 33^15
+     0xa3476dc1,   // 33^14
+     0x3b4039a1,   // 33^13
+     0x4f5f0981},  // 33^12
+    {0x30f35d61,   // 33^11
+     0x855cb541,   // 33^10
+     0x040a9121,   // 33^9
+     0x747c7101},  // 33^8
+    {0xec41d4e1,   // 33^7
+     0x4cfa3cc1,   // 33^6
+     0x025528a1,   // 33^5
+     0x00121881},  // 33^4
+    {0x00008c61,   // 33^3
+     0x00000441,   // 33^2
+     0x00000021,   // 33^1
+     0x00000001},  // 33^0
+};
+
+static const uvec32 kDjb2WidenIndices[] = {
+    {0xffffff00U, 0xffffff01U, 0xffffff02U, 0xffffff03U},
+    {0xffffff04U, 0xffffff05U, 0xffffff06U, 0xffffff07U},
+    {0xffffff08U, 0xffffff09U, 0xffffff0aU, 0xffffff0bU},
+    {0xffffff0cU, 0xffffff0dU, 0xffffff0eU, 0xffffff0fU},
+};
+
+uint32_t HashDjb2_NEON(const uint8_t* src, int count, uint32_t seed) {
+  uint32_t hash = seed;
+  const uint32_t c16 = 0x92d9e201;  // 33^16
+  uint32_t tmp, tmp2;
+  asm("ld1   {v16.4s, v17.4s, v18.4s, v19.4s}, [%[kIdx]] \n"
+      "ld1   {v4.4s, v5.4s, v6.4s, v7.4s}, [%[kMuls]]    \n"
+
+      // count is always a multiple of 16.
+      // maintain two accumulators, reduce and then final sum in scalar since
+      // this has better performance on little cores.
+      "1:                                \n"
+      "ldr   q0, [%[src]], #16           \n"
+      "subs  %w[count], %w[count], #16   \n"
+      "tbl   v3.16b, {v0.16b}, v19.16b   \n"
+      "tbl   v2.16b, {v0.16b}, v18.16b   \n"
+      "tbl   v1.16b, {v0.16b}, v17.16b   \n"
+      "tbl   v0.16b, {v0.16b}, v16.16b   \n"
+      "mul   v3.4s, v3.4s, v7.4s         \n"
+      "mul   v2.4s, v2.4s, v6.4s         \n"
+      "mla   v3.4s, v1.4s, v5.4s         \n"
+      "mla   v2.4s, v0.4s, v4.4s         \n"
+      "addv  s1, v3.4s                   \n"
+      "addv  s0, v2.4s                   \n"
+      "fmov  %w[tmp2], s1                \n"
+      "fmov  %w[tmp], s0                 \n"
+      "add   %w[tmp], %w[tmp], %w[tmp2]  \n"
+      "madd  %w[hash], %w[hash], %w[c16], %w[tmp] \n"
+      "b.gt  1b                          \n"
+      : [hash] "+r"(hash),                // %[hash]
+        [count] "+r"(count),              // %[count]
+        [tmp] "=&r"(tmp),                 // %[tmp]
+        [tmp2] "=&r"(tmp2)                // %[tmp2]
+      : [src] "r"(src),                   // %[src]
+        [kMuls] "r"(kDjb2Multiplicands),  // %[kMuls]
+        [kIdx] "r"(kDjb2WidenIndices),    // %[kIdx]
+        [c16] "r"(c16)                    // %[c16]
+      : "cc", "memory", "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v16",
+        "v17", "v18", "v19");
+  return hash;
+}
+
+uint32_t HammingDistance_NEON_DotProd(const uint8_t* src_a,
+                                      const uint8_t* src_b,
+                                      int count) {
+  uint32_t diff;
+  asm volatile (
+      "movi        v4.4s, #0                     \n"
+      "movi        v5.4s, #0                     \n"
+      "movi        v6.16b, #1                    \n"
+
+      "1:                                        \n"
+      "ldp         q0, q1, [%0], #32             \n"
+      "ldp         q2, q3, [%1], #32             \n"
+      "eor         v0.16b, v0.16b, v2.16b        \n"
+      "prfm        pldl1keep, [%0, 448]          \n"  // prefetch 7 lines ahead
+      "eor         v1.16b, v1.16b, v3.16b        \n"
+      "cnt         v0.16b, v0.16b                \n"
+      "prfm        pldl1keep, [%1, 448]          \n"
+      "cnt         v1.16b, v1.16b                \n"
+      "subs        %w2, %w2, #32                 \n"
+      "udot        v4.4s, v0.16b, v6.16b         \n"
+      "udot        v5.4s, v1.16b, v6.16b         \n"
+      "b.gt        1b                            \n"
+
+      "add         v0.4s, v4.4s, v5.4s           \n"
+      "addv        s0, v0.4s                     \n"
+      "fmov        %w3, s0                       \n"
+      : "+r"(src_a), "+r"(src_b), "+r"(count), "=r"(diff)
+      :
+      : "memory", "cc", "v0", "v1", "v2", "v3", "v4", "v5", "v6");
+  return diff;
+}
+
+uint32_t SumSquareError_NEON_DotProd(const uint8_t* src_a,
+                                     const uint8_t* src_b,
+                                     int count) {
+  // count is guaranteed to be a multiple of 32.
+  uint32_t sse;
+  asm volatile (
+      "movi        v4.4s, #0                     \n"
+      "movi        v5.4s, #0                     \n"
+
+      "1:                                        \n"
+      "ldp         q0, q2, [%0], #32             \n"
+      "ldp         q1, q3, [%1], #32             \n"
+      "subs        %w2, %w2, #32                 \n"
+      "uabd        v0.16b, v0.16b, v1.16b        \n"
+      "uabd        v1.16b, v2.16b, v3.16b        \n"
+      "prfm        pldl1keep, [%0, 448]          \n"  // prefetch 7 lines ahead
+      "udot        v4.4s, v0.16b, v0.16b         \n"
+      "udot        v5.4s, v1.16b, v1.16b         \n"
+      "prfm        pldl1keep, [%1, 448]          \n"
+      "b.gt        1b                            \n"
+
+      "add         v0.4s, v4.4s, v5.4s           \n"
+      "addv        s0, v0.4s                     \n"
+      "fmov        %w3, s0                       \n"
+      : "+r"(src_a), "+r"(src_b), "+r"(count), "=r"(sse)
+      :
+      : "memory", "cc", "v0", "v1", "v2", "v3", "v4", "v5");
+  return sse;
+}
+
+#endif  // !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__)
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/compare_win.cc

@@ -0,0 +1,241 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/basic_types.h"
+
+#include "libyuv/compare_row.h"
+#include "libyuv/row.h"
+
+#if defined(_MSC_VER)
+#include <intrin.h>  // For __popcnt
+#endif
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// This module is for 32 bit Visual C x86
+#if !defined(LIBYUV_DISABLE_X86) && defined(_MSC_VER) && \
+    !defined(__clang__) && defined(_M_IX86)
+
+uint32_t HammingDistance_SSE42(const uint8_t* src_a,
+                               const uint8_t* src_b,
+                               int count) {
+  uint32_t diff = 0u;
+
+  int i;
+  for (i = 0; i < count - 3; i += 4) {
+    uint32_t x = *((uint32_t*)src_a) ^ *((uint32_t*)src_b);  // NOLINT
+    src_a += 4;
+    src_b += 4;
+    diff += __popcnt(x);
+  }
+  return diff;
+}
+
+__declspec(naked) uint32_t
+    SumSquareError_SSE2(const uint8_t* src_a, const uint8_t* src_b, int count) {
+  __asm {
+    mov        eax, [esp + 4]  // src_a
+    mov        edx, [esp + 8]  // src_b
+    mov        ecx, [esp + 12]  // count
+    pxor       xmm0, xmm0
+    pxor       xmm5, xmm5
+
+  wloop:
+    movdqu     xmm1, [eax]
+    lea        eax,  [eax + 16]
+    movdqu     xmm2, [edx]
+    lea        edx,  [edx + 16]
+    movdqa     xmm3, xmm1  // abs trick
+    psubusb    xmm1, xmm2
+    psubusb    xmm2, xmm3
+    por        xmm1, xmm2
+    movdqa     xmm2, xmm1
+    punpcklbw  xmm1, xmm5
+    punpckhbw  xmm2, xmm5
+    pmaddwd    xmm1, xmm1
+    pmaddwd    xmm2, xmm2
+    paddd      xmm0, xmm1
+    paddd      xmm0, xmm2
+    sub        ecx, 16
+    jg         wloop
+
+    pshufd     xmm1, xmm0, 0xee
+    paddd      xmm0, xmm1
+    pshufd     xmm1, xmm0, 0x01
+    paddd      xmm0, xmm1
+    movd       eax, xmm0
+    ret
+  }
+}
+
+#ifdef HAS_SUMSQUAREERROR_AVX2
+// C4752: found Intel(R) Advanced Vector Extensions; consider using /arch:AVX.
+#pragma warning(disable : 4752)
+__declspec(naked) uint32_t
+    SumSquareError_AVX2(const uint8_t* src_a, const uint8_t* src_b, int count) {
+  __asm {
+    mov        eax, [esp + 4]  // src_a
+    mov        edx, [esp + 8]  // src_b
+    mov        ecx, [esp + 12]  // count
+    vpxor      ymm0, ymm0, ymm0  // sum
+    vpxor      ymm5, ymm5, ymm5  // constant 0 for unpck
+    sub        edx, eax
+
+  wloop:
+    vmovdqu    ymm1, [eax]
+    vmovdqu    ymm2, [eax + edx]
+    lea        eax,  [eax + 32]
+    vpsubusb   ymm3, ymm1, ymm2  // abs difference trick
+    vpsubusb   ymm2, ymm2, ymm1
+    vpor       ymm1, ymm2, ymm3
+    vpunpcklbw ymm2, ymm1, ymm5  // u16.  mutates order.
+    vpunpckhbw ymm1, ymm1, ymm5
+    vpmaddwd   ymm2, ymm2, ymm2  // square + hadd to u32.
+    vpmaddwd   ymm1, ymm1, ymm1
+    vpaddd     ymm0, ymm0, ymm1
+    vpaddd     ymm0, ymm0, ymm2
+    sub        ecx, 32
+    jg         wloop
+
+    vpshufd    ymm1, ymm0, 0xee  // 3, 2 + 1, 0 both lanes.
+    vpaddd     ymm0, ymm0, ymm1
+    vpshufd    ymm1, ymm0, 0x01  // 1 + 0 both lanes.
+    vpaddd     ymm0, ymm0, ymm1
+    vpermq     ymm1, ymm0, 0x02  // high + low lane.
+    vpaddd     ymm0, ymm0, ymm1
+    vmovd      eax, xmm0
+    vzeroupper
+    ret
+  }
+}
+#endif  // HAS_SUMSQUAREERROR_AVX2
+
+uvec32 kHash16x33 = {0x92d9e201, 0, 0, 0};  // 33 ^ 16
+uvec32 kHashMul0 = {
+    0x0c3525e1,  // 33 ^ 15
+    0xa3476dc1,  // 33 ^ 14
+    0x3b4039a1,  // 33 ^ 13
+    0x4f5f0981,  // 33 ^ 12
+};
+uvec32 kHashMul1 = {
+    0x30f35d61,  // 33 ^ 11
+    0x855cb541,  // 33 ^ 10
+    0x040a9121,  // 33 ^ 9
+    0x747c7101,  // 33 ^ 8
+};
+uvec32 kHashMul2 = {
+    0xec41d4e1,  // 33 ^ 7
+    0x4cfa3cc1,  // 33 ^ 6
+    0x025528a1,  // 33 ^ 5
+    0x00121881,  // 33 ^ 4
+};
+uvec32 kHashMul3 = {
+    0x00008c61,  // 33 ^ 3
+    0x00000441,  // 33 ^ 2
+    0x00000021,  // 33 ^ 1
+    0x00000001,  // 33 ^ 0
+};
+
+__declspec(naked) uint32_t
+    HashDjb2_SSE41(const uint8_t* src, int count, uint32_t seed) {
+  __asm {
+    mov        eax, [esp + 4]  // src
+    mov        ecx, [esp + 8]  // count
+    movd       xmm0, [esp + 12]  // seed
+
+    pxor       xmm7, xmm7  // constant 0 for unpck
+    movdqa     xmm6, xmmword ptr kHash16x33
+
+  wloop:
+    movdqu     xmm1, [eax]  // src[0-15]
+    lea        eax, [eax + 16]
+    pmulld     xmm0, xmm6  // hash *= 33 ^ 16
+    movdqa     xmm5, xmmword ptr kHashMul0
+    movdqa     xmm2, xmm1
+    punpcklbw  xmm2, xmm7  // src[0-7]
+    movdqa     xmm3, xmm2
+    punpcklwd  xmm3, xmm7  // src[0-3]
+    pmulld     xmm3, xmm5
+    movdqa     xmm5, xmmword ptr kHashMul1
+    movdqa     xmm4, xmm2
+    punpckhwd  xmm4, xmm7  // src[4-7]
+    pmulld     xmm4, xmm5
+    movdqa     xmm5, xmmword ptr kHashMul2
+    punpckhbw  xmm1, xmm7  // src[8-15]
+    movdqa     xmm2, xmm1
+    punpcklwd  xmm2, xmm7  // src[8-11]
+    pmulld     xmm2, xmm5
+    movdqa     xmm5, xmmword ptr kHashMul3
+    punpckhwd  xmm1, xmm7  // src[12-15]
+    pmulld     xmm1, xmm5
+    paddd      xmm3, xmm4  // add 16 results
+    paddd      xmm1, xmm2
+    paddd      xmm1, xmm3
+
+    pshufd     xmm2, xmm1, 0x0e  // upper 2 dwords
+    paddd      xmm1, xmm2
+    pshufd     xmm2, xmm1, 0x01
+    paddd      xmm1, xmm2
+    paddd      xmm0, xmm1
+    sub        ecx, 16
+    jg         wloop
+
+    movd       eax, xmm0  // return hash
+    ret
+  }
+}
+
+// Visual C 2012 required for AVX2.
+#ifdef HAS_HASHDJB2_AVX2
+__declspec(naked) uint32_t
+    HashDjb2_AVX2(const uint8_t* src, int count, uint32_t seed) {
+  __asm {
+    mov        eax, [esp + 4]  // src
+    mov        ecx, [esp + 8]  // count
+    vmovd      xmm0, [esp + 12]  // seed
+
+  wloop:
+    vpmovzxbd  xmm3, [eax]  // src[0-3]
+    vpmulld    xmm0, xmm0, xmmword ptr kHash16x33  // hash *= 33 ^ 16
+    vpmovzxbd  xmm4, [eax + 4]  // src[4-7]
+    vpmulld    xmm3, xmm3, xmmword ptr kHashMul0
+    vpmovzxbd  xmm2, [eax + 8]  // src[8-11]
+    vpmulld    xmm4, xmm4, xmmword ptr kHashMul1
+    vpmovzxbd  xmm1, [eax + 12]  // src[12-15]
+    vpmulld    xmm2, xmm2, xmmword ptr kHashMul2
+    lea        eax, [eax + 16]
+    vpmulld    xmm1, xmm1, xmmword ptr kHashMul3
+    vpaddd     xmm3, xmm3, xmm4  // add 16 results
+    vpaddd     xmm1, xmm1, xmm2
+    vpaddd     xmm1, xmm1, xmm3
+    vpshufd    xmm2, xmm1, 0x0e  // upper 2 dwords
+    vpaddd     xmm1, xmm1,xmm2
+    vpshufd    xmm2, xmm1, 0x01
+    vpaddd     xmm1, xmm1, xmm2
+    vpaddd     xmm0, xmm0, xmm1
+    sub        ecx, 16
+    jg         wloop
+
+    vmovd      eax, xmm0  // return hash
+    vzeroupper
+    ret
+  }
+}
+#endif  // HAS_HASHDJB2_AVX2
+
+#endif  // !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86)
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/convert.cc

@@ -0,0 +1,4364 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/convert.h"
+
+#include "libyuv/basic_types.h"
+#include "libyuv/cpu_id.h"
+#include "libyuv/planar_functions.h"
+#include "libyuv/rotate.h"
+#include "libyuv/row.h"
+#include "libyuv/scale.h"      // For ScalePlane()
+#include "libyuv/scale_row.h"  // For FixedDiv
+#include "libyuv/scale_uv.h"   // For UVScale()
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Subsample amount uses a shift.
+//   v is value
+//   a is amount to add to round up
+//   s is shift to subsample down
+#define SUBSAMPLE(v, a, s) (v < 0) ? (-((-v + a) >> s)) : ((v + a) >> s)
+static __inline int Abs(int v) {
+  return v >= 0 ? v : -v;
+}
+
+// Any I4xx To I420 format with mirroring.
+static int I4xxToI420(const uint8_t* src_y,
+                      int src_stride_y,
+                      const uint8_t* src_u,
+                      int src_stride_u,
+                      const uint8_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_y,
+                      int dst_stride_y,
+                      uint8_t* dst_u,
+                      int dst_stride_u,
+                      uint8_t* dst_v,
+                      int dst_stride_v,
+                      int src_y_width,
+                      int src_y_height,
+                      int src_uv_width,
+                      int src_uv_height) {
+  const int dst_y_width = Abs(src_y_width);
+  const int dst_y_height = Abs(src_y_height);
+  const int dst_uv_width = SUBSAMPLE(dst_y_width, 1, 1);
+  const int dst_uv_height = SUBSAMPLE(dst_y_height, 1, 1);
+  int r;
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v ||
+      src_y_width <= 0 || src_y_height == 0 || src_uv_width <= 0 ||
+      src_uv_height == 0) {
+    return -1;
+  }
+  if (dst_y) {
+    r = ScalePlane(src_y, src_stride_y, src_y_width, src_y_height, dst_y,
+                   dst_stride_y, dst_y_width, dst_y_height, kFilterBilinear);
+    if (r != 0) {
+      return r;
+    }
+  }
+  r = ScalePlane(src_u, src_stride_u, src_uv_width, src_uv_height, dst_u,
+                 dst_stride_u, dst_uv_width, dst_uv_height, kFilterBilinear);
+  if (r != 0) {
+    return r;
+  }
+  r = ScalePlane(src_v, src_stride_v, src_uv_width, src_uv_height, dst_v,
+                 dst_stride_v, dst_uv_width, dst_uv_height, kFilterBilinear);
+  return r;
+}
+
+// Copy I420 with optional flipping.
+// TODO(fbarchard): Use Scale plane which supports mirroring, but ensure
+// is does row coalescing.
+LIBYUV_API
+int I420Copy(const uint8_t* src_y,
+             int src_stride_y,
+             const uint8_t* src_u,
+             int src_stride_u,
+             const uint8_t* src_v,
+             int src_stride_v,
+             uint8_t* dst_y,
+             int dst_stride_y,
+             uint8_t* dst_u,
+             int dst_stride_u,
+             uint8_t* dst_v,
+             int dst_stride_v,
+             int width,
+             int height) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (halfheight - 1) * src_stride_u;
+    src_v = src_v + (halfheight - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  if (dst_y) {
+    CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+  // Copy UV planes.
+  CopyPlane(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth, halfheight);
+  CopyPlane(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth, halfheight);
+  return 0;
+}
+
+// Copy I010 with optional flipping.
+LIBYUV_API
+int I010Copy(const uint16_t* src_y,
+             int src_stride_y,
+             const uint16_t* src_u,
+             int src_stride_u,
+             const uint16_t* src_v,
+             int src_stride_v,
+             uint16_t* dst_y,
+             int dst_stride_y,
+             uint16_t* dst_u,
+             int dst_stride_u,
+             uint16_t* dst_v,
+             int dst_stride_v,
+             int width,
+             int height) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (halfheight - 1) * src_stride_u;
+    src_v = src_v + (halfheight - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  if (dst_y) {
+    CopyPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+  // Copy UV planes.
+  CopyPlane_16(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth, halfheight);
+  CopyPlane_16(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth, halfheight);
+  return 0;
+}
+
+static int Planar16bitTo8bit(const uint16_t* src_y,
+                             int src_stride_y,
+                             const uint16_t* src_u,
+                             int src_stride_u,
+                             const uint16_t* src_v,
+                             int src_stride_v,
+                             uint8_t* dst_y,
+                             int dst_stride_y,
+                             uint8_t* dst_u,
+                             int dst_stride_u,
+                             uint8_t* dst_v,
+                             int dst_stride_v,
+                             int width,
+                             int height,
+                             int subsample_x,
+                             int subsample_y,
+                             int depth) {
+  int uv_width = SUBSAMPLE(width, subsample_x, subsample_x);
+  int uv_height = SUBSAMPLE(height, subsample_y, subsample_y);
+  int scale = 1 << (24 - depth);
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    uv_height = -uv_height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (uv_height - 1) * src_stride_u;
+    src_v = src_v + (uv_height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  // Convert Y plane.
+  if (dst_y) {
+    Convert16To8Plane(src_y, src_stride_y, dst_y, dst_stride_y, scale, width,
+                      height);
+  }
+  // Convert UV planes.
+  Convert16To8Plane(src_u, src_stride_u, dst_u, dst_stride_u, scale, uv_width,
+                    uv_height);
+  Convert16To8Plane(src_v, src_stride_v, dst_v, dst_stride_v, scale, uv_width,
+                    uv_height);
+  return 0;
+}
+
+static int I41xToI420(const uint16_t* src_y,
+                      int src_stride_y,
+                      const uint16_t* src_u,
+                      int src_stride_u,
+                      const uint16_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_y,
+                      int dst_stride_y,
+                      uint8_t* dst_u,
+                      int dst_stride_u,
+                      uint8_t* dst_v,
+                      int dst_stride_v,
+                      int width,
+                      int height,
+                      int depth) {
+  const int scale = 1 << (24 - depth);
+
+  if (width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  {
+    const int uv_width = SUBSAMPLE(width, 1, 1);
+    const int uv_height = SUBSAMPLE(height, 1, 1);
+
+    Convert16To8Plane(src_y, src_stride_y, dst_y, dst_stride_y, scale, width,
+                      height);
+    ScalePlaneDown2_16To8(width, height, uv_width, uv_height, src_stride_u,
+                          dst_stride_u, src_u, dst_u, scale, kFilterBilinear);
+    ScalePlaneDown2_16To8(width, height, uv_width, uv_height, src_stride_v,
+                          dst_stride_v, src_v, dst_v, scale, kFilterBilinear);
+  }
+  return 0;
+}
+
+static int I21xToI420(const uint16_t* src_y,
+                      int src_stride_y,
+                      const uint16_t* src_u,
+                      int src_stride_u,
+                      const uint16_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_y,
+                      int dst_stride_y,
+                      uint8_t* dst_u,
+                      int dst_stride_u,
+                      uint8_t* dst_v,
+                      int dst_stride_v,
+                      int width,
+                      int height,
+                      int depth) {
+  const int scale = 1 << (24 - depth);
+
+  if (width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  {
+    const int uv_width = SUBSAMPLE(width, 1, 1);
+    const int uv_height = SUBSAMPLE(height, 1, 1);
+    const int dy = FixedDiv(height, uv_height);
+
+    Convert16To8Plane(src_y, src_stride_y, dst_y, dst_stride_y, scale, width,
+                      height);
+    ScalePlaneVertical_16To8(height, uv_width, uv_height, src_stride_u,
+                             dst_stride_u, src_u, dst_u, 0, 32768, dy,
+                             /*bpp=*/1, scale, kFilterBilinear);
+    ScalePlaneVertical_16To8(height, uv_width, uv_height, src_stride_v,
+                             dst_stride_v, src_v, dst_v, 0, 32768, dy,
+                             /*bpp=*/1, scale, kFilterBilinear);
+  }
+  return 0;
+}
+
+// Convert 10 bit YUV to 8 bit.
+LIBYUV_API
+int I010ToI420(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return Planar16bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                           src_stride_v, dst_y, dst_stride_y, dst_u,
+                           dst_stride_u, dst_v, dst_stride_v, width, height, 1,
+                           1, 10);
+}
+
+LIBYUV_API
+int I210ToI420(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return I21xToI420(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                    dst_v, dst_stride_v, width, height, 10);
+}
+
+LIBYUV_API
+int I210ToI422(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return Planar16bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                           src_stride_v, dst_y, dst_stride_y, dst_u,
+                           dst_stride_u, dst_v, dst_stride_v, width, height, 1,
+                           0, 10);
+}
+
+LIBYUV_API
+int I410ToI420(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return I41xToI420(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                    dst_v, dst_stride_v, width, height, 10);
+}
+
+LIBYUV_API
+int I410ToI444(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return Planar16bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                           src_stride_v, dst_y, dst_stride_y, dst_u,
+                           dst_stride_u, dst_v, dst_stride_v, width, height, 0,
+                           0, 10);
+}
+
+LIBYUV_API
+int I012ToI420(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return Planar16bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                           src_stride_v, dst_y, dst_stride_y, dst_u,
+                           dst_stride_u, dst_v, dst_stride_v, width, height, 1,
+                           1, 12);
+}
+
+LIBYUV_API
+int I212ToI422(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return Planar16bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                           src_stride_v, dst_y, dst_stride_y, dst_u,
+                           dst_stride_u, dst_v, dst_stride_v, width, height, 1,
+                           0, 12);
+}
+
+LIBYUV_API
+int I212ToI420(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return I21xToI420(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                    dst_v, dst_stride_v, width, height, 12);
+}
+
+LIBYUV_API
+int I412ToI444(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return Planar16bitTo8bit(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                           src_stride_v, dst_y, dst_stride_y, dst_u,
+                           dst_stride_u, dst_v, dst_stride_v, width, height, 0,
+                           0, 12);
+}
+
+LIBYUV_API
+int I412ToI420(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return I41xToI420(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                    dst_v, dst_stride_v, width, height, 12);
+}
+
+// Any Ix10 To I010 format with mirroring.
+static int Ix10ToI010(const uint16_t* src_y,
+                      int src_stride_y,
+                      const uint16_t* src_u,
+                      int src_stride_u,
+                      const uint16_t* src_v,
+                      int src_stride_v,
+                      uint16_t* dst_y,
+                      int dst_stride_y,
+                      uint16_t* dst_u,
+                      int dst_stride_u,
+                      uint16_t* dst_v,
+                      int dst_stride_v,
+                      int width,
+                      int height,
+                      int subsample_x,
+                      int subsample_y) {
+  const int dst_y_width = Abs(width);
+  const int dst_y_height = Abs(height);
+  const int src_uv_width = SUBSAMPLE(width, subsample_x, subsample_x);
+  const int src_uv_height = SUBSAMPLE(height, subsample_y, subsample_y);
+  const int dst_uv_width = SUBSAMPLE(dst_y_width, 1, 1);
+  const int dst_uv_height = SUBSAMPLE(dst_y_height, 1, 1);
+  int r;
+  if (width <= 0 || height == 0) {
+    return -1;
+  }
+  if (dst_y) {
+    r = ScalePlane_12(src_y, src_stride_y, width, height, dst_y, dst_stride_y,
+                      dst_y_width, dst_y_height, kFilterBilinear);
+    if (r != 0) {
+      return r;
+    }
+  }
+  r = ScalePlane_12(src_u, src_stride_u, src_uv_width, src_uv_height, dst_u,
+                    dst_stride_u, dst_uv_width, dst_uv_height, kFilterBilinear);
+  if (r != 0) {
+    return r;
+  }
+  r = ScalePlane_12(src_v, src_stride_v, src_uv_width, src_uv_height, dst_v,
+                    dst_stride_v, dst_uv_width, dst_uv_height, kFilterBilinear);
+  return r;
+}
+
+LIBYUV_API
+int I410ToI010(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return Ix10ToI010(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                    dst_v, dst_stride_v, width, height, 0, 0);
+}
+
+LIBYUV_API
+int I210ToI010(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return Ix10ToI010(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                    dst_v, dst_stride_v, width, height, 1, 0);
+}
+
+// Any I[420]1[02] to P[420]1[02] format with mirroring.
+static int IxxxToPxxx(const uint16_t* src_y,
+                      int src_stride_y,
+                      const uint16_t* src_u,
+                      int src_stride_u,
+                      const uint16_t* src_v,
+                      int src_stride_v,
+                      uint16_t* dst_y,
+                      int dst_stride_y,
+                      uint16_t* dst_uv,
+                      int dst_stride_uv,
+                      int width,
+                      int height,
+                      int subsample_x,
+                      int subsample_y,
+                      int depth) {
+  const int uv_width = SUBSAMPLE(width, subsample_x, subsample_x);
+  const int uv_height = SUBSAMPLE(height, subsample_y, subsample_y);
+  if (width <= 0 || height == 0) {
+    return -1;
+  }
+
+  ConvertToMSBPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height,
+                       depth);
+  MergeUVPlane_16(src_u, src_stride_u, src_v, src_stride_v, dst_uv,
+                  dst_stride_uv, uv_width, uv_height, depth);
+  return 0;
+}
+
+LIBYUV_API
+int I010ToP010(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  return IxxxToPxxx(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_uv, dst_stride_uv,
+                    width, height, 1, 1, 10);
+}
+
+LIBYUV_API
+int I010ToNV12(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int y;
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  const int scale = 16385;  // 16384 for 10 bits
+  void (*Convert16To8Row)(const uint16_t* src_y, uint8_t* dst_y, int scale,
+                          int width) = Convert16To8Row_C;
+  void (*MergeUVRow)(const uint8_t* src_u, const uint8_t* src_v,
+                      uint8_t* dst_uv, int width) = MergeUVRow_C;
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_uv || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (halfheight - 1) * src_stride_u;
+    src_v = src_v + (halfheight - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+#if defined(HAS_CONVERT16TO8ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    Convert16To8Row = Convert16To8Row_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      Convert16To8Row = Convert16To8Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_CONVERT16TO8ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    Convert16To8Row = Convert16To8Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      Convert16To8Row = Convert16To8Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_CONVERT16TO8ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    Convert16To8Row = Convert16To8Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      Convert16To8Row = Convert16To8Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_CONVERT16TO8ROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW)) {
+    Convert16To8Row = Convert16To8Row_Any_AVX512BW;
+    if (IS_ALIGNED(width, 64)) {
+      Convert16To8Row = Convert16To8Row_AVX512BW;
+    }
+  }
+#endif
+
+#if defined(HAS_MERGEUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    MergeUVRow = MergeUVRow_Any_SSE2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow = MergeUVRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeUVRow = MergeUVRow_Any_AVX2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow = MergeUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW)) {
+    MergeUVRow = MergeUVRow_Any_AVX512BW;
+    if (IS_ALIGNED(halfwidth, 32)) {
+      MergeUVRow = MergeUVRow_AVX512BW;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeUVRow = MergeUVRow_Any_NEON;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow = MergeUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    MergeUVRow = MergeUVRow_Any_MSA;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow = MergeUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    MergeUVRow = MergeUVRow_Any_LSX;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow = MergeUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    MergeUVRow = MergeUVRow_RVV;
+  }
+#endif
+
+  // Convert Y plane.
+  if (dst_y) {
+    Convert16To8Plane(src_y, src_stride_y, dst_y, dst_stride_y, scale, width,
+                      height);
+  }
+
+  {
+    // Allocate a row of uv.
+    align_buffer_64(row_u, ((halfwidth + 31) & ~31) * 2);
+    uint8_t* row_v = row_u + ((halfwidth + 31) & ~31);
+    if (!row_u)
+      return 1;
+
+    for (y = 0; y < halfheight; ++y) {
+      Convert16To8Row(src_u, row_u, scale, halfwidth);
+      Convert16To8Row(src_v, row_v, scale, halfwidth);
+      MergeUVRow(row_u, row_v, dst_uv, halfwidth);
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+      dst_uv += dst_stride_uv;
+    }
+    free_aligned_buffer_64(row_u);
+  }
+  return 0;
+}
+
+LIBYUV_API
+int I210ToP210(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  return IxxxToPxxx(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_uv, dst_stride_uv,
+                    width, height, 1, 0, 10);
+}
+
+LIBYUV_API
+int I012ToP012(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  return IxxxToPxxx(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_uv, dst_stride_uv,
+                    width, height, 1, 1, 12);
+}
+
+LIBYUV_API
+int I212ToP212(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  return IxxxToPxxx(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_uv, dst_stride_uv,
+                    width, height, 1, 0, 12);
+}
+
+// 422 chroma is 1/2 width, 1x height
+// 420 chroma is 1/2 width, 1/2 height
+LIBYUV_API
+int I422ToI420(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  const int src_uv_width = SUBSAMPLE(width, 1, 1);
+  return I4xxToI420(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                    dst_v, dst_stride_v, width, height, src_uv_width, height);
+}
+
+LIBYUV_API
+int I422ToI210(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int halfwidth = (width + 1) >> 1;
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  // Convert Y plane.
+  Convert8To16Plane(src_y, src_stride_y, dst_y, dst_stride_y, 1024, width,
+                    height);
+  // Convert UV planes.
+  Convert8To16Plane(src_u, src_stride_u, dst_u, dst_stride_u, 1024, halfwidth,
+                    height);
+  Convert8To16Plane(src_v, src_stride_v, dst_v, dst_stride_v, 1024, halfwidth,
+                    height);
+  return 0;
+}
+
+// TODO(fbarchard): Implement row conversion.
+LIBYUV_API
+int I422ToNV21(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  // Allocate u and v buffers
+  align_buffer_64(plane_u, halfwidth * halfheight * 2);
+  uint8_t* plane_v = plane_u + halfwidth * halfheight;
+  if (!plane_u)
+    return 1;
+
+  I422ToI420(src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v,
+             dst_y, dst_stride_y, plane_u, halfwidth, plane_v, halfwidth, width,
+             height);
+  MergeUVPlane(plane_v, halfwidth, plane_u, halfwidth, dst_vu, dst_stride_vu,
+               halfwidth, halfheight);
+  free_aligned_buffer_64(plane_u);
+  return 0;
+}
+
+LIBYUV_API
+int MM21ToNV12(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  if (!src_uv || !dst_uv || width <= 0) {
+    return -1;
+  }
+
+  int sign = height < 0 ? -1 : 1;
+
+  if (dst_y) {
+    DetilePlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height, 32);
+  }
+  DetilePlane(src_uv, src_stride_uv, dst_uv, dst_stride_uv, (width + 1) & ~1,
+              (height + sign) / 2, 16);
+
+  return 0;
+}
+
+LIBYUV_API
+int MM21ToI420(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int sign = height < 0 ? -1 : 1;
+
+  if (!src_uv || !dst_u || !dst_v || width <= 0) {
+    return -1;
+  }
+
+  if (dst_y) {
+    DetilePlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height, 32);
+  }
+  DetileSplitUVPlane(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v,
+                     dst_stride_v, (width + 1) & ~1, (height + sign) / 2, 16);
+
+  return 0;
+}
+
+LIBYUV_API
+int MM21ToYUY2(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_yuy2,
+               int dst_stride_yuy2,
+               int width,
+               int height) {
+  if (!src_y || !src_uv || !dst_yuy2 || width <= 0) {
+    return -1;
+  }
+
+  DetileToYUY2(src_y, src_stride_y, src_uv, src_stride_uv, dst_yuy2,
+               dst_stride_yuy2, width, height, 32);
+
+  return 0;
+}
+
+// Convert MT2T into P010. See tinyurl.com/mtk-10bit-video-format for format
+// documentation.
+// TODO(greenjustin): Add an MT2T to I420 conversion.
+LIBYUV_API
+int MT2TToP010(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  if (width <= 0 || !height || !src_uv || !dst_uv) {
+    return -1;
+  }
+
+  {
+    int uv_width = (width + 1) & ~1;
+    int uv_height = (height + 1) / 2;
+    int y = 0;
+    const int tile_width = 16;
+    const int y_tile_height = 32;
+    const int uv_tile_height = 16;
+    int padded_width = (width + tile_width - 1) & ~(tile_width - 1);
+    int y_tile_row_size = padded_width * y_tile_height * 10 / 8;
+    int uv_tile_row_size = padded_width * uv_tile_height * 10 / 8;
+    size_t row_buf_size = padded_width * y_tile_height * sizeof(uint16_t);
+    void (*UnpackMT2T)(const uint8_t* src, uint16_t* dst, size_t size) =
+        UnpackMT2T_C;
+    align_buffer_64(row_buf, row_buf_size);
+    if (!row_buf)
+      return 1;
+
+#if defined(HAS_UNPACKMT2T_NEON)
+    if (TestCpuFlag(kCpuHasNEON)) {
+      UnpackMT2T = UnpackMT2T_NEON;
+    }
+#endif
+    // Negative height means invert the image.
+    if (height < 0) {
+      height = -height;
+      uv_height = (height + 1) / 2;
+      if (dst_y) {
+        dst_y = dst_y + (height - 1) * dst_stride_y;
+        dst_stride_y = -dst_stride_y;
+      }
+      dst_uv = dst_uv + (uv_height - 1) * dst_stride_uv;
+      dst_stride_uv = -dst_stride_uv;
+    }
+
+    // Unpack and detile Y in rows of tiles
+    if (src_y && dst_y) {
+      for (y = 0; y < (height & ~(y_tile_height - 1)); y += y_tile_height) {
+        UnpackMT2T(src_y, (uint16_t*)row_buf, y_tile_row_size);
+        DetilePlane_16((uint16_t*)row_buf, padded_width, dst_y, dst_stride_y,
+                       width, y_tile_height, y_tile_height);
+        src_y += src_stride_y * y_tile_height;
+        dst_y += dst_stride_y * y_tile_height;
+      }
+      if (height & (y_tile_height - 1)) {
+        UnpackMT2T(src_y, (uint16_t*)row_buf, y_tile_row_size);
+        DetilePlane_16((uint16_t*)row_buf, padded_width, dst_y, dst_stride_y,
+                       width, height & (y_tile_height - 1), y_tile_height);
+      }
+    }
+
+    // Unpack and detile UV plane
+    for (y = 0; y < (uv_height & ~(uv_tile_height - 1)); y += uv_tile_height) {
+      UnpackMT2T(src_uv, (uint16_t*)row_buf, uv_tile_row_size);
+      DetilePlane_16((uint16_t*)row_buf, padded_width, dst_uv, dst_stride_uv,
+                     uv_width, uv_tile_height, uv_tile_height);
+      src_uv += src_stride_uv * uv_tile_height;
+      dst_uv += dst_stride_uv * uv_tile_height;
+    }
+    if (uv_height & (uv_tile_height - 1)) {
+      UnpackMT2T(src_uv, (uint16_t*)row_buf, uv_tile_row_size);
+      DetilePlane_16((uint16_t*)row_buf, padded_width, dst_uv, dst_stride_uv,
+                     uv_width, uv_height & (uv_tile_height - 1),
+                     uv_tile_height);
+    }
+    free_aligned_buffer_64(row_buf);
+  }
+  return 0;
+}
+
+#ifdef I422TONV21_ROW_VERSION
+// Unittest fails for this version.
+// 422 chroma is 1/2 width, 1x height
+// 420 chroma is 1/2 width, 1/2 height
+// Swap src_u and src_v to implement I422ToNV12
+LIBYUV_API
+int I422ToNV21(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height) {
+  int y;
+  void (*MergeUVRow)(const uint8_t* src_u, const uint8_t* src_v,
+                     uint8_t* dst_uv, int width) = MergeUVRow_C;
+  void (*InterpolateRow)(uint8_t* dst_ptr, const uint8_t* src_ptr,
+                         ptrdiff_t src_stride, int dst_width,
+                         int source_y_fraction) = InterpolateRow_C;
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  if (!src_u || !src_v || !dst_vu || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (halfheight - 1) * src_stride_u;
+    src_v = src_v + (halfheight - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+#if defined(HAS_MERGEUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    MergeUVRow = MergeUVRow_Any_SSE2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow = MergeUVRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeUVRow = MergeUVRow_Any_AVX2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow = MergeUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW)) {
+    MergeUVRow = MergeUVRow_Any_AVX512BW;
+    if (IS_ALIGNED(halfwidth, 32)) {
+      MergeUVRow = MergeUVRow_AVX512BW;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeUVRow = MergeUVRow_Any_NEON;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow = MergeUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    MergeUVRow = MergeUVRow_Any_MSA;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow = MergeUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    MergeUVRow = MergeUVRow_Any_LSX;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow = MergeUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    MergeUVRow = MergeUVRow_RVV;
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    InterpolateRow = InterpolateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      InterpolateRow = InterpolateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    InterpolateRow = InterpolateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      InterpolateRow = InterpolateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    InterpolateRow = InterpolateRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      InterpolateRow = InterpolateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    InterpolateRow = InterpolateRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      InterpolateRow = InterpolateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    InterpolateRow = InterpolateRow_Any_LSX;
+    if (IS_ALIGNED(width, 32)) {
+      InterpolateRow = InterpolateRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    InterpolateRow = InterpolateRow_RVV;
+  }
+#endif
+
+  if (dst_y) {
+    CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, halfwidth, height);
+  }
+  {
+    // Allocate 2 rows of vu.
+    int awidth = halfwidth * 2;
+    align_buffer_64(row_vu_0, awidth * 2);
+    uint8_t* row_vu_1 = row_vu_0 + awidth;
+    if (!row_vu_0)
+      return 1;
+
+    for (y = 0; y < height - 1; y += 2) {
+      MergeUVRow(src_v, src_u, row_vu_0, halfwidth);
+      MergeUVRow(src_v + src_stride_v, src_u + src_stride_u, row_vu_1,
+                 halfwidth);
+      InterpolateRow(dst_vu, row_vu_0, awidth, awidth, 128);
+      src_u += src_stride_u * 2;
+      src_v += src_stride_v * 2;
+      dst_vu += dst_stride_vu;
+    }
+    if (height & 1) {
+      MergeUVRow(src_v, src_u, dst_vu, halfwidth);
+    }
+    free_aligned_buffer_64(row_vu_0);
+  }
+  return 0;
+}
+#endif  // I422TONV21_ROW_VERSION
+
+// 444 chroma is 1x width, 1x height
+// 420 chroma is 1/2 width, 1/2 height
+LIBYUV_API
+int I444ToI420(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return I4xxToI420(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                    dst_v, dst_stride_v, width, height, width, height);
+}
+
+LIBYUV_API
+int I444ToNV12(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  if (!src_y || !src_u || !src_v || !dst_uv || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+  if (dst_y) {
+    CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+  HalfMergeUVPlane(src_u, src_stride_u, src_v, src_stride_v, dst_uv,
+                   dst_stride_uv, width, height);
+  return 0;
+}
+
+LIBYUV_API
+int I444ToNV21(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height) {
+  return I444ToNV12(src_y, src_stride_y, src_v, src_stride_v, src_u,
+                    src_stride_u, dst_y, dst_stride_y, dst_vu, dst_stride_vu,
+                    width, height);
+}
+
+// I400 is greyscale typically used in MJPG
+LIBYUV_API
+int I400ToI420(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  if (!dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_stride_y = -src_stride_y;
+  }
+  if (dst_y) {
+    CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+  SetPlane(dst_u, dst_stride_u, halfwidth, halfheight, 128);
+  SetPlane(dst_v, dst_stride_v, halfwidth, halfheight, 128);
+  return 0;
+}
+
+// I400 is greyscale typically used in MJPG
+LIBYUV_API
+int I400ToNV21(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  if (!dst_vu || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_stride_y = -src_stride_y;
+  }
+  if (dst_y) {
+    CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+  SetPlane(dst_vu, dst_stride_vu, halfwidth * 2, halfheight, 128);
+  return 0;
+}
+
+// Convert NV12 to I420.
+// TODO(fbarchard): Consider inverting destination. Faster on ARM with prfm.
+LIBYUV_API
+int NV12ToI420(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  if (!src_uv || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_uv = src_uv + (halfheight - 1) * src_stride_uv;
+    src_stride_y = -src_stride_y;
+    src_stride_uv = -src_stride_uv;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && dst_stride_y == width) {
+    width *= height;
+    height = 1;
+    src_stride_y = dst_stride_y = 0;
+  }
+  // Coalesce rows.
+  if (src_stride_uv == halfwidth * 2 && dst_stride_u == halfwidth &&
+      dst_stride_v == halfwidth) {
+    halfwidth *= halfheight;
+    halfheight = 1;
+    src_stride_uv = dst_stride_u = dst_stride_v = 0;
+  }
+
+  if (dst_y) {
+    CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+
+  // Split UV plane - NV12 / NV21
+  SplitUVPlane(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v, dst_stride_v,
+               halfwidth, halfheight);
+
+  return 0;
+}
+
+// Convert NV21 to I420.  Same as NV12 but u and v pointers swapped.
+LIBYUV_API
+int NV21ToI420(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_vu,
+               int src_stride_vu,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return NV12ToI420(src_y, src_stride_y, src_vu, src_stride_vu, dst_y,
+                    dst_stride_y, dst_v, dst_stride_v, dst_u, dst_stride_u,
+                    width, height);
+}
+
+LIBYUV_API
+int NV12ToNV24(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int r;
+  if (width <= 0 || height == 0) {
+    return -1;
+  }
+
+  if (dst_y) {
+    r = ScalePlane(src_y, src_stride_y, width, height, dst_y, dst_stride_y,
+                   Abs(width), Abs(height), kFilterBilinear);
+    if (r != 0) {
+      return r;
+    }
+  }
+  r = UVScale(src_uv, src_stride_uv, SUBSAMPLE(width, 1, 1),
+              SUBSAMPLE(height, 1, 1), dst_uv, dst_stride_uv, Abs(width),
+              Abs(height), kFilterBilinear);
+  return r;
+}
+
+LIBYUV_API
+int NV16ToNV24(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int r;
+  if (width <= 0 || height == 0) {
+    return -1;
+  }
+
+  if (dst_y) {
+    r = ScalePlane(src_y, src_stride_y, width, height, dst_y, dst_stride_y,
+                   Abs(width), Abs(height), kFilterBilinear);
+    if (r != 0) {
+      return r;
+    }
+  }
+  r = UVScale(src_uv, src_stride_uv, SUBSAMPLE(width, 1, 1), height, dst_uv,
+              dst_stride_uv, Abs(width), Abs(height), kFilterBilinear);
+  return r;
+}
+
+// Any P[420]1[02] to I[420]1[02] format with mirroring.
+static int PxxxToIxxx(const uint16_t* src_y,
+                      int src_stride_y,
+                      const uint16_t* src_uv,
+                      int src_stride_uv,
+                      uint16_t* dst_y,
+                      int dst_stride_y,
+                      uint16_t* dst_u,
+                      int dst_stride_u,
+                      uint16_t* dst_v,
+                      int dst_stride_v,
+                      int width,
+                      int height,
+                      int subsample_x,
+                      int subsample_y,
+                      int depth) {
+  const int uv_width = SUBSAMPLE(width, subsample_x, subsample_x);
+  const int uv_height = SUBSAMPLE(height, subsample_y, subsample_y);
+  if (width <= 0 || height == 0) {
+    return -1;
+  }
+  ConvertToLSBPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height,
+                       depth);
+  SplitUVPlane_16(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v,
+                  dst_stride_v, uv_width, uv_height, depth);
+  return 0;
+}
+
+LIBYUV_API
+int P010ToI010(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_uv,
+               int src_stride_uv,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return PxxxToIxxx(src_y, src_stride_y, src_uv, src_stride_uv, dst_y,
+                    dst_stride_y, dst_u, dst_stride_u, dst_v, dst_stride_v,
+                    width, height, 1, 1, 10);
+}
+
+LIBYUV_API
+int P012ToI012(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_uv,
+               int src_stride_uv,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  return PxxxToIxxx(src_y, src_stride_y, src_uv, src_stride_uv, dst_y,
+                    dst_stride_y, dst_u, dst_stride_u, dst_v, dst_stride_v,
+                    width, height, 1, 1, 12);
+}
+
+LIBYUV_API
+int P010ToP410(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_uv,
+               int src_stride_uv,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int r;
+  if (width <= 0 || height == 0) {
+    return -1;
+  }
+
+  if (dst_y) {
+    r = ScalePlane_16(src_y, src_stride_y, width, height, dst_y, dst_stride_y,
+                      Abs(width), Abs(height), kFilterBilinear);
+    if (r != 0) {
+      return r;
+    }
+  }
+  r = UVScale_16(src_uv, src_stride_uv, SUBSAMPLE(width, 1, 1),
+                 SUBSAMPLE(height, 1, 1), dst_uv, dst_stride_uv, Abs(width),
+                 Abs(height), kFilterBilinear);
+  return r;
+}
+
+LIBYUV_API
+int P210ToP410(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_uv,
+               int src_stride_uv,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int r;
+  if (width <= 0 || height == 0) {
+    return -1;
+  }
+
+  if (dst_y) {
+    r = ScalePlane_16(src_y, src_stride_y, width, height, dst_y, dst_stride_y,
+                      Abs(width), Abs(height), kFilterBilinear);
+    if (r != 0) {
+      return r;
+    }
+  }
+  r = UVScale_16(src_uv, src_stride_uv, SUBSAMPLE(width, 1, 1), height, dst_uv,
+                 dst_stride_uv, Abs(width), Abs(height), kFilterBilinear);
+  return r;
+}
+
+// Convert YUY2 to I420.
+LIBYUV_API
+int YUY2ToI420(const uint8_t* src_yuy2,
+               int src_stride_yuy2,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int y;
+  void (*YUY2ToUVRow)(const uint8_t* src_yuy2, int src_stride_yuy2,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      YUY2ToUVRow_C;
+  void (*YUY2ToYRow)(const uint8_t* src_yuy2, uint8_t* dst_y, int width) =
+      YUY2ToYRow_C;
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_yuy2 = src_yuy2 + (height - 1) * src_stride_yuy2;
+    src_stride_yuy2 = -src_stride_yuy2;
+  }
+#if defined(HAS_YUY2TOYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    YUY2ToUVRow = YUY2ToUVRow_Any_SSE2;
+    YUY2ToYRow = YUY2ToYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToUVRow = YUY2ToUVRow_SSE2;
+      YUY2ToYRow = YUY2ToYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    YUY2ToUVRow = YUY2ToUVRow_Any_AVX2;
+    YUY2ToYRow = YUY2ToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToUVRow = YUY2ToUVRow_AVX2;
+      YUY2ToYRow = YUY2ToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    YUY2ToYRow = YUY2ToYRow_Any_NEON;
+    YUY2ToUVRow = YUY2ToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToYRow = YUY2ToYRow_NEON;
+      YUY2ToUVRow = YUY2ToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_MSA) && defined(HAS_YUY2TOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    YUY2ToYRow = YUY2ToYRow_Any_MSA;
+    YUY2ToUVRow = YUY2ToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToYRow = YUY2ToYRow_MSA;
+      YUY2ToUVRow = YUY2ToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_LSX) && defined(HAS_YUY2TOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    YUY2ToYRow = YUY2ToYRow_Any_LSX;
+    YUY2ToUVRow = YUY2ToUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToYRow = YUY2ToYRow_LSX;
+      YUY2ToUVRow = YUY2ToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_LASX) && defined(HAS_YUY2TOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    YUY2ToYRow = YUY2ToYRow_Any_LASX;
+    YUY2ToUVRow = YUY2ToUVRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToYRow = YUY2ToYRow_LASX;
+      YUY2ToUVRow = YUY2ToUVRow_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    YUY2ToUVRow(src_yuy2, src_stride_yuy2, dst_u, dst_v, width);
+    YUY2ToYRow(src_yuy2, dst_y, width);
+    YUY2ToYRow(src_yuy2 + src_stride_yuy2, dst_y + dst_stride_y, width);
+    src_yuy2 += src_stride_yuy2 * 2;
+    dst_y += dst_stride_y * 2;
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+  }
+  if (height & 1) {
+    YUY2ToUVRow(src_yuy2, 0, dst_u, dst_v, width);
+    YUY2ToYRow(src_yuy2, dst_y, width);
+  }
+  return 0;
+}
+
+// Convert UYVY to I420.
+LIBYUV_API
+int UYVYToI420(const uint8_t* src_uyvy,
+               int src_stride_uyvy,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int y;
+  void (*UYVYToUVRow)(const uint8_t* src_uyvy, int src_stride_uyvy,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      UYVYToUVRow_C;
+  void (*UYVYToYRow)(const uint8_t* src_uyvy, uint8_t* dst_y, int width) =
+      UYVYToYRow_C;
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_uyvy = src_uyvy + (height - 1) * src_stride_uyvy;
+    src_stride_uyvy = -src_stride_uyvy;
+  }
+#if defined(HAS_UYVYTOYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    UYVYToUVRow = UYVYToUVRow_Any_SSE2;
+    UYVYToYRow = UYVYToYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      UYVYToUVRow = UYVYToUVRow_SSE2;
+      UYVYToYRow = UYVYToYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    UYVYToUVRow = UYVYToUVRow_Any_AVX2;
+    UYVYToYRow = UYVYToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      UYVYToUVRow = UYVYToUVRow_AVX2;
+      UYVYToYRow = UYVYToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    UYVYToYRow = UYVYToYRow_Any_NEON;
+    UYVYToUVRow = UYVYToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      UYVYToYRow = UYVYToYRow_NEON;
+      UYVYToUVRow = UYVYToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    UYVYToYRow = UYVYToYRow_Any_MSA;
+    UYVYToUVRow = UYVYToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      UYVYToYRow = UYVYToYRow_MSA;
+      UYVYToUVRow = UYVYToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    UYVYToYRow = UYVYToYRow_Any_LSX;
+    UYVYToUVRow = UYVYToUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      UYVYToYRow = UYVYToYRow_LSX;
+      UYVYToUVRow = UYVYToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    UYVYToYRow = UYVYToYRow_Any_LSX;
+    UYVYToUVRow = UYVYToUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      UYVYToYRow = UYVYToYRow_LSX;
+      UYVYToUVRow = UYVYToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    UYVYToYRow = UYVYToYRow_Any_LASX;
+    UYVYToUVRow = UYVYToUVRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      UYVYToYRow = UYVYToYRow_LASX;
+      UYVYToUVRow = UYVYToUVRow_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    UYVYToUVRow(src_uyvy, src_stride_uyvy, dst_u, dst_v, width);
+    UYVYToYRow(src_uyvy, dst_y, width);
+    UYVYToYRow(src_uyvy + src_stride_uyvy, dst_y + dst_stride_y, width);
+    src_uyvy += src_stride_uyvy * 2;
+    dst_y += dst_stride_y * 2;
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+  }
+  if (height & 1) {
+    UYVYToUVRow(src_uyvy, 0, dst_u, dst_v, width);
+    UYVYToYRow(src_uyvy, dst_y, width);
+  }
+  return 0;
+}
+
+// Convert AYUV to NV12.
+LIBYUV_API
+int AYUVToNV12(const uint8_t* src_ayuv,
+               int src_stride_ayuv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int y;
+  void (*AYUVToUVRow)(const uint8_t* src_ayuv, int src_stride_ayuv,
+                      uint8_t* dst_uv, int width) = AYUVToUVRow_C;
+  void (*AYUVToYRow)(const uint8_t* src_ayuv, uint8_t* dst_y, int width) =
+      AYUVToYRow_C;
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_ayuv = src_ayuv + (height - 1) * src_stride_ayuv;
+    src_stride_ayuv = -src_stride_ayuv;
+  }
+// place holders for future intel code
+#if defined(HAS_AYUVTOYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    AYUVToUVRow = AYUVToUVRow_Any_SSE2;
+    AYUVToYRow = AYUVToYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      AYUVToUVRow = AYUVToUVRow_SSE2;
+      AYUVToYRow = AYUVToYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_AYUVTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    AYUVToUVRow = AYUVToUVRow_Any_AVX2;
+    AYUVToYRow = AYUVToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      AYUVToUVRow = AYUVToUVRow_AVX2;
+      AYUVToYRow = AYUVToYRow_AVX2;
+    }
+  }
+#endif
+
+#if defined(HAS_AYUVTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    AYUVToYRow = AYUVToYRow_Any_NEON;
+    AYUVToUVRow = AYUVToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      AYUVToYRow = AYUVToYRow_NEON;
+      AYUVToUVRow = AYUVToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_AYUVTOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    AYUVToUVRow = AYUVToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      AYUVToUVRow = AYUVToUVRow_SVE2;
+    }
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    AYUVToUVRow(src_ayuv, src_stride_ayuv, dst_uv, width);
+    AYUVToYRow(src_ayuv, dst_y, width);
+    AYUVToYRow(src_ayuv + src_stride_ayuv, dst_y + dst_stride_y, width);
+    src_ayuv += src_stride_ayuv * 2;
+    dst_y += dst_stride_y * 2;
+    dst_uv += dst_stride_uv;
+  }
+  if (height & 1) {
+    AYUVToUVRow(src_ayuv, 0, dst_uv, width);
+    AYUVToYRow(src_ayuv, dst_y, width);
+  }
+  return 0;
+}
+
+// Convert AYUV to NV21.
+LIBYUV_API
+int AYUVToNV21(const uint8_t* src_ayuv,
+               int src_stride_ayuv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height) {
+  int y;
+  void (*AYUVToVURow)(const uint8_t* src_ayuv, int src_stride_ayuv,
+                      uint8_t* dst_vu, int width) = AYUVToVURow_C;
+  void (*AYUVToYRow)(const uint8_t* src_ayuv, uint8_t* dst_y, int width) =
+      AYUVToYRow_C;
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_ayuv = src_ayuv + (height - 1) * src_stride_ayuv;
+    src_stride_ayuv = -src_stride_ayuv;
+  }
+// place holders for future intel code
+#if defined(HAS_AYUVTOYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    AYUVToVURow = AYUVToVURow_Any_SSE2;
+    AYUVToYRow = AYUVToYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      AYUVToVURow = AYUVToVURow_SSE2;
+      AYUVToYRow = AYUVToYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_AYUVTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    AYUVToVURow = AYUVToVURow_Any_AVX2;
+    AYUVToYRow = AYUVToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      AYUVToVURow = AYUVToVURow_AVX2;
+      AYUVToYRow = AYUVToYRow_AVX2;
+    }
+  }
+#endif
+
+#if defined(HAS_AYUVTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    AYUVToYRow = AYUVToYRow_Any_NEON;
+    AYUVToVURow = AYUVToVURow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      AYUVToYRow = AYUVToYRow_NEON;
+      AYUVToVURow = AYUVToVURow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_AYUVTOVUROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    AYUVToVURow = AYUVToVURow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      AYUVToVURow = AYUVToVURow_SVE2;
+    }
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    AYUVToVURow(src_ayuv, src_stride_ayuv, dst_vu, width);
+    AYUVToYRow(src_ayuv, dst_y, width);
+    AYUVToYRow(src_ayuv + src_stride_ayuv, dst_y + dst_stride_y, width);
+    src_ayuv += src_stride_ayuv * 2;
+    dst_y += dst_stride_y * 2;
+    dst_vu += dst_stride_vu;
+  }
+  if (height & 1) {
+    AYUVToVURow(src_ayuv, 0, dst_vu, width);
+    AYUVToYRow(src_ayuv, dst_y, width);
+  }
+  return 0;
+}
+
+// Convert ARGB to I420.
+LIBYUV_API
+int ARGBToI420(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVRow_C;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+  if (!src_argb || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+#if defined(HAS_ARGBTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToUVRow = ARGBToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ARGBToUVRow = ARGBToUVRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYRow = ARGBToYRow_Any_MSA;
+    ARGBToUVRow = ARGBToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    ARGBToUVRow = ARGBToUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+      ARGBToUVRow = ARGBToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYRow = ARGBToYRow_Any_LASX;
+    ARGBToUVRow = ARGBToUVRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_LASX;
+      ARGBToUVRow = ARGBToUVRow_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    ARGBToUVRow(src_argb, src_stride_argb, dst_u, dst_v, width);
+    ARGBToYRow(src_argb, dst_y, width);
+    ARGBToYRow(src_argb + src_stride_argb, dst_y + dst_stride_y, width);
+    src_argb += src_stride_argb * 2;
+    dst_y += dst_stride_y * 2;
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+  }
+  if (height & 1) {
+    ARGBToUVRow(src_argb, 0, dst_u, dst_v, width);
+    ARGBToYRow(src_argb, dst_y, width);
+  }
+  return 0;
+}
+
+#ifdef USE_EXTRACTALPHA
+// Convert ARGB to I420 with Alpha
+// The following version calls ARGBExtractAlpha on the full image.
+LIBYUV_API
+int ARGBToI420Alpha(const uint8_t* src_argb,
+                    int src_stride_argb,
+                    uint8_t* dst_y,
+                    int dst_stride_y,
+                    uint8_t* dst_u,
+                    int dst_stride_u,
+                    uint8_t* dst_v,
+                    int dst_stride_v,
+                    uint8_t* dst_a,
+                    int dst_stride_a,
+                    int width,
+                    int height) {
+  int r = ARGBToI420(src_argb, src_stride_argb, dst_y, dst_stride_y, dst_u,
+                     dst_stride_u, dst_v, dst_stride_v, width, height);
+  if (r == 0) {
+    r = ARGBExtractAlpha(src_argb, src_stride_argb, dst_a, dst_stride_a, width,
+                         height);
+  }
+  return r;
+}
+#else  // USE_EXTRACTALPHA
+// Convert ARGB to I420 with Alpha
+LIBYUV_API
+int ARGBToI420Alpha(const uint8_t* src_argb,
+                    int src_stride_argb,
+                    uint8_t* dst_y,
+                    int dst_stride_y,
+                    uint8_t* dst_u,
+                    int dst_stride_u,
+                    uint8_t* dst_v,
+                    int dst_stride_v,
+                    uint8_t* dst_a,
+                    int dst_stride_a,
+                    int width,
+                    int height) {
+  int y;
+  void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVRow_C;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+  void (*ARGBExtractAlphaRow)(const uint8_t* src_argb, uint8_t* dst_a,
+                              int width) = ARGBExtractAlphaRow_C;
+  if (!src_argb || !dst_y || !dst_u || !dst_v || !dst_a || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+#if defined(HAS_ARGBTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToUVRow = ARGBToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ARGBToUVRow = ARGBToUVRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYRow = ARGBToYRow_Any_MSA;
+    ARGBToUVRow = ARGBToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYRow = ARGBToYRow_Any_LASX;
+    ARGBToUVRow = ARGBToUVRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_LASX;
+      ARGBToUVRow = ARGBToUVRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBEXTRACTALPHAROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBExtractAlphaRow = IS_ALIGNED(width, 8) ? ARGBExtractAlphaRow_SSE2
+                                               : ARGBExtractAlphaRow_Any_SSE2;
+  }
+#endif
+#if defined(HAS_ARGBEXTRACTALPHAROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBExtractAlphaRow = IS_ALIGNED(width, 32) ? ARGBExtractAlphaRow_AVX2
+                                                : ARGBExtractAlphaRow_Any_AVX2;
+  }
+#endif
+#if defined(HAS_ARGBEXTRACTALPHAROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_NEON
+                                                : ARGBExtractAlphaRow_Any_NEON;
+  }
+#endif
+#if defined(HAS_ARGBEXTRACTALPHAROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_MSA
+                                                : ARGBExtractAlphaRow_Any_MSA;
+  }
+#endif
+#if defined(HAS_ARGBEXTRACTALPHAROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_LSX
+                                                : ARGBExtractAlphaRow_Any_LSX;
+  }
+#endif
+#if defined(HAS_ARGBEXTRACTALPHAROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBExtractAlphaRow = ARGBExtractAlphaRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    ARGBToUVRow(src_argb, src_stride_argb, dst_u, dst_v, width);
+    ARGBToYRow(src_argb, dst_y, width);
+    ARGBToYRow(src_argb + src_stride_argb, dst_y + dst_stride_y, width);
+    ARGBExtractAlphaRow(src_argb, dst_a, width);
+    ARGBExtractAlphaRow(src_argb + src_stride_argb, dst_a + dst_stride_a,
+                        width);
+    src_argb += src_stride_argb * 2;
+    dst_y += dst_stride_y * 2;
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+    dst_a += dst_stride_a * 2;
+  }
+  if (height & 1) {
+    ARGBToUVRow(src_argb, 0, dst_u, dst_v, width);
+    ARGBToYRow(src_argb, dst_y, width);
+    ARGBExtractAlphaRow(src_argb, dst_a, width);
+  }
+  return 0;
+}
+#endif  // USE_EXTRACTALPHA
+
+// Convert BGRA to I420.
+LIBYUV_API
+int BGRAToI420(const uint8_t* src_bgra,
+               int src_stride_bgra,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int y;
+  void (*BGRAToUVRow)(const uint8_t* src_bgra0, int src_stride_bgra,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      BGRAToUVRow_C;
+  void (*BGRAToYRow)(const uint8_t* src_bgra, uint8_t* dst_y, int width) =
+      BGRAToYRow_C;
+  if (!src_bgra || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_bgra = src_bgra + (height - 1) * src_stride_bgra;
+    src_stride_bgra = -src_stride_bgra;
+  }
+#if defined(HAS_BGRATOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    BGRAToYRow = BGRAToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      BGRAToYRow = BGRAToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_BGRATOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    BGRAToYRow = BGRAToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      BGRAToYRow = BGRAToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_BGRATOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    BGRAToUVRow = BGRAToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      BGRAToUVRow = BGRAToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_BGRATOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    BGRAToUVRow = BGRAToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      BGRAToUVRow = BGRAToUVRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_BGRATOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    BGRAToYRow = BGRAToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      BGRAToYRow = BGRAToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_BGRATOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    BGRAToUVRow = BGRAToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      BGRAToUVRow = BGRAToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_BGRATOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    BGRAToYRow = BGRAToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      BGRAToYRow = BGRAToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_BGRATOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    BGRAToUVRow = BGRAToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      BGRAToUVRow = BGRAToUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_BGRATOYROW_MSA) && defined(HAS_BGRATOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    BGRAToYRow = BGRAToYRow_Any_MSA;
+    BGRAToUVRow = BGRAToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      BGRAToYRow = BGRAToYRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      BGRAToUVRow = BGRAToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_BGRATOYROW_LSX) && defined(HAS_BGRATOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    BGRAToYRow = BGRAToYRow_Any_LSX;
+    BGRAToUVRow = BGRAToUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      BGRAToYRow = BGRAToYRow_LSX;
+      BGRAToUVRow = BGRAToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_BGRATOYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    BGRAToYRow = BGRAToYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      BGRAToYRow = BGRAToYRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_BGRATOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    BGRAToYRow = BGRAToYRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    BGRAToUVRow(src_bgra, src_stride_bgra, dst_u, dst_v, width);
+    BGRAToYRow(src_bgra, dst_y, width);
+    BGRAToYRow(src_bgra + src_stride_bgra, dst_y + dst_stride_y, width);
+    src_bgra += src_stride_bgra * 2;
+    dst_y += dst_stride_y * 2;
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+  }
+  if (height & 1) {
+    BGRAToUVRow(src_bgra, 0, dst_u, dst_v, width);
+    BGRAToYRow(src_bgra, dst_y, width);
+  }
+  return 0;
+}
+
+// Convert ABGR to I420.
+LIBYUV_API
+int ABGRToI420(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int y;
+  void (*ABGRToUVRow)(const uint8_t* src_abgr0, int src_stride_abgr,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ABGRToUVRow_C;
+  void (*ABGRToYRow)(const uint8_t* src_abgr, uint8_t* dst_y, int width) =
+      ABGRToYRow_C;
+  if (!src_abgr || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_abgr = src_abgr + (height - 1) * src_stride_abgr;
+    src_stride_abgr = -src_stride_abgr;
+  }
+#if defined(HAS_ABGRTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ABGRToYRow = ABGRToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ABGRToUVRow = ABGRToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToUVRow = ABGRToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ABGRToYRow = ABGRToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToYRow = ABGRToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ABGRToUVRow = ABGRToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToUVRow = ABGRToUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ABGRToYRow = ABGRToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ABGRToYRow = ABGRToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ABGRToUVRow = ABGRToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToUVRow = ABGRToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ABGRToUVRow = ABGRToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ABGRToUVRow = ABGRToUVRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_MSA) && defined(HAS_ABGRTOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ABGRToYRow = ABGRToYRow_Any_MSA;
+    ABGRToUVRow = ABGRToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_MSA;
+      ABGRToUVRow = ABGRToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_LSX) && defined(HAS_ABGRTOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ABGRToYRow = ABGRToYRow_Any_LSX;
+    ABGRToUVRow = ABGRToUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_LSX;
+      ABGRToUVRow = ABGRToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ABGRToYRow = ABGRToYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToYRow = ABGRToYRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ABGRToYRow = ABGRToYRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    ABGRToUVRow(src_abgr, src_stride_abgr, dst_u, dst_v, width);
+    ABGRToYRow(src_abgr, dst_y, width);
+    ABGRToYRow(src_abgr + src_stride_abgr, dst_y + dst_stride_y, width);
+    src_abgr += src_stride_abgr * 2;
+    dst_y += dst_stride_y * 2;
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+  }
+  if (height & 1) {
+    ABGRToUVRow(src_abgr, 0, dst_u, dst_v, width);
+    ABGRToYRow(src_abgr, dst_y, width);
+  }
+  return 0;
+}
+
+// Convert RGBA to I420.
+LIBYUV_API
+int RGBAToI420(const uint8_t* src_rgba,
+               int src_stride_rgba,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int y;
+  void (*RGBAToUVRow)(const uint8_t* src_rgba0, int src_stride_rgba,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      RGBAToUVRow_C;
+  void (*RGBAToYRow)(const uint8_t* src_rgba, uint8_t* dst_y, int width) =
+      RGBAToYRow_C;
+  if (!src_rgba || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_rgba = src_rgba + (height - 1) * src_stride_rgba;
+    src_stride_rgba = -src_stride_rgba;
+  }
+#if defined(HAS_RGBATOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RGBAToYRow = RGBAToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToYRow = RGBAToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RGBAToUVRow = RGBAToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToUVRow = RGBAToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RGBAToYRow = RGBAToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToYRow = RGBAToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    RGBAToYRow = RGBAToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToYRow = RGBAToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RGBAToUVRow = RGBAToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToUVRow = RGBAToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    RGBAToUVRow = RGBAToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      RGBAToUVRow = RGBAToUVRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYROW_MSA) && defined(HAS_RGBATOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RGBAToYRow = RGBAToYRow_Any_MSA;
+    RGBAToUVRow = RGBAToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToYRow = RGBAToYRow_MSA;
+      RGBAToUVRow = RGBAToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYROW_LSX) && defined(HAS_RGBATOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RGBAToYRow = RGBAToYRow_Any_LSX;
+    RGBAToUVRow = RGBAToUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToYRow = RGBAToYRow_LSX;
+      RGBAToUVRow = RGBAToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYROW_LASX)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RGBAToYRow = RGBAToYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RGBAToYRow = RGBAToYRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RGBAToYRow = RGBAToYRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    RGBAToUVRow(src_rgba, src_stride_rgba, dst_u, dst_v, width);
+    RGBAToYRow(src_rgba, dst_y, width);
+    RGBAToYRow(src_rgba + src_stride_rgba, dst_y + dst_stride_y, width);
+    src_rgba += src_stride_rgba * 2;
+    dst_y += dst_stride_y * 2;
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+  }
+  if (height & 1) {
+    RGBAToUVRow(src_rgba, 0, dst_u, dst_v, width);
+    RGBAToYRow(src_rgba, dst_y, width);
+  }
+  return 0;
+}
+
+// Enabled if 1 pass is available
+#if (defined(HAS_RGB24TOYROW_NEON) || defined(HAS_RGB24TOYROW_MSA) || \
+     defined(HAS_RGB24TOYROW_LSX) || defined(HAS_RGB24TOYROW_RVV))
+#define HAS_RGB24TOYROW
+#endif
+
+// Convert RGB24 to I420.
+LIBYUV_API
+int RGB24ToI420(const uint8_t* src_rgb24,
+                int src_stride_rgb24,
+                uint8_t* dst_y,
+                int dst_stride_y,
+                uint8_t* dst_u,
+                int dst_stride_u,
+                uint8_t* dst_v,
+                int dst_stride_v,
+                int width,
+                int height) {
+  int y;
+#if defined(HAS_RGB24TOYROW)
+  void (*RGB24ToUVRow)(const uint8_t* src_rgb24, int src_stride_rgb24,
+                       uint8_t* dst_u, uint8_t* dst_v, int width) =
+      RGB24ToUVRow_C;
+  void (*RGB24ToYRow)(const uint8_t* src_rgb24, uint8_t* dst_y, int width) =
+      RGB24ToYRow_C;
+#else
+  void (*RGB24ToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) =
+      RGB24ToARGBRow_C;
+  void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVRow_C;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+#endif
+  if (!src_rgb24 || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_rgb24 = src_rgb24 + (height - 1) * src_stride_rgb24;
+    src_stride_rgb24 = -src_stride_rgb24;
+  }
+
+#if defined(HAS_RGB24TOYROW)
+
+// Neon version does direct RGB24 to YUV.
+#if defined(HAS_RGB24TOYROW_NEON) && defined(HAS_RGB24TOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RGB24ToUVRow = RGB24ToUVRow_Any_NEON;
+    RGB24ToYRow = RGB24ToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToYRow = RGB24ToYRow_NEON;
+      RGB24ToUVRow = RGB24ToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYROW_MSA) && defined(HAS_RGB24TOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RGB24ToUVRow = RGB24ToUVRow_Any_MSA;
+    RGB24ToYRow = RGB24ToYRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToYRow = RGB24ToYRow_MSA;
+      RGB24ToUVRow = RGB24ToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYROW_LSX) && defined(HAS_RGB24TOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RGB24ToUVRow = RGB24ToUVRow_Any_LSX;
+    RGB24ToYRow = RGB24ToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToYRow = RGB24ToYRow_LSX;
+      RGB24ToUVRow = RGB24ToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYROW_LASX) && defined(HAS_RGB24TOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    RGB24ToUVRow = RGB24ToUVRow_Any_LASX;
+    RGB24ToYRow = RGB24ToYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RGB24ToYRow = RGB24ToYRow_LASX;
+      RGB24ToUVRow = RGB24ToUVRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RGB24ToYRow = RGB24ToYRow_RVV;
+  }
+#endif
+
+// Other platforms do intermediate conversion from RGB24 to ARGB.
+#else  // HAS_RGB24TOYROW
+
+#if defined(HAS_RGB24TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RGB24ToARGBRow = RGB24ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToARGBRow = RGB24ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_AVX2;
+    }
+  }
+#endif
+#endif  // HAS_RGB24TOYROW
+
+  {
+#if !defined(HAS_RGB24TOYROW)
+    // Allocate 2 rows of ARGB.
+    const int row_size = (width * 4 + 31) & ~31;
+    align_buffer_64(row, row_size * 2);
+    if (!row)
+      return 1;
+#endif
+
+    for (y = 0; y < height - 1; y += 2) {
+#if defined(HAS_RGB24TOYROW)
+      RGB24ToUVRow(src_rgb24, src_stride_rgb24, dst_u, dst_v, width);
+      RGB24ToYRow(src_rgb24, dst_y, width);
+      RGB24ToYRow(src_rgb24 + src_stride_rgb24, dst_y + dst_stride_y, width);
+#else
+      RGB24ToARGBRow(src_rgb24, row, width);
+      RGB24ToARGBRow(src_rgb24 + src_stride_rgb24, row + row_size, width);
+      ARGBToUVRow(row, row_size, dst_u, dst_v, width);
+      ARGBToYRow(row, dst_y, width);
+      ARGBToYRow(row + row_size, dst_y + dst_stride_y, width);
+#endif
+      src_rgb24 += src_stride_rgb24 * 2;
+      dst_y += dst_stride_y * 2;
+      dst_u += dst_stride_u;
+      dst_v += dst_stride_v;
+    }
+    if (height & 1) {
+#if defined(HAS_RGB24TOYROW)
+      RGB24ToUVRow(src_rgb24, 0, dst_u, dst_v, width);
+      RGB24ToYRow(src_rgb24, dst_y, width);
+#else
+      RGB24ToARGBRow(src_rgb24, row, width);
+      ARGBToUVRow(row, 0, dst_u, dst_v, width);
+      ARGBToYRow(row, dst_y, width);
+#endif
+    }
+#if !defined(HAS_RGB24TOYROW)
+    free_aligned_buffer_64(row);
+#endif
+  }
+  return 0;
+}
+#undef HAS_RGB24TOYROW
+
+// Enabled if 1 pass is available
+#if defined(HAS_RGB24TOYJROW_NEON) || defined(HAS_RGB24TOYJROW_MSA) || \
+    defined(HAS_RGB24TOYJROW_RVV)
+#define HAS_RGB24TOYJROW
+#endif
+
+// Convert RGB24 to J420.
+LIBYUV_API
+int RGB24ToJ420(const uint8_t* src_rgb24,
+                int src_stride_rgb24,
+                uint8_t* dst_y,
+                int dst_stride_y,
+                uint8_t* dst_u,
+                int dst_stride_u,
+                uint8_t* dst_v,
+                int dst_stride_v,
+                int width,
+                int height) {
+  int y;
+#if defined(HAS_RGB24TOYJROW)
+  void (*RGB24ToUVJRow)(const uint8_t* src_rgb24, int src_stride_rgb24,
+                        uint8_t* dst_u, uint8_t* dst_v, int width) =
+      RGB24ToUVJRow_C;
+  void (*RGB24ToYJRow)(const uint8_t* src_rgb24, uint8_t* dst_y, int width) =
+      RGB24ToYJRow_C;
+#else
+  void (*RGB24ToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) =
+      RGB24ToARGBRow_C;
+  void (*ARGBToUVJRow)(const uint8_t* src_argb0, int src_stride_argb,
+                       uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVJRow_C;
+  void (*ARGBToYJRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYJRow_C;
+#endif
+  if (!src_rgb24 || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_rgb24 = src_rgb24 + (height - 1) * src_stride_rgb24;
+    src_stride_rgb24 = -src_stride_rgb24;
+  }
+
+#if defined(HAS_RGB24TOYJROW)
+
+// Neon version does direct RGB24 to YUV.
+#if defined(HAS_RGB24TOYJROW_NEON) && defined(HAS_RGB24TOUVJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RGB24ToUVJRow = RGB24ToUVJRow_Any_NEON;
+    RGB24ToYJRow = RGB24ToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToYJRow = RGB24ToYJRow_NEON;
+      RGB24ToUVJRow = RGB24ToUVJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYJROW_MSA) && defined(HAS_RGB24TOUVJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RGB24ToUVJRow = RGB24ToUVJRow_Any_MSA;
+    RGB24ToYJRow = RGB24ToYJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToYJRow = RGB24ToYJRow_MSA;
+      RGB24ToUVJRow = RGB24ToUVJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYJROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RGB24ToYJRow = RGB24ToYJRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToYJRow = RGB24ToYJRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYJROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    RGB24ToYJRow = RGB24ToYJRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RGB24ToYJRow = RGB24ToYJRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RGB24ToYJRow = RGB24ToYJRow_RVV;
+  }
+#endif
+
+// Other platforms do intermediate conversion from RGB24 to ARGB.
+#else  // HAS_RGB24TOYJROW
+
+#if defined(HAS_RGB24TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RGB24ToARGBRow = RGB24ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToARGBRow = RGB24ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYJRow = ARGBToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYJRow = ARGBToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYJRow = ARGBToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVJRow = ARGBToUVJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVJRow = ARGBToUVJRow_AVX2;
+    }
+  }
+#endif
+#endif  // HAS_RGB24TOYJROW
+
+  {
+#if !defined(HAS_RGB24TOYJROW)
+    // Allocate 2 rows of ARGB.
+    const int row_size = (width * 4 + 31) & ~31;
+    align_buffer_64(row, row_size * 2);
+    if (!row)
+      return 1;
+#endif
+
+    for (y = 0; y < height - 1; y += 2) {
+#if defined(HAS_RGB24TOYJROW)
+      RGB24ToUVJRow(src_rgb24, src_stride_rgb24, dst_u, dst_v, width);
+      RGB24ToYJRow(src_rgb24, dst_y, width);
+      RGB24ToYJRow(src_rgb24 + src_stride_rgb24, dst_y + dst_stride_y, width);
+#else
+      RGB24ToARGBRow(src_rgb24, row, width);
+      RGB24ToARGBRow(src_rgb24 + src_stride_rgb24, row + row_size, width);
+      ARGBToUVJRow(row, row_size, dst_u, dst_v, width);
+      ARGBToYJRow(row, dst_y, width);
+      ARGBToYJRow(row + row_size, dst_y + dst_stride_y, width);
+#endif
+      src_rgb24 += src_stride_rgb24 * 2;
+      dst_y += dst_stride_y * 2;
+      dst_u += dst_stride_u;
+      dst_v += dst_stride_v;
+    }
+    if (height & 1) {
+#if defined(HAS_RGB24TOYJROW)
+      RGB24ToUVJRow(src_rgb24, 0, dst_u, dst_v, width);
+      RGB24ToYJRow(src_rgb24, dst_y, width);
+#else
+      RGB24ToARGBRow(src_rgb24, row, width);
+      ARGBToUVJRow(row, 0, dst_u, dst_v, width);
+      ARGBToYJRow(row, dst_y, width);
+#endif
+    }
+#if !defined(HAS_RGB24TOYJROW)
+    free_aligned_buffer_64(row);
+#endif
+  }
+  return 0;
+}
+#undef HAS_RGB24TOYJROW
+
+// Enabled if 1 pass is available
+#if (defined(HAS_RAWTOYROW_NEON) || defined(HAS_RAWTOYROW_MSA) || \
+     defined(HAS_RAWTOYROW_LSX) || defined(HAS_RAWTOYROW_RVV))
+#define HAS_RAWTOYROW
+#endif
+
+// Convert RAW to I420.
+LIBYUV_API
+int RAWToI420(const uint8_t* src_raw,
+              int src_stride_raw,
+              uint8_t* dst_y,
+              int dst_stride_y,
+              uint8_t* dst_u,
+              int dst_stride_u,
+              uint8_t* dst_v,
+              int dst_stride_v,
+              int width,
+              int height) {
+  int y;
+#if defined(HAS_RAWTOYROW)
+  void (*RAWToUVRow)(const uint8_t* src_raw, int src_stride_raw, uint8_t* dst_u,
+                     uint8_t* dst_v, int width) = RAWToUVRow_C;
+  void (*RAWToYRow)(const uint8_t* src_raw, uint8_t* dst_y, int width) =
+      RAWToYRow_C;
+#else
+  void (*RAWToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) =
+      RAWToARGBRow_C;
+  void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVRow_C;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+#endif
+  if (!src_raw || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_raw = src_raw + (height - 1) * src_stride_raw;
+    src_stride_raw = -src_stride_raw;
+  }
+
+#if defined(HAS_RAWTOYROW)
+
+// Neon version does direct RAW to YUV.
+#if defined(HAS_RAWTOYROW_NEON) && defined(HAS_RAWTOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RAWToUVRow = RAWToUVRow_Any_NEON;
+    RAWToYRow = RAWToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYRow = RAWToYRow_NEON;
+      RAWToUVRow = RAWToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYROW_MSA) && defined(HAS_RAWTOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RAWToUVRow = RAWToUVRow_Any_MSA;
+    RAWToYRow = RAWToYRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYRow = RAWToYRow_MSA;
+      RAWToUVRow = RAWToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYROW_LSX) && defined(HAS_RAWTOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RAWToUVRow = RAWToUVRow_Any_LSX;
+    RAWToYRow = RAWToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYRow = RAWToYRow_LSX;
+      RAWToUVRow = RAWToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYROW_LASX) && defined(HAS_RAWTOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    RAWToUVRow = RAWToUVRow_Any_LASX;
+    RAWToYRow = RAWToYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RAWToYRow = RAWToYRow_LASX;
+      RAWToUVRow = RAWToUVRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RAWToYRow = RAWToYRow_RVV;
+  }
+#endif
+
+// Other platforms do intermediate conversion from RAW to ARGB.
+#else  // HAS_RAWTOYROW
+
+#if defined(HAS_RAWTOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RAWToARGBRow = RAWToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToARGBRow = RAWToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_AVX2;
+    }
+  }
+#endif
+#endif  // HAS_RAWTOYROW
+
+  {
+#if !defined(HAS_RAWTOYROW)
+    // Allocate 2 rows of ARGB.
+    const int row_size = (width * 4 + 31) & ~31;
+    align_buffer_64(row, row_size * 2);
+    if (!row)
+      return 1;
+#endif
+
+    for (y = 0; y < height - 1; y += 2) {
+#if defined(HAS_RAWTOYROW)
+      RAWToUVRow(src_raw, src_stride_raw, dst_u, dst_v, width);
+      RAWToYRow(src_raw, dst_y, width);
+      RAWToYRow(src_raw + src_stride_raw, dst_y + dst_stride_y, width);
+#else
+      RAWToARGBRow(src_raw, row, width);
+      RAWToARGBRow(src_raw + src_stride_raw, row + row_size, width);
+      ARGBToUVRow(row, row_size, dst_u, dst_v, width);
+      ARGBToYRow(row, dst_y, width);
+      ARGBToYRow(row + row_size, dst_y + dst_stride_y, width);
+#endif
+      src_raw += src_stride_raw * 2;
+      dst_y += dst_stride_y * 2;
+      dst_u += dst_stride_u;
+      dst_v += dst_stride_v;
+    }
+    if (height & 1) {
+#if defined(HAS_RAWTOYROW)
+      RAWToUVRow(src_raw, 0, dst_u, dst_v, width);
+      RAWToYRow(src_raw, dst_y, width);
+#else
+      RAWToARGBRow(src_raw, row, width);
+      ARGBToUVRow(row, 0, dst_u, dst_v, width);
+      ARGBToYRow(row, dst_y, width);
+#endif
+    }
+#if !defined(HAS_RAWTOYROW)
+    free_aligned_buffer_64(row);
+#endif
+  }
+  return 0;
+}
+#undef HAS_RAWTOYROW
+
+// Enabled if 1 pass is available
+#if defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_MSA) || \
+    defined(HAS_RAWTOYJROW_RVV)
+#define HAS_RAWTOYJROW
+#endif
+
+// Convert RAW to J420.
+LIBYUV_API
+int RAWToJ420(const uint8_t* src_raw,
+              int src_stride_raw,
+              uint8_t* dst_y,
+              int dst_stride_y,
+              uint8_t* dst_u,
+              int dst_stride_u,
+              uint8_t* dst_v,
+              int dst_stride_v,
+              int width,
+              int height) {
+  int y;
+#if defined(HAS_RAWTOYJROW)
+  void (*RAWToUVJRow)(const uint8_t* src_raw, int src_stride_raw,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      RAWToUVJRow_C;
+  void (*RAWToYJRow)(const uint8_t* src_raw, uint8_t* dst_y, int width) =
+      RAWToYJRow_C;
+#else
+  void (*RAWToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) =
+      RAWToARGBRow_C;
+  void (*ARGBToUVJRow)(const uint8_t* src_argb0, int src_stride_argb,
+                       uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVJRow_C;
+  void (*ARGBToYJRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYJRow_C;
+#endif
+  if (!src_raw || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_raw = src_raw + (height - 1) * src_stride_raw;
+    src_stride_raw = -src_stride_raw;
+  }
+
+#if defined(HAS_RAWTOYJROW)
+
+// Neon version does direct RAW to YUV.
+#if defined(HAS_RAWTOYJROW_NEON) && defined(HAS_RAWTOUVJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RAWToUVJRow = RAWToUVJRow_Any_NEON;
+    RAWToYJRow = RAWToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYJRow = RAWToYJRow_NEON;
+      RAWToUVJRow = RAWToUVJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_MSA) && defined(HAS_RAWTOUVJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RAWToUVJRow = RAWToUVJRow_Any_MSA;
+    RAWToYJRow = RAWToYJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYJRow = RAWToYJRow_MSA;
+      RAWToUVJRow = RAWToUVJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RAWToYJRow = RAWToYJRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYJRow = RAWToYJRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    RAWToYJRow = RAWToYJRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RAWToYJRow = RAWToYJRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RAWToYJRow = RAWToYJRow_RVV;
+  }
+#endif
+
+// Other platforms do intermediate conversion from RAW to ARGB.
+#else  // HAS_RAWTOYJROW
+
+#if defined(HAS_RAWTOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RAWToARGBRow = RAWToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToARGBRow = RAWToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYJRow = ARGBToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYJRow = ARGBToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYJRow = ARGBToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVJRow = ARGBToUVJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVJRow = ARGBToUVJRow_AVX2;
+    }
+  }
+#endif
+#endif  // HAS_RAWTOYJROW
+
+  {
+#if !defined(HAS_RAWTOYJROW)
+    // Allocate 2 rows of ARGB.
+    const int row_size = (width * 4 + 31) & ~31;
+    align_buffer_64(row, row_size * 2);
+    if (!row)
+      return 1;
+#endif
+
+    for (y = 0; y < height - 1; y += 2) {
+#if defined(HAS_RAWTOYJROW)
+      RAWToUVJRow(src_raw, src_stride_raw, dst_u, dst_v, width);
+      RAWToYJRow(src_raw, dst_y, width);
+      RAWToYJRow(src_raw + src_stride_raw, dst_y + dst_stride_y, width);
+#else
+      RAWToARGBRow(src_raw, row, width);
+      RAWToARGBRow(src_raw + src_stride_raw, row + row_size, width);
+      ARGBToUVJRow(row, row_size, dst_u, dst_v, width);
+      ARGBToYJRow(row, dst_y, width);
+      ARGBToYJRow(row + row_size, dst_y + dst_stride_y, width);
+#endif
+      src_raw += src_stride_raw * 2;
+      dst_y += dst_stride_y * 2;
+      dst_u += dst_stride_u;
+      dst_v += dst_stride_v;
+    }
+    if (height & 1) {
+#if defined(HAS_RAWTOYJROW)
+      RAWToUVJRow(src_raw, 0, dst_u, dst_v, width);
+      RAWToYJRow(src_raw, dst_y, width);
+#else
+      RAWToARGBRow(src_raw, row, width);
+      ARGBToUVJRow(row, 0, dst_u, dst_v, width);
+      ARGBToYJRow(row, dst_y, width);
+#endif
+    }
+#if !defined(HAS_RAWTOYJROW)
+    free_aligned_buffer_64(row);
+#endif
+  }
+  return 0;
+}
+#undef HAS_RAWTOYJROW
+
+// Convert RGB565 to I420.
+LIBYUV_API
+int RGB565ToI420(const uint8_t* src_rgb565,
+                 int src_stride_rgb565,
+                 uint8_t* dst_y,
+                 int dst_stride_y,
+                 uint8_t* dst_u,
+                 int dst_stride_u,
+                 uint8_t* dst_v,
+                 int dst_stride_v,
+                 int width,
+                 int height) {
+  int y;
+#if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
+     defined(HAS_RGB565TOYROW_LSX) || defined(HAS_RGB565TOYROW_LASX))
+  void (*RGB565ToUVRow)(const uint8_t* src_rgb565, int src_stride_rgb565,
+                        uint8_t* dst_u, uint8_t* dst_v, int width) =
+      RGB565ToUVRow_C;
+  void (*RGB565ToYRow)(const uint8_t* src_rgb565, uint8_t* dst_y, int width) =
+      RGB565ToYRow_C;
+#else
+  void (*RGB565ToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb,
+                          int width) = RGB565ToARGBRow_C;
+  void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVRow_C;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+#endif
+  if (!src_rgb565 || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_rgb565 = src_rgb565 + (height - 1) * src_stride_rgb565;
+    src_stride_rgb565 = -src_stride_rgb565;
+  }
+
+// Neon version does direct RGB565 to YUV.
+#if defined(HAS_RGB565TOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RGB565ToUVRow = RGB565ToUVRow_Any_NEON;
+    RGB565ToYRow = RGB565ToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      RGB565ToYRow = RGB565ToYRow_NEON;
+      if (IS_ALIGNED(width, 16)) {
+        RGB565ToUVRow = RGB565ToUVRow_NEON;
+      }
+    }
+  }
+// MSA version does direct RGB565 to YUV.
+#elif (defined(HAS_RGB565TOYROW_MSA) || defined(HAS_RGB565TOYROW_LSX) || \
+       defined(HAS_RGB565TOYROW_LASX))
+#if defined(HAS_RGB565TOYROW_MSA) && defined(HAS_RGB565TOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RGB565ToUVRow = RGB565ToUVRow_Any_MSA;
+    RGB565ToYRow = RGB565ToYRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RGB565ToYRow = RGB565ToYRow_MSA;
+      RGB565ToUVRow = RGB565ToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RGB565TOYROW_LSX) && defined(HAS_RGB565TOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RGB565ToUVRow = RGB565ToUVRow_Any_LSX;
+    RGB565ToYRow = RGB565ToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RGB565ToYRow = RGB565ToYRow_LSX;
+      RGB565ToUVRow = RGB565ToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RGB565TOYROW_LASX) && defined(HAS_RGB565TOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    RGB565ToUVRow = RGB565ToUVRow_Any_LASX;
+    RGB565ToYRow = RGB565ToYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RGB565ToYRow = RGB565ToYRow_LASX;
+      RGB565ToUVRow = RGB565ToUVRow_LASX;
+    }
+  }
+#endif
+// Other platforms do intermediate conversion from RGB565 to ARGB.
+#else
+#if defined(HAS_RGB565TOARGBROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    RGB565ToARGBRow = RGB565ToARGBRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      RGB565ToARGBRow = RGB565ToARGBRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_RGB565TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    RGB565ToARGBRow = RGB565ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      RGB565ToARGBRow = RGB565ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_AVX2;
+    }
+  }
+#endif
+#endif
+  {
+#if !(defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
+      defined(HAS_RGB565TOYROW_LSX) || defined(HAS_RGB565TOYROW_LASX))
+    // Allocate 2 rows of ARGB.
+    const int row_size = (width * 4 + 31) & ~31;
+    align_buffer_64(row, row_size * 2);
+    if (!row)
+      return 1;
+#endif
+    for (y = 0; y < height - 1; y += 2) {
+#if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
+     defined(HAS_RGB565TOYROW_LSX) || defined(HAS_RGB565TOYROW_LASX))
+      RGB565ToUVRow(src_rgb565, src_stride_rgb565, dst_u, dst_v, width);
+      RGB565ToYRow(src_rgb565, dst_y, width);
+      RGB565ToYRow(src_rgb565 + src_stride_rgb565, dst_y + dst_stride_y, width);
+#else
+      RGB565ToARGBRow(src_rgb565, row, width);
+      RGB565ToARGBRow(src_rgb565 + src_stride_rgb565, row + row_size, width);
+      ARGBToUVRow(row, row_size, dst_u, dst_v, width);
+      ARGBToYRow(row, dst_y, width);
+      ARGBToYRow(row + row_size, dst_y + dst_stride_y, width);
+#endif
+      src_rgb565 += src_stride_rgb565 * 2;
+      dst_y += dst_stride_y * 2;
+      dst_u += dst_stride_u;
+      dst_v += dst_stride_v;
+    }
+    if (height & 1) {
+#if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
+     defined(HAS_RGB565TOYROW_LSX) || defined(HAS_RGB565TOYROW_LASX))
+      RGB565ToUVRow(src_rgb565, 0, dst_u, dst_v, width);
+      RGB565ToYRow(src_rgb565, dst_y, width);
+#else
+      RGB565ToARGBRow(src_rgb565, row, width);
+      ARGBToUVRow(row, 0, dst_u, dst_v, width);
+      ARGBToYRow(row, dst_y, width);
+#endif
+    }
+#if !(defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
+      defined(HAS_RGB565TOYROW_LSX) || defined(HAS_RGB565TOYROW_LASX))
+    free_aligned_buffer_64(row);
+#endif
+  }
+  return 0;
+}
+
+// Convert ARGB1555 to I420.
+LIBYUV_API
+int ARGB1555ToI420(const uint8_t* src_argb1555,
+                   int src_stride_argb1555,
+                   uint8_t* dst_y,
+                   int dst_stride_y,
+                   uint8_t* dst_u,
+                   int dst_stride_u,
+                   uint8_t* dst_v,
+                   int dst_stride_v,
+                   int width,
+                   int height) {
+  int y;
+#if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
+     defined(HAS_ARGB1555TOYROW_LSX) || defined(HAS_ARGB1555TOYROW_LASX))
+  void (*ARGB1555ToUVRow)(const uint8_t* src_argb1555, int src_stride_argb1555,
+                          uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGB1555ToUVRow_C;
+  void (*ARGB1555ToYRow)(const uint8_t* src_argb1555, uint8_t* dst_y,
+                         int width) = ARGB1555ToYRow_C;
+#else
+  void (*ARGB1555ToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb,
+                            int width) = ARGB1555ToARGBRow_C;
+  void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVRow_C;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+#endif
+  if (!src_argb1555 || !dst_y || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb1555 = src_argb1555 + (height - 1) * src_stride_argb1555;
+    src_stride_argb1555 = -src_stride_argb1555;
+  }
+
+// Neon version does direct ARGB1555 to YUV.
+#if defined(HAS_ARGB1555TOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGB1555ToUVRow = ARGB1555ToUVRow_Any_NEON;
+    ARGB1555ToYRow = ARGB1555ToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGB1555ToYRow = ARGB1555ToYRow_NEON;
+      if (IS_ALIGNED(width, 16)) {
+        ARGB1555ToUVRow = ARGB1555ToUVRow_NEON;
+      }
+    }
+  }
+// MSA version does direct ARGB1555 to YUV.
+#elif (defined(HAS_ARGB1555TOYROW_MSA) || defined(HAS_ARGB1555TOYROW_LSX) || \
+       defined(HAS_ARGB1555TOYROW_LASX))
+#if defined(HAS_ARGB1555TOYROW_MSA) && defined(HAS_ARGB1555TOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGB1555ToUVRow = ARGB1555ToUVRow_Any_MSA;
+    ARGB1555ToYRow = ARGB1555ToYRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB1555ToYRow = ARGB1555ToYRow_MSA;
+      ARGB1555ToUVRow = ARGB1555ToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGB1555TOYROW_LSX) && defined(HAS_ARGB1555TOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGB1555ToUVRow = ARGB1555ToUVRow_Any_LSX;
+    ARGB1555ToYRow = ARGB1555ToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB1555ToYRow = ARGB1555ToYRow_LSX;
+      ARGB1555ToUVRow = ARGB1555ToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGB1555TOYROW_LASX) && defined(HAS_ARGB1555TOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGB1555ToUVRow = ARGB1555ToUVRow_Any_LASX;
+    ARGB1555ToYRow = ARGB1555ToYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGB1555ToYRow = ARGB1555ToYRow_LASX;
+      ARGB1555ToUVRow = ARGB1555ToUVRow_LASX;
+    }
+  }
+#endif
+// Other platforms do intermediate conversion from ARGB1555 to ARGB.
+#else
+#if defined(HAS_ARGB1555TOARGBROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGB1555ToARGBRow = ARGB1555ToARGBRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGB1555TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB1555ToARGBRow = ARGB1555ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_AVX2;
+    }
+  }
+#endif
+#endif
+  {
+#if !(defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
+      defined(HAS_ARGB1555TOYROW_LSX) || defined(HAS_ARGB1555TOYROW_LASX))
+    // Allocate 2 rows of ARGB.
+    const int row_size = (width * 4 + 31) & ~31;
+    align_buffer_64(row, row_size * 2);
+    if (!row)
+      return 1;
+#endif
+
+    for (y = 0; y < height - 1; y += 2) {
+#if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
+     defined(HAS_ARGB1555TOYROW_LSX) || defined(HAS_ARGB1555TOYROW_LASX))
+      ARGB1555ToUVRow(src_argb1555, src_stride_argb1555, dst_u, dst_v, width);
+      ARGB1555ToYRow(src_argb1555, dst_y, width);
+      ARGB1555ToYRow(src_argb1555 + src_stride_argb1555, dst_y + dst_stride_y,
+                     width);
+#else
+      ARGB1555ToARGBRow(src_argb1555, row, width);
+      ARGB1555ToARGBRow(src_argb1555 + src_stride_argb1555, row + row_size,
+                        width);
+      ARGBToUVRow(row, row_size, dst_u, dst_v, width);
+      ARGBToYRow(row, dst_y, width);
+      ARGBToYRow(row + row_size, dst_y + dst_stride_y, width);
+#endif
+      src_argb1555 += src_stride_argb1555 * 2;
+      dst_y += dst_stride_y * 2;
+      dst_u += dst_stride_u;
+      dst_v += dst_stride_v;
+    }
+    if (height & 1) {
+#if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
+     defined(HAS_ARGB1555TOYROW_LSX) || defined(HAS_ARGB1555TOYROW_LASX))
+      ARGB1555ToUVRow(src_argb1555, 0, dst_u, dst_v, width);
+      ARGB1555ToYRow(src_argb1555, dst_y, width);
+#else
+      ARGB1555ToARGBRow(src_argb1555, row, width);
+      ARGBToUVRow(row, 0, dst_u, dst_v, width);
+      ARGBToYRow(row, dst_y, width);
+#endif
+    }
+#if !(defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
+      defined(HAS_ARGB1555TOYROW_LSX) || defined(HAS_ARGB1555TOYROW_LASX))
+    free_aligned_buffer_64(row);
+#endif
+  }
+  return 0;
+}
+
+// Convert ARGB4444 to I420.
+LIBYUV_API
+int ARGB4444ToI420(const uint8_t* src_argb4444,
+                   int src_stride_argb4444,
+                   uint8_t* dst_y,
+                   int dst_stride_y,
+                   uint8_t* dst_u,
+                   int dst_stride_u,
+                   uint8_t* dst_v,
+                   int dst_stride_v,
+                   int width,
+                   int height) {
+  int y;
+#if defined(HAS_ARGB4444TOYROW_NEON)
+  void (*ARGB4444ToUVRow)(const uint8_t* src_argb4444, int src_stride_argb4444,
+                          uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGB4444ToUVRow_C;
+  void (*ARGB4444ToYRow)(const uint8_t* src_argb4444, uint8_t* dst_y,
+                         int width) = ARGB4444ToYRow_C;
+#else
+  void (*ARGB4444ToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb,
+                            int width) = ARGB4444ToARGBRow_C;
+  void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVRow_C;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+#endif
+  if (!src_argb4444 || !dst_y || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb4444 = src_argb4444 + (height - 1) * src_stride_argb4444;
+    src_stride_argb4444 = -src_stride_argb4444;
+  }
+
+// Neon version does direct ARGB4444 to YUV.
+#if defined(HAS_ARGB4444TOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGB4444ToUVRow = ARGB4444ToUVRow_Any_NEON;
+    ARGB4444ToYRow = ARGB4444ToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGB4444ToYRow = ARGB4444ToYRow_NEON;
+      if (IS_ALIGNED(width, 16)) {
+        ARGB4444ToUVRow = ARGB4444ToUVRow_NEON;
+      }
+    }
+  }
+// Other platforms do intermediate conversion from ARGB4444 to ARGB.
+#else
+#if defined(HAS_ARGB4444TOARGBROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGB4444ToARGBRow = ARGB4444ToARGBRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGB4444TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB4444ToARGBRow = ARGB4444ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGB4444TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB4444ToARGBRow = ARGB4444ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGB4444TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB4444ToARGBRow = ARGB4444ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGB4444TOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGB4444ToARGBRow = ARGB4444ToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToUVRow = ARGBToUVRow_Any_MSA;
+    ARGBToYRow = ARGBToYRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_MSA;
+      if (IS_ALIGNED(width, 32)) {
+        ARGBToUVRow = ARGBToUVRow_MSA;
+      }
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    ARGBToUVRow = ARGBToUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+      ARGBToUVRow = ARGBToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYRow = ARGBToYRow_Any_LASX;
+    ARGBToUVRow = ARGBToUVRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_LASX;
+      ARGBToUVRow = ARGBToUVRow_LASX;
+    }
+  }
+#endif
+#endif
+
+  {
+#if !(defined(HAS_ARGB4444TOYROW_NEON))
+    // Allocate 2 rows of ARGB.
+    const int row_size = (width * 4 + 31) & ~31;
+    align_buffer_64(row, row_size * 2);
+    if (!row)
+      return 1;
+#endif
+
+    for (y = 0; y < height - 1; y += 2) {
+#if defined(HAS_ARGB4444TOYROW_NEON)
+      ARGB4444ToUVRow(src_argb4444, src_stride_argb4444, dst_u, dst_v, width);
+      ARGB4444ToYRow(src_argb4444, dst_y, width);
+      ARGB4444ToYRow(src_argb4444 + src_stride_argb4444, dst_y + dst_stride_y,
+                     width);
+#else
+      ARGB4444ToARGBRow(src_argb4444, row, width);
+      ARGB4444ToARGBRow(src_argb4444 + src_stride_argb4444, row + row_size,
+                        width);
+      ARGBToUVRow(row, row_size, dst_u, dst_v, width);
+      ARGBToYRow(row, dst_y, width);
+      ARGBToYRow(row + row_size, dst_y + dst_stride_y, width);
+#endif
+      src_argb4444 += src_stride_argb4444 * 2;
+      dst_y += dst_stride_y * 2;
+      dst_u += dst_stride_u;
+      dst_v += dst_stride_v;
+    }
+    if (height & 1) {
+#if defined(HAS_ARGB4444TOYROW_NEON)
+      ARGB4444ToUVRow(src_argb4444, 0, dst_u, dst_v, width);
+      ARGB4444ToYRow(src_argb4444, dst_y, width);
+#else
+      ARGB4444ToARGBRow(src_argb4444, row, width);
+      ARGBToUVRow(row, 0, dst_u, dst_v, width);
+      ARGBToYRow(row, dst_y, width);
+#endif
+    }
+#if !(defined(HAS_ARGB4444TOYROW_NEON))
+    free_aligned_buffer_64(row);
+#endif
+  }
+  return 0;
+}
+
+// Convert RGB24 to J400.
+LIBYUV_API
+int RGB24ToJ400(const uint8_t* src_rgb24,
+                int src_stride_rgb24,
+                uint8_t* dst_yj,
+                int dst_stride_yj,
+                int width,
+                int height) {
+  int y;
+  void (*RGB24ToYJRow)(const uint8_t* src_rgb24, uint8_t* dst_yj, int width) =
+      RGB24ToYJRow_C;
+  if (!src_rgb24 || !dst_yj || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_rgb24 = src_rgb24 + (height - 1) * src_stride_rgb24;
+    src_stride_rgb24 = -src_stride_rgb24;
+  }
+  // Coalesce rows.
+  if (src_stride_rgb24 == width * 3 && dst_stride_yj == width) {
+    width *= height;
+    height = 1;
+    src_stride_rgb24 = dst_stride_yj = 0;
+  }
+#if defined(HAS_RGB24TOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RGB24ToYJRow = RGB24ToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToYJRow = RGB24ToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    RGB24ToYJRow = RGB24ToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      RGB24ToYJRow = RGB24ToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RGB24ToYJRow = RGB24ToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToYJRow = RGB24ToYJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RGB24ToYJRow = RGB24ToYJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToYJRow = RGB24ToYJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYJROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RGB24ToYJRow = RGB24ToYJRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToYJRow = RGB24ToYJRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYJROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    RGB24ToYJRow = RGB24ToYJRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RGB24ToYJRow = RGB24ToYJRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RGB24ToYJRow = RGB24ToYJRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    RGB24ToYJRow(src_rgb24, dst_yj, width);
+    src_rgb24 += src_stride_rgb24;
+    dst_yj += dst_stride_yj;
+  }
+  return 0;
+}
+
+// Convert RAW to J400.
+LIBYUV_API
+int RAWToJ400(const uint8_t* src_raw,
+              int src_stride_raw,
+              uint8_t* dst_yj,
+              int dst_stride_yj,
+              int width,
+              int height) {
+  int y;
+  void (*RAWToYJRow)(const uint8_t* src_raw, uint8_t* dst_yj, int width) =
+      RAWToYJRow_C;
+  if (!src_raw || !dst_yj || width <= 0 || height == 0) {
+    return -1;
+  }
+
+  if (height < 0) {
+    height = -height;
+    src_raw = src_raw + (height - 1) * src_stride_raw;
+    src_stride_raw = -src_stride_raw;
+  }
+  // Coalesce rows.
+  if (src_stride_raw == width * 3 && dst_stride_yj == width) {
+    width *= height;
+    height = 1;
+    src_stride_raw = dst_stride_yj = 0;
+  }
+
+#if defined(HAS_RAWTOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RAWToYJRow = RAWToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYJRow = RAWToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    RAWToYJRow = RAWToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      RAWToYJRow = RAWToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RAWToYJRow = RAWToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYJRow = RAWToYJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RAWToYJRow = RAWToYJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYJRow = RAWToYJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RAWToYJRow = RAWToYJRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYJRow = RAWToYJRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    RAWToYJRow = RAWToYJRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RAWToYJRow = RAWToYJRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RAWToYJRow = RAWToYJRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    RAWToYJRow(src_raw, dst_yj, width);
+    src_raw += src_stride_raw;
+    dst_yj += dst_stride_yj;
+  }
+  return 0;
+}
+
+// Convert Android420 to I420.
+LIBYUV_API
+int Android420ToI420(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     int src_pixel_stride_uv,
+                     uint8_t* dst_y,
+                     int dst_stride_y,
+                     uint8_t* dst_u,
+                     int dst_stride_u,
+                     uint8_t* dst_v,
+                     int dst_stride_v,
+                     int width,
+                     int height) {
+  return Android420ToI420Rotate(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                                src_stride_v, src_pixel_stride_uv, dst_y,
+                                dst_stride_y, dst_u, dst_stride_u, dst_v,
+                                dst_stride_v, width, height, kRotate0);
+}
+
+// depth is source bits measured from lsb; For msb use 16
+static int Biplanar16bitTo8bit(const uint16_t* src_y,
+                               int src_stride_y,
+                               const uint16_t* src_uv,
+                               int src_stride_uv,
+                               uint8_t* dst_y,
+                               int dst_stride_y,
+                               uint8_t* dst_uv,
+                               int dst_stride_uv,
+                               int width,
+                               int height,
+                               int subsample_x,
+                               int subsample_y,
+                               int depth) {
+  int uv_width = SUBSAMPLE(width, subsample_x, subsample_x);
+  int uv_height = SUBSAMPLE(height, subsample_y, subsample_y);
+  int scale = 1 << (24 - depth);
+  if ((!src_y && dst_y) || !src_uv || !dst_uv || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    uv_height = -uv_height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_uv = src_uv + (uv_height - 1) * src_stride_uv;
+    src_stride_y = -src_stride_y;
+    src_stride_uv = -src_stride_uv;
+  }
+
+  // Convert Y plane.
+  if (dst_y) {
+    Convert16To8Plane(src_y, src_stride_y, dst_y, dst_stride_y, scale, width,
+                      height);
+  }
+  // Convert UV planes.
+  Convert16To8Plane(src_uv, src_stride_uv, dst_uv, dst_stride_uv, scale,
+                    uv_width * 2, uv_height);
+  return 0;
+}
+
+// Convert 10 bit P010 to 8 bit NV12.
+// Depth set to 16 because P010 uses 10 msb and this function keeps the upper 8
+// bits of the specified number of bits.
+LIBYUV_API
+int P010ToNV12(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  return Biplanar16bitTo8bit(src_y, src_stride_y, src_uv, src_stride_uv, dst_y,
+                             dst_stride_y, dst_uv, dst_stride_uv, width, height,
+                             1, 1, 16);
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/convert_argb.cc

@@ -0,0 +1,8903 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/convert_argb.h"
+
+#include <assert.h>
+
+#include "libyuv/convert_from_argb.h"
+#include "libyuv/cpu_id.h"
+#include "libyuv/planar_functions.h"  // For CopyPlane and ARGBShuffle.
+#include "libyuv/rotate_argb.h"
+#include "libyuv/row.h"
+#include "libyuv/scale_row.h"  // For ScaleRowUp2_Linear and ScaleRowUp2_Bilinear
+#include "libyuv/video_common.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Copy ARGB with optional flipping
+LIBYUV_API
+int ARGBCopy(const uint8_t* src_argb,
+             int src_stride_argb,
+             uint8_t* dst_argb,
+             int dst_stride_argb,
+             int width,
+             int height) {
+  if (!src_argb || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+
+  CopyPlane(src_argb, src_stride_argb, dst_argb, dst_stride_argb, width * 4,
+            height);
+  return 0;
+}
+
+// Convert I420 to ARGB with matrix.
+LIBYUV_API
+int I420ToARGBMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I422ToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                        const uint8_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I422ToARGBRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I422TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToARGBRow = I422ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGBRow = I422ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToARGBRow = I422ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToARGBRow = I422ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW | kCpuHasAVX512VL) ==
+      (kCpuHasAVX512BW | kCpuHasAVX512VL)) {
+    I422ToARGBRow = I422ToARGBRow_Any_AVX512BW;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToARGBRow = I422ToARGBRow_AVX512BW;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToARGBRow = I422ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGBRow = I422ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I422ToARGBRow = I422ToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToARGBRow = I422ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGBRow = I422ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToARGBRow = I422ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToARGBRow = I422ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToARGBRow = I422ToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToARGBRow = I422ToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I422ToARGBRow = I422ToARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToARGBRow(src_y, src_u, src_v, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert I420 to ARGB.
+LIBYUV_API
+int I420ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I420ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvI601Constants, width, height);
+}
+
+// Convert I420 to ABGR.
+LIBYUV_API
+int I420ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I420ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuI601Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert J420 to ARGB.
+LIBYUV_API
+int J420ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I420ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvJPEGConstants, width, height);
+}
+
+// Convert J420 to ABGR.
+LIBYUV_API
+int J420ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I420ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuJPEGConstants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert H420 to ARGB.
+LIBYUV_API
+int H420ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I420ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvH709Constants, width, height);
+}
+
+// Convert H420 to ABGR.
+LIBYUV_API
+int H420ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I420ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuH709Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert U420 to ARGB.
+LIBYUV_API
+int U420ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I420ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuv2020Constants, width, height);
+}
+
+// Convert U420 to ABGR.
+LIBYUV_API
+int U420ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I420ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvu2020Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert I422 to ARGB with matrix.
+LIBYUV_API
+int I422ToARGBMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I422ToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                        const uint8_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I422ToARGBRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && src_stride_u * 2 == width &&
+      src_stride_v * 2 == width && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_y = src_stride_u = src_stride_v = dst_stride_argb = 0;
+  }
+#if defined(HAS_I422TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToARGBRow = I422ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGBRow = I422ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToARGBRow = I422ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToARGBRow = I422ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW | kCpuHasAVX512VL) ==
+      (kCpuHasAVX512BW | kCpuHasAVX512VL)) {
+    I422ToARGBRow = I422ToARGBRow_Any_AVX512BW;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToARGBRow = I422ToARGBRow_AVX512BW;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToARGBRow = I422ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGBRow = I422ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I422ToARGBRow = I422ToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToARGBRow = I422ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGBRow = I422ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToARGBRow = I422ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToARGBRow = I422ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToARGBRow = I422ToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToARGBRow = I422ToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I422ToARGBRow = I422ToARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToARGBRow(src_y, src_u, src_v, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert I422 to ARGB.
+LIBYUV_API
+int I422ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I422ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvI601Constants, width, height);
+}
+
+// Convert I422 to ABGR.
+LIBYUV_API
+int I422ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I422ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuI601Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert J422 to ARGB.
+LIBYUV_API
+int J422ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I422ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvJPEGConstants, width, height);
+}
+
+// Convert J422 to ABGR.
+LIBYUV_API
+int J422ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I422ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuJPEGConstants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert H422 to ARGB.
+LIBYUV_API
+int H422ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I422ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvH709Constants, width, height);
+}
+
+// Convert H422 to ABGR.
+LIBYUV_API
+int H422ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I422ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuH709Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert U422 to ARGB.
+LIBYUV_API
+int U422ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I422ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuv2020Constants, width, height);
+}
+
+// Convert U422 to ABGR.
+LIBYUV_API
+int U422ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I422ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvu2020Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert I444 to ARGB with matrix.
+LIBYUV_API
+int I444ToARGBMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I444ToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                        const uint8_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I444ToARGBRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && src_stride_u == width && src_stride_v == width &&
+      dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_y = src_stride_u = src_stride_v = dst_stride_argb = 0;
+  }
+#if defined(HAS_I444TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I444ToARGBRow = I444ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToARGBRow = I444ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I444ToARGBRow = I444ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I444ToARGBRow = I444ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I444ToARGBRow = I444ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToARGBRow = I444ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I444ToARGBRow = I444ToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I444ToARGBRow = I444ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToARGBRow = I444ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I444ToARGBRow = I444ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I444ToARGBRow = I444ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I444ToARGBRow = I444ToARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I444ToARGBRow(src_y, src_u, src_v, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert I444 to ARGB.
+LIBYUV_API
+int I444ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I444ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvI601Constants, width, height);
+}
+
+// Convert I444 to ABGR.
+LIBYUV_API
+int I444ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I444ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuI601Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert J444 to ARGB.
+LIBYUV_API
+int J444ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I444ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvJPEGConstants, width, height);
+}
+
+// Convert J444 to ABGR.
+LIBYUV_API
+int J444ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I444ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuJPEGConstants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert H444 to ARGB.
+LIBYUV_API
+int H444ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I444ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvH709Constants, width, height);
+}
+
+// Convert H444 to ABGR.
+LIBYUV_API
+int H444ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I444ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuH709Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert U444 to ARGB.
+LIBYUV_API
+int U444ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I444ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuv2020Constants, width, height);
+}
+
+// Convert U444 to ABGR.
+LIBYUV_API
+int U444ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I444ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvu2020Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert I444 to RGB24 with matrix.
+LIBYUV_API
+int I444ToRGB24Matrix(const uint8_t* src_y,
+                      int src_stride_y,
+                      const uint8_t* src_u,
+                      int src_stride_u,
+                      const uint8_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_rgb24,
+                      int dst_stride_rgb24,
+                      const struct YuvConstants* yuvconstants,
+                      int width,
+                      int height) {
+  int y;
+  void (*I444ToRGB24Row)(const uint8_t* y_buf, const uint8_t* u_buf,
+                         const uint8_t* v_buf, uint8_t* rgb_buf,
+                         const struct YuvConstants* yuvconstants, int width) =
+      I444ToRGB24Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_rgb24 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb24 = dst_rgb24 + (height - 1) * dst_stride_rgb24;
+    dst_stride_rgb24 = -dst_stride_rgb24;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && src_stride_u == width && src_stride_v == width &&
+      dst_stride_rgb24 == width * 3) {
+    width *= height;
+    height = 1;
+    src_stride_y = src_stride_u = src_stride_v = dst_stride_rgb24 = 0;
+  }
+#if defined(HAS_I444TORGB24ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      I444ToRGB24Row = I444ToRGB24Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      I444ToRGB24Row = I444ToRGB24Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToRGB24Row = I444ToRGB24Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToRGB24Row = I444ToRGB24Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I444ToRGB24Row = I444ToRGB24Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I444ToRGB24Row = I444ToRGB24Row_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I444ToRGB24Row(src_y, src_u, src_v, dst_rgb24, yuvconstants, width);
+    dst_rgb24 += dst_stride_rgb24;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert I444 to RGB24.
+LIBYUV_API
+int I444ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_u,
+                int src_stride_u,
+                const uint8_t* src_v,
+                int src_stride_v,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height) {
+  return I444ToRGB24Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                           src_stride_v, dst_rgb24, dst_stride_rgb24,
+                           &kYuvI601Constants, width, height);
+}
+
+// Convert I444 to RAW.
+LIBYUV_API
+int I444ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height) {
+  return I444ToRGB24Matrix(src_y, src_stride_y, src_v,
+                           src_stride_v,  // Swap U and V
+                           src_u, src_stride_u, dst_raw, dst_stride_raw,
+                           &kYvuI601Constants,  // Use Yvu matrix
+                           width, height);
+}
+
+// Convert 10 bit YUV to ARGB with matrix.
+// TODO(fbarchard): Consider passing scale multiplier to I210ToARGB to
+// multiply 10 bit yuv into high bits to allow any number of bits.
+LIBYUV_API
+int I010ToAR30Matrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I210ToAR30Row)(const uint16_t* y_buf, const uint16_t* u_buf,
+                        const uint16_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I210ToAR30Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar30 = dst_ar30 + (height - 1) * dst_stride_ar30;
+    dst_stride_ar30 = -dst_stride_ar30;
+  }
+#if defined(HAS_I210TOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I210ToAR30Row = I210ToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I210ToAR30Row = I210ToAR30Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I210TOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I210ToAR30Row = I210ToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I210ToAR30Row = I210ToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I210TOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I210ToAR30Row = I210ToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I210ToAR30Row = I210ToAR30Row_AVX2;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    I210ToAR30Row(src_y, src_u, src_v, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert I010 to AR30.
+LIBYUV_API
+int I010ToAR30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height) {
+  return I010ToAR30Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_ar30, dst_stride_ar30,
+                          &kYuvI601Constants, width, height);
+}
+
+// Convert H010 to AR30.
+LIBYUV_API
+int H010ToAR30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height) {
+  return I010ToAR30Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_ar30, dst_stride_ar30,
+                          &kYuvH709Constants, width, height);
+}
+
+// Convert U010 to AR30.
+LIBYUV_API
+int U010ToAR30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height) {
+  return I010ToAR30Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_ar30, dst_stride_ar30,
+                          &kYuv2020Constants, width, height);
+}
+
+// Convert I010 to AB30.
+LIBYUV_API
+int I010ToAB30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height) {
+  return I010ToAR30Matrix(src_y, src_stride_y, src_v, src_stride_v, src_u,
+                          src_stride_u, dst_ab30, dst_stride_ab30,
+                          &kYvuI601Constants, width, height);
+}
+
+// Convert H010 to AB30.
+LIBYUV_API
+int H010ToAB30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height) {
+  return I010ToAR30Matrix(src_y, src_stride_y, src_v, src_stride_v, src_u,
+                          src_stride_u, dst_ab30, dst_stride_ab30,
+                          &kYvuH709Constants, width, height);
+}
+
+// Convert U010 to AB30.
+LIBYUV_API
+int U010ToAB30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height) {
+  return I010ToAR30Matrix(src_y, src_stride_y, src_v, src_stride_v, src_u,
+                          src_stride_u, dst_ab30, dst_stride_ab30,
+                          &kYuv2020Constants, width, height);
+}
+
+// Convert 12 bit YUV to ARGB with matrix.
+// TODO(fbarchard): Consider passing scale multiplier to I212ToARGB to
+// multiply 12 bit yuv into high bits to allow any number of bits.
+LIBYUV_API
+int I012ToAR30Matrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I212ToAR30Row)(const uint16_t* y_buf, const uint16_t* u_buf,
+                        const uint16_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I212ToAR30Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar30 = dst_ar30 + (height - 1) * dst_stride_ar30;
+    dst_stride_ar30 = -dst_stride_ar30;
+  }
+#if defined(HAS_I212TOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I212ToAR30Row = I212ToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I212ToAR30Row = I212ToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I212TOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I212ToAR30Row = I212ToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I212ToAR30Row = I212ToAR30Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I212TOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I212ToAR30Row = I212ToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I212ToAR30Row = I212ToAR30Row_NEON;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    I212ToAR30Row(src_y, src_u, src_v, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert 10 bit YUV to ARGB with matrix.
+// TODO(fbarchard): Consider passing scale multiplier to I210ToARGB to
+// multiply 10 bit yuv into high bits to allow any number of bits.
+LIBYUV_API
+int I210ToAR30Matrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I210ToAR30Row)(const uint16_t* y_buf, const uint16_t* u_buf,
+                        const uint16_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I210ToAR30Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar30 = dst_ar30 + (height - 1) * dst_stride_ar30;
+    dst_stride_ar30 = -dst_stride_ar30;
+  }
+#if defined(HAS_I210TOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I210ToAR30Row = I210ToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I210ToAR30Row = I210ToAR30Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I210TOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I210ToAR30Row = I210ToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I210ToAR30Row = I210ToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I210TOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I210ToAR30Row = I210ToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I210ToAR30Row = I210ToAR30Row_AVX2;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    I210ToAR30Row(src_y, src_u, src_v, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert I210 to AR30.
+LIBYUV_API
+int I210ToAR30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height) {
+  return I210ToAR30Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_ar30, dst_stride_ar30,
+                          &kYuvI601Constants, width, height);
+}
+
+// Convert H210 to AR30.
+LIBYUV_API
+int H210ToAR30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height) {
+  return I210ToAR30Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_ar30, dst_stride_ar30,
+                          &kYuvH709Constants, width, height);
+}
+
+// Convert U210 to AR30.
+LIBYUV_API
+int U210ToAR30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height) {
+  return I210ToAR30Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_ar30, dst_stride_ar30,
+                          &kYuv2020Constants, width, height);
+}
+
+// Convert I210 to AB30.
+LIBYUV_API
+int I210ToAB30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height) {
+  return I210ToAR30Matrix(src_y, src_stride_y, src_v, src_stride_v, src_u,
+                          src_stride_u, dst_ab30, dst_stride_ab30,
+                          &kYvuI601Constants, width, height);
+}
+
+// Convert H210 to AB30.
+LIBYUV_API
+int H210ToAB30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height) {
+  return I210ToAR30Matrix(src_y, src_stride_y, src_v, src_stride_v, src_u,
+                          src_stride_u, dst_ab30, dst_stride_ab30,
+                          &kYvuH709Constants, width, height);
+}
+
+// Convert U210 to AB30.
+LIBYUV_API
+int U210ToAB30(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height) {
+  return I210ToAR30Matrix(src_y, src_stride_y, src_v, src_stride_v, src_u,
+                          src_stride_u, dst_ab30, dst_stride_ab30,
+                          &kYuv2020Constants, width, height);
+}
+
+LIBYUV_API
+int I410ToAR30Matrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I410ToAR30Row)(const uint16_t* y_buf, const uint16_t* u_buf,
+                        const uint16_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I410ToAR30Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar30 = dst_ar30 + (height - 1) * dst_stride_ar30;
+    dst_stride_ar30 = -dst_stride_ar30;
+  }
+#if defined(HAS_I410TOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I410ToAR30Row = I410ToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I410ToAR30Row = I410ToAR30Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I410TOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I410ToAR30Row = I410ToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I410ToAR30Row = I410ToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I410TOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I410ToAR30Row = I410ToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I410ToAR30Row = I410ToAR30Row_AVX2;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    I410ToAR30Row(src_y, src_u, src_v, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert 10 bit YUV to ARGB with matrix.
+LIBYUV_API
+int I010ToARGBMatrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I210ToARGBRow)(const uint16_t* y_buf, const uint16_t* u_buf,
+                        const uint16_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I210ToARGBRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I210TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I210ToARGBRow = I210ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I210ToARGBRow = I210ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I210TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I210ToARGBRow = I210ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I210ToARGBRow = I210ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I210TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I210ToARGBRow = I210ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I210ToARGBRow = I210ToARGBRow_AVX2;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    I210ToARGBRow(src_y, src_u, src_v, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert I010 to ARGB.
+LIBYUV_API
+int I010ToARGB(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I010ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvI601Constants, width, height);
+}
+
+// Convert I010 to ABGR.
+LIBYUV_API
+int I010ToABGR(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I010ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuI601Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert H010 to ARGB.
+LIBYUV_API
+int H010ToARGB(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I010ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvH709Constants, width, height);
+}
+
+// Convert H010 to ABGR.
+LIBYUV_API
+int H010ToABGR(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I010ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuH709Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert U010 to ARGB.
+LIBYUV_API
+int U010ToARGB(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I010ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuv2020Constants, width, height);
+}
+
+// Convert U010 to ABGR.
+LIBYUV_API
+int U010ToABGR(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I010ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvu2020Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert 12 bit YUV to ARGB with matrix.
+LIBYUV_API
+int I012ToARGBMatrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I212ToARGBRow)(const uint16_t* y_buf, const uint16_t* u_buf,
+                        const uint16_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I212ToARGBRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I212TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I212ToARGBRow = I212ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I212ToARGBRow = I212ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I212TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I212ToARGBRow = I212ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I212ToARGBRow = I212ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I212TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I212ToARGBRow = I212ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I212ToARGBRow = I212ToARGBRow_NEON;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    I212ToARGBRow(src_y, src_u, src_v, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert 10 bit 422 YUV to ARGB with matrix.
+LIBYUV_API
+int I210ToARGBMatrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I210ToARGBRow)(const uint16_t* y_buf, const uint16_t* u_buf,
+                        const uint16_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I210ToARGBRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I210TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I210ToARGBRow = I210ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I210ToARGBRow = I210ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I210TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I210ToARGBRow = I210ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I210ToARGBRow = I210ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I210TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I210ToARGBRow = I210ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I210ToARGBRow = I210ToARGBRow_AVX2;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    I210ToARGBRow(src_y, src_u, src_v, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert I210 to ARGB.
+LIBYUV_API
+int I210ToARGB(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I210ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvI601Constants, width, height);
+}
+
+// Convert I210 to ABGR.
+LIBYUV_API
+int I210ToABGR(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I210ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuI601Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert H210 to ARGB.
+LIBYUV_API
+int H210ToARGB(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I210ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuvH709Constants, width, height);
+}
+
+// Convert H210 to ABGR.
+LIBYUV_API
+int H210ToABGR(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I210ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvuH709Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert U210 to ARGB.
+LIBYUV_API
+int U210ToARGB(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I210ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_argb, dst_stride_argb,
+                          &kYuv2020Constants, width, height);
+}
+
+// Convert U210 to ABGR.
+LIBYUV_API
+int U210ToABGR(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return I210ToARGBMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_abgr, dst_stride_abgr,
+                          &kYvu2020Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+LIBYUV_API
+int I410ToARGBMatrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I410ToARGBRow)(const uint16_t* y_buf, const uint16_t* u_buf,
+                        const uint16_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I410ToARGBRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I410TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I410ToARGBRow = I410ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I410ToARGBRow = I410ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I410TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I410ToARGBRow = I410ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I410ToARGBRow = I410ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I410TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I410ToARGBRow = I410ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I410ToARGBRow = I410ToARGBRow_AVX2;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    I410ToARGBRow(src_y, src_u, src_v, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+LIBYUV_API
+int P010ToARGBMatrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_uv,
+                     int src_stride_uv,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*P210ToARGBRow)(
+      const uint16_t* y_buf, const uint16_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = P210ToARGBRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_uv || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_P210TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    P210ToARGBRow = P210ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      P210ToARGBRow = P210ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_P210TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    P210ToARGBRow = P210ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      P210ToARGBRow = P210ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_P210TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    P210ToARGBRow = P210ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      P210ToARGBRow = P210ToARGBRow_NEON;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    P210ToARGBRow(src_y, src_uv, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_uv += src_stride_uv;
+    }
+  }
+  return 0;
+}
+
+LIBYUV_API
+int P210ToARGBMatrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_uv,
+                     int src_stride_uv,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*P210ToARGBRow)(
+      const uint16_t* y_buf, const uint16_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = P210ToARGBRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_uv || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_P210TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    P210ToARGBRow = P210ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      P210ToARGBRow = P210ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_P210TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    P210ToARGBRow = P210ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      P210ToARGBRow = P210ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_P210TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    P210ToARGBRow = P210ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      P210ToARGBRow = P210ToARGBRow_NEON;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    P210ToARGBRow(src_y, src_uv, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_uv += src_stride_uv;
+  }
+  return 0;
+}
+
+LIBYUV_API
+int P010ToAR30Matrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_uv,
+                     int src_stride_uv,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*P210ToAR30Row)(
+      const uint16_t* y_buf, const uint16_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = P210ToAR30Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_uv || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar30 = dst_ar30 + (height - 1) * dst_stride_ar30;
+    dst_stride_ar30 = -dst_stride_ar30;
+  }
+#if defined(HAS_P210TOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    P210ToAR30Row = P210ToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      P210ToAR30Row = P210ToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_P210TOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    P210ToAR30Row = P210ToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      P210ToAR30Row = P210ToAR30Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_P210TOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    P210ToAR30Row = P210ToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      P210ToAR30Row = P210ToAR30Row_NEON;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    P210ToAR30Row(src_y, src_uv, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_uv += src_stride_uv;
+    }
+  }
+  return 0;
+}
+
+LIBYUV_API
+int P210ToAR30Matrix(const uint16_t* src_y,
+                     int src_stride_y,
+                     const uint16_t* src_uv,
+                     int src_stride_uv,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*P210ToAR30Row)(
+      const uint16_t* y_buf, const uint16_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = P210ToAR30Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_uv || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar30 = dst_ar30 + (height - 1) * dst_stride_ar30;
+    dst_stride_ar30 = -dst_stride_ar30;
+  }
+#if defined(HAS_P210TOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    P210ToAR30Row = P210ToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      P210ToAR30Row = P210ToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_P210TOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    P210ToAR30Row = P210ToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      P210ToAR30Row = P210ToAR30Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_P210TOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    P210ToAR30Row = P210ToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      P210ToAR30Row = P210ToAR30Row_NEON;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    P210ToAR30Row(src_y, src_uv, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    src_uv += src_stride_uv;
+  }
+  return 0;
+}
+
+// Convert I420 with Alpha to preattenuated ARGB with matrix.
+LIBYUV_API
+int I420AlphaToARGBMatrix(const uint8_t* src_y,
+                          int src_stride_y,
+                          const uint8_t* src_u,
+                          int src_stride_u,
+                          const uint8_t* src_v,
+                          int src_stride_v,
+                          const uint8_t* src_a,
+                          int src_stride_a,
+                          uint8_t* dst_argb,
+                          int dst_stride_argb,
+                          const struct YuvConstants* yuvconstants,
+                          int width,
+                          int height,
+                          int attenuate) {
+  int y;
+  void (*I422AlphaToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                             const uint8_t* v_buf, const uint8_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width) = I422AlphaToARGBRow_C;
+  void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                           int width) = ARGBAttenuateRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !src_a || !dst_argb || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I422ALPHATOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_RVV;
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422AlphaToARGBRow(src_y, src_u, src_v, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+    dst_argb += dst_stride_argb;
+    src_a += src_stride_a;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert I422 with Alpha to preattenuated ARGB with matrix.
+LIBYUV_API
+int I422AlphaToARGBMatrix(const uint8_t* src_y,
+                          int src_stride_y,
+                          const uint8_t* src_u,
+                          int src_stride_u,
+                          const uint8_t* src_v,
+                          int src_stride_v,
+                          const uint8_t* src_a,
+                          int src_stride_a,
+                          uint8_t* dst_argb,
+                          int dst_stride_argb,
+                          const struct YuvConstants* yuvconstants,
+                          int width,
+                          int height,
+                          int attenuate) {
+  int y;
+  void (*I422AlphaToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                             const uint8_t* v_buf, const uint8_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width) = I422AlphaToARGBRow_C;
+  void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                           int width) = ARGBAttenuateRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !src_a || !dst_argb || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I422ALPHATOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I422ALPHATOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_RVV;
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422AlphaToARGBRow(src_y, src_u, src_v, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+    dst_argb += dst_stride_argb;
+    src_a += src_stride_a;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert I444 with Alpha to preattenuated ARGB with matrix.
+LIBYUV_API
+int I444AlphaToARGBMatrix(const uint8_t* src_y,
+                          int src_stride_y,
+                          const uint8_t* src_u,
+                          int src_stride_u,
+                          const uint8_t* src_v,
+                          int src_stride_v,
+                          const uint8_t* src_a,
+                          int src_stride_a,
+                          uint8_t* dst_argb,
+                          int dst_stride_argb,
+                          const struct YuvConstants* yuvconstants,
+                          int width,
+                          int height,
+                          int attenuate) {
+  int y;
+  void (*I444AlphaToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                             const uint8_t* v_buf, const uint8_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width) = I444AlphaToARGBRow_C;
+  void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                           int width) = ARGBAttenuateRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !src_a || !dst_argb || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I444ALPHATOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_RVV;
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I444AlphaToARGBRow(src_y, src_u, src_v, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+    dst_argb += dst_stride_argb;
+    src_a += src_stride_a;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert I420 with Alpha to ARGB.
+LIBYUV_API
+int I420AlphaToARGB(const uint8_t* src_y,
+                    int src_stride_y,
+                    const uint8_t* src_u,
+                    int src_stride_u,
+                    const uint8_t* src_v,
+                    int src_stride_v,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    int width,
+                    int height,
+                    int attenuate) {
+  return I420AlphaToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                               src_stride_v, src_a, src_stride_a, dst_argb,
+                               dst_stride_argb, &kYuvI601Constants, width,
+                               height, attenuate);
+}
+
+// Convert I420 with Alpha to ABGR.
+LIBYUV_API
+int I420AlphaToABGR(const uint8_t* src_y,
+                    int src_stride_y,
+                    const uint8_t* src_u,
+                    int src_stride_u,
+                    const uint8_t* src_v,
+                    int src_stride_v,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_abgr,
+                    int dst_stride_abgr,
+                    int width,
+                    int height,
+                    int attenuate) {
+  return I420AlphaToARGBMatrix(
+      src_y, src_stride_y, src_v, src_stride_v,  // Swap U and V
+      src_u, src_stride_u, src_a, src_stride_a, dst_abgr, dst_stride_abgr,
+      &kYvuI601Constants,  // Use Yvu matrix
+      width, height, attenuate);
+}
+
+// Convert I422 with Alpha to ARGB.
+LIBYUV_API
+int I422AlphaToARGB(const uint8_t* src_y,
+                    int src_stride_y,
+                    const uint8_t* src_u,
+                    int src_stride_u,
+                    const uint8_t* src_v,
+                    int src_stride_v,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    int width,
+                    int height,
+                    int attenuate) {
+  return I422AlphaToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                               src_stride_v, src_a, src_stride_a, dst_argb,
+                               dst_stride_argb, &kYuvI601Constants, width,
+                               height, attenuate);
+}
+
+// Convert I422 with Alpha to ABGR.
+LIBYUV_API
+int I422AlphaToABGR(const uint8_t* src_y,
+                    int src_stride_y,
+                    const uint8_t* src_u,
+                    int src_stride_u,
+                    const uint8_t* src_v,
+                    int src_stride_v,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_abgr,
+                    int dst_stride_abgr,
+                    int width,
+                    int height,
+                    int attenuate) {
+  return I422AlphaToARGBMatrix(
+      src_y, src_stride_y, src_v, src_stride_v,  // Swap U and V
+      src_u, src_stride_u, src_a, src_stride_a, dst_abgr, dst_stride_abgr,
+      &kYvuI601Constants,  // Use Yvu matrix
+      width, height, attenuate);
+}
+
+// Convert I444 with Alpha to ARGB.
+LIBYUV_API
+int I444AlphaToARGB(const uint8_t* src_y,
+                    int src_stride_y,
+                    const uint8_t* src_u,
+                    int src_stride_u,
+                    const uint8_t* src_v,
+                    int src_stride_v,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    int width,
+                    int height,
+                    int attenuate) {
+  return I444AlphaToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                               src_stride_v, src_a, src_stride_a, dst_argb,
+                               dst_stride_argb, &kYuvI601Constants, width,
+                               height, attenuate);
+}
+
+// Convert I444 with Alpha to ABGR.
+LIBYUV_API
+int I444AlphaToABGR(const uint8_t* src_y,
+                    int src_stride_y,
+                    const uint8_t* src_u,
+                    int src_stride_u,
+                    const uint8_t* src_v,
+                    int src_stride_v,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_abgr,
+                    int dst_stride_abgr,
+                    int width,
+                    int height,
+                    int attenuate) {
+  return I444AlphaToARGBMatrix(
+      src_y, src_stride_y, src_v, src_stride_v,  // Swap U and V
+      src_u, src_stride_u, src_a, src_stride_a, dst_abgr, dst_stride_abgr,
+      &kYvuI601Constants,  // Use Yvu matrix
+      width, height, attenuate);
+}
+
+// Convert I010 with Alpha to preattenuated ARGB with matrix.
+LIBYUV_API
+int I010AlphaToARGBMatrix(const uint16_t* src_y,
+                          int src_stride_y,
+                          const uint16_t* src_u,
+                          int src_stride_u,
+                          const uint16_t* src_v,
+                          int src_stride_v,
+                          const uint16_t* src_a,
+                          int src_stride_a,
+                          uint8_t* dst_argb,
+                          int dst_stride_argb,
+                          const struct YuvConstants* yuvconstants,
+                          int width,
+                          int height,
+                          int attenuate) {
+  int y;
+  void (*I210AlphaToARGBRow)(const uint16_t* y_buf, const uint16_t* u_buf,
+                             const uint16_t* v_buf, const uint16_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width) = I210AlphaToARGBRow_C;
+  void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                           int width) = ARGBAttenuateRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !src_a || !dst_argb || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I210ALPHATOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I210AlphaToARGBRow = I210AlphaToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I210AlphaToARGBRow = I210AlphaToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I210ALPHATOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I210AlphaToARGBRow = I210AlphaToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I210AlphaToARGBRow = I210AlphaToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I210ALPHATOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I210AlphaToARGBRow = I210AlphaToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I210AlphaToARGBRow = I210AlphaToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I210AlphaToARGBRow(src_y, src_u, src_v, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+    dst_argb += dst_stride_argb;
+    src_a += src_stride_a;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert I210 with Alpha to preattenuated ARGB with matrix.
+LIBYUV_API
+int I210AlphaToARGBMatrix(const uint16_t* src_y,
+                          int src_stride_y,
+                          const uint16_t* src_u,
+                          int src_stride_u,
+                          const uint16_t* src_v,
+                          int src_stride_v,
+                          const uint16_t* src_a,
+                          int src_stride_a,
+                          uint8_t* dst_argb,
+                          int dst_stride_argb,
+                          const struct YuvConstants* yuvconstants,
+                          int width,
+                          int height,
+                          int attenuate) {
+  int y;
+  void (*I210AlphaToARGBRow)(const uint16_t* y_buf, const uint16_t* u_buf,
+                             const uint16_t* v_buf, const uint16_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width) = I210AlphaToARGBRow_C;
+  void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                           int width) = ARGBAttenuateRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !src_a || !dst_argb || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I210ALPHATOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I210AlphaToARGBRow = I210AlphaToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I210AlphaToARGBRow = I210AlphaToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I210ALPHATOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I210AlphaToARGBRow = I210AlphaToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I210AlphaToARGBRow = I210AlphaToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I210ALPHATOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I210AlphaToARGBRow = I210AlphaToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I210AlphaToARGBRow = I210AlphaToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I210AlphaToARGBRow(src_y, src_u, src_v, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+    dst_argb += dst_stride_argb;
+    src_a += src_stride_a;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert I410 with Alpha to preattenuated ARGB with matrix.
+LIBYUV_API
+int I410AlphaToARGBMatrix(const uint16_t* src_y,
+                          int src_stride_y,
+                          const uint16_t* src_u,
+                          int src_stride_u,
+                          const uint16_t* src_v,
+                          int src_stride_v,
+                          const uint16_t* src_a,
+                          int src_stride_a,
+                          uint8_t* dst_argb,
+                          int dst_stride_argb,
+                          const struct YuvConstants* yuvconstants,
+                          int width,
+                          int height,
+                          int attenuate) {
+  int y;
+  void (*I410AlphaToARGBRow)(const uint16_t* y_buf, const uint16_t* u_buf,
+                             const uint16_t* v_buf, const uint16_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width) = I410AlphaToARGBRow_C;
+  void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                           int width) = ARGBAttenuateRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !src_a || !dst_argb || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I410ALPHATOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I410AlphaToARGBRow = I410AlphaToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I410AlphaToARGBRow = I410AlphaToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I410ALPHATOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I410AlphaToARGBRow = I410AlphaToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I410AlphaToARGBRow = I410AlphaToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I410ALPHATOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I410AlphaToARGBRow = I410AlphaToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I410AlphaToARGBRow = I410AlphaToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I410AlphaToARGBRow(src_y, src_u, src_v, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+    dst_argb += dst_stride_argb;
+    src_a += src_stride_a;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert I400 to ARGB with matrix.
+LIBYUV_API
+int I400ToARGBMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I400ToARGBRow)(const uint8_t* y_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I400ToARGBRow_C;
+  assert(yuvconstants);
+  if (!src_y || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_y = dst_stride_argb = 0;
+  }
+#if defined(HAS_I400TOARGBROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    I400ToARGBRow = I400ToARGBRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      I400ToARGBRow = I400ToARGBRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_I400TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I400ToARGBRow = I400ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I400ToARGBRow = I400ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I400TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I400ToARGBRow = I400ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I400ToARGBRow = I400ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I400TOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I400ToARGBRow = I400ToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_I400TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I400ToARGBRow = I400ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      I400ToARGBRow = I400ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I400TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I400ToARGBRow = I400ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I400ToARGBRow = I400ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I400TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I400ToARGBRow = I400ToARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I400ToARGBRow(src_y, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+  }
+  return 0;
+}
+
+// Convert I400 to ARGB.
+LIBYUV_API
+int I400ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return I400ToARGBMatrix(src_y, src_stride_y, dst_argb, dst_stride_argb,
+                          &kYuvI601Constants, width, height);
+}
+
+// Convert J400 to ARGB.
+LIBYUV_API
+int J400ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  int y;
+  void (*J400ToARGBRow)(const uint8_t* src_y, uint8_t* dst_argb, int width) =
+      J400ToARGBRow_C;
+  if (!src_y || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_stride_y = -src_stride_y;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_y = dst_stride_argb = 0;
+  }
+#if defined(HAS_J400TOARGBROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    J400ToARGBRow = J400ToARGBRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      J400ToARGBRow = J400ToARGBRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_J400TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    J400ToARGBRow = J400ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      J400ToARGBRow = J400ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_J400TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    J400ToARGBRow = J400ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      J400ToARGBRow = J400ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_J400TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    J400ToARGBRow = J400ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      J400ToARGBRow = J400ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_J400TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    J400ToARGBRow = J400ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      J400ToARGBRow = J400ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_J400TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    J400ToARGBRow = J400ToARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    J400ToARGBRow(src_y, dst_argb, width);
+    src_y += src_stride_y;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+#ifndef __riscv
+// Shuffle table for converting BGRA to ARGB.
+static const uvec8 kShuffleMaskBGRAToARGB = {
+    3u, 2u, 1u, 0u, 7u, 6u, 5u, 4u, 11u, 10u, 9u, 8u, 15u, 14u, 13u, 12u};
+
+// Shuffle table for converting ABGR to ARGB.
+static const uvec8 kShuffleMaskABGRToARGB = {
+    2u, 1u, 0u, 3u, 6u, 5u, 4u, 7u, 10u, 9u, 8u, 11u, 14u, 13u, 12u, 15u};
+
+// Shuffle table for converting RGBA to ARGB.
+static const uvec8 kShuffleMaskRGBAToARGB = {
+    1u, 2u, 3u, 0u, 5u, 6u, 7u, 4u, 9u, 10u, 11u, 8u, 13u, 14u, 15u, 12u};
+
+// Shuffle table for converting AR64 to AB64.
+static const uvec8 kShuffleMaskAR64ToAB64 = {
+    4u, 5u, 2u, 3u, 0u, 1u, 6u, 7u, 12u, 13u, 10u, 11u, 8u, 9u, 14u, 15u};
+
+// Convert BGRA to ARGB.
+LIBYUV_API
+int BGRAToARGB(const uint8_t* src_bgra,
+               int src_stride_bgra,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return ARGBShuffle(src_bgra, src_stride_bgra, dst_argb, dst_stride_argb,
+                     (const uint8_t*)&kShuffleMaskBGRAToARGB, width, height);
+}
+
+// Convert ARGB to BGRA (same as BGRAToARGB).
+LIBYUV_API
+int ARGBToBGRA(const uint8_t* src_bgra,
+               int src_stride_bgra,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return ARGBShuffle(src_bgra, src_stride_bgra, dst_argb, dst_stride_argb,
+                     (const uint8_t*)&kShuffleMaskBGRAToARGB, width, height);
+}
+
+// Convert ABGR to ARGB.
+LIBYUV_API
+int ABGRToARGB(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return ARGBShuffle(src_abgr, src_stride_abgr, dst_argb, dst_stride_argb,
+                     (const uint8_t*)&kShuffleMaskABGRToARGB, width, height);
+}
+
+// Convert ARGB to ABGR to (same as ABGRToARGB).
+LIBYUV_API
+int ARGBToABGR(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return ARGBShuffle(src_abgr, src_stride_abgr, dst_argb, dst_stride_argb,
+                     (const uint8_t*)&kShuffleMaskABGRToARGB, width, height);
+}
+
+// Convert RGBA to ARGB.
+LIBYUV_API
+int RGBAToARGB(const uint8_t* src_rgba,
+               int src_stride_rgba,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return ARGBShuffle(src_rgba, src_stride_rgba, dst_argb, dst_stride_argb,
+                     (const uint8_t*)&kShuffleMaskRGBAToARGB, width, height);
+}
+
+// Convert AR64 To AB64.
+LIBYUV_API
+int AR64ToAB64(const uint16_t* src_ar64,
+               int src_stride_ar64,
+               uint16_t* dst_ab64,
+               int dst_stride_ab64,
+               int width,
+               int height) {
+  return AR64Shuffle(src_ar64, src_stride_ar64, dst_ab64, dst_stride_ab64,
+                     (const uint8_t*)&kShuffleMaskAR64ToAB64, width, height);
+}
+#else
+// Convert BGRA to ARGB (same as ARGBToBGRA).
+LIBYUV_API
+int BGRAToARGB(const uint8_t* src_bgra,
+               int src_stride_bgra,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return ARGBToBGRA(src_bgra, src_stride_bgra, dst_argb, dst_stride_argb, width,
+                    height);
+}
+
+// Convert ARGB to BGRA.
+LIBYUV_API
+int ARGBToBGRA(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_bgra,
+               int dst_stride_bgra,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToBGRARow)(const uint8_t* src_argb, uint8_t* dst_bgra, int width) =
+      ARGBToBGRARow_C;
+  if (!src_argb || !dst_bgra || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_bgra == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_bgra = 0;
+  }
+
+#if defined(HAS_ARGBTOBGRAROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToBGRARow = ARGBToBGRARow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToBGRARow(src_argb, dst_bgra, width);
+    src_argb += src_stride_argb;
+    dst_bgra += dst_stride_bgra;
+  }
+  return 0;
+}
+
+// Convert ARGB to ABGR.
+LIBYUV_API
+int ARGBToABGR(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToABGRRow)(const uint8_t* src_argb, uint8_t* dst_abgr, int width) =
+      ARGBToABGRRow_C;
+  if (!src_argb || !dst_abgr || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_abgr == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_abgr = 0;
+  }
+
+#if defined(HAS_ARGBTOABGRROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToABGRRow = ARGBToABGRRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToABGRRow(src_argb, dst_abgr, width);
+    src_argb += src_stride_argb;
+    dst_abgr += dst_stride_abgr;
+  }
+  return 0;
+}
+
+// Convert ABGR to ARGB (same as ARGBToABGR).
+LIBYUV_API
+int ABGRToARGB(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return ARGBToABGR(src_abgr, src_stride_abgr, dst_argb, dst_stride_argb, width,
+                    height);
+}
+
+// Convert RGBA to ARGB.
+LIBYUV_API
+int RGBAToARGB(const uint8_t* src_rgba,
+               int src_stride_rgba,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  int y;
+  void (*RGBAToARGBRow)(const uint8_t* src_rgba, uint8_t* dst_argb, int width) =
+      RGBAToARGBRow_C;
+  if (!src_rgba || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_rgba = src_rgba + (height - 1) * src_stride_rgba;
+    src_stride_rgba = -src_stride_rgba;
+  }
+  // Coalesce rows.
+  if (src_stride_rgba == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_rgba = dst_stride_argb = 0;
+  }
+
+#if defined(HAS_RGBATOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RGBAToARGBRow = RGBAToARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    RGBAToARGBRow(src_rgba, dst_argb, width);
+    src_rgba += src_stride_rgba;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert AR64 To AB64.
+LIBYUV_API
+int AR64ToAB64(const uint16_t* src_ar64,
+               int src_stride_ar64,
+               uint16_t* dst_ab64,
+               int dst_stride_ab64,
+               int width,
+               int height) {
+  int y;
+  void (*AR64ToAB64Row)(const uint16_t* src_ar64, uint16_t* dst_ab64,
+                        int width) = AR64ToAB64Row_C;
+  if (!src_ar64 || !dst_ab64 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_ar64 = src_ar64 + (height - 1) * src_stride_ar64;
+    src_stride_ar64 = -src_stride_ar64;
+  }
+  // Coalesce rows.
+  if (src_stride_ar64 == width * 4 && dst_stride_ab64 == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_ar64 = dst_stride_ab64 = 0;
+  }
+
+#if defined(HAS_AR64TOAB64ROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    AR64ToAB64Row = AR64ToAB64Row_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    AR64ToAB64Row(src_ar64, dst_ab64, width);
+    src_ar64 += src_stride_ar64;
+    dst_ab64 += dst_stride_ab64;
+  }
+  return 0;
+}
+#endif
+
+// Convert RGB24 to ARGB.
+LIBYUV_API
+int RGB24ToARGB(const uint8_t* src_rgb24,
+                int src_stride_rgb24,
+                uint8_t* dst_argb,
+                int dst_stride_argb,
+                int width,
+                int height) {
+  int y;
+  void (*RGB24ToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) =
+      RGB24ToARGBRow_C;
+  if (!src_rgb24 || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_rgb24 = src_rgb24 + (height - 1) * src_stride_rgb24;
+    src_stride_rgb24 = -src_stride_rgb24;
+  }
+  // Coalesce rows.
+  if (src_stride_rgb24 == width * 3 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_rgb24 = dst_stride_argb = 0;
+  }
+#if defined(HAS_RGB24TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RGB24ToARGBRow = RGB24ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToARGBRow = RGB24ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RGB24ToARGBRow = RGB24ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      RGB24ToARGBRow = RGB24ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    RGB24ToARGBRow = RGB24ToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_RGB24TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RGB24ToARGBRow = RGB24ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToARGBRow = RGB24ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RGB24ToARGBRow = RGB24ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24ToARGBRow = RGB24ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    RGB24ToARGBRow = RGB24ToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RGB24ToARGBRow = RGB24ToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_RGB24TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RGB24ToARGBRow = RGB24ToARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    RGB24ToARGBRow(src_rgb24, dst_argb, width);
+    src_rgb24 += src_stride_rgb24;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert RAW to ARGB.
+LIBYUV_API
+int RAWToARGB(const uint8_t* src_raw,
+              int src_stride_raw,
+              uint8_t* dst_argb,
+              int dst_stride_argb,
+              int width,
+              int height) {
+  int y;
+  void (*RAWToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) =
+      RAWToARGBRow_C;
+  if (!src_raw || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_raw = src_raw + (height - 1) * src_stride_raw;
+    src_stride_raw = -src_stride_raw;
+  }
+  // Coalesce rows.
+  if (src_stride_raw == width * 3 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_raw = dst_stride_argb = 0;
+  }
+#if defined(HAS_RAWTOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RAWToARGBRow = RAWToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToARGBRow = RAWToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RAWToARGBRow = RAWToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      RAWToARGBRow = RAWToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    RAWToARGBRow = RAWToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_RAWTOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RAWToARGBRow = RAWToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToARGBRow = RAWToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RAWToARGBRow = RAWToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToARGBRow = RAWToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    RAWToARGBRow = RAWToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RAWToARGBRow = RAWToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RAWToARGBRow = RAWToARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    RAWToARGBRow(src_raw, dst_argb, width);
+    src_raw += src_stride_raw;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert RAW to RGBA.
+LIBYUV_API
+int RAWToRGBA(const uint8_t* src_raw,
+              int src_stride_raw,
+              uint8_t* dst_rgba,
+              int dst_stride_rgba,
+              int width,
+              int height) {
+  int y;
+  void (*RAWToRGBARow)(const uint8_t* src_rgb, uint8_t* dst_rgba, int width) =
+      RAWToRGBARow_C;
+  if (!src_raw || !dst_rgba || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_raw = src_raw + (height - 1) * src_stride_raw;
+    src_stride_raw = -src_stride_raw;
+  }
+  // Coalesce rows.
+  if (src_stride_raw == width * 3 && dst_stride_rgba == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_raw = dst_stride_rgba = 0;
+  }
+#if defined(HAS_RAWTORGBAROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RAWToRGBARow = RAWToRGBARow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToRGBARow = RAWToRGBARow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_RAWTORGBAROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RAWToRGBARow = RAWToRGBARow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      RAWToRGBARow = RAWToRGBARow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RAWTORGBAROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    RAWToRGBARow = RAWToRGBARow_SVE2;
+  }
+#endif
+#if defined(HAS_RAWTORGBAROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RAWToRGBARow = RAWToRGBARow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    RAWToRGBARow(src_raw, dst_rgba, width);
+    src_raw += src_stride_raw;
+    dst_rgba += dst_stride_rgba;
+  }
+  return 0;
+}
+
+// Convert RGB565 to ARGB.
+LIBYUV_API
+int RGB565ToARGB(const uint8_t* src_rgb565,
+                 int src_stride_rgb565,
+                 uint8_t* dst_argb,
+                 int dst_stride_argb,
+                 int width,
+                 int height) {
+  int y;
+  void (*RGB565ToARGBRow)(const uint8_t* src_rgb565, uint8_t* dst_argb,
+                          int width) = RGB565ToARGBRow_C;
+  if (!src_rgb565 || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_rgb565 = src_rgb565 + (height - 1) * src_stride_rgb565;
+    src_stride_rgb565 = -src_stride_rgb565;
+  }
+  // Coalesce rows.
+  if (src_stride_rgb565 == width * 2 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_rgb565 = dst_stride_argb = 0;
+  }
+#if defined(HAS_RGB565TOARGBROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    RGB565ToARGBRow = RGB565ToARGBRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      RGB565ToARGBRow = RGB565ToARGBRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_RGB565TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    RGB565ToARGBRow = RGB565ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      RGB565ToARGBRow = RGB565ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_RGB565TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RGB565ToARGBRow = RGB565ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      RGB565ToARGBRow = RGB565ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RGB565TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RGB565ToARGBRow = RGB565ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RGB565ToARGBRow = RGB565ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RGB565TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RGB565ToARGBRow = RGB565ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RGB565ToARGBRow = RGB565ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RGB565TOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    RGB565ToARGBRow = RGB565ToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RGB565ToARGBRow = RGB565ToARGBRow_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    RGB565ToARGBRow(src_rgb565, dst_argb, width);
+    src_rgb565 += src_stride_rgb565;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert ARGB1555 to ARGB.
+LIBYUV_API
+int ARGB1555ToARGB(const uint8_t* src_argb1555,
+                   int src_stride_argb1555,
+                   uint8_t* dst_argb,
+                   int dst_stride_argb,
+                   int width,
+                   int height) {
+  int y;
+  void (*ARGB1555ToARGBRow)(const uint8_t* src_argb1555, uint8_t* dst_argb,
+                            int width) = ARGB1555ToARGBRow_C;
+  if (!src_argb1555 || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb1555 = src_argb1555 + (height - 1) * src_stride_argb1555;
+    src_stride_argb1555 = -src_stride_argb1555;
+  }
+  // Coalesce rows.
+  if (src_stride_argb1555 == width * 2 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb1555 = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGB1555TOARGBROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGB1555ToARGBRow = ARGB1555ToARGBRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGB1555TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB1555ToARGBRow = ARGB1555ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGB1555TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGB1555ToARGBRow = ARGB1555ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGB1555TOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGB1555ToARGBRow = ARGB1555ToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_ARGB1555TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB1555ToARGBRow = ARGB1555ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGB1555TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB1555ToARGBRow = ARGB1555ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGB1555TOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGB1555ToARGBRow = ARGB1555ToARGBRow_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGB1555ToARGBRow(src_argb1555, dst_argb, width);
+    src_argb1555 += src_stride_argb1555;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert ARGB4444 to ARGB.
+LIBYUV_API
+int ARGB4444ToARGB(const uint8_t* src_argb4444,
+                   int src_stride_argb4444,
+                   uint8_t* dst_argb,
+                   int dst_stride_argb,
+                   int width,
+                   int height) {
+  int y;
+  void (*ARGB4444ToARGBRow)(const uint8_t* src_argb4444, uint8_t* dst_argb,
+                            int width) = ARGB4444ToARGBRow_C;
+  if (!src_argb4444 || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb4444 = src_argb4444 + (height - 1) * src_stride_argb4444;
+    src_stride_argb4444 = -src_stride_argb4444;
+  }
+  // Coalesce rows.
+  if (src_stride_argb4444 == width * 2 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb4444 = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGB4444TOARGBROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGB4444ToARGBRow = ARGB4444ToARGBRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGB4444TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB4444ToARGBRow = ARGB4444ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGB4444TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGB4444ToARGBRow = ARGB4444ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGB4444TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB4444ToARGBRow = ARGB4444ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGB4444TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB4444ToARGBRow = ARGB4444ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGB4444TOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGB4444ToARGBRow = ARGB4444ToARGBRow_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGB4444ToARGBRow(src_argb4444, dst_argb, width);
+    src_argb4444 += src_stride_argb4444;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert AR30 to ARGB.
+LIBYUV_API
+int AR30ToARGB(const uint8_t* src_ar30,
+               int src_stride_ar30,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  int y;
+  if (!src_ar30 || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_ar30 = src_ar30 + (height - 1) * src_stride_ar30;
+    src_stride_ar30 = -src_stride_ar30;
+  }
+  // Coalesce rows.
+  if (src_stride_ar30 == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_ar30 = dst_stride_argb = 0;
+  }
+  for (y = 0; y < height; ++y) {
+    AR30ToARGBRow_C(src_ar30, dst_argb, width);
+    src_ar30 += src_stride_ar30;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert AR30 to ABGR.
+LIBYUV_API
+int AR30ToABGR(const uint8_t* src_ar30,
+               int src_stride_ar30,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  int y;
+  if (!src_ar30 || !dst_abgr || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_ar30 = src_ar30 + (height - 1) * src_stride_ar30;
+    src_stride_ar30 = -src_stride_ar30;
+  }
+  // Coalesce rows.
+  if (src_stride_ar30 == width * 4 && dst_stride_abgr == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_ar30 = dst_stride_abgr = 0;
+  }
+  for (y = 0; y < height; ++y) {
+    AR30ToABGRRow_C(src_ar30, dst_abgr, width);
+    src_ar30 += src_stride_ar30;
+    dst_abgr += dst_stride_abgr;
+  }
+  return 0;
+}
+
+// Convert AR30 to AB30.
+LIBYUV_API
+int AR30ToAB30(const uint8_t* src_ar30,
+               int src_stride_ar30,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height) {
+  int y;
+  if (!src_ar30 || !dst_ab30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_ar30 = src_ar30 + (height - 1) * src_stride_ar30;
+    src_stride_ar30 = -src_stride_ar30;
+  }
+  // Coalesce rows.
+  if (src_stride_ar30 == width * 4 && dst_stride_ab30 == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_ar30 = dst_stride_ab30 = 0;
+  }
+  for (y = 0; y < height; ++y) {
+    AR30ToAB30Row_C(src_ar30, dst_ab30, width);
+    src_ar30 += src_stride_ar30;
+    dst_ab30 += dst_stride_ab30;
+  }
+  return 0;
+}
+
+// Convert AR64 to ARGB.
+LIBYUV_API
+int AR64ToARGB(const uint16_t* src_ar64,
+               int src_stride_ar64,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  int y;
+  void (*AR64ToARGBRow)(const uint16_t* src_ar64, uint8_t* dst_argb,
+                        int width) = AR64ToARGBRow_C;
+  if (!src_ar64 || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_ar64 = src_ar64 + (height - 1) * src_stride_ar64;
+    src_stride_ar64 = -src_stride_ar64;
+  }
+  // Coalesce rows.
+  if (src_stride_ar64 == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_ar64 = dst_stride_argb = 0;
+  }
+#if defined(HAS_AR64TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    AR64ToARGBRow = AR64ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      AR64ToARGBRow = AR64ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_AR64TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    AR64ToARGBRow = AR64ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      AR64ToARGBRow = AR64ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_AR64TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    AR64ToARGBRow = AR64ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      AR64ToARGBRow = AR64ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_AR64TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    AR64ToARGBRow = AR64ToARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    AR64ToARGBRow(src_ar64, dst_argb, width);
+    src_ar64 += src_stride_ar64;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert AB64 to ARGB.
+LIBYUV_API
+int AB64ToARGB(const uint16_t* src_ab64,
+               int src_stride_ab64,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  int y;
+  void (*AB64ToARGBRow)(const uint16_t* src_ar64, uint8_t* dst_argb,
+                        int width) = AB64ToARGBRow_C;
+  if (!src_ab64 || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_ab64 = src_ab64 + (height - 1) * src_stride_ab64;
+    src_stride_ab64 = -src_stride_ab64;
+  }
+  // Coalesce rows.
+  if (src_stride_ab64 == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_ab64 = dst_stride_argb = 0;
+  }
+#if defined(HAS_AB64TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    AB64ToARGBRow = AB64ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      AB64ToARGBRow = AB64ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_AB64TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    AB64ToARGBRow = AB64ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      AB64ToARGBRow = AB64ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_AB64TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    AB64ToARGBRow = AB64ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      AB64ToARGBRow = AB64ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_AB64TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    AB64ToARGBRow = AB64ToARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    AB64ToARGBRow(src_ab64, dst_argb, width);
+    src_ab64 += src_stride_ab64;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert NV12 to ARGB with matrix.
+LIBYUV_API
+int NV12ToARGBMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_uv,
+                     int src_stride_uv,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*NV12ToARGBRow)(
+      const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = NV12ToARGBRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_uv || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_NV12TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    NV12ToARGBRow = NV12ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      NV12ToARGBRow = NV12ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_NV12TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    NV12ToARGBRow = NV12ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      NV12ToARGBRow = NV12ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_NV12TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    NV12ToARGBRow = NV12ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      NV12ToARGBRow = NV12ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_NV12TOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    NV12ToARGBRow = NV12ToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_NV12TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    NV12ToARGBRow = NV12ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      NV12ToARGBRow = NV12ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_NV12TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    NV12ToARGBRow = NV12ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      NV12ToARGBRow = NV12ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_NV12TOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    NV12ToARGBRow = NV12ToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      NV12ToARGBRow = NV12ToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_NV12TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    NV12ToARGBRow = NV12ToARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    NV12ToARGBRow(src_y, src_uv, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_uv += src_stride_uv;
+    }
+  }
+  return 0;
+}
+
+// Convert NV21 to ARGB with matrix.
+LIBYUV_API
+int NV21ToARGBMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_vu,
+                     int src_stride_vu,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*NV21ToARGBRow)(
+      const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = NV21ToARGBRow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_vu || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_NV21TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    NV21ToARGBRow = NV21ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      NV21ToARGBRow = NV21ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_NV21TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    NV21ToARGBRow = NV21ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      NV21ToARGBRow = NV21ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_NV21TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    NV21ToARGBRow = NV21ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      NV21ToARGBRow = NV21ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_NV21TOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    NV21ToARGBRow = NV21ToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_NV21TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    NV21ToARGBRow = NV21ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      NV21ToARGBRow = NV21ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_NV21TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    NV21ToARGBRow = NV21ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      NV21ToARGBRow = NV21ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_NV21TOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    NV21ToARGBRow = NV21ToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      NV21ToARGBRow = NV21ToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_NV21TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    NV21ToARGBRow = NV21ToARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    NV21ToARGBRow(src_y, src_vu, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_vu += src_stride_vu;
+    }
+  }
+  return 0;
+}
+
+// Convert NV12 to ARGB.
+LIBYUV_API
+int NV12ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return NV12ToARGBMatrix(src_y, src_stride_y, src_uv, src_stride_uv, dst_argb,
+                          dst_stride_argb, &kYuvI601Constants, width, height);
+}
+
+// Convert NV21 to ARGB.
+LIBYUV_API
+int NV21ToARGB(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_vu,
+               int src_stride_vu,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return NV21ToARGBMatrix(src_y, src_stride_y, src_vu, src_stride_vu, dst_argb,
+                          dst_stride_argb, &kYuvI601Constants, width, height);
+}
+
+// Convert NV12 to ABGR.
+// To output ABGR instead of ARGB swap the UV and use a mirrored yuv matrix.
+// To swap the UV use NV12 instead of NV21.LIBYUV_API
+LIBYUV_API
+int NV12ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return NV21ToARGBMatrix(src_y, src_stride_y, src_uv, src_stride_uv, dst_abgr,
+                          dst_stride_abgr, &kYvuI601Constants, width, height);
+}
+
+// Convert NV21 to ABGR.
+LIBYUV_API
+int NV21ToABGR(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_vu,
+               int src_stride_vu,
+               uint8_t* dst_abgr,
+               int dst_stride_abgr,
+               int width,
+               int height) {
+  return NV12ToARGBMatrix(src_y, src_stride_y, src_vu, src_stride_vu, dst_abgr,
+                          dst_stride_abgr, &kYvuI601Constants, width, height);
+}
+
+// TODO(fbarchard): Consider SSSE3 2 step conversion.
+// Convert NV12 to RGB24 with matrix.
+LIBYUV_API
+int NV12ToRGB24Matrix(const uint8_t* src_y,
+                      int src_stride_y,
+                      const uint8_t* src_uv,
+                      int src_stride_uv,
+                      uint8_t* dst_rgb24,
+                      int dst_stride_rgb24,
+                      const struct YuvConstants* yuvconstants,
+                      int width,
+                      int height) {
+  int y;
+  void (*NV12ToRGB24Row)(
+      const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = NV12ToRGB24Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_uv || !dst_rgb24 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb24 = dst_rgb24 + (height - 1) * dst_stride_rgb24;
+    dst_stride_rgb24 = -dst_stride_rgb24;
+  }
+#if defined(HAS_NV12TORGB24ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    NV12ToRGB24Row = NV12ToRGB24Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      NV12ToRGB24Row = NV12ToRGB24Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_NV12TORGB24ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    NV12ToRGB24Row = NV12ToRGB24Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      NV12ToRGB24Row = NV12ToRGB24Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_NV12TORGB24ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    NV12ToRGB24Row = NV12ToRGB24Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      NV12ToRGB24Row = NV12ToRGB24Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_NV12TORGB24ROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    NV12ToRGB24Row = NV12ToRGB24Row_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    NV12ToRGB24Row(src_y, src_uv, dst_rgb24, yuvconstants, width);
+    dst_rgb24 += dst_stride_rgb24;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_uv += src_stride_uv;
+    }
+  }
+  return 0;
+}
+
+// Convert NV21 to RGB24 with matrix.
+LIBYUV_API
+int NV21ToRGB24Matrix(const uint8_t* src_y,
+                      int src_stride_y,
+                      const uint8_t* src_vu,
+                      int src_stride_vu,
+                      uint8_t* dst_rgb24,
+                      int dst_stride_rgb24,
+                      const struct YuvConstants* yuvconstants,
+                      int width,
+                      int height) {
+  int y;
+  void (*NV21ToRGB24Row)(
+      const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = NV21ToRGB24Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_vu || !dst_rgb24 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb24 = dst_rgb24 + (height - 1) * dst_stride_rgb24;
+    dst_stride_rgb24 = -dst_stride_rgb24;
+  }
+#if defined(HAS_NV21TORGB24ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    NV21ToRGB24Row = NV21ToRGB24Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      NV21ToRGB24Row = NV21ToRGB24Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_NV21TORGB24ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    NV21ToRGB24Row = NV21ToRGB24Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      NV21ToRGB24Row = NV21ToRGB24Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_NV21TORGB24ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    NV21ToRGB24Row = NV21ToRGB24Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      NV21ToRGB24Row = NV21ToRGB24Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_NV21TORGB24ROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    NV21ToRGB24Row = NV21ToRGB24Row_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    NV21ToRGB24Row(src_y, src_vu, dst_rgb24, yuvconstants, width);
+    dst_rgb24 += dst_stride_rgb24;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_vu += src_stride_vu;
+    }
+  }
+  return 0;
+}
+
+// Convert NV12 to RGB24.
+LIBYUV_API
+int NV12ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_uv,
+                int src_stride_uv,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height) {
+  return NV12ToRGB24Matrix(src_y, src_stride_y, src_uv, src_stride_uv,
+                           dst_rgb24, dst_stride_rgb24, &kYuvI601Constants,
+                           width, height);
+}
+
+// Convert NV21 to RGB24.
+LIBYUV_API
+int NV21ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_vu,
+                int src_stride_vu,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height) {
+  return NV21ToRGB24Matrix(src_y, src_stride_y, src_vu, src_stride_vu,
+                           dst_rgb24, dst_stride_rgb24, &kYuvI601Constants,
+                           width, height);
+}
+
+// Convert NV12 to RAW.
+LIBYUV_API
+int NV12ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_uv,
+              int src_stride_uv,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height) {
+  return NV21ToRGB24Matrix(src_y, src_stride_y, src_uv, src_stride_uv, dst_raw,
+                           dst_stride_raw, &kYvuI601Constants, width, height);
+}
+
+// Convert NV21 to RAW.
+LIBYUV_API
+int NV21ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_vu,
+              int src_stride_vu,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height) {
+  return NV12ToRGB24Matrix(src_y, src_stride_y, src_vu, src_stride_vu, dst_raw,
+                           dst_stride_raw, &kYvuI601Constants, width, height);
+}
+
+// Convert NV21 to YUV24
+int NV21ToYUV24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_vu,
+                int src_stride_vu,
+                uint8_t* dst_yuv24,
+                int dst_stride_yuv24,
+                int width,
+                int height) {
+  int y;
+  void (*NV21ToYUV24Row)(const uint8_t* src_y, const uint8_t* src_vu,
+                         uint8_t* dst_yuv24, int width) = NV21ToYUV24Row_C;
+  if (!src_y || !src_vu || !dst_yuv24 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_yuv24 = dst_yuv24 + (height - 1) * dst_stride_yuv24;
+    dst_stride_yuv24 = -dst_stride_yuv24;
+  }
+#if defined(HAS_NV21TOYUV24ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    NV21ToYUV24Row = NV21ToYUV24Row_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      NV21ToYUV24Row = NV21ToYUV24Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_NV21TOYUV24ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    NV21ToYUV24Row = NV21ToYUV24Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      NV21ToYUV24Row = NV21ToYUV24Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_NV21TOYUV24ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    NV21ToYUV24Row = NV21ToYUV24Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      NV21ToYUV24Row = NV21ToYUV24Row_AVX2;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    NV21ToYUV24Row(src_y, src_vu, dst_yuv24, width);
+    dst_yuv24 += dst_stride_yuv24;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_vu += src_stride_vu;
+    }
+  }
+  return 0;
+}
+
+// Convert YUY2 to ARGB with matrix.
+LIBYUV_API
+int YUY2ToARGBMatrix(const uint8_t* src_yuy2,
+                     int src_stride_yuy2,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*YUY2ToARGBRow)(const uint8_t* src_yuy2, uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants, int width) =
+      YUY2ToARGBRow_C;
+  if (!src_yuy2 || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_yuy2 = src_yuy2 + (height - 1) * src_stride_yuy2;
+    src_stride_yuy2 = -src_stride_yuy2;
+  }
+  // Coalesce rows.
+  if (src_stride_yuy2 == width * 2 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_yuy2 = dst_stride_argb = 0;
+  }
+#if defined(HAS_YUY2TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    YUY2ToARGBRow = YUY2ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToARGBRow = YUY2ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    YUY2ToARGBRow = YUY2ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToARGBRow = YUY2ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    YUY2ToARGBRow = YUY2ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      YUY2ToARGBRow = YUY2ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    YUY2ToARGBRow = YUY2ToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_YUY2TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    YUY2ToARGBRow = YUY2ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      YUY2ToARGBRow = YUY2ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    YUY2ToARGBRow = YUY2ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      YUY2ToARGBRow = YUY2ToARGBRow_LSX;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    YUY2ToARGBRow(src_yuy2, dst_argb, yuvconstants, width);
+    src_yuy2 += src_stride_yuy2;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert YUY2 to ARGB.
+LIBYUV_API
+int YUY2ToARGB(const uint8_t* src_yuy2,
+               int src_stride_yuy2,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return YUY2ToARGBMatrix(src_yuy2, src_stride_yuy2, dst_argb, dst_stride_argb,
+                          &kYuvI601Constants, width, height);
+}
+
+// Convert UYVY to ARGB with matrix.
+LIBYUV_API
+int UYVYToARGBMatrix(const uint8_t* src_uyvy,
+                     int src_stride_uyvy,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*UYVYToARGBRow)(const uint8_t* src_uyvy, uint8_t* dst_argb,
+                        const struct YuvConstants* yuvconstants, int width) =
+      UYVYToARGBRow_C;
+  if (!src_uyvy || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_uyvy = src_uyvy + (height - 1) * src_stride_uyvy;
+    src_stride_uyvy = -src_stride_uyvy;
+  }
+  // Coalesce rows.
+  if (src_stride_uyvy == width * 2 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_uyvy = dst_stride_argb = 0;
+  }
+#if defined(HAS_UYVYTOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    UYVYToARGBRow = UYVYToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      UYVYToARGBRow = UYVYToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    UYVYToARGBRow = UYVYToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      UYVYToARGBRow = UYVYToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    UYVYToARGBRow = UYVYToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      UYVYToARGBRow = UYVYToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    UYVYToARGBRow = UYVYToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_UYVYTOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    UYVYToARGBRow = UYVYToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      UYVYToARGBRow = UYVYToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    UYVYToARGBRow = UYVYToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      UYVYToARGBRow = UYVYToARGBRow_LSX;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    UYVYToARGBRow(src_uyvy, dst_argb, yuvconstants, width);
+    src_uyvy += src_stride_uyvy;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert UYVY to ARGB.
+LIBYUV_API
+int UYVYToARGB(const uint8_t* src_uyvy,
+               int src_stride_uyvy,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  return UYVYToARGBMatrix(src_uyvy, src_stride_uyvy, dst_argb, dst_stride_argb,
+                          &kYuvI601Constants, width, height);
+}
+
+static void WeavePixels(const uint8_t* src_u,
+                        const uint8_t* src_v,
+                        int src_pixel_stride_uv,
+                        uint8_t* dst_uv,
+                        int width) {
+  int i;
+  for (i = 0; i < width; ++i) {
+    dst_uv[0] = *src_u;
+    dst_uv[1] = *src_v;
+    dst_uv += 2;
+    src_u += src_pixel_stride_uv;
+    src_v += src_pixel_stride_uv;
+  }
+}
+
+// Convert Android420 to ARGB with matrix.
+LIBYUV_API
+int Android420ToARGBMatrix(const uint8_t* src_y,
+                           int src_stride_y,
+                           const uint8_t* src_u,
+                           int src_stride_u,
+                           const uint8_t* src_v,
+                           int src_stride_v,
+                           int src_pixel_stride_uv,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height) {
+  int y;
+  uint8_t* dst_uv;
+  const ptrdiff_t vu_off = src_v - src_u;
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+
+  // I420
+  if (src_pixel_stride_uv == 1) {
+    return I420ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                            src_stride_v, dst_argb, dst_stride_argb,
+                            yuvconstants, width, height);
+    // NV21
+  }
+  if (src_pixel_stride_uv == 2 && vu_off == -1 &&
+      src_stride_u == src_stride_v) {
+    return NV21ToARGBMatrix(src_y, src_stride_y, src_v, src_stride_v, dst_argb,
+                            dst_stride_argb, yuvconstants, width, height);
+    // NV12
+  }
+  if (src_pixel_stride_uv == 2 && vu_off == 1 && src_stride_u == src_stride_v) {
+    return NV12ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, dst_argb,
+                            dst_stride_argb, yuvconstants, width, height);
+  }
+
+  // General case fallback creates NV12
+  align_buffer_64(plane_uv, halfwidth * 2 * halfheight);
+  if (!plane_uv)
+    return 1;
+  dst_uv = plane_uv;
+  for (y = 0; y < halfheight; ++y) {
+    WeavePixels(src_u, src_v, src_pixel_stride_uv, dst_uv, halfwidth);
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+    dst_uv += halfwidth * 2;
+  }
+  NV12ToARGBMatrix(src_y, src_stride_y, plane_uv, halfwidth * 2, dst_argb,
+                   dst_stride_argb, yuvconstants, width, height);
+  free_aligned_buffer_64(plane_uv);
+  return 0;
+}
+
+// Convert Android420 to ARGB.
+LIBYUV_API
+int Android420ToARGB(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     int src_pixel_stride_uv,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     int width,
+                     int height) {
+  return Android420ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                                src_stride_v, src_pixel_stride_uv, dst_argb,
+                                dst_stride_argb, &kYuvI601Constants, width,
+                                height);
+}
+
+// Convert Android420 to ABGR.
+LIBYUV_API
+int Android420ToABGR(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     int src_pixel_stride_uv,
+                     uint8_t* dst_abgr,
+                     int dst_stride_abgr,
+                     int width,
+                     int height) {
+  return Android420ToARGBMatrix(src_y, src_stride_y, src_v, src_stride_v, src_u,
+                                src_stride_u, src_pixel_stride_uv, dst_abgr,
+                                dst_stride_abgr, &kYvuI601Constants, width,
+                                height);
+}
+
+// Convert I422 to RGBA with matrix.
+LIBYUV_API
+int I422ToRGBAMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_rgba,
+                     int dst_stride_rgba,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I422ToRGBARow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                        const uint8_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I422ToRGBARow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_rgba || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgba = dst_rgba + (height - 1) * dst_stride_rgba;
+    dst_stride_rgba = -dst_stride_rgba;
+  }
+#if defined(HAS_I422TORGBAROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToRGBARow = I422ToRGBARow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGBARow = I422ToRGBARow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToRGBARow = I422ToRGBARow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGBARow = I422ToRGBARow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToRGBARow = I422ToRGBARow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGBARow = I422ToRGBARow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I422ToRGBARow = I422ToRGBARow_SVE2;
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToRGBARow = I422ToRGBARow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGBARow = I422ToRGBARow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToRGBARow = I422ToRGBARow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGBARow = I422ToRGBARow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToRGBARow = I422ToRGBARow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToRGBARow = I422ToRGBARow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I422ToRGBARow = I422ToRGBARow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToRGBARow(src_y, src_u, src_v, dst_rgba, yuvconstants, width);
+    dst_rgba += dst_stride_rgba;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert I422 to RGBA.
+LIBYUV_API
+int I422ToRGBA(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_rgba,
+               int dst_stride_rgba,
+               int width,
+               int height) {
+  return I422ToRGBAMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_rgba, dst_stride_rgba,
+                          &kYuvI601Constants, width, height);
+}
+
+// Convert I422 to BGRA.
+LIBYUV_API
+int I422ToBGRA(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_bgra,
+               int dst_stride_bgra,
+               int width,
+               int height) {
+  return I422ToRGBAMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_bgra, dst_stride_bgra,
+                          &kYvuI601Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert NV12 to RGB565 with matrix.
+LIBYUV_API
+int NV12ToRGB565Matrix(const uint8_t* src_y,
+                       int src_stride_y,
+                       const uint8_t* src_uv,
+                       int src_stride_uv,
+                       uint8_t* dst_rgb565,
+                       int dst_stride_rgb565,
+                       const struct YuvConstants* yuvconstants,
+                       int width,
+                       int height) {
+  int y;
+  void (*NV12ToRGB565Row)(
+      const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = NV12ToRGB565Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_uv || !dst_rgb565 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb565 = dst_rgb565 + (height - 1) * dst_stride_rgb565;
+    dst_stride_rgb565 = -dst_stride_rgb565;
+  }
+#if defined(HAS_NV12TORGB565ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    NV12ToRGB565Row = NV12ToRGB565Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      NV12ToRGB565Row = NV12ToRGB565Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_NV12TORGB565ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    NV12ToRGB565Row = NV12ToRGB565Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      NV12ToRGB565Row = NV12ToRGB565Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_NV12TORGB565ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    NV12ToRGB565Row = NV12ToRGB565Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      NV12ToRGB565Row = NV12ToRGB565Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_NV12TORGB565ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    NV12ToRGB565Row = NV12ToRGB565Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      NV12ToRGB565Row = NV12ToRGB565Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_NV12TORGB565ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    NV12ToRGB565Row = NV12ToRGB565Row_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      NV12ToRGB565Row = NV12ToRGB565Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_NV12TORGB565ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    NV12ToRGB565Row = NV12ToRGB565Row_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      NV12ToRGB565Row = NV12ToRGB565Row_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    NV12ToRGB565Row(src_y, src_uv, dst_rgb565, yuvconstants, width);
+    dst_rgb565 += dst_stride_rgb565;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_uv += src_stride_uv;
+    }
+  }
+  return 0;
+}
+
+// Convert NV12 to RGB565.
+LIBYUV_API
+int NV12ToRGB565(const uint8_t* src_y,
+                 int src_stride_y,
+                 const uint8_t* src_uv,
+                 int src_stride_uv,
+                 uint8_t* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height) {
+  return NV12ToRGB565Matrix(src_y, src_stride_y, src_uv, src_stride_uv,
+                            dst_rgb565, dst_stride_rgb565, &kYuvI601Constants,
+                            width, height);
+}
+
+// Convert I422 to RGBA with matrix.
+LIBYUV_API
+int I420ToRGBAMatrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_rgba,
+                     int dst_stride_rgba,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I422ToRGBARow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                        const uint8_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I422ToRGBARow_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_rgba || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgba = dst_rgba + (height - 1) * dst_stride_rgba;
+    dst_stride_rgba = -dst_stride_rgba;
+  }
+#if defined(HAS_I422TORGBAROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToRGBARow = I422ToRGBARow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGBARow = I422ToRGBARow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToRGBARow = I422ToRGBARow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGBARow = I422ToRGBARow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToRGBARow = I422ToRGBARow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGBARow = I422ToRGBARow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I422ToRGBARow = I422ToRGBARow_SVE2;
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToRGBARow = I422ToRGBARow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGBARow = I422ToRGBARow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToRGBARow = I422ToRGBARow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGBARow = I422ToRGBARow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToRGBARow = I422ToRGBARow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToRGBARow = I422ToRGBARow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGBAROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I422ToRGBARow = I422ToRGBARow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToRGBARow(src_y, src_u, src_v, dst_rgba, yuvconstants, width);
+    dst_rgba += dst_stride_rgba;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert I420 to RGBA.
+LIBYUV_API
+int I420ToRGBA(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_rgba,
+               int dst_stride_rgba,
+               int width,
+               int height) {
+  return I420ToRGBAMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_rgba, dst_stride_rgba,
+                          &kYuvI601Constants, width, height);
+}
+
+// Convert I420 to BGRA.
+LIBYUV_API
+int I420ToBGRA(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_bgra,
+               int dst_stride_bgra,
+               int width,
+               int height) {
+  return I420ToRGBAMatrix(src_y, src_stride_y, src_v,
+                          src_stride_v,  // Swap U and V
+                          src_u, src_stride_u, dst_bgra, dst_stride_bgra,
+                          &kYvuI601Constants,  // Use Yvu matrix
+                          width, height);
+}
+
+// Convert I420 to RGB24 with matrix.
+LIBYUV_API
+int I420ToRGB24Matrix(const uint8_t* src_y,
+                      int src_stride_y,
+                      const uint8_t* src_u,
+                      int src_stride_u,
+                      const uint8_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_rgb24,
+                      int dst_stride_rgb24,
+                      const struct YuvConstants* yuvconstants,
+                      int width,
+                      int height) {
+  int y;
+  void (*I422ToRGB24Row)(const uint8_t* y_buf, const uint8_t* u_buf,
+                         const uint8_t* v_buf, uint8_t* rgb_buf,
+                         const struct YuvConstants* yuvconstants, int width) =
+      I422ToRGB24Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_rgb24 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb24 = dst_rgb24 + (height - 1) * dst_stride_rgb24;
+    dst_stride_rgb24 = -dst_stride_rgb24;
+  }
+#if defined(HAS_I422TORGB24ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGB24Row = I422ToRGB24Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB24ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToRGB24Row = I422ToRGB24Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB24ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGB24Row = I422ToRGB24Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB24ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGB24Row = I422ToRGB24Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB24ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGB24Row = I422ToRGB24Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB24ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToRGB24Row = I422ToRGB24Row_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB24ROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I422ToRGB24Row = I422ToRGB24Row_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToRGB24Row(src_y, src_u, src_v, dst_rgb24, yuvconstants, width);
+    dst_rgb24 += dst_stride_rgb24;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert I420 to RGB24.
+LIBYUV_API
+int I420ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_u,
+                int src_stride_u,
+                const uint8_t* src_v,
+                int src_stride_v,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height) {
+  return I420ToRGB24Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                           src_stride_v, dst_rgb24, dst_stride_rgb24,
+                           &kYuvI601Constants, width, height);
+}
+
+// Convert I420 to RAW.
+LIBYUV_API
+int I420ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height) {
+  return I420ToRGB24Matrix(src_y, src_stride_y, src_v,
+                           src_stride_v,  // Swap U and V
+                           src_u, src_stride_u, dst_raw, dst_stride_raw,
+                           &kYvuI601Constants,  // Use Yvu matrix
+                           width, height);
+}
+
+// Convert J420 to RGB24.
+LIBYUV_API
+int J420ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_u,
+                int src_stride_u,
+                const uint8_t* src_v,
+                int src_stride_v,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height) {
+  return I420ToRGB24Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                           src_stride_v, dst_rgb24, dst_stride_rgb24,
+                           &kYuvJPEGConstants, width, height);
+}
+
+// Convert J420 to RAW.
+LIBYUV_API
+int J420ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height) {
+  return I420ToRGB24Matrix(src_y, src_stride_y, src_v,
+                           src_stride_v,  // Swap U and V
+                           src_u, src_stride_u, dst_raw, dst_stride_raw,
+                           &kYvuJPEGConstants,  // Use Yvu matrix
+                           width, height);
+}
+
+// Convert H420 to RGB24.
+LIBYUV_API
+int H420ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_u,
+                int src_stride_u,
+                const uint8_t* src_v,
+                int src_stride_v,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height) {
+  return I420ToRGB24Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                           src_stride_v, dst_rgb24, dst_stride_rgb24,
+                           &kYuvH709Constants, width, height);
+}
+
+// Convert H420 to RAW.
+LIBYUV_API
+int H420ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height) {
+  return I420ToRGB24Matrix(src_y, src_stride_y, src_v,
+                           src_stride_v,  // Swap U and V
+                           src_u, src_stride_u, dst_raw, dst_stride_raw,
+                           &kYvuH709Constants,  // Use Yvu matrix
+                           width, height);
+}
+
+// Convert I422 to RGB24 with matrix.
+LIBYUV_API
+int I422ToRGB24Matrix(const uint8_t* src_y,
+                      int src_stride_y,
+                      const uint8_t* src_u,
+                      int src_stride_u,
+                      const uint8_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_rgb24,
+                      int dst_stride_rgb24,
+                      const struct YuvConstants* yuvconstants,
+                      int width,
+                      int height) {
+  int y;
+  void (*I422ToRGB24Row)(const uint8_t* y_buf, const uint8_t* u_buf,
+                         const uint8_t* v_buf, uint8_t* rgb_buf,
+                         const struct YuvConstants* yuvconstants, int width) =
+      I422ToRGB24Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_rgb24 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb24 = dst_rgb24 + (height - 1) * dst_stride_rgb24;
+    dst_stride_rgb24 = -dst_stride_rgb24;
+  }
+#if defined(HAS_I422TORGB24ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGB24Row = I422ToRGB24Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB24ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToRGB24Row = I422ToRGB24Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB24ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGB24Row = I422ToRGB24Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB24ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGB24Row = I422ToRGB24Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB24ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGB24Row = I422ToRGB24Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB24ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToRGB24Row = I422ToRGB24Row_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB24ROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I422ToRGB24Row = I422ToRGB24Row_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToRGB24Row(src_y, src_u, src_v, dst_rgb24, yuvconstants, width);
+    dst_rgb24 += dst_stride_rgb24;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert I422 to RGB24.
+LIBYUV_API
+int I422ToRGB24(const uint8_t* src_y,
+                int src_stride_y,
+                const uint8_t* src_u,
+                int src_stride_u,
+                const uint8_t* src_v,
+                int src_stride_v,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height) {
+  return I422ToRGB24Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                           src_stride_v, dst_rgb24, dst_stride_rgb24,
+                           &kYuvI601Constants, width, height);
+}
+
+// Convert I422 to RAW.
+LIBYUV_API
+int I422ToRAW(const uint8_t* src_y,
+              int src_stride_y,
+              const uint8_t* src_u,
+              int src_stride_u,
+              const uint8_t* src_v,
+              int src_stride_v,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height) {
+  return I422ToRGB24Matrix(src_y, src_stride_y, src_v,
+                           src_stride_v,  // Swap U and V
+                           src_u, src_stride_u, dst_raw, dst_stride_raw,
+                           &kYvuI601Constants,  // Use Yvu matrix
+                           width, height);
+}
+
+// Convert I420 to ARGB1555.
+LIBYUV_API
+int I420ToARGB1555(const uint8_t* src_y,
+                   int src_stride_y,
+                   const uint8_t* src_u,
+                   int src_stride_u,
+                   const uint8_t* src_v,
+                   int src_stride_v,
+                   uint8_t* dst_argb1555,
+                   int dst_stride_argb1555,
+                   int width,
+                   int height) {
+  int y;
+  void (*I422ToARGB1555Row)(const uint8_t* y_buf, const uint8_t* u_buf,
+                            const uint8_t* v_buf, uint8_t* rgb_buf,
+                            const struct YuvConstants* yuvconstants,
+                            int width) = I422ToARGB1555Row_C;
+  if (!src_y || !src_u || !src_v || !dst_argb1555 || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb1555 = dst_argb1555 + (height - 1) * dst_stride_argb1555;
+    dst_stride_argb1555 = -dst_stride_argb1555;
+  }
+#if defined(HAS_I422TOARGB1555ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToARGB1555Row = I422ToARGB1555Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGB1555Row = I422ToARGB1555Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGB1555ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToARGB1555Row = I422ToARGB1555Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToARGB1555Row = I422ToARGB1555Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGB1555ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToARGB1555Row = I422ToARGB1555Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGB1555Row = I422ToARGB1555Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGB1555ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToARGB1555Row = I422ToARGB1555Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGB1555Row = I422ToARGB1555Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGB1555ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToARGB1555Row = I422ToARGB1555Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToARGB1555Row = I422ToARGB1555Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGB1555ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToARGB1555Row = I422ToARGB1555Row_Any_LASX;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGB1555Row = I422ToARGB1555Row_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToARGB1555Row(src_y, src_u, src_v, dst_argb1555, &kYuvI601Constants,
+                      width);
+    dst_argb1555 += dst_stride_argb1555;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert I420 to ARGB4444.
+LIBYUV_API
+int I420ToARGB4444(const uint8_t* src_y,
+                   int src_stride_y,
+                   const uint8_t* src_u,
+                   int src_stride_u,
+                   const uint8_t* src_v,
+                   int src_stride_v,
+                   uint8_t* dst_argb4444,
+                   int dst_stride_argb4444,
+                   int width,
+                   int height) {
+  int y;
+  void (*I422ToARGB4444Row)(const uint8_t* y_buf, const uint8_t* u_buf,
+                            const uint8_t* v_buf, uint8_t* rgb_buf,
+                            const struct YuvConstants* yuvconstants,
+                            int width) = I422ToARGB4444Row_C;
+  if (!src_y || !src_u || !src_v || !dst_argb4444 || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb4444 = dst_argb4444 + (height - 1) * dst_stride_argb4444;
+    dst_stride_argb4444 = -dst_stride_argb4444;
+  }
+#if defined(HAS_I422TOARGB4444ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToARGB4444Row = I422ToARGB4444Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGB4444Row = I422ToARGB4444Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGB4444ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToARGB4444Row = I422ToARGB4444Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToARGB4444Row = I422ToARGB4444Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGB4444ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToARGB4444Row = I422ToARGB4444Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGB4444Row = I422ToARGB4444Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGB4444ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToARGB4444Row = I422ToARGB4444Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGB4444Row = I422ToARGB4444Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGB4444ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToARGB4444Row = I422ToARGB4444Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToARGB4444Row = I422ToARGB4444Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGB4444ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToARGB4444Row = I422ToARGB4444Row_Any_LASX;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGB4444Row = I422ToARGB4444Row_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToARGB4444Row(src_y, src_u, src_v, dst_argb4444, &kYuvI601Constants,
+                      width);
+    dst_argb4444 += dst_stride_argb4444;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert I420 to RGB565 with specified color matrix.
+LIBYUV_API
+int I420ToRGB565Matrix(const uint8_t* src_y,
+                       int src_stride_y,
+                       const uint8_t* src_u,
+                       int src_stride_u,
+                       const uint8_t* src_v,
+                       int src_stride_v,
+                       uint8_t* dst_rgb565,
+                       int dst_stride_rgb565,
+                       const struct YuvConstants* yuvconstants,
+                       int width,
+                       int height) {
+  int y;
+  void (*I422ToRGB565Row)(const uint8_t* y_buf, const uint8_t* u_buf,
+                          const uint8_t* v_buf, uint8_t* rgb_buf,
+                          const struct YuvConstants* yuvconstants, int width) =
+      I422ToRGB565Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_rgb565 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb565 = dst_rgb565 + (height - 1) * dst_stride_rgb565;
+    dst_stride_rgb565 = -dst_stride_rgb565;
+  }
+#if defined(HAS_I422TORGB565ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGB565Row = I422ToRGB565Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGB565Row = I422ToRGB565Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGB565Row = I422ToRGB565Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGB565Row = I422ToRGB565Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGB565Row = I422ToRGB565Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToRGB565Row = I422ToRGB565Row_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToRGB565Row(src_y, src_u, src_v, dst_rgb565, yuvconstants, width);
+    dst_rgb565 += dst_stride_rgb565;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert I420 to RGB565.
+LIBYUV_API
+int I420ToRGB565(const uint8_t* src_y,
+                 int src_stride_y,
+                 const uint8_t* src_u,
+                 int src_stride_u,
+                 const uint8_t* src_v,
+                 int src_stride_v,
+                 uint8_t* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height) {
+  return I420ToRGB565Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                            src_stride_v, dst_rgb565, dst_stride_rgb565,
+                            &kYuvI601Constants, width, height);
+}
+
+// Convert J420 to RGB565.
+LIBYUV_API
+int J420ToRGB565(const uint8_t* src_y,
+                 int src_stride_y,
+                 const uint8_t* src_u,
+                 int src_stride_u,
+                 const uint8_t* src_v,
+                 int src_stride_v,
+                 uint8_t* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height) {
+  return I420ToRGB565Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                            src_stride_v, dst_rgb565, dst_stride_rgb565,
+                            &kYuvJPEGConstants, width, height);
+}
+
+// Convert H420 to RGB565.
+LIBYUV_API
+int H420ToRGB565(const uint8_t* src_y,
+                 int src_stride_y,
+                 const uint8_t* src_u,
+                 int src_stride_u,
+                 const uint8_t* src_v,
+                 int src_stride_v,
+                 uint8_t* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height) {
+  return I420ToRGB565Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                            src_stride_v, dst_rgb565, dst_stride_rgb565,
+                            &kYuvH709Constants, width, height);
+}
+
+// Convert I422 to RGB565 with specified color matrix.
+LIBYUV_API
+int I422ToRGB565Matrix(const uint8_t* src_y,
+                       int src_stride_y,
+                       const uint8_t* src_u,
+                       int src_stride_u,
+                       const uint8_t* src_v,
+                       int src_stride_v,
+                       uint8_t* dst_rgb565,
+                       int dst_stride_rgb565,
+                       const struct YuvConstants* yuvconstants,
+                       int width,
+                       int height) {
+  int y;
+  void (*I422ToRGB565Row)(const uint8_t* y_buf, const uint8_t* u_buf,
+                          const uint8_t* v_buf, uint8_t* rgb_buf,
+                          const struct YuvConstants* yuvconstants, int width) =
+      I422ToRGB565Row_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_rgb565 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb565 = dst_rgb565 + (height - 1) * dst_stride_rgb565;
+    dst_stride_rgb565 = -dst_stride_rgb565;
+  }
+#if defined(HAS_I422TORGB565ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGB565Row = I422ToRGB565Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGB565Row = I422ToRGB565Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGB565Row = I422ToRGB565Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGB565Row = I422ToRGB565Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGB565Row = I422ToRGB565Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToRGB565Row = I422ToRGB565Row_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToRGB565Row(src_y, src_u, src_v, dst_rgb565, yuvconstants, width);
+    dst_rgb565 += dst_stride_rgb565;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert I422 to RGB565.
+LIBYUV_API
+int I422ToRGB565(const uint8_t* src_y,
+                 int src_stride_y,
+                 const uint8_t* src_u,
+                 int src_stride_u,
+                 const uint8_t* src_v,
+                 int src_stride_v,
+                 uint8_t* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height) {
+  return I422ToRGB565Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                            src_stride_v, dst_rgb565, dst_stride_rgb565,
+                            &kYuvI601Constants, width, height);
+}
+
+// Ordered 8x8 dither for 888 to 565.  Values from 0 to 7.
+static const uint8_t kDither565_4x4[16] = {
+    0, 4, 1, 5, 6, 2, 7, 3, 1, 5, 0, 4, 7, 3, 6, 2,
+};
+
+// Convert I420 to RGB565 with dithering.
+LIBYUV_API
+int I420ToRGB565Dither(const uint8_t* src_y,
+                       int src_stride_y,
+                       const uint8_t* src_u,
+                       int src_stride_u,
+                       const uint8_t* src_v,
+                       int src_stride_v,
+                       uint8_t* dst_rgb565,
+                       int dst_stride_rgb565,
+                       const uint8_t* dither4x4,
+                       int width,
+                       int height) {
+  int y;
+  void (*I422ToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                        const uint8_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I422ToARGBRow_C;
+  void (*ARGBToRGB565DitherRow)(const uint8_t* src_argb, uint8_t* dst_rgb,
+                                uint32_t dither4, int width) =
+      ARGBToRGB565DitherRow_C;
+  if (!src_y || !src_u || !src_v || !dst_rgb565 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb565 = dst_rgb565 + (height - 1) * dst_stride_rgb565;
+    dst_stride_rgb565 = -dst_stride_rgb565;
+  }
+  if (!dither4x4) {
+    dither4x4 = kDither565_4x4;
+  }
+#if defined(HAS_I422TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToARGBRow = I422ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGBRow = I422ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToARGBRow = I422ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToARGBRow = I422ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW | kCpuHasAVX512VL) ==
+      (kCpuHasAVX512BW | kCpuHasAVX512VL)) {
+    I422ToARGBRow = I422ToARGBRow_Any_AVX512BW;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToARGBRow = I422ToARGBRow_AVX512BW;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToARGBRow = I422ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGBRow = I422ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I422ToARGBRow = I422ToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToARGBRow = I422ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGBRow = I422ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToARGBRow = I422ToARGBRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToARGBRow = I422ToARGBRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToARGBRow = I422ToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToARGBRow = I422ToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I422ToARGBRow = I422ToARGBRow_RVV;
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_SSE2;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_SVE2;
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_LASX;
+    }
+  }
+#endif
+  {
+    // Allocate a row of argb.
+    align_buffer_64(row_argb, width * 4);
+    if (!row_argb)
+      return 1;
+    for (y = 0; y < height; ++y) {
+      I422ToARGBRow(src_y, src_u, src_v, row_argb, &kYuvI601Constants, width);
+      ARGBToRGB565DitherRow(row_argb, dst_rgb565,
+                            *(const uint32_t*)(dither4x4 + ((y & 3) << 2)),
+                            width);
+      dst_rgb565 += dst_stride_rgb565;
+      src_y += src_stride_y;
+      if (y & 1) {
+        src_u += src_stride_u;
+        src_v += src_stride_v;
+      }
+    }
+    free_aligned_buffer_64(row_argb);
+  }
+  return 0;
+}
+
+// Convert I420 to AR30 with matrix.
+LIBYUV_API
+int I420ToAR30Matrix(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_u,
+                     int src_stride_u,
+                     const uint8_t* src_v,
+                     int src_stride_v,
+                     uint8_t* dst_ar30,
+                     int dst_stride_ar30,
+                     const struct YuvConstants* yuvconstants,
+                     int width,
+                     int height) {
+  int y;
+  void (*I422ToAR30Row)(const uint8_t* y_buf, const uint8_t* u_buf,
+                        const uint8_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I422ToAR30Row_C;
+
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar30 = dst_ar30 + (height - 1) * dst_stride_ar30;
+    dst_stride_ar30 = -dst_stride_ar30;
+  }
+
+#if defined(HAS_I422TOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToAR30Row = I422ToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToAR30Row = I422ToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToAR30Row = I422ToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToAR30Row = I422ToAR30Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToAR30Row = I422ToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToAR30Row = I422ToAR30Row_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToAR30Row(src_y, src_u, src_v, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert I420 to AR30.
+LIBYUV_API
+int I420ToAR30(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height) {
+  return I420ToAR30Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_ar30, dst_stride_ar30,
+                          &kYuvI601Constants, width, height);
+}
+
+// Convert H420 to AR30.
+LIBYUV_API
+int H420ToAR30(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height) {
+  return I420ToAR30Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                          src_stride_v, dst_ar30, dst_stride_ar30,
+                          &kYvuH709Constants, width, height);
+}
+
+// Convert I420 to AB30.
+LIBYUV_API
+int I420ToAB30(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height) {
+  return I420ToAR30Matrix(src_y, src_stride_y, src_v, src_stride_v, src_u,
+                          src_stride_u, dst_ab30, dst_stride_ab30,
+                          &kYvuI601Constants, width, height);
+}
+
+// Convert H420 to AB30.
+LIBYUV_API
+int H420ToAB30(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_ab30,
+               int dst_stride_ab30,
+               int width,
+               int height) {
+  return I420ToAR30Matrix(src_y, src_stride_y, src_v, src_stride_v, src_u,
+                          src_stride_u, dst_ab30, dst_stride_ab30,
+                          &kYvuH709Constants, width, height);
+}
+
+static int I420ToARGBMatrixBilinear(const uint8_t* src_y,
+                                    int src_stride_y,
+                                    const uint8_t* src_u,
+                                    int src_stride_u,
+                                    const uint8_t* src_v,
+                                    int src_stride_v,
+                                    uint8_t* dst_argb,
+                                    int dst_stride_argb,
+                                    const struct YuvConstants* yuvconstants,
+                                    int width,
+                                    int height) {
+  int y;
+  void (*I444ToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                        const uint8_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I444ToARGBRow_C;
+  void (*Scale2RowUp_Bilinear)(const uint8_t* src_ptr, ptrdiff_t src_stride,
+                               uint8_t* dst_ptr, ptrdiff_t dst_stride,
+                               int dst_width) = ScaleRowUp2_Bilinear_Any_C;
+  void (*ScaleRowUp2_Linear)(const uint8_t* src_ptr, uint8_t* dst_ptr,
+                             int dst_width) = ScaleRowUp2_Linear_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I444TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I444ToARGBRow = I444ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToARGBRow = I444ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I444ToARGBRow = I444ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I444ToARGBRow = I444ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I444ToARGBRow = I444ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToARGBRow = I444ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I444ToARGBRow = I444ToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I444ToARGBRow = I444ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToARGBRow = I444ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I444ToARGBRow = I444ToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I444ToARGBRow = I444ToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I444ToARGBRow = I444ToARGBRow_RVV;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_Any_SSE2;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_SSE2;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_Any_SSSE3;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_SSSE3;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_Any_AVX2;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_AVX2;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_Any_NEON;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_NEON;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_BILINEAR_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_RVV;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_RVV;
+  }
+#endif
+
+  // alloc 4 lines temp
+  const int row_size = (width + 31) & ~31;
+  align_buffer_64(row, row_size * 4);
+  uint8_t* temp_u_1 = row;
+  uint8_t* temp_u_2 = row + row_size;
+  uint8_t* temp_v_1 = row + row_size * 2;
+  uint8_t* temp_v_2 = row + row_size * 3;
+  if (!row)
+    return 1;
+
+  ScaleRowUp2_Linear(src_u, temp_u_1, width);
+  ScaleRowUp2_Linear(src_v, temp_v_1, width);
+  I444ToARGBRow(src_y, temp_u_1, temp_v_1, dst_argb, yuvconstants, width);
+  dst_argb += dst_stride_argb;
+  src_y += src_stride_y;
+
+  for (y = 0; y < height - 2; y += 2) {
+    Scale2RowUp_Bilinear(src_u, src_stride_u, temp_u_1, row_size, width);
+    Scale2RowUp_Bilinear(src_v, src_stride_v, temp_v_1, row_size, width);
+    I444ToARGBRow(src_y, temp_u_1, temp_v_1, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    I444ToARGBRow(src_y, temp_u_2, temp_v_2, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+
+  if (!(height & 1)) {
+    ScaleRowUp2_Linear(src_u, temp_u_1, width);
+    ScaleRowUp2_Linear(src_v, temp_v_1, width);
+    I444ToARGBRow(src_y, temp_u_1, temp_v_1, dst_argb, yuvconstants, width);
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int I422ToARGBMatrixLinear(const uint8_t* src_y,
+                                  int src_stride_y,
+                                  const uint8_t* src_u,
+                                  int src_stride_u,
+                                  const uint8_t* src_v,
+                                  int src_stride_v,
+                                  uint8_t* dst_argb,
+                                  int dst_stride_argb,
+                                  const struct YuvConstants* yuvconstants,
+                                  int width,
+                                  int height) {
+  int y;
+  void (*I444ToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                        const uint8_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I444ToARGBRow_C;
+  void (*ScaleRowUp2_Linear)(const uint8_t* src_ptr, uint8_t* dst_ptr,
+                             int dst_width) = ScaleRowUp2_Linear_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I444TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I444ToARGBRow = I444ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToARGBRow = I444ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I444ToARGBRow = I444ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I444ToARGBRow = I444ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I444ToARGBRow = I444ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToARGBRow = I444ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I444ToARGBRow = I444ToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I444ToARGBRow = I444ToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToARGBRow = I444ToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I444ToARGBRow = I444ToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I444ToARGBRow = I444ToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I444TOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I444ToARGBRow = I444ToARGBRow_RVV;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_SSE2;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_SSSE3;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_AVX2;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_NEON;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_RVV;
+  }
+#endif
+
+  // alloc 2 lines temp
+  const int row_size = (width + 31) & ~31;
+  align_buffer_64(row, row_size * 2);
+  uint8_t* temp_u = row;
+  uint8_t* temp_v = row + row_size;
+  if (!row)
+    return 1;
+
+  for (y = 0; y < height; ++y) {
+    ScaleRowUp2_Linear(src_u, temp_u, width);
+    ScaleRowUp2_Linear(src_v, temp_v, width);
+    I444ToARGBRow(src_y, temp_u, temp_v, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int I420ToRGB24MatrixBilinear(const uint8_t* src_y,
+                                     int src_stride_y,
+                                     const uint8_t* src_u,
+                                     int src_stride_u,
+                                     const uint8_t* src_v,
+                                     int src_stride_v,
+                                     uint8_t* dst_rgb24,
+                                     int dst_stride_rgb24,
+                                     const struct YuvConstants* yuvconstants,
+                                     int width,
+                                     int height) {
+  int y;
+  void (*I444ToRGB24Row)(const uint8_t* y_buf, const uint8_t* u_buf,
+                         const uint8_t* v_buf, uint8_t* rgb_buf,
+                         const struct YuvConstants* yuvconstants, int width) =
+      I444ToRGB24Row_C;
+  void (*Scale2RowUp_Bilinear)(const uint8_t* src_ptr, ptrdiff_t src_stride,
+                               uint8_t* dst_ptr, ptrdiff_t dst_stride,
+                               int dst_width) = ScaleRowUp2_Bilinear_Any_C;
+  void (*ScaleRowUp2_Linear)(const uint8_t* src_ptr, uint8_t* dst_ptr,
+                             int dst_width) = ScaleRowUp2_Linear_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_rgb24 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb24 = dst_rgb24 + (height - 1) * dst_stride_rgb24;
+    dst_stride_rgb24 = -dst_stride_rgb24;
+  }
+#if defined(HAS_I444TORGB24ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      I444ToRGB24Row = I444ToRGB24Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      I444ToRGB24Row = I444ToRGB24Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToRGB24Row = I444ToRGB24Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToRGB24Row = I444ToRGB24Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I444ToRGB24Row = I444ToRGB24Row_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I444ToRGB24Row = I444ToRGB24Row_RVV;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_Any_SSE2;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_SSE2;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_Any_SSSE3;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_SSSE3;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_Any_AVX2;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_AVX2;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_Any_NEON;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_NEON;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_BILINEAR_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_RVV;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_RVV;
+  }
+#endif
+
+  // alloc 4 lines temp
+  const int row_size = (width + 31) & ~31;
+  align_buffer_64(row, row_size * 4);
+  uint8_t* temp_u_1 = row;
+  uint8_t* temp_u_2 = row + row_size;
+  uint8_t* temp_v_1 = row + row_size * 2;
+  uint8_t* temp_v_2 = row + row_size * 3;
+  if (!row)
+    return 1;
+
+  ScaleRowUp2_Linear(src_u, temp_u_1, width);
+  ScaleRowUp2_Linear(src_v, temp_v_1, width);
+  I444ToRGB24Row(src_y, temp_u_1, temp_v_1, dst_rgb24, yuvconstants, width);
+  dst_rgb24 += dst_stride_rgb24;
+  src_y += src_stride_y;
+
+  for (y = 0; y < height - 2; y += 2) {
+    Scale2RowUp_Bilinear(src_u, src_stride_u, temp_u_1, row_size, width);
+    Scale2RowUp_Bilinear(src_v, src_stride_v, temp_v_1, row_size, width);
+    I444ToRGB24Row(src_y, temp_u_1, temp_v_1, dst_rgb24, yuvconstants, width);
+    dst_rgb24 += dst_stride_rgb24;
+    src_y += src_stride_y;
+    I444ToRGB24Row(src_y, temp_u_2, temp_v_2, dst_rgb24, yuvconstants, width);
+    dst_rgb24 += dst_stride_rgb24;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+
+  if (!(height & 1)) {
+    ScaleRowUp2_Linear(src_u, temp_u_1, width);
+    ScaleRowUp2_Linear(src_v, temp_v_1, width);
+    I444ToRGB24Row(src_y, temp_u_1, temp_v_1, dst_rgb24, yuvconstants, width);
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int I010ToAR30MatrixBilinear(const uint16_t* src_y,
+                                    int src_stride_y,
+                                    const uint16_t* src_u,
+                                    int src_stride_u,
+                                    const uint16_t* src_v,
+                                    int src_stride_v,
+                                    uint8_t* dst_ar30,
+                                    int dst_stride_ar30,
+                                    const struct YuvConstants* yuvconstants,
+                                    int width,
+                                    int height) {
+  int y;
+  void (*I410ToAR30Row)(const uint16_t* y_buf, const uint16_t* u_buf,
+                        const uint16_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I410ToAR30Row_C;
+  void (*Scale2RowUp_Bilinear_12)(
+      const uint16_t* src_ptr, ptrdiff_t src_stride, uint16_t* dst_ptr,
+      ptrdiff_t dst_stride, int dst_width) = ScaleRowUp2_Bilinear_16_Any_C;
+  void (*ScaleRowUp2_Linear_12)(const uint16_t* src_ptr, uint16_t* dst_ptr,
+                                int dst_width) = ScaleRowUp2_Linear_16_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar30 = dst_ar30 + (height - 1) * dst_stride_ar30;
+    dst_stride_ar30 = -dst_stride_ar30;
+  }
+#if defined(HAS_I410TOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I410ToAR30Row = I410ToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I410ToAR30Row = I410ToAR30Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I410TOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I410ToAR30Row = I410ToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I410ToAR30Row = I410ToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I410TOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I410ToAR30Row = I410ToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I410ToAR30Row = I410ToAR30Row_AVX2;
+    }
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_12_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    Scale2RowUp_Bilinear_12 = ScaleRowUp2_Bilinear_12_Any_SSSE3;
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_SSSE3;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_12_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    Scale2RowUp_Bilinear_12 = ScaleRowUp2_Bilinear_12_Any_AVX2;
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_AVX2;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_12_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    Scale2RowUp_Bilinear_12 = ScaleRowUp2_Bilinear_12_Any_NEON;
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_NEON;
+  }
+#endif
+
+  // alloc 4 lines temp
+  const int row_size = (width + 31) & ~31;
+  align_buffer_64(row, row_size * 4 * sizeof(uint16_t));
+  uint16_t* temp_u_1 = (uint16_t*)(row);
+  uint16_t* temp_u_2 = (uint16_t*)(row) + row_size;
+  uint16_t* temp_v_1 = (uint16_t*)(row) + row_size * 2;
+  uint16_t* temp_v_2 = (uint16_t*)(row) + row_size * 3;
+  if (!row)
+    return 1;
+
+  ScaleRowUp2_Linear_12(src_u, temp_u_1, width);
+  ScaleRowUp2_Linear_12(src_v, temp_v_1, width);
+  I410ToAR30Row(src_y, temp_u_1, temp_v_1, dst_ar30, yuvconstants, width);
+  dst_ar30 += dst_stride_ar30;
+  src_y += src_stride_y;
+
+  for (y = 0; y < height - 2; y += 2) {
+    Scale2RowUp_Bilinear_12(src_u, src_stride_u, temp_u_1, row_size, width);
+    Scale2RowUp_Bilinear_12(src_v, src_stride_v, temp_v_1, row_size, width);
+    I410ToAR30Row(src_y, temp_u_1, temp_v_1, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    I410ToAR30Row(src_y, temp_u_2, temp_v_2, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+
+  if (!(height & 1)) {
+    ScaleRowUp2_Linear_12(src_u, temp_u_1, width);
+    ScaleRowUp2_Linear_12(src_v, temp_v_1, width);
+    I410ToAR30Row(src_y, temp_u_1, temp_v_1, dst_ar30, yuvconstants, width);
+  }
+
+  free_aligned_buffer_64(row);
+
+  return 0;
+}
+
+static int I210ToAR30MatrixLinear(const uint16_t* src_y,
+                                  int src_stride_y,
+                                  const uint16_t* src_u,
+                                  int src_stride_u,
+                                  const uint16_t* src_v,
+                                  int src_stride_v,
+                                  uint8_t* dst_ar30,
+                                  int dst_stride_ar30,
+                                  const struct YuvConstants* yuvconstants,
+                                  int width,
+                                  int height) {
+  int y;
+  void (*I410ToAR30Row)(const uint16_t* y_buf, const uint16_t* u_buf,
+                        const uint16_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I410ToAR30Row_C;
+  void (*ScaleRowUp2_Linear_12)(const uint16_t* src_ptr, uint16_t* dst_ptr,
+                                int dst_width) = ScaleRowUp2_Linear_16_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar30 = dst_ar30 + (height - 1) * dst_stride_ar30;
+    dst_stride_ar30 = -dst_stride_ar30;
+  }
+#if defined(HAS_I410TOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I410ToAR30Row = I410ToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I410ToAR30Row = I410ToAR30Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I410TOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I410ToAR30Row = I410ToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I410ToAR30Row = I410ToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I410TOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I410ToAR30Row = I410ToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I410ToAR30Row = I410ToAR30Row_AVX2;
+    }
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_LINEAR_12_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_SSSE3;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_12_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_AVX2;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_12_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_NEON;
+  }
+#endif
+
+  // alloc 2 lines temp
+  const int row_size = (width + 31) & ~31;
+  align_buffer_64(row, row_size * 2 * sizeof(uint16_t));
+  uint16_t* temp_u = (uint16_t*)(row);
+  uint16_t* temp_v = (uint16_t*)(row) + row_size;
+  if (!row)
+    return 1;
+
+  for (y = 0; y < height; ++y) {
+    ScaleRowUp2_Linear_12(src_u, temp_u, width);
+    ScaleRowUp2_Linear_12(src_v, temp_v, width);
+    I410ToAR30Row(src_y, temp_u, temp_v, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int I010ToARGBMatrixBilinear(const uint16_t* src_y,
+                                    int src_stride_y,
+                                    const uint16_t* src_u,
+                                    int src_stride_u,
+                                    const uint16_t* src_v,
+                                    int src_stride_v,
+                                    uint8_t* dst_argb,
+                                    int dst_stride_argb,
+                                    const struct YuvConstants* yuvconstants,
+                                    int width,
+                                    int height) {
+  int y;
+  void (*I410ToARGBRow)(const uint16_t* y_buf, const uint16_t* u_buf,
+                        const uint16_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I410ToARGBRow_C;
+  void (*Scale2RowUp_Bilinear_12)(
+      const uint16_t* src_ptr, ptrdiff_t src_stride, uint16_t* dst_ptr,
+      ptrdiff_t dst_stride, int dst_width) = ScaleRowUp2_Bilinear_16_Any_C;
+  void (*ScaleRowUp2_Linear_12)(const uint16_t* src_ptr, uint16_t* dst_ptr,
+                                int dst_width) = ScaleRowUp2_Linear_16_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I410TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I410ToARGBRow = I410ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I410ToARGBRow = I410ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I410TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I410ToARGBRow = I410ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I410ToARGBRow = I410ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I410TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I410ToARGBRow = I410ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I410ToARGBRow = I410ToARGBRow_AVX2;
+    }
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_12_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    Scale2RowUp_Bilinear_12 = ScaleRowUp2_Bilinear_12_Any_SSSE3;
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_SSSE3;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_12_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    Scale2RowUp_Bilinear_12 = ScaleRowUp2_Bilinear_12_Any_AVX2;
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_AVX2;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_12_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    Scale2RowUp_Bilinear_12 = ScaleRowUp2_Bilinear_12_Any_NEON;
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_NEON;
+  }
+#endif
+
+  // alloc 4 lines temp
+  const int row_size = (width + 31) & ~31;
+  align_buffer_64(row, row_size * 4 * sizeof(uint16_t));
+  uint16_t* temp_u_1 = (uint16_t*)(row);
+  uint16_t* temp_u_2 = (uint16_t*)(row) + row_size;
+  uint16_t* temp_v_1 = (uint16_t*)(row) + row_size * 2;
+  uint16_t* temp_v_2 = (uint16_t*)(row) + row_size * 3;
+  if (!row)
+    return 1;
+
+  ScaleRowUp2_Linear_12(src_u, temp_u_1, width);
+  ScaleRowUp2_Linear_12(src_v, temp_v_1, width);
+  I410ToARGBRow(src_y, temp_u_1, temp_v_1, dst_argb, yuvconstants, width);
+  dst_argb += dst_stride_argb;
+  src_y += src_stride_y;
+
+  for (y = 0; y < height - 2; y += 2) {
+    Scale2RowUp_Bilinear_12(src_u, src_stride_u, temp_u_1, row_size, width);
+    Scale2RowUp_Bilinear_12(src_v, src_stride_v, temp_v_1, row_size, width);
+    I410ToARGBRow(src_y, temp_u_1, temp_v_1, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    I410ToARGBRow(src_y, temp_u_2, temp_v_2, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+
+  if (!(height & 1)) {
+    ScaleRowUp2_Linear_12(src_u, temp_u_1, width);
+    ScaleRowUp2_Linear_12(src_v, temp_v_1, width);
+    I410ToARGBRow(src_y, temp_u_1, temp_v_1, dst_argb, yuvconstants, width);
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int I210ToARGBMatrixLinear(const uint16_t* src_y,
+                                  int src_stride_y,
+                                  const uint16_t* src_u,
+                                  int src_stride_u,
+                                  const uint16_t* src_v,
+                                  int src_stride_v,
+                                  uint8_t* dst_argb,
+                                  int dst_stride_argb,
+                                  const struct YuvConstants* yuvconstants,
+                                  int width,
+                                  int height) {
+  int y;
+  void (*I410ToARGBRow)(const uint16_t* y_buf, const uint16_t* u_buf,
+                        const uint16_t* v_buf, uint8_t* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int width) =
+      I410ToARGBRow_C;
+  void (*ScaleRowUp2_Linear_12)(const uint16_t* src_ptr, uint16_t* dst_ptr,
+                                int dst_width) = ScaleRowUp2_Linear_16_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I410TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I410ToARGBRow = I410ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I410ToARGBRow = I410ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I410TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I410ToARGBRow = I410ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I410ToARGBRow = I410ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I410TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I410ToARGBRow = I410ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I410ToARGBRow = I410ToARGBRow_AVX2;
+    }
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_LINEAR_12_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_SSSE3;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_12_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_AVX2;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_12_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_NEON;
+  }
+#endif
+
+  // alloc 2 lines temp
+  const int row_size = (width + 31) & ~31;
+  align_buffer_64(row, row_size * 2 * sizeof(uint16_t));
+  uint16_t* temp_u = (uint16_t*)(row);
+  uint16_t* temp_v = (uint16_t*)(row) + row_size;
+  if (!row)
+    return 1;
+
+  for (y = 0; y < height; ++y) {
+    ScaleRowUp2_Linear_12(src_u, temp_u, width);
+    ScaleRowUp2_Linear_12(src_v, temp_v, width);
+    I410ToARGBRow(src_y, temp_u, temp_v, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int I420AlphaToARGBMatrixBilinear(
+    const uint8_t* src_y,
+    int src_stride_y,
+    const uint8_t* src_u,
+    int src_stride_u,
+    const uint8_t* src_v,
+    int src_stride_v,
+    const uint8_t* src_a,
+    int src_stride_a,
+    uint8_t* dst_argb,
+    int dst_stride_argb,
+    const struct YuvConstants* yuvconstants,
+    int width,
+    int height,
+    int attenuate) {
+  int y;
+  void (*I444AlphaToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                             const uint8_t* v_buf, const uint8_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width) = I444AlphaToARGBRow_C;
+  void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                           int width) = ARGBAttenuateRow_C;
+  void (*Scale2RowUp_Bilinear)(const uint8_t* src_ptr, ptrdiff_t src_stride,
+                               uint8_t* dst_ptr, ptrdiff_t dst_stride,
+                               int dst_width) = ScaleRowUp2_Bilinear_Any_C;
+  void (*ScaleRowUp2_Linear)(const uint8_t* src_ptr, uint8_t* dst_ptr,
+                             int dst_width) = ScaleRowUp2_Linear_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !src_a || !dst_argb || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I444ALPHATOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_RVV;
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_RVV;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_Any_SSE2;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_SSE2;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_Any_SSSE3;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_SSSE3;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_Any_AVX2;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_AVX2;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_Any_NEON;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_NEON;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_BILINEAR_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    Scale2RowUp_Bilinear = ScaleRowUp2_Bilinear_RVV;
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_RVV;
+  }
+#endif
+
+  // alloc 4 lines temp
+  const int row_size = (width + 31) & ~31;
+  align_buffer_64(row, row_size * 4);
+  uint8_t* temp_u_1 = row;
+  uint8_t* temp_u_2 = row + row_size;
+  uint8_t* temp_v_1 = row + row_size * 2;
+  uint8_t* temp_v_2 = row + row_size * 3;
+  if (!row)
+    return 1;
+
+  ScaleRowUp2_Linear(src_u, temp_u_1, width);
+  ScaleRowUp2_Linear(src_v, temp_v_1, width);
+  I444AlphaToARGBRow(src_y, temp_u_1, temp_v_1, src_a, dst_argb, yuvconstants,
+                     width);
+  if (attenuate) {
+    ARGBAttenuateRow(dst_argb, dst_argb, width);
+  }
+  dst_argb += dst_stride_argb;
+  src_y += src_stride_y;
+  src_a += src_stride_a;
+
+  for (y = 0; y < height - 2; y += 2) {
+    Scale2RowUp_Bilinear(src_u, src_stride_u, temp_u_1, row_size, width);
+    Scale2RowUp_Bilinear(src_v, src_stride_v, temp_v_1, row_size, width);
+    I444AlphaToARGBRow(src_y, temp_u_1, temp_v_1, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_a += src_stride_a;
+    I444AlphaToARGBRow(src_y, temp_u_2, temp_v_2, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+    src_a += src_stride_a;
+  }
+
+  if (!(height & 1)) {
+    ScaleRowUp2_Linear(src_u, temp_u_1, width);
+    ScaleRowUp2_Linear(src_v, temp_v_1, width);
+    I444AlphaToARGBRow(src_y, temp_u_1, temp_v_1, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int I422AlphaToARGBMatrixLinear(const uint8_t* src_y,
+                                       int src_stride_y,
+                                       const uint8_t* src_u,
+                                       int src_stride_u,
+                                       const uint8_t* src_v,
+                                       int src_stride_v,
+                                       const uint8_t* src_a,
+                                       int src_stride_a,
+                                       uint8_t* dst_argb,
+                                       int dst_stride_argb,
+                                       const struct YuvConstants* yuvconstants,
+                                       int width,
+                                       int height,
+                                       int attenuate) {
+  int y;
+  void (*I444AlphaToARGBRow)(const uint8_t* y_buf, const uint8_t* u_buf,
+                             const uint8_t* v_buf, const uint8_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width) = I444AlphaToARGBRow_C;
+  void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                           int width) = ARGBAttenuateRow_C;
+  void (*ScaleRowUp2_Linear)(const uint8_t* src_ptr, uint8_t* dst_ptr,
+                             int dst_width) = ScaleRowUp2_Linear_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !src_a || !dst_argb || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I444ALPHATOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_SVE2;
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      I444AlphaToARGBRow = I444AlphaToARGBRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_I444ALPHATOARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I444AlphaToARGBRow = I444AlphaToARGBRow_RVV;
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_RVV;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_LINEAR_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_SSE2;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_SSSE3;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_AVX2;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_NEON;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_RVV;
+  }
+#endif
+
+  // alloc 2 lines temp
+  const int row_size = (width + 31) & ~31;
+  align_buffer_64(row, row_size * 2);
+  uint8_t* temp_u = row;
+  uint8_t* temp_v = row + row_size;
+  if (!row)
+    return 1;
+
+  for (y = 0; y < height; ++y) {
+    ScaleRowUp2_Linear(src_u, temp_u, width);
+    ScaleRowUp2_Linear(src_v, temp_v, width);
+    I444AlphaToARGBRow(src_y, temp_u, temp_v, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+    dst_argb += dst_stride_argb;
+    src_a += src_stride_a;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int I010AlphaToARGBMatrixBilinear(
+    const uint16_t* src_y,
+    int src_stride_y,
+    const uint16_t* src_u,
+    int src_stride_u,
+    const uint16_t* src_v,
+    int src_stride_v,
+    const uint16_t* src_a,
+    int src_stride_a,
+    uint8_t* dst_argb,
+    int dst_stride_argb,
+    const struct YuvConstants* yuvconstants,
+    int width,
+    int height,
+    int attenuate) {
+  int y;
+  void (*I410AlphaToARGBRow)(const uint16_t* y_buf, const uint16_t* u_buf,
+                             const uint16_t* v_buf, const uint16_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width) = I410AlphaToARGBRow_C;
+  void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                           int width) = ARGBAttenuateRow_C;
+  void (*Scale2RowUp_Bilinear_12)(
+      const uint16_t* src_ptr, ptrdiff_t src_stride, uint16_t* dst_ptr,
+      ptrdiff_t dst_stride, int dst_width) = ScaleRowUp2_Bilinear_16_Any_C;
+  void (*ScaleRowUp2_Linear_12)(const uint16_t* src_ptr, uint16_t* dst_ptr,
+                                int dst_width) = ScaleRowUp2_Linear_16_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !src_a || !dst_argb || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I410ALPHATOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I410AlphaToARGBRow = I410AlphaToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I410AlphaToARGBRow = I410AlphaToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I410ALPHATOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I410AlphaToARGBRow = I410AlphaToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I410AlphaToARGBRow = I410AlphaToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I410ALPHATOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I410AlphaToARGBRow = I410AlphaToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I410AlphaToARGBRow = I410AlphaToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_RVV;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_12_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    Scale2RowUp_Bilinear_12 = ScaleRowUp2_Bilinear_12_Any_SSSE3;
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_SSSE3;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_12_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    Scale2RowUp_Bilinear_12 = ScaleRowUp2_Bilinear_12_Any_AVX2;
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_AVX2;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_BILINEAR_12_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    Scale2RowUp_Bilinear_12 = ScaleRowUp2_Bilinear_12_Any_NEON;
+    ScaleRowUp2_Linear_12 = ScaleRowUp2_Linear_12_Any_NEON;
+  }
+#endif
+
+  // alloc 4 lines temp
+  const int row_size = (width + 31) & ~31;
+  align_buffer_64(row, row_size * 4 * sizeof(uint16_t));
+  uint16_t* temp_u_1 = (uint16_t*)(row);
+  uint16_t* temp_u_2 = (uint16_t*)(row) + row_size;
+  uint16_t* temp_v_1 = (uint16_t*)(row) + row_size * 2;
+  uint16_t* temp_v_2 = (uint16_t*)(row) + row_size * 3;
+  if (!row)
+    return 1;
+
+  ScaleRowUp2_Linear_12(src_u, temp_u_1, width);
+  ScaleRowUp2_Linear_12(src_v, temp_v_1, width);
+  I410AlphaToARGBRow(src_y, temp_u_1, temp_v_1, src_a, dst_argb, yuvconstants,
+                     width);
+  if (attenuate) {
+    ARGBAttenuateRow(dst_argb, dst_argb, width);
+  }
+  dst_argb += dst_stride_argb;
+  src_y += src_stride_y;
+  src_a += src_stride_a;
+
+  for (y = 0; y < height - 2; y += 2) {
+    Scale2RowUp_Bilinear_12(src_u, src_stride_u, temp_u_1, row_size, width);
+    Scale2RowUp_Bilinear_12(src_v, src_stride_v, temp_v_1, row_size, width);
+    I410AlphaToARGBRow(src_y, temp_u_1, temp_v_1, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_a += src_stride_a;
+    I410AlphaToARGBRow(src_y, temp_u_2, temp_v_2, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_a += src_stride_a;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+
+  if (!(height & 1)) {
+    ScaleRowUp2_Linear_12(src_u, temp_u_1, width);
+    ScaleRowUp2_Linear_12(src_v, temp_v_1, width);
+    I410AlphaToARGBRow(src_y, temp_u_1, temp_v_1, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int I210AlphaToARGBMatrixLinear(const uint16_t* src_y,
+                                       int src_stride_y,
+                                       const uint16_t* src_u,
+                                       int src_stride_u,
+                                       const uint16_t* src_v,
+                                       int src_stride_v,
+                                       const uint16_t* src_a,
+                                       int src_stride_a,
+                                       uint8_t* dst_argb,
+                                       int dst_stride_argb,
+                                       const struct YuvConstants* yuvconstants,
+                                       int width,
+                                       int height,
+                                       int attenuate) {
+  int y;
+  void (*I410AlphaToARGBRow)(const uint16_t* y_buf, const uint16_t* u_buf,
+                             const uint16_t* v_buf, const uint16_t* a_buf,
+                             uint8_t* dst_argb,
+                             const struct YuvConstants* yuvconstants,
+                             int width) = I410AlphaToARGBRow_C;
+  void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                           int width) = ARGBAttenuateRow_C;
+  void (*ScaleRowUp2_Linear)(const uint16_t* src_ptr, uint16_t* dst_ptr,
+                             int dst_width) = ScaleRowUp2_Linear_16_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !src_a || !dst_argb || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_I410ALPHATOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I410AlphaToARGBRow = I410AlphaToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I410AlphaToARGBRow = I410AlphaToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I410ALPHATOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I410AlphaToARGBRow = I410AlphaToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I410AlphaToARGBRow = I410AlphaToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I410ALPHATOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I410AlphaToARGBRow = I410AlphaToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I410AlphaToARGBRow = I410AlphaToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_RVV;
+  }
+#endif
+
+#if defined(HAS_SCALEROWUP2_LINEAR_12_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_12_Any_SSSE3;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_12_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_12_Any_AVX2;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_12_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_12_Any_NEON;
+  }
+#endif
+
+  // alloc 2 lines temp
+  const int row_size = (width + 31) & ~31;
+  align_buffer_64(row, row_size * 2 * sizeof(uint16_t));
+  uint16_t* temp_u = (uint16_t*)(row);
+  uint16_t* temp_v = (uint16_t*)(row) + row_size;
+  if (!row)
+    return 1;
+
+  for (y = 0; y < height; ++y) {
+    ScaleRowUp2_Linear(src_u, temp_u, width);
+    ScaleRowUp2_Linear(src_v, temp_v, width);
+    I410AlphaToARGBRow(src_y, temp_u, temp_v, src_a, dst_argb, yuvconstants,
+                       width);
+    if (attenuate) {
+      ARGBAttenuateRow(dst_argb, dst_argb, width);
+    }
+    dst_argb += dst_stride_argb;
+    src_a += src_stride_a;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int P010ToARGBMatrixBilinear(const uint16_t* src_y,
+                                    int src_stride_y,
+                                    const uint16_t* src_uv,
+                                    int src_stride_uv,
+                                    uint8_t* dst_argb,
+                                    int dst_stride_argb,
+                                    const struct YuvConstants* yuvconstants,
+                                    int width,
+                                    int height) {
+  int y;
+  void (*P410ToARGBRow)(
+      const uint16_t* y_buf, const uint16_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = P410ToARGBRow_C;
+  void (*Scale2RowUp_Bilinear_16)(
+      const uint16_t* src_ptr, ptrdiff_t src_stride, uint16_t* dst_ptr,
+      ptrdiff_t dst_stride, int dst_width) = ScaleUVRowUp2_Bilinear_16_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_uv || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_P410TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    P410ToARGBRow = P410ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      P410ToARGBRow = P410ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_P410TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    P410ToARGBRow = P410ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      P410ToARGBRow = P410ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_P410TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    P410ToARGBRow = P410ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      P410ToARGBRow = P410ToARGBRow_NEON;
+    }
+  }
+#endif
+
+#ifdef HAS_SCALEUVROWUP2_BILINEAR_16_SSE41
+  if (TestCpuFlag(kCpuHasSSE41)) {
+    Scale2RowUp_Bilinear_16 = ScaleUVRowUp2_Bilinear_16_Any_SSE41;
+  }
+#endif
+
+#ifdef HAS_SCALEUVROWUP2_BILINEAR_16_AVX2
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    Scale2RowUp_Bilinear_16 = ScaleUVRowUp2_Bilinear_16_Any_AVX2;
+  }
+#endif
+
+#ifdef HAS_SCALEUVROWUP2_BILINEAR_16_NEON
+  if (TestCpuFlag(kCpuHasNEON)) {
+    Scale2RowUp_Bilinear_16 = ScaleUVRowUp2_Bilinear_16_Any_NEON;
+  }
+#endif
+
+  // alloc 2 lines temp
+  const int row_size = (2 * width + 31) & ~31;
+  align_buffer_64(row, row_size * 2 * sizeof(uint16_t));
+  uint16_t* temp_uv_1 = (uint16_t*)(row);
+  uint16_t* temp_uv_2 = (uint16_t*)(row) + row_size;
+  if (!row)
+    return 1;
+
+  Scale2RowUp_Bilinear_16(src_uv, 0, temp_uv_1, row_size, width);
+  P410ToARGBRow(src_y, temp_uv_1, dst_argb, yuvconstants, width);
+  dst_argb += dst_stride_argb;
+  src_y += src_stride_y;
+
+  for (y = 0; y < height - 2; y += 2) {
+    Scale2RowUp_Bilinear_16(src_uv, src_stride_uv, temp_uv_1, row_size, width);
+    P410ToARGBRow(src_y, temp_uv_1, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    P410ToARGBRow(src_y, temp_uv_2, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_uv += src_stride_uv;
+  }
+
+  if (!(height & 1)) {
+    Scale2RowUp_Bilinear_16(src_uv, 0, temp_uv_1, row_size, width);
+    P410ToARGBRow(src_y, temp_uv_1, dst_argb, yuvconstants, width);
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int P210ToARGBMatrixLinear(const uint16_t* src_y,
+                                  int src_stride_y,
+                                  const uint16_t* src_uv,
+                                  int src_stride_uv,
+                                  uint8_t* dst_argb,
+                                  int dst_stride_argb,
+                                  const struct YuvConstants* yuvconstants,
+                                  int width,
+                                  int height) {
+  int y;
+  void (*P410ToARGBRow)(
+      const uint16_t* y_buf, const uint16_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = P410ToARGBRow_C;
+  void (*ScaleRowUp2_Linear)(const uint16_t* src_uv, uint16_t* dst_uv,
+                             int dst_width) = ScaleUVRowUp2_Linear_16_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_uv || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_P410TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    P410ToARGBRow = P410ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      P410ToARGBRow = P410ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_P410TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    P410ToARGBRow = P410ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      P410ToARGBRow = P410ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_P410TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    P410ToARGBRow = P410ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      P410ToARGBRow = P410ToARGBRow_NEON;
+    }
+  }
+#endif
+
+#ifdef HAS_SCALEUVROWUP2_LINEAR_16_SSE41
+  if (TestCpuFlag(kCpuHasSSE41)) {
+    ScaleRowUp2_Linear = ScaleUVRowUp2_Linear_16_Any_SSE41;
+  }
+#endif
+
+#ifdef HAS_SCALEUVROWUP2_LINEAR_16_AVX2
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ScaleRowUp2_Linear = ScaleUVRowUp2_Linear_16_Any_AVX2;
+  }
+#endif
+
+#ifdef HAS_SCALEUVROWUP2_LINEAR_16_NEON
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ScaleRowUp2_Linear = ScaleUVRowUp2_Linear_16_Any_NEON;
+  }
+#endif
+
+  const int row_size = (2 * width + 31) & ~31;
+  align_buffer_64(row, row_size * sizeof(uint16_t));
+  uint16_t* temp_uv = (uint16_t*)(row);
+  if (!row)
+    return 1;
+
+  for (y = 0; y < height; ++y) {
+    ScaleRowUp2_Linear(src_uv, temp_uv, width);
+    P410ToARGBRow(src_y, temp_uv, dst_argb, yuvconstants, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_uv += src_stride_uv;
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int P010ToAR30MatrixBilinear(const uint16_t* src_y,
+                                    int src_stride_y,
+                                    const uint16_t* src_uv,
+                                    int src_stride_uv,
+                                    uint8_t* dst_ar30,
+                                    int dst_stride_ar30,
+                                    const struct YuvConstants* yuvconstants,
+                                    int width,
+                                    int height) {
+  int y;
+  void (*P410ToAR30Row)(
+      const uint16_t* y_buf, const uint16_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = P410ToAR30Row_C;
+  void (*Scale2RowUp_Bilinear_16)(
+      const uint16_t* src_ptr, ptrdiff_t src_stride, uint16_t* dst_ptr,
+      ptrdiff_t dst_stride, int dst_width) = ScaleUVRowUp2_Bilinear_16_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_uv || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar30 = dst_ar30 + (height - 1) * dst_stride_ar30;
+    dst_stride_ar30 = -dst_stride_ar30;
+  }
+#if defined(HAS_P410TOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    P410ToAR30Row = P410ToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      P410ToAR30Row = P410ToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_P410TOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    P410ToAR30Row = P410ToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      P410ToAR30Row = P410ToAR30Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_P410TOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    P410ToAR30Row = P410ToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      P410ToAR30Row = P410ToAR30Row_NEON;
+    }
+  }
+#endif
+
+#ifdef HAS_SCALEUVROWUP2_BILINEAR_16_SSE41
+  if (TestCpuFlag(kCpuHasSSE41)) {
+    Scale2RowUp_Bilinear_16 = ScaleUVRowUp2_Bilinear_16_Any_SSE41;
+  }
+#endif
+
+#ifdef HAS_SCALEUVROWUP2_BILINEAR_16_AVX2
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    Scale2RowUp_Bilinear_16 = ScaleUVRowUp2_Bilinear_16_Any_AVX2;
+  }
+#endif
+
+#ifdef HAS_SCALEUVROWUP2_BILINEAR_16_NEON
+  if (TestCpuFlag(kCpuHasNEON)) {
+    Scale2RowUp_Bilinear_16 = ScaleUVRowUp2_Bilinear_16_Any_NEON;
+  }
+#endif
+
+  // alloc 2 lines temp
+  const int row_size = (2 * width + 31) & ~31;
+  align_buffer_64(row, row_size * 2 * sizeof(uint16_t));
+  uint16_t* temp_uv_1 = (uint16_t*)(row);
+  uint16_t* temp_uv_2 = (uint16_t*)(row) + row_size;
+  if (!row)
+    return 1;
+
+  Scale2RowUp_Bilinear_16(src_uv, 0, temp_uv_1, row_size, width);
+  P410ToAR30Row(src_y, temp_uv_1, dst_ar30, yuvconstants, width);
+  dst_ar30 += dst_stride_ar30;
+  src_y += src_stride_y;
+
+  for (y = 0; y < height - 2; y += 2) {
+    Scale2RowUp_Bilinear_16(src_uv, src_stride_uv, temp_uv_1, row_size, width);
+    P410ToAR30Row(src_y, temp_uv_1, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    P410ToAR30Row(src_y, temp_uv_2, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    src_uv += src_stride_uv;
+  }
+
+  if (!(height & 1)) {
+    Scale2RowUp_Bilinear_16(src_uv, 0, temp_uv_1, row_size, width);
+    P410ToAR30Row(src_y, temp_uv_1, dst_ar30, yuvconstants, width);
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int P210ToAR30MatrixLinear(const uint16_t* src_y,
+                                  int src_stride_y,
+                                  const uint16_t* src_uv,
+                                  int src_stride_uv,
+                                  uint8_t* dst_ar30,
+                                  int dst_stride_ar30,
+                                  const struct YuvConstants* yuvconstants,
+                                  int width,
+                                  int height) {
+  int y;
+  void (*P410ToAR30Row)(
+      const uint16_t* y_buf, const uint16_t* uv_buf, uint8_t* rgb_buf,
+      const struct YuvConstants* yuvconstants, int width) = P410ToAR30Row_C;
+  void (*ScaleRowUp2_Linear)(const uint16_t* src_uv, uint16_t* dst_uv,
+                             int dst_width) = ScaleUVRowUp2_Linear_16_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_uv || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar30 = dst_ar30 + (height - 1) * dst_stride_ar30;
+    dst_stride_ar30 = -dst_stride_ar30;
+  }
+#if defined(HAS_P410TOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    P410ToAR30Row = P410ToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      P410ToAR30Row = P410ToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_P410TOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    P410ToAR30Row = P410ToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      P410ToAR30Row = P410ToAR30Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_P410TOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    P410ToAR30Row = P410ToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      P410ToAR30Row = P410ToAR30Row_NEON;
+    }
+  }
+#endif
+
+#ifdef HAS_SCALEUVROWUP2_LINEAR_16_SSE41
+  if (TestCpuFlag(kCpuHasSSE41)) {
+    ScaleRowUp2_Linear = ScaleUVRowUp2_Linear_16_Any_SSE41;
+  }
+#endif
+
+#ifdef HAS_SCALEUVROWUP2_LINEAR_16_AVX2
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ScaleRowUp2_Linear = ScaleUVRowUp2_Linear_16_Any_AVX2;
+  }
+#endif
+
+#ifdef HAS_SCALEUVROWUP2_LINEAR_16_NEON
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ScaleRowUp2_Linear = ScaleUVRowUp2_Linear_16_Any_NEON;
+  }
+#endif
+
+  const int row_size = (2 * width + 31) & ~31;
+  align_buffer_64(row, row_size * sizeof(uint16_t));
+  uint16_t* temp_uv = (uint16_t*)(row);
+  if (!row)
+    return 1;
+
+  for (y = 0; y < height; ++y) {
+    ScaleRowUp2_Linear(src_uv, temp_uv, width);
+    P410ToAR30Row(src_y, temp_uv, dst_ar30, yuvconstants, width);
+    dst_ar30 += dst_stride_ar30;
+    src_y += src_stride_y;
+    src_uv += src_stride_uv;
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+static int I422ToRGB24MatrixLinear(const uint8_t* src_y,
+                                   int src_stride_y,
+                                   const uint8_t* src_u,
+                                   int src_stride_u,
+                                   const uint8_t* src_v,
+                                   int src_stride_v,
+                                   uint8_t* dst_rgb24,
+                                   int dst_stride_rgb24,
+                                   const struct YuvConstants* yuvconstants,
+                                   int width,
+                                   int height) {
+  int y;
+  void (*I444ToRGB24Row)(const uint8_t* y_buf, const uint8_t* u_buf,
+                         const uint8_t* v_buf, uint8_t* rgb_buf,
+                         const struct YuvConstants* yuvconstants, int width) =
+      I444ToRGB24Row_C;
+  void (*ScaleRowUp2_Linear)(const uint8_t* src_ptr, uint8_t* dst_ptr,
+                             int dst_width) = ScaleRowUp2_Linear_Any_C;
+  assert(yuvconstants);
+  if (!src_y || !src_u || !src_v || !dst_rgb24 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb24 = dst_rgb24 + (height - 1) * dst_stride_rgb24;
+    dst_stride_rgb24 = -dst_stride_rgb24;
+  }
+#if defined(HAS_I444TORGB24ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      I444ToRGB24Row = I444ToRGB24Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      I444ToRGB24Row = I444ToRGB24Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I444ToRGB24Row = I444ToRGB24Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I444ToRGB24Row = I444ToRGB24Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I444TORGB24ROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    I444ToRGB24Row = I444ToRGB24Row_RVV;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_SSE2;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_SSSE3;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_AVX2;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_Any_NEON;
+  }
+#endif
+#if defined(HAS_SCALEROWUP2_LINEAR_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ScaleRowUp2_Linear = ScaleRowUp2_Linear_RVV;
+  }
+#endif
+
+  // alloc 2 lines temp
+  const int row_size = (width + 31) & ~31;
+  align_buffer_64(row, row_size * 2);
+  uint8_t* temp_u = row;
+  uint8_t* temp_v = row + row_size;
+  if (!row)
+    return 1;
+
+  for (y = 0; y < height; ++y) {
+    ScaleRowUp2_Linear(src_u, temp_u, width);
+    ScaleRowUp2_Linear(src_v, temp_v, width);
+    I444ToRGB24Row(src_y, temp_u, temp_v, dst_rgb24, yuvconstants, width);
+    dst_rgb24 += dst_stride_rgb24;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+LIBYUV_API
+int I422ToRGB24MatrixFilter(const uint8_t* src_y,
+                            int src_stride_y,
+                            const uint8_t* src_u,
+                            int src_stride_u,
+                            const uint8_t* src_v,
+                            int src_stride_v,
+                            uint8_t* dst_rgb24,
+                            int dst_stride_rgb24,
+                            const struct YuvConstants* yuvconstants,
+                            int width,
+                            int height,
+                            enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return I422ToRGB24Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                               src_stride_v, dst_rgb24, dst_stride_rgb24,
+                               yuvconstants, width, height);
+    case kFilterBilinear:
+    case kFilterBox:
+    case kFilterLinear:
+      return I422ToRGB24MatrixLinear(
+          src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v,
+          dst_rgb24, dst_stride_rgb24, yuvconstants, width, height);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int I420ToARGBMatrixFilter(const uint8_t* src_y,
+                           int src_stride_y,
+                           const uint8_t* src_u,
+                           int src_stride_u,
+                           const uint8_t* src_v,
+                           int src_stride_v,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return I420ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                              src_stride_v, dst_argb, dst_stride_argb,
+                              yuvconstants, width, height);
+    case kFilterBilinear:
+    case kFilterBox:
+      return I420ToARGBMatrixBilinear(
+          src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v,
+          dst_argb, dst_stride_argb, yuvconstants, width, height);
+    case kFilterLinear:
+      // Actually we can do this, but probably there's no usage.
+      return -1;
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int I422ToARGBMatrixFilter(const uint8_t* src_y,
+                           int src_stride_y,
+                           const uint8_t* src_u,
+                           int src_stride_u,
+                           const uint8_t* src_v,
+                           int src_stride_v,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return I422ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                              src_stride_v, dst_argb, dst_stride_argb,
+                              yuvconstants, width, height);
+    case kFilterBilinear:
+    case kFilterBox:
+    case kFilterLinear:
+      return I422ToARGBMatrixLinear(
+          src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v,
+          dst_argb, dst_stride_argb, yuvconstants, width, height);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int I420ToRGB24MatrixFilter(const uint8_t* src_y,
+                            int src_stride_y,
+                            const uint8_t* src_u,
+                            int src_stride_u,
+                            const uint8_t* src_v,
+                            int src_stride_v,
+                            uint8_t* dst_rgb24,
+                            int dst_stride_rgb24,
+                            const struct YuvConstants* yuvconstants,
+                            int width,
+                            int height,
+                            enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return I420ToRGB24Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                               src_stride_v, dst_rgb24, dst_stride_rgb24,
+                               yuvconstants, width, height);
+    case kFilterLinear:  // TODO(fb): Implement Linear using Bilinear stride 0
+    case kFilterBilinear:
+    case kFilterBox:
+      return I420ToRGB24MatrixBilinear(
+          src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v,
+          dst_rgb24, dst_stride_rgb24, yuvconstants, width, height);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int I010ToAR30MatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_u,
+                           int src_stride_u,
+                           const uint16_t* src_v,
+                           int src_stride_v,
+                           uint8_t* dst_ar30,
+                           int dst_stride_ar30,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return I010ToAR30Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                              src_stride_v, dst_ar30, dst_stride_ar30,
+                              yuvconstants, width, height);
+    case kFilterLinear:  // TODO(fb): Implement Linear using Bilinear stride 0
+    case kFilterBilinear:
+    case kFilterBox:
+      return I010ToAR30MatrixBilinear(
+          src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v,
+          dst_ar30, dst_stride_ar30, yuvconstants, width, height);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int I210ToAR30MatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_u,
+                           int src_stride_u,
+                           const uint16_t* src_v,
+                           int src_stride_v,
+                           uint8_t* dst_ar30,
+                           int dst_stride_ar30,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return I210ToAR30Matrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                              src_stride_v, dst_ar30, dst_stride_ar30,
+                              yuvconstants, width, height);
+    case kFilterBilinear:
+    case kFilterBox:
+    case kFilterLinear:
+      return I210ToAR30MatrixLinear(
+          src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v,
+          dst_ar30, dst_stride_ar30, yuvconstants, width, height);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int I010ToARGBMatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_u,
+                           int src_stride_u,
+                           const uint16_t* src_v,
+                           int src_stride_v,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return I010ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                              src_stride_v, dst_argb, dst_stride_argb,
+                              yuvconstants, width, height);
+    case kFilterLinear:  // TODO(fb): Implement Linear using Bilinear stride 0
+    case kFilterBilinear:
+    case kFilterBox:
+      return I010ToARGBMatrixBilinear(
+          src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v,
+          dst_argb, dst_stride_argb, yuvconstants, width, height);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int I210ToARGBMatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_u,
+                           int src_stride_u,
+                           const uint16_t* src_v,
+                           int src_stride_v,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return I210ToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                              src_stride_v, dst_argb, dst_stride_argb,
+                              yuvconstants, width, height);
+    case kFilterBilinear:
+    case kFilterBox:
+    case kFilterLinear:
+      return I210ToARGBMatrixLinear(
+          src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v,
+          dst_argb, dst_stride_argb, yuvconstants, width, height);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int I420AlphaToARGBMatrixFilter(const uint8_t* src_y,
+                                int src_stride_y,
+                                const uint8_t* src_u,
+                                int src_stride_u,
+                                const uint8_t* src_v,
+                                int src_stride_v,
+                                const uint8_t* src_a,
+                                int src_stride_a,
+                                uint8_t* dst_argb,
+                                int dst_stride_argb,
+                                const struct YuvConstants* yuvconstants,
+                                int width,
+                                int height,
+                                int attenuate,
+                                enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return I420AlphaToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u,
+                                   src_v, src_stride_v, src_a, src_stride_a,
+                                   dst_argb, dst_stride_argb, yuvconstants,
+                                   width, height, attenuate);
+    case kFilterLinear:  // TODO(fb): Implement Linear using Bilinear stride 0
+    case kFilterBilinear:
+    case kFilterBox:
+      return I420AlphaToARGBMatrixBilinear(
+          src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v, src_a,
+          src_stride_a, dst_argb, dst_stride_argb, yuvconstants, width, height,
+          attenuate);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int I422AlphaToARGBMatrixFilter(const uint8_t* src_y,
+                                int src_stride_y,
+                                const uint8_t* src_u,
+                                int src_stride_u,
+                                const uint8_t* src_v,
+                                int src_stride_v,
+                                const uint8_t* src_a,
+                                int src_stride_a,
+                                uint8_t* dst_argb,
+                                int dst_stride_argb,
+                                const struct YuvConstants* yuvconstants,
+                                int width,
+                                int height,
+                                int attenuate,
+                                enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return I422AlphaToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u,
+                                   src_v, src_stride_v, src_a, src_stride_a,
+                                   dst_argb, dst_stride_argb, yuvconstants,
+                                   width, height, attenuate);
+    case kFilterBilinear:
+    case kFilterBox:
+    case kFilterLinear:
+      return I422AlphaToARGBMatrixLinear(
+          src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v, src_a,
+          src_stride_a, dst_argb, dst_stride_argb, yuvconstants, width, height,
+          attenuate);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int I010AlphaToARGBMatrixFilter(const uint16_t* src_y,
+                                int src_stride_y,
+                                const uint16_t* src_u,
+                                int src_stride_u,
+                                const uint16_t* src_v,
+                                int src_stride_v,
+                                const uint16_t* src_a,
+                                int src_stride_a,
+                                uint8_t* dst_argb,
+                                int dst_stride_argb,
+                                const struct YuvConstants* yuvconstants,
+                                int width,
+                                int height,
+                                int attenuate,
+                                enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return I010AlphaToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u,
+                                   src_v, src_stride_v, src_a, src_stride_a,
+                                   dst_argb, dst_stride_argb, yuvconstants,
+                                   width, height, attenuate);
+    case kFilterLinear:  // TODO(fb): Implement Linear using Bilinear stride 0
+    case kFilterBilinear:
+    case kFilterBox:
+      return I010AlphaToARGBMatrixBilinear(
+          src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v, src_a,
+          src_stride_a, dst_argb, dst_stride_argb, yuvconstants, width, height,
+          attenuate);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int I210AlphaToARGBMatrixFilter(const uint16_t* src_y,
+                                int src_stride_y,
+                                const uint16_t* src_u,
+                                int src_stride_u,
+                                const uint16_t* src_v,
+                                int src_stride_v,
+                                const uint16_t* src_a,
+                                int src_stride_a,
+                                uint8_t* dst_argb,
+                                int dst_stride_argb,
+                                const struct YuvConstants* yuvconstants,
+                                int width,
+                                int height,
+                                int attenuate,
+                                enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return I210AlphaToARGBMatrix(src_y, src_stride_y, src_u, src_stride_u,
+                                   src_v, src_stride_v, src_a, src_stride_a,
+                                   dst_argb, dst_stride_argb, yuvconstants,
+                                   width, height, attenuate);
+    case kFilterBilinear:
+    case kFilterBox:
+    case kFilterLinear:
+      return I210AlphaToARGBMatrixLinear(
+          src_y, src_stride_y, src_u, src_stride_u, src_v, src_stride_v, src_a,
+          src_stride_a, dst_argb, dst_stride_argb, yuvconstants, width, height,
+          attenuate);
+  }
+
+  return -1;
+}
+
+// TODO(fb): Verify this function works correctly.  P010 is like NV12 but 10 bit
+// UV is biplanar.
+LIBYUV_API
+int P010ToARGBMatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_uv,
+                           int src_stride_uv,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return P010ToARGBMatrix(src_y, src_stride_y, src_uv, src_stride_uv,
+                              dst_argb, dst_stride_argb, yuvconstants, width,
+                              height);
+    case kFilterLinear:  // TODO(fb): Implement Linear using Bilinear stride 0
+    case kFilterBilinear:
+    case kFilterBox:
+      return P010ToARGBMatrixBilinear(src_y, src_stride_y, src_uv,
+                                      src_stride_uv, dst_argb, dst_stride_argb,
+                                      yuvconstants, width, height);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int P210ToARGBMatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_uv,
+                           int src_stride_uv,
+                           uint8_t* dst_argb,
+                           int dst_stride_argb,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return P210ToARGBMatrix(src_y, src_stride_y, src_uv, src_stride_uv,
+                              dst_argb, dst_stride_argb, yuvconstants, width,
+                              height);
+    case kFilterBilinear:
+    case kFilterBox:
+    case kFilterLinear:
+      return P210ToARGBMatrixLinear(src_y, src_stride_y, src_uv, src_stride_uv,
+                                    dst_argb, dst_stride_argb, yuvconstants,
+                                    width, height);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int P010ToAR30MatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_uv,
+                           int src_stride_uv,
+                           uint8_t* dst_ar30,
+                           int dst_stride_ar30,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return P010ToAR30Matrix(src_y, src_stride_y, src_uv, src_stride_uv,
+                              dst_ar30, dst_stride_ar30, yuvconstants, width,
+                              height);
+    case kFilterLinear:  // TODO(fb): Implement Linear using Bilinear stride 0
+    case kFilterBilinear:
+    case kFilterBox:
+      return P010ToAR30MatrixBilinear(src_y, src_stride_y, src_uv,
+                                      src_stride_uv, dst_ar30, dst_stride_ar30,
+                                      yuvconstants, width, height);
+  }
+
+  return -1;
+}
+
+LIBYUV_API
+int P210ToAR30MatrixFilter(const uint16_t* src_y,
+                           int src_stride_y,
+                           const uint16_t* src_uv,
+                           int src_stride_uv,
+                           uint8_t* dst_ar30,
+                           int dst_stride_ar30,
+                           const struct YuvConstants* yuvconstants,
+                           int width,
+                           int height,
+                           enum FilterMode filter) {
+  switch (filter) {
+    case kFilterNone:
+      return P210ToAR30Matrix(src_y, src_stride_y, src_uv, src_stride_uv,
+                              dst_ar30, dst_stride_ar30, yuvconstants, width,
+                              height);
+    case kFilterBilinear:
+    case kFilterBox:
+    case kFilterLinear:
+      return P210ToAR30MatrixLinear(src_y, src_stride_y, src_uv, src_stride_uv,
+                                    dst_ar30, dst_stride_ar30, yuvconstants,
+                                    width, height);
+  }
+
+  return -1;
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/convert_from.cc

@@ -0,0 +1,910 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/convert_from.h"
+
+#include "libyuv/basic_types.h"
+#include "libyuv/convert.h"  // For I420Copy
+#include "libyuv/cpu_id.h"
+#include "libyuv/planar_functions.h"
+#include "libyuv/rotate.h"
+#include "libyuv/row.h"
+#include "libyuv/scale.h"  // For ScalePlane()
+#include "libyuv/video_common.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+#define SUBSAMPLE(v, a, s) (v < 0) ? (-((-v + a) >> s)) : ((v + a) >> s)
+static __inline int Abs(int v) {
+  return v >= 0 ? v : -v;
+}
+
+// I420 To any I4xx YUV format with mirroring.
+// TODO(fbarchard): Consider kFilterNone for Y, or CopyPlane
+
+static int I420ToI4xx(const uint8_t* src_y,
+                      int src_stride_y,
+                      const uint8_t* src_u,
+                      int src_stride_u,
+                      const uint8_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_y,
+                      int dst_stride_y,
+                      uint8_t* dst_u,
+                      int dst_stride_u,
+                      uint8_t* dst_v,
+                      int dst_stride_v,
+                      int src_y_width,
+                      int src_y_height,
+                      int dst_uv_width,
+                      int dst_uv_height) {
+  const int dst_y_width = Abs(src_y_width);
+  const int dst_y_height = Abs(src_y_height);
+  const int src_uv_width = SUBSAMPLE(src_y_width, 1, 1);
+  const int src_uv_height = SUBSAMPLE(src_y_height, 1, 1);
+  int r;
+  if (src_y_width == 0 || src_y_height == 0 || dst_uv_width <= 0 ||
+      dst_uv_height <= 0) {
+    return -1;
+  }
+  if (dst_y) {
+    r = ScalePlane(src_y, src_stride_y, src_y_width, src_y_height, dst_y,
+                   dst_stride_y, dst_y_width, dst_y_height, kFilterBilinear);
+    if (r != 0) {
+      return r;
+    }
+  }
+  r = ScalePlane(src_u, src_stride_u, src_uv_width, src_uv_height, dst_u,
+                 dst_stride_u, dst_uv_width, dst_uv_height, kFilterBilinear);
+  if (r != 0) {
+    return r;
+  }
+  r = ScalePlane(src_v, src_stride_v, src_uv_width, src_uv_height, dst_v,
+                 dst_stride_v, dst_uv_width, dst_uv_height, kFilterBilinear);
+  return r;
+}
+
+// Convert 8 bit YUV to 10 bit.
+LIBYUV_API
+int I420ToI010(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (halfheight - 1) * src_stride_u;
+    src_v = src_v + (halfheight - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  // Convert Y plane.
+  Convert8To16Plane(src_y, src_stride_y, dst_y, dst_stride_y, 1024, width,
+                    height);
+  // Convert UV planes.
+  Convert8To16Plane(src_u, src_stride_u, dst_u, dst_stride_u, 1024, halfwidth,
+                    halfheight);
+  Convert8To16Plane(src_v, src_stride_v, dst_v, dst_stride_v, 1024, halfwidth,
+                    halfheight);
+  return 0;
+}
+
+// Convert 8 bit YUV to 12 bit.
+LIBYUV_API
+int I420ToI012(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (halfheight - 1) * src_stride_u;
+    src_v = src_v + (halfheight - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  // Convert Y plane.
+  Convert8To16Plane(src_y, src_stride_y, dst_y, dst_stride_y, 4096, width,
+                    height);
+  // Convert UV planes.
+  Convert8To16Plane(src_u, src_stride_u, dst_u, dst_stride_u, 4096, halfwidth,
+                    halfheight);
+  Convert8To16Plane(src_v, src_stride_v, dst_v, dst_stride_v, 4096, halfwidth,
+                    halfheight);
+  return 0;
+}
+
+// 420 chroma is 1/2 width, 1/2 height
+// 422 chroma is 1/2 width, 1x height
+LIBYUV_API
+int I420ToI422(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  const int dst_uv_width = (Abs(width) + 1) >> 1;
+  const int dst_uv_height = Abs(height);
+  return I420ToI4xx(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                    dst_v, dst_stride_v, width, height, dst_uv_width,
+                    dst_uv_height);
+}
+
+// 420 chroma is 1/2 width, 1/2 height
+// 444 chroma is 1x width, 1x height
+LIBYUV_API
+int I420ToI444(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  const int dst_uv_width = Abs(width);
+  const int dst_uv_height = Abs(height);
+  return I420ToI4xx(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                    src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                    dst_v, dst_stride_v, width, height, dst_uv_width,
+                    dst_uv_height);
+}
+
+// 420 chroma to 444 chroma, 10/12 bit version
+LIBYUV_API
+int I010ToI410(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int r;
+  if (width == 0 || height == 0) {
+    return -1;
+  }
+
+  if (dst_y) {
+    r = ScalePlane_12(src_y, src_stride_y, width, height, dst_y, dst_stride_y,
+                      Abs(width), Abs(height), kFilterBilinear);
+    if (r != 0) {
+      return r;
+    }
+  }
+  r = ScalePlane_12(src_u, src_stride_u, SUBSAMPLE(width, 1, 1),
+                    SUBSAMPLE(height, 1, 1), dst_u, dst_stride_u, Abs(width),
+                    Abs(height), kFilterBilinear);
+  if (r != 0) {
+    return r;
+  }
+  r = ScalePlane_12(src_v, src_stride_v, SUBSAMPLE(width, 1, 1),
+                    SUBSAMPLE(height, 1, 1), dst_v, dst_stride_v, Abs(width),
+                    Abs(height), kFilterBilinear);
+  return r;
+}
+
+// 422 chroma to 444 chroma, 10/12 bit version
+LIBYUV_API
+int I210ToI410(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int r;
+  if (width == 0 || height == 0) {
+    return -1;
+  }
+
+  if (dst_y) {
+    r = ScalePlane_12(src_y, src_stride_y, width, height, dst_y, dst_stride_y,
+                      Abs(width), Abs(height), kFilterBilinear);
+    if (r != 0) {
+      return r;
+    }
+  }
+  r = ScalePlane_12(src_u, src_stride_u, SUBSAMPLE(width, 1, 1), height, dst_u,
+                    dst_stride_u, Abs(width), Abs(height), kFilterBilinear);
+  if (r != 0) {
+    return r;
+  }
+  r = ScalePlane_12(src_v, src_stride_v, SUBSAMPLE(width, 1, 1), height, dst_v,
+                    dst_stride_v, Abs(width), Abs(height), kFilterBilinear);
+  return r;
+}
+
+// 422 chroma is 1/2 width, 1x height
+// 444 chroma is 1x width, 1x height
+LIBYUV_API
+int I422ToI444(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int r;
+  if (width == 0 || height == 0) {
+    return -1;
+  }
+
+  if (dst_y) {
+    r = ScalePlane(src_y, src_stride_y, width, height, dst_y, dst_stride_y,
+                   Abs(width), Abs(height), kFilterBilinear);
+    if (r != 0) {
+      return r;
+    }
+  }
+  r = ScalePlane(src_u, src_stride_u, SUBSAMPLE(width, 1, 1), height, dst_u,
+                 dst_stride_u, Abs(width), Abs(height), kFilterBilinear);
+  if (r != 0) {
+    return r;
+  }
+  r = ScalePlane(src_v, src_stride_v, SUBSAMPLE(width, 1, 1), height, dst_v,
+                 dst_stride_v, Abs(width), Abs(height), kFilterBilinear);
+  return r;
+}
+
+// Copy to I400. Source can be I420,422,444,400,NV12,NV21
+LIBYUV_API
+int I400Copy(const uint8_t* src_y,
+             int src_stride_y,
+             uint8_t* dst_y,
+             int dst_stride_y,
+             int width,
+             int height) {
+  if (!src_y || !dst_y || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_stride_y = -src_stride_y;
+  }
+  CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  return 0;
+}
+
+LIBYUV_API
+int I422ToYUY2(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_yuy2,
+               int dst_stride_yuy2,
+               int width,
+               int height) {
+  int y;
+  void (*I422ToYUY2Row)(const uint8_t* src_y, const uint8_t* src_u,
+                        const uint8_t* src_v, uint8_t* dst_yuy2, int width) =
+      I422ToYUY2Row_C;
+  if (!src_y || !src_u || !src_v || !dst_yuy2 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_yuy2 = dst_yuy2 + (height - 1) * dst_stride_yuy2;
+    dst_stride_yuy2 = -dst_stride_yuy2;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && src_stride_u * 2 == width &&
+      src_stride_v * 2 == width && dst_stride_yuy2 == width * 2) {
+    width *= height;
+    height = 1;
+    src_stride_y = src_stride_u = src_stride_v = dst_stride_yuy2 = 0;
+  }
+#if defined(HAS_I422TOYUY2ROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToYUY2Row = I422ToYUY2Row_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToYUY2Row = I422ToYUY2Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToYUY2Row = I422ToYUY2Row_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToYUY2Row(src_y, src_u, src_v, dst_yuy2, width);
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+    dst_yuy2 += dst_stride_yuy2;
+  }
+  return 0;
+}
+
+LIBYUV_API
+int I420ToYUY2(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_yuy2,
+               int dst_stride_yuy2,
+               int width,
+               int height) {
+  int y;
+  void (*I422ToYUY2Row)(const uint8_t* src_y, const uint8_t* src_u,
+                        const uint8_t* src_v, uint8_t* dst_yuy2, int width) =
+      I422ToYUY2Row_C;
+  if (!src_y || !src_u || !src_v || !dst_yuy2 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_yuy2 = dst_yuy2 + (height - 1) * dst_stride_yuy2;
+    dst_stride_yuy2 = -dst_stride_yuy2;
+  }
+#if defined(HAS_I422TOYUY2ROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToYUY2Row = I422ToYUY2Row_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToYUY2Row = I422ToYUY2Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToYUY2Row = I422ToYUY2Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToYUY2Row = I422ToYUY2Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToYUY2Row = I422ToYUY2Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToYUY2Row = I422ToYUY2Row_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    I422ToYUY2Row(src_y, src_u, src_v, dst_yuy2, width);
+    I422ToYUY2Row(src_y + src_stride_y, src_u, src_v,
+                  dst_yuy2 + dst_stride_yuy2, width);
+    src_y += src_stride_y * 2;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+    dst_yuy2 += dst_stride_yuy2 * 2;
+  }
+  if (height & 1) {
+    I422ToYUY2Row(src_y, src_u, src_v, dst_yuy2, width);
+  }
+  return 0;
+}
+
+LIBYUV_API
+int I422ToUYVY(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_uyvy,
+               int dst_stride_uyvy,
+               int width,
+               int height) {
+  int y;
+  void (*I422ToUYVYRow)(const uint8_t* src_y, const uint8_t* src_u,
+                        const uint8_t* src_v, uint8_t* dst_uyvy, int width) =
+      I422ToUYVYRow_C;
+  if (!src_y || !src_u || !src_v || !dst_uyvy || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_uyvy = dst_uyvy + (height - 1) * dst_stride_uyvy;
+    dst_stride_uyvy = -dst_stride_uyvy;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && src_stride_u * 2 == width &&
+      src_stride_v * 2 == width && dst_stride_uyvy == width * 2) {
+    width *= height;
+    height = 1;
+    src_stride_y = src_stride_u = src_stride_v = dst_stride_uyvy = 0;
+  }
+#if defined(HAS_I422TOUYVYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToUYVYRow = I422ToUYVYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToUYVYRow = I422ToUYVYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToUYVYRow = I422ToUYVYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToUYVYRow = I422ToUYVYRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToUYVYRow = I422ToUYVYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToUYVYRow = I422ToUYVYRow_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToUYVYRow(src_y, src_u, src_v, dst_uyvy, width);
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+    dst_uyvy += dst_stride_uyvy;
+  }
+  return 0;
+}
+
+LIBYUV_API
+int I420ToUYVY(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_uyvy,
+               int dst_stride_uyvy,
+               int width,
+               int height) {
+  int y;
+  void (*I422ToUYVYRow)(const uint8_t* src_y, const uint8_t* src_u,
+                        const uint8_t* src_v, uint8_t* dst_uyvy, int width) =
+      I422ToUYVYRow_C;
+  if (!src_y || !src_u || !src_v || !dst_uyvy || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_uyvy = dst_uyvy + (height - 1) * dst_stride_uyvy;
+    dst_stride_uyvy = -dst_stride_uyvy;
+  }
+#if defined(HAS_I422TOUYVYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToUYVYRow = I422ToUYVYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToUYVYRow = I422ToUYVYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToUYVYRow = I422ToUYVYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToUYVYRow = I422ToUYVYRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToUYVYRow = I422ToUYVYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToUYVYRow = I422ToUYVYRow_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    I422ToUYVYRow(src_y, src_u, src_v, dst_uyvy, width);
+    I422ToUYVYRow(src_y + src_stride_y, src_u, src_v,
+                  dst_uyvy + dst_stride_uyvy, width);
+    src_y += src_stride_y * 2;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+    dst_uyvy += dst_stride_uyvy * 2;
+  }
+  if (height & 1) {
+    I422ToUYVYRow(src_y, src_u, src_v, dst_uyvy, width);
+  }
+  return 0;
+}
+
+LIBYUV_API
+int I420ToNV12(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int halfwidth = (width + 1) / 2;
+  int halfheight = (height + 1) / 2;
+  if (!src_y || !src_u || !src_v || !dst_uv || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (halfheight - 1) * src_stride_u;
+    src_v = src_v + (halfheight - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+  if (dst_y) {
+    CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+  MergeUVPlane(src_u, src_stride_u, src_v, src_stride_v, dst_uv, dst_stride_uv,
+               halfwidth, halfheight);
+  return 0;
+}
+
+LIBYUV_API
+int I420ToNV21(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height) {
+  return I420ToNV12(src_y, src_stride_y, src_v, src_stride_v, src_u,
+                    src_stride_u, dst_y, dst_stride_y, dst_vu, dst_stride_vu,
+                    width, height);
+}
+
+// Convert I420 to specified format
+LIBYUV_API
+int ConvertFromI420(const uint8_t* y,
+                    int y_stride,
+                    const uint8_t* u,
+                    int u_stride,
+                    const uint8_t* v,
+                    int v_stride,
+                    uint8_t* dst_sample,
+                    int dst_sample_stride,
+                    int width,
+                    int height,
+                    uint32_t fourcc) {
+  uint32_t format = CanonicalFourCC(fourcc);
+  int r = 0;
+  if (!y || !u || !v || !dst_sample || width <= 0 || height == 0) {
+    return -1;
+  }
+  switch (format) {
+    // Single plane formats
+    case FOURCC_YUY2:
+      r = I420ToYUY2(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                     dst_sample_stride ? dst_sample_stride : width * 2, width,
+                     height);
+      break;
+    case FOURCC_UYVY:
+      r = I420ToUYVY(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                     dst_sample_stride ? dst_sample_stride : width * 2, width,
+                     height);
+      break;
+    case FOURCC_RGBP:
+      r = I420ToRGB565(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                       dst_sample_stride ? dst_sample_stride : width * 2, width,
+                       height);
+      break;
+    case FOURCC_RGBO:
+      r = I420ToARGB1555(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                         dst_sample_stride ? dst_sample_stride : width * 2,
+                         width, height);
+      break;
+    case FOURCC_R444:
+      r = I420ToARGB4444(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                         dst_sample_stride ? dst_sample_stride : width * 2,
+                         width, height);
+      break;
+    case FOURCC_24BG:
+      r = I420ToRGB24(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                      dst_sample_stride ? dst_sample_stride : width * 3, width,
+                      height);
+      break;
+    case FOURCC_RAW:
+      r = I420ToRAW(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                    dst_sample_stride ? dst_sample_stride : width * 3, width,
+                    height);
+      break;
+    case FOURCC_ARGB:
+      r = I420ToARGB(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                     dst_sample_stride ? dst_sample_stride : width * 4, width,
+                     height);
+      break;
+    case FOURCC_BGRA:
+      r = I420ToBGRA(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                     dst_sample_stride ? dst_sample_stride : width * 4, width,
+                     height);
+      break;
+    case FOURCC_ABGR:
+      r = I420ToABGR(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                     dst_sample_stride ? dst_sample_stride : width * 4, width,
+                     height);
+      break;
+    case FOURCC_RGBA:
+      r = I420ToRGBA(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                     dst_sample_stride ? dst_sample_stride : width * 4, width,
+                     height);
+      break;
+    case FOURCC_AR30:
+      r = I420ToAR30(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                     dst_sample_stride ? dst_sample_stride : width * 4, width,
+                     height);
+      break;
+    case FOURCC_I400:
+      r = I400Copy(y, y_stride, dst_sample,
+                   dst_sample_stride ? dst_sample_stride : width, width,
+                   height);
+      break;
+    case FOURCC_NV12: {
+      int dst_y_stride = dst_sample_stride ? dst_sample_stride : width;
+      uint8_t* dst_uv = dst_sample + dst_y_stride * height;
+      r = I420ToNV12(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                     dst_sample_stride ? dst_sample_stride : width, dst_uv,
+                     dst_sample_stride ? dst_sample_stride : width, width,
+                     height);
+      break;
+    }
+    case FOURCC_NV21: {
+      int dst_y_stride = dst_sample_stride ? dst_sample_stride : width;
+      uint8_t* dst_vu = dst_sample + dst_y_stride * height;
+      r = I420ToNV21(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                     dst_sample_stride ? dst_sample_stride : width, dst_vu,
+                     dst_sample_stride ? dst_sample_stride : width, width,
+                     height);
+      break;
+    }
+    // Triplanar formats
+    case FOURCC_I420:
+    case FOURCC_YV12: {
+      dst_sample_stride = dst_sample_stride ? dst_sample_stride : width;
+      int halfstride = (dst_sample_stride + 1) / 2;
+      int halfheight = (height + 1) / 2;
+      uint8_t* dst_u;
+      uint8_t* dst_v;
+      if (format == FOURCC_YV12) {
+        dst_v = dst_sample + dst_sample_stride * height;
+        dst_u = dst_v + halfstride * halfheight;
+      } else {
+        dst_u = dst_sample + dst_sample_stride * height;
+        dst_v = dst_u + halfstride * halfheight;
+      }
+      r = I420Copy(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                   dst_sample_stride, dst_u, halfstride, dst_v, halfstride,
+                   width, height);
+      break;
+    }
+    case FOURCC_I422:
+    case FOURCC_YV16: {
+      dst_sample_stride = dst_sample_stride ? dst_sample_stride : width;
+      int halfstride = (dst_sample_stride + 1) / 2;
+      uint8_t* dst_u;
+      uint8_t* dst_v;
+      if (format == FOURCC_YV16) {
+        dst_v = dst_sample + dst_sample_stride * height;
+        dst_u = dst_v + halfstride * height;
+      } else {
+        dst_u = dst_sample + dst_sample_stride * height;
+        dst_v = dst_u + halfstride * height;
+      }
+      r = I420ToI422(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                     dst_sample_stride, dst_u, halfstride, dst_v, halfstride,
+                     width, height);
+      break;
+    }
+    case FOURCC_I444:
+    case FOURCC_YV24: {
+      dst_sample_stride = dst_sample_stride ? dst_sample_stride : width;
+      uint8_t* dst_u;
+      uint8_t* dst_v;
+      if (format == FOURCC_YV24) {
+        dst_v = dst_sample + dst_sample_stride * height;
+        dst_u = dst_v + dst_sample_stride * height;
+      } else {
+        dst_u = dst_sample + dst_sample_stride * height;
+        dst_v = dst_u + dst_sample_stride * height;
+      }
+      r = I420ToI444(y, y_stride, u, u_stride, v, v_stride, dst_sample,
+                     dst_sample_stride, dst_u, dst_sample_stride, dst_v,
+                     dst_sample_stride, width, height);
+      break;
+    }
+    // Formats not supported - MJPG, biplanar, some rgb formats.
+    default:
+      return -1;  // unknown fourcc - return failure code.
+  }
+  return r;
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/convert_from_argb.cc

@@ -0,0 +1,3602 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/convert_from_argb.h"
+
+#include "libyuv/basic_types.h"
+#include "libyuv/cpu_id.h"
+#include "libyuv/planar_functions.h"
+#include "libyuv/row.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// ARGB little endian (bgra in memory) to I444
+LIBYUV_API
+int ARGBToI444(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+  void (*ARGBToUV444Row)(const uint8_t* src_argb, uint8_t* dst_u,
+                         uint8_t* dst_v, int width) = ARGBToUV444Row_C;
+  if (!src_argb || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_y == width &&
+      dst_stride_u == width && dst_stride_v == width) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_y = dst_stride_u = dst_stride_v = 0;
+  }
+#if defined(HAS_ARGBTOUV444ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUV444Row = ARGBToUV444Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUV444Row = ARGBToUV444Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUV444ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToUV444Row = ARGBToUV444Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToUV444Row = ARGBToUV444Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUV444ROW_NEON_I8MM)
+  if (TestCpuFlag(kCpuHasNeonI8MM)) {
+    ARGBToUV444Row = ARGBToUV444Row_Any_NEON_I8MM;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToUV444Row = ARGBToUV444Row_NEON_I8MM;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUV444ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToUV444Row = ARGBToUV444Row_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUV444Row = ARGBToUV444Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUV444ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToUV444Row = ARGBToUV444Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUV444Row = ARGBToUV444Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUV444ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToUV444Row = ARGBToUV444Row_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUV444Row = ARGBToUV444Row_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYRow = ARGBToYRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYRow = ARGBToYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToUV444Row(src_argb, dst_u, dst_v, width);
+    ARGBToYRow(src_argb, dst_y, width);
+    src_argb += src_stride_argb;
+    dst_y += dst_stride_y;
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+  }
+  return 0;
+}
+
+// ARGB little endian (bgra in memory) to I422
+LIBYUV_API
+int ARGBToI422(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVRow_C;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+  if (!src_argb || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_y == width &&
+      dst_stride_u * 2 == width && dst_stride_v * 2 == width) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_y = dst_stride_u = dst_stride_v = 0;
+  }
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToUVRow = ARGBToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ARGBToUVRow = ARGBToUVRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYRow = ARGBToYRow_Any_MSA;
+    ARGBToUVRow = ARGBToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    ARGBToUVRow = ARGBToUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+      ARGBToUVRow = ARGBToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYRow = ARGBToYRow_Any_LASX;
+    ARGBToUVRow = ARGBToUVRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_LASX;
+      ARGBToUVRow = ARGBToUVRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToUVRow(src_argb, 0, dst_u, dst_v, width);
+    ARGBToYRow(src_argb, dst_y, width);
+    src_argb += src_stride_argb;
+    dst_y += dst_stride_y;
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+  }
+  return 0;
+}
+
+LIBYUV_API
+int ARGBToNV12(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int y;
+  int halfwidth = (width + 1) >> 1;
+  void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVRow_C;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+  void (*MergeUVRow_)(const uint8_t* src_u, const uint8_t* src_v,
+                      uint8_t* dst_uv, int width) = MergeUVRow_C;
+  if (!src_argb || !dst_y || !dst_uv || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+#if defined(HAS_ARGBTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToUVRow = ARGBToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ARGBToUVRow = ARGBToUVRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYRow = ARGBToYRow_Any_MSA;
+    ARGBToUVRow = ARGBToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYRow = ARGBToYRow_Any_LASX;
+    ARGBToUVRow = ARGBToUVRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_LASX;
+      ARGBToUVRow = ARGBToUVRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
+#if defined(HAS_MERGEUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    MergeUVRow_ = MergeUVRow_Any_SSE2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeUVRow_ = MergeUVRow_Any_AVX2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW)) {
+    MergeUVRow_ = MergeUVRow_Any_AVX512BW;
+    if (IS_ALIGNED(halfwidth, 32)) {
+      MergeUVRow_ = MergeUVRow_AVX512BW;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeUVRow_ = MergeUVRow_Any_NEON;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    MergeUVRow_ = MergeUVRow_Any_MSA;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    MergeUVRow_ = MergeUVRow_Any_LSX;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    MergeUVRow_ = MergeUVRow_RVV;
+  }
+#endif
+  {
+    // Allocate a rows of uv.
+    align_buffer_64(row_u, ((halfwidth + 31) & ~31) * 2);
+    uint8_t* row_v = row_u + ((halfwidth + 31) & ~31);
+    if (!row_u)
+      return 1;
+
+    for (y = 0; y < height - 1; y += 2) {
+      ARGBToUVRow(src_argb, src_stride_argb, row_u, row_v, width);
+      MergeUVRow_(row_u, row_v, dst_uv, halfwidth);
+      ARGBToYRow(src_argb, dst_y, width);
+      ARGBToYRow(src_argb + src_stride_argb, dst_y + dst_stride_y, width);
+      src_argb += src_stride_argb * 2;
+      dst_y += dst_stride_y * 2;
+      dst_uv += dst_stride_uv;
+    }
+    if (height & 1) {
+      ARGBToUVRow(src_argb, 0, row_u, row_v, width);
+      MergeUVRow_(row_u, row_v, dst_uv, halfwidth);
+      ARGBToYRow(src_argb, dst_y, width);
+    }
+    free_aligned_buffer_64(row_u);
+  }
+  return 0;
+}
+
+// Same as NV12 but U and V swapped.
+LIBYUV_API
+int ARGBToNV21(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height) {
+  int y;
+  int halfwidth = (width + 1) >> 1;
+  void (*ARGBToUVRow)(const uint8_t* src_argb0, int src_stride_argb,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVRow_C;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+  void (*MergeUVRow_)(const uint8_t* src_u, const uint8_t* src_v,
+                      uint8_t* dst_vu, int width) = MergeUVRow_C;
+  if (!src_argb || !dst_y || !dst_vu || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToUVRow = ARGBToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ARGBToUVRow = ARGBToUVRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYRow = ARGBToYRow_Any_MSA;
+    ARGBToUVRow = ARGBToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    ARGBToUVRow = ARGBToUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+      ARGBToUVRow = ARGBToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYRow = ARGBToYRow_Any_LASX;
+    ARGBToUVRow = ARGBToUVRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_LASX;
+      ARGBToUVRow = ARGBToUVRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
+#if defined(HAS_MERGEUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    MergeUVRow_ = MergeUVRow_Any_SSE2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeUVRow_ = MergeUVRow_Any_AVX2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW)) {
+    MergeUVRow_ = MergeUVRow_Any_AVX512BW;
+    if (IS_ALIGNED(halfwidth, 64)) {
+      MergeUVRow_ = MergeUVRow_AVX512BW;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeUVRow_ = MergeUVRow_Any_NEON;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    MergeUVRow_ = MergeUVRow_Any_MSA;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    MergeUVRow_ = MergeUVRow_Any_LSX;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    MergeUVRow_ = MergeUVRow_RVV;
+  }
+#endif
+  {
+    // Allocate a rows of uv.
+    align_buffer_64(row_u, ((halfwidth + 31) & ~31) * 2);
+    uint8_t* row_v = row_u + ((halfwidth + 31) & ~31);
+    if (!row_u)
+      return 1;
+
+    for (y = 0; y < height - 1; y += 2) {
+      ARGBToUVRow(src_argb, src_stride_argb, row_u, row_v, width);
+      MergeUVRow_(row_v, row_u, dst_vu, halfwidth);
+      ARGBToYRow(src_argb, dst_y, width);
+      ARGBToYRow(src_argb + src_stride_argb, dst_y + dst_stride_y, width);
+      src_argb += src_stride_argb * 2;
+      dst_y += dst_stride_y * 2;
+      dst_vu += dst_stride_vu;
+    }
+    if (height & 1) {
+      ARGBToUVRow(src_argb, 0, row_u, row_v, width);
+      MergeUVRow_(row_v, row_u, dst_vu, halfwidth);
+      ARGBToYRow(src_argb, dst_y, width);
+    }
+    free_aligned_buffer_64(row_u);
+  }
+  return 0;
+}
+
+LIBYUV_API
+int ABGRToNV12(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int y;
+  int halfwidth = (width + 1) >> 1;
+  void (*ABGRToUVRow)(const uint8_t* src_abgr0, int src_stride_abgr,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ABGRToUVRow_C;
+  void (*ABGRToYRow)(const uint8_t* src_abgr, uint8_t* dst_y, int width) =
+      ABGRToYRow_C;
+  void (*MergeUVRow_)(const uint8_t* src_u, const uint8_t* src_v,
+                      uint8_t* dst_uv, int width) = MergeUVRow_C;
+  if (!src_abgr || !dst_y || !dst_uv || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_abgr = src_abgr + (height - 1) * src_stride_abgr;
+    src_stride_abgr = -src_stride_abgr;
+  }
+#if defined(HAS_ABGRTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ABGRToYRow = ABGRToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ABGRToUVRow = ABGRToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToUVRow = ABGRToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ABGRToYRow = ABGRToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToYRow = ABGRToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ABGRToUVRow = ABGRToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToUVRow = ABGRToUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ABGRToYRow = ABGRToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ABGRToYRow = ABGRToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ABGRToUVRow = ABGRToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToUVRow = ABGRToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ABGRToUVRow = ABGRToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ABGRToUVRow = ABGRToUVRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_MSA) && defined(HAS_ABGRTOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ABGRToYRow = ABGRToYRow_Any_MSA;
+    ABGRToUVRow = ABGRToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToUVRow = ABGRToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ABGRToYRow = ABGRToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ABGRToYRow = ABGRToYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToYRow = ABGRToYRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ABGRToYRow = ABGRToYRow_RVV;
+  }
+#endif
+#if defined(HAS_MERGEUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    MergeUVRow_ = MergeUVRow_Any_SSE2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeUVRow_ = MergeUVRow_Any_AVX2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW)) {
+    MergeUVRow_ = MergeUVRow_Any_AVX512BW;
+    if (IS_ALIGNED(halfwidth, 64)) {
+      MergeUVRow_ = MergeUVRow_AVX512BW;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeUVRow_ = MergeUVRow_Any_NEON;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    MergeUVRow_ = MergeUVRow_Any_MSA;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    MergeUVRow_ = MergeUVRow_Any_LSX;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    MergeUVRow_ = MergeUVRow_RVV;
+  }
+#endif
+  {
+    // Allocate a rows of uv.
+    align_buffer_64(row_u, ((halfwidth + 31) & ~31) * 2);
+    uint8_t* row_v = row_u + ((halfwidth + 31) & ~31);
+    if (!row_u)
+      return 1;
+
+    for (y = 0; y < height - 1; y += 2) {
+      ABGRToUVRow(src_abgr, src_stride_abgr, row_u, row_v, width);
+      MergeUVRow_(row_u, row_v, dst_uv, halfwidth);
+      ABGRToYRow(src_abgr, dst_y, width);
+      ABGRToYRow(src_abgr + src_stride_abgr, dst_y + dst_stride_y, width);
+      src_abgr += src_stride_abgr * 2;
+      dst_y += dst_stride_y * 2;
+      dst_uv += dst_stride_uv;
+    }
+    if (height & 1) {
+      ABGRToUVRow(src_abgr, 0, row_u, row_v, width);
+      MergeUVRow_(row_u, row_v, dst_uv, halfwidth);
+      ABGRToYRow(src_abgr, dst_y, width);
+    }
+    free_aligned_buffer_64(row_u);
+  }
+  return 0;
+}
+
+// Same as NV12 but U and V swapped.
+LIBYUV_API
+int ABGRToNV21(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height) {
+  int y;
+  int halfwidth = (width + 1) >> 1;
+  void (*ABGRToUVRow)(const uint8_t* src_abgr0, int src_stride_abgr,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ABGRToUVRow_C;
+  void (*ABGRToYRow)(const uint8_t* src_abgr, uint8_t* dst_y, int width) =
+      ABGRToYRow_C;
+  void (*MergeUVRow_)(const uint8_t* src_u, const uint8_t* src_v,
+                      uint8_t* dst_vu, int width) = MergeUVRow_C;
+  if (!src_abgr || !dst_y || !dst_vu || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_abgr = src_abgr + (height - 1) * src_stride_abgr;
+    src_stride_abgr = -src_stride_abgr;
+  }
+#if defined(HAS_ABGRTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ABGRToYRow = ABGRToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ABGRToUVRow = ABGRToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToUVRow = ABGRToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ABGRToYRow = ABGRToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToYRow = ABGRToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ABGRToUVRow = ABGRToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToUVRow = ABGRToUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ABGRToYRow = ABGRToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ABGRToYRow = ABGRToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ABGRToUVRow = ABGRToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToUVRow = ABGRToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ABGRToUVRow = ABGRToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ABGRToUVRow = ABGRToUVRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_MSA) && defined(HAS_ABGRTOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ABGRToYRow = ABGRToYRow_Any_MSA;
+    ABGRToUVRow = ABGRToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToUVRow = ABGRToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ABGRToYRow = ABGRToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ABGRToYRow = ABGRToYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToYRow = ABGRToYRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ABGRToYRow = ABGRToYRow_RVV;
+  }
+#endif
+#if defined(HAS_MERGEUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    MergeUVRow_ = MergeUVRow_Any_SSE2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeUVRow_ = MergeUVRow_Any_AVX2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW)) {
+    MergeUVRow_ = MergeUVRow_Any_AVX512BW;
+    if (IS_ALIGNED(halfwidth, 64)) {
+      MergeUVRow_ = MergeUVRow_AVX512BW;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeUVRow_ = MergeUVRow_Any_NEON;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    MergeUVRow_ = MergeUVRow_Any_MSA;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    MergeUVRow_ = MergeUVRow_Any_LSX;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    MergeUVRow_ = MergeUVRow_RVV;
+  }
+#endif
+  {
+    // Allocate a rows of uv.
+    align_buffer_64(row_u, ((halfwidth + 31) & ~31) * 2);
+    uint8_t* row_v = row_u + ((halfwidth + 31) & ~31);
+    if (!row_u)
+      return 1;
+
+    for (y = 0; y < height - 1; y += 2) {
+      ABGRToUVRow(src_abgr, src_stride_abgr, row_u, row_v, width);
+      MergeUVRow_(row_v, row_u, dst_vu, halfwidth);
+      ABGRToYRow(src_abgr, dst_y, width);
+      ABGRToYRow(src_abgr + src_stride_abgr, dst_y + dst_stride_y, width);
+      src_abgr += src_stride_abgr * 2;
+      dst_y += dst_stride_y * 2;
+      dst_vu += dst_stride_vu;
+    }
+    if (height & 1) {
+      ABGRToUVRow(src_abgr, 0, row_u, row_v, width);
+      MergeUVRow_(row_v, row_u, dst_vu, halfwidth);
+      ABGRToYRow(src_abgr, dst_y, width);
+    }
+    free_aligned_buffer_64(row_u);
+  }
+  return 0;
+}
+
+// Convert ARGB to YUY2.
+LIBYUV_API
+int ARGBToYUY2(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_yuy2,
+               int dst_stride_yuy2,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToUVRow)(const uint8_t* src_argb, int src_stride_argb,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVRow_C;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+  void (*I422ToYUY2Row)(const uint8_t* src_y, const uint8_t* src_u,
+                        const uint8_t* src_v, uint8_t* dst_yuy2, int width) =
+      I422ToYUY2Row_C;
+
+  if (!src_argb || !dst_yuy2 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_yuy2 = dst_yuy2 + (height - 1) * dst_stride_yuy2;
+    dst_stride_yuy2 = -dst_stride_yuy2;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_yuy2 == width * 2) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_yuy2 = 0;
+  }
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToUVRow = ARGBToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ARGBToUVRow = ARGBToUVRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYRow = ARGBToYRow_Any_MSA;
+    ARGBToUVRow = ARGBToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    ARGBToUVRow = ARGBToUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+      ARGBToUVRow = ARGBToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYRow = ARGBToYRow_Any_LASX;
+    ARGBToUVRow = ARGBToUVRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_LASX;
+      ARGBToUVRow = ARGBToUVRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToYUY2Row = I422ToYUY2Row_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToYUY2Row = I422ToYUY2Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToYUY2Row = I422ToYUY2Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToYUY2Row = I422ToYUY2Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToYUY2Row = I422ToYUY2Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOYUY2ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToYUY2Row = I422ToYUY2Row_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToYUY2Row = I422ToYUY2Row_LASX;
+    }
+  }
+#endif
+
+  {
+    // Allocate a rows of yuv.
+    align_buffer_64(row_y, ((width + 63) & ~63) * 2);
+    uint8_t* row_u = row_y + ((width + 63) & ~63);
+    uint8_t* row_v = row_u + ((width + 63) & ~63) / 2;
+    if (!row_y)
+      return 1;
+
+    for (y = 0; y < height; ++y) {
+      ARGBToUVRow(src_argb, 0, row_u, row_v, width);
+      ARGBToYRow(src_argb, row_y, width);
+      I422ToYUY2Row(row_y, row_u, row_v, dst_yuy2, width);
+      src_argb += src_stride_argb;
+      dst_yuy2 += dst_stride_yuy2;
+    }
+
+    free_aligned_buffer_64(row_y);
+  }
+  return 0;
+}
+
+// Convert ARGB to UYVY.
+LIBYUV_API
+int ARGBToUYVY(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_uyvy,
+               int dst_stride_uyvy,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToUVRow)(const uint8_t* src_argb, int src_stride_argb,
+                      uint8_t* dst_u, uint8_t* dst_v, int width) =
+      ARGBToUVRow_C;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+  void (*I422ToUYVYRow)(const uint8_t* src_y, const uint8_t* src_u,
+                        const uint8_t* src_v, uint8_t* dst_uyvy, int width) =
+      I422ToUYVYRow_C;
+
+  if (!src_argb || !dst_uyvy || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_uyvy = dst_uyvy + (height - 1) * dst_stride_uyvy;
+    dst_stride_uyvy = -dst_stride_uyvy;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_uyvy == width * 2) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_uyvy = 0;
+  }
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToUVRow = ARGBToUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVRow = ARGBToUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToUVRow = ARGBToUVRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ARGBToUVRow = ARGBToUVRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYRow = ARGBToYRow_Any_MSA;
+    ARGBToUVRow = ARGBToUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVRow = ARGBToUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    ARGBToUVRow = ARGBToUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+      ARGBToUVRow = ARGBToUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYRow = ARGBToYRow_Any_LASX;
+    ARGBToUVRow = ARGBToUVRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_LASX;
+      ARGBToUVRow = ARGBToUVRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToUYVYRow = I422ToUYVYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToUYVYRow = I422ToUYVYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToUYVYRow = I422ToUYVYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToUYVYRow = I422ToUYVYRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToUYVYRow = I422ToUYVYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_I422TOUYVYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    I422ToUYVYRow = I422ToUYVYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      I422ToUYVYRow = I422ToUYVYRow_LASX;
+    }
+  }
+#endif
+
+  {
+    // Allocate a rows of yuv.
+    align_buffer_64(row_y, ((width + 63) & ~63) * 2);
+    uint8_t* row_u = row_y + ((width + 63) & ~63);
+    uint8_t* row_v = row_u + ((width + 63) & ~63) / 2;
+    if (!row_y)
+      return 1;
+
+    for (y = 0; y < height; ++y) {
+      ARGBToUVRow(src_argb, 0, row_u, row_v, width);
+      ARGBToYRow(src_argb, row_y, width);
+      I422ToUYVYRow(row_y, row_u, row_v, dst_uyvy, width);
+      src_argb += src_stride_argb;
+      dst_uyvy += dst_stride_uyvy;
+    }
+
+    free_aligned_buffer_64(row_y);
+  }
+  return 0;
+}
+
+// Convert ARGB to I400.
+LIBYUV_API
+int ARGBToI400(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToYRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYRow_C;
+  if (!src_argb || !dst_y || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_y == width) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_y = 0;
+  }
+#if defined(HAS_ARGBTOYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYRow = ARGBToYRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYRow = ARGBToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYRow = ARGBToYRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYRow = ARGBToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYRow = ARGBToYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYRow = ARGBToYRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToYRow(src_argb, dst_y, width);
+    src_argb += src_stride_argb;
+    dst_y += dst_stride_y;
+  }
+  return 0;
+}
+
+#ifndef __riscv
+// Shuffle table for converting ARGB to RGBA.
+static const uvec8 kShuffleMaskARGBToRGBA = {
+    3u, 0u, 1u, 2u, 7u, 4u, 5u, 6u, 11u, 8u, 9u, 10u, 15u, 12u, 13u, 14u};
+
+// Convert ARGB to RGBA.
+LIBYUV_API
+int ARGBToRGBA(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_rgba,
+               int dst_stride_rgba,
+               int width,
+               int height) {
+  return ARGBShuffle(src_argb, src_stride_argb, dst_rgba, dst_stride_rgba,
+                     (const uint8_t*)(&kShuffleMaskARGBToRGBA), width, height);
+}
+#else
+// Convert ARGB to RGBA.
+LIBYUV_API
+int ARGBToRGBA(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_rgba,
+               int dst_stride_rgba,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToRGBARow)(const uint8_t* src_argb, uint8_t* dst_rgba, int width) =
+      ARGBToRGBARow_C;
+  if (!src_argb || !dst_rgba || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_rgba == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_rgba = 0;
+  }
+
+#if defined(HAS_ARGBTORGBAROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToRGBARow = ARGBToRGBARow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToRGBARow(src_argb, dst_rgba, width);
+    src_argb += src_stride_argb;
+    dst_rgba += dst_stride_rgba;
+  }
+  return 0;
+}
+#endif
+
+// Convert ARGB To RGB24.
+LIBYUV_API
+int ARGBToRGB24(const uint8_t* src_argb,
+                int src_stride_argb,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height) {
+  int y;
+  void (*ARGBToRGB24Row)(const uint8_t* src_argb, uint8_t* dst_rgb, int width) =
+      ARGBToRGB24Row_C;
+  if (!src_argb || !dst_rgb24 || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_rgb24 == width * 3) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_rgb24 = 0;
+  }
+#if defined(HAS_ARGBTORGB24ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToRGB24Row = ARGBToRGB24Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToRGB24Row = ARGBToRGB24Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB24ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToRGB24Row = ARGBToRGB24Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToRGB24Row = ARGBToRGB24Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB24ROW_AVX512VBMI)
+  if (TestCpuFlag(kCpuHasAVX512VBMI)) {
+    ARGBToRGB24Row = ARGBToRGB24Row_Any_AVX512VBMI;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToRGB24Row = ARGBToRGB24Row_AVX512VBMI;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB24ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToRGB24Row = ARGBToRGB24Row_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToRGB24Row = ARGBToRGB24Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB24ROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToRGB24Row = ARGBToRGB24Row_SVE2;
+  }
+#endif
+#if defined(HAS_ARGBTORGB24ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToRGB24Row = ARGBToRGB24Row_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToRGB24Row = ARGBToRGB24Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB24ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToRGB24Row = ARGBToRGB24Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToRGB24Row = ARGBToRGB24Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB24ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToRGB24Row = ARGBToRGB24Row_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToRGB24Row = ARGBToRGB24Row_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB24ROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToRGB24Row = ARGBToRGB24Row_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToRGB24Row(src_argb, dst_rgb24, width);
+    src_argb += src_stride_argb;
+    dst_rgb24 += dst_stride_rgb24;
+  }
+  return 0;
+}
+
+// Convert ARGB To RAW.
+LIBYUV_API
+int ARGBToRAW(const uint8_t* src_argb,
+              int src_stride_argb,
+              uint8_t* dst_raw,
+              int dst_stride_raw,
+              int width,
+              int height) {
+  int y;
+  void (*ARGBToRAWRow)(const uint8_t* src_argb, uint8_t* dst_rgb, int width) =
+      ARGBToRAWRow_C;
+  if (!src_argb || !dst_raw || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_raw == width * 3) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_raw = 0;
+  }
+#if defined(HAS_ARGBTORAWROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToRAWRow = ARGBToRAWRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToRAWRow = ARGBToRAWRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORAWROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToRAWRow = ARGBToRAWRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToRAWRow = ARGBToRAWRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORAWROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToRAWRow = ARGBToRAWRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRAWRow = ARGBToRAWRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORAWROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToRAWRow = ARGBToRAWRow_SVE2;
+  }
+#endif
+#if defined(HAS_ARGBTORAWROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToRAWRow = ARGBToRAWRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToRAWRow = ARGBToRAWRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORAWROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToRAWRow = ARGBToRAWRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToRAWRow = ARGBToRAWRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORAWROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToRAWRow = ARGBToRAWRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToRAWRow = ARGBToRAWRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORAWROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToRAWRow = ARGBToRAWRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToRAWRow(src_argb, dst_raw, width);
+    src_argb += src_stride_argb;
+    dst_raw += dst_stride_raw;
+  }
+  return 0;
+}
+
+// Ordered 8x8 dither for 888 to 565.  Values from 0 to 7.
+static const uint8_t kDither565_4x4[16] = {
+    0, 4, 1, 5, 6, 2, 7, 3, 1, 5, 0, 4, 7, 3, 6, 2,
+};
+
+// Convert ARGB To RGB565 with 4x4 dither matrix (16 bytes).
+LIBYUV_API
+int ARGBToRGB565Dither(const uint8_t* src_argb,
+                       int src_stride_argb,
+                       uint8_t* dst_rgb565,
+                       int dst_stride_rgb565,
+                       const uint8_t* dither4x4,
+                       int width,
+                       int height) {
+  int y;
+  void (*ARGBToRGB565DitherRow)(const uint8_t* src_argb, uint8_t* dst_rgb,
+                                uint32_t dither4, int width) =
+      ARGBToRGB565DitherRow_C;
+  if (!src_argb || !dst_rgb565 || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  if (!dither4x4) {
+    dither4x4 = kDither565_4x4;
+  }
+#if defined(HAS_ARGBTORGB565DITHERROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_SSE2;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_SVE2;
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565DITHERROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToRGB565DitherRow(src_argb, dst_rgb565,
+                          *(const uint32_t*)(dither4x4 + ((y & 3) << 2)),
+                          width);
+    src_argb += src_stride_argb;
+    dst_rgb565 += dst_stride_rgb565;
+  }
+  return 0;
+}
+
+// Convert ARGB To RGB565.
+// TODO(fbarchard): Consider using dither function low level with zeros.
+LIBYUV_API
+int ARGBToRGB565(const uint8_t* src_argb,
+                 int src_stride_argb,
+                 uint8_t* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height) {
+  int y;
+  void (*ARGBToRGB565Row)(const uint8_t* src_argb, uint8_t* dst_rgb,
+                          int width) = ARGBToRGB565Row_C;
+  if (!src_argb || !dst_rgb565 || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_rgb565 == width * 2) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_rgb565 = 0;
+  }
+#if defined(HAS_ARGBTORGB565ROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBToRGB565Row = ARGBToRGB565Row_Any_SSE2;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBToRGB565Row = ARGBToRGB565Row_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToRGB565Row = ARGBToRGB565Row_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRGB565Row = ARGBToRGB565Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToRGB565Row = ARGBToRGB565Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRGB565Row = ARGBToRGB565Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565ROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToRGB565Row = ARGBToRGB565Row_SVE2;
+  }
+#endif
+#if defined(HAS_ARGBTORGB565ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToRGB565Row = ARGBToRGB565Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRGB565Row = ARGBToRGB565Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTORGB565ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToRGB565Row = ARGBToRGB565Row_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToRGB565Row = ARGBToRGB565Row_LSX;
+    }
+  }
+#endif
+
+#if defined(HAS_ARGBTORGB565ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToRGB565Row = ARGBToRGB565Row_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToRGB565Row = ARGBToRGB565Row_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToRGB565Row(src_argb, dst_rgb565, width);
+    src_argb += src_stride_argb;
+    dst_rgb565 += dst_stride_rgb565;
+  }
+  return 0;
+}
+
+// Convert ARGB To ARGB1555.
+LIBYUV_API
+int ARGBToARGB1555(const uint8_t* src_argb,
+                   int src_stride_argb,
+                   uint8_t* dst_argb1555,
+                   int dst_stride_argb1555,
+                   int width,
+                   int height) {
+  int y;
+  void (*ARGBToARGB1555Row)(const uint8_t* src_argb, uint8_t* dst_rgb,
+                            int width) = ARGBToARGB1555Row_C;
+  if (!src_argb || !dst_argb1555 || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_argb1555 == width * 2) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_argb1555 = 0;
+  }
+#if defined(HAS_ARGBTOARGB1555ROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBToARGB1555Row = ARGBToARGB1555Row_Any_SSE2;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBToARGB1555Row = ARGBToARGB1555Row_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOARGB1555ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToARGB1555Row = ARGBToARGB1555Row_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToARGB1555Row = ARGBToARGB1555Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOARGB1555ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToARGB1555Row = ARGBToARGB1555Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToARGB1555Row = ARGBToARGB1555Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOARGB1555ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToARGB1555Row = ARGBToARGB1555Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToARGB1555Row = ARGBToARGB1555Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOARGB1555ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToARGB1555Row = ARGBToARGB1555Row_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToARGB1555Row = ARGBToARGB1555Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOARGB1555ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToARGB1555Row = ARGBToARGB1555Row_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToARGB1555Row = ARGBToARGB1555Row_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToARGB1555Row(src_argb, dst_argb1555, width);
+    src_argb += src_stride_argb;
+    dst_argb1555 += dst_stride_argb1555;
+  }
+  return 0;
+}
+
+// Convert ARGB To ARGB4444.
+LIBYUV_API
+int ARGBToARGB4444(const uint8_t* src_argb,
+                   int src_stride_argb,
+                   uint8_t* dst_argb4444,
+                   int dst_stride_argb4444,
+                   int width,
+                   int height) {
+  int y;
+  void (*ARGBToARGB4444Row)(const uint8_t* src_argb, uint8_t* dst_rgb,
+                            int width) = ARGBToARGB4444Row_C;
+  if (!src_argb || !dst_argb4444 || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_argb4444 == width * 2) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_argb4444 = 0;
+  }
+#if defined(HAS_ARGBTOARGB4444ROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBToARGB4444Row = ARGBToARGB4444Row_Any_SSE2;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBToARGB4444Row = ARGBToARGB4444Row_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOARGB4444ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToARGB4444Row = ARGBToARGB4444Row_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToARGB4444Row = ARGBToARGB4444Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOARGB4444ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToARGB4444Row = ARGBToARGB4444Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToARGB4444Row = ARGBToARGB4444Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOARGB4444ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToARGB4444Row = ARGBToARGB4444Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToARGB4444Row = ARGBToARGB4444Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOARGB4444ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToARGB4444Row = ARGBToARGB4444Row_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToARGB4444Row = ARGBToARGB4444Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOARGB4444ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToARGB4444Row = ARGBToARGB4444Row_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToARGB4444Row = ARGBToARGB4444Row_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToARGB4444Row(src_argb, dst_argb4444, width);
+    src_argb += src_stride_argb;
+    dst_argb4444 += dst_stride_argb4444;
+  }
+  return 0;
+}
+
+// Convert ABGR To AR30.
+LIBYUV_API
+int ABGRToAR30(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height) {
+  int y;
+  void (*ABGRToAR30Row)(const uint8_t* src_abgr, uint8_t* dst_rgb, int width) =
+      ABGRToAR30Row_C;
+  if (!src_abgr || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_abgr = src_abgr + (height - 1) * src_stride_abgr;
+    src_stride_abgr = -src_stride_abgr;
+  }
+  // Coalesce rows.
+  if (src_stride_abgr == width * 4 && dst_stride_ar30 == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_abgr = dst_stride_ar30 = 0;
+  }
+#if defined(HAS_ABGRTOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ABGRToAR30Row = ABGRToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ABGRToAR30Row = ABGRToAR30Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ABGRToAR30Row = ABGRToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ABGRToAR30Row = ABGRToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ABGRToAR30Row = ABGRToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ABGRToAR30Row = ABGRToAR30Row_AVX2;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    ABGRToAR30Row(src_abgr, dst_ar30, width);
+    src_abgr += src_stride_abgr;
+    dst_ar30 += dst_stride_ar30;
+  }
+  return 0;
+}
+
+// Convert ARGB To AR30.
+LIBYUV_API
+int ARGBToAR30(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_ar30,
+               int dst_stride_ar30,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToAR30Row)(const uint8_t* src_argb, uint8_t* dst_rgb, int width) =
+      ARGBToAR30Row_C;
+  if (!src_argb || !dst_ar30 || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_ar30 == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_ar30 = 0;
+  }
+#if defined(HAS_ARGBTOAR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToAR30Row = ARGBToAR30Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToAR30Row = ARGBToAR30Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOAR30ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToAR30Row = ARGBToAR30Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBToAR30Row = ARGBToAR30Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOAR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToAR30Row = ARGBToAR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToAR30Row = ARGBToAR30Row_AVX2;
+    }
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    ARGBToAR30Row(src_argb, dst_ar30, width);
+    src_argb += src_stride_argb;
+    dst_ar30 += dst_stride_ar30;
+  }
+  return 0;
+}
+
+// Convert ARGB to J420. (JPeg full range I420).
+LIBYUV_API
+int ARGBToJ420(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               uint8_t* dst_uj,
+               int dst_stride_uj,
+               uint8_t* dst_vj,
+               int dst_stride_vj,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToUVJRow)(const uint8_t* src_argb0, int src_stride_argb,
+                       uint8_t* dst_uj, uint8_t* dst_vj, int width) =
+      ARGBToUVJRow_C;
+  void (*ARGBToYJRow)(const uint8_t* src_argb, uint8_t* dst_yj, int width) =
+      ARGBToYJRow_C;
+  if (!src_argb || !dst_yj || !dst_uj || !dst_vj || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+#if defined(HAS_ARGBTOYJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYJRow = ARGBToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYJRow = ARGBToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVJRow = ARGBToUVJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ARGBToUVJRow = ARGBToUVJRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYJRow = ARGBToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVJRow = ARGBToUVJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYJRow = ARGBToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYJRow = ARGBToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVJRow = ARGBToUVJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_MSA) && defined(HAS_ARGBTOUVJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYJRow = ARGBToYJRow_Any_MSA;
+    ARGBToUVJRow = ARGBToUVJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVJRow = ARGBToUVJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_LSX) && defined(HAS_ARGBTOUVJROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYJRow = ARGBToYJRow_Any_LSX;
+    ARGBToUVJRow = ARGBToUVJRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_LSX;
+      ARGBToUVJRow = ARGBToUVJRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_LASX) && defined(HAS_ARGBTOUVJROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYJRow = ARGBToYJRow_Any_LASX;
+    ARGBToUVJRow = ARGBToUVJRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYJRow = ARGBToYJRow_LASX;
+      ARGBToUVJRow = ARGBToUVJRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYJRow = ARGBToYJRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    ARGBToUVJRow(src_argb, src_stride_argb, dst_uj, dst_vj, width);
+    ARGBToYJRow(src_argb, dst_yj, width);
+    ARGBToYJRow(src_argb + src_stride_argb, dst_yj + dst_stride_yj, width);
+    src_argb += src_stride_argb * 2;
+    dst_yj += dst_stride_yj * 2;
+    dst_uj += dst_stride_uj;
+    dst_vj += dst_stride_vj;
+  }
+  if (height & 1) {
+    ARGBToUVJRow(src_argb, 0, dst_uj, dst_vj, width);
+    ARGBToYJRow(src_argb, dst_yj, width);
+  }
+  return 0;
+}
+
+// Convert ARGB to J422. (JPeg full range I422).
+LIBYUV_API
+int ARGBToJ422(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               uint8_t* dst_uj,
+               int dst_stride_uj,
+               uint8_t* dst_vj,
+               int dst_stride_vj,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToUVJRow)(const uint8_t* src_argb0, int src_stride_argb,
+                       uint8_t* dst_uj, uint8_t* dst_vj, int width) =
+      ARGBToUVJRow_C;
+  void (*ARGBToYJRow)(const uint8_t* src_argb, uint8_t* dst_yj, int width) =
+      ARGBToYJRow_C;
+  if (!src_argb || !dst_yj || !dst_uj || !dst_vj || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_yj == width &&
+      dst_stride_uj * 2 == width && dst_stride_vj * 2 == width) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_yj = dst_stride_uj = dst_stride_vj = 0;
+  }
+#if defined(HAS_ARGBTOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYJRow = ARGBToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVJRow = ARGBToUVJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYJRow = ARGBToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYJRow = ARGBToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVJRow = ARGBToUVJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYJRow = ARGBToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYJRow = ARGBToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVJRow = ARGBToUVJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ARGBToUVJRow = ARGBToUVJRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_MSA) && defined(HAS_ARGBTOUVJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYJRow = ARGBToYJRow_Any_MSA;
+    ARGBToUVJRow = ARGBToUVJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVJRow = ARGBToUVJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_LSX) && defined(HAS_ARGBTOUVJROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYJRow = ARGBToYJRow_Any_LSX;
+    ARGBToUVJRow = ARGBToUVJRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_LSX;
+      ARGBToUVJRow = ARGBToUVJRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_LASX) && defined(HAS_ARGBTOUVJROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYJRow = ARGBToYJRow_Any_LASX;
+    ARGBToUVJRow = ARGBToUVJRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYJRow = ARGBToYJRow_LASX;
+      ARGBToUVJRow = ARGBToUVJRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYJRow = ARGBToYJRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToUVJRow(src_argb, 0, dst_uj, dst_vj, width);
+    ARGBToYJRow(src_argb, dst_yj, width);
+    src_argb += src_stride_argb;
+    dst_yj += dst_stride_yj;
+    dst_uj += dst_stride_uj;
+    dst_vj += dst_stride_vj;
+  }
+  return 0;
+}
+
+// Convert ARGB to J400.
+LIBYUV_API
+int ARGBToJ400(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToYJRow)(const uint8_t* src_argb, uint8_t* dst_yj, int width) =
+      ARGBToYJRow_C;
+  if (!src_argb || !dst_yj || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_yj == width) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_yj = 0;
+  }
+#if defined(HAS_ARGBTOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYJRow = ARGBToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYJRow = ARGBToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYJRow = ARGBToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYJRow = ARGBToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYJRow = ARGBToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYJRow = ARGBToYJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYJRow = ARGBToYJRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToYJRow(src_argb, dst_yj, width);
+    src_argb += src_stride_argb;
+    dst_yj += dst_stride_yj;
+  }
+  return 0;
+}
+
+// Convert RGBA to J400.
+LIBYUV_API
+int RGBAToJ400(const uint8_t* src_rgba,
+               int src_stride_rgba,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               int width,
+               int height) {
+  int y;
+  void (*RGBAToYJRow)(const uint8_t* src_rgba, uint8_t* dst_yj, int width) =
+      RGBAToYJRow_C;
+  if (!src_rgba || !dst_yj || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_rgba = src_rgba + (height - 1) * src_stride_rgba;
+    src_stride_rgba = -src_stride_rgba;
+  }
+  // Coalesce rows.
+  if (src_stride_rgba == width * 4 && dst_stride_yj == width) {
+    width *= height;
+    height = 1;
+    src_stride_rgba = dst_stride_yj = 0;
+  }
+#if defined(HAS_RGBATOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RGBAToYJRow = RGBAToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToYJRow = RGBAToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    RGBAToYJRow = RGBAToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      RGBAToYJRow = RGBAToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RGBAToYJRow = RGBAToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToYJRow = RGBAToYJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    RGBAToYJRow = RGBAToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToYJRow = RGBAToYJRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RGBAToYJRow = RGBAToYJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToYJRow = RGBAToYJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYJROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RGBAToYJRow = RGBAToYJRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToYJRow = RGBAToYJRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYJROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    RGBAToYJRow = RGBAToYJRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RGBAToYJRow = RGBAToYJRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_RGBATOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RGBAToYJRow = RGBAToYJRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    RGBAToYJRow(src_rgba, dst_yj, width);
+    src_rgba += src_stride_rgba;
+    dst_yj += dst_stride_yj;
+  }
+  return 0;
+}
+
+// Convert ABGR to J420. (JPeg full range I420).
+LIBYUV_API
+int ABGRToJ420(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               uint8_t* dst_uj,
+               int dst_stride_uj,
+               uint8_t* dst_vj,
+               int dst_stride_vj,
+               int width,
+               int height) {
+  int y;
+  void (*ABGRToUVJRow)(const uint8_t* src_abgr0, int src_stride_abgr,
+                       uint8_t* dst_uj, uint8_t* dst_vj, int width) =
+      ABGRToUVJRow_C;
+  void (*ABGRToYJRow)(const uint8_t* src_abgr, uint8_t* dst_yj, int width) =
+      ABGRToYJRow_C;
+  if (!src_abgr || !dst_yj || !dst_uj || !dst_vj || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_abgr = src_abgr + (height - 1) * src_stride_abgr;
+    src_stride_abgr = -src_stride_abgr;
+  }
+#if defined(HAS_ABGRTOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ABGRToYJRow = ABGRToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ABGRToUVJRow = ABGRToUVJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToUVJRow = ABGRToUVJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ABGRToYJRow = ABGRToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToYJRow = ABGRToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ABGRToUVJRow = ABGRToUVJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToUVJRow = ABGRToUVJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ABGRToYJRow = ABGRToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ABGRToYJRow = ABGRToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ABGRToUVJRow = ABGRToUVJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToUVJRow = ABGRToUVJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVJROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ABGRToUVJRow = ABGRToUVJRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ABGRToUVJRow = ABGRToUVJRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_MSA) && defined(HAS_ABGRTOUVJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ABGRToYJRow = ABGRToYJRow_Any_MSA;
+    ABGRToUVJRow = ABGRToUVJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_MSA;
+      ABGRToUVJRow = ABGRToUVJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ABGRToYJRow = ABGRToYJRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ABGRToYJRow = ABGRToYJRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToYJRow = ABGRToYJRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ABGRToYJRow = ABGRToYJRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    ABGRToUVJRow(src_abgr, src_stride_abgr, dst_uj, dst_vj, width);
+    ABGRToYJRow(src_abgr, dst_yj, width);
+    ABGRToYJRow(src_abgr + src_stride_abgr, dst_yj + dst_stride_yj, width);
+    src_abgr += src_stride_abgr * 2;
+    dst_yj += dst_stride_yj * 2;
+    dst_uj += dst_stride_uj;
+    dst_vj += dst_stride_vj;
+  }
+  if (height & 1) {
+    ABGRToUVJRow(src_abgr, 0, dst_uj, dst_vj, width);
+    ABGRToYJRow(src_abgr, dst_yj, width);
+  }
+  return 0;
+}
+
+// Convert ABGR to J422. (JPeg full range I422).
+LIBYUV_API
+int ABGRToJ422(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               uint8_t* dst_uj,
+               int dst_stride_uj,
+               uint8_t* dst_vj,
+               int dst_stride_vj,
+               int width,
+               int height) {
+  int y;
+  void (*ABGRToUVJRow)(const uint8_t* src_abgr0, int src_stride_abgr,
+                       uint8_t* dst_uj, uint8_t* dst_vj, int width) =
+      ABGRToUVJRow_C;
+  void (*ABGRToYJRow)(const uint8_t* src_abgr, uint8_t* dst_yj, int width) =
+      ABGRToYJRow_C;
+  if (!src_abgr || !dst_yj || !dst_uj || !dst_vj || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_abgr = src_abgr + (height - 1) * src_stride_abgr;
+    src_stride_abgr = -src_stride_abgr;
+  }
+  // Coalesce rows.
+  if (src_stride_abgr == width * 4 && dst_stride_yj == width &&
+      dst_stride_uj * 2 == width && dst_stride_vj * 2 == width) {
+    width *= height;
+    height = 1;
+    src_stride_abgr = dst_stride_yj = dst_stride_uj = dst_stride_vj = 0;
+  }
+#if defined(HAS_ABGRTOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ABGRToYJRow = ABGRToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ABGRToUVJRow = ABGRToUVJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToUVJRow = ABGRToUVJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ABGRToYJRow = ABGRToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToYJRow = ABGRToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ABGRToUVJRow = ABGRToUVJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToUVJRow = ABGRToUVJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ABGRToYJRow = ABGRToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ABGRToYJRow = ABGRToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ABGRToUVJRow = ABGRToUVJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToUVJRow = ABGRToUVJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOUVJROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    ABGRToUVJRow = ABGRToUVJRow_Any_SVE2;
+    if (IS_ALIGNED(width, 2)) {
+      ABGRToUVJRow = ABGRToUVJRow_SVE2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_MSA) && defined(HAS_ABGRTOUVJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ABGRToYJRow = ABGRToYJRow_Any_MSA;
+    ABGRToUVJRow = ABGRToUVJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_MSA;
+    }
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToUVJRow = ABGRToUVJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ABGRToYJRow = ABGRToYJRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ABGRToYJRow = ABGRToYJRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToYJRow = ABGRToYJRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ABGRToYJRow = ABGRToYJRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ABGRToUVJRow(src_abgr, 0, dst_uj, dst_vj, width);
+    ABGRToYJRow(src_abgr, dst_yj, width);
+    src_abgr += src_stride_abgr;
+    dst_yj += dst_stride_yj;
+    dst_uj += dst_stride_uj;
+    dst_vj += dst_stride_vj;
+  }
+  return 0;
+}
+
+// Convert ABGR to J400.
+LIBYUV_API
+int ABGRToJ400(const uint8_t* src_abgr,
+               int src_stride_abgr,
+               uint8_t* dst_yj,
+               int dst_stride_yj,
+               int width,
+               int height) {
+  int y;
+  void (*ABGRToYJRow)(const uint8_t* src_abgr, uint8_t* dst_yj, int width) =
+      ABGRToYJRow_C;
+  if (!src_abgr || !dst_yj || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_abgr = src_abgr + (height - 1) * src_stride_abgr;
+    src_stride_abgr = -src_stride_abgr;
+  }
+  // Coalesce rows.
+  if (src_stride_abgr == width * 4 && dst_stride_yj == width) {
+    width *= height;
+    height = 1;
+    src_stride_abgr = dst_stride_yj = 0;
+  }
+#if defined(HAS_ABGRTOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ABGRToYJRow = ABGRToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ABGRToYJRow = ABGRToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToYJRow = ABGRToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ABGRToYJRow = ABGRToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ABGRToYJRow = ABGRToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_NEON_DotProd;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ABGRToYJRow = ABGRToYJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ABGRToYJRow = ABGRToYJRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ABGRToYJRow = ABGRToYJRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ABGRToYJRow = ABGRToYJRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ABGRTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ABGRToYJRow = ABGRToYJRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ABGRToYJRow(src_abgr, dst_yj, width);
+    src_abgr += src_stride_abgr;
+    dst_yj += dst_stride_yj;
+  }
+  return 0;
+}
+
+// Convert ARGB to AR64.
+LIBYUV_API
+int ARGBToAR64(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint16_t* dst_ar64,
+               int dst_stride_ar64,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToAR64Row)(const uint8_t* src_argb, uint16_t* dst_ar64,
+                        int width) = ARGBToAR64Row_C;
+  if (!src_argb || !dst_ar64 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_ar64 == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_ar64 = 0;
+  }
+#if defined(HAS_ARGBTOAR64ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToAR64Row = ARGBToAR64Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBToAR64Row = ARGBToAR64Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOAR64ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToAR64Row = ARGBToAR64Row_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToAR64Row = ARGBToAR64Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOAR64ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToAR64Row = ARGBToAR64Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToAR64Row = ARGBToAR64Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOAR64ROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToAR64Row = ARGBToAR64Row_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToAR64Row(src_argb, dst_ar64, width);
+    src_argb += src_stride_argb;
+    dst_ar64 += dst_stride_ar64;
+  }
+  return 0;
+}
+
+// Convert ARGB to AB64.
+LIBYUV_API
+int ARGBToAB64(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint16_t* dst_ab64,
+               int dst_stride_ab64,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBToAB64Row)(const uint8_t* src_argb, uint16_t* dst_ar64,
+                        int width) = ARGBToAB64Row_C;
+  if (!src_argb || !dst_ab64 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_ab64 == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_ab64 = 0;
+  }
+#if defined(HAS_ARGBTOAB64ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToAB64Row = ARGBToAB64Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBToAB64Row = ARGBToAB64Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOAB64ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToAB64Row = ARGBToAB64Row_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToAB64Row = ARGBToAB64Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOAB64ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToAB64Row = ARGBToAB64Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBToAB64Row = ARGBToAB64Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOAB64ROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToAB64Row = ARGBToAB64Row_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBToAB64Row(src_argb, dst_ab64, width);
+    src_argb += src_stride_argb;
+    dst_ab64 += dst_stride_ab64;
+  }
+  return 0;
+}
+
+// Enabled if 1 pass is available
+#if defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_MSA) || \
+    defined(HAS_RAWTOYJROW_RVV)
+#define HAS_RAWTOYJROW
+#endif
+
+// RAW to JNV21 full range NV21
+LIBYUV_API
+int RAWToJNV21(const uint8_t* src_raw,
+               int src_stride_raw,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int width,
+               int height) {
+  int y;
+  int halfwidth = (width + 1) >> 1;
+#if defined(HAS_RAWTOYJROW)
+  void (*RAWToUVJRow)(const uint8_t* src_raw, int src_stride_raw,
+                      uint8_t* dst_uj, uint8_t* dst_vj, int width) =
+      RAWToUVJRow_C;
+  void (*RAWToYJRow)(const uint8_t* src_raw, uint8_t* dst_y, int width) =
+      RAWToYJRow_C;
+#else
+  void (*RAWToARGBRow)(const uint8_t* src_rgb, uint8_t* dst_argb, int width) =
+      RAWToARGBRow_C;
+  void (*ARGBToUVJRow)(const uint8_t* src_argb0, int src_stride_argb,
+                       uint8_t* dst_uj, uint8_t* dst_vj, int width) =
+      ARGBToUVJRow_C;
+  void (*ARGBToYJRow)(const uint8_t* src_argb, uint8_t* dst_y, int width) =
+      ARGBToYJRow_C;
+#endif
+  void (*MergeUVRow_)(const uint8_t* src_uj, const uint8_t* src_vj,
+                      uint8_t* dst_vu, int width) = MergeUVRow_C;
+  if (!src_raw || !dst_y || !dst_vu || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_raw = src_raw + (height - 1) * src_stride_raw;
+    src_stride_raw = -src_stride_raw;
+  }
+
+#if defined(HAS_RAWTOYJROW)
+
+// Neon version does direct RAW to YUV.
+#if defined(HAS_RAWTOYJROW_NEON) && defined(HAS_RAWTOUVJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RAWToUVJRow = RAWToUVJRow_Any_NEON;
+    RAWToYJRow = RAWToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYJRow = RAWToYJRow_NEON;
+      RAWToUVJRow = RAWToUVJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_MSA) && defined(HAS_RAWTOUVJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RAWToUVJRow = RAWToUVJRow_Any_MSA;
+    RAWToYJRow = RAWToYJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYJRow = RAWToYJRow_MSA;
+      RAWToUVJRow = RAWToUVJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RAWToYJRow = RAWToYJRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToYJRow = RAWToYJRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    RAWToYJRow = RAWToYJRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      RAWToYJRow = RAWToYJRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_RAWTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RAWToYJRow = RAWToYJRow_RVV;
+  }
+#endif
+
+// Other platforms do intermediate conversion from RAW to ARGB.
+#else  // HAS_RAWTOYJROW
+
+#if defined(HAS_RAWTOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RAWToARGBRow = RAWToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToARGBRow = RAWToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYJRow = ARGBToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYJRow = ARGBToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYJRow = ARGBToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToUVJRow = ARGBToUVJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOUVJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToUVJRow = ARGBToUVJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToUVJRow = ARGBToUVJRow_AVX2;
+    }
+  }
+#endif
+#endif  // HAS_RAWTOYJROW
+#if defined(HAS_MERGEUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    MergeUVRow_ = MergeUVRow_Any_SSE2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeUVRow_ = MergeUVRow_Any_AVX2;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW)) {
+    MergeUVRow_ = MergeUVRow_Any_AVX512BW;
+    if (IS_ALIGNED(halfwidth, 64)) {
+      MergeUVRow_ = MergeUVRow_AVX512BW;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeUVRow_ = MergeUVRow_Any_NEON;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    MergeUVRow_ = MergeUVRow_Any_MSA;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    MergeUVRow_ = MergeUVRow_Any_LSX;
+    if (IS_ALIGNED(halfwidth, 16)) {
+      MergeUVRow_ = MergeUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    MergeUVRow_ = MergeUVRow_RVV;
+  }
+#endif
+  {
+#if defined(HAS_RAWTOYJROW)
+    // Allocate a row of uv.
+    const int row_uv_size = ((halfwidth + 31) & ~31);
+    align_buffer_64(row_uj, row_uv_size * 2);
+    uint8_t* row_vj = row_uj + row_uv_size;
+#else
+    // Allocate row of uv and 2 rows of ARGB.
+    const int row_size = ((width * 4 + 31) & ~31);
+    const int row_uv_size = ((halfwidth + 31) & ~31);
+    align_buffer_64(row_uj, row_uv_size * 2 + row_size * 2);
+    uint8_t* row_vj = row_uj + row_uv_size;
+    uint8_t* row = row_vj + row_uv_size;
+#endif
+    if (!row_uj)
+      return 1;
+
+    for (y = 0; y < height - 1; y += 2) {
+#if defined(HAS_RAWTOYJROW)
+      RAWToUVJRow(src_raw, src_stride_raw, row_uj, row_vj, width);
+      MergeUVRow_(row_vj, row_uj, dst_vu, halfwidth);
+      RAWToYJRow(src_raw, dst_y, width);
+      RAWToYJRow(src_raw + src_stride_raw, dst_y + dst_stride_y, width);
+#else
+      RAWToARGBRow(src_raw, row, width);
+      RAWToARGBRow(src_raw + src_stride_raw, row + row_size, width);
+      ARGBToUVJRow(row, row_size, row_uj, row_vj, width);
+      MergeUVRow_(row_vj, row_uj, dst_vu, halfwidth);
+      ARGBToYJRow(row, dst_y, width);
+      ARGBToYJRow(row + row_size, dst_y + dst_stride_y, width);
+#endif
+      src_raw += src_stride_raw * 2;
+      dst_y += dst_stride_y * 2;
+      dst_vu += dst_stride_vu;
+    }
+    if (height & 1) {
+#if defined(HAS_RAWTOYJROW)
+      RAWToUVJRow(src_raw, 0, row_uj, row_vj, width);
+      MergeUVRow_(row_vj, row_uj, dst_vu, halfwidth);
+      RAWToYJRow(src_raw, dst_y, width);
+#else
+      RAWToARGBRow(src_raw, row, width);
+      ARGBToUVJRow(row, 0, row_uj, row_vj, width);
+      MergeUVRow_(row_vj, row_uj, dst_vu, halfwidth);
+      ARGBToYJRow(row, dst_y, width);
+#endif
+    }
+    free_aligned_buffer_64(row_uj);
+  }
+  return 0;
+}
+#undef HAS_RAWTOYJROW
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/convert_jpeg.cc

@@ -0,0 +1,602 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/convert.h"
+#include "libyuv/convert_argb.h"
+
+#ifdef HAVE_JPEG
+#include "libyuv/mjpeg_decoder.h"
+#endif
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+#ifdef HAVE_JPEG
+struct I420Buffers {
+  uint8_t* y;
+  int y_stride;
+  uint8_t* u;
+  int u_stride;
+  uint8_t* v;
+  int v_stride;
+  int w;
+  int h;
+};
+
+static void JpegCopyI420(void* opaque,
+                         const uint8_t* const* data,
+                         const int* strides,
+                         int rows) {
+  I420Buffers* dest = (I420Buffers*)(opaque);
+  I420Copy(data[0], strides[0], data[1], strides[1], data[2], strides[2],
+           dest->y, dest->y_stride, dest->u, dest->u_stride, dest->v,
+           dest->v_stride, dest->w, rows);
+  dest->y += rows * dest->y_stride;
+  dest->u += ((rows + 1) >> 1) * dest->u_stride;
+  dest->v += ((rows + 1) >> 1) * dest->v_stride;
+  dest->h -= rows;
+}
+
+static void JpegI422ToI420(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  I420Buffers* dest = (I420Buffers*)(opaque);
+  I422ToI420(data[0], strides[0], data[1], strides[1], data[2], strides[2],
+             dest->y, dest->y_stride, dest->u, dest->u_stride, dest->v,
+             dest->v_stride, dest->w, rows);
+  dest->y += rows * dest->y_stride;
+  dest->u += ((rows + 1) >> 1) * dest->u_stride;
+  dest->v += ((rows + 1) >> 1) * dest->v_stride;
+  dest->h -= rows;
+}
+
+static void JpegI444ToI420(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  I420Buffers* dest = (I420Buffers*)(opaque);
+  I444ToI420(data[0], strides[0], data[1], strides[1], data[2], strides[2],
+             dest->y, dest->y_stride, dest->u, dest->u_stride, dest->v,
+             dest->v_stride, dest->w, rows);
+  dest->y += rows * dest->y_stride;
+  dest->u += ((rows + 1) >> 1) * dest->u_stride;
+  dest->v += ((rows + 1) >> 1) * dest->v_stride;
+  dest->h -= rows;
+}
+
+static void JpegI400ToI420(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  I420Buffers* dest = (I420Buffers*)(opaque);
+  I400ToI420(data[0], strides[0], dest->y, dest->y_stride, dest->u,
+             dest->u_stride, dest->v, dest->v_stride, dest->w, rows);
+  dest->y += rows * dest->y_stride;
+  dest->u += ((rows + 1) >> 1) * dest->u_stride;
+  dest->v += ((rows + 1) >> 1) * dest->v_stride;
+  dest->h -= rows;
+}
+
+// Query size of MJPG in pixels.
+LIBYUV_API
+int MJPGSize(const uint8_t* src_mjpg,
+             size_t src_size_mjpg,
+             int* width,
+             int* height) {
+  MJpegDecoder mjpeg_decoder;
+  LIBYUV_BOOL ret = mjpeg_decoder.LoadFrame(src_mjpg, src_size_mjpg);
+  if (ret) {
+    *width = mjpeg_decoder.GetWidth();
+    *height = mjpeg_decoder.GetHeight();
+  }
+  mjpeg_decoder.UnloadFrame();
+  return ret ? 0 : -1;  // -1 for runtime failure.
+}
+
+// MJPG (Motion JPeg) to I420
+// TODO(fbarchard): review src_width and src_height requirement. dst_width and
+// dst_height may be enough.
+LIBYUV_API
+int MJPGToI420(const uint8_t* src_mjpg,
+               size_t src_size_mjpg,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int src_width,
+               int src_height,
+               int dst_width,
+               int dst_height) {
+  if (src_size_mjpg == kUnknownDataSize) {
+    // ERROR: MJPEG frame size unknown
+    return -1;
+  }
+
+  // TODO(fbarchard): Port MJpeg to C.
+  MJpegDecoder mjpeg_decoder;
+  LIBYUV_BOOL ret = mjpeg_decoder.LoadFrame(src_mjpg, src_size_mjpg);
+  if (ret && (mjpeg_decoder.GetWidth() != src_width ||
+              mjpeg_decoder.GetHeight() != src_height)) {
+    // ERROR: MJPEG frame has unexpected dimensions
+    mjpeg_decoder.UnloadFrame();
+    return 1;  // runtime failure
+  }
+  if (ret) {
+    I420Buffers bufs = {dst_y, dst_stride_y, dst_u,     dst_stride_u,
+                        dst_v, dst_stride_v, dst_width, dst_height};
+    // YUV420
+    if (mjpeg_decoder.GetColorSpace() == MJpegDecoder::kColorSpaceYCbCr &&
+        mjpeg_decoder.GetNumComponents() == 3 &&
+        mjpeg_decoder.GetVertSampFactor(0) == 2 &&
+        mjpeg_decoder.GetHorizSampFactor(0) == 2 &&
+        mjpeg_decoder.GetVertSampFactor(1) == 1 &&
+        mjpeg_decoder.GetHorizSampFactor(1) == 1 &&
+        mjpeg_decoder.GetVertSampFactor(2) == 1 &&
+        mjpeg_decoder.GetHorizSampFactor(2) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegCopyI420, &bufs, dst_width,
+                                           dst_height);
+      // YUV422
+    } else if (mjpeg_decoder.GetColorSpace() ==
+                   MJpegDecoder::kColorSpaceYCbCr &&
+               mjpeg_decoder.GetNumComponents() == 3 &&
+               mjpeg_decoder.GetVertSampFactor(0) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(0) == 2 &&
+               mjpeg_decoder.GetVertSampFactor(1) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(1) == 1 &&
+               mjpeg_decoder.GetVertSampFactor(2) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(2) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI422ToI420, &bufs, dst_width,
+                                           dst_height);
+      // YUV444
+    } else if (mjpeg_decoder.GetColorSpace() ==
+                   MJpegDecoder::kColorSpaceYCbCr &&
+               mjpeg_decoder.GetNumComponents() == 3 &&
+               mjpeg_decoder.GetVertSampFactor(0) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(0) == 1 &&
+               mjpeg_decoder.GetVertSampFactor(1) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(1) == 1 &&
+               mjpeg_decoder.GetVertSampFactor(2) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(2) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI444ToI420, &bufs, dst_width,
+                                           dst_height);
+      // YUV400
+    } else if (mjpeg_decoder.GetColorSpace() ==
+                   MJpegDecoder::kColorSpaceGrayscale &&
+               mjpeg_decoder.GetNumComponents() == 1 &&
+               mjpeg_decoder.GetVertSampFactor(0) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(0) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI400ToI420, &bufs, dst_width,
+                                           dst_height);
+    } else {
+      // TODO(fbarchard): Implement conversion for any other
+      // colorspace/subsample factors that occur in practice. ERROR: Unable to
+      // convert MJPEG frame because format is not supported
+      mjpeg_decoder.UnloadFrame();
+      return 1;
+    }
+  }
+  return ret ? 0 : 1;
+}
+
+struct NV21Buffers {
+  uint8_t* y;
+  int y_stride;
+  uint8_t* vu;
+  int vu_stride;
+  int w;
+  int h;
+};
+
+static void JpegI420ToNV21(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  NV21Buffers* dest = (NV21Buffers*)(opaque);
+  I420ToNV21(data[0], strides[0], data[1], strides[1], data[2], strides[2],
+             dest->y, dest->y_stride, dest->vu, dest->vu_stride, dest->w, rows);
+  dest->y += rows * dest->y_stride;
+  dest->vu += ((rows + 1) >> 1) * dest->vu_stride;
+  dest->h -= rows;
+}
+
+static void JpegI422ToNV21(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  NV21Buffers* dest = (NV21Buffers*)(opaque);
+  I422ToNV21(data[0], strides[0], data[1], strides[1], data[2], strides[2],
+             dest->y, dest->y_stride, dest->vu, dest->vu_stride, dest->w, rows);
+  dest->y += rows * dest->y_stride;
+  dest->vu += ((rows + 1) >> 1) * dest->vu_stride;
+  dest->h -= rows;
+}
+
+static void JpegI444ToNV21(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  NV21Buffers* dest = (NV21Buffers*)(opaque);
+  I444ToNV21(data[0], strides[0], data[1], strides[1], data[2], strides[2],
+             dest->y, dest->y_stride, dest->vu, dest->vu_stride, dest->w, rows);
+  dest->y += rows * dest->y_stride;
+  dest->vu += ((rows + 1) >> 1) * dest->vu_stride;
+  dest->h -= rows;
+}
+
+static void JpegI400ToNV21(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  NV21Buffers* dest = (NV21Buffers*)(opaque);
+  I400ToNV21(data[0], strides[0], dest->y, dest->y_stride, dest->vu,
+             dest->vu_stride, dest->w, rows);
+  dest->y += rows * dest->y_stride;
+  dest->vu += ((rows + 1) >> 1) * dest->vu_stride;
+  dest->h -= rows;
+}
+
+// MJPG (Motion JPeg) to NV21
+LIBYUV_API
+int MJPGToNV21(const uint8_t* src_mjpg,
+               size_t src_size_mjpg,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_vu,
+               int dst_stride_vu,
+               int src_width,
+               int src_height,
+               int dst_width,
+               int dst_height) {
+  if (src_size_mjpg == kUnknownDataSize) {
+    // ERROR: MJPEG frame size unknown
+    return -1;
+  }
+
+  // TODO(fbarchard): Port MJpeg to C.
+  MJpegDecoder mjpeg_decoder;
+  LIBYUV_BOOL ret = mjpeg_decoder.LoadFrame(src_mjpg, src_size_mjpg);
+  if (ret && (mjpeg_decoder.GetWidth() != src_width ||
+              mjpeg_decoder.GetHeight() != src_height)) {
+    // ERROR: MJPEG frame has unexpected dimensions
+    mjpeg_decoder.UnloadFrame();
+    return 1;  // runtime failure
+  }
+  if (ret) {
+    NV21Buffers bufs = {dst_y,         dst_stride_y, dst_vu,
+                        dst_stride_vu, dst_width,    dst_height};
+    // YUV420
+    if (mjpeg_decoder.GetColorSpace() == MJpegDecoder::kColorSpaceYCbCr &&
+        mjpeg_decoder.GetNumComponents() == 3 &&
+        mjpeg_decoder.GetVertSampFactor(0) == 2 &&
+        mjpeg_decoder.GetHorizSampFactor(0) == 2 &&
+        mjpeg_decoder.GetVertSampFactor(1) == 1 &&
+        mjpeg_decoder.GetHorizSampFactor(1) == 1 &&
+        mjpeg_decoder.GetVertSampFactor(2) == 1 &&
+        mjpeg_decoder.GetHorizSampFactor(2) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI420ToNV21, &bufs, dst_width,
+                                           dst_height);
+      // YUV422
+    } else if (mjpeg_decoder.GetColorSpace() ==
+                   MJpegDecoder::kColorSpaceYCbCr &&
+               mjpeg_decoder.GetNumComponents() == 3 &&
+               mjpeg_decoder.GetVertSampFactor(0) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(0) == 2 &&
+               mjpeg_decoder.GetVertSampFactor(1) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(1) == 1 &&
+               mjpeg_decoder.GetVertSampFactor(2) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(2) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI422ToNV21, &bufs, dst_width,
+                                           dst_height);
+      // YUV444
+    } else if (mjpeg_decoder.GetColorSpace() ==
+                   MJpegDecoder::kColorSpaceYCbCr &&
+               mjpeg_decoder.GetNumComponents() == 3 &&
+               mjpeg_decoder.GetVertSampFactor(0) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(0) == 1 &&
+               mjpeg_decoder.GetVertSampFactor(1) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(1) == 1 &&
+               mjpeg_decoder.GetVertSampFactor(2) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(2) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI444ToNV21, &bufs, dst_width,
+                                           dst_height);
+      // YUV400
+    } else if (mjpeg_decoder.GetColorSpace() ==
+                   MJpegDecoder::kColorSpaceGrayscale &&
+               mjpeg_decoder.GetNumComponents() == 1 &&
+               mjpeg_decoder.GetVertSampFactor(0) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(0) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI400ToNV21, &bufs, dst_width,
+                                           dst_height);
+    } else {
+      // Unknown colorspace.
+      mjpeg_decoder.UnloadFrame();
+      return 1;
+    }
+  }
+  return ret ? 0 : 1;
+}
+
+static void JpegI420ToNV12(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  NV21Buffers* dest = (NV21Buffers*)(opaque);
+  // Use NV21 with VU swapped.
+  I420ToNV21(data[0], strides[0], data[2], strides[2], data[1], strides[1],
+             dest->y, dest->y_stride, dest->vu, dest->vu_stride, dest->w, rows);
+  dest->y += rows * dest->y_stride;
+  dest->vu += ((rows + 1) >> 1) * dest->vu_stride;
+  dest->h -= rows;
+}
+
+static void JpegI422ToNV12(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  NV21Buffers* dest = (NV21Buffers*)(opaque);
+  // Use NV21 with VU swapped.
+  I422ToNV21(data[0], strides[0], data[2], strides[2], data[1], strides[1],
+             dest->y, dest->y_stride, dest->vu, dest->vu_stride, dest->w, rows);
+  dest->y += rows * dest->y_stride;
+  dest->vu += ((rows + 1) >> 1) * dest->vu_stride;
+  dest->h -= rows;
+}
+
+static void JpegI444ToNV12(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  NV21Buffers* dest = (NV21Buffers*)(opaque);
+  // Use NV21 with VU swapped.
+  I444ToNV21(data[0], strides[0], data[2], strides[2], data[1], strides[1],
+             dest->y, dest->y_stride, dest->vu, dest->vu_stride, dest->w, rows);
+  dest->y += rows * dest->y_stride;
+  dest->vu += ((rows + 1) >> 1) * dest->vu_stride;
+  dest->h -= rows;
+}
+
+static void JpegI400ToNV12(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  NV21Buffers* dest = (NV21Buffers*)(opaque);
+  // Use NV21 since there is no UV plane.
+  I400ToNV21(data[0], strides[0], dest->y, dest->y_stride, dest->vu,
+             dest->vu_stride, dest->w, rows);
+  dest->y += rows * dest->y_stride;
+  dest->vu += ((rows + 1) >> 1) * dest->vu_stride;
+  dest->h -= rows;
+}
+
+// MJPG (Motion JPEG) to NV12.
+LIBYUV_API
+int MJPGToNV12(const uint8_t* sample,
+               size_t sample_size,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int src_width,
+               int src_height,
+               int dst_width,
+               int dst_height) {
+  if (sample_size == kUnknownDataSize) {
+    // ERROR: MJPEG frame size unknown
+    return -1;
+  }
+
+  // TODO(fbarchard): Port MJpeg to C.
+  MJpegDecoder mjpeg_decoder;
+  LIBYUV_BOOL ret = mjpeg_decoder.LoadFrame(sample, sample_size);
+  if (ret && (mjpeg_decoder.GetWidth() != src_width ||
+              mjpeg_decoder.GetHeight() != src_height)) {
+    // ERROR: MJPEG frame has unexpected dimensions
+    mjpeg_decoder.UnloadFrame();
+    return 1;  // runtime failure
+  }
+  if (ret) {
+    // Use NV21Buffers but with UV instead of VU.
+    NV21Buffers bufs = {dst_y,         dst_stride_y, dst_uv,
+                        dst_stride_uv, dst_width,    dst_height};
+    // YUV420
+    if (mjpeg_decoder.GetColorSpace() == MJpegDecoder::kColorSpaceYCbCr &&
+        mjpeg_decoder.GetNumComponents() == 3 &&
+        mjpeg_decoder.GetVertSampFactor(0) == 2 &&
+        mjpeg_decoder.GetHorizSampFactor(0) == 2 &&
+        mjpeg_decoder.GetVertSampFactor(1) == 1 &&
+        mjpeg_decoder.GetHorizSampFactor(1) == 1 &&
+        mjpeg_decoder.GetVertSampFactor(2) == 1 &&
+        mjpeg_decoder.GetHorizSampFactor(2) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI420ToNV12, &bufs, dst_width,
+                                           dst_height);
+      // YUV422
+    } else if (mjpeg_decoder.GetColorSpace() ==
+                   MJpegDecoder::kColorSpaceYCbCr &&
+               mjpeg_decoder.GetNumComponents() == 3 &&
+               mjpeg_decoder.GetVertSampFactor(0) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(0) == 2 &&
+               mjpeg_decoder.GetVertSampFactor(1) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(1) == 1 &&
+               mjpeg_decoder.GetVertSampFactor(2) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(2) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI422ToNV12, &bufs, dst_width,
+                                           dst_height);
+      // YUV444
+    } else if (mjpeg_decoder.GetColorSpace() ==
+                   MJpegDecoder::kColorSpaceYCbCr &&
+               mjpeg_decoder.GetNumComponents() == 3 &&
+               mjpeg_decoder.GetVertSampFactor(0) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(0) == 1 &&
+               mjpeg_decoder.GetVertSampFactor(1) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(1) == 1 &&
+               mjpeg_decoder.GetVertSampFactor(2) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(2) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI444ToNV12, &bufs, dst_width,
+                                           dst_height);
+      // YUV400
+    } else if (mjpeg_decoder.GetColorSpace() ==
+                   MJpegDecoder::kColorSpaceGrayscale &&
+               mjpeg_decoder.GetNumComponents() == 1 &&
+               mjpeg_decoder.GetVertSampFactor(0) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(0) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI400ToNV12, &bufs, dst_width,
+                                           dst_height);
+    } else {
+      // Unknown colorspace.
+      mjpeg_decoder.UnloadFrame();
+      return 1;
+    }
+  }
+  return ret ? 0 : 1;
+}
+
+struct ARGBBuffers {
+  uint8_t* argb;
+  int argb_stride;
+  int w;
+  int h;
+};
+
+static void JpegI420ToARGB(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  ARGBBuffers* dest = (ARGBBuffers*)(opaque);
+  I420ToARGB(data[0], strides[0], data[1], strides[1], data[2], strides[2],
+             dest->argb, dest->argb_stride, dest->w, rows);
+  dest->argb += rows * dest->argb_stride;
+  dest->h -= rows;
+}
+
+static void JpegI422ToARGB(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  ARGBBuffers* dest = (ARGBBuffers*)(opaque);
+  I422ToARGB(data[0], strides[0], data[1], strides[1], data[2], strides[2],
+             dest->argb, dest->argb_stride, dest->w, rows);
+  dest->argb += rows * dest->argb_stride;
+  dest->h -= rows;
+}
+
+static void JpegI444ToARGB(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  ARGBBuffers* dest = (ARGBBuffers*)(opaque);
+  I444ToARGB(data[0], strides[0], data[1], strides[1], data[2], strides[2],
+             dest->argb, dest->argb_stride, dest->w, rows);
+  dest->argb += rows * dest->argb_stride;
+  dest->h -= rows;
+}
+
+static void JpegI400ToARGB(void* opaque,
+                           const uint8_t* const* data,
+                           const int* strides,
+                           int rows) {
+  ARGBBuffers* dest = (ARGBBuffers*)(opaque);
+  I400ToARGB(data[0], strides[0], dest->argb, dest->argb_stride, dest->w, rows);
+  dest->argb += rows * dest->argb_stride;
+  dest->h -= rows;
+}
+
+// MJPG (Motion JPeg) to ARGB
+// TODO(fbarchard): review src_width and src_height requirement. dst_width and
+// dst_height may be enough.
+LIBYUV_API
+int MJPGToARGB(const uint8_t* src_mjpg,
+               size_t src_size_mjpg,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int src_width,
+               int src_height,
+               int dst_width,
+               int dst_height) {
+  if (src_size_mjpg == kUnknownDataSize) {
+    // ERROR: MJPEG frame size unknown
+    return -1;
+  }
+
+  // TODO(fbarchard): Port MJpeg to C.
+  MJpegDecoder mjpeg_decoder;
+  LIBYUV_BOOL ret = mjpeg_decoder.LoadFrame(src_mjpg, src_size_mjpg);
+  if (ret && (mjpeg_decoder.GetWidth() != src_width ||
+              mjpeg_decoder.GetHeight() != src_height)) {
+    // ERROR: MJPEG frame has unexpected dimensions
+    mjpeg_decoder.UnloadFrame();
+    return 1;  // runtime failure
+  }
+  if (ret) {
+    ARGBBuffers bufs = {dst_argb, dst_stride_argb, dst_width, dst_height};
+    // YUV420
+    if (mjpeg_decoder.GetColorSpace() == MJpegDecoder::kColorSpaceYCbCr &&
+        mjpeg_decoder.GetNumComponents() == 3 &&
+        mjpeg_decoder.GetVertSampFactor(0) == 2 &&
+        mjpeg_decoder.GetHorizSampFactor(0) == 2 &&
+        mjpeg_decoder.GetVertSampFactor(1) == 1 &&
+        mjpeg_decoder.GetHorizSampFactor(1) == 1 &&
+        mjpeg_decoder.GetVertSampFactor(2) == 1 &&
+        mjpeg_decoder.GetHorizSampFactor(2) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI420ToARGB, &bufs, dst_width,
+                                           dst_height);
+      // YUV422
+    } else if (mjpeg_decoder.GetColorSpace() ==
+                   MJpegDecoder::kColorSpaceYCbCr &&
+               mjpeg_decoder.GetNumComponents() == 3 &&
+               mjpeg_decoder.GetVertSampFactor(0) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(0) == 2 &&
+               mjpeg_decoder.GetVertSampFactor(1) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(1) == 1 &&
+               mjpeg_decoder.GetVertSampFactor(2) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(2) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI422ToARGB, &bufs, dst_width,
+                                           dst_height);
+      // YUV444
+    } else if (mjpeg_decoder.GetColorSpace() ==
+                   MJpegDecoder::kColorSpaceYCbCr &&
+               mjpeg_decoder.GetNumComponents() == 3 &&
+               mjpeg_decoder.GetVertSampFactor(0) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(0) == 1 &&
+               mjpeg_decoder.GetVertSampFactor(1) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(1) == 1 &&
+               mjpeg_decoder.GetVertSampFactor(2) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(2) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI444ToARGB, &bufs, dst_width,
+                                           dst_height);
+      // YUV400
+    } else if (mjpeg_decoder.GetColorSpace() ==
+                   MJpegDecoder::kColorSpaceGrayscale &&
+               mjpeg_decoder.GetNumComponents() == 1 &&
+               mjpeg_decoder.GetVertSampFactor(0) == 1 &&
+               mjpeg_decoder.GetHorizSampFactor(0) == 1) {
+      ret = mjpeg_decoder.DecodeToCallback(&JpegI400ToARGB, &bufs, dst_width,
+                                           dst_height);
+    } else {
+      // TODO(fbarchard): Implement conversion for any other
+      // colorspace/subsample factors that occur in practice. ERROR: Unable to
+      // convert MJPEG frame because format is not supported
+      mjpeg_decoder.UnloadFrame();
+      return 1;
+    }
+  }
+  return ret ? 0 : 1;
+}
+
+#endif  // HAVE_JPEG
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/convert_to_argb.cc

@@ -0,0 +1,382 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/convert_argb.h"
+
+#include "libyuv/cpu_id.h"
+#ifdef HAVE_JPEG
+#include "libyuv/mjpeg_decoder.h"
+#endif
+#include "libyuv/rotate_argb.h"
+#include "libyuv/row.h"
+#include "libyuv/video_common.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Convert camera sample to ARGB with cropping, rotation and vertical flip.
+// src_width is used for source stride computation
+// src_height is used to compute location of planes, and indicate inversion
+// sample_size is measured in bytes and is the size of the frame.
+//   With MJPEG it is the compressed size of the frame.
+
+// TODO(fbarchard): Add the following:
+// H010ToARGB
+// I010ToARGB
+
+LIBYUV_API
+int ConvertToARGB(const uint8_t* sample,
+                  size_t sample_size,
+                  uint8_t* dst_argb,
+                  int dst_stride_argb,
+                  int crop_x,
+                  int crop_y,
+                  int src_width,
+                  int src_height,
+                  int crop_width,
+                  int crop_height,
+                  enum RotationMode rotation,
+                  uint32_t fourcc) {
+  uint32_t format = CanonicalFourCC(fourcc);
+  int aligned_src_width = (src_width + 1) & ~1;
+  const uint8_t* src;
+  const uint8_t* src_uv;
+  int abs_src_height = (src_height < 0) ? -src_height : src_height;
+  int inv_crop_height = (crop_height < 0) ? -crop_height : crop_height;
+  int r = 0;
+
+  // One pass rotation is available for some formats. For the rest, convert
+  // to ARGB (with optional vertical flipping) into a temporary ARGB buffer,
+  // and then rotate the ARGB to the final destination buffer.
+  // For in-place conversion, if destination dst_argb is same as source sample,
+  // also enable temporary buffer.
+  LIBYUV_BOOL need_buf =
+      (rotation && format != FOURCC_ARGB) || dst_argb == sample;
+  uint8_t* dest_argb = dst_argb;
+  int dest_dst_stride_argb = dst_stride_argb;
+  uint8_t* rotate_buffer = NULL;
+  int abs_crop_height = (crop_height < 0) ? -crop_height : crop_height;
+
+  if (dst_argb == NULL || sample == NULL || src_width <= 0 || crop_width <= 0 ||
+      src_height == 0 || crop_height == 0) {
+    return -1;
+  }
+  if (src_height < 0) {
+    inv_crop_height = -inv_crop_height;
+  }
+
+  if (need_buf) {
+    int argb_size = crop_width * 4 * abs_crop_height;
+    rotate_buffer = (uint8_t*)malloc(argb_size); /* NOLINT */
+    if (!rotate_buffer) {
+      return 1;  // Out of memory runtime error.
+    }
+    dst_argb = rotate_buffer;
+    dst_stride_argb = crop_width * 4;
+  }
+
+  switch (format) {
+    // Single plane formats
+    case FOURCC_YUY2:
+      src = sample + (aligned_src_width * crop_y + crop_x) * 2;
+      r = YUY2ToARGB(src, aligned_src_width * 2, dst_argb, dst_stride_argb,
+                     crop_width, inv_crop_height);
+      break;
+    case FOURCC_UYVY:
+      src = sample + (aligned_src_width * crop_y + crop_x) * 2;
+      r = UYVYToARGB(src, aligned_src_width * 2, dst_argb, dst_stride_argb,
+                     crop_width, inv_crop_height);
+      break;
+    case FOURCC_24BG:
+      src = sample + (src_width * crop_y + crop_x) * 3;
+      r = RGB24ToARGB(src, src_width * 3, dst_argb, dst_stride_argb, crop_width,
+                      inv_crop_height);
+      break;
+    case FOURCC_RAW:
+      src = sample + (src_width * crop_y + crop_x) * 3;
+      r = RAWToARGB(src, src_width * 3, dst_argb, dst_stride_argb, crop_width,
+                    inv_crop_height);
+      break;
+    case FOURCC_ARGB:
+      if (!need_buf && !rotation) {
+        src = sample + (src_width * crop_y + crop_x) * 4;
+        r = ARGBToARGB(src, src_width * 4, dst_argb, dst_stride_argb,
+                       crop_width, inv_crop_height);
+      }
+      break;
+    case FOURCC_BGRA:
+      src = sample + (src_width * crop_y + crop_x) * 4;
+      r = BGRAToARGB(src, src_width * 4, dst_argb, dst_stride_argb, crop_width,
+                     inv_crop_height);
+      break;
+    case FOURCC_ABGR:
+      src = sample + (src_width * crop_y + crop_x) * 4;
+      r = ABGRToARGB(src, src_width * 4, dst_argb, dst_stride_argb, crop_width,
+                     inv_crop_height);
+      break;
+    case FOURCC_RGBA:
+      src = sample + (src_width * crop_y + crop_x) * 4;
+      r = RGBAToARGB(src, src_width * 4, dst_argb, dst_stride_argb, crop_width,
+                     inv_crop_height);
+      break;
+    case FOURCC_AR30:
+      src = sample + (src_width * crop_y + crop_x) * 4;
+      r = AR30ToARGB(src, src_width * 4, dst_argb, dst_stride_argb, crop_width,
+                     inv_crop_height);
+      break;
+    case FOURCC_AB30:
+      src = sample + (src_width * crop_y + crop_x) * 4;
+      r = AB30ToARGB(src, src_width * 4, dst_argb, dst_stride_argb, crop_width,
+                     inv_crop_height);
+      break;
+    case FOURCC_RGBP:
+      src = sample + (src_width * crop_y + crop_x) * 2;
+      r = RGB565ToARGB(src, src_width * 2, dst_argb, dst_stride_argb,
+                       crop_width, inv_crop_height);
+      break;
+    case FOURCC_RGBO:
+      src = sample + (src_width * crop_y + crop_x) * 2;
+      r = ARGB1555ToARGB(src, src_width * 2, dst_argb, dst_stride_argb,
+                         crop_width, inv_crop_height);
+      break;
+    case FOURCC_R444:
+      src = sample + (src_width * crop_y + crop_x) * 2;
+      r = ARGB4444ToARGB(src, src_width * 2, dst_argb, dst_stride_argb,
+                         crop_width, inv_crop_height);
+      break;
+    case FOURCC_I400:
+      src = sample + src_width * crop_y + crop_x;
+      r = I400ToARGB(src, src_width, dst_argb, dst_stride_argb, crop_width,
+                     inv_crop_height);
+      break;
+    case FOURCC_J400:
+      src = sample + src_width * crop_y + crop_x;
+      r = J400ToARGB(src, src_width, dst_argb, dst_stride_argb, crop_width,
+                     inv_crop_height);
+      break;
+
+    // Biplanar formats
+    case FOURCC_NV12:
+      src = sample + (src_width * crop_y + crop_x);
+      src_uv =
+          sample + aligned_src_width * (abs_src_height + crop_y / 2) + crop_x;
+      r = NV12ToARGB(src, src_width, src_uv, aligned_src_width, dst_argb,
+                     dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    case FOURCC_NV21:
+      src = sample + (src_width * crop_y + crop_x);
+      src_uv =
+          sample + aligned_src_width * (abs_src_height + crop_y / 2) + crop_x;
+      // Call NV12 but with u and v parameters swapped.
+      r = NV21ToARGB(src, src_width, src_uv, aligned_src_width, dst_argb,
+                     dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    // Triplanar formats
+    case FOURCC_I420:
+    case FOURCC_YV12: {
+      const uint8_t* src_y = sample + (src_width * crop_y + crop_x);
+      const uint8_t* src_u;
+      const uint8_t* src_v;
+      int halfwidth = (src_width + 1) / 2;
+      int halfheight = (abs_src_height + 1) / 2;
+      if (format == FOURCC_YV12) {
+        src_v = sample + src_width * abs_src_height +
+                (halfwidth * crop_y + crop_x) / 2;
+        src_u = sample + src_width * abs_src_height +
+                halfwidth * (halfheight + crop_y / 2) + crop_x / 2;
+      } else {
+        src_u = sample + src_width * abs_src_height +
+                (halfwidth * crop_y + crop_x) / 2;
+        src_v = sample + src_width * abs_src_height +
+                halfwidth * (halfheight + crop_y / 2) + crop_x / 2;
+      }
+      r = I420ToARGB(src_y, src_width, src_u, halfwidth, src_v, halfwidth,
+                     dst_argb, dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    }
+
+    case FOURCC_J420: {
+      int halfwidth = (src_width + 1) / 2;
+      int halfheight = (abs_src_height + 1) / 2;
+      const uint8_t* src_y = sample + (src_width * crop_y + crop_x);
+      const uint8_t* src_u = sample + src_width * abs_src_height +
+                             (halfwidth * crop_y + crop_x) / 2;
+      const uint8_t* src_v = sample + src_width * abs_src_height +
+                             halfwidth * (halfheight + crop_y / 2) + crop_x / 2;
+      r = J420ToARGB(src_y, src_width, src_u, halfwidth, src_v, halfwidth,
+                     dst_argb, dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    }
+
+    case FOURCC_H420: {
+      int halfwidth = (src_width + 1) / 2;
+      int halfheight = (abs_src_height + 1) / 2;
+      const uint8_t* src_y = sample + (src_width * crop_y + crop_x);
+      const uint8_t* src_u = sample + src_width * abs_src_height +
+                             (halfwidth * crop_y + crop_x) / 2;
+      const uint8_t* src_v = sample + src_width * abs_src_height +
+                             halfwidth * (halfheight + crop_y / 2) + crop_x / 2;
+      r = H420ToARGB(src_y, src_width, src_u, halfwidth, src_v, halfwidth,
+                     dst_argb, dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    }
+
+    case FOURCC_U420: {
+      int halfwidth = (src_width + 1) / 2;
+      int halfheight = (abs_src_height + 1) / 2;
+      const uint8_t* src_y = sample + (src_width * crop_y + crop_x);
+      const uint8_t* src_u = sample + src_width * abs_src_height +
+                             (halfwidth * crop_y + crop_x) / 2;
+      const uint8_t* src_v = sample + src_width * abs_src_height +
+                             halfwidth * (halfheight + crop_y / 2) + crop_x / 2;
+      r = U420ToARGB(src_y, src_width, src_u, halfwidth, src_v, halfwidth,
+                     dst_argb, dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    }
+
+    case FOURCC_I422:
+    case FOURCC_YV16: {
+      int halfwidth = (src_width + 1) / 2;
+      const uint8_t* src_y = sample + src_width * crop_y + crop_x;
+      const uint8_t* src_u;
+      const uint8_t* src_v;
+      if (format == FOURCC_YV16) {
+        src_v = sample + src_width * abs_src_height + halfwidth * crop_y +
+                crop_x / 2;
+        src_u = sample + src_width * abs_src_height +
+                halfwidth * (abs_src_height + crop_y) + crop_x / 2;
+      } else {
+        src_u = sample + src_width * abs_src_height + halfwidth * crop_y +
+                crop_x / 2;
+        src_v = sample + src_width * abs_src_height +
+                halfwidth * (abs_src_height + crop_y) + crop_x / 2;
+      }
+      r = I422ToARGB(src_y, src_width, src_u, halfwidth, src_v, halfwidth,
+                     dst_argb, dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    }
+
+    case FOURCC_J422: {
+      int halfwidth = (src_width + 1) / 2;
+      const uint8_t* src_y = sample + src_width * crop_y + crop_x;
+      const uint8_t* src_u =
+          sample + src_width * abs_src_height + halfwidth * crop_y + crop_x / 2;
+      const uint8_t* src_v = sample + src_width * abs_src_height +
+                             halfwidth * (abs_src_height + crop_y) + crop_x / 2;
+      r = J422ToARGB(src_y, src_width, src_u, halfwidth, src_v, halfwidth,
+                     dst_argb, dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    }
+
+    case FOURCC_H422: {
+      int halfwidth = (src_width + 1) / 2;
+      const uint8_t* src_y = sample + src_width * crop_y + crop_x;
+      const uint8_t* src_u =
+          sample + src_width * abs_src_height + halfwidth * crop_y + crop_x / 2;
+      const uint8_t* src_v = sample + src_width * abs_src_height +
+                             halfwidth * (abs_src_height + crop_y) + crop_x / 2;
+      r = H422ToARGB(src_y, src_width, src_u, halfwidth, src_v, halfwidth,
+                     dst_argb, dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    }
+
+    case FOURCC_U422: {
+      int halfwidth = (src_width + 1) / 2;
+      const uint8_t* src_y = sample + src_width * crop_y + crop_x;
+      const uint8_t* src_u =
+          sample + src_width * abs_src_height + halfwidth * crop_y + crop_x / 2;
+      const uint8_t* src_v = sample + src_width * abs_src_height +
+                             halfwidth * (abs_src_height + crop_y) + crop_x / 2;
+      r = H422ToARGB(src_y, src_width, src_u, halfwidth, src_v, halfwidth,
+                     dst_argb, dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    }
+
+    case FOURCC_I444:
+    case FOURCC_YV24: {
+      const uint8_t* src_y = sample + src_width * crop_y + crop_x;
+      const uint8_t* src_u;
+      const uint8_t* src_v;
+      if (format == FOURCC_YV24) {
+        src_v = sample + src_width * (abs_src_height + crop_y) + crop_x;
+        src_u = sample + src_width * (abs_src_height * 2 + crop_y) + crop_x;
+      } else {
+        src_u = sample + src_width * (abs_src_height + crop_y) + crop_x;
+        src_v = sample + src_width * (abs_src_height * 2 + crop_y) + crop_x;
+      }
+      r = I444ToARGB(src_y, src_width, src_u, src_width, src_v, src_width,
+                     dst_argb, dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    }
+
+    case FOURCC_J444: {
+      const uint8_t* src_y = sample + src_width * crop_y + crop_x;
+      const uint8_t* src_u;
+      const uint8_t* src_v;
+      src_u = sample + src_width * (abs_src_height + crop_y) + crop_x;
+      src_v = sample + src_width * (abs_src_height * 2 + crop_y) + crop_x;
+      r = J444ToARGB(src_y, src_width, src_u, src_width, src_v, src_width,
+                     dst_argb, dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    }
+
+    case FOURCC_H444: {
+      const uint8_t* src_y = sample + src_width * crop_y + crop_x;
+      const uint8_t* src_u;
+      const uint8_t* src_v;
+      src_u = sample + src_width * (abs_src_height + crop_y) + crop_x;
+      src_v = sample + src_width * (abs_src_height * 2 + crop_y) + crop_x;
+      r = H444ToARGB(src_y, src_width, src_u, src_width, src_v, src_width,
+                     dst_argb, dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    }
+
+    case FOURCC_U444: {
+      const uint8_t* src_y = sample + src_width * crop_y + crop_x;
+      const uint8_t* src_u;
+      const uint8_t* src_v;
+      src_u = sample + src_width * (abs_src_height + crop_y) + crop_x;
+      src_v = sample + src_width * (abs_src_height * 2 + crop_y) + crop_x;
+      r = U444ToARGB(src_y, src_width, src_u, src_width, src_v, src_width,
+                     dst_argb, dst_stride_argb, crop_width, inv_crop_height);
+      break;
+    }
+
+#ifdef HAVE_JPEG
+    case FOURCC_MJPG:
+      r = MJPGToARGB(sample, sample_size, dst_argb, dst_stride_argb, src_width,
+                     abs_src_height, crop_width, inv_crop_height);
+      break;
+#endif
+    default:
+      r = -1;  // unknown fourcc - return failure code.
+  }
+
+  if (need_buf) {
+    if (!r) {
+      r = ARGBRotate(dst_argb, dst_stride_argb, dest_argb, dest_dst_stride_argb,
+                     crop_width, abs_crop_height, rotation);
+    }
+    free(rotate_buffer);
+  } else if (rotation) {
+    src = sample + (src_width * crop_y + crop_x) * 4;
+    r = ARGBRotate(src, src_width * 4, dst_argb, dst_stride_argb, crop_width,
+                   inv_crop_height, rotation);
+  }
+
+  return r;
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/convert_to_i420.cc

@@ -0,0 +1,280 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include <stdlib.h>
+
+#include "libyuv/convert.h"
+
+#include "libyuv/video_common.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Convert camera sample to I420 with cropping, rotation and vertical flip.
+// src_width is used for source stride computation
+// src_height is used to compute location of planes, and indicate inversion
+// sample_size is measured in bytes and is the size of the frame.
+//   With MJPEG it is the compressed size of the frame.
+LIBYUV_API
+int ConvertToI420(const uint8_t* sample,
+                  size_t sample_size,
+                  uint8_t* dst_y,
+                  int dst_stride_y,
+                  uint8_t* dst_u,
+                  int dst_stride_u,
+                  uint8_t* dst_v,
+                  int dst_stride_v,
+                  int crop_x,
+                  int crop_y,
+                  int src_width,
+                  int src_height,
+                  int crop_width,
+                  int crop_height,
+                  enum RotationMode rotation,
+                  uint32_t fourcc) {
+  uint32_t format = CanonicalFourCC(fourcc);
+  int aligned_src_width = (src_width + 1) & ~1;
+  const uint8_t* src;
+  const uint8_t* src_uv;
+  const int abs_src_height = (src_height < 0) ? -src_height : src_height;
+  // TODO(nisse): Why allow crop_height < 0?
+  const int abs_crop_height = (crop_height < 0) ? -crop_height : crop_height;
+  int r = 0;
+  LIBYUV_BOOL need_buf =
+      (rotation && format != FOURCC_I420 && format != FOURCC_NV12 &&
+       format != FOURCC_NV21 && format != FOURCC_YV12) ||
+      dst_y == sample;
+  uint8_t* tmp_y = dst_y;
+  uint8_t* tmp_u = dst_u;
+  uint8_t* tmp_v = dst_v;
+  int tmp_y_stride = dst_stride_y;
+  int tmp_u_stride = dst_stride_u;
+  int tmp_v_stride = dst_stride_v;
+  uint8_t* rotate_buffer = NULL;
+  const int inv_crop_height =
+      (src_height < 0) ? -abs_crop_height : abs_crop_height;
+
+  if (!dst_y || !dst_u || !dst_v || !sample || src_width <= 0 ||
+      crop_width <= 0 || src_height == 0 || crop_height == 0) {
+    return -1;
+  }
+
+  // One pass rotation is available for some formats. For the rest, convert
+  // to I420 (with optional vertical flipping) into a temporary I420 buffer,
+  // and then rotate the I420 to the final destination buffer.
+  // For in-place conversion, if destination dst_y is same as source sample,
+  // also enable temporary buffer.
+  if (need_buf) {
+    int y_size = crop_width * abs_crop_height;
+    int uv_size = ((crop_width + 1) / 2) * ((abs_crop_height + 1) / 2);
+    rotate_buffer = (uint8_t*)malloc(y_size + uv_size * 2); /* NOLINT */
+    if (!rotate_buffer) {
+      return 1;  // Out of memory runtime error.
+    }
+    dst_y = rotate_buffer;
+    dst_u = dst_y + y_size;
+    dst_v = dst_u + uv_size;
+    dst_stride_y = crop_width;
+    dst_stride_u = dst_stride_v = ((crop_width + 1) / 2);
+  }
+
+  switch (format) {
+    // Single plane formats
+    case FOURCC_YUY2: {  // TODO(fbarchard): Find better odd crop fix.
+      uint8_t* u = (crop_x & 1) ? dst_v : dst_u;
+      uint8_t* v = (crop_x & 1) ? dst_u : dst_v;
+      int stride_u = (crop_x & 1) ? dst_stride_v : dst_stride_u;
+      int stride_v = (crop_x & 1) ? dst_stride_u : dst_stride_v;
+      src = sample + (aligned_src_width * crop_y + crop_x) * 2;
+      r = YUY2ToI420(src, aligned_src_width * 2, dst_y, dst_stride_y, u,
+                     stride_u, v, stride_v, crop_width, inv_crop_height);
+      break;
+    }
+    case FOURCC_UYVY: {
+      uint8_t* u = (crop_x & 1) ? dst_v : dst_u;
+      uint8_t* v = (crop_x & 1) ? dst_u : dst_v;
+      int stride_u = (crop_x & 1) ? dst_stride_v : dst_stride_u;
+      int stride_v = (crop_x & 1) ? dst_stride_u : dst_stride_v;
+      src = sample + (aligned_src_width * crop_y + crop_x) * 2;
+      r = UYVYToI420(src, aligned_src_width * 2, dst_y, dst_stride_y, u,
+                     stride_u, v, stride_v, crop_width, inv_crop_height);
+      break;
+    }
+    case FOURCC_RGBP:
+      src = sample + (src_width * crop_y + crop_x) * 2;
+      r = RGB565ToI420(src, src_width * 2, dst_y, dst_stride_y, dst_u,
+                       dst_stride_u, dst_v, dst_stride_v, crop_width,
+                       inv_crop_height);
+      break;
+    case FOURCC_RGBO:
+      src = sample + (src_width * crop_y + crop_x) * 2;
+      r = ARGB1555ToI420(src, src_width * 2, dst_y, dst_stride_y, dst_u,
+                         dst_stride_u, dst_v, dst_stride_v, crop_width,
+                         inv_crop_height);
+      break;
+    case FOURCC_R444:
+      src = sample + (src_width * crop_y + crop_x) * 2;
+      r = ARGB4444ToI420(src, src_width * 2, dst_y, dst_stride_y, dst_u,
+                         dst_stride_u, dst_v, dst_stride_v, crop_width,
+                         inv_crop_height);
+      break;
+    case FOURCC_24BG:
+      src = sample + (src_width * crop_y + crop_x) * 3;
+      r = RGB24ToI420(src, src_width * 3, dst_y, dst_stride_y, dst_u,
+                      dst_stride_u, dst_v, dst_stride_v, crop_width,
+                      inv_crop_height);
+      break;
+    case FOURCC_RAW:
+      src = sample + (src_width * crop_y + crop_x) * 3;
+      r = RAWToI420(src, src_width * 3, dst_y, dst_stride_y, dst_u,
+                    dst_stride_u, dst_v, dst_stride_v, crop_width,
+                    inv_crop_height);
+      break;
+    case FOURCC_ARGB:
+      src = sample + (src_width * crop_y + crop_x) * 4;
+      r = ARGBToI420(src, src_width * 4, dst_y, dst_stride_y, dst_u,
+                     dst_stride_u, dst_v, dst_stride_v, crop_width,
+                     inv_crop_height);
+      break;
+    case FOURCC_BGRA:
+      src = sample + (src_width * crop_y + crop_x) * 4;
+      r = BGRAToI420(src, src_width * 4, dst_y, dst_stride_y, dst_u,
+                     dst_stride_u, dst_v, dst_stride_v, crop_width,
+                     inv_crop_height);
+      break;
+    case FOURCC_ABGR:
+      src = sample + (src_width * crop_y + crop_x) * 4;
+      r = ABGRToI420(src, src_width * 4, dst_y, dst_stride_y, dst_u,
+                     dst_stride_u, dst_v, dst_stride_v, crop_width,
+                     inv_crop_height);
+      break;
+    case FOURCC_RGBA:
+      src = sample + (src_width * crop_y + crop_x) * 4;
+      r = RGBAToI420(src, src_width * 4, dst_y, dst_stride_y, dst_u,
+                     dst_stride_u, dst_v, dst_stride_v, crop_width,
+                     inv_crop_height);
+      break;
+    // TODO(fbarchard): Add AR30 and AB30
+    case FOURCC_I400:
+      src = sample + src_width * crop_y + crop_x;
+      r = I400ToI420(src, src_width, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                     dst_v, dst_stride_v, crop_width, inv_crop_height);
+      break;
+    // Biplanar formats
+    case FOURCC_NV12:
+      src = sample + (src_width * crop_y + crop_x);
+      src_uv = sample + (src_width * abs_src_height) +
+               ((crop_y / 2) * aligned_src_width) + ((crop_x / 2) * 2);
+      r = NV12ToI420Rotate(src, src_width, src_uv, aligned_src_width, dst_y,
+                           dst_stride_y, dst_u, dst_stride_u, dst_v,
+                           dst_stride_v, crop_width, inv_crop_height, rotation);
+      break;
+    case FOURCC_NV21:
+      src = sample + (src_width * crop_y + crop_x);
+      src_uv = sample + (src_width * abs_src_height) +
+               ((crop_y / 2) * aligned_src_width) + ((crop_x / 2) * 2);
+      // Call NV12 but with dst_u and dst_v parameters swapped.
+      r = NV12ToI420Rotate(src, src_width, src_uv, aligned_src_width, dst_y,
+                           dst_stride_y, dst_v, dst_stride_v, dst_u,
+                           dst_stride_u, crop_width, inv_crop_height, rotation);
+      break;
+    // Triplanar formats
+    case FOURCC_I420:
+    case FOURCC_YV12: {
+      const uint8_t* src_y = sample + (src_width * crop_y + crop_x);
+      const uint8_t* src_u;
+      const uint8_t* src_v;
+      int halfwidth = (src_width + 1) / 2;
+      int halfheight = (abs_src_height + 1) / 2;
+      if (format == FOURCC_YV12) {
+        src_v = sample + src_width * abs_src_height + halfwidth * (crop_y / 2) +
+                (crop_x / 2);
+        src_u = sample + src_width * abs_src_height +
+                halfwidth * (halfheight + (crop_y / 2)) + (crop_x / 2);
+      } else {
+        src_u = sample + src_width * abs_src_height + halfwidth * (crop_y / 2) +
+                (crop_x / 2);
+        src_v = sample + src_width * abs_src_height +
+                halfwidth * (halfheight + (crop_y / 2)) + (crop_x / 2);
+      }
+      r = I420Rotate(src_y, src_width, src_u, halfwidth, src_v, halfwidth,
+                     dst_y, dst_stride_y, dst_u, dst_stride_u, dst_v,
+                     dst_stride_v, crop_width, inv_crop_height, rotation);
+      break;
+    }
+    case FOURCC_I422:
+    case FOURCC_YV16: {
+      const uint8_t* src_y = sample + src_width * crop_y + crop_x;
+      const uint8_t* src_u;
+      const uint8_t* src_v;
+      int halfwidth = (src_width + 1) / 2;
+      if (format == FOURCC_YV16) {
+        src_v = sample + src_width * abs_src_height + halfwidth * crop_y +
+                (crop_x / 2);
+        src_u = sample + src_width * abs_src_height +
+                halfwidth * (abs_src_height + crop_y) + (crop_x / 2);
+      } else {
+        src_u = sample + src_width * abs_src_height + halfwidth * crop_y +
+                (crop_x / 2);
+        src_v = sample + src_width * abs_src_height +
+                halfwidth * (abs_src_height + crop_y) + (crop_x / 2);
+      }
+      r = I422ToI420(src_y, src_width, src_u, halfwidth, src_v, halfwidth,
+                     dst_y, dst_stride_y, dst_u, dst_stride_u, dst_v,
+                     dst_stride_v, crop_width, inv_crop_height);
+      break;
+    }
+    case FOURCC_I444:
+    case FOURCC_YV24: {
+      const uint8_t* src_y = sample + src_width * crop_y + crop_x;
+      const uint8_t* src_u;
+      const uint8_t* src_v;
+      if (format == FOURCC_YV24) {
+        src_v = sample + src_width * (abs_src_height + crop_y) + crop_x;
+        src_u = sample + src_width * (abs_src_height * 2 + crop_y) + crop_x;
+      } else {
+        src_u = sample + src_width * (abs_src_height + crop_y) + crop_x;
+        src_v = sample + src_width * (abs_src_height * 2 + crop_y) + crop_x;
+      }
+      r = I444ToI420(src_y, src_width, src_u, src_width, src_v, src_width,
+                     dst_y, dst_stride_y, dst_u, dst_stride_u, dst_v,
+                     dst_stride_v, crop_width, inv_crop_height);
+      break;
+    }
+#ifdef HAVE_JPEG
+    case FOURCC_MJPG:
+      r = MJPGToI420(sample, sample_size, dst_y, dst_stride_y, dst_u,
+                     dst_stride_u, dst_v, dst_stride_v, src_width,
+                     abs_src_height, crop_width, inv_crop_height);
+      break;
+#endif
+    default:
+      r = -1;  // unknown fourcc - return failure code.
+  }
+
+  if (need_buf) {
+    if (!r) {
+      r = I420Rotate(dst_y, dst_stride_y, dst_u, dst_stride_u, dst_v,
+                     dst_stride_v, tmp_y, tmp_y_stride, tmp_u, tmp_u_stride,
+                     tmp_v, tmp_v_stride, crop_width, abs_crop_height,
+                     rotation);
+    }
+    free(rotate_buffer);
+  }
+
+  return r;
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/cpu_id.cc

@@ -0,0 +1,498 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/cpu_id.h"
+
+#if defined(_MSC_VER)
+#include <intrin.h>  // For __cpuidex()
+#endif
+#if !defined(__pnacl__) && !defined(__CLR_VER) &&                           \
+    !defined(__native_client__) && (defined(_M_IX86) || defined(_M_X64)) && \
+    defined(_MSC_FULL_VER) && (_MSC_FULL_VER >= 160040219)
+#include <immintrin.h>  // For _xgetbv()
+#endif
+
+// For ArmCpuCaps() but unittested on all platforms
+#include <stdio.h>  // For fopen()
+#include <string.h>
+
+#if defined(__linux__) && defined(__aarch64__)
+#include <sys/auxv.h>  // For getauxval()
+#endif
+
+#if defined(_WIN32) && defined(__aarch64__)
+#undef WIN32_LEAN_AND_MEAN
+#define WIN32_LEAN_AND_MEAN
+#undef WIN32_EXTRA_LEAN
+#define WIN32_EXTRA_LEAN
+#include <windows.h>  // For IsProcessorFeaturePresent()
+#endif
+
+#if defined(__APPLE__) && defined(__aarch64__)
+#include <sys/sysctl.h>  // For sysctlbyname()
+#endif
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// For functions that use the stack and have runtime checks for overflow,
+// use SAFEBUFFERS to avoid additional check.
+#if defined(_MSC_FULL_VER) && (_MSC_FULL_VER >= 160040219) && \
+    !defined(__clang__)
+#define SAFEBUFFERS __declspec(safebuffers)
+#else
+#define SAFEBUFFERS
+#endif
+
+// cpu_info_ variable for SIMD instruction sets detected.
+LIBYUV_API int cpu_info_ = 0;
+
+// Low level cpuid for X86.
+#if (defined(_M_IX86) || defined(_M_X64) || defined(__i386__) || \
+     defined(__x86_64__)) &&                                     \
+    !defined(__pnacl__) && !defined(__CLR_VER)
+LIBYUV_API
+void CpuId(int info_eax, int info_ecx, int* cpu_info) {
+#if defined(_MSC_VER)
+// Visual C version uses intrinsic or inline x86 assembly.
+#if defined(_MSC_FULL_VER) && (_MSC_FULL_VER >= 160040219)
+  __cpuidex(cpu_info, info_eax, info_ecx);
+#elif defined(_M_IX86)
+  __asm {
+    mov        eax, info_eax
+    mov        ecx, info_ecx
+    mov        edi, cpu_info
+    cpuid
+    mov        [edi], eax
+    mov        [edi + 4], ebx
+    mov        [edi + 8], ecx
+    mov        [edi + 12], edx
+  }
+#else  // Visual C but not x86
+  if (info_ecx == 0) {
+    __cpuid(cpu_info, info_eax);
+  } else {
+    cpu_info[3] = cpu_info[2] = cpu_info[1] = cpu_info[0] = 0u;
+  }
+#endif
+// GCC version uses inline x86 assembly.
+#else  // defined(_MSC_VER)
+  int info_ebx, info_edx;
+  asm volatile(
+#if defined(__i386__) && defined(__PIC__)
+      // Preserve ebx for fpic 32 bit.
+      "mov         %%ebx, %%edi                  \n"
+      "cpuid                                     \n"
+      "xchg        %%edi, %%ebx                  \n"
+      : "=D"(info_ebx),
+#else
+      "cpuid                                     \n"
+      : "=b"(info_ebx),
+#endif  //  defined( __i386__) && defined(__PIC__)
+        "+a"(info_eax), "+c"(info_ecx), "=d"(info_edx));
+  cpu_info[0] = info_eax;
+  cpu_info[1] = info_ebx;
+  cpu_info[2] = info_ecx;
+  cpu_info[3] = info_edx;
+#endif  // defined(_MSC_VER)
+}
+#else  // (defined(_M_IX86) || defined(_M_X64) ...
+LIBYUV_API
+void CpuId(int eax, int ecx, int* cpu_info) {
+  (void)eax;
+  (void)ecx;
+  cpu_info[0] = cpu_info[1] = cpu_info[2] = cpu_info[3] = 0;
+}
+#endif
+
+// For VS2010 and earlier emit can be used:
+//   _asm _emit 0x0f _asm _emit 0x01 _asm _emit 0xd0  // For VS2010 and earlier.
+//  __asm {
+//    xor        ecx, ecx    // xcr 0
+//    xgetbv
+//    mov        xcr0, eax
+//  }
+// For VS2013 and earlier 32 bit, the _xgetbv(0) optimizer produces bad code.
+// https://code.google.com/p/libyuv/issues/detail?id=529
+#if defined(_M_IX86) && defined(_MSC_VER) && (_MSC_VER < 1900)
+#pragma optimize("g", off)
+#endif
+#if (defined(_M_IX86) || defined(_M_X64) || defined(__i386__) || \
+     defined(__x86_64__)) &&                                     \
+    !defined(__pnacl__) && !defined(__CLR_VER) && !defined(__native_client__)
+// X86 CPUs have xgetbv to detect OS saves high parts of ymm registers.
+static int GetXCR0() {
+  int xcr0 = 0;
+#if defined(_MSC_FULL_VER) && (_MSC_FULL_VER >= 160040219)
+  xcr0 = (int)_xgetbv(0);  // VS2010 SP1 required.  NOLINT
+#elif defined(__i386__) || defined(__x86_64__)
+  asm(".byte 0x0f, 0x01, 0xd0" : "=a"(xcr0) : "c"(0) : "%edx");
+#endif  // defined(__i386__) || defined(__x86_64__)
+  return xcr0;
+}
+#else
+// xgetbv unavailable to query for OSSave support.  Return 0.
+#define GetXCR0() 0
+#endif  // defined(_M_IX86) || defined(_M_X64) ..
+// Return optimization to previous setting.
+#if defined(_M_IX86) && defined(_MSC_VER) && (_MSC_VER < 1900)
+#pragma optimize("g", on)
+#endif
+
+static int cpuinfo_search(const char* cpuinfo_line,
+                          const char* needle,
+                          int needle_len) {
+  const char* p = strstr(cpuinfo_line, needle);
+  return p && (p[needle_len] == ' ' || p[needle_len] == '\n');
+}
+
+// Based on libvpx arm_cpudetect.c
+// For Arm, but public to allow testing on any CPU
+LIBYUV_API SAFEBUFFERS int ArmCpuCaps(const char* cpuinfo_name) {
+  char cpuinfo_line[512];
+  FILE* f = fopen(cpuinfo_name, "re");
+  if (!f) {
+    // Assume Neon if /proc/cpuinfo is unavailable.
+    // This will occur for Chrome sandbox for Pepper or Render process.
+    return kCpuHasNEON;
+  }
+  memset(cpuinfo_line, 0, sizeof(cpuinfo_line));
+  int features = 0;
+  while (fgets(cpuinfo_line, sizeof(cpuinfo_line), f)) {
+    if (memcmp(cpuinfo_line, "Features", 8) == 0) {
+      if (cpuinfo_search(cpuinfo_line, " neon", 5)) {
+        features |= kCpuHasNEON;
+      }
+    }
+  }
+  fclose(f);
+  return features;
+}
+
+#ifdef __aarch64__
+#ifdef __linux__
+// Define hwcap values ourselves: building with an old auxv header where these
+// hwcap values are not defined should not prevent features from being enabled.
+#define YUV_AARCH64_HWCAP_ASIMDDP (1 << 20)
+#define YUV_AARCH64_HWCAP_SVE (1 << 22)
+#define YUV_AARCH64_HWCAP2_SVE2 (1 << 1)
+#define YUV_AARCH64_HWCAP2_I8MM (1 << 13)
+#define YUV_AARCH64_HWCAP2_SME (1 << 23)
+
+// For AArch64, but public to allow testing on any CPU.
+LIBYUV_API SAFEBUFFERS int AArch64CpuCaps(unsigned long hwcap,
+                                          unsigned long hwcap2) {
+  // Neon is mandatory on AArch64, so enable regardless of hwcaps.
+  int features = kCpuHasNEON;
+
+  // Don't try to enable later extensions unless earlier extensions are also
+  // reported available. Some of these constraints aren't strictly required by
+  // the architecture, but are satisfied by all micro-architectures of
+  // interest. This also avoids an issue on some emulators where true
+  // architectural constraints are not satisfied, e.g. SVE2 may be reported as
+  // available while SVE is not.
+  if (hwcap & YUV_AARCH64_HWCAP_ASIMDDP) {
+    features |= kCpuHasNeonDotProd;
+    if (hwcap2 & YUV_AARCH64_HWCAP2_I8MM) {
+      features |= kCpuHasNeonI8MM;
+      if (hwcap & YUV_AARCH64_HWCAP_SVE) {
+        features |= kCpuHasSVE;
+        if (hwcap2 & YUV_AARCH64_HWCAP2_SVE2) {
+          features |= kCpuHasSVE2;
+          if (hwcap2 & YUV_AARCH64_HWCAP2_SME) {
+            features |= kCpuHasSME;
+          }
+        }
+      }
+    }
+  }
+  return features;
+}
+
+#elif defined(_WIN32)
+// For AArch64, but public to allow testing on any CPU.
+LIBYUV_API SAFEBUFFERS int AArch64CpuCaps() {
+  // Neon is mandatory on AArch64, so enable unconditionally.
+  int features = kCpuHasNEON;
+
+  // For more information on IsProcessorFeaturePresent(), see:
+  // https://learn.microsoft.com/en-us/windows/win32/api/processthreadsapi/nf-processthreadsapi-isprocessorfeaturepresent#parameters
+#ifdef PF_ARM_V82_DP_INSTRUCTIONS_AVAILABLE
+  if (IsProcessorFeaturePresent(PF_ARM_V82_DP_INSTRUCTIONS_AVAILABLE)) {
+    features |= kCpuHasNeonDotProd;
+  }
+#endif
+  // No Neon I8MM or SVE feature detection available here at time of writing.
+  return features;
+}
+
+#elif defined(__APPLE__)
+static bool have_feature(const char* feature) {
+  // For more information on sysctlbyname(), see:
+  // https://developer.apple.com/documentation/kernel/1387446-sysctlbyname/determining_instruction_set_characteristics
+  int64_t feature_present = 0;
+  size_t size = sizeof(feature_present);
+  if (sysctlbyname(feature, &feature_present, &size, NULL, 0) != 0) {
+    return false;
+  }
+  return feature_present;
+}
+
+// For AArch64, but public to allow testing on any CPU.
+LIBYUV_API SAFEBUFFERS int AArch64CpuCaps() {
+  // Neon is mandatory on AArch64, so enable unconditionally.
+  int features = kCpuHasNEON;
+
+  if (have_feature("hw.optional.arm.FEAT_DotProd")) {
+    features |= kCpuHasNeonDotProd;
+    if (have_feature("hw.optional.arm.FEAT_I8MM")) {
+      features |= kCpuHasNeonI8MM;
+      if (have_feature("hw.optional.arm.FEAT_SME2")) {
+        features |= kCpuHasSME;
+      }
+    }
+  }
+  // No SVE feature detection available here at time of writing.
+  return features;
+}
+
+#else  // !defined(__linux__) && !defined(_WIN32) && !defined(__APPLE__)
+// For AArch64, but public to allow testing on any CPU.
+LIBYUV_API SAFEBUFFERS int AArch64CpuCaps() {
+  // Neon is mandatory on AArch64, so enable unconditionally.
+  int features = kCpuHasNEON;
+
+  // TODO(libyuv:980) support feature detection on other platforms.
+
+  return features;
+}
+#endif
+#endif  // defined(__aarch64__)
+
+LIBYUV_API SAFEBUFFERS int RiscvCpuCaps(const char* cpuinfo_name) {
+  char cpuinfo_line[512];
+  int flag = 0;
+  FILE* f = fopen(cpuinfo_name, "re");
+  if (!f) {
+#if defined(__riscv_vector)
+    // Assume RVV if /proc/cpuinfo is unavailable.
+    // This will occur for Chrome sandbox for Pepper or Render process.
+    return kCpuHasRVV;
+#else
+    return 0;
+#endif
+  }
+  memset(cpuinfo_line, 0, sizeof(cpuinfo_line));
+  while (fgets(cpuinfo_line, sizeof(cpuinfo_line), f)) {
+    if (memcmp(cpuinfo_line, "isa", 3) == 0) {
+      // ISA string must begin with rv64{i,e,g} for a 64-bit processor.
+      char* isa = strstr(cpuinfo_line, "rv64");
+      if (isa) {
+        size_t isa_len = strlen(isa);
+        char* extensions;
+        size_t extensions_len = 0;
+        size_t std_isa_len;
+        // Remove the new-line character at the end of string
+        if (isa[isa_len - 1] == '\n') {
+          isa[--isa_len] = '\0';
+        }
+        // 5 ISA characters
+        if (isa_len < 5) {
+          fclose(f);
+          return 0;
+        }
+        // Skip {i,e,g} canonical checking.
+        // Skip rvxxx
+        isa += 5;
+        // Find the very first occurrence of 's', 'x' or 'z'.
+        // To detect multi-letter standard, non-standard, and
+        // supervisor-level extensions.
+        extensions = strpbrk(isa, "zxs");
+        if (extensions) {
+          // Multi-letter extensions are seperated by a single underscore
+          // as described in RISC-V User-Level ISA V2.2.
+          char* ext = strtok(extensions, "_");
+          extensions_len = strlen(extensions);
+          while (ext) {
+            // Search for the ZVFH (Vector FP16) extension.
+            if (!strcmp(ext, "zvfh")) {
+              flag |= kCpuHasRVVZVFH;
+            }
+            ext = strtok(NULL, "_");
+          }
+        }
+        std_isa_len = isa_len - extensions_len - 5;
+        // Detect the v in the standard single-letter extensions.
+        if (memchr(isa, 'v', std_isa_len)) {
+          // The RVV implied the F extension.
+          flag |= kCpuHasRVV;
+        }
+      }
+    }
+#if defined(__riscv_vector)
+    // Assume RVV if /proc/cpuinfo is from x86 host running QEMU.
+    else if ((memcmp(cpuinfo_line, "vendor_id\t: GenuineIntel", 24) == 0) ||
+             (memcmp(cpuinfo_line, "vendor_id\t: AuthenticAMD", 24) == 0)) {
+      fclose(f);
+      return kCpuHasRVV;
+    }
+#endif
+  }
+  fclose(f);
+  return flag;
+}
+
+LIBYUV_API SAFEBUFFERS int MipsCpuCaps(const char* cpuinfo_name) {
+  char cpuinfo_line[512];
+  int flag = 0;
+  FILE* f = fopen(cpuinfo_name, "re");
+  if (!f) {
+    // Assume nothing if /proc/cpuinfo is unavailable.
+    // This will occur for Chrome sandbox for Pepper or Render process.
+    return 0;
+  }
+  memset(cpuinfo_line, 0, sizeof(cpuinfo_line));
+  while (fgets(cpuinfo_line, sizeof(cpuinfo_line), f)) {
+    if (memcmp(cpuinfo_line, "cpu model", 9) == 0) {
+      // Workaround early kernel without MSA in ASEs line.
+      if (strstr(cpuinfo_line, "Loongson-2K")) {
+        flag |= kCpuHasMSA;
+      }
+    }
+    if (memcmp(cpuinfo_line, "ASEs implemented", 16) == 0) {
+      if (strstr(cpuinfo_line, "msa")) {
+        flag |= kCpuHasMSA;
+      }
+      // ASEs is the last line, so we can break here.
+      break;
+    }
+  }
+  fclose(f);
+  return flag;
+}
+
+#define LOONGARCH_CFG2 0x2
+#define LOONGARCH_CFG2_LSX (1 << 6)
+#define LOONGARCH_CFG2_LASX (1 << 7)
+
+#if defined(__loongarch__)
+LIBYUV_API SAFEBUFFERS int LoongarchCpuCaps(void) {
+  int flag = 0;
+  uint32_t cfg2 = 0;
+
+  __asm__ volatile("cpucfg %0, %1 \n\t" : "+&r"(cfg2) : "r"(LOONGARCH_CFG2));
+
+  if (cfg2 & LOONGARCH_CFG2_LSX)
+    flag |= kCpuHasLSX;
+
+  if (cfg2 & LOONGARCH_CFG2_LASX)
+    flag |= kCpuHasLASX;
+  return flag;
+}
+#endif
+
+static SAFEBUFFERS int GetCpuFlags(void) {
+  int cpu_info = 0;
+#if !defined(__pnacl__) && !defined(__CLR_VER) &&                   \
+    (defined(__x86_64__) || defined(_M_X64) || defined(__i386__) || \
+     defined(_M_IX86))
+  int cpu_info0[4] = {0, 0, 0, 0};
+  int cpu_info1[4] = {0, 0, 0, 0};
+  int cpu_info7[4] = {0, 0, 0, 0};
+  int cpu_einfo7[4] = {0, 0, 0, 0};
+  CpuId(0, 0, cpu_info0);
+  CpuId(1, 0, cpu_info1);
+  if (cpu_info0[0] >= 7) {
+    CpuId(7, 0, cpu_info7);
+    CpuId(7, 1, cpu_einfo7);
+  }
+  cpu_info = kCpuHasX86 | ((cpu_info1[3] & 0x04000000) ? kCpuHasSSE2 : 0) |
+             ((cpu_info1[2] & 0x00000200) ? kCpuHasSSSE3 : 0) |
+             ((cpu_info1[2] & 0x00080000) ? kCpuHasSSE41 : 0) |
+             ((cpu_info1[2] & 0x00100000) ? kCpuHasSSE42 : 0) |
+             ((cpu_info7[1] & 0x00000200) ? kCpuHasERMS : 0);
+
+  // AVX requires OS saves YMM registers.
+  if (((cpu_info1[2] & 0x1c000000) == 0x1c000000) &&  // AVX and OSXSave
+      ((GetXCR0() & 6) == 6)) {  // Test OS saves YMM registers
+    cpu_info |= kCpuHasAVX | ((cpu_info7[1] & 0x00000020) ? kCpuHasAVX2 : 0) |
+                ((cpu_info1[2] & 0x00001000) ? kCpuHasFMA3 : 0) |
+                ((cpu_info1[2] & 0x20000000) ? kCpuHasF16C : 0) |
+                ((cpu_einfo7[0] & 0x00000010) ? kCpuHasAVXVNNI : 0) |
+                ((cpu_einfo7[3] & 0x00000010) ? kCpuHasAVXVNNIINT8 : 0);
+
+    // Detect AVX512bw
+    if ((GetXCR0() & 0xe0) == 0xe0) {
+      cpu_info |= (cpu_info7[1] & 0x40000000) ? kCpuHasAVX512BW : 0;
+      cpu_info |= (cpu_info7[1] & 0x80000000) ? kCpuHasAVX512VL : 0;
+      cpu_info |= (cpu_info7[2] & 0x00000002) ? kCpuHasAVX512VBMI : 0;
+      cpu_info |= (cpu_info7[2] & 0x00000040) ? kCpuHasAVX512VBMI2 : 0;
+      cpu_info |= (cpu_info7[2] & 0x00000800) ? kCpuHasAVX512VNNI : 0;
+      cpu_info |= (cpu_info7[2] & 0x00001000) ? kCpuHasAVX512VBITALG : 0;
+      cpu_info |= (cpu_einfo7[3] & 0x00080000) ? kCpuHasAVX10 : 0;
+      cpu_info |= (cpu_info7[3] & 0x02000000) ? kCpuHasAMXINT8 : 0;
+    }
+  }
+#endif
+#if defined(__mips__) && defined(__linux__)
+  cpu_info = MipsCpuCaps("/proc/cpuinfo");
+  cpu_info |= kCpuHasMIPS;
+#endif
+#if defined(__loongarch__) && defined(__linux__)
+  cpu_info = LoongarchCpuCaps();
+  cpu_info |= kCpuHasLOONGARCH;
+#endif
+#if defined(__arm__) || defined(__aarch64__)
+#if defined(__aarch64__) && defined(__linux__)
+  // getauxval is supported since Android SDK version 18, minimum at time of
+  // writing is 21, so should be safe to always use this. If getauxval is
+  // somehow disabled then getauxval returns 0, which will leave Neon enabled
+  // since Neon is mandatory on AArch64.
+  unsigned long hwcap = getauxval(AT_HWCAP);
+  unsigned long hwcap2 = getauxval(AT_HWCAP2);
+  cpu_info = AArch64CpuCaps(hwcap, hwcap2);
+#elif defined(__aarch64__)
+  cpu_info = AArch64CpuCaps();
+#else
+  // gcc -mfpu=neon defines __ARM_NEON__
+  // __ARM_NEON__ generates code that requires Neon.  NaCL also requires Neon.
+  // For Linux, /proc/cpuinfo can be tested but without that assume Neon.
+  // Linux arm parse text file for neon detect.
+  cpu_info = ArmCpuCaps("/proc/cpuinfo");
+#endif
+  cpu_info |= kCpuHasARM;
+#endif  // __arm__
+#if defined(__riscv) && defined(__linux__)
+  cpu_info = RiscvCpuCaps("/proc/cpuinfo");
+  cpu_info |= kCpuHasRISCV;
+#endif  // __riscv
+  cpu_info |= kCpuInitialized;
+  return cpu_info;
+}
+
+// Note that use of this function is not thread safe.
+LIBYUV_API
+int MaskCpuFlags(int enable_flags) {
+  int cpu_info = GetCpuFlags() & enable_flags;
+  SetCpuFlags(cpu_info);
+  return cpu_info;
+}
+
+LIBYUV_API
+int InitCpuFlags(void) {
+  return MaskCpuFlags(-1);
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/mjpeg_decoder.cc

@@ -0,0 +1,580 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/mjpeg_decoder.h"
+
+#ifdef HAVE_JPEG
+#include <assert.h>
+
+#if !defined(__pnacl__) && !defined(__CLR_VER) && !defined(COVERAGE_ENABLED)
+// Must be included before jpeglib.
+#include <setjmp.h>
+#define HAVE_SETJMP
+
+#if defined(_MSC_VER)
+// disable warning 4324: structure was padded due to __declspec(align())
+#pragma warning(disable : 4324)
+#endif
+
+#endif
+
+#include <stdio.h>  // For jpeglib.h.
+
+// C++ build requires extern C for jpeg internals.
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <jpeglib.h>
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+
+#include "libyuv/planar_functions.h"  // For CopyPlane().
+
+namespace libyuv {
+
+#ifdef HAVE_SETJMP
+struct SetJmpErrorMgr {
+  jpeg_error_mgr base;  // Must be at the top
+  jmp_buf setjmp_buffer;
+};
+#endif
+
+const int MJpegDecoder::kColorSpaceUnknown = JCS_UNKNOWN;
+const int MJpegDecoder::kColorSpaceGrayscale = JCS_GRAYSCALE;
+const int MJpegDecoder::kColorSpaceRgb = JCS_RGB;
+const int MJpegDecoder::kColorSpaceYCbCr = JCS_YCbCr;
+const int MJpegDecoder::kColorSpaceCMYK = JCS_CMYK;
+const int MJpegDecoder::kColorSpaceYCCK = JCS_YCCK;
+
+// Methods that are passed to jpeglib.
+boolean fill_input_buffer(jpeg_decompress_struct* cinfo);
+void init_source(jpeg_decompress_struct* cinfo);
+void skip_input_data(jpeg_decompress_struct* cinfo, long num_bytes);  // NOLINT
+void term_source(jpeg_decompress_struct* cinfo);
+void ErrorHandler(jpeg_common_struct* cinfo);
+void OutputHandler(jpeg_common_struct* cinfo);
+
+MJpegDecoder::MJpegDecoder()
+    : has_scanline_padding_(LIBYUV_FALSE),
+      num_outbufs_(0),
+      scanlines_(NULL),
+      scanlines_sizes_(NULL),
+      databuf_(NULL),
+      databuf_strides_(NULL) {
+  decompress_struct_ = new jpeg_decompress_struct;
+  source_mgr_ = new jpeg_source_mgr;
+#ifdef HAVE_SETJMP
+  error_mgr_ = new SetJmpErrorMgr;
+  decompress_struct_->err = jpeg_std_error(&error_mgr_->base);
+  // Override standard exit()-based error handler.
+  error_mgr_->base.error_exit = &ErrorHandler;
+  error_mgr_->base.output_message = &OutputHandler;
+#endif
+  decompress_struct_->client_data = NULL;
+  source_mgr_->init_source = &init_source;
+  source_mgr_->fill_input_buffer = &fill_input_buffer;
+  source_mgr_->skip_input_data = &skip_input_data;
+  source_mgr_->resync_to_restart = &jpeg_resync_to_restart;
+  source_mgr_->term_source = &term_source;
+  jpeg_create_decompress(decompress_struct_);
+  decompress_struct_->src = source_mgr_;
+  buf_vec_.buffers = &buf_;
+  buf_vec_.len = 1;
+}
+
+MJpegDecoder::~MJpegDecoder() {
+  jpeg_destroy_decompress(decompress_struct_);
+  delete decompress_struct_;
+  delete source_mgr_;
+#ifdef HAVE_SETJMP
+  delete error_mgr_;
+#endif
+  DestroyOutputBuffers();
+}
+
+LIBYUV_BOOL MJpegDecoder::LoadFrame(const uint8_t* src, size_t src_len) {
+  if (!ValidateJpeg(src, src_len)) {
+    return LIBYUV_FALSE;
+  }
+
+  buf_.data = src;
+  buf_.len = (int)src_len;
+  buf_vec_.pos = 0;
+  decompress_struct_->client_data = &buf_vec_;
+#ifdef HAVE_SETJMP
+  if (setjmp(error_mgr_->setjmp_buffer)) {
+    // We called jpeg_read_header, it experienced an error, and we called
+    // longjmp() and rewound the stack to here. Return error.
+    return LIBYUV_FALSE;
+  }
+#endif
+  if (jpeg_read_header(decompress_struct_, TRUE) != JPEG_HEADER_OK) {
+    // ERROR: Bad MJPEG header
+    return LIBYUV_FALSE;
+  }
+  AllocOutputBuffers(GetNumComponents());
+  for (int i = 0; i < num_outbufs_; ++i) {
+    int scanlines_size = GetComponentScanlinesPerImcuRow(i);
+    if (scanlines_sizes_[i] != scanlines_size) {
+      if (scanlines_[i]) {
+        delete scanlines_[i];
+      }
+      scanlines_[i] = new uint8_t*[scanlines_size];
+      scanlines_sizes_[i] = scanlines_size;
+    }
+
+    // We allocate padding for the final scanline to pad it up to DCTSIZE bytes
+    // to avoid memory errors, since jpeglib only reads full MCUs blocks. For
+    // the preceding scanlines, the padding is not needed/wanted because the
+    // following addresses will already be valid (they are the initial bytes of
+    // the next scanline) and will be overwritten when jpeglib writes out that
+    // next scanline.
+    int databuf_stride = GetComponentStride(i);
+    int databuf_size = scanlines_size * databuf_stride;
+    if (databuf_strides_[i] != databuf_stride) {
+      if (databuf_[i]) {
+        delete databuf_[i];
+      }
+      databuf_[i] = new uint8_t[databuf_size];
+      databuf_strides_[i] = databuf_stride;
+    }
+
+    if (GetComponentStride(i) != GetComponentWidth(i)) {
+      has_scanline_padding_ = LIBYUV_TRUE;
+    }
+  }
+  return LIBYUV_TRUE;
+}
+
+static int DivideAndRoundUp(int numerator, int denominator) {
+  return (numerator + denominator - 1) / denominator;
+}
+
+static int DivideAndRoundDown(int numerator, int denominator) {
+  return numerator / denominator;
+}
+
+// Returns width of the last loaded frame.
+int MJpegDecoder::GetWidth() {
+  return decompress_struct_->image_width;
+}
+
+// Returns height of the last loaded frame.
+int MJpegDecoder::GetHeight() {
+  return decompress_struct_->image_height;
+}
+
+// Returns format of the last loaded frame. The return value is one of the
+// kColorSpace* constants.
+int MJpegDecoder::GetColorSpace() {
+  return decompress_struct_->jpeg_color_space;
+}
+
+// Number of color components in the color space.
+int MJpegDecoder::GetNumComponents() {
+  return decompress_struct_->num_components;
+}
+
+// Sample factors of the n-th component.
+int MJpegDecoder::GetHorizSampFactor(int component) {
+  return decompress_struct_->comp_info[component].h_samp_factor;
+}
+
+int MJpegDecoder::GetVertSampFactor(int component) {
+  return decompress_struct_->comp_info[component].v_samp_factor;
+}
+
+int MJpegDecoder::GetHorizSubSampFactor(int component) {
+  return decompress_struct_->max_h_samp_factor / GetHorizSampFactor(component);
+}
+
+int MJpegDecoder::GetVertSubSampFactor(int component) {
+  return decompress_struct_->max_v_samp_factor / GetVertSampFactor(component);
+}
+
+int MJpegDecoder::GetImageScanlinesPerImcuRow() {
+  return decompress_struct_->max_v_samp_factor * DCTSIZE;
+}
+
+int MJpegDecoder::GetComponentScanlinesPerImcuRow(int component) {
+  int vs = GetVertSubSampFactor(component);
+  return DivideAndRoundUp(GetImageScanlinesPerImcuRow(), vs);
+}
+
+int MJpegDecoder::GetComponentWidth(int component) {
+  int hs = GetHorizSubSampFactor(component);
+  return DivideAndRoundUp(GetWidth(), hs);
+}
+
+int MJpegDecoder::GetComponentHeight(int component) {
+  int vs = GetVertSubSampFactor(component);
+  return DivideAndRoundUp(GetHeight(), vs);
+}
+
+// Get width in bytes padded out to a multiple of DCTSIZE
+int MJpegDecoder::GetComponentStride(int component) {
+  return (GetComponentWidth(component) + DCTSIZE - 1) & ~(DCTSIZE - 1);
+}
+
+int MJpegDecoder::GetComponentSize(int component) {
+  return GetComponentWidth(component) * GetComponentHeight(component);
+}
+
+LIBYUV_BOOL MJpegDecoder::UnloadFrame() {
+#ifdef HAVE_SETJMP
+  if (setjmp(error_mgr_->setjmp_buffer)) {
+    // We called jpeg_abort_decompress, it experienced an error, and we called
+    // longjmp() and rewound the stack to here. Return error.
+    return LIBYUV_FALSE;
+  }
+#endif
+  jpeg_abort_decompress(decompress_struct_);
+  return LIBYUV_TRUE;
+}
+
+// TODO(fbarchard): Allow rectangle to be specified: x, y, width, height.
+LIBYUV_BOOL MJpegDecoder::DecodeToBuffers(uint8_t** planes,
+                                          int dst_width,
+                                          int dst_height) {
+  if (dst_width != GetWidth() || dst_height > GetHeight()) {
+    // ERROR: Bad dimensions
+    return LIBYUV_FALSE;
+  }
+#ifdef HAVE_SETJMP
+  if (setjmp(error_mgr_->setjmp_buffer)) {
+    // We called into jpeglib, it experienced an error sometime during this
+    // function call, and we called longjmp() and rewound the stack to here.
+    // Return error.
+    return LIBYUV_FALSE;
+  }
+#endif
+  if (!StartDecode()) {
+    return LIBYUV_FALSE;
+  }
+  SetScanlinePointers(databuf_);
+  int lines_left = dst_height;
+  // Compute amount of lines to skip to implement vertical crop.
+  // TODO(fbarchard): Ensure skip is a multiple of maximum component
+  // subsample. ie 2
+  int skip = (GetHeight() - dst_height) / 2;
+  if (skip > 0) {
+    // There is no API to skip lines in the output data, so we read them
+    // into the temp buffer.
+    while (skip >= GetImageScanlinesPerImcuRow()) {
+      if (!DecodeImcuRow()) {
+        FinishDecode();
+        return LIBYUV_FALSE;
+      }
+      skip -= GetImageScanlinesPerImcuRow();
+    }
+    if (skip > 0) {
+      // Have a partial iMCU row left over to skip. Must read it and then
+      // copy the parts we want into the destination.
+      if (!DecodeImcuRow()) {
+        FinishDecode();
+        return LIBYUV_FALSE;
+      }
+      for (int i = 0; i < num_outbufs_; ++i) {
+        // TODO(fbarchard): Compute skip to avoid this
+        assert(skip % GetVertSubSampFactor(i) == 0);
+        int rows_to_skip = DivideAndRoundDown(skip, GetVertSubSampFactor(i));
+        int scanlines_to_copy =
+            GetComponentScanlinesPerImcuRow(i) - rows_to_skip;
+        int data_to_skip = rows_to_skip * GetComponentStride(i);
+        CopyPlane(databuf_[i] + data_to_skip, GetComponentStride(i), planes[i],
+                  GetComponentWidth(i), GetComponentWidth(i),
+                  scanlines_to_copy);
+        planes[i] += scanlines_to_copy * GetComponentWidth(i);
+      }
+      lines_left -= (GetImageScanlinesPerImcuRow() - skip);
+    }
+  }
+
+  // Read full MCUs but cropped horizontally
+  for (; lines_left > GetImageScanlinesPerImcuRow();
+       lines_left -= GetImageScanlinesPerImcuRow()) {
+    if (!DecodeImcuRow()) {
+      FinishDecode();
+      return LIBYUV_FALSE;
+    }
+    for (int i = 0; i < num_outbufs_; ++i) {
+      int scanlines_to_copy = GetComponentScanlinesPerImcuRow(i);
+      CopyPlane(databuf_[i], GetComponentStride(i), planes[i],
+                GetComponentWidth(i), GetComponentWidth(i), scanlines_to_copy);
+      planes[i] += scanlines_to_copy * GetComponentWidth(i);
+    }
+  }
+
+  if (lines_left > 0) {
+    // Have a partial iMCU row left over to decode.
+    if (!DecodeImcuRow()) {
+      FinishDecode();
+      return LIBYUV_FALSE;
+    }
+    for (int i = 0; i < num_outbufs_; ++i) {
+      int scanlines_to_copy =
+          DivideAndRoundUp(lines_left, GetVertSubSampFactor(i));
+      CopyPlane(databuf_[i], GetComponentStride(i), planes[i],
+                GetComponentWidth(i), GetComponentWidth(i), scanlines_to_copy);
+      planes[i] += scanlines_to_copy * GetComponentWidth(i);
+    }
+  }
+  return FinishDecode();
+}
+
+LIBYUV_BOOL MJpegDecoder::DecodeToCallback(CallbackFunction fn,
+                                           void* opaque,
+                                           int dst_width,
+                                           int dst_height) {
+  if (dst_width != GetWidth() || dst_height > GetHeight()) {
+    // ERROR: Bad dimensions
+    return LIBYUV_FALSE;
+  }
+#ifdef HAVE_SETJMP
+  if (setjmp(error_mgr_->setjmp_buffer)) {
+    // We called into jpeglib, it experienced an error sometime during this
+    // function call, and we called longjmp() and rewound the stack to here.
+    // Return error.
+    return LIBYUV_FALSE;
+  }
+#endif
+  if (!StartDecode()) {
+    return LIBYUV_FALSE;
+  }
+  SetScanlinePointers(databuf_);
+  int lines_left = dst_height;
+  // TODO(fbarchard): Compute amount of lines to skip to implement vertical crop
+  int skip = (GetHeight() - dst_height) / 2;
+  if (skip > 0) {
+    while (skip >= GetImageScanlinesPerImcuRow()) {
+      if (!DecodeImcuRow()) {
+        FinishDecode();
+        return LIBYUV_FALSE;
+      }
+      skip -= GetImageScanlinesPerImcuRow();
+    }
+    if (skip > 0) {
+      // Have a partial iMCU row left over to skip.
+      if (!DecodeImcuRow()) {
+        FinishDecode();
+        return LIBYUV_FALSE;
+      }
+      for (int i = 0; i < num_outbufs_; ++i) {
+        // TODO(fbarchard): Compute skip to avoid this
+        assert(skip % GetVertSubSampFactor(i) == 0);
+        int rows_to_skip = DivideAndRoundDown(skip, GetVertSubSampFactor(i));
+        int data_to_skip = rows_to_skip * GetComponentStride(i);
+        // Change our own data buffer pointers so we can pass them to the
+        // callback.
+        databuf_[i] += data_to_skip;
+      }
+      int scanlines_to_copy = GetImageScanlinesPerImcuRow() - skip;
+      (*fn)(opaque, databuf_, databuf_strides_, scanlines_to_copy);
+      // Now change them back.
+      for (int i = 0; i < num_outbufs_; ++i) {
+        int rows_to_skip = DivideAndRoundDown(skip, GetVertSubSampFactor(i));
+        int data_to_skip = rows_to_skip * GetComponentStride(i);
+        databuf_[i] -= data_to_skip;
+      }
+      lines_left -= scanlines_to_copy;
+    }
+  }
+  // Read full MCUs until we get to the crop point.
+  for (; lines_left >= GetImageScanlinesPerImcuRow();
+       lines_left -= GetImageScanlinesPerImcuRow()) {
+    if (!DecodeImcuRow()) {
+      FinishDecode();
+      return LIBYUV_FALSE;
+    }
+    (*fn)(opaque, databuf_, databuf_strides_, GetImageScanlinesPerImcuRow());
+  }
+  if (lines_left > 0) {
+    // Have a partial iMCU row left over to decode.
+    if (!DecodeImcuRow()) {
+      FinishDecode();
+      return LIBYUV_FALSE;
+    }
+    (*fn)(opaque, databuf_, databuf_strides_, lines_left);
+  }
+  return FinishDecode();
+}
+
+void init_source(j_decompress_ptr cinfo) {
+  fill_input_buffer(cinfo);
+}
+
+boolean fill_input_buffer(j_decompress_ptr cinfo) {
+  BufferVector* buf_vec = reinterpret_cast<BufferVector*>(cinfo->client_data);
+  if (buf_vec->pos >= buf_vec->len) {
+    // ERROR: No more data
+    return FALSE;
+  }
+  cinfo->src->next_input_byte = buf_vec->buffers[buf_vec->pos].data;
+  cinfo->src->bytes_in_buffer = buf_vec->buffers[buf_vec->pos].len;
+  ++buf_vec->pos;
+  return TRUE;
+}
+
+void skip_input_data(j_decompress_ptr cinfo, long num_bytes) {  // NOLINT
+  jpeg_source_mgr* src = cinfo->src;
+  size_t bytes = (size_t)num_bytes;
+  if (bytes > src->bytes_in_buffer) {
+    src->next_input_byte = nullptr;
+    src->bytes_in_buffer = 0;
+  } else {
+    src->next_input_byte += bytes;
+    src->bytes_in_buffer -= bytes;
+  }
+}
+
+void term_source(j_decompress_ptr cinfo) {
+  (void)cinfo;  // Nothing to do.
+}
+
+#ifdef HAVE_SETJMP
+void ErrorHandler(j_common_ptr cinfo) {
+// This is called when a jpeglib command experiences an error. Unfortunately
+// jpeglib's error handling model is not very flexible, because it expects the
+// error handler to not return--i.e., it wants the program to terminate. To
+// recover from errors we use setjmp() as shown in their example. setjmp() is
+// C's implementation for the "call with current continuation" functionality
+// seen in some functional programming languages.
+// A formatted message can be output, but is unsafe for release.
+#ifdef DEBUG
+  char buf[JMSG_LENGTH_MAX];
+  (*cinfo->err->format_message)(cinfo, buf);
+// ERROR: Error in jpeglib: buf
+#endif
+
+  SetJmpErrorMgr* mgr = reinterpret_cast<SetJmpErrorMgr*>(cinfo->err);
+  // This rewinds the call stack to the point of the corresponding setjmp()
+  // and causes it to return (for a second time) with value 1.
+  longjmp(mgr->setjmp_buffer, 1);
+}
+
+// Suppress fprintf warnings.
+void OutputHandler(j_common_ptr cinfo) {
+  (void)cinfo;
+}
+
+#endif  // HAVE_SETJMP
+
+void MJpegDecoder::AllocOutputBuffers(int num_outbufs) {
+  if (num_outbufs != num_outbufs_) {
+    // We could perhaps optimize this case to resize the output buffers without
+    // necessarily having to delete and recreate each one, but it's not worth
+    // it.
+    DestroyOutputBuffers();
+
+    scanlines_ = new uint8_t**[num_outbufs];
+    scanlines_sizes_ = new int[num_outbufs];
+    databuf_ = new uint8_t*[num_outbufs];
+    databuf_strides_ = new int[num_outbufs];
+
+    for (int i = 0; i < num_outbufs; ++i) {
+      scanlines_[i] = NULL;
+      scanlines_sizes_[i] = 0;
+      databuf_[i] = NULL;
+      databuf_strides_[i] = 0;
+    }
+
+    num_outbufs_ = num_outbufs;
+  }
+}
+
+void MJpegDecoder::DestroyOutputBuffers() {
+  for (int i = 0; i < num_outbufs_; ++i) {
+    delete[] scanlines_[i];
+    delete[] databuf_[i];
+  }
+  delete[] scanlines_;
+  delete[] databuf_;
+  delete[] scanlines_sizes_;
+  delete[] databuf_strides_;
+  scanlines_ = NULL;
+  databuf_ = NULL;
+  scanlines_sizes_ = NULL;
+  databuf_strides_ = NULL;
+  num_outbufs_ = 0;
+}
+
+// JDCT_IFAST and do_block_smoothing improve performance substantially.
+LIBYUV_BOOL MJpegDecoder::StartDecode() {
+  decompress_struct_->raw_data_out = TRUE;
+  decompress_struct_->dct_method = JDCT_IFAST;  // JDCT_ISLOW is default
+  decompress_struct_->dither_mode = JDITHER_NONE;
+  // Not applicable to 'raw':
+  decompress_struct_->do_fancy_upsampling = (boolean)(LIBYUV_FALSE);
+  // Only for buffered mode:
+  decompress_struct_->enable_2pass_quant = (boolean)(LIBYUV_FALSE);
+  // Blocky but fast:
+  decompress_struct_->do_block_smoothing = (boolean)(LIBYUV_FALSE);
+
+  if (!jpeg_start_decompress(decompress_struct_)) {
+    // ERROR: Couldn't start JPEG decompressor";
+    return LIBYUV_FALSE;
+  }
+  return LIBYUV_TRUE;
+}
+
+LIBYUV_BOOL MJpegDecoder::FinishDecode() {
+  // jpeglib considers it an error if we finish without decoding the whole
+  // image, so we call "abort" rather than "finish".
+  jpeg_abort_decompress(decompress_struct_);
+  return LIBYUV_TRUE;
+}
+
+void MJpegDecoder::SetScanlinePointers(uint8_t** data) {
+  for (int i = 0; i < num_outbufs_; ++i) {
+    uint8_t* data_i = data[i];
+    for (int j = 0; j < scanlines_sizes_[i]; ++j) {
+      scanlines_[i][j] = data_i;
+      data_i += GetComponentStride(i);
+    }
+  }
+}
+
+inline LIBYUV_BOOL MJpegDecoder::DecodeImcuRow() {
+  return (unsigned int)(GetImageScanlinesPerImcuRow()) ==
+         jpeg_read_raw_data(decompress_struct_, scanlines_,
+                            GetImageScanlinesPerImcuRow());
+}
+
+// The helper function which recognizes the jpeg sub-sampling type.
+JpegSubsamplingType MJpegDecoder::JpegSubsamplingTypeHelper(
+    int* subsample_x,
+    int* subsample_y,
+    int number_of_components) {
+  if (number_of_components == 3) {  // Color images.
+    if (subsample_x[0] == 1 && subsample_y[0] == 1 && subsample_x[1] == 2 &&
+        subsample_y[1] == 2 && subsample_x[2] == 2 && subsample_y[2] == 2) {
+      return kJpegYuv420;
+    }
+    if (subsample_x[0] == 1 && subsample_y[0] == 1 && subsample_x[1] == 2 &&
+        subsample_y[1] == 1 && subsample_x[2] == 2 && subsample_y[2] == 1) {
+      return kJpegYuv422;
+    }
+    if (subsample_x[0] == 1 && subsample_y[0] == 1 && subsample_x[1] == 1 &&
+        subsample_y[1] == 1 && subsample_x[2] == 1 && subsample_y[2] == 1) {
+      return kJpegYuv444;
+    }
+  } else if (number_of_components == 1) {  // Grey-scale images.
+    if (subsample_x[0] == 1 && subsample_y[0] == 1) {
+      return kJpegYuv400;
+    }
+  }
+  return kJpegUnknown;
+}
+
+}  // namespace libyuv
+#endif  // HAVE_JPEG

yuv.mod/libyuv/source/mjpeg_validate.cc

@@ -0,0 +1,71 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/mjpeg_decoder.h"
+
+#include <string.h>  // For memchr.
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Helper function to scan for EOI marker (0xff 0xd9).
+static LIBYUV_BOOL ScanEOI(const uint8_t* src_mjpg, size_t src_size_mjpg) {
+  if (src_size_mjpg >= 2) {
+    const uint8_t* end = src_mjpg + src_size_mjpg - 1;
+    const uint8_t* it = src_mjpg;
+    while (it < end) {
+      // TODO(fbarchard): scan for 0xd9 instead.
+      it = (const uint8_t*)(memchr(it, 0xff, end - it));
+      if (it == NULL) {
+        break;
+      }
+      if (it[1] == 0xd9) {
+        return LIBYUV_TRUE;  // Success: Valid jpeg.
+      }
+      ++it;  // Skip over current 0xff.
+    }
+  }
+  // ERROR: Invalid jpeg end code not found. Size src_size_mjpg
+  return LIBYUV_FALSE;
+}
+
+// Helper function to validate the jpeg appears intact.
+LIBYUV_BOOL ValidateJpeg(const uint8_t* src_mjpg, size_t src_size_mjpg) {
+  // Maximum size that ValidateJpeg will consider valid.
+  const size_t kMaxJpegSize = 0x7fffffffull;
+  const size_t kBackSearchSize = 1024;
+  if (src_size_mjpg < 64 || src_size_mjpg > kMaxJpegSize || !src_mjpg) {
+    // ERROR: Invalid jpeg size: src_size_mjpg
+    return LIBYUV_FALSE;
+  }
+  // SOI marker
+  if (src_mjpg[0] != 0xff || src_mjpg[1] != 0xd8 || src_mjpg[2] != 0xff) {
+    // ERROR: Invalid jpeg initial start code
+    return LIBYUV_FALSE;
+  }
+
+  // Look for the End Of Image (EOI) marker near the end of the buffer.
+  if (src_size_mjpg > kBackSearchSize) {
+    if (ScanEOI(src_mjpg + src_size_mjpg - kBackSearchSize, kBackSearchSize)) {
+      return LIBYUV_TRUE;  // Success: Valid jpeg.
+    }
+    // Reduce search size for forward search.
+    src_size_mjpg = src_size_mjpg - kBackSearchSize + 1;
+  }
+  // Step over SOI marker and scan for EOI.
+  return ScanEOI(src_mjpg + 2, src_size_mjpg - 2);
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/planar_functions.cc

@@ -0,0 +1,5756 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/planar_functions.h"
+
+#include <assert.h>
+#include <string.h>  // for memset()
+
+#include "libyuv/cpu_id.h"
+#include "libyuv/row.h"
+#include "libyuv/scale_row.h"  // for ScaleRowDown2
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// Copy a plane of data
+LIBYUV_API
+void CopyPlane(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               int width,
+               int height) {
+  int y;
+  void (*CopyRow)(const uint8_t* src, uint8_t* dst, int width) = CopyRow_C;
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_y = dst_y + (height - 1) * dst_stride_y;
+    dst_stride_y = -dst_stride_y;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && dst_stride_y == width) {
+    width *= height;
+    height = 1;
+    src_stride_y = dst_stride_y = 0;
+  }
+  // Nothing to do.
+  if (src_y == dst_y && src_stride_y == dst_stride_y) {
+    return;
+  }
+
+#if defined(HAS_COPYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    CopyRow = IS_ALIGNED(width, 32) ? CopyRow_SSE2 : CopyRow_Any_SSE2;
+  }
+#endif
+#if defined(HAS_COPYROW_AVX)
+  if (TestCpuFlag(kCpuHasAVX)) {
+    CopyRow = IS_ALIGNED(width, 64) ? CopyRow_AVX : CopyRow_Any_AVX;
+  }
+#endif
+#if defined(HAS_COPYROW_ERMS)
+  if (TestCpuFlag(kCpuHasERMS)) {
+    CopyRow = CopyRow_ERMS;
+  }
+#endif
+#if defined(HAS_COPYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    CopyRow = IS_ALIGNED(width, 32) ? CopyRow_NEON : CopyRow_Any_NEON;
+  }
+#endif
+#if defined(HAS_COPYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    CopyRow = CopyRow_RVV;
+  }
+#endif
+
+  // Copy plane
+  for (y = 0; y < height; ++y) {
+    CopyRow(src_y, dst_y, width);
+    src_y += src_stride_y;
+    dst_y += dst_stride_y;
+  }
+}
+
+LIBYUV_API
+void CopyPlane_16(const uint16_t* src_y,
+                  int src_stride_y,
+                  uint16_t* dst_y,
+                  int dst_stride_y,
+                  int width,
+                  int height) {
+  CopyPlane((const uint8_t*)src_y, src_stride_y * 2, (uint8_t*)dst_y,
+            dst_stride_y * 2, width * 2, height);
+}
+
+// Convert a plane of 16 bit data to 8 bit
+LIBYUV_API
+void Convert16To8Plane(const uint16_t* src_y,
+                       int src_stride_y,
+                       uint8_t* dst_y,
+                       int dst_stride_y,
+                       int scale,  // 16384 for 10 bits
+                       int width,
+                       int height) {
+  int y;
+  void (*Convert16To8Row)(const uint16_t* src_y, uint8_t* dst_y, int scale,
+                          int width) = Convert16To8Row_C;
+
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_y = dst_y + (height - 1) * dst_stride_y;
+    dst_stride_y = -dst_stride_y;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && dst_stride_y == width) {
+    width *= height;
+    height = 1;
+    src_stride_y = dst_stride_y = 0;
+  }
+#if defined(HAS_CONVERT16TO8ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    Convert16To8Row = Convert16To8Row_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      Convert16To8Row = Convert16To8Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_CONVERT16TO8ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    Convert16To8Row = Convert16To8Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      Convert16To8Row = Convert16To8Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_CONVERT16TO8ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    Convert16To8Row = Convert16To8Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      Convert16To8Row = Convert16To8Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_CONVERT16TO8ROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW)) {
+    Convert16To8Row = Convert16To8Row_Any_AVX512BW;
+    if (IS_ALIGNED(width, 64)) {
+      Convert16To8Row = Convert16To8Row_AVX512BW;
+    }
+  }
+#endif
+
+  // Convert plane
+  for (y = 0; y < height; ++y) {
+    Convert16To8Row(src_y, dst_y, scale, width);
+    src_y += src_stride_y;
+    dst_y += dst_stride_y;
+  }
+}
+
+// Convert a plane of 8 bit data to 16 bit
+LIBYUV_API
+void Convert8To16Plane(const uint8_t* src_y,
+                       int src_stride_y,
+                       uint16_t* dst_y,
+                       int dst_stride_y,
+                       int scale,  // 1024 for 10 bits
+                       int width,
+                       int height) {
+  int y;
+  void (*Convert8To16Row)(const uint8_t* src_y, uint16_t* dst_y, int scale,
+                          int width) = Convert8To16Row_C;
+
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_y = dst_y + (height - 1) * dst_stride_y;
+    dst_stride_y = -dst_stride_y;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && dst_stride_y == width) {
+    width *= height;
+    height = 1;
+    src_stride_y = dst_stride_y = 0;
+  }
+#if defined(HAS_CONVERT8TO16ROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    Convert8To16Row = Convert8To16Row_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      Convert8To16Row = Convert8To16Row_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_CONVERT8TO16ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    Convert8To16Row = Convert8To16Row_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      Convert8To16Row = Convert8To16Row_AVX2;
+    }
+  }
+#endif
+
+  // Convert plane
+  for (y = 0; y < height; ++y) {
+    Convert8To16Row(src_y, dst_y, scale, width);
+    src_y += src_stride_y;
+    dst_y += dst_stride_y;
+  }
+}
+
+// Copy I422.
+LIBYUV_API
+int I422Copy(const uint8_t* src_y,
+             int src_stride_y,
+             const uint8_t* src_u,
+             int src_stride_u,
+             const uint8_t* src_v,
+             int src_stride_v,
+             uint8_t* dst_y,
+             int dst_stride_y,
+             uint8_t* dst_u,
+             int dst_stride_u,
+             uint8_t* dst_v,
+             int dst_stride_v,
+             int width,
+             int height) {
+  int halfwidth = (width + 1) >> 1;
+
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  if (dst_y) {
+    CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+  CopyPlane(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth, height);
+  CopyPlane(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth, height);
+  return 0;
+}
+
+// Copy I444.
+LIBYUV_API
+int I444Copy(const uint8_t* src_y,
+             int src_stride_y,
+             const uint8_t* src_u,
+             int src_stride_u,
+             const uint8_t* src_v,
+             int src_stride_v,
+             uint8_t* dst_y,
+             int dst_stride_y,
+             uint8_t* dst_u,
+             int dst_stride_u,
+             uint8_t* dst_v,
+             int dst_stride_v,
+             int width,
+             int height) {
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  if (dst_y) {
+    CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+  CopyPlane(src_u, src_stride_u, dst_u, dst_stride_u, width, height);
+  CopyPlane(src_v, src_stride_v, dst_v, dst_stride_v, width, height);
+  return 0;
+}
+
+// Copy I210.
+LIBYUV_API
+int I210Copy(const uint16_t* src_y,
+             int src_stride_y,
+             const uint16_t* src_u,
+             int src_stride_u,
+             const uint16_t* src_v,
+             int src_stride_v,
+             uint16_t* dst_y,
+             int dst_stride_y,
+             uint16_t* dst_u,
+             int dst_stride_u,
+             uint16_t* dst_v,
+             int dst_stride_v,
+             int width,
+             int height) {
+  int halfwidth = (width + 1) >> 1;
+
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  if (dst_y) {
+    CopyPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+  // Copy UV planes.
+  CopyPlane_16(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth, height);
+  CopyPlane_16(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth, height);
+  return 0;
+}
+
+// Copy I410.
+LIBYUV_API
+int I410Copy(const uint16_t* src_y,
+             int src_stride_y,
+             const uint16_t* src_u,
+             int src_stride_u,
+             const uint16_t* src_v,
+             int src_stride_v,
+             uint16_t* dst_y,
+             int dst_stride_y,
+             uint16_t* dst_u,
+             int dst_stride_u,
+             uint16_t* dst_v,
+             int dst_stride_v,
+             int width,
+             int height) {
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  if (dst_y) {
+    CopyPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+  CopyPlane_16(src_u, src_stride_u, dst_u, dst_stride_u, width, height);
+  CopyPlane_16(src_v, src_stride_v, dst_v, dst_stride_v, width, height);
+  return 0;
+}
+
+// Copy I400.
+LIBYUV_API
+int I400ToI400(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               int width,
+               int height) {
+  if (!src_y || !dst_y || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_stride_y = -src_stride_y;
+  }
+  CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  return 0;
+}
+
+// Convert I420 to I400.
+LIBYUV_API
+int I420ToI400(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               int width,
+               int height) {
+  (void)src_u;
+  (void)src_stride_u;
+  (void)src_v;
+  (void)src_stride_v;
+  if (!src_y || !dst_y || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_stride_y = -src_stride_y;
+  }
+
+  CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  return 0;
+}
+
+// Copy NV12. Supports inverting.
+LIBYUV_API
+int NV12Copy(const uint8_t* src_y,
+             int src_stride_y,
+             const uint8_t* src_uv,
+             int src_stride_uv,
+             uint8_t* dst_y,
+             int dst_stride_y,
+             uint8_t* dst_uv,
+             int dst_stride_uv,
+             int width,
+             int height) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+
+  if (!src_y || !dst_y || !src_uv || !dst_uv || width <= 0 || height == 0) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_uv = src_uv + (halfheight - 1) * src_stride_uv;
+    src_stride_y = -src_stride_y;
+    src_stride_uv = -src_stride_uv;
+  }
+  CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  CopyPlane(src_uv, src_stride_uv, dst_uv, dst_stride_uv, halfwidth * 2,
+            halfheight);
+  return 0;
+}
+
+// Copy NV21. Supports inverting.
+LIBYUV_API
+int NV21Copy(const uint8_t* src_y,
+             int src_stride_y,
+             const uint8_t* src_vu,
+             int src_stride_vu,
+             uint8_t* dst_y,
+             int dst_stride_y,
+             uint8_t* dst_vu,
+             int dst_stride_vu,
+             int width,
+             int height) {
+  return NV12Copy(src_y, src_stride_y, src_vu, src_stride_vu, dst_y,
+                  dst_stride_y, dst_vu, dst_stride_vu, width, height);
+}
+
+// Support function for NV12 etc UV channels.
+// Width and height are plane sizes (typically half pixel width).
+LIBYUV_API
+void SplitUVPlane(const uint8_t* src_uv,
+                  int src_stride_uv,
+                  uint8_t* dst_u,
+                  int dst_stride_u,
+                  uint8_t* dst_v,
+                  int dst_stride_v,
+                  int width,
+                  int height) {
+  int y;
+  void (*SplitUVRow)(const uint8_t* src_uv, uint8_t* dst_u, uint8_t* dst_v,
+                     int width) = SplitUVRow_C;
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_u = dst_u + (height - 1) * dst_stride_u;
+    dst_v = dst_v + (height - 1) * dst_stride_v;
+    dst_stride_u = -dst_stride_u;
+    dst_stride_v = -dst_stride_v;
+  }
+  // Coalesce rows.
+  if (src_stride_uv == width * 2 && dst_stride_u == width &&
+      dst_stride_v == width) {
+    width *= height;
+    height = 1;
+    src_stride_uv = dst_stride_u = dst_stride_v = 0;
+  }
+#if defined(HAS_SPLITUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    SplitUVRow = SplitUVRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      SplitUVRow = SplitUVRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_SPLITUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    SplitUVRow = SplitUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      SplitUVRow = SplitUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_SPLITUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SplitUVRow = SplitUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      SplitUVRow = SplitUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_SPLITUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    SplitUVRow = SplitUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      SplitUVRow = SplitUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_SPLITUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    SplitUVRow = SplitUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 32)) {
+      SplitUVRow = SplitUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_SPLITUVROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    SplitUVRow = SplitUVRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    // Copy a row of UV.
+    SplitUVRow(src_uv, dst_u, dst_v, width);
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+    src_uv += src_stride_uv;
+  }
+}
+
+LIBYUV_API
+void MergeUVPlane(const uint8_t* src_u,
+                  int src_stride_u,
+                  const uint8_t* src_v,
+                  int src_stride_v,
+                  uint8_t* dst_uv,
+                  int dst_stride_uv,
+                  int width,
+                  int height) {
+  int y;
+  void (*MergeUVRow)(const uint8_t* src_u, const uint8_t* src_v,
+                     uint8_t* dst_uv, int width) = MergeUVRow_C;
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_uv = dst_uv + (height - 1) * dst_stride_uv;
+    dst_stride_uv = -dst_stride_uv;
+  }
+  // Coalesce rows.
+  if (src_stride_u == width && src_stride_v == width &&
+      dst_stride_uv == width * 2) {
+    width *= height;
+    height = 1;
+    src_stride_u = src_stride_v = dst_stride_uv = 0;
+  }
+#if defined(HAS_MERGEUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    MergeUVRow = MergeUVRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      MergeUVRow = MergeUVRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeUVRow = MergeUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      MergeUVRow = MergeUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_AVX512BW)
+  if (TestCpuFlag(kCpuHasAVX512BW)) {
+    MergeUVRow = MergeUVRow_Any_AVX512BW;
+    if (IS_ALIGNED(width, 32)) {
+      MergeUVRow = MergeUVRow_AVX512BW;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeUVRow = MergeUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      MergeUVRow = MergeUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    MergeUVRow = MergeUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      MergeUVRow = MergeUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    MergeUVRow = MergeUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      MergeUVRow = MergeUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    MergeUVRow = MergeUVRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    // Merge a row of U and V into a row of UV.
+    MergeUVRow(src_u, src_v, dst_uv, width);
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+    dst_uv += dst_stride_uv;
+  }
+}
+
+// Support function for P010 etc UV channels.
+// Width and height are plane sizes (typically half pixel width).
+LIBYUV_API
+void SplitUVPlane_16(const uint16_t* src_uv,
+                     int src_stride_uv,
+                     uint16_t* dst_u,
+                     int dst_stride_u,
+                     uint16_t* dst_v,
+                     int dst_stride_v,
+                     int width,
+                     int height,
+                     int depth) {
+  int y;
+  void (*SplitUVRow_16)(const uint16_t* src_uv, uint16_t* dst_u,
+                        uint16_t* dst_v, int depth, int width) =
+      SplitUVRow_16_C;
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_u = dst_u + (height - 1) * dst_stride_u;
+    dst_v = dst_v + (height - 1) * dst_stride_v;
+    dst_stride_u = -dst_stride_u;
+    dst_stride_v = -dst_stride_v;
+  }
+  // Coalesce rows.
+  if (src_stride_uv == width * 2 && dst_stride_u == width &&
+      dst_stride_v == width) {
+    width *= height;
+    height = 1;
+    src_stride_uv = dst_stride_u = dst_stride_v = 0;
+  }
+#if defined(HAS_SPLITUVROW_16_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    SplitUVRow_16 = SplitUVRow_16_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      SplitUVRow_16 = SplitUVRow_16_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_SPLITUVROW_16_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SplitUVRow_16 = SplitUVRow_16_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      SplitUVRow_16 = SplitUVRow_16_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    // Copy a row of UV.
+    SplitUVRow_16(src_uv, dst_u, dst_v, depth, width);
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+    src_uv += src_stride_uv;
+  }
+}
+
+LIBYUV_API
+void MergeUVPlane_16(const uint16_t* src_u,
+                     int src_stride_u,
+                     const uint16_t* src_v,
+                     int src_stride_v,
+                     uint16_t* dst_uv,
+                     int dst_stride_uv,
+                     int width,
+                     int height,
+                     int depth) {
+  int y;
+  void (*MergeUVRow_16)(const uint16_t* src_u, const uint16_t* src_v,
+                        uint16_t* dst_uv, int depth, int width) =
+      MergeUVRow_16_C;
+  assert(depth >= 8);
+  assert(depth <= 16);
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_uv = dst_uv + (height - 1) * dst_stride_uv;
+    dst_stride_uv = -dst_stride_uv;
+  }
+  // Coalesce rows.
+  if (src_stride_u == width && src_stride_v == width &&
+      dst_stride_uv == width * 2) {
+    width *= height;
+    height = 1;
+    src_stride_u = src_stride_v = dst_stride_uv = 0;
+  }
+#if defined(HAS_MERGEUVROW_16_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeUVRow_16 = MergeUVRow_16_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      MergeUVRow_16 = MergeUVRow_16_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEUVROW_16_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeUVRow_16 = MergeUVRow_16_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      MergeUVRow_16 = MergeUVRow_16_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    // Merge a row of U and V into a row of UV.
+    MergeUVRow_16(src_u, src_v, dst_uv, depth, width);
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+    dst_uv += dst_stride_uv;
+  }
+}
+
+// Convert plane from lsb to msb
+LIBYUV_API
+void ConvertToMSBPlane_16(const uint16_t* src_y,
+                          int src_stride_y,
+                          uint16_t* dst_y,
+                          int dst_stride_y,
+                          int width,
+                          int height,
+                          int depth) {
+  int y;
+  int scale = 1 << (16 - depth);
+  void (*MultiplyRow_16)(const uint16_t* src_y, uint16_t* dst_y, int scale,
+                         int width) = MultiplyRow_16_C;
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_y = dst_y + (height - 1) * dst_stride_y;
+    dst_stride_y = -dst_stride_y;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && dst_stride_y == width) {
+    width *= height;
+    height = 1;
+    src_stride_y = dst_stride_y = 0;
+  }
+
+#if defined(HAS_MULTIPLYROW_16_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MultiplyRow_16 = MultiplyRow_16_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      MultiplyRow_16 = MultiplyRow_16_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MULTIPLYROW_16_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MultiplyRow_16 = MultiplyRow_16_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      MultiplyRow_16 = MultiplyRow_16_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    MultiplyRow_16(src_y, dst_y, scale, width);
+    src_y += src_stride_y;
+    dst_y += dst_stride_y;
+  }
+}
+
+// Convert plane from msb to lsb
+LIBYUV_API
+void ConvertToLSBPlane_16(const uint16_t* src_y,
+                          int src_stride_y,
+                          uint16_t* dst_y,
+                          int dst_stride_y,
+                          int width,
+                          int height,
+                          int depth) {
+  int y;
+  int scale = 1 << depth;
+  void (*DivideRow)(const uint16_t* src_y, uint16_t* dst_y, int scale,
+                    int width) = DivideRow_16_C;
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_y = dst_y + (height - 1) * dst_stride_y;
+    dst_stride_y = -dst_stride_y;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && dst_stride_y == width) {
+    width *= height;
+    height = 1;
+    src_stride_y = dst_stride_y = 0;
+  }
+
+#if defined(HAS_DIVIDEROW_16_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    DivideRow = DivideRow_16_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      DivideRow = DivideRow_16_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_DIVIDEROW_16_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    DivideRow = DivideRow_16_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      DivideRow = DivideRow_16_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    DivideRow(src_y, dst_y, scale, width);
+    src_y += src_stride_y;
+    dst_y += dst_stride_y;
+  }
+}
+
+// Swap U and V channels in interleaved UV plane.
+LIBYUV_API
+void SwapUVPlane(const uint8_t* src_uv,
+                 int src_stride_uv,
+                 uint8_t* dst_vu,
+                 int dst_stride_vu,
+                 int width,
+                 int height) {
+  int y;
+  void (*SwapUVRow)(const uint8_t* src_uv, uint8_t* dst_vu, int width) =
+      SwapUVRow_C;
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_uv = src_uv + (height - 1) * src_stride_uv;
+    src_stride_uv = -src_stride_uv;
+  }
+  // Coalesce rows.
+  if (src_stride_uv == width * 2 && dst_stride_vu == width * 2) {
+    width *= height;
+    height = 1;
+    src_stride_uv = dst_stride_vu = 0;
+  }
+
+#if defined(HAS_SWAPUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    SwapUVRow = SwapUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      SwapUVRow = SwapUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_SWAPUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    SwapUVRow = SwapUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      SwapUVRow = SwapUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_SWAPUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SwapUVRow = SwapUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      SwapUVRow = SwapUVRow_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    SwapUVRow(src_uv, dst_vu, width);
+    src_uv += src_stride_uv;
+    dst_vu += dst_stride_vu;
+  }
+}
+
+// Convert NV21 to NV12.
+LIBYUV_API
+int NV21ToNV12(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_vu,
+               int src_stride_vu,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+
+  if (!src_vu || !dst_uv || width <= 0 || height == 0) {
+    return -1;
+  }
+
+  if (dst_y) {
+    CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_vu = src_vu + (halfheight - 1) * src_stride_vu;
+    src_stride_vu = -src_stride_vu;
+  }
+
+  SwapUVPlane(src_vu, src_stride_vu, dst_uv, dst_stride_uv, halfwidth,
+              halfheight);
+  return 0;
+}
+
+// Test if tile_height is a power of 2 (16 or 32)
+#define IS_POWEROFTWO(x) (!((x) & ((x)-1)))
+
+// Detile a plane of data
+// tile width is 16 and assumed.
+// tile_height is 16 or 32 for MM21.
+// src_stride_y is bytes per row of source ignoring tiling. e.g. 640
+// TODO: More detile row functions.
+LIBYUV_API
+int DetilePlane(const uint8_t* src_y,
+                int src_stride_y,
+                uint8_t* dst_y,
+                int dst_stride_y,
+                int width,
+                int height,
+                int tile_height) {
+  const ptrdiff_t src_tile_stride = 16 * tile_height;
+  int y;
+  void (*DetileRow)(const uint8_t* src, ptrdiff_t src_tile_stride, uint8_t* dst,
+                    int width) = DetileRow_C;
+  if (!src_y || !dst_y || width <= 0 || height == 0 ||
+      !IS_POWEROFTWO(tile_height)) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_y = dst_y + (height - 1) * dst_stride_y;
+    dst_stride_y = -dst_stride_y;
+  }
+
+#if defined(HAS_DETILEROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    DetileRow = DetileRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      DetileRow = DetileRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_DETILEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    DetileRow = DetileRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      DetileRow = DetileRow_NEON;
+    }
+  }
+#endif
+
+  // Detile plane
+  for (y = 0; y < height; ++y) {
+    DetileRow(src_y, src_tile_stride, dst_y, width);
+    dst_y += dst_stride_y;
+    src_y += 16;
+    // Advance to next row of tiles.
+    if ((y & (tile_height - 1)) == (tile_height - 1)) {
+      src_y = src_y - src_tile_stride + src_stride_y * tile_height;
+    }
+  }
+  return 0;
+}
+
+// Convert a plane of 16 bit tiles of 16 x H to linear.
+// tile width is 16 and assumed.
+// tile_height is 16 or 32 for MT2T.
+LIBYUV_API
+int DetilePlane_16(const uint16_t* src_y,
+                   int src_stride_y,
+                   uint16_t* dst_y,
+                   int dst_stride_y,
+                   int width,
+                   int height,
+                   int tile_height) {
+  const ptrdiff_t src_tile_stride = 16 * tile_height;
+  int y;
+  void (*DetileRow_16)(const uint16_t* src, ptrdiff_t src_tile_stride,
+                       uint16_t* dst, int width) = DetileRow_16_C;
+  if (!src_y || !dst_y || width <= 0 || height == 0 ||
+      !IS_POWEROFTWO(tile_height)) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_y = dst_y + (height - 1) * dst_stride_y;
+    dst_stride_y = -dst_stride_y;
+  }
+
+#if defined(HAS_DETILEROW_16_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    DetileRow_16 = DetileRow_16_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      DetileRow_16 = DetileRow_16_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_DETILEROW_16_AVX)
+  if (TestCpuFlag(kCpuHasAVX)) {
+    DetileRow_16 = DetileRow_16_Any_AVX;
+    if (IS_ALIGNED(width, 16)) {
+      DetileRow_16 = DetileRow_16_AVX;
+    }
+  }
+#endif
+#if defined(HAS_DETILEROW_16_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    DetileRow_16 = DetileRow_16_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      DetileRow_16 = DetileRow_16_NEON;
+    }
+  }
+#endif
+
+  // Detile plane
+  for (y = 0; y < height; ++y) {
+    DetileRow_16(src_y, src_tile_stride, dst_y, width);
+    dst_y += dst_stride_y;
+    src_y += 16;
+    // Advance to next row of tiles.
+    if ((y & (tile_height - 1)) == (tile_height - 1)) {
+      src_y = src_y - src_tile_stride + src_stride_y * tile_height;
+    }
+  }
+  return 0;
+}
+
+LIBYUV_API
+void DetileSplitUVPlane(const uint8_t* src_uv,
+                        int src_stride_uv,
+                        uint8_t* dst_u,
+                        int dst_stride_u,
+                        uint8_t* dst_v,
+                        int dst_stride_v,
+                        int width,
+                        int height,
+                        int tile_height) {
+  const ptrdiff_t src_tile_stride = 16 * tile_height;
+  int y;
+  void (*DetileSplitUVRow)(const uint8_t* src, ptrdiff_t src_tile_stride,
+                           uint8_t* dst_u, uint8_t* dst_v, int width) =
+      DetileSplitUVRow_C;
+  assert(src_stride_uv >= 0);
+  assert(tile_height > 0);
+  assert(src_stride_uv > 0);
+
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_u = dst_u + (height - 1) * dst_stride_u;
+    dst_stride_u = -dst_stride_u;
+    dst_v = dst_v + (height - 1) * dst_stride_v;
+    dst_stride_v = -dst_stride_v;
+  }
+
+#if defined(HAS_DETILESPLITUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    DetileSplitUVRow = DetileSplitUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      DetileSplitUVRow = DetileSplitUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_DETILESPLITUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    DetileSplitUVRow = DetileSplitUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      DetileSplitUVRow = DetileSplitUVRow_NEON;
+    }
+  }
+#endif
+
+  // Detile plane
+  for (y = 0; y < height; ++y) {
+    DetileSplitUVRow(src_uv, src_tile_stride, dst_u, dst_v, width);
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+    src_uv += 16;
+    // Advance to next row of tiles.
+    if ((y & (tile_height - 1)) == (tile_height - 1)) {
+      src_uv = src_uv - src_tile_stride + src_stride_uv * tile_height;
+    }
+  }
+}
+
+LIBYUV_API
+void DetileToYUY2(const uint8_t* src_y,
+                  int src_stride_y,
+                  const uint8_t* src_uv,
+                  int src_stride_uv,
+                  uint8_t* dst_yuy2,
+                  int dst_stride_yuy2,
+                  int width,
+                  int height,
+                  int tile_height) {
+  const ptrdiff_t src_y_tile_stride = 16 * tile_height;
+  const ptrdiff_t src_uv_tile_stride = src_y_tile_stride / 2;
+  int y;
+  void (*DetileToYUY2)(const uint8_t* src_y, ptrdiff_t src_y_tile_stride,
+                       const uint8_t* src_uv, ptrdiff_t src_uv_tile_stride,
+                       uint8_t* dst_yuy2, int width) = DetileToYUY2_C;
+  assert(src_stride_y >= 0);
+  assert(src_stride_y > 0);
+  assert(src_stride_uv >= 0);
+  assert(src_stride_uv > 0);
+  assert(tile_height > 0);
+
+  if (width <= 0 || height == 0 || tile_height <= 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_yuy2 = dst_yuy2 + (height - 1) * dst_stride_yuy2;
+    dst_stride_yuy2 = -dst_stride_yuy2;
+  }
+
+#if defined(HAS_DETILETOYUY2_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    DetileToYUY2 = DetileToYUY2_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      DetileToYUY2 = DetileToYUY2_NEON;
+    }
+  }
+#endif
+
+#if defined(HAS_DETILETOYUY2_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    DetileToYUY2 = DetileToYUY2_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      DetileToYUY2 = DetileToYUY2_SSE2;
+    }
+  }
+#endif
+
+  // Detile plane
+  for (y = 0; y < height; ++y) {
+    DetileToYUY2(src_y, src_y_tile_stride, src_uv, src_uv_tile_stride, dst_yuy2,
+                 width);
+    dst_yuy2 += dst_stride_yuy2;
+    src_y += 16;
+
+    if (y & 0x1)
+      src_uv += 16;
+
+    // Advance to next row of tiles.
+    if ((y & (tile_height - 1)) == (tile_height - 1)) {
+      src_y = src_y - src_y_tile_stride + src_stride_y * tile_height;
+      src_uv = src_uv - src_uv_tile_stride + src_stride_uv * (tile_height / 2);
+    }
+  }
+}
+
+// Support function for NV12 etc RGB channels.
+// Width and height are plane sizes (typically half pixel width).
+LIBYUV_API
+void SplitRGBPlane(const uint8_t* src_rgb,
+                   int src_stride_rgb,
+                   uint8_t* dst_r,
+                   int dst_stride_r,
+                   uint8_t* dst_g,
+                   int dst_stride_g,
+                   uint8_t* dst_b,
+                   int dst_stride_b,
+                   int width,
+                   int height) {
+  int y;
+  void (*SplitRGBRow)(const uint8_t* src_rgb, uint8_t* dst_r, uint8_t* dst_g,
+                      uint8_t* dst_b, int width) = SplitRGBRow_C;
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_r = dst_r + (height - 1) * dst_stride_r;
+    dst_g = dst_g + (height - 1) * dst_stride_g;
+    dst_b = dst_b + (height - 1) * dst_stride_b;
+    dst_stride_r = -dst_stride_r;
+    dst_stride_g = -dst_stride_g;
+    dst_stride_b = -dst_stride_b;
+  }
+  // Coalesce rows.
+  if (src_stride_rgb == width * 3 && dst_stride_r == width &&
+      dst_stride_g == width && dst_stride_b == width) {
+    width *= height;
+    height = 1;
+    src_stride_rgb = dst_stride_r = dst_stride_g = dst_stride_b = 0;
+  }
+#if defined(HAS_SPLITRGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    SplitRGBRow = SplitRGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      SplitRGBRow = SplitRGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_SPLITRGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SplitRGBRow = SplitRGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      SplitRGBRow = SplitRGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_SPLITRGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    SplitRGBRow = SplitRGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    // Copy a row of RGB.
+    SplitRGBRow(src_rgb, dst_r, dst_g, dst_b, width);
+    dst_r += dst_stride_r;
+    dst_g += dst_stride_g;
+    dst_b += dst_stride_b;
+    src_rgb += src_stride_rgb;
+  }
+}
+
+LIBYUV_API
+void MergeRGBPlane(const uint8_t* src_r,
+                   int src_stride_r,
+                   const uint8_t* src_g,
+                   int src_stride_g,
+                   const uint8_t* src_b,
+                   int src_stride_b,
+                   uint8_t* dst_rgb,
+                   int dst_stride_rgb,
+                   int width,
+                   int height) {
+  int y;
+  void (*MergeRGBRow)(const uint8_t* src_r, const uint8_t* src_g,
+                      const uint8_t* src_b, uint8_t* dst_rgb, int width) =
+      MergeRGBRow_C;
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  // Coalesce rows.
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb = dst_rgb + (height - 1) * dst_stride_rgb;
+    dst_stride_rgb = -dst_stride_rgb;
+  }
+  // Coalesce rows.
+  if (src_stride_r == width && src_stride_g == width && src_stride_b == width &&
+      dst_stride_rgb == width * 3) {
+    width *= height;
+    height = 1;
+    src_stride_r = src_stride_g = src_stride_b = dst_stride_rgb = 0;
+  }
+#if defined(HAS_MERGERGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    MergeRGBRow = MergeRGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      MergeRGBRow = MergeRGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_MERGERGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeRGBRow = MergeRGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      MergeRGBRow = MergeRGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MERGERGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    MergeRGBRow = MergeRGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    // Merge a row of U and V into a row of RGB.
+    MergeRGBRow(src_r, src_g, src_b, dst_rgb, width);
+    src_r += src_stride_r;
+    src_g += src_stride_g;
+    src_b += src_stride_b;
+    dst_rgb += dst_stride_rgb;
+  }
+}
+
+LIBYUV_NOINLINE
+static void SplitARGBPlaneAlpha(const uint8_t* src_argb,
+                                int src_stride_argb,
+                                uint8_t* dst_r,
+                                int dst_stride_r,
+                                uint8_t* dst_g,
+                                int dst_stride_g,
+                                uint8_t* dst_b,
+                                int dst_stride_b,
+                                uint8_t* dst_a,
+                                int dst_stride_a,
+                                int width,
+                                int height) {
+  int y;
+  void (*SplitARGBRow)(const uint8_t* src_rgb, uint8_t* dst_r, uint8_t* dst_g,
+                       uint8_t* dst_b, uint8_t* dst_a, int width) =
+      SplitARGBRow_C;
+
+  assert(height > 0);
+
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  if (src_stride_argb == width * 4 && dst_stride_r == width &&
+      dst_stride_g == width && dst_stride_b == width && dst_stride_a == width) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_r = dst_stride_g = dst_stride_b =
+        dst_stride_a = 0;
+  }
+
+#if defined(HAS_SPLITARGBROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    SplitARGBRow = SplitARGBRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      SplitARGBRow = SplitARGBRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_SPLITARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    SplitARGBRow = SplitARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      SplitARGBRow = SplitARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_SPLITARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    SplitARGBRow = SplitARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      SplitARGBRow = SplitARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_SPLITARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SplitARGBRow = SplitARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      SplitARGBRow = SplitARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_SPLITARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    SplitARGBRow = SplitARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    SplitARGBRow(src_argb, dst_r, dst_g, dst_b, dst_a, width);
+    dst_r += dst_stride_r;
+    dst_g += dst_stride_g;
+    dst_b += dst_stride_b;
+    dst_a += dst_stride_a;
+    src_argb += src_stride_argb;
+  }
+}
+
+LIBYUV_NOINLINE
+static void SplitARGBPlaneOpaque(const uint8_t* src_argb,
+                                 int src_stride_argb,
+                                 uint8_t* dst_r,
+                                 int dst_stride_r,
+                                 uint8_t* dst_g,
+                                 int dst_stride_g,
+                                 uint8_t* dst_b,
+                                 int dst_stride_b,
+                                 int width,
+                                 int height) {
+  int y;
+  void (*SplitXRGBRow)(const uint8_t* src_rgb, uint8_t* dst_r, uint8_t* dst_g,
+                       uint8_t* dst_b, int width) = SplitXRGBRow_C;
+  assert(height > 0);
+
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  if (src_stride_argb == width * 4 && dst_stride_r == width &&
+      dst_stride_g == width && dst_stride_b == width) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_r = dst_stride_g = dst_stride_b = 0;
+  }
+
+#if defined(HAS_SPLITXRGBROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    SplitXRGBRow = SplitXRGBRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      SplitXRGBRow = SplitXRGBRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_SPLITXRGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    SplitXRGBRow = SplitXRGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      SplitXRGBRow = SplitXRGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_SPLITXRGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    SplitXRGBRow = SplitXRGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      SplitXRGBRow = SplitXRGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_SPLITXRGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SplitXRGBRow = SplitXRGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      SplitXRGBRow = SplitXRGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_SPLITXRGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    SplitXRGBRow = SplitXRGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    SplitXRGBRow(src_argb, dst_r, dst_g, dst_b, width);
+    dst_r += dst_stride_r;
+    dst_g += dst_stride_g;
+    dst_b += dst_stride_b;
+    src_argb += src_stride_argb;
+  }
+}
+
+LIBYUV_API
+void SplitARGBPlane(const uint8_t* src_argb,
+                    int src_stride_argb,
+                    uint8_t* dst_r,
+                    int dst_stride_r,
+                    uint8_t* dst_g,
+                    int dst_stride_g,
+                    uint8_t* dst_b,
+                    int dst_stride_b,
+                    uint8_t* dst_a,
+                    int dst_stride_a,
+                    int width,
+                    int height) {
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_r = dst_r + (height - 1) * dst_stride_r;
+    dst_g = dst_g + (height - 1) * dst_stride_g;
+    dst_b = dst_b + (height - 1) * dst_stride_b;
+    dst_a = dst_a + (height - 1) * dst_stride_a;
+    dst_stride_r = -dst_stride_r;
+    dst_stride_g = -dst_stride_g;
+    dst_stride_b = -dst_stride_b;
+    dst_stride_a = -dst_stride_a;
+  }
+
+  if (dst_a == NULL) {
+    SplitARGBPlaneOpaque(src_argb, src_stride_argb, dst_r, dst_stride_r, dst_g,
+                         dst_stride_g, dst_b, dst_stride_b, width, height);
+  } else {
+    SplitARGBPlaneAlpha(src_argb, src_stride_argb, dst_r, dst_stride_r, dst_g,
+                        dst_stride_g, dst_b, dst_stride_b, dst_a, dst_stride_a,
+                        width, height);
+  }
+}
+
+LIBYUV_NOINLINE
+static void MergeARGBPlaneAlpha(const uint8_t* src_r,
+                                int src_stride_r,
+                                const uint8_t* src_g,
+                                int src_stride_g,
+                                const uint8_t* src_b,
+                                int src_stride_b,
+                                const uint8_t* src_a,
+                                int src_stride_a,
+                                uint8_t* dst_argb,
+                                int dst_stride_argb,
+                                int width,
+                                int height) {
+  int y;
+  void (*MergeARGBRow)(const uint8_t* src_r, const uint8_t* src_g,
+                       const uint8_t* src_b, const uint8_t* src_a,
+                       uint8_t* dst_argb, int width) = MergeARGBRow_C;
+
+  assert(height > 0);
+
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  if (src_stride_r == width && src_stride_g == width && src_stride_b == width &&
+      src_stride_a == width && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_r = src_stride_g = src_stride_b = src_stride_a =
+        dst_stride_argb = 0;
+  }
+#if defined(HAS_MERGEARGBROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    MergeARGBRow = MergeARGBRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      MergeARGBRow = MergeARGBRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeARGBRow = MergeARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      MergeARGBRow = MergeARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeARGBRow = MergeARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      MergeARGBRow = MergeARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MERGEARGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    MergeARGBRow = MergeARGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    MergeARGBRow(src_r, src_g, src_b, src_a, dst_argb, width);
+    src_r += src_stride_r;
+    src_g += src_stride_g;
+    src_b += src_stride_b;
+    src_a += src_stride_a;
+    dst_argb += dst_stride_argb;
+  }
+}
+
+LIBYUV_NOINLINE
+static void MergeARGBPlaneOpaque(const uint8_t* src_r,
+                                 int src_stride_r,
+                                 const uint8_t* src_g,
+                                 int src_stride_g,
+                                 const uint8_t* src_b,
+                                 int src_stride_b,
+                                 uint8_t* dst_argb,
+                                 int dst_stride_argb,
+                                 int width,
+                                 int height) {
+  int y;
+  void (*MergeXRGBRow)(const uint8_t* src_r, const uint8_t* src_g,
+                       const uint8_t* src_b, uint8_t* dst_argb, int width) =
+      MergeXRGBRow_C;
+
+  assert(height > 0);
+
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  if (src_stride_r == width && src_stride_g == width && src_stride_b == width &&
+      dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_r = src_stride_g = src_stride_b = dst_stride_argb = 0;
+  }
+#if defined(HAS_MERGEXRGBROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    MergeXRGBRow = MergeXRGBRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      MergeXRGBRow = MergeXRGBRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEXRGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeXRGBRow = MergeXRGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      MergeXRGBRow = MergeXRGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEXRGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeXRGBRow = MergeXRGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      MergeXRGBRow = MergeXRGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MERGEXRGBROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    MergeXRGBRow = MergeXRGBRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    MergeXRGBRow(src_r, src_g, src_b, dst_argb, width);
+    src_r += src_stride_r;
+    src_g += src_stride_g;
+    src_b += src_stride_b;
+    dst_argb += dst_stride_argb;
+  }
+}
+
+LIBYUV_API
+void MergeARGBPlane(const uint8_t* src_r,
+                    int src_stride_r,
+                    const uint8_t* src_g,
+                    int src_stride_g,
+                    const uint8_t* src_b,
+                    int src_stride_b,
+                    const uint8_t* src_a,
+                    int src_stride_a,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    int width,
+                    int height) {
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+
+  if (src_a == NULL) {
+    MergeARGBPlaneOpaque(src_r, src_stride_r, src_g, src_stride_g, src_b,
+                         src_stride_b, dst_argb, dst_stride_argb, width,
+                         height);
+  } else {
+    MergeARGBPlaneAlpha(src_r, src_stride_r, src_g, src_stride_g, src_b,
+                        src_stride_b, src_a, src_stride_a, dst_argb,
+                        dst_stride_argb, width, height);
+  }
+}
+
+// TODO(yuan): Support 2 bit alpha channel.
+LIBYUV_API
+void MergeXR30Plane(const uint16_t* src_r,
+                    int src_stride_r,
+                    const uint16_t* src_g,
+                    int src_stride_g,
+                    const uint16_t* src_b,
+                    int src_stride_b,
+                    uint8_t* dst_ar30,
+                    int dst_stride_ar30,
+                    int width,
+                    int height,
+                    int depth) {
+  int y;
+  void (*MergeXR30Row)(const uint16_t* src_r, const uint16_t* src_g,
+                       const uint16_t* src_b, uint8_t* dst_ar30, int depth,
+                       int width) = MergeXR30Row_C;
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar30 = dst_ar30 + (height - 1) * dst_stride_ar30;
+    dst_stride_ar30 = -dst_stride_ar30;
+  }
+  // Coalesce rows.
+  if (src_stride_r == width && src_stride_g == width && src_stride_b == width &&
+      dst_stride_ar30 == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_r = src_stride_g = src_stride_b = dst_stride_ar30 = 0;
+  }
+#if defined(HAS_MERGEXR30ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeXR30Row = MergeXR30Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      MergeXR30Row = MergeXR30Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEXR30ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    if (depth == 10) {
+      MergeXR30Row = MergeXR30Row_10_Any_NEON;
+      if (IS_ALIGNED(width, 8)) {
+        MergeXR30Row = MergeXR30Row_10_NEON;
+      }
+    } else {
+      MergeXR30Row = MergeXR30Row_Any_NEON;
+      if (IS_ALIGNED(width, 8)) {
+        MergeXR30Row = MergeXR30Row_NEON;
+      }
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    MergeXR30Row(src_r, src_g, src_b, dst_ar30, depth, width);
+    src_r += src_stride_r;
+    src_g += src_stride_g;
+    src_b += src_stride_b;
+    dst_ar30 += dst_stride_ar30;
+  }
+}
+
+LIBYUV_NOINLINE
+static void MergeAR64PlaneAlpha(const uint16_t* src_r,
+                                int src_stride_r,
+                                const uint16_t* src_g,
+                                int src_stride_g,
+                                const uint16_t* src_b,
+                                int src_stride_b,
+                                const uint16_t* src_a,
+                                int src_stride_a,
+                                uint16_t* dst_ar64,
+                                int dst_stride_ar64,
+                                int width,
+                                int height,
+                                int depth) {
+  int y;
+  void (*MergeAR64Row)(const uint16_t* src_r, const uint16_t* src_g,
+                       const uint16_t* src_b, const uint16_t* src_a,
+                       uint16_t* dst_argb, int depth, int width) =
+      MergeAR64Row_C;
+
+  if (src_stride_r == width && src_stride_g == width && src_stride_b == width &&
+      src_stride_a == width && dst_stride_ar64 == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_r = src_stride_g = src_stride_b = src_stride_a =
+        dst_stride_ar64 = 0;
+  }
+#if defined(HAS_MERGEAR64ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeAR64Row = MergeAR64Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      MergeAR64Row = MergeAR64Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEAR64ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeAR64Row = MergeAR64Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      MergeAR64Row = MergeAR64Row_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    MergeAR64Row(src_r, src_g, src_b, src_a, dst_ar64, depth, width);
+    src_r += src_stride_r;
+    src_g += src_stride_g;
+    src_b += src_stride_b;
+    src_a += src_stride_a;
+    dst_ar64 += dst_stride_ar64;
+  }
+}
+
+LIBYUV_NOINLINE
+static void MergeAR64PlaneOpaque(const uint16_t* src_r,
+                                 int src_stride_r,
+                                 const uint16_t* src_g,
+                                 int src_stride_g,
+                                 const uint16_t* src_b,
+                                 int src_stride_b,
+                                 uint16_t* dst_ar64,
+                                 int dst_stride_ar64,
+                                 int width,
+                                 int height,
+                                 int depth) {
+  int y;
+  void (*MergeXR64Row)(const uint16_t* src_r, const uint16_t* src_g,
+                       const uint16_t* src_b, uint16_t* dst_argb, int depth,
+                       int width) = MergeXR64Row_C;
+
+  // Coalesce rows.
+  if (src_stride_r == width && src_stride_g == width && src_stride_b == width &&
+      dst_stride_ar64 == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_r = src_stride_g = src_stride_b = dst_stride_ar64 = 0;
+  }
+#if defined(HAS_MERGEXR64ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeXR64Row = MergeXR64Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      MergeXR64Row = MergeXR64Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEXR64ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeXR64Row = MergeXR64Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      MergeXR64Row = MergeXR64Row_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    MergeXR64Row(src_r, src_g, src_b, dst_ar64, depth, width);
+    src_r += src_stride_r;
+    src_g += src_stride_g;
+    src_b += src_stride_b;
+    dst_ar64 += dst_stride_ar64;
+  }
+}
+
+LIBYUV_API
+void MergeAR64Plane(const uint16_t* src_r,
+                    int src_stride_r,
+                    const uint16_t* src_g,
+                    int src_stride_g,
+                    const uint16_t* src_b,
+                    int src_stride_b,
+                    const uint16_t* src_a,
+                    int src_stride_a,
+                    uint16_t* dst_ar64,
+                    int dst_stride_ar64,
+                    int width,
+                    int height,
+                    int depth) {
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_ar64 = dst_ar64 + (height - 1) * dst_stride_ar64;
+    dst_stride_ar64 = -dst_stride_ar64;
+  }
+
+  if (src_a == NULL) {
+    MergeAR64PlaneOpaque(src_r, src_stride_r, src_g, src_stride_g, src_b,
+                         src_stride_b, dst_ar64, dst_stride_ar64, width, height,
+                         depth);
+  } else {
+    MergeAR64PlaneAlpha(src_r, src_stride_r, src_g, src_stride_g, src_b,
+                        src_stride_b, src_a, src_stride_a, dst_ar64,
+                        dst_stride_ar64, width, height, depth);
+  }
+}
+
+LIBYUV_NOINLINE
+static void MergeARGB16To8PlaneAlpha(const uint16_t* src_r,
+                                     int src_stride_r,
+                                     const uint16_t* src_g,
+                                     int src_stride_g,
+                                     const uint16_t* src_b,
+                                     int src_stride_b,
+                                     const uint16_t* src_a,
+                                     int src_stride_a,
+                                     uint8_t* dst_argb,
+                                     int dst_stride_argb,
+                                     int width,
+                                     int height,
+                                     int depth) {
+  int y;
+  void (*MergeARGB16To8Row)(const uint16_t* src_r, const uint16_t* src_g,
+                            const uint16_t* src_b, const uint16_t* src_a,
+                            uint8_t* dst_argb, int depth, int width) =
+      MergeARGB16To8Row_C;
+
+  if (src_stride_r == width && src_stride_g == width && src_stride_b == width &&
+      src_stride_a == width && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_r = src_stride_g = src_stride_b = src_stride_a =
+        dst_stride_argb = 0;
+  }
+#if defined(HAS_MERGEARGB16TO8ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeARGB16To8Row = MergeARGB16To8Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      MergeARGB16To8Row = MergeARGB16To8Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEARGB16TO8ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeARGB16To8Row = MergeARGB16To8Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      MergeARGB16To8Row = MergeARGB16To8Row_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    MergeARGB16To8Row(src_r, src_g, src_b, src_a, dst_argb, depth, width);
+    src_r += src_stride_r;
+    src_g += src_stride_g;
+    src_b += src_stride_b;
+    src_a += src_stride_a;
+    dst_argb += dst_stride_argb;
+  }
+}
+
+LIBYUV_NOINLINE
+static void MergeARGB16To8PlaneOpaque(const uint16_t* src_r,
+                                      int src_stride_r,
+                                      const uint16_t* src_g,
+                                      int src_stride_g,
+                                      const uint16_t* src_b,
+                                      int src_stride_b,
+                                      uint8_t* dst_argb,
+                                      int dst_stride_argb,
+                                      int width,
+                                      int height,
+                                      int depth) {
+  int y;
+  void (*MergeXRGB16To8Row)(const uint16_t* src_r, const uint16_t* src_g,
+                            const uint16_t* src_b, uint8_t* dst_argb, int depth,
+                            int width) = MergeXRGB16To8Row_C;
+
+  // Coalesce rows.
+  if (src_stride_r == width && src_stride_g == width && src_stride_b == width &&
+      dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_r = src_stride_g = src_stride_b = dst_stride_argb = 0;
+  }
+#if defined(HAS_MERGEXRGB16TO8ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MergeXRGB16To8Row = MergeXRGB16To8Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      MergeXRGB16To8Row = MergeXRGB16To8Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MERGEXRGB16TO8ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MergeXRGB16To8Row = MergeXRGB16To8Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      MergeXRGB16To8Row = MergeXRGB16To8Row_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    MergeXRGB16To8Row(src_r, src_g, src_b, dst_argb, depth, width);
+    src_r += src_stride_r;
+    src_g += src_stride_g;
+    src_b += src_stride_b;
+    dst_argb += dst_stride_argb;
+  }
+}
+
+LIBYUV_API
+void MergeARGB16To8Plane(const uint16_t* src_r,
+                         int src_stride_r,
+                         const uint16_t* src_g,
+                         int src_stride_g,
+                         const uint16_t* src_b,
+                         int src_stride_b,
+                         const uint16_t* src_a,
+                         int src_stride_a,
+                         uint8_t* dst_argb,
+                         int dst_stride_argb,
+                         int width,
+                         int height,
+                         int depth) {
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+
+  if (src_a == NULL) {
+    MergeARGB16To8PlaneOpaque(src_r, src_stride_r, src_g, src_stride_g, src_b,
+                              src_stride_b, dst_argb, dst_stride_argb, width,
+                              height, depth);
+  } else {
+    MergeARGB16To8PlaneAlpha(src_r, src_stride_r, src_g, src_stride_g, src_b,
+                             src_stride_b, src_a, src_stride_a, dst_argb,
+                             dst_stride_argb, width, height, depth);
+  }
+}
+
+// Convert YUY2 to I422.
+LIBYUV_API
+int YUY2ToI422(const uint8_t* src_yuy2,
+               int src_stride_yuy2,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int y;
+  void (*YUY2ToUV422Row)(const uint8_t* src_yuy2, uint8_t* dst_u,
+                         uint8_t* dst_v, int width) = YUY2ToUV422Row_C;
+  void (*YUY2ToYRow)(const uint8_t* src_yuy2, uint8_t* dst_y, int width) =
+      YUY2ToYRow_C;
+  if (!src_yuy2 || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_yuy2 = src_yuy2 + (height - 1) * src_stride_yuy2;
+    src_stride_yuy2 = -src_stride_yuy2;
+  }
+  // Coalesce rows.
+  if (src_stride_yuy2 == width * 2 && dst_stride_y == width &&
+      dst_stride_u * 2 == width && dst_stride_v * 2 == width &&
+      width * height <= 32768) {
+    width *= height;
+    height = 1;
+    src_stride_yuy2 = dst_stride_y = dst_stride_u = dst_stride_v = 0;
+  }
+#if defined(HAS_YUY2TOYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    YUY2ToUV422Row = YUY2ToUV422Row_Any_SSE2;
+    YUY2ToYRow = YUY2ToYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToUV422Row = YUY2ToUV422Row_SSE2;
+      YUY2ToYRow = YUY2ToYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    YUY2ToUV422Row = YUY2ToUV422Row_Any_AVX2;
+    YUY2ToYRow = YUY2ToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToUV422Row = YUY2ToUV422Row_AVX2;
+      YUY2ToYRow = YUY2ToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    YUY2ToYRow = YUY2ToYRow_Any_NEON;
+    YUY2ToUV422Row = YUY2ToUV422Row_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToYRow = YUY2ToYRow_NEON;
+      YUY2ToUV422Row = YUY2ToUV422Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_MSA) && defined(HAS_YUY2TOUV422ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    YUY2ToYRow = YUY2ToYRow_Any_MSA;
+    YUY2ToUV422Row = YUY2ToUV422Row_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToYRow = YUY2ToYRow_MSA;
+      YUY2ToUV422Row = YUY2ToUV422Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_LSX) && defined(HAS_YUY2TOUV422ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    YUY2ToYRow = YUY2ToYRow_Any_LSX;
+    YUY2ToUV422Row = YUY2ToUV422Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToYRow = YUY2ToYRow_LSX;
+      YUY2ToUV422Row = YUY2ToUV422Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_LASX) && defined(HAS_YUY2TOUV422ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    YUY2ToYRow = YUY2ToYRow_Any_LASX;
+    YUY2ToUV422Row = YUY2ToUV422Row_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToYRow = YUY2ToYRow_LASX;
+      YUY2ToUV422Row = YUY2ToUV422Row_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    YUY2ToUV422Row(src_yuy2, dst_u, dst_v, width);
+    YUY2ToYRow(src_yuy2, dst_y, width);
+    src_yuy2 += src_stride_yuy2;
+    dst_y += dst_stride_y;
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+  }
+  return 0;
+}
+
+// Convert UYVY to I422.
+LIBYUV_API
+int UYVYToI422(const uint8_t* src_uyvy,
+               int src_stride_uyvy,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int y;
+  void (*UYVYToUV422Row)(const uint8_t* src_uyvy, uint8_t* dst_u,
+                         uint8_t* dst_v, int width) = UYVYToUV422Row_C;
+  void (*UYVYToYRow)(const uint8_t* src_uyvy, uint8_t* dst_y, int width) =
+      UYVYToYRow_C;
+  if (!src_uyvy || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_uyvy = src_uyvy + (height - 1) * src_stride_uyvy;
+    src_stride_uyvy = -src_stride_uyvy;
+  }
+  // Coalesce rows.
+  if (src_stride_uyvy == width * 2 && dst_stride_y == width &&
+      dst_stride_u * 2 == width && dst_stride_v * 2 == width &&
+      width * height <= 32768) {
+    width *= height;
+    height = 1;
+    src_stride_uyvy = dst_stride_y = dst_stride_u = dst_stride_v = 0;
+  }
+#if defined(HAS_UYVYTOYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    UYVYToUV422Row = UYVYToUV422Row_Any_SSE2;
+    UYVYToYRow = UYVYToYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      UYVYToUV422Row = UYVYToUV422Row_SSE2;
+      UYVYToYRow = UYVYToYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    UYVYToUV422Row = UYVYToUV422Row_Any_AVX2;
+    UYVYToYRow = UYVYToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      UYVYToUV422Row = UYVYToUV422Row_AVX2;
+      UYVYToYRow = UYVYToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    UYVYToYRow = UYVYToYRow_Any_NEON;
+    UYVYToUV422Row = UYVYToUV422Row_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      UYVYToYRow = UYVYToYRow_NEON;
+      UYVYToUV422Row = UYVYToUV422Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_MSA) && defined(HAS_UYVYTOUV422ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    UYVYToYRow = UYVYToYRow_Any_MSA;
+    UYVYToUV422Row = UYVYToUV422Row_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      UYVYToYRow = UYVYToYRow_MSA;
+      UYVYToUV422Row = UYVYToUV422Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_LSX) && defined(HAS_UYVYTOUV422ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    UYVYToYRow = UYVYToYRow_Any_LSX;
+    UYVYToUV422Row = UYVYToUV422Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      UYVYToYRow = UYVYToYRow_LSX;
+      UYVYToUV422Row = UYVYToUV422Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_LASX) && defined(HAS_UYVYTOUV422ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    UYVYToYRow = UYVYToYRow_Any_LASX;
+    UYVYToUV422Row = UYVYToUV422Row_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      UYVYToYRow = UYVYToYRow_LASX;
+      UYVYToUV422Row = UYVYToUV422Row_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    UYVYToUV422Row(src_uyvy, dst_u, dst_v, width);
+    UYVYToYRow(src_uyvy, dst_y, width);
+    src_uyvy += src_stride_uyvy;
+    dst_y += dst_stride_y;
+    dst_u += dst_stride_u;
+    dst_v += dst_stride_v;
+  }
+  return 0;
+}
+
+// Convert YUY2 to Y.
+LIBYUV_API
+int YUY2ToY(const uint8_t* src_yuy2,
+            int src_stride_yuy2,
+            uint8_t* dst_y,
+            int dst_stride_y,
+            int width,
+            int height) {
+  int y;
+  void (*YUY2ToYRow)(const uint8_t* src_yuy2, uint8_t* dst_y, int width) =
+      YUY2ToYRow_C;
+  if (!src_yuy2 || !dst_y || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_yuy2 = src_yuy2 + (height - 1) * src_stride_yuy2;
+    src_stride_yuy2 = -src_stride_yuy2;
+  }
+  // Coalesce rows.
+  if (src_stride_yuy2 == width * 2 && dst_stride_y == width) {
+    width *= height;
+    height = 1;
+    src_stride_yuy2 = dst_stride_y = 0;
+  }
+#if defined(HAS_YUY2TOYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    YUY2ToYRow = YUY2ToYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToYRow = YUY2ToYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    YUY2ToYRow = YUY2ToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToYRow = YUY2ToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    YUY2ToYRow = YUY2ToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToYRow = YUY2ToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    YUY2ToYRow = YUY2ToYRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToYRow = YUY2ToYRow_MSA;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    YUY2ToYRow(src_yuy2, dst_y, width);
+    src_yuy2 += src_stride_yuy2;
+    dst_y += dst_stride_y;
+  }
+  return 0;
+}
+
+// Convert UYVY to Y.
+LIBYUV_API
+int UYVYToY(const uint8_t* src_uyvy,
+            int src_stride_uyvy,
+            uint8_t* dst_y,
+            int dst_stride_y,
+            int width,
+            int height) {
+  int y;
+  void (*UYVYToYRow)(const uint8_t* src_uyvy, uint8_t* dst_y, int width) =
+      UYVYToYRow_C;
+  if (!src_uyvy || !dst_y || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_uyvy = src_uyvy + (height - 1) * src_stride_uyvy;
+    src_stride_uyvy = -src_stride_uyvy;
+  }
+  // Coalesce rows.
+  if (src_stride_uyvy == width * 2 && dst_stride_y == width) {
+    width *= height;
+    height = 1;
+    src_stride_uyvy = dst_stride_y = 0;
+  }
+#if defined(HAS_UYVYTOYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    UYVYToYRow = UYVYToYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      UYVYToYRow = UYVYToYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    UYVYToYRow = UYVYToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      UYVYToYRow = UYVYToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    UYVYToYRow = UYVYToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      UYVYToYRow = UYVYToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    UYVYToYRow = UYVYToYRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      UYVYToYRow = UYVYToYRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_UYVYTOYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    UYVYToYRow = UYVYToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      UYVYToYRow = UYVYToYRow_LSX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    UYVYToYRow(src_uyvy, dst_y, width);
+    src_uyvy += src_stride_uyvy;
+    dst_y += dst_stride_y;
+  }
+  return 0;
+}
+
+// Mirror a plane of data.
+// See Also I400Mirror
+LIBYUV_API
+void MirrorPlane(const uint8_t* src_y,
+                 int src_stride_y,
+                 uint8_t* dst_y,
+                 int dst_stride_y,
+                 int width,
+                 int height) {
+  int y;
+  void (*MirrorRow)(const uint8_t* src, uint8_t* dst, int width) = MirrorRow_C;
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_stride_y = -src_stride_y;
+  }
+#if defined(HAS_MIRRORROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MirrorRow = MirrorRow_Any_NEON;
+    if (IS_ALIGNED(width, 32)) {
+      MirrorRow = MirrorRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    MirrorRow = MirrorRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      MirrorRow = MirrorRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MirrorRow = MirrorRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      MirrorRow = MirrorRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    MirrorRow = MirrorRow_Any_MSA;
+    if (IS_ALIGNED(width, 64)) {
+      MirrorRow = MirrorRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    MirrorRow = MirrorRow_Any_LSX;
+    if (IS_ALIGNED(width, 32)) {
+      MirrorRow = MirrorRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    MirrorRow = MirrorRow_Any_LASX;
+    if (IS_ALIGNED(width, 64)) {
+      MirrorRow = MirrorRow_LASX;
+    }
+  }
+#endif
+
+  // Mirror plane
+  for (y = 0; y < height; ++y) {
+    MirrorRow(src_y, dst_y, width);
+    src_y += src_stride_y;
+    dst_y += dst_stride_y;
+  }
+}
+
+// Mirror a plane of UV data.
+LIBYUV_API
+void MirrorUVPlane(const uint8_t* src_uv,
+                   int src_stride_uv,
+                   uint8_t* dst_uv,
+                   int dst_stride_uv,
+                   int width,
+                   int height) {
+  int y;
+  void (*MirrorUVRow)(const uint8_t* src, uint8_t* dst, int width) =
+      MirrorUVRow_C;
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_uv = src_uv + (height - 1) * src_stride_uv;
+    src_stride_uv = -src_stride_uv;
+  }
+#if defined(HAS_MIRRORUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MirrorUVRow = MirrorUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 32)) {
+      MirrorUVRow = MirrorUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    MirrorUVRow = MirrorUVRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      MirrorUVRow = MirrorUVRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MirrorUVRow = MirrorUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      MirrorUVRow = MirrorUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    MirrorUVRow = MirrorUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      MirrorUVRow = MirrorUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    MirrorUVRow = MirrorUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      MirrorUVRow = MirrorUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORUVROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    MirrorUVRow = MirrorUVRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      MirrorUVRow = MirrorUVRow_LASX;
+    }
+  }
+#endif
+
+  // MirrorUV plane
+  for (y = 0; y < height; ++y) {
+    MirrorUVRow(src_uv, dst_uv, width);
+    src_uv += src_stride_uv;
+    dst_uv += dst_stride_uv;
+  }
+}
+
+// Mirror I400 with optional flipping
+LIBYUV_API
+int I400Mirror(const uint8_t* src_y,
+               int src_stride_y,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               int width,
+               int height) {
+  if (!src_y || !dst_y || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_stride_y = -src_stride_y;
+  }
+
+  MirrorPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  return 0;
+}
+
+// Mirror I420 with optional flipping
+LIBYUV_API
+int I420Mirror(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+
+  if (!src_y || !src_u || !src_v || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (halfheight - 1) * src_stride_u;
+    src_v = src_v + (halfheight - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  if (dst_y) {
+    MirrorPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+  MirrorPlane(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth, halfheight);
+  MirrorPlane(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth, halfheight);
+  return 0;
+}
+
+// NV12 mirror.
+LIBYUV_API
+int NV12Mirror(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_uv,
+               int src_stride_uv,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+
+  if (!src_y || !src_uv || !dst_uv || width <= 0 || height == 0) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_uv = src_uv + (halfheight - 1) * src_stride_uv;
+    src_stride_y = -src_stride_y;
+    src_stride_uv = -src_stride_uv;
+  }
+
+  if (dst_y) {
+    MirrorPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+  }
+  MirrorUVPlane(src_uv, src_stride_uv, dst_uv, dst_stride_uv, halfwidth,
+                halfheight);
+  return 0;
+}
+
+// ARGB mirror.
+LIBYUV_API
+int ARGBMirror(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBMirrorRow)(const uint8_t* src, uint8_t* dst, int width) =
+      ARGBMirrorRow_C;
+  if (!src_argb || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+#if defined(HAS_ARGBMIRRORROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBMirrorRow = ARGBMirrorRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBMirrorRow = ARGBMirrorRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMIRRORROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBMirrorRow = ARGBMirrorRow_Any_SSE2;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBMirrorRow = ARGBMirrorRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMIRRORROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBMirrorRow = ARGBMirrorRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBMirrorRow = ARGBMirrorRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMIRRORROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBMirrorRow = ARGBMirrorRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBMirrorRow = ARGBMirrorRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMIRRORROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBMirrorRow = ARGBMirrorRow_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBMirrorRow = ARGBMirrorRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMIRRORROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBMirrorRow = ARGBMirrorRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBMirrorRow = ARGBMirrorRow_LASX;
+    }
+  }
+#endif
+
+  // Mirror plane
+  for (y = 0; y < height; ++y) {
+    ARGBMirrorRow(src_argb, dst_argb, width);
+    src_argb += src_stride_argb;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// RGB24 mirror.
+LIBYUV_API
+int RGB24Mirror(const uint8_t* src_rgb24,
+                int src_stride_rgb24,
+                uint8_t* dst_rgb24,
+                int dst_stride_rgb24,
+                int width,
+                int height) {
+  int y;
+  void (*RGB24MirrorRow)(const uint8_t* src, uint8_t* dst, int width) =
+      RGB24MirrorRow_C;
+  if (!src_rgb24 || !dst_rgb24 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_rgb24 = src_rgb24 + (height - 1) * src_stride_rgb24;
+    src_stride_rgb24 = -src_stride_rgb24;
+  }
+#if defined(HAS_RGB24MIRRORROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RGB24MirrorRow = RGB24MirrorRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24MirrorRow = RGB24MirrorRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RGB24MIRRORROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RGB24MirrorRow = RGB24MirrorRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      RGB24MirrorRow = RGB24MirrorRow_SSSE3;
+    }
+  }
+#endif
+
+  // Mirror plane
+  for (y = 0; y < height; ++y) {
+    RGB24MirrorRow(src_rgb24, dst_rgb24, width);
+    src_rgb24 += src_stride_rgb24;
+    dst_rgb24 += dst_stride_rgb24;
+  }
+  return 0;
+}
+
+// Alpha Blend 2 ARGB images and store to destination.
+LIBYUV_API
+int ARGBBlend(const uint8_t* src_argb0,
+              int src_stride_argb0,
+              const uint8_t* src_argb1,
+              int src_stride_argb1,
+              uint8_t* dst_argb,
+              int dst_stride_argb,
+              int width,
+              int height) {
+  int y;
+  void (*ARGBBlendRow)(const uint8_t* src_argb, const uint8_t* src_argb1,
+                       uint8_t* dst_argb, int width) = ARGBBlendRow_C;
+  if (!src_argb0 || !src_argb1 || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb0 == width * 4 && src_stride_argb1 == width * 4 &&
+      dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb0 = src_stride_argb1 = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBBLENDROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBBlendRow = ARGBBlendRow_SSSE3;
+  }
+#endif
+#if defined(HAS_ARGBBLENDROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBBlendRow = ARGBBlendRow_NEON;
+  }
+#endif
+#if defined(HAS_ARGBBLENDROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBBlendRow = ARGBBlendRow_MSA;
+  }
+#endif
+#if defined(HAS_ARGBBLENDROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBBlendRow = ARGBBlendRow_LSX;
+  }
+#endif
+#if defined(HAS_ARGBBLENDROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBBlendRow = ARGBBlendRow_RVV;
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    ARGBBlendRow(src_argb0, src_argb1, dst_argb, width);
+    src_argb0 += src_stride_argb0;
+    src_argb1 += src_stride_argb1;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Alpha Blend plane and store to destination.
+LIBYUV_API
+int BlendPlane(const uint8_t* src_y0,
+               int src_stride_y0,
+               const uint8_t* src_y1,
+               int src_stride_y1,
+               const uint8_t* alpha,
+               int alpha_stride,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               int width,
+               int height) {
+  int y;
+  void (*BlendPlaneRow)(const uint8_t* src0, const uint8_t* src1,
+                        const uint8_t* alpha, uint8_t* dst, int width) =
+      BlendPlaneRow_C;
+  if (!src_y0 || !src_y1 || !alpha || !dst_y || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_y = dst_y + (height - 1) * dst_stride_y;
+    dst_stride_y = -dst_stride_y;
+  }
+
+  // Coalesce rows for Y plane.
+  if (src_stride_y0 == width && src_stride_y1 == width &&
+      alpha_stride == width && dst_stride_y == width) {
+    width *= height;
+    height = 1;
+    src_stride_y0 = src_stride_y1 = alpha_stride = dst_stride_y = 0;
+  }
+
+#if defined(HAS_BLENDPLANEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    BlendPlaneRow = BlendPlaneRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      BlendPlaneRow = BlendPlaneRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_BLENDPLANEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    BlendPlaneRow = BlendPlaneRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      BlendPlaneRow = BlendPlaneRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_BLENDPLANEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    BlendPlaneRow = BlendPlaneRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    BlendPlaneRow(src_y0, src_y1, alpha, dst_y, width);
+    src_y0 += src_stride_y0;
+    src_y1 += src_stride_y1;
+    alpha += alpha_stride;
+    dst_y += dst_stride_y;
+  }
+  return 0;
+}
+
+#define MAXTWIDTH 2048
+// Alpha Blend YUV images and store to destination.
+LIBYUV_API
+int I420Blend(const uint8_t* src_y0,
+              int src_stride_y0,
+              const uint8_t* src_u0,
+              int src_stride_u0,
+              const uint8_t* src_v0,
+              int src_stride_v0,
+              const uint8_t* src_y1,
+              int src_stride_y1,
+              const uint8_t* src_u1,
+              int src_stride_u1,
+              const uint8_t* src_v1,
+              int src_stride_v1,
+              const uint8_t* alpha,
+              int alpha_stride,
+              uint8_t* dst_y,
+              int dst_stride_y,
+              uint8_t* dst_u,
+              int dst_stride_u,
+              uint8_t* dst_v,
+              int dst_stride_v,
+              int width,
+              int height) {
+  int y;
+  // Half width/height for UV.
+  int halfwidth = (width + 1) >> 1;
+  void (*BlendPlaneRow)(const uint8_t* src0, const uint8_t* src1,
+                        const uint8_t* alpha, uint8_t* dst, int width) =
+      BlendPlaneRow_C;
+  void (*ScaleRowDown2)(const uint8_t* src_ptr, ptrdiff_t src_stride,
+                        uint8_t* dst_ptr, int dst_width) = ScaleRowDown2Box_C;
+
+  if (!src_y0 || !src_u0 || !src_v0 || !src_y1 || !src_u1 || !src_v1 ||
+      !alpha || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_y = dst_y + (height - 1) * dst_stride_y;
+    dst_stride_y = -dst_stride_y;
+  }
+
+  // Blend Y plane.
+  BlendPlane(src_y0, src_stride_y0, src_y1, src_stride_y1, alpha, alpha_stride,
+             dst_y, dst_stride_y, width, height);
+
+#if defined(HAS_BLENDPLANEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    BlendPlaneRow = BlendPlaneRow_Any_SSSE3;
+    if (IS_ALIGNED(halfwidth, 8)) {
+      BlendPlaneRow = BlendPlaneRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_BLENDPLANEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    BlendPlaneRow = BlendPlaneRow_Any_AVX2;
+    if (IS_ALIGNED(halfwidth, 32)) {
+      BlendPlaneRow = BlendPlaneRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_BLENDPLANEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    BlendPlaneRow = BlendPlaneRow_RVV;
+  }
+#endif
+  if (!IS_ALIGNED(width, 2)) {
+    ScaleRowDown2 = ScaleRowDown2Box_Odd_C;
+  }
+#if defined(HAS_SCALEROWDOWN2_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ScaleRowDown2 = ScaleRowDown2Box_Odd_NEON;
+    if (IS_ALIGNED(width, 2)) {
+      ScaleRowDown2 = ScaleRowDown2Box_Any_NEON;
+      if (IS_ALIGNED(halfwidth, 16)) {
+        ScaleRowDown2 = ScaleRowDown2Box_NEON;
+      }
+    }
+  }
+#endif
+#if defined(HAS_SCALEROWDOWN2_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ScaleRowDown2 = ScaleRowDown2Box_Odd_SSSE3;
+    if (IS_ALIGNED(width, 2)) {
+      ScaleRowDown2 = ScaleRowDown2Box_Any_SSSE3;
+      if (IS_ALIGNED(halfwidth, 16)) {
+        ScaleRowDown2 = ScaleRowDown2Box_SSSE3;
+      }
+    }
+  }
+#endif
+#if defined(HAS_SCALEROWDOWN2_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ScaleRowDown2 = ScaleRowDown2Box_Odd_AVX2;
+    if (IS_ALIGNED(width, 2)) {
+      ScaleRowDown2 = ScaleRowDown2Box_Any_AVX2;
+      if (IS_ALIGNED(halfwidth, 32)) {
+        ScaleRowDown2 = ScaleRowDown2Box_AVX2;
+      }
+    }
+  }
+#endif
+#if defined(HAS_SCALEROWDOWN2_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ScaleRowDown2 = ScaleRowDown2Box_RVV;
+  }
+#endif
+
+  // Row buffer for intermediate alpha pixels.
+  align_buffer_64(halfalpha, halfwidth);
+  if (!halfalpha)
+    return 1;
+  for (y = 0; y < height; y += 2) {
+    // last row of odd height image use 1 row of alpha instead of 2.
+    if (y == (height - 1)) {
+      alpha_stride = 0;
+    }
+    // Subsample 2 rows of UV to half width and half height.
+    ScaleRowDown2(alpha, alpha_stride, halfalpha, halfwidth);
+    alpha += alpha_stride * 2;
+    BlendPlaneRow(src_u0, src_u1, halfalpha, dst_u, halfwidth);
+    BlendPlaneRow(src_v0, src_v1, halfalpha, dst_v, halfwidth);
+    src_u0 += src_stride_u0;
+    src_u1 += src_stride_u1;
+    dst_u += dst_stride_u;
+    src_v0 += src_stride_v0;
+    src_v1 += src_stride_v1;
+    dst_v += dst_stride_v;
+  }
+  free_aligned_buffer_64(halfalpha);
+  return 0;
+}
+
+// Multiply 2 ARGB images and store to destination.
+LIBYUV_API
+int ARGBMultiply(const uint8_t* src_argb0,
+                 int src_stride_argb0,
+                 const uint8_t* src_argb1,
+                 int src_stride_argb1,
+                 uint8_t* dst_argb,
+                 int dst_stride_argb,
+                 int width,
+                 int height) {
+  int y;
+  void (*ARGBMultiplyRow)(const uint8_t* src0, const uint8_t* src1,
+                          uint8_t* dst, int width) = ARGBMultiplyRow_C;
+  if (!src_argb0 || !src_argb1 || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb0 == width * 4 && src_stride_argb1 == width * 4 &&
+      dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb0 = src_stride_argb1 = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBMULTIPLYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBMultiplyRow = ARGBMultiplyRow_Any_SSE2;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBMultiplyRow = ARGBMultiplyRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMULTIPLYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBMultiplyRow = ARGBMultiplyRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBMultiplyRow = ARGBMultiplyRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMULTIPLYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBMultiplyRow = ARGBMultiplyRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBMultiplyRow = ARGBMultiplyRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMULTIPLYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBMultiplyRow = ARGBMultiplyRow_Any_MSA;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBMultiplyRow = ARGBMultiplyRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMULTIPLYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBMultiplyRow = ARGBMultiplyRow_Any_LSX;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBMultiplyRow = ARGBMultiplyRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMULTIPLYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBMultiplyRow = ARGBMultiplyRow_Any_LASX;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBMultiplyRow = ARGBMultiplyRow_LASX;
+    }
+  }
+#endif
+
+  // Multiply plane
+  for (y = 0; y < height; ++y) {
+    ARGBMultiplyRow(src_argb0, src_argb1, dst_argb, width);
+    src_argb0 += src_stride_argb0;
+    src_argb1 += src_stride_argb1;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Add 2 ARGB images and store to destination.
+LIBYUV_API
+int ARGBAdd(const uint8_t* src_argb0,
+            int src_stride_argb0,
+            const uint8_t* src_argb1,
+            int src_stride_argb1,
+            uint8_t* dst_argb,
+            int dst_stride_argb,
+            int width,
+            int height) {
+  int y;
+  void (*ARGBAddRow)(const uint8_t* src0, const uint8_t* src1, uint8_t* dst,
+                     int width) = ARGBAddRow_C;
+  if (!src_argb0 || !src_argb1 || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb0 == width * 4 && src_stride_argb1 == width * 4 &&
+      dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb0 = src_stride_argb1 = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBADDROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBAddRow = ARGBAddRow_SSE2;
+  }
+#endif
+#if defined(HAS_ARGBADDROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBAddRow = ARGBAddRow_Any_SSE2;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAddRow = ARGBAddRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBADDROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBAddRow = ARGBAddRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAddRow = ARGBAddRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBADDROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBAddRow = ARGBAddRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAddRow = ARGBAddRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBADDROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBAddRow = ARGBAddRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAddRow = ARGBAddRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBADDROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBAddRow = ARGBAddRow_Any_LSX;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAddRow = ARGBAddRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBADDROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBAddRow = ARGBAddRow_Any_LASX;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAddRow = ARGBAddRow_LASX;
+    }
+  }
+#endif
+
+  // Add plane
+  for (y = 0; y < height; ++y) {
+    ARGBAddRow(src_argb0, src_argb1, dst_argb, width);
+    src_argb0 += src_stride_argb0;
+    src_argb1 += src_stride_argb1;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Subtract 2 ARGB images and store to destination.
+LIBYUV_API
+int ARGBSubtract(const uint8_t* src_argb0,
+                 int src_stride_argb0,
+                 const uint8_t* src_argb1,
+                 int src_stride_argb1,
+                 uint8_t* dst_argb,
+                 int dst_stride_argb,
+                 int width,
+                 int height) {
+  int y;
+  void (*ARGBSubtractRow)(const uint8_t* src0, const uint8_t* src1,
+                          uint8_t* dst, int width) = ARGBSubtractRow_C;
+  if (!src_argb0 || !src_argb1 || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb0 == width * 4 && src_stride_argb1 == width * 4 &&
+      dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb0 = src_stride_argb1 = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBSUBTRACTROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBSubtractRow = ARGBSubtractRow_Any_SSE2;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBSubtractRow = ARGBSubtractRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSUBTRACTROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBSubtractRow = ARGBSubtractRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBSubtractRow = ARGBSubtractRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSUBTRACTROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBSubtractRow = ARGBSubtractRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBSubtractRow = ARGBSubtractRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSUBTRACTROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBSubtractRow = ARGBSubtractRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBSubtractRow = ARGBSubtractRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSUBTRACTROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBSubtractRow = ARGBSubtractRow_Any_LSX;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBSubtractRow = ARGBSubtractRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSUBTRACTROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBSubtractRow = ARGBSubtractRow_Any_LASX;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBSubtractRow = ARGBSubtractRow_LASX;
+    }
+  }
+#endif
+
+  // Subtract plane
+  for (y = 0; y < height; ++y) {
+    ARGBSubtractRow(src_argb0, src_argb1, dst_argb, width);
+    src_argb0 += src_stride_argb0;
+    src_argb1 += src_stride_argb1;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert RAW to RGB24.
+LIBYUV_API
+int RAWToRGB24(const uint8_t* src_raw,
+               int src_stride_raw,
+               uint8_t* dst_rgb24,
+               int dst_stride_rgb24,
+               int width,
+               int height) {
+  int y;
+  void (*RAWToRGB24Row)(const uint8_t* src_rgb, uint8_t* dst_rgb24, int width) =
+      RAWToRGB24Row_C;
+  if (!src_raw || !dst_rgb24 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_raw = src_raw + (height - 1) * src_stride_raw;
+    src_stride_raw = -src_stride_raw;
+  }
+  // Coalesce rows.
+  if (src_stride_raw == width * 3 && dst_stride_rgb24 == width * 3) {
+    width *= height;
+    height = 1;
+    src_stride_raw = dst_stride_rgb24 = 0;
+  }
+#if defined(HAS_RAWTORGB24ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    RAWToRGB24Row = RAWToRGB24Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      RAWToRGB24Row = RAWToRGB24Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_RAWTORGB24ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    RAWToRGB24Row = RAWToRGB24Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      RAWToRGB24Row = RAWToRGB24Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_RAWTORGB24ROW_SVE2)
+  if (TestCpuFlag(kCpuHasSVE2)) {
+    RAWToRGB24Row = RAWToRGB24Row_SVE2;
+  }
+#endif
+#if defined(HAS_RAWTORGB24ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    RAWToRGB24Row = RAWToRGB24Row_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToRGB24Row = RAWToRGB24Row_MSA;
+    }
+  }
+#endif
+#if defined(HAS_RAWTORGB24ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    RAWToRGB24Row = RAWToRGB24Row_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      RAWToRGB24Row = RAWToRGB24Row_LSX;
+    }
+  }
+#endif
+#if defined(HAS_RAWTORGB24ROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RAWToRGB24Row = RAWToRGB24Row_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    RAWToRGB24Row(src_raw, dst_rgb24, width);
+    src_raw += src_stride_raw;
+    dst_rgb24 += dst_stride_rgb24;
+  }
+  return 0;
+}
+
+// TODO(fbarchard): Consider uint8_t value
+LIBYUV_API
+void SetPlane(uint8_t* dst_y,
+              int dst_stride_y,
+              int width,
+              int height,
+              uint32_t value) {
+  int y;
+  void (*SetRow)(uint8_t* dst, uint8_t value, int width) = SetRow_C;
+
+  if (width <= 0 || height == 0) {
+    return;
+  }
+  if (height < 0) {
+    height = -height;
+    dst_y = dst_y + (height - 1) * dst_stride_y;
+    dst_stride_y = -dst_stride_y;
+  }
+  // Coalesce rows.
+  if (dst_stride_y == width) {
+    width *= height;
+    height = 1;
+    dst_stride_y = 0;
+  }
+#if defined(HAS_SETROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SetRow = SetRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      SetRow = SetRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_SETROW_X86)
+  if (TestCpuFlag(kCpuHasX86)) {
+    SetRow = SetRow_Any_X86;
+    if (IS_ALIGNED(width, 4)) {
+      SetRow = SetRow_X86;
+    }
+  }
+#endif
+#if defined(HAS_SETROW_ERMS)
+  if (TestCpuFlag(kCpuHasERMS)) {
+    SetRow = SetRow_ERMS;
+  }
+#endif
+#if defined(HAS_SETROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 16)) {
+    SetRow = SetRow_MSA;
+  }
+#endif
+#if defined(HAS_SETROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    SetRow = SetRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      SetRow = SetRow_LSX;
+    }
+  }
+#endif
+
+  // Set plane
+  for (y = 0; y < height; ++y) {
+    SetRow(dst_y, (uint8_t)value, width);
+    dst_y += dst_stride_y;
+  }
+}
+
+// Draw a rectangle into I420
+LIBYUV_API
+int I420Rect(uint8_t* dst_y,
+             int dst_stride_y,
+             uint8_t* dst_u,
+             int dst_stride_u,
+             uint8_t* dst_v,
+             int dst_stride_v,
+             int x,
+             int y,
+             int width,
+             int height,
+             int value_y,
+             int value_u,
+             int value_v) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  uint8_t* start_y = dst_y + y * dst_stride_y + x;
+  uint8_t* start_u = dst_u + (y / 2) * dst_stride_u + (x / 2);
+  uint8_t* start_v = dst_v + (y / 2) * dst_stride_v + (x / 2);
+
+  if (!dst_y || !dst_u || !dst_v || width <= 0 || height == 0 || x < 0 ||
+      y < 0 || value_y < 0 || value_y > 255 || value_u < 0 || value_u > 255 ||
+      value_v < 0 || value_v > 255) {
+    return -1;
+  }
+
+  SetPlane(start_y, dst_stride_y, width, height, value_y);
+  SetPlane(start_u, dst_stride_u, halfwidth, halfheight, value_u);
+  SetPlane(start_v, dst_stride_v, halfwidth, halfheight, value_v);
+  return 0;
+}
+
+// Draw a rectangle into ARGB
+LIBYUV_API
+int ARGBRect(uint8_t* dst_argb,
+             int dst_stride_argb,
+             int dst_x,
+             int dst_y,
+             int width,
+             int height,
+             uint32_t value) {
+  int y;
+  void (*ARGBSetRow)(uint8_t* dst_argb, uint32_t value, int width) =
+      ARGBSetRow_C;
+  if (!dst_argb || width <= 0 || height == 0 || dst_x < 0 || dst_y < 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+  dst_argb += dst_y * dst_stride_argb + dst_x * 4;
+  // Coalesce rows.
+  if (dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    dst_stride_argb = 0;
+  }
+
+#if defined(HAS_ARGBSETROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBSetRow = ARGBSetRow_Any_NEON;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBSetRow = ARGBSetRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSETROW_X86)
+  if (TestCpuFlag(kCpuHasX86)) {
+    ARGBSetRow = ARGBSetRow_X86;
+  }
+#endif
+#if defined(HAS_ARGBSETROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBSetRow = ARGBSetRow_Any_MSA;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBSetRow = ARGBSetRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSETROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBSetRow = ARGBSetRow_Any_LSX;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBSetRow = ARGBSetRow_LSX;
+    }
+  }
+#endif
+
+  // Set plane
+  for (y = 0; y < height; ++y) {
+    ARGBSetRow(dst_argb, value, width);
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert unattentuated ARGB to preattenuated ARGB.
+// An unattenutated ARGB alpha blend uses the formula
+// p = a * f + (1 - a) * b
+// where
+//   p is output pixel
+//   f is foreground pixel
+//   b is background pixel
+//   a is alpha value from foreground pixel
+// An preattenutated ARGB alpha blend uses the formula
+// p = f + (1 - a) * b
+// where
+//   f is foreground pixel premultiplied by alpha
+
+LIBYUV_API
+int ARGBAttenuate(const uint8_t* src_argb,
+                  int src_stride_argb,
+                  uint8_t* dst_argb,
+                  int dst_stride_argb,
+                  int width,
+                  int height) {
+  int y;
+  void (*ARGBAttenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                           int width) = ARGBAttenuateRow_C;
+  if (!src_argb || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBATTENUATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBAttenuateRow = ARGBAttenuateRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBATTENUATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBAttenuateRow = ARGBAttenuateRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBAttenuateRow(src_argb, dst_argb, width);
+    src_argb += src_stride_argb;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert preattentuated ARGB to unattenuated ARGB.
+LIBYUV_API
+int ARGBUnattenuate(const uint8_t* src_argb,
+                    int src_stride_argb,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    int width,
+                    int height) {
+  int y;
+  void (*ARGBUnattenuateRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                             int width) = ARGBUnattenuateRow_C;
+  if (!src_argb || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBUNATTENUATEROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBUnattenuateRow = ARGBUnattenuateRow_Any_SSE2;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBUnattenuateRow = ARGBUnattenuateRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBUNATTENUATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBUnattenuateRow = ARGBUnattenuateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBUnattenuateRow = ARGBUnattenuateRow_AVX2;
+    }
+  }
+#endif
+  // TODO(fbarchard): Neon version.
+
+  for (y = 0; y < height; ++y) {
+    ARGBUnattenuateRow(src_argb, dst_argb, width);
+    src_argb += src_stride_argb;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert ARGB to Grayed ARGB.
+LIBYUV_API
+int ARGBGrayTo(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height) {
+  int y;
+  void (*ARGBGrayRow)(const uint8_t* src_argb, uint8_t* dst_argb, int width) =
+      ARGBGrayRow_C;
+  if (!src_argb || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBGRAYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3) && IS_ALIGNED(width, 8)) {
+    ARGBGrayRow = ARGBGrayRow_SSSE3;
+  }
+#endif
+#if defined(HAS_ARGBGRAYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(width, 8)) {
+    ARGBGrayRow = ARGBGrayRow_NEON;
+  }
+#endif
+#if defined(HAS_ARGBGRAYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 8)) {
+    ARGBGrayRow = ARGBGrayRow_MSA;
+  }
+#endif
+#if defined(HAS_ARGBGRAYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
+    ARGBGrayRow = ARGBGrayRow_LSX;
+  }
+#endif
+#if defined(HAS_ARGBGRAYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX) && IS_ALIGNED(width, 16)) {
+    ARGBGrayRow = ARGBGrayRow_LASX;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBGrayRow(src_argb, dst_argb, width);
+    src_argb += src_stride_argb;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Make a rectangle of ARGB gray scale.
+LIBYUV_API
+int ARGBGray(uint8_t* dst_argb,
+             int dst_stride_argb,
+             int dst_x,
+             int dst_y,
+             int width,
+             int height) {
+  int y;
+  void (*ARGBGrayRow)(const uint8_t* src_argb, uint8_t* dst_argb, int width) =
+      ARGBGrayRow_C;
+  uint8_t* dst = dst_argb + dst_y * dst_stride_argb + dst_x * 4;
+  if (!dst_argb || width <= 0 || height <= 0 || dst_x < 0 || dst_y < 0) {
+    return -1;
+  }
+  // Coalesce rows.
+  if (dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBGRAYROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3) && IS_ALIGNED(width, 8)) {
+    ARGBGrayRow = ARGBGrayRow_SSSE3;
+  }
+#endif
+#if defined(HAS_ARGBGRAYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(width, 8)) {
+    ARGBGrayRow = ARGBGrayRow_NEON;
+  }
+#endif
+#if defined(HAS_ARGBGRAYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 8)) {
+    ARGBGrayRow = ARGBGrayRow_MSA;
+  }
+#endif
+#if defined(HAS_ARGBGRAYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
+    ARGBGrayRow = ARGBGrayRow_LSX;
+  }
+#endif
+#if defined(HAS_ARGBGRAYROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX) && IS_ALIGNED(width, 16)) {
+    ARGBGrayRow = ARGBGrayRow_LASX;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBGrayRow(dst, dst, width);
+    dst += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Make a rectangle of ARGB Sepia tone.
+LIBYUV_API
+int ARGBSepia(uint8_t* dst_argb,
+              int dst_stride_argb,
+              int dst_x,
+              int dst_y,
+              int width,
+              int height) {
+  int y;
+  void (*ARGBSepiaRow)(uint8_t* dst_argb, int width) = ARGBSepiaRow_C;
+  uint8_t* dst = dst_argb + dst_y * dst_stride_argb + dst_x * 4;
+  if (!dst_argb || width <= 0 || height <= 0 || dst_x < 0 || dst_y < 0) {
+    return -1;
+  }
+  // Coalesce rows.
+  if (dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBSEPIAROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3) && IS_ALIGNED(width, 8)) {
+    ARGBSepiaRow = ARGBSepiaRow_SSSE3;
+  }
+#endif
+#if defined(HAS_ARGBSEPIAROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(width, 8)) {
+    ARGBSepiaRow = ARGBSepiaRow_NEON;
+  }
+#endif
+#if defined(HAS_ARGBSEPIAROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 8)) {
+    ARGBSepiaRow = ARGBSepiaRow_MSA;
+  }
+#endif
+#if defined(HAS_ARGBSEPIAROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
+    ARGBSepiaRow = ARGBSepiaRow_LSX;
+  }
+#endif
+#if defined(HAS_ARGBSEPIAROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX) && IS_ALIGNED(width, 16)) {
+    ARGBSepiaRow = ARGBSepiaRow_LASX;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBSepiaRow(dst, width);
+    dst += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Apply a 4x4 matrix to each ARGB pixel.
+// Note: Normally for shading, but can be used to swizzle or invert.
+LIBYUV_API
+int ARGBColorMatrix(const uint8_t* src_argb,
+                    int src_stride_argb,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    const int8_t* matrix_argb,
+                    int width,
+                    int height) {
+  int y;
+  void (*ARGBColorMatrixRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                             const int8_t* matrix_argb, int width) =
+      ARGBColorMatrixRow_C;
+  if (!src_argb || !dst_argb || !matrix_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBCOLORMATRIXROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3) && IS_ALIGNED(width, 8)) {
+    ARGBColorMatrixRow = ARGBColorMatrixRow_SSSE3;
+  }
+#endif
+#if defined(HAS_ARGBCOLORMATRIXROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(width, 8)) {
+    ARGBColorMatrixRow = ARGBColorMatrixRow_NEON;
+  }
+#endif
+#if defined(HAS_ARGBCOLORMATRIXROW_NEON_I8MM)
+  if (TestCpuFlag(kCpuHasNeonI8MM) && IS_ALIGNED(width, 8)) {
+    ARGBColorMatrixRow = ARGBColorMatrixRow_NEON_I8MM;
+  }
+#endif
+#if defined(HAS_ARGBCOLORMATRIXROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 8)) {
+    ARGBColorMatrixRow = ARGBColorMatrixRow_MSA;
+  }
+#endif
+#if defined(HAS_ARGBCOLORMATRIXROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
+    ARGBColorMatrixRow = ARGBColorMatrixRow_LSX;
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    ARGBColorMatrixRow(src_argb, dst_argb, matrix_argb, width);
+    src_argb += src_stride_argb;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Apply a 4x3 matrix to each ARGB pixel.
+// Deprecated.
+LIBYUV_API
+int RGBColorMatrix(uint8_t* dst_argb,
+                   int dst_stride_argb,
+                   const int8_t* matrix_rgb,
+                   int dst_x,
+                   int dst_y,
+                   int width,
+                   int height) {
+  SIMD_ALIGNED(int8_t matrix_argb[16]);
+  uint8_t* dst = dst_argb + dst_y * dst_stride_argb + dst_x * 4;
+  if (!dst_argb || !matrix_rgb || width <= 0 || height <= 0 || dst_x < 0 ||
+      dst_y < 0) {
+    return -1;
+  }
+
+  // Convert 4x3 7 bit matrix to 4x4 6 bit matrix.
+  matrix_argb[0] = matrix_rgb[0] / 2;
+  matrix_argb[1] = matrix_rgb[1] / 2;
+  matrix_argb[2] = matrix_rgb[2] / 2;
+  matrix_argb[3] = matrix_rgb[3] / 2;
+  matrix_argb[4] = matrix_rgb[4] / 2;
+  matrix_argb[5] = matrix_rgb[5] / 2;
+  matrix_argb[6] = matrix_rgb[6] / 2;
+  matrix_argb[7] = matrix_rgb[7] / 2;
+  matrix_argb[8] = matrix_rgb[8] / 2;
+  matrix_argb[9] = matrix_rgb[9] / 2;
+  matrix_argb[10] = matrix_rgb[10] / 2;
+  matrix_argb[11] = matrix_rgb[11] / 2;
+  matrix_argb[14] = matrix_argb[13] = matrix_argb[12] = 0;
+  matrix_argb[15] = 64;  // 1.0
+
+  return ARGBColorMatrix((const uint8_t*)(dst), dst_stride_argb, dst,
+                         dst_stride_argb, &matrix_argb[0], width, height);
+}
+
+// Apply a color table each ARGB pixel.
+// Table contains 256 ARGB values.
+LIBYUV_API
+int ARGBColorTable(uint8_t* dst_argb,
+                   int dst_stride_argb,
+                   const uint8_t* table_argb,
+                   int dst_x,
+                   int dst_y,
+                   int width,
+                   int height) {
+  int y;
+  void (*ARGBColorTableRow)(uint8_t* dst_argb, const uint8_t* table_argb,
+                            int width) = ARGBColorTableRow_C;
+  uint8_t* dst = dst_argb + dst_y * dst_stride_argb + dst_x * 4;
+  if (!dst_argb || !table_argb || width <= 0 || height <= 0 || dst_x < 0 ||
+      dst_y < 0) {
+    return -1;
+  }
+  // Coalesce rows.
+  if (dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBCOLORTABLEROW_X86)
+  if (TestCpuFlag(kCpuHasX86)) {
+    ARGBColorTableRow = ARGBColorTableRow_X86;
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    ARGBColorTableRow(dst, table_argb, width);
+    dst += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Apply a color table each ARGB pixel but preserve destination alpha.
+// Table contains 256 ARGB values.
+LIBYUV_API
+int RGBColorTable(uint8_t* dst_argb,
+                  int dst_stride_argb,
+                  const uint8_t* table_argb,
+                  int dst_x,
+                  int dst_y,
+                  int width,
+                  int height) {
+  int y;
+  void (*RGBColorTableRow)(uint8_t* dst_argb, const uint8_t* table_argb,
+                           int width) = RGBColorTableRow_C;
+  uint8_t* dst = dst_argb + dst_y * dst_stride_argb + dst_x * 4;
+  if (!dst_argb || !table_argb || width <= 0 || height <= 0 || dst_x < 0 ||
+      dst_y < 0) {
+    return -1;
+  }
+  // Coalesce rows.
+  if (dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    dst_stride_argb = 0;
+  }
+#if defined(HAS_RGBCOLORTABLEROW_X86)
+  if (TestCpuFlag(kCpuHasX86)) {
+    RGBColorTableRow = RGBColorTableRow_X86;
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    RGBColorTableRow(dst, table_argb, width);
+    dst += dst_stride_argb;
+  }
+  return 0;
+}
+
+// ARGBQuantize is used to posterize art.
+// e.g. rgb / qvalue * qvalue + qvalue / 2
+// But the low levels implement efficiently with 3 parameters, and could be
+// used for other high level operations.
+// dst_argb[0] = (b * scale >> 16) * interval_size + interval_offset;
+// where scale is 1 / interval_size as a fixed point value.
+// The divide is replaces with a multiply by reciprocal fixed point multiply.
+// Caveat - although SSE2 saturates, the C function does not and should be used
+// with care if doing anything but quantization.
+LIBYUV_API
+int ARGBQuantize(uint8_t* dst_argb,
+                 int dst_stride_argb,
+                 int scale,
+                 int interval_size,
+                 int interval_offset,
+                 int dst_x,
+                 int dst_y,
+                 int width,
+                 int height) {
+  int y;
+  void (*ARGBQuantizeRow)(uint8_t* dst_argb, int scale, int interval_size,
+                          int interval_offset, int width) = ARGBQuantizeRow_C;
+  uint8_t* dst = dst_argb + dst_y * dst_stride_argb + dst_x * 4;
+  if (!dst_argb || width <= 0 || height <= 0 || dst_x < 0 || dst_y < 0 ||
+      interval_size < 1 || interval_size > 255) {
+    return -1;
+  }
+  // Coalesce rows.
+  if (dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBQUANTIZEROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2) && IS_ALIGNED(width, 4)) {
+    ARGBQuantizeRow = ARGBQuantizeRow_SSE2;
+  }
+#endif
+#if defined(HAS_ARGBQUANTIZEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(width, 8)) {
+    ARGBQuantizeRow = ARGBQuantizeRow_NEON;
+  }
+#endif
+#if defined(HAS_ARGBQUANTIZEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 8)) {
+    ARGBQuantizeRow = ARGBQuantizeRow_MSA;
+  }
+#endif
+#if defined(HAS_ARGBQUANTIZEROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
+    ARGBQuantizeRow = ARGBQuantizeRow_LSX;
+  }
+#endif
+  for (y = 0; y < height; ++y) {
+    ARGBQuantizeRow(dst, scale, interval_size, interval_offset, width);
+    dst += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Computes table of cumulative sum for image where the value is the sum
+// of all values above and to the left of the entry. Used by ARGBBlur.
+LIBYUV_API
+int ARGBComputeCumulativeSum(const uint8_t* src_argb,
+                             int src_stride_argb,
+                             int32_t* dst_cumsum,
+                             int dst_stride32_cumsum,
+                             int width,
+                             int height) {
+  int y;
+  void (*ComputeCumulativeSumRow)(const uint8_t* row, int32_t* cumsum,
+                                  const int32_t* previous_cumsum, int width) =
+      ComputeCumulativeSumRow_C;
+  int32_t* previous_cumsum = dst_cumsum;
+  if (!dst_cumsum || !src_argb || width <= 0 || height <= 0) {
+    return -1;
+  }
+#if defined(HAS_CUMULATIVESUMTOAVERAGEROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ComputeCumulativeSumRow = ComputeCumulativeSumRow_SSE2;
+  }
+#endif
+
+  memset(dst_cumsum, 0, width * sizeof(dst_cumsum[0]) * 4);  // 4 int per pixel.
+  for (y = 0; y < height; ++y) {
+    ComputeCumulativeSumRow(src_argb, dst_cumsum, previous_cumsum, width);
+    previous_cumsum = dst_cumsum;
+    dst_cumsum += dst_stride32_cumsum;
+    src_argb += src_stride_argb;
+  }
+  return 0;
+}
+
+// Blur ARGB image.
+// Caller should allocate CumulativeSum table of width * height * 16 bytes
+// aligned to 16 byte boundary. height can be radius * 2 + 2 to save memory
+// as the buffer is treated as circular.
+LIBYUV_API
+int ARGBBlur(const uint8_t* src_argb,
+             int src_stride_argb,
+             uint8_t* dst_argb,
+             int dst_stride_argb,
+             int32_t* dst_cumsum,
+             int dst_stride32_cumsum,
+             int width,
+             int height,
+             int radius) {
+  int y;
+  void (*ComputeCumulativeSumRow)(const uint8_t* row, int32_t* cumsum,
+                                  const int32_t* previous_cumsum, int width) =
+      ComputeCumulativeSumRow_C;
+  void (*CumulativeSumToAverageRow)(
+      const int32_t* topleft, const int32_t* botleft, int width, int area,
+      uint8_t* dst, int count) = CumulativeSumToAverageRow_C;
+  int32_t* cumsum_bot_row;
+  int32_t* max_cumsum_bot_row;
+  int32_t* cumsum_top_row;
+
+  if (!src_argb || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  if (radius > height) {
+    radius = height;
+  }
+  if (radius > (width / 2 - 1)) {
+    radius = width / 2 - 1;
+  }
+  if (radius <= 0 || height <= 1) {
+    return -1;
+  }
+#if defined(HAS_CUMULATIVESUMTOAVERAGEROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ComputeCumulativeSumRow = ComputeCumulativeSumRow_SSE2;
+    CumulativeSumToAverageRow = CumulativeSumToAverageRow_SSE2;
+  }
+#endif
+  // Compute enough CumulativeSum for first row to be blurred. After this
+  // one row of CumulativeSum is updated at a time.
+  ARGBComputeCumulativeSum(src_argb, src_stride_argb, dst_cumsum,
+                           dst_stride32_cumsum, width, radius);
+
+  src_argb = src_argb + radius * src_stride_argb;
+  cumsum_bot_row = &dst_cumsum[(radius - 1) * dst_stride32_cumsum];
+
+  max_cumsum_bot_row = &dst_cumsum[(radius * 2 + 2) * dst_stride32_cumsum];
+  cumsum_top_row = &dst_cumsum[0];
+
+  for (y = 0; y < height; ++y) {
+    int top_y = ((y - radius - 1) >= 0) ? (y - radius - 1) : 0;
+    int bot_y = ((y + radius) < height) ? (y + radius) : (height - 1);
+    int area = radius * (bot_y - top_y);
+    int boxwidth = radius * 4;
+    int x;
+    int n;
+
+    // Increment cumsum_top_row pointer with circular buffer wrap around.
+    if (top_y) {
+      cumsum_top_row += dst_stride32_cumsum;
+      if (cumsum_top_row >= max_cumsum_bot_row) {
+        cumsum_top_row = dst_cumsum;
+      }
+    }
+    // Increment cumsum_bot_row pointer with circular buffer wrap around and
+    // then fill in a row of CumulativeSum.
+    if ((y + radius) < height) {
+      const int32_t* prev_cumsum_bot_row = cumsum_bot_row;
+      cumsum_bot_row += dst_stride32_cumsum;
+      if (cumsum_bot_row >= max_cumsum_bot_row) {
+        cumsum_bot_row = dst_cumsum;
+      }
+      ComputeCumulativeSumRow(src_argb, cumsum_bot_row, prev_cumsum_bot_row,
+                              width);
+      src_argb += src_stride_argb;
+    }
+
+    // Left clipped.
+    for (x = 0; x < radius + 1; ++x) {
+      CumulativeSumToAverageRow(cumsum_top_row, cumsum_bot_row, boxwidth, area,
+                                &dst_argb[x * 4], 1);
+      area += (bot_y - top_y);
+      boxwidth += 4;
+    }
+
+    // Middle unclipped.
+    n = (width - 1) - radius - x + 1;
+    CumulativeSumToAverageRow(cumsum_top_row, cumsum_bot_row, boxwidth, area,
+                              &dst_argb[x * 4], n);
+
+    // Right clipped.
+    for (x += n; x <= width - 1; ++x) {
+      area -= (bot_y - top_y);
+      boxwidth -= 4;
+      CumulativeSumToAverageRow(cumsum_top_row + (x - radius - 1) * 4,
+                                cumsum_bot_row + (x - radius - 1) * 4, boxwidth,
+                                area, &dst_argb[x * 4], 1);
+    }
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Multiply ARGB image by a specified ARGB value.
+LIBYUV_API
+int ARGBShade(const uint8_t* src_argb,
+              int src_stride_argb,
+              uint8_t* dst_argb,
+              int dst_stride_argb,
+              int width,
+              int height,
+              uint32_t value) {
+  int y;
+  void (*ARGBShadeRow)(const uint8_t* src_argb, uint8_t* dst_argb, int width,
+                       uint32_t value) = ARGBShadeRow_C;
+  if (!src_argb || !dst_argb || width <= 0 || height == 0 || value == 0u) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBSHADEROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2) && IS_ALIGNED(width, 4)) {
+    ARGBShadeRow = ARGBShadeRow_SSE2;
+  }
+#endif
+#if defined(HAS_ARGBSHADEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(width, 8)) {
+    ARGBShadeRow = ARGBShadeRow_NEON;
+  }
+#endif
+#if defined(HAS_ARGBSHADEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 4)) {
+    ARGBShadeRow = ARGBShadeRow_MSA;
+  }
+#endif
+#if defined(HAS_ARGBSHADEROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 4)) {
+    ARGBShadeRow = ARGBShadeRow_LSX;
+  }
+#endif
+#if defined(HAS_ARGBSHADEROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX) && IS_ALIGNED(width, 8)) {
+    ARGBShadeRow = ARGBShadeRow_LASX;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBShadeRow(src_argb, dst_argb, width, value);
+    src_argb += src_stride_argb;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Interpolate 2 planes by specified amount (0 to 255).
+LIBYUV_API
+int InterpolatePlane(const uint8_t* src0,
+                     int src_stride0,
+                     const uint8_t* src1,
+                     int src_stride1,
+                     uint8_t* dst,
+                     int dst_stride,
+                     int width,
+                     int height,
+                     int interpolation) {
+  int y;
+  void (*InterpolateRow)(uint8_t* dst_ptr, const uint8_t* src_ptr,
+                         ptrdiff_t src_stride, int dst_width,
+                         int source_y_fraction) = InterpolateRow_C;
+  if (!src0 || !src1 || !dst || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst = dst + (height - 1) * dst_stride;
+    dst_stride = -dst_stride;
+  }
+  // Coalesce rows.
+  if (src_stride0 == width && src_stride1 == width && dst_stride == width) {
+    width *= height;
+    height = 1;
+    src_stride0 = src_stride1 = dst_stride = 0;
+  }
+#if defined(HAS_INTERPOLATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    InterpolateRow = InterpolateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      InterpolateRow = InterpolateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    InterpolateRow = InterpolateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      InterpolateRow = InterpolateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    InterpolateRow = InterpolateRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      InterpolateRow = InterpolateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    InterpolateRow = InterpolateRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      InterpolateRow = InterpolateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    InterpolateRow = InterpolateRow_Any_LSX;
+    if (IS_ALIGNED(width, 32)) {
+      InterpolateRow = InterpolateRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    InterpolateRow = InterpolateRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    InterpolateRow(dst, src0, src1 - src0, width, interpolation);
+    src0 += src_stride0;
+    src1 += src_stride1;
+    dst += dst_stride;
+  }
+  return 0;
+}
+
+// Interpolate 2 planes by specified amount (0 to 255).
+LIBYUV_API
+int InterpolatePlane_16(const uint16_t* src0,
+                        int src_stride0,
+                        const uint16_t* src1,
+                        int src_stride1,
+                        uint16_t* dst,
+                        int dst_stride,
+                        int width,
+                        int height,
+                        int interpolation) {
+  int y;
+  void (*InterpolateRow_16)(uint16_t* dst_ptr, const uint16_t* src_ptr,
+                            ptrdiff_t src_stride, int dst_width,
+                            int source_y_fraction) = InterpolateRow_16_C;
+  if (!src0 || !src1 || !dst || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst = dst + (height - 1) * dst_stride;
+    dst_stride = -dst_stride;
+  }
+  // Coalesce rows.
+  if (src_stride0 == width && src_stride1 == width && dst_stride == width) {
+    width *= height;
+    height = 1;
+    src_stride0 = src_stride1 = dst_stride = 0;
+  }
+#if defined(HAS_INTERPOLATEROW_16_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    InterpolateRow_16 = InterpolateRow_16_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      InterpolateRow_16 = InterpolateRow_16_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_16_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    InterpolateRow_16 = InterpolateRow_16_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      InterpolateRow_16 = InterpolateRow_16_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_16_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    InterpolateRow_16 = InterpolateRow_16_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      InterpolateRow_16 = InterpolateRow_16_NEON;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_16_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    InterpolateRow_16 = InterpolateRow_16_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      InterpolateRow_16 = InterpolateRow_16_MSA;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_16_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    InterpolateRow_16 = InterpolateRow_16_Any_LSX;
+    if (IS_ALIGNED(width, 32)) {
+      InterpolateRow_16 = InterpolateRow_16_LSX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    InterpolateRow_16(dst, src0, src1 - src0, width, interpolation);
+    src0 += src_stride0;
+    src1 += src_stride1;
+    dst += dst_stride;
+  }
+  return 0;
+}
+
+// Interpolate 2 ARGB images by specified amount (0 to 255).
+LIBYUV_API
+int ARGBInterpolate(const uint8_t* src_argb0,
+                    int src_stride_argb0,
+                    const uint8_t* src_argb1,
+                    int src_stride_argb1,
+                    uint8_t* dst_argb,
+                    int dst_stride_argb,
+                    int width,
+                    int height,
+                    int interpolation) {
+  return InterpolatePlane(src_argb0, src_stride_argb0, src_argb1,
+                          src_stride_argb1, dst_argb, dst_stride_argb,
+                          width * 4, height, interpolation);
+}
+
+// Interpolate 2 YUV images by specified amount (0 to 255).
+LIBYUV_API
+int I420Interpolate(const uint8_t* src0_y,
+                    int src0_stride_y,
+                    const uint8_t* src0_u,
+                    int src0_stride_u,
+                    const uint8_t* src0_v,
+                    int src0_stride_v,
+                    const uint8_t* src1_y,
+                    int src1_stride_y,
+                    const uint8_t* src1_u,
+                    int src1_stride_u,
+                    const uint8_t* src1_v,
+                    int src1_stride_v,
+                    uint8_t* dst_y,
+                    int dst_stride_y,
+                    uint8_t* dst_u,
+                    int dst_stride_u,
+                    uint8_t* dst_v,
+                    int dst_stride_v,
+                    int width,
+                    int height,
+                    int interpolation) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+
+  if (!src0_y || !src0_u || !src0_v || !src1_y || !src1_u || !src1_v ||
+      !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
+
+  InterpolatePlane(src0_y, src0_stride_y, src1_y, src1_stride_y, dst_y,
+                   dst_stride_y, width, height, interpolation);
+  InterpolatePlane(src0_u, src0_stride_u, src1_u, src1_stride_u, dst_u,
+                   dst_stride_u, halfwidth, halfheight, interpolation);
+  InterpolatePlane(src0_v, src0_stride_v, src1_v, src1_stride_v, dst_v,
+                   dst_stride_v, halfwidth, halfheight, interpolation);
+  return 0;
+}
+
+// Shuffle ARGB channel order.  e.g. BGRA to ARGB.
+LIBYUV_API
+int ARGBShuffle(const uint8_t* src_bgra,
+                int src_stride_bgra,
+                uint8_t* dst_argb,
+                int dst_stride_argb,
+                const uint8_t* shuffler,
+                int width,
+                int height) {
+  int y;
+  void (*ARGBShuffleRow)(const uint8_t* src_bgra, uint8_t* dst_argb,
+                         const uint8_t* shuffler, int width) = ARGBShuffleRow_C;
+  if (!src_bgra || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_bgra = src_bgra + (height - 1) * src_stride_bgra;
+    src_stride_bgra = -src_stride_bgra;
+  }
+  // Coalesce rows.
+  if (src_stride_bgra == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_bgra = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBSHUFFLEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBShuffleRow = ARGBShuffleRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBShuffleRow = ARGBShuffleRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSHUFFLEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBShuffleRow = ARGBShuffleRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBShuffleRow = ARGBShuffleRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSHUFFLEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBShuffleRow = ARGBShuffleRow_Any_NEON;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBShuffleRow = ARGBShuffleRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSHUFFLEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBShuffleRow = ARGBShuffleRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBShuffleRow = ARGBShuffleRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSHUFFLEROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBShuffleRow = ARGBShuffleRow_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBShuffleRow = ARGBShuffleRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSHUFFLEROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBShuffleRow = ARGBShuffleRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBShuffleRow = ARGBShuffleRow_LASX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBShuffleRow(src_bgra, dst_argb, shuffler, width);
+    src_bgra += src_stride_bgra;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Shuffle AR64 channel order.  e.g. AR64 to AB64.
+LIBYUV_API
+int AR64Shuffle(const uint16_t* src_ar64,
+                int src_stride_ar64,
+                uint16_t* dst_ar64,
+                int dst_stride_ar64,
+                const uint8_t* shuffler,
+                int width,
+                int height) {
+  int y;
+  void (*AR64ShuffleRow)(const uint8_t* src_ar64, uint8_t* dst_ar64,
+                         const uint8_t* shuffler, int width) = AR64ShuffleRow_C;
+  if (!src_ar64 || !dst_ar64 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_ar64 = src_ar64 + (height - 1) * src_stride_ar64;
+    src_stride_ar64 = -src_stride_ar64;
+  }
+  // Coalesce rows.
+  if (src_stride_ar64 == width * 4 && dst_stride_ar64 == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_ar64 = dst_stride_ar64 = 0;
+  }
+  // Assembly versions can be reused if it's implemented with shuffle.
+#if defined(HAS_ARGBSHUFFLEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    AR64ShuffleRow = ARGBShuffleRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      AR64ShuffleRow = ARGBShuffleRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSHUFFLEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    AR64ShuffleRow = ARGBShuffleRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      AR64ShuffleRow = ARGBShuffleRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBSHUFFLEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    AR64ShuffleRow = ARGBShuffleRow_Any_NEON;
+    if (IS_ALIGNED(width, 4)) {
+      AR64ShuffleRow = ARGBShuffleRow_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    AR64ShuffleRow((uint8_t*)(src_ar64), (uint8_t*)(dst_ar64), shuffler,
+                   width * 2);
+    src_ar64 += src_stride_ar64;
+    dst_ar64 += dst_stride_ar64;
+  }
+  return 0;
+}
+
+// Gauss blur a float plane using Gaussian 5x5 filter with
+// coefficients of 1, 4, 6, 4, 1.
+// Each destination pixel is a blur of the 5x5
+// pixels from the source.
+// Source edges are clamped.
+// Edge is 2 pixels on each side, and interior is multiple of 4.
+LIBYUV_API
+int GaussPlane_F32(const float* src,
+                   int src_stride,
+                   float* dst,
+                   int dst_stride,
+                   int width,
+                   int height) {
+  int y;
+  void (*GaussCol_F32)(const float* src0, const float* src1, const float* src2,
+                       const float* src3, const float* src4, float* dst,
+                       int width) = GaussCol_F32_C;
+  void (*GaussRow_F32)(const float* src, float* dst, int width) =
+      GaussRow_F32_C;
+  if (!src || !dst || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src = src + (height - 1) * src_stride;
+    src_stride = -src_stride;
+  }
+
+#if defined(HAS_GAUSSCOL_F32_NEON)
+  if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(width, 8)) {
+    GaussCol_F32 = GaussCol_F32_NEON;
+  }
+#endif
+#if defined(HAS_GAUSSROW_F32_NEON)
+  if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(width, 8)) {
+    GaussRow_F32 = GaussRow_F32_NEON;
+  }
+#endif
+  {
+    // 2 pixels on each side, but aligned out to 16 bytes.
+    align_buffer_64(rowbuf, (4 + width + 4) * 4);
+    if (!rowbuf)
+      return 1;
+    memset(rowbuf, 0, 16);
+    memset(rowbuf + (4 + width) * 4, 0, 16);
+    float* row = (float*)(rowbuf + 16);
+    const float* src0 = src;
+    const float* src1 = src;
+    const float* src2 = src;
+    const float* src3 = src2 + ((height > 1) ? src_stride : 0);
+    const float* src4 = src3 + ((height > 2) ? src_stride : 0);
+
+    for (y = 0; y < height; ++y) {
+      GaussCol_F32(src0, src1, src2, src3, src4, row, width);
+
+      // Extrude edge by 2 floats
+      row[-2] = row[-1] = row[0];
+      row[width + 1] = row[width] = row[width - 1];
+
+      GaussRow_F32(row - 2, dst, width);
+
+      src0 = src1;
+      src1 = src2;
+      src2 = src3;
+      src3 = src4;
+      if ((y + 2) < (height - 1)) {
+        src4 += src_stride;
+      }
+      dst += dst_stride;
+    }
+    free_aligned_buffer_64(rowbuf);
+  }
+  return 0;
+}
+
+// Sobel ARGB effect.
+static int ARGBSobelize(const uint8_t* src_argb,
+                        int src_stride_argb,
+                        uint8_t* dst_argb,
+                        int dst_stride_argb,
+                        int width,
+                        int height,
+                        void (*SobelRow)(const uint8_t* src_sobelx,
+                                         const uint8_t* src_sobely,
+                                         uint8_t* dst,
+                                         int width)) {
+  int y;
+  void (*ARGBToYJRow)(const uint8_t* src_argb, uint8_t* dst_g, int width) =
+      ARGBToYJRow_C;
+  void (*SobelYRow)(const uint8_t* src_y0, const uint8_t* src_y1,
+                    uint8_t* dst_sobely, int width) = SobelYRow_C;
+  void (*SobelXRow)(const uint8_t* src_y0, const uint8_t* src_y1,
+                    const uint8_t* src_y2, uint8_t* dst_sobely, int width) =
+      SobelXRow_C;
+  const int kEdge = 16;  // Extra pixels at start of row for extrude/align.
+  if (!src_argb || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+
+#if defined(HAS_ARGBTOYJROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    ARGBToYJRow = ARGBToYJRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBToYJRow = ARGBToYJRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYJRow = ARGBToYJRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBToYJRow = ARGBToYJRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBToYJRow = ARGBToYJRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBToYJRow = ARGBToYJRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBToYJRow = ARGBToYJRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      ARGBToYJRow = ARGBToYJRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYJRow = ARGBToYJRow_RVV;
+  }
+#endif
+
+#if defined(HAS_SOBELYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    SobelYRow = SobelYRow_SSE2;
+  }
+#endif
+#if defined(HAS_SOBELYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SobelYRow = SobelYRow_NEON;
+  }
+#endif
+#if defined(HAS_SOBELYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    SobelYRow = SobelYRow_MSA;
+  }
+#endif
+#if defined(HAS_SOBELXROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    SobelXRow = SobelXRow_SSE2;
+  }
+#endif
+#if defined(HAS_SOBELXROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SobelXRow = SobelXRow_NEON;
+  }
+#endif
+#if defined(HAS_SOBELXROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    SobelXRow = SobelXRow_MSA;
+  }
+#endif
+  {
+    // 3 rows with edges before/after.
+    const int row_size = (width + kEdge + 31) & ~31;
+    align_buffer_64(rows, row_size * 2 + (kEdge + row_size * 3 + kEdge));
+    uint8_t* row_sobelx = rows;
+    uint8_t* row_sobely = rows + row_size;
+    uint8_t* row_y = rows + row_size * 2;
+
+    // Convert first row.
+    uint8_t* row_y0 = row_y + kEdge;
+    uint8_t* row_y1 = row_y0 + row_size;
+    uint8_t* row_y2 = row_y1 + row_size;
+    if (!rows)
+      return 1;
+    ARGBToYJRow(src_argb, row_y0, width);
+    row_y0[-1] = row_y0[0];
+    memset(row_y0 + width, row_y0[width - 1], 16);  // Extrude 16 for valgrind.
+    ARGBToYJRow(src_argb, row_y1, width);
+    row_y1[-1] = row_y1[0];
+    memset(row_y1 + width, row_y1[width - 1], 16);
+    memset(row_y2 + width, 0, 16);
+
+    for (y = 0; y < height; ++y) {
+      // Convert next row of ARGB to G.
+      if (y < (height - 1)) {
+        src_argb += src_stride_argb;
+      }
+      ARGBToYJRow(src_argb, row_y2, width);
+      row_y2[-1] = row_y2[0];
+      row_y2[width] = row_y2[width - 1];
+
+      SobelXRow(row_y0 - 1, row_y1 - 1, row_y2 - 1, row_sobelx, width);
+      SobelYRow(row_y0 - 1, row_y2 - 1, row_sobely, width);
+      SobelRow(row_sobelx, row_sobely, dst_argb, width);
+
+      // Cycle thru circular queue of 3 row_y buffers.
+      {
+        uint8_t* row_yt = row_y0;
+        row_y0 = row_y1;
+        row_y1 = row_y2;
+        row_y2 = row_yt;
+      }
+
+      dst_argb += dst_stride_argb;
+    }
+    free_aligned_buffer_64(rows);
+  }
+  return 0;
+}
+
+// Sobel ARGB effect.
+LIBYUV_API
+int ARGBSobel(const uint8_t* src_argb,
+              int src_stride_argb,
+              uint8_t* dst_argb,
+              int dst_stride_argb,
+              int width,
+              int height) {
+  void (*SobelRow)(const uint8_t* src_sobelx, const uint8_t* src_sobely,
+                   uint8_t* dst_argb, int width) = SobelRow_C;
+#if defined(HAS_SOBELROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    SobelRow = SobelRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      SobelRow = SobelRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_SOBELROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SobelRow = SobelRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      SobelRow = SobelRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_SOBELROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    SobelRow = SobelRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      SobelRow = SobelRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_SOBELROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    SobelRow = SobelRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      SobelRow = SobelRow_LSX;
+    }
+  }
+#endif
+  return ARGBSobelize(src_argb, src_stride_argb, dst_argb, dst_stride_argb,
+                      width, height, SobelRow);
+}
+
+// Sobel ARGB effect with planar output.
+LIBYUV_API
+int ARGBSobelToPlane(const uint8_t* src_argb,
+                     int src_stride_argb,
+                     uint8_t* dst_y,
+                     int dst_stride_y,
+                     int width,
+                     int height) {
+  void (*SobelToPlaneRow)(const uint8_t* src_sobelx, const uint8_t* src_sobely,
+                          uint8_t* dst_, int width) = SobelToPlaneRow_C;
+#if defined(HAS_SOBELTOPLANEROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    SobelToPlaneRow = SobelToPlaneRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      SobelToPlaneRow = SobelToPlaneRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_SOBELTOPLANEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SobelToPlaneRow = SobelToPlaneRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      SobelToPlaneRow = SobelToPlaneRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_SOBELTOPLANEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    SobelToPlaneRow = SobelToPlaneRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      SobelToPlaneRow = SobelToPlaneRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_SOBELTOPLANEROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    SobelToPlaneRow = SobelToPlaneRow_Any_LSX;
+    if (IS_ALIGNED(width, 32)) {
+      SobelToPlaneRow = SobelToPlaneRow_LSX;
+    }
+  }
+#endif
+  return ARGBSobelize(src_argb, src_stride_argb, dst_y, dst_stride_y, width,
+                      height, SobelToPlaneRow);
+}
+
+// SobelXY ARGB effect.
+// Similar to Sobel, but also stores Sobel X in R and Sobel Y in B.  G = Sobel.
+LIBYUV_API
+int ARGBSobelXY(const uint8_t* src_argb,
+                int src_stride_argb,
+                uint8_t* dst_argb,
+                int dst_stride_argb,
+                int width,
+                int height) {
+  void (*SobelXYRow)(const uint8_t* src_sobelx, const uint8_t* src_sobely,
+                     uint8_t* dst_argb, int width) = SobelXYRow_C;
+#if defined(HAS_SOBELXYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    SobelXYRow = SobelXYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      SobelXYRow = SobelXYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_SOBELXYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SobelXYRow = SobelXYRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      SobelXYRow = SobelXYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_SOBELXYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    SobelXYRow = SobelXYRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      SobelXYRow = SobelXYRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_SOBELXYROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    SobelXYRow = SobelXYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      SobelXYRow = SobelXYRow_LSX;
+    }
+  }
+#endif
+  return ARGBSobelize(src_argb, src_stride_argb, dst_argb, dst_stride_argb,
+                      width, height, SobelXYRow);
+}
+
+// Apply a 4x4 polynomial to each ARGB pixel.
+LIBYUV_API
+int ARGBPolynomial(const uint8_t* src_argb,
+                   int src_stride_argb,
+                   uint8_t* dst_argb,
+                   int dst_stride_argb,
+                   const float* poly,
+                   int width,
+                   int height) {
+  int y;
+  void (*ARGBPolynomialRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                            const float* poly, int width) = ARGBPolynomialRow_C;
+  if (!src_argb || !dst_argb || !poly || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBPOLYNOMIALROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2) && IS_ALIGNED(width, 2)) {
+    ARGBPolynomialRow = ARGBPolynomialRow_SSE2;
+  }
+#endif
+#if defined(HAS_ARGBPOLYNOMIALROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2) && TestCpuFlag(kCpuHasFMA3) &&
+      IS_ALIGNED(width, 2)) {
+    ARGBPolynomialRow = ARGBPolynomialRow_AVX2;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBPolynomialRow(src_argb, dst_argb, poly, width);
+    src_argb += src_stride_argb;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Convert plane of 16 bit shorts to half floats.
+// Source values are multiplied by scale before storing as half float.
+LIBYUV_API
+int HalfFloatPlane(const uint16_t* src_y,
+                   int src_stride_y,
+                   uint16_t* dst_y,
+                   int dst_stride_y,
+                   float scale,
+                   int width,
+                   int height) {
+  int y;
+  void (*HalfFloatRow)(const uint16_t* src, uint16_t* dst, float scale,
+                       int width) = HalfFloatRow_C;
+  if (!src_y || !dst_y || width <= 0 || height == 0) {
+    return -1;
+  }
+  src_stride_y >>= 1;
+  dst_stride_y >>= 1;
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_stride_y = -src_stride_y;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && dst_stride_y == width) {
+    width *= height;
+    height = 1;
+    src_stride_y = dst_stride_y = 0;
+  }
+#if defined(HAS_HALFFLOATROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    HalfFloatRow = HalfFloatRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      HalfFloatRow = HalfFloatRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_HALFFLOATROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    HalfFloatRow = HalfFloatRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      HalfFloatRow = HalfFloatRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_HALFFLOATROW_F16C)
+  if (TestCpuFlag(kCpuHasAVX2) && TestCpuFlag(kCpuHasF16C)) {
+    HalfFloatRow =
+        (scale == 1.0f) ? HalfFloat1Row_Any_F16C : HalfFloatRow_Any_F16C;
+    if (IS_ALIGNED(width, 16)) {
+      HalfFloatRow = (scale == 1.0f) ? HalfFloat1Row_F16C : HalfFloatRow_F16C;
+    }
+  }
+#endif
+#if defined(HAS_HALFFLOATROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    HalfFloatRow =
+        (scale == 1.0f) ? HalfFloat1Row_Any_NEON : HalfFloatRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      HalfFloatRow = (scale == 1.0f) ? HalfFloat1Row_NEON : HalfFloatRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_HALFFLOATROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    HalfFloatRow = HalfFloatRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      HalfFloatRow = HalfFloatRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_HALFFLOATROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    HalfFloatRow = HalfFloatRow_Any_LSX;
+    if (IS_ALIGNED(width, 32)) {
+      HalfFloatRow = HalfFloatRow_LSX;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    HalfFloatRow(src_y, dst_y, scale, width);
+    src_y += src_stride_y;
+    dst_y += dst_stride_y;
+  }
+  return 0;
+}
+
+// Convert a buffer of bytes to floats, scale the values and store as floats.
+LIBYUV_API
+int ByteToFloat(const uint8_t* src_y, float* dst_y, float scale, int width) {
+  void (*ByteToFloatRow)(const uint8_t* src, float* dst, float scale,
+                         int width) = ByteToFloatRow_C;
+  if (!src_y || !dst_y || width <= 0) {
+    return -1;
+  }
+#if defined(HAS_BYTETOFLOATROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ByteToFloatRow = ByteToFloatRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ByteToFloatRow = ByteToFloatRow_NEON;
+    }
+  }
+#endif
+
+  ByteToFloatRow(src_y, dst_y, scale, width);
+  return 0;
+}
+
+// Apply a lumacolortable to each ARGB pixel.
+LIBYUV_API
+int ARGBLumaColorTable(const uint8_t* src_argb,
+                       int src_stride_argb,
+                       uint8_t* dst_argb,
+                       int dst_stride_argb,
+                       const uint8_t* luma,
+                       int width,
+                       int height) {
+  int y;
+  void (*ARGBLumaColorTableRow)(
+      const uint8_t* src_argb, uint8_t* dst_argb, int width,
+      const uint8_t* luma, const uint32_t lumacoeff) = ARGBLumaColorTableRow_C;
+  if (!src_argb || !dst_argb || !luma || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBLUMACOLORTABLEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3) && IS_ALIGNED(width, 4)) {
+    ARGBLumaColorTableRow = ARGBLumaColorTableRow_SSSE3;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBLumaColorTableRow(src_argb, dst_argb, width, luma, 0x00264b0f);
+    src_argb += src_stride_argb;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Copy Alpha from one ARGB image to another.
+LIBYUV_API
+int ARGBCopyAlpha(const uint8_t* src_argb,
+                  int src_stride_argb,
+                  uint8_t* dst_argb,
+                  int dst_stride_argb,
+                  int width,
+                  int height) {
+  int y;
+  void (*ARGBCopyAlphaRow)(const uint8_t* src_argb, uint8_t* dst_argb,
+                           int width) = ARGBCopyAlphaRow_C;
+  if (!src_argb || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBCOPYALPHAROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBCopyAlphaRow = ARGBCopyAlphaRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBCopyAlphaRow = ARGBCopyAlphaRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBCOPYALPHAROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBCopyAlphaRow = ARGBCopyAlphaRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBCopyAlphaRow = ARGBCopyAlphaRow_AVX2;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBCopyAlphaRow(src_argb, dst_argb, width);
+    src_argb += src_stride_argb;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+// Extract just the alpha channel from ARGB.
+LIBYUV_API
+int ARGBExtractAlpha(const uint8_t* src_argb,
+                     int src_stride_argb,
+                     uint8_t* dst_a,
+                     int dst_stride_a,
+                     int width,
+                     int height) {
+  if (!src_argb || !dst_a || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb += (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_argb == width * 4 && dst_stride_a == width) {
+    width *= height;
+    height = 1;
+    src_stride_argb = dst_stride_a = 0;
+  }
+  void (*ARGBExtractAlphaRow)(const uint8_t* src_argb, uint8_t* dst_a,
+                              int width) = ARGBExtractAlphaRow_C;
+#if defined(HAS_ARGBEXTRACTALPHAROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBExtractAlphaRow = IS_ALIGNED(width, 8) ? ARGBExtractAlphaRow_SSE2
+                                               : ARGBExtractAlphaRow_Any_SSE2;
+  }
+#endif
+#if defined(HAS_ARGBEXTRACTALPHAROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBExtractAlphaRow = IS_ALIGNED(width, 32) ? ARGBExtractAlphaRow_AVX2
+                                                : ARGBExtractAlphaRow_Any_AVX2;
+  }
+#endif
+#if defined(HAS_ARGBEXTRACTALPHAROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_NEON
+                                                : ARGBExtractAlphaRow_Any_NEON;
+  }
+#endif
+#if defined(HAS_ARGBEXTRACTALPHAROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_MSA
+                                                : ARGBExtractAlphaRow_Any_MSA;
+  }
+#endif
+#if defined(HAS_ARGBEXTRACTALPHAROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_LSX
+                                                : ARGBExtractAlphaRow_Any_LSX;
+  }
+#endif
+#if defined(HAS_ARGBEXTRACTALPHAROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBExtractAlphaRow = ARGBExtractAlphaRow_RVV;
+  }
+#endif
+
+  for (int y = 0; y < height; ++y) {
+    ARGBExtractAlphaRow(src_argb, dst_a, width);
+    src_argb += src_stride_argb;
+    dst_a += dst_stride_a;
+  }
+  return 0;
+}
+
+// Copy a planar Y channel to the alpha channel of a destination ARGB image.
+LIBYUV_API
+int ARGBCopyYToAlpha(const uint8_t* src_y,
+                     int src_stride_y,
+                     uint8_t* dst_argb,
+                     int dst_stride_argb,
+                     int width,
+                     int height) {
+  int y;
+  void (*ARGBCopyYToAlphaRow)(const uint8_t* src_y, uint8_t* dst_argb,
+                              int width) = ARGBCopyYToAlphaRow_C;
+  if (!src_y || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_stride_y = -src_stride_y;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width && dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_y = dst_stride_argb = 0;
+  }
+#if defined(HAS_ARGBCOPYYTOALPHAROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBCopyYToAlphaRow = ARGBCopyYToAlphaRow_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBCopyYToAlphaRow = ARGBCopyYToAlphaRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBCOPYYTOALPHAROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBCopyYToAlphaRow = ARGBCopyYToAlphaRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBCopyYToAlphaRow = ARGBCopyYToAlphaRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBCOPYYTOALPHAROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBCopyYToAlphaRow = ARGBCopyYToAlphaRow_RVV;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    ARGBCopyYToAlphaRow(src_y, dst_argb, width);
+    src_y += src_stride_y;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
+}
+
+LIBYUV_API
+int YUY2ToNV12(const uint8_t* src_yuy2,
+               int src_stride_yuy2,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int y;
+  void (*YUY2ToYRow)(const uint8_t* src_yuy2, uint8_t* dst_y, int width) =
+      YUY2ToYRow_C;
+  void (*YUY2ToNVUVRow)(const uint8_t* src_yuy2, int stride_yuy2,
+                        uint8_t* dst_uv, int width) = YUY2ToNVUVRow_C;
+  if (!src_yuy2 || !dst_y || !dst_uv || width <= 0 || height == 0) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_yuy2 = src_yuy2 + (height - 1) * src_stride_yuy2;
+    src_stride_yuy2 = -src_stride_yuy2;
+  }
+#if defined(HAS_YUY2TOYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    YUY2ToYRow = YUY2ToYRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToYRow = YUY2ToYRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    YUY2ToYRow = YUY2ToYRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToYRow = YUY2ToYRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    YUY2ToYRow = YUY2ToYRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToYRow = YUY2ToYRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_MSA) && defined(HAS_YUY2TOUV422ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    YUY2ToYRow = YUY2ToYRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToYRow = YUY2ToYRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_LSX) && defined(HAS_YUY2TOUV422ROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    YUY2ToYRow = YUY2ToYRow_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToYRow = YUY2ToYRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TOYROW_LASX) && defined(HAS_YUY2TOUV422ROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    YUY2ToYRow = YUY2ToYRow_Any_LASX;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToYRow = YUY2ToYRow_LASX;
+    }
+  }
+#endif
+
+#if defined(HAS_YUY2TONVUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    YUY2ToNVUVRow = YUY2ToNVUVRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToNVUVRow = YUY2ToNVUVRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TONVUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    YUY2ToNVUVRow = YUY2ToNVUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      YUY2ToNVUVRow = YUY2ToNVUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_YUY2TONVUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    YUY2ToNVUVRow = YUY2ToNVUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      YUY2ToNVUVRow = YUY2ToNVUVRow_NEON;
+    }
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    YUY2ToYRow(src_yuy2, dst_y, width);
+    YUY2ToYRow(src_yuy2 + src_stride_yuy2, dst_y + dst_stride_y, width);
+    YUY2ToNVUVRow(src_yuy2, src_stride_yuy2, dst_uv, width);
+    src_yuy2 += src_stride_yuy2 * 2;
+    dst_y += dst_stride_y * 2;
+    dst_uv += dst_stride_uv;
+  }
+  if (height & 1) {
+    YUY2ToYRow(src_yuy2, dst_y, width);
+    YUY2ToNVUVRow(src_yuy2, 0, dst_uv, width);
+  }
+  return 0;
+}
+
+LIBYUV_API
+int UYVYToNV12(const uint8_t* src_uyvy,
+               int src_stride_uyvy,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_uv,
+               int dst_stride_uv,
+               int width,
+               int height) {
+  int y;
+  int halfwidth = (width + 1) >> 1;
+  void (*SplitUVRow)(const uint8_t* src_uv, uint8_t* dst_u, uint8_t* dst_v,
+                     int width) = SplitUVRow_C;
+  void (*InterpolateRow)(uint8_t* dst_ptr, const uint8_t* src_ptr,
+                         ptrdiff_t src_stride, int dst_width,
+                         int source_y_fraction) = InterpolateRow_C;
+
+  if (!src_uyvy || !dst_y || !dst_uv || width <= 0 || height == 0) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_uyvy = src_uyvy + (height - 1) * src_stride_uyvy;
+    src_stride_uyvy = -src_stride_uyvy;
+  }
+#if defined(HAS_SPLITUVROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    SplitUVRow = SplitUVRow_Any_SSE2;
+    if (IS_ALIGNED(width, 16)) {
+      SplitUVRow = SplitUVRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_SPLITUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    SplitUVRow = SplitUVRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      SplitUVRow = SplitUVRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_SPLITUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    SplitUVRow = SplitUVRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      SplitUVRow = SplitUVRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_SPLITUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    SplitUVRow = SplitUVRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      SplitUVRow = SplitUVRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_SPLITUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    SplitUVRow = SplitUVRow_Any_LSX;
+    if (IS_ALIGNED(width, 32)) {
+      SplitUVRow = SplitUVRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_SPLITUVROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    SplitUVRow = SplitUVRow_RVV;
+  }
+#endif
+
+#if defined(HAS_INTERPOLATEROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    InterpolateRow = InterpolateRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      InterpolateRow = InterpolateRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    InterpolateRow = InterpolateRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      InterpolateRow = InterpolateRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    InterpolateRow = InterpolateRow_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      InterpolateRow = InterpolateRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    InterpolateRow = InterpolateRow_Any_MSA;
+    if (IS_ALIGNED(width, 32)) {
+      InterpolateRow = InterpolateRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    InterpolateRow = InterpolateRow_Any_LSX;
+    if (IS_ALIGNED(width, 32)) {
+      InterpolateRow = InterpolateRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_INTERPOLATEROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    InterpolateRow = InterpolateRow_RVV;
+  }
+#endif
+
+  {
+    int awidth = halfwidth * 2;
+    // row of y and 2 rows of uv
+    align_buffer_64(rows, awidth * 3);
+    if (!rows)
+      return 1;
+
+    for (y = 0; y < height - 1; y += 2) {
+      // Split Y from UV.
+      SplitUVRow(src_uyvy, rows + awidth, rows, awidth);
+      memcpy(dst_y, rows, width);
+      SplitUVRow(src_uyvy + src_stride_uyvy, rows + awidth * 2, rows, awidth);
+      memcpy(dst_y + dst_stride_y, rows, width);
+      InterpolateRow(dst_uv, rows + awidth, awidth, awidth, 128);
+      src_uyvy += src_stride_uyvy * 2;
+      dst_y += dst_stride_y * 2;
+      dst_uv += dst_stride_uv;
+    }
+    if (height & 1) {
+      // Split Y from UV.
+      SplitUVRow(src_uyvy, dst_uv, rows, awidth);
+      memcpy(dst_y, rows, width);
+    }
+    free_aligned_buffer_64(rows);
+  }
+  return 0;
+}
+
+// width and height are src size allowing odd size handling.
+LIBYUV_API
+void HalfMergeUVPlane(const uint8_t* src_u,
+                      int src_stride_u,
+                      const uint8_t* src_v,
+                      int src_stride_v,
+                      uint8_t* dst_uv,
+                      int dst_stride_uv,
+                      int width,
+                      int height) {
+  int y;
+  void (*HalfMergeUVRow)(const uint8_t* src_u, int src_stride_u,
+                         const uint8_t* src_v, int src_stride_v,
+                         uint8_t* dst_uv, int width) = HalfMergeUVRow_C;
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+#if defined(HAS_HALFMERGEUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(width, 16)) {
+    HalfMergeUVRow = HalfMergeUVRow_NEON;
+  }
+#endif
+#if defined(HAS_HALFMERGEUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3) && IS_ALIGNED(width, 16)) {
+    HalfMergeUVRow = HalfMergeUVRow_SSSE3;
+  }
+#endif
+#if defined(HAS_HALFMERGEUVROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2) && IS_ALIGNED(width, 32)) {
+    HalfMergeUVRow = HalfMergeUVRow_AVX2;
+  }
+#endif
+
+  for (y = 0; y < height - 1; y += 2) {
+    // Merge a row of U and V into a row of UV.
+    HalfMergeUVRow(src_u, src_stride_u, src_v, src_stride_v, dst_uv, width);
+    src_u += src_stride_u * 2;
+    src_v += src_stride_v * 2;
+    dst_uv += dst_stride_uv;
+  }
+  if (height & 1) {
+    HalfMergeUVRow(src_u, 0, src_v, 0, dst_uv, width);
+  }
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/rotate.cc

@@ -0,0 +1,1276 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include <assert.h>
+
+#include "libyuv/rotate.h"
+
+#include "libyuv/convert.h"
+#include "libyuv/cpu_id.h"
+#include "libyuv/planar_functions.h"
+#include "libyuv/rotate_row.h"
+#include "libyuv/row.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+LIBYUV_API
+void TransposePlane(const uint8_t* src,
+                    int src_stride,
+                    uint8_t* dst,
+                    int dst_stride,
+                    int width,
+                    int height) {
+  int i = height;
+#if defined(HAS_TRANSPOSEWXH_SME)
+  void (*TransposeWxH)(const uint8_t* src, int src_stride, uint8_t* dst,
+                       int dst_stride, int width, int height) = NULL;
+#endif
+#if defined(HAS_TRANSPOSEWX16_MSA) || defined(HAS_TRANSPOSEWX16_LSX) || \
+    defined(HAS_TRANSPOSEWX16_NEON)
+  void (*TransposeWx16)(const uint8_t* src, int src_stride, uint8_t* dst,
+                        int dst_stride, int width) = TransposeWx16_C;
+#else
+  void (*TransposeWx8)(const uint8_t* src, int src_stride, uint8_t* dst,
+                       int dst_stride, int width) = TransposeWx8_C;
+#endif
+
+#if defined(HAS_TRANSPOSEWX8_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    TransposeWx8 = TransposeWx8_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      TransposeWx8 = TransposeWx8_NEON;
+    }
+  }
+#endif
+#if defined(HAS_TRANSPOSEWX16_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    TransposeWx16 = TransposeWx16_Any_NEON;
+    if (IS_ALIGNED(width, 16)) {
+      TransposeWx16 = TransposeWx16_NEON;
+    }
+  }
+#endif
+#if defined(HAS_TRANSPOSEWXH_SME)
+  if (TestCpuFlag(kCpuHasSME)) {
+    TransposeWxH = TransposeWxH_SME;
+  }
+#endif
+#if defined(HAS_TRANSPOSEWX8_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    TransposeWx8 = TransposeWx8_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      TransposeWx8 = TransposeWx8_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_TRANSPOSEWX8_FAST_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    TransposeWx8 = TransposeWx8_Fast_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      TransposeWx8 = TransposeWx8_Fast_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_TRANSPOSEWX16_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    TransposeWx16 = TransposeWx16_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      TransposeWx16 = TransposeWx16_MSA;
+    }
+  }
+#endif
+#if defined(HAS_TRANSPOSEWX16_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    TransposeWx16 = TransposeWx16_Any_LSX;
+    if (IS_ALIGNED(width, 16)) {
+      TransposeWx16 = TransposeWx16_LSX;
+    }
+  }
+#endif
+
+#if defined(HAS_TRANSPOSEWXH_SME)
+  if (TransposeWxH) {
+    TransposeWxH(src, src_stride, dst, dst_stride, width, height);
+    return;
+  }
+#endif
+#if defined(HAS_TRANSPOSEWX16_MSA) || defined(HAS_TRANSPOSEWX16_LSX) || \
+    defined(HAS_TRANSPOSEWX16_NEON)
+  // Work across the source in 16x16 tiles
+  while (i >= 16) {
+    TransposeWx16(src, src_stride, dst, dst_stride, width);
+    src += 16 * src_stride;  // Go down 16 rows.
+    dst += 16;               // Move over 16 columns.
+    i -= 16;
+  }
+#else
+  // Work across the source in 8x8 tiles
+  while (i >= 8) {
+    TransposeWx8(src, src_stride, dst, dst_stride, width);
+    src += 8 * src_stride;  // Go down 8 rows.
+    dst += 8;               // Move over 8 columns.
+    i -= 8;
+  }
+#endif
+
+  if (i > 0) {
+    TransposeWxH_C(src, src_stride, dst, dst_stride, width, i);
+  }
+}
+
+LIBYUV_API
+void RotatePlane90(const uint8_t* src,
+                   int src_stride,
+                   uint8_t* dst,
+                   int dst_stride,
+                   int width,
+                   int height) {
+  // Rotate by 90 is a transpose with the source read
+  // from bottom to top. So set the source pointer to the end
+  // of the buffer and flip the sign of the source stride.
+  src += src_stride * (height - 1);
+  src_stride = -src_stride;
+  TransposePlane(src, src_stride, dst, dst_stride, width, height);
+}
+
+LIBYUV_API
+void RotatePlane270(const uint8_t* src,
+                    int src_stride,
+                    uint8_t* dst,
+                    int dst_stride,
+                    int width,
+                    int height) {
+  // Rotate by 270 is a transpose with the destination written
+  // from bottom to top. So set the destination pointer to the end
+  // of the buffer and flip the sign of the destination stride.
+  dst += dst_stride * (width - 1);
+  dst_stride = -dst_stride;
+  TransposePlane(src, src_stride, dst, dst_stride, width, height);
+}
+
+LIBYUV_API
+void RotatePlane180(const uint8_t* src,
+                    int src_stride,
+                    uint8_t* dst,
+                    int dst_stride,
+                    int width,
+                    int height) {
+  // Swap top and bottom row and mirror the content. Uses a temporary row.
+  align_buffer_64(row, width);
+  assert(row);
+  if (!row)
+    return;
+  const uint8_t* src_bot = src + src_stride * (height - 1);
+  uint8_t* dst_bot = dst + dst_stride * (height - 1);
+  int half_height = (height + 1) >> 1;
+  int y;
+  void (*MirrorRow)(const uint8_t* src, uint8_t* dst, int width) = MirrorRow_C;
+  void (*CopyRow)(const uint8_t* src, uint8_t* dst, int width) = CopyRow_C;
+#if defined(HAS_MIRRORROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    MirrorRow = MirrorRow_Any_NEON;
+    if (IS_ALIGNED(width, 32)) {
+      MirrorRow = MirrorRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    MirrorRow = MirrorRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 16)) {
+      MirrorRow = MirrorRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    MirrorRow = MirrorRow_Any_AVX2;
+    if (IS_ALIGNED(width, 32)) {
+      MirrorRow = MirrorRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    MirrorRow = MirrorRow_Any_MSA;
+    if (IS_ALIGNED(width, 64)) {
+      MirrorRow = MirrorRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    MirrorRow = MirrorRow_Any_LSX;
+    if (IS_ALIGNED(width, 32)) {
+      MirrorRow = MirrorRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_MIRRORROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    MirrorRow = MirrorRow_Any_LASX;
+    if (IS_ALIGNED(width, 64)) {
+      MirrorRow = MirrorRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_COPYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    CopyRow = IS_ALIGNED(width, 32) ? CopyRow_SSE2 : CopyRow_Any_SSE2;
+  }
+#endif
+#if defined(HAS_COPYROW_AVX)
+  if (TestCpuFlag(kCpuHasAVX)) {
+    CopyRow = IS_ALIGNED(width, 64) ? CopyRow_AVX : CopyRow_Any_AVX;
+  }
+#endif
+#if defined(HAS_COPYROW_ERMS)
+  if (TestCpuFlag(kCpuHasERMS)) {
+    CopyRow = CopyRow_ERMS;
+  }
+#endif
+#if defined(HAS_COPYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    CopyRow = IS_ALIGNED(width, 32) ? CopyRow_NEON : CopyRow_Any_NEON;
+  }
+#endif
+#if defined(HAS_COPYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    CopyRow = CopyRow_RVV;
+  }
+#endif
+
+  // Odd height will harmlessly mirror the middle row twice.
+  for (y = 0; y < half_height; ++y) {
+    CopyRow(src, row, width);        // Copy top row into buffer
+    MirrorRow(src_bot, dst, width);  // Mirror bottom row into top row
+    MirrorRow(row, dst_bot, width);  // Mirror buffer into bottom row
+    src += src_stride;
+    dst += dst_stride;
+    src_bot -= src_stride;
+    dst_bot -= dst_stride;
+  }
+  free_aligned_buffer_64(row);
+}
+
+LIBYUV_API
+void SplitTransposeUV(const uint8_t* src,
+                      int src_stride,
+                      uint8_t* dst_a,
+                      int dst_stride_a,
+                      uint8_t* dst_b,
+                      int dst_stride_b,
+                      int width,
+                      int height) {
+  int i = height;
+#if defined(HAS_TRANSPOSEUVWXH_SME)
+  void (*TransposeUVWxH)(const uint8_t* src, int src_stride, uint8_t* dst_a,
+                         int dst_stride_a, uint8_t* dst_b, int dst_stride_b,
+                         int width, int height) = TransposeUVWxH_C;
+#endif
+#if defined(HAS_TRANSPOSEUVWX16_MSA)
+  void (*TransposeUVWx16)(const uint8_t* src, int src_stride, uint8_t* dst_a,
+                          int dst_stride_a, uint8_t* dst_b, int dst_stride_b,
+                          int width) = TransposeUVWx16_C;
+#elif defined(HAS_TRANSPOSEUVWX16_LSX)
+  void (*TransposeUVWx16)(const uint8_t* src, int src_stride, uint8_t* dst_a,
+                          int dst_stride_a, uint8_t* dst_b, int dst_stride_b,
+                          int width) = TransposeUVWx16_C;
+#else
+  void (*TransposeUVWx8)(const uint8_t* src, int src_stride, uint8_t* dst_a,
+                         int dst_stride_a, uint8_t* dst_b, int dst_stride_b,
+                         int width) = TransposeUVWx8_C;
+#endif
+
+#if defined(HAS_TRANSPOSEUVWX16_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    TransposeUVWx16 = TransposeUVWx16_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      TransposeUVWx16 = TransposeUVWx16_MSA;
+    }
+  }
+#elif defined(HAS_TRANSPOSEUVWX16_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    TransposeUVWx16 = TransposeUVWx16_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      TransposeUVWx16 = TransposeUVWx16_LSX;
+    }
+  }
+#else
+#if defined(HAS_TRANSPOSEUVWX8_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    TransposeUVWx8 = TransposeUVWx8_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      TransposeUVWx8 = TransposeUVWx8_NEON;
+    }
+  }
+#endif
+#if defined(HAS_TRANSPOSEUVWXH_SME)
+  if (TestCpuFlag(kCpuHasSME)) {
+    TransposeUVWxH = TransposeUVWxH_SME;
+  }
+#endif
+#if defined(HAS_TRANSPOSEUVWX8_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    TransposeUVWx8 = TransposeUVWx8_Any_SSE2;
+    if (IS_ALIGNED(width, 8)) {
+      TransposeUVWx8 = TransposeUVWx8_SSE2;
+    }
+  }
+#endif
+#endif /* defined(HAS_TRANSPOSEUVWX16_MSA) */
+
+#if defined(HAS_TRANSPOSEUVWXH_SME)
+  if (TestCpuFlag(kCpuHasSME)) {
+    TransposeUVWxH(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
+                   width, i);
+    return;
+  }
+#endif
+#if defined(HAS_TRANSPOSEUVWX16_MSA)
+  // Work through the source in 8x8 tiles.
+  while (i >= 16) {
+    TransposeUVWx16(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
+                    width);
+    src += 16 * src_stride;  // Go down 16 rows.
+    dst_a += 16;             // Move over 8 columns.
+    dst_b += 16;             // Move over 8 columns.
+    i -= 16;
+  }
+#elif defined(HAS_TRANSPOSEUVWX16_LSX)
+  // Work through the source in 8x8 tiles.
+  while (i >= 16) {
+    TransposeUVWx16(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
+                    width);
+    src += 16 * src_stride;  // Go down 16 rows.
+    dst_a += 16;             // Move over 8 columns.
+    dst_b += 16;             // Move over 8 columns.
+    i -= 16;
+  }
+#else
+  // Work through the source in 8x8 tiles.
+  while (i >= 8) {
+    TransposeUVWx8(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
+                   width);
+    src += 8 * src_stride;  // Go down 8 rows.
+    dst_a += 8;             // Move over 8 columns.
+    dst_b += 8;             // Move over 8 columns.
+    i -= 8;
+  }
+#endif
+
+  if (i > 0) {
+    TransposeUVWxH_C(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
+                     width, i);
+  }
+}
+
+LIBYUV_API
+void SplitRotateUV90(const uint8_t* src,
+                     int src_stride,
+                     uint8_t* dst_a,
+                     int dst_stride_a,
+                     uint8_t* dst_b,
+                     int dst_stride_b,
+                     int width,
+                     int height) {
+  src += src_stride * (height - 1);
+  src_stride = -src_stride;
+
+  SplitTransposeUV(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
+                   width, height);
+}
+
+LIBYUV_API
+void SplitRotateUV270(const uint8_t* src,
+                      int src_stride,
+                      uint8_t* dst_a,
+                      int dst_stride_a,
+                      uint8_t* dst_b,
+                      int dst_stride_b,
+                      int width,
+                      int height) {
+  dst_a += dst_stride_a * (width - 1);
+  dst_b += dst_stride_b * (width - 1);
+  dst_stride_a = -dst_stride_a;
+  dst_stride_b = -dst_stride_b;
+
+  SplitTransposeUV(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
+                   width, height);
+}
+
+// Rotate 180 is a horizontal and vertical flip.
+LIBYUV_API
+void SplitRotateUV180(const uint8_t* src,
+                      int src_stride,
+                      uint8_t* dst_a,
+                      int dst_stride_a,
+                      uint8_t* dst_b,
+                      int dst_stride_b,
+                      int width,
+                      int height) {
+  int i;
+  void (*MirrorSplitUVRow)(const uint8_t* src, uint8_t* dst_u, uint8_t* dst_v,
+                           int width) = MirrorSplitUVRow_C;
+#if defined(HAS_MIRRORSPLITUVROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(width, 16)) {
+    MirrorSplitUVRow = MirrorSplitUVRow_NEON;
+  }
+#endif
+#if defined(HAS_MIRRORSPLITUVROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3) && IS_ALIGNED(width, 16)) {
+    MirrorSplitUVRow = MirrorSplitUVRow_SSSE3;
+  }
+#endif
+#if defined(HAS_MIRRORSPLITUVROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 32)) {
+    MirrorSplitUVRow = MirrorSplitUVRow_MSA;
+  }
+#endif
+#if defined(HAS_MIRRORSPLITUVROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 32)) {
+    MirrorSplitUVRow = MirrorSplitUVRow_LSX;
+  }
+#endif
+
+  dst_a += dst_stride_a * (height - 1);
+  dst_b += dst_stride_b * (height - 1);
+
+  for (i = 0; i < height; ++i) {
+    MirrorSplitUVRow(src, dst_a, dst_b, width);
+    src += src_stride;
+    dst_a -= dst_stride_a;
+    dst_b -= dst_stride_b;
+  }
+}
+
+// Rotate UV and split into planar.
+// width and height expected to be half size for NV12
+LIBYUV_API
+int SplitRotateUV(const uint8_t* src_uv,
+                  int src_stride_uv,
+                  uint8_t* dst_u,
+                  int dst_stride_u,
+                  uint8_t* dst_v,
+                  int dst_stride_v,
+                  int width,
+                  int height,
+                  enum RotationMode mode) {
+  if (!src_uv || width <= 0 || height == 0 || !dst_u || !dst_v) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_uv = src_uv + (height - 1) * src_stride_uv;
+    src_stride_uv = -src_stride_uv;
+  }
+
+  switch (mode) {
+    case kRotate0:
+      SplitUVPlane(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v,
+                   dst_stride_v, width, height);
+      return 0;
+    case kRotate90:
+      SplitRotateUV90(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v,
+                      dst_stride_v, width, height);
+      return 0;
+    case kRotate270:
+      SplitRotateUV270(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v,
+                       dst_stride_v, width, height);
+      return 0;
+    case kRotate180:
+      SplitRotateUV180(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v,
+                       dst_stride_v, width, height);
+      return 0;
+    default:
+      break;
+  }
+  return -1;
+}
+
+LIBYUV_API
+int RotatePlane(const uint8_t* src,
+                int src_stride,
+                uint8_t* dst,
+                int dst_stride,
+                int width,
+                int height,
+                enum RotationMode mode) {
+  if (!src || width <= 0 || height == 0 || !dst) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src = src + (height - 1) * src_stride;
+    src_stride = -src_stride;
+  }
+
+  switch (mode) {
+    case kRotate0:
+      // copy frame
+      CopyPlane(src, src_stride, dst, dst_stride, width, height);
+      return 0;
+    case kRotate90:
+      RotatePlane90(src, src_stride, dst, dst_stride, width, height);
+      return 0;
+    case kRotate270:
+      RotatePlane270(src, src_stride, dst, dst_stride, width, height);
+      return 0;
+    case kRotate180:
+      RotatePlane180(src, src_stride, dst, dst_stride, width, height);
+      return 0;
+    default:
+      break;
+  }
+  return -1;
+}
+
+static void TransposePlane_16(const uint16_t* src,
+                              int src_stride,
+                              uint16_t* dst,
+                              int dst_stride,
+                              int width,
+                              int height) {
+  int i = height;
+  // Work across the source in 8x8 tiles
+  while (i >= 8) {
+    TransposeWx8_16_C(src, src_stride, dst, dst_stride, width);
+    src += 8 * src_stride;  // Go down 8 rows.
+    dst += 8;               // Move over 8 columns.
+    i -= 8;
+  }
+
+  if (i > 0) {
+    TransposeWxH_16_C(src, src_stride, dst, dst_stride, width, i);
+  }
+}
+
+static void RotatePlane90_16(const uint16_t* src,
+                             int src_stride,
+                             uint16_t* dst,
+                             int dst_stride,
+                             int width,
+                             int height) {
+  // Rotate by 90 is a transpose with the source read
+  // from bottom to top. So set the source pointer to the end
+  // of the buffer and flip the sign of the source stride.
+  src += src_stride * (height - 1);
+  src_stride = -src_stride;
+  TransposePlane_16(src, src_stride, dst, dst_stride, width, height);
+}
+
+static void RotatePlane270_16(const uint16_t* src,
+                              int src_stride,
+                              uint16_t* dst,
+                              int dst_stride,
+                              int width,
+                              int height) {
+  // Rotate by 270 is a transpose with the destination written
+  // from bottom to top. So set the destination pointer to the end
+  // of the buffer and flip the sign of the destination stride.
+  dst += dst_stride * (width - 1);
+  dst_stride = -dst_stride;
+  TransposePlane_16(src, src_stride, dst, dst_stride, width, height);
+}
+
+static void RotatePlane180_16(const uint16_t* src,
+                              int src_stride,
+                              uint16_t* dst,
+                              int dst_stride,
+                              int width,
+                              int height) {
+  const uint16_t* src_bot = src + src_stride * (height - 1);
+  uint16_t* dst_bot = dst + dst_stride * (height - 1);
+  int half_height = (height + 1) >> 1;
+  int y;
+
+  // Swap top and bottom row and mirror the content. Uses a temporary row.
+  align_buffer_64(row, width * 2);
+  uint16_t* row_tmp = (uint16_t*)row;
+  assert(row);
+  if (!row)
+    return;
+
+  // Odd height will harmlessly mirror the middle row twice.
+  for (y = 0; y < half_height; ++y) {
+    CopyRow_16_C(src, row_tmp, width);        // Copy top row into buffer
+    MirrorRow_16_C(src_bot, dst, width);      // Mirror bottom row into top row
+    MirrorRow_16_C(row_tmp, dst_bot, width);  // Mirror buffer into bottom row
+    src += src_stride;
+    dst += dst_stride;
+    src_bot -= src_stride;
+    dst_bot -= dst_stride;
+  }
+  free_aligned_buffer_64(row);
+}
+
+LIBYUV_API
+int RotatePlane_16(const uint16_t* src,
+                   int src_stride,
+                   uint16_t* dst,
+                   int dst_stride,
+                   int width,
+                   int height,
+                   enum RotationMode mode) {
+  if (!src || width <= 0 || height == 0 || !dst) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src = src + (height - 1) * src_stride;
+    src_stride = -src_stride;
+  }
+
+  switch (mode) {
+    case kRotate0:
+      // copy frame
+      CopyPlane_16(src, src_stride, dst, dst_stride, width, height);
+      return 0;
+    case kRotate90:
+      RotatePlane90_16(src, src_stride, dst, dst_stride, width, height);
+      return 0;
+    case kRotate270:
+      RotatePlane270_16(src, src_stride, dst, dst_stride, width, height);
+      return 0;
+    case kRotate180:
+      RotatePlane180_16(src, src_stride, dst, dst_stride, width, height);
+      return 0;
+    default:
+      break;
+  }
+  return -1;
+}
+
+LIBYUV_API
+int I420Rotate(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height,
+               enum RotationMode mode) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  if ((!src_y && dst_y) || !src_u || !src_v || width <= 0 || height == 0 ||
+      !dst_y || !dst_u || !dst_v) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (halfheight - 1) * src_stride_u;
+    src_v = src_v + (halfheight - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  switch (mode) {
+    case kRotate0:
+      // copy frame
+      return I420Copy(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                      src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                      dst_v, dst_stride_v, width, height);
+    case kRotate90:
+      RotatePlane90(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      RotatePlane90(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth,
+                    halfheight);
+      RotatePlane90(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth,
+                    halfheight);
+      return 0;
+    case kRotate270:
+      RotatePlane270(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      RotatePlane270(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth,
+                     halfheight);
+      RotatePlane270(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth,
+                     halfheight);
+      return 0;
+    case kRotate180:
+      RotatePlane180(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      RotatePlane180(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth,
+                     halfheight);
+      RotatePlane180(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth,
+                     halfheight);
+      return 0;
+    default:
+      break;
+  }
+  return -1;
+}
+
+// I422 has half width x full height UV planes, so rotate by 90 and 270
+// require scaling to maintain 422 subsampling.
+LIBYUV_API
+int I422Rotate(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height,
+               enum RotationMode mode) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  int r;
+  if (!src_y || !src_u || !src_v || width <= 0 || height == 0 || !dst_y ||
+      !dst_u || !dst_v) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  switch (mode) {
+    case kRotate0:
+      // Copy frame
+      CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      CopyPlane(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth, height);
+      CopyPlane(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth, height);
+      return 0;
+
+      // Note on temporary Y plane for UV.
+      // Rotation of UV first fits within the Y destination plane rows.
+      // Y plane is width x height
+      // Y plane rotated is height x width
+      // UV plane is (width / 2) x height
+      // UV plane rotated is height x (width / 2)
+      // UV plane rotated+scaled is (height / 2) x width.
+      // UV plane rotated is a temporary that fits within the Y plane rotated.
+
+    case kRotate90:
+      RotatePlane90(src_u, src_stride_u, dst_y, dst_stride_y, halfwidth,
+                    height);
+      r = ScalePlane(dst_y, dst_stride_y, height, halfwidth, dst_u,
+                     dst_stride_u, halfheight, width, kFilterBilinear);
+      if (r != 0) {
+        return r;
+      }
+      RotatePlane90(src_v, src_stride_v, dst_y, dst_stride_y, halfwidth,
+                    height);
+      r = ScalePlane(dst_y, dst_stride_y, height, halfwidth, dst_v,
+                     dst_stride_v, halfheight, width, kFilterLinear);
+      if (r != 0) {
+        return r;
+      }
+      RotatePlane90(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      return 0;
+    case kRotate270:
+      RotatePlane270(src_u, src_stride_u, dst_y, dst_stride_y, halfwidth,
+                     height);
+      r = ScalePlane(dst_y, dst_stride_y, height, halfwidth, dst_u,
+                     dst_stride_u, halfheight, width, kFilterBilinear);
+      if (r != 0) {
+        return r;
+      }
+      RotatePlane270(src_v, src_stride_v, dst_y, dst_stride_y, halfwidth,
+                     height);
+      r = ScalePlane(dst_y, dst_stride_y, height, halfwidth, dst_v,
+                     dst_stride_v, halfheight, width, kFilterLinear);
+      if (r != 0) {
+        return r;
+      }
+      RotatePlane270(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      return 0;
+    case kRotate180:
+      RotatePlane180(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      RotatePlane180(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth,
+                     height);
+      RotatePlane180(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth,
+                     height);
+      return 0;
+    default:
+      break;
+  }
+  return -1;
+}
+
+LIBYUV_API
+int I444Rotate(const uint8_t* src_y,
+               int src_stride_y,
+               const uint8_t* src_u,
+               int src_stride_u,
+               const uint8_t* src_v,
+               int src_stride_v,
+               uint8_t* dst_y,
+               int dst_stride_y,
+               uint8_t* dst_u,
+               int dst_stride_u,
+               uint8_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height,
+               enum RotationMode mode) {
+  if (!src_y || !src_u || !src_v || width <= 0 || height == 0 || !dst_y ||
+      !dst_u || !dst_v) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  switch (mode) {
+    case kRotate0:
+      // copy frame
+      CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      CopyPlane(src_u, src_stride_u, dst_u, dst_stride_u, width, height);
+      CopyPlane(src_v, src_stride_v, dst_v, dst_stride_v, width, height);
+      return 0;
+    case kRotate90:
+      RotatePlane90(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      RotatePlane90(src_u, src_stride_u, dst_u, dst_stride_u, width, height);
+      RotatePlane90(src_v, src_stride_v, dst_v, dst_stride_v, width, height);
+      return 0;
+    case kRotate270:
+      RotatePlane270(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      RotatePlane270(src_u, src_stride_u, dst_u, dst_stride_u, width, height);
+      RotatePlane270(src_v, src_stride_v, dst_v, dst_stride_v, width, height);
+      return 0;
+    case kRotate180:
+      RotatePlane180(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      RotatePlane180(src_u, src_stride_u, dst_u, dst_stride_u, width, height);
+      RotatePlane180(src_v, src_stride_v, dst_v, dst_stride_v, width, height);
+      return 0;
+    default:
+      break;
+  }
+  return -1;
+}
+
+LIBYUV_API
+int NV12ToI420Rotate(const uint8_t* src_y,
+                     int src_stride_y,
+                     const uint8_t* src_uv,
+                     int src_stride_uv,
+                     uint8_t* dst_y,
+                     int dst_stride_y,
+                     uint8_t* dst_u,
+                     int dst_stride_u,
+                     uint8_t* dst_v,
+                     int dst_stride_v,
+                     int width,
+                     int height,
+                     enum RotationMode mode) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  if (!src_y || !src_uv || width <= 0 || height == 0 || !dst_y || !dst_u ||
+      !dst_v) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_uv = src_uv + (halfheight - 1) * src_stride_uv;
+    src_stride_y = -src_stride_y;
+    src_stride_uv = -src_stride_uv;
+  }
+
+  switch (mode) {
+    case kRotate0:
+      // copy frame
+      return NV12ToI420(src_y, src_stride_y, src_uv, src_stride_uv, dst_y,
+                        dst_stride_y, dst_u, dst_stride_u, dst_v, dst_stride_v,
+                        width, height);
+    case kRotate90:
+      RotatePlane90(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      SplitRotateUV90(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v,
+                      dst_stride_v, halfwidth, halfheight);
+      return 0;
+    case kRotate270:
+      RotatePlane270(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      SplitRotateUV270(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v,
+                       dst_stride_v, halfwidth, halfheight);
+      return 0;
+    case kRotate180:
+      RotatePlane180(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      SplitRotateUV180(src_uv, src_stride_uv, dst_u, dst_stride_u, dst_v,
+                       dst_stride_v, halfwidth, halfheight);
+      return 0;
+    default:
+      break;
+  }
+  return -1;
+}
+
+static void SplitPixels(const uint8_t* src_u,
+                        int src_pixel_stride_uv,
+                        uint8_t* dst_u,
+                        int width) {
+  int i;
+  for (i = 0; i < width; ++i) {
+    *dst_u = *src_u;
+    ++dst_u;
+    src_u += src_pixel_stride_uv;
+  }
+}
+
+// Convert Android420 to I420 with Rotate
+LIBYUV_API
+int Android420ToI420Rotate(const uint8_t* src_y,
+                           int src_stride_y,
+                           const uint8_t* src_u,
+                           int src_stride_u,
+                           const uint8_t* src_v,
+                           int src_stride_v,
+                           int src_pixel_stride_uv,
+                           uint8_t* dst_y,
+                           int dst_stride_y,
+                           uint8_t* dst_u,
+                           int dst_stride_u,
+                           uint8_t* dst_v,
+                           int dst_stride_v,
+                           int width,
+                           int height,
+                           enum RotationMode rotation) {
+  int y;
+  const ptrdiff_t vu_off = src_v - src_u;
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  if ((!src_y && dst_y) || !src_u || !src_v || !dst_u || !dst_v || width <= 0 ||
+      height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    halfheight = (height + 1) >> 1;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (halfheight - 1) * src_stride_u;
+    src_v = src_v + (halfheight - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  if (dst_y) {
+    RotatePlane(src_y, src_stride_y, dst_y, dst_stride_y, width, height,
+                rotation);
+  }
+
+  // Copy UV planes - I420
+  if (src_pixel_stride_uv == 1) {
+    RotatePlane(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth, halfheight,
+                rotation);
+    RotatePlane(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth, halfheight,
+                rotation);
+    return 0;
+  }
+  // Split UV planes - NV21
+  if (src_pixel_stride_uv == 2 && vu_off == -1 &&
+      src_stride_u == src_stride_v) {
+    SplitRotateUV(src_v, src_stride_v, dst_v, dst_stride_v, dst_u, dst_stride_u,
+                  halfwidth, halfheight, rotation);
+    return 0;
+  }
+  // Split UV planes - NV12
+  if (src_pixel_stride_uv == 2 && vu_off == 1 && src_stride_u == src_stride_v) {
+    SplitRotateUV(src_u, src_stride_u, dst_u, dst_stride_u, dst_v, dst_stride_v,
+                  halfwidth, halfheight, rotation);
+    return 0;
+  }
+
+  if (rotation == 0) {
+    for (y = 0; y < halfheight; ++y) {
+      SplitPixels(src_u, src_pixel_stride_uv, dst_u, halfwidth);
+      SplitPixels(src_v, src_pixel_stride_uv, dst_v, halfwidth);
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+      dst_u += dst_stride_u;
+      dst_v += dst_stride_v;
+    }
+    return 0;
+  }
+  // unsupported type and/or rotation.
+  return -1;
+}
+
+LIBYUV_API
+int I010Rotate(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height,
+               enum RotationMode mode) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  if (!src_y || !src_u || !src_v || width <= 0 || height == 0 || !dst_y ||
+      !dst_u || !dst_v || dst_stride_y < 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  switch (mode) {
+    case kRotate0:
+      // copy frame
+      return I010Copy(src_y, src_stride_y, src_u, src_stride_u, src_v,
+                      src_stride_v, dst_y, dst_stride_y, dst_u, dst_stride_u,
+                      dst_v, dst_stride_v, width, height);
+    case kRotate90:
+      RotatePlane90_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      RotatePlane90_16(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth,
+                       halfheight);
+      RotatePlane90_16(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth,
+                       halfheight);
+      return 0;
+    case kRotate270:
+      RotatePlane270_16(src_y, src_stride_y, dst_y, dst_stride_y, width,
+                        height);
+      RotatePlane270_16(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth,
+                        halfheight);
+      RotatePlane270_16(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth,
+                        halfheight);
+      return 0;
+    case kRotate180:
+      RotatePlane180_16(src_y, src_stride_y, dst_y, dst_stride_y, width,
+                        height);
+      RotatePlane180_16(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth,
+                        halfheight);
+      RotatePlane180_16(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth,
+                        halfheight);
+      return 0;
+    default:
+      break;
+  }
+  return -1;
+}
+
+// I210 has half width x full height UV planes, so rotate by 90 and 270
+// require scaling to maintain 422 subsampling.
+LIBYUV_API
+int I210Rotate(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height,
+               enum RotationMode mode) {
+  int halfwidth = (width + 1) >> 1;
+  int halfheight = (height + 1) >> 1;
+  int r;
+  if (!src_y || !src_u || !src_v || width <= 0 || height == 0 || !dst_y ||
+      !dst_u || !dst_v) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  switch (mode) {
+    case kRotate0:
+      // Copy frame
+      CopyPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      CopyPlane_16(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth, height);
+      CopyPlane_16(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth, height);
+      return 0;
+
+      // Note on temporary Y plane for UV.
+      // Rotation of UV first fits within the Y destination plane rows.
+      // Y plane is width x height
+      // Y plane rotated is height x width
+      // UV plane is (width / 2) x height
+      // UV plane rotated is height x (width / 2)
+      // UV plane rotated+scaled is (height / 2) x width.
+      // UV plane rotated is a temporary that fits within the Y plane rotated.
+
+    case kRotate90:
+      RotatePlane90_16(src_u, src_stride_u, dst_y, dst_stride_y, halfwidth,
+                       height);
+      r = ScalePlane_16(dst_y, dst_stride_y, height, halfwidth, dst_u,
+                        dst_stride_u, halfheight, width, kFilterBilinear);
+      if (r != 0) {
+        return r;
+      }
+      RotatePlane90_16(src_v, src_stride_v, dst_y, dst_stride_y, halfwidth,
+                       height);
+      r = ScalePlane_16(dst_y, dst_stride_y, height, halfwidth, dst_v,
+                        dst_stride_v, halfheight, width, kFilterLinear);
+      if (r != 0) {
+        return r;
+      }
+      RotatePlane90_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      return 0;
+    case kRotate270:
+      RotatePlane270_16(src_u, src_stride_u, dst_y, dst_stride_y, halfwidth,
+                        height);
+      r = ScalePlane_16(dst_y, dst_stride_y, height, halfwidth, dst_u,
+                        dst_stride_u, halfheight, width, kFilterBilinear);
+      if (r != 0) {
+        return r;
+      }
+      RotatePlane270_16(src_v, src_stride_v, dst_y, dst_stride_y, halfwidth,
+                        height);
+      r = ScalePlane_16(dst_y, dst_stride_y, height, halfwidth, dst_v,
+                        dst_stride_v, halfheight, width, kFilterLinear);
+      if (r != 0) {
+        return r;
+      }
+      RotatePlane270_16(src_y, src_stride_y, dst_y, dst_stride_y, width,
+                        height);
+      return 0;
+    case kRotate180:
+      RotatePlane180_16(src_y, src_stride_y, dst_y, dst_stride_y, width,
+                        height);
+      RotatePlane180_16(src_u, src_stride_u, dst_u, dst_stride_u, halfwidth,
+                        height);
+      RotatePlane180_16(src_v, src_stride_v, dst_v, dst_stride_v, halfwidth,
+                        height);
+      return 0;
+    default:
+      break;
+  }
+  return -1;
+}
+
+LIBYUV_API
+int I410Rotate(const uint16_t* src_y,
+               int src_stride_y,
+               const uint16_t* src_u,
+               int src_stride_u,
+               const uint16_t* src_v,
+               int src_stride_v,
+               uint16_t* dst_y,
+               int dst_stride_y,
+               uint16_t* dst_u,
+               int dst_stride_u,
+               uint16_t* dst_v,
+               int dst_stride_v,
+               int width,
+               int height,
+               enum RotationMode mode) {
+  if (!src_y || !src_u || !src_v || width <= 0 || height == 0 || !dst_y ||
+      !dst_u || !dst_v || dst_stride_y < 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_y = src_y + (height - 1) * src_stride_y;
+    src_u = src_u + (height - 1) * src_stride_u;
+    src_v = src_v + (height - 1) * src_stride_v;
+    src_stride_y = -src_stride_y;
+    src_stride_u = -src_stride_u;
+    src_stride_v = -src_stride_v;
+  }
+
+  switch (mode) {
+    case kRotate0:
+      // copy frame
+      CopyPlane_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      CopyPlane_16(src_u, src_stride_u, dst_u, dst_stride_u, width, height);
+      CopyPlane_16(src_v, src_stride_v, dst_v, dst_stride_v, width, height);
+      return 0;
+    case kRotate90:
+      RotatePlane90_16(src_y, src_stride_y, dst_y, dst_stride_y, width, height);
+      RotatePlane90_16(src_u, src_stride_u, dst_u, dst_stride_u, width, height);
+      RotatePlane90_16(src_v, src_stride_v, dst_v, dst_stride_v, width, height);
+      return 0;
+    case kRotate270:
+      RotatePlane270_16(src_y, src_stride_y, dst_y, dst_stride_y, width,
+                        height);
+      RotatePlane270_16(src_u, src_stride_u, dst_u, dst_stride_u, width,
+                        height);
+      RotatePlane270_16(src_v, src_stride_v, dst_v, dst_stride_v, width,
+                        height);
+      return 0;
+    case kRotate180:
+      RotatePlane180_16(src_y, src_stride_y, dst_y, dst_stride_y, width,
+                        height);
+      RotatePlane180_16(src_u, src_stride_u, dst_u, dst_stride_u, width,
+                        height);
+      RotatePlane180_16(src_v, src_stride_v, dst_v, dst_stride_v, width,
+                        height);
+      return 0;
+    default:
+      break;
+  }
+  return -1;
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/rotate_any.cc

@@ -0,0 +1,82 @@
+/*
+ *  Copyright 2015 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/rotate.h"
+#include "libyuv/rotate_row.h"
+
+#include "libyuv/basic_types.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+#define TANY(NAMEANY, TPOS_SIMD, TPOS_C, MASK)                        \
+  void NAMEANY(const uint8_t* src, int src_stride, uint8_t* dst,      \
+               int dst_stride, int width) {                           \
+    int r = width & MASK;                                             \
+    int n = width - r;                                                \
+    if (n > 0) {                                                      \
+      TPOS_SIMD(src, src_stride, dst, dst_stride, n);                 \
+    }                                                                 \
+    TPOS_C(src + n, src_stride, dst + n * dst_stride, dst_stride, r); \
+  }
+
+#ifdef HAS_TRANSPOSEWX8_NEON
+TANY(TransposeWx8_Any_NEON, TransposeWx8_NEON, TransposeWx8_C, 7)
+#endif
+#ifdef HAS_TRANSPOSEWX16_NEON
+TANY(TransposeWx16_Any_NEON, TransposeWx16_NEON, TransposeWx16_C, 15)
+#endif
+#ifdef HAS_TRANSPOSEWX8_SSSE3
+TANY(TransposeWx8_Any_SSSE3, TransposeWx8_SSSE3, TransposeWx8_C, 7)
+#endif
+#ifdef HAS_TRANSPOSEWX8_FAST_SSSE3
+TANY(TransposeWx8_Fast_Any_SSSE3, TransposeWx8_Fast_SSSE3, TransposeWx8_C, 15)
+#endif
+#ifdef HAS_TRANSPOSEWX16_MSA
+TANY(TransposeWx16_Any_MSA, TransposeWx16_MSA, TransposeWx16_C, 15)
+#endif
+#ifdef HAS_TRANSPOSEWX16_LSX
+TANY(TransposeWx16_Any_LSX, TransposeWx16_LSX, TransposeWx16_C, 15)
+#endif
+#undef TANY
+
+#define TUVANY(NAMEANY, TPOS_SIMD, MASK)                                       \
+  void NAMEANY(const uint8_t* src, int src_stride, uint8_t* dst_a,             \
+               int dst_stride_a, uint8_t* dst_b, int dst_stride_b,             \
+               int width) {                                                    \
+    int r = width & MASK;                                                      \
+    int n = width - r;                                                         \
+    if (n > 0) {                                                               \
+      TPOS_SIMD(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b, n); \
+    }                                                                          \
+    TransposeUVWx8_C(src + n * 2, src_stride, dst_a + n * dst_stride_a,        \
+                     dst_stride_a, dst_b + n * dst_stride_b, dst_stride_b, r); \
+  }
+
+#ifdef HAS_TRANSPOSEUVWX8_NEON
+TUVANY(TransposeUVWx8_Any_NEON, TransposeUVWx8_NEON, 7)
+#endif
+#ifdef HAS_TRANSPOSEUVWX8_SSE2
+TUVANY(TransposeUVWx8_Any_SSE2, TransposeUVWx8_SSE2, 7)
+#endif
+#ifdef HAS_TRANSPOSEUVWX16_MSA
+TUVANY(TransposeUVWx16_Any_MSA, TransposeUVWx16_MSA, 7)
+#endif
+#ifdef HAS_TRANSPOSEUVWX16_LSX
+TUVANY(TransposeUVWx16_Any_LSX, TransposeUVWx16_LSX, 7)
+#endif
+#undef TUVANY
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/rotate_argb.cc

@@ -0,0 +1,264 @@
+/*
+ *  Copyright 2012 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/rotate_argb.h"
+
+#include "libyuv/convert.h"
+#include "libyuv/cpu_id.h"
+#include "libyuv/planar_functions.h"
+#include "libyuv/rotate.h"
+#include "libyuv/row.h"
+#include "libyuv/scale_row.h" /* for ScaleARGBRowDownEven_ */
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+static int ARGBTranspose(const uint8_t* src_argb,
+                         int src_stride_argb,
+                         uint8_t* dst_argb,
+                         int dst_stride_argb,
+                         int width,
+                         int height) {
+  int i;
+  int src_pixel_step = src_stride_argb >> 2;
+  void (*ScaleARGBRowDownEven)(
+      const uint8_t* src_argb, ptrdiff_t src_stride_argb, int src_step,
+      uint8_t* dst_argb, int dst_width) = ScaleARGBRowDownEven_C;
+  // Check stride is a multiple of 4.
+  if (src_stride_argb & 3) {
+    return -1;
+  }
+#if defined(HAS_SCALEARGBROWDOWNEVEN_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ScaleARGBRowDownEven = ScaleARGBRowDownEven_Any_SSE2;
+    if (IS_ALIGNED(height, 4)) {  // Width of dest.
+      ScaleARGBRowDownEven = ScaleARGBRowDownEven_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_SCALEARGBROWDOWNEVEN_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ScaleARGBRowDownEven = ScaleARGBRowDownEven_Any_NEON;
+    if (IS_ALIGNED(height, 4)) {  // Width of dest.
+      ScaleARGBRowDownEven = ScaleARGBRowDownEven_NEON;
+    }
+  }
+#endif
+#if defined(HAS_SCALEARGBROWDOWNEVEN_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ScaleARGBRowDownEven = ScaleARGBRowDownEven_Any_MSA;
+    if (IS_ALIGNED(height, 4)) {  // Width of dest.
+      ScaleARGBRowDownEven = ScaleARGBRowDownEven_MSA;
+    }
+  }
+#endif
+#if defined(HAS_SCALEARGBROWDOWNEVEN_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ScaleARGBRowDownEven = ScaleARGBRowDownEven_Any_LSX;
+    if (IS_ALIGNED(height, 4)) {  // Width of dest.
+      ScaleARGBRowDownEven = ScaleARGBRowDownEven_LSX;
+    }
+  }
+#endif
+#if defined(HAS_SCALEARGBROWDOWNEVEN_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ScaleARGBRowDownEven = ScaleARGBRowDownEven_RVV;
+  }
+#endif
+
+  for (i = 0; i < width; ++i) {  // column of source to row of dest.
+    ScaleARGBRowDownEven(src_argb, 0, src_pixel_step, dst_argb, height);
+    dst_argb += dst_stride_argb;
+    src_argb += 4;
+  }
+  return 0;
+}
+
+static int ARGBRotate90(const uint8_t* src_argb,
+                        int src_stride_argb,
+                        uint8_t* dst_argb,
+                        int dst_stride_argb,
+                        int width,
+                        int height) {
+  // Rotate by 90 is a ARGBTranspose with the source read
+  // from bottom to top. So set the source pointer to the end
+  // of the buffer and flip the sign of the source stride.
+  src_argb += src_stride_argb * (height - 1);
+  src_stride_argb = -src_stride_argb;
+  return ARGBTranspose(src_argb, src_stride_argb, dst_argb, dst_stride_argb,
+                       width, height);
+}
+
+static int ARGBRotate270(const uint8_t* src_argb,
+                         int src_stride_argb,
+                         uint8_t* dst_argb,
+                         int dst_stride_argb,
+                         int width,
+                         int height) {
+  // Rotate by 270 is a ARGBTranspose with the destination written
+  // from bottom to top. So set the destination pointer to the end
+  // of the buffer and flip the sign of the destination stride.
+  dst_argb += dst_stride_argb * (width - 1);
+  dst_stride_argb = -dst_stride_argb;
+  return ARGBTranspose(src_argb, src_stride_argb, dst_argb, dst_stride_argb,
+                       width, height);
+}
+
+static int ARGBRotate180(const uint8_t* src_argb,
+                         int src_stride_argb,
+                         uint8_t* dst_argb,
+                         int dst_stride_argb,
+                         int width,
+                         int height) {
+  // Swap first and last row and mirror the content. Uses a temporary row.
+  const uint8_t* src_bot = src_argb + src_stride_argb * (height - 1);
+  uint8_t* dst_bot = dst_argb + dst_stride_argb * (height - 1);
+  int half_height = (height + 1) >> 1;
+  int y;
+  void (*ARGBMirrorRow)(const uint8_t* src_argb, uint8_t* dst_argb, int width) =
+      ARGBMirrorRow_C;
+  void (*CopyRow)(const uint8_t* src_argb, uint8_t* dst_argb, int width) =
+      CopyRow_C;
+  align_buffer_64(row, width * 4);
+  if (!row)
+    return 1;
+#if defined(HAS_ARGBMIRRORROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    ARGBMirrorRow = ARGBMirrorRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBMirrorRow = ARGBMirrorRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMIRRORROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    ARGBMirrorRow = ARGBMirrorRow_Any_SSE2;
+    if (IS_ALIGNED(width, 4)) {
+      ARGBMirrorRow = ARGBMirrorRow_SSE2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMIRRORROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    ARGBMirrorRow = ARGBMirrorRow_Any_AVX2;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBMirrorRow = ARGBMirrorRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMIRRORROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGBMirrorRow = ARGBMirrorRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBMirrorRow = ARGBMirrorRow_MSA;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMIRRORROW_LSX)
+  if (TestCpuFlag(kCpuHasLSX)) {
+    ARGBMirrorRow = ARGBMirrorRow_Any_LSX;
+    if (IS_ALIGNED(width, 8)) {
+      ARGBMirrorRow = ARGBMirrorRow_LSX;
+    }
+  }
+#endif
+#if defined(HAS_ARGBMIRRORROW_LASX)
+  if (TestCpuFlag(kCpuHasLASX)) {
+    ARGBMirrorRow = ARGBMirrorRow_Any_LASX;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBMirrorRow = ARGBMirrorRow_LASX;
+    }
+  }
+#endif
+#if defined(HAS_COPYROW_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2)) {
+    CopyRow = IS_ALIGNED(width * 4, 32) ? CopyRow_SSE2 : CopyRow_Any_SSE2;
+  }
+#endif
+#if defined(HAS_COPYROW_AVX)
+  if (TestCpuFlag(kCpuHasAVX)) {
+    CopyRow = IS_ALIGNED(width * 4, 64) ? CopyRow_AVX : CopyRow_Any_AVX;
+  }
+#endif
+#if defined(HAS_COPYROW_ERMS)
+  if (TestCpuFlag(kCpuHasERMS)) {
+    CopyRow = CopyRow_ERMS;
+  }
+#endif
+#if defined(HAS_COPYROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    CopyRow = IS_ALIGNED(width * 4, 32) ? CopyRow_NEON : CopyRow_Any_NEON;
+  }
+#endif
+#if defined(HAS_COPYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    CopyRow = CopyRow_RVV;
+  }
+#endif
+
+  // Odd height will harmlessly mirror the middle row twice.
+  for (y = 0; y < half_height; ++y) {
+    ARGBMirrorRow(src_argb, row, width);      // Mirror first row into a buffer
+    ARGBMirrorRow(src_bot, dst_argb, width);  // Mirror last row into first row
+    CopyRow(row, dst_bot, width * 4);  // Copy first mirrored row into last
+    src_argb += src_stride_argb;
+    dst_argb += dst_stride_argb;
+    src_bot -= src_stride_argb;
+    dst_bot -= dst_stride_argb;
+  }
+  free_aligned_buffer_64(row);
+  return 0;
+}
+
+LIBYUV_API
+int ARGBRotate(const uint8_t* src_argb,
+               int src_stride_argb,
+               uint8_t* dst_argb,
+               int dst_stride_argb,
+               int width,
+               int height,
+               enum RotationMode mode) {
+  if (!src_argb || width <= 0 || height == 0 || !dst_argb) {
+    return -1;
+  }
+
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+
+  switch (mode) {
+    case kRotate0:
+      // copy frame
+      return ARGBCopy(src_argb, src_stride_argb, dst_argb, dst_stride_argb,
+                      width, height);
+    case kRotate90:
+      return ARGBRotate90(src_argb, src_stride_argb, dst_argb, dst_stride_argb,
+                          width, height);
+    case kRotate270:
+      return ARGBRotate270(src_argb, src_stride_argb, dst_argb, dst_stride_argb,
+                           width, height);
+    case kRotate180:
+      return ARGBRotate180(src_argb, src_stride_argb, dst_argb, dst_stride_argb,
+                           width, height);
+    default:
+      break;
+  }
+  return -1;
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/rotate_common.cc

@@ -0,0 +1,208 @@
+/*
+ *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/rotate_row.h"
+#include "libyuv/row.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+void TransposeWx8_C(const uint8_t* src,
+                    int src_stride,
+                    uint8_t* dst,
+                    int dst_stride,
+                    int width) {
+  int i;
+  for (i = 0; i < width; ++i) {
+    dst[0] = src[0 * src_stride];
+    dst[1] = src[1 * src_stride];
+    dst[2] = src[2 * src_stride];
+    dst[3] = src[3 * src_stride];
+    dst[4] = src[4 * src_stride];
+    dst[5] = src[5 * src_stride];
+    dst[6] = src[6 * src_stride];
+    dst[7] = src[7 * src_stride];
+    ++src;
+    dst += dst_stride;
+  }
+}
+
+void TransposeWx16_C(const uint8_t* src,
+                     int src_stride,
+                     uint8_t* dst,
+                     int dst_stride,
+                     int width) {
+  TransposeWx8_C(src, src_stride, dst, dst_stride, width);
+  TransposeWx8_C((src + 8 * src_stride), src_stride, (dst + 8), dst_stride,
+                 width);
+}
+
+void TransposeUVWx8_C(const uint8_t* src,
+                      int src_stride,
+                      uint8_t* dst_a,
+                      int dst_stride_a,
+                      uint8_t* dst_b,
+                      int dst_stride_b,
+                      int width) {
+  int i;
+  for (i = 0; i < width; ++i) {
+    dst_a[0] = src[0 * src_stride + 0];
+    dst_b[0] = src[0 * src_stride + 1];
+    dst_a[1] = src[1 * src_stride + 0];
+    dst_b[1] = src[1 * src_stride + 1];
+    dst_a[2] = src[2 * src_stride + 0];
+    dst_b[2] = src[2 * src_stride + 1];
+    dst_a[3] = src[3 * src_stride + 0];
+    dst_b[3] = src[3 * src_stride + 1];
+    dst_a[4] = src[4 * src_stride + 0];
+    dst_b[4] = src[4 * src_stride + 1];
+    dst_a[5] = src[5 * src_stride + 0];
+    dst_b[5] = src[5 * src_stride + 1];
+    dst_a[6] = src[6 * src_stride + 0];
+    dst_b[6] = src[6 * src_stride + 1];
+    dst_a[7] = src[7 * src_stride + 0];
+    dst_b[7] = src[7 * src_stride + 1];
+    src += 2;
+    dst_a += dst_stride_a;
+    dst_b += dst_stride_b;
+  }
+}
+
+void TransposeWxH_C(const uint8_t* src,
+                    int src_stride,
+                    uint8_t* dst,
+                    int dst_stride,
+                    int width,
+                    int height) {
+  int i;
+  for (i = 0; i < width; ++i) {
+    int j;
+    for (j = 0; j < height; ++j) {
+      dst[i * dst_stride + j] = src[j * src_stride + i];
+    }
+  }
+}
+
+void TransposeUVWxH_C(const uint8_t* src,
+                      int src_stride,
+                      uint8_t* dst_a,
+                      int dst_stride_a,
+                      uint8_t* dst_b,
+                      int dst_stride_b,
+                      int width,
+                      int height) {
+  int i;
+  for (i = 0; i < width * 2; i += 2) {
+    int j;
+    for (j = 0; j < height; ++j) {
+      dst_a[((i >> 1) * dst_stride_a) + j] = src[i + (j * src_stride)];
+      dst_b[((i >> 1) * dst_stride_b) + j] = src[i + (j * src_stride) + 1];
+    }
+  }
+}
+
+void TransposeWx8_16_C(const uint16_t* src,
+                       int src_stride,
+                       uint16_t* dst,
+                       int dst_stride,
+                       int width) {
+  int i;
+  for (i = 0; i < width; ++i) {
+    dst[0] = src[0 * src_stride];
+    dst[1] = src[1 * src_stride];
+    dst[2] = src[2 * src_stride];
+    dst[3] = src[3 * src_stride];
+    dst[4] = src[4 * src_stride];
+    dst[5] = src[5 * src_stride];
+    dst[6] = src[6 * src_stride];
+    dst[7] = src[7 * src_stride];
+    ++src;
+    dst += dst_stride;
+  }
+}
+
+void TransposeWxH_16_C(const uint16_t* src,
+                       int src_stride,
+                       uint16_t* dst,
+                       int dst_stride,
+                       int width,
+                       int height) {
+  int i;
+  for (i = 0; i < width; ++i) {
+    int j;
+    for (j = 0; j < height; ++j) {
+      dst[i * dst_stride + j] = src[j * src_stride + i];
+    }
+  }
+}
+
+// Transpose 32 bit values (ARGB)
+void Transpose4x4_32_C(const uint8_t* src,
+                       int src_stride,
+                       uint8_t* dst,
+                       int dst_stride,
+                       int width) {
+  const uint8_t* src1 = src + src_stride;
+  const uint8_t* src2 = src1 + src_stride;
+  const uint8_t* src3 = src2 + src_stride;
+  uint8_t* dst1 = dst + dst_stride;
+  uint8_t* dst2 = dst1 + dst_stride;
+  uint8_t* dst3 = dst2 + dst_stride;
+  int i;
+  for (i = 0; i < width; i += 4) {
+    uint32_t p00 = ((uint32_t*)(src))[0];
+    uint32_t p10 = ((uint32_t*)(src))[1];
+    uint32_t p20 = ((uint32_t*)(src))[2];
+    uint32_t p30 = ((uint32_t*)(src))[3];
+    uint32_t p01 = ((uint32_t*)(src1))[0];
+    uint32_t p11 = ((uint32_t*)(src1))[1];
+    uint32_t p21 = ((uint32_t*)(src1))[2];
+    uint32_t p31 = ((uint32_t*)(src1))[3];
+    uint32_t p02 = ((uint32_t*)(src2))[0];
+    uint32_t p12 = ((uint32_t*)(src2))[1];
+    uint32_t p22 = ((uint32_t*)(src2))[2];
+    uint32_t p32 = ((uint32_t*)(src2))[3];
+    uint32_t p03 = ((uint32_t*)(src3))[0];
+    uint32_t p13 = ((uint32_t*)(src3))[1];
+    uint32_t p23 = ((uint32_t*)(src3))[2];
+    uint32_t p33 = ((uint32_t*)(src3))[3];
+    ((uint32_t*)(dst))[0] = p00;
+    ((uint32_t*)(dst))[1] = p01;
+    ((uint32_t*)(dst))[2] = p02;
+    ((uint32_t*)(dst))[3] = p03;
+    ((uint32_t*)(dst1))[0] = p10;
+    ((uint32_t*)(dst1))[1] = p11;
+    ((uint32_t*)(dst1))[2] = p12;
+    ((uint32_t*)(dst1))[3] = p13;
+    ((uint32_t*)(dst2))[0] = p20;
+    ((uint32_t*)(dst2))[1] = p21;
+    ((uint32_t*)(dst2))[2] = p22;
+    ((uint32_t*)(dst2))[3] = p23;
+    ((uint32_t*)(dst3))[0] = p30;
+    ((uint32_t*)(dst3))[1] = p31;
+    ((uint32_t*)(dst3))[2] = p32;
+    ((uint32_t*)(dst3))[3] = p33;
+    src += src_stride * 4;  // advance 4 rows
+    src1 += src_stride * 4;
+    src2 += src_stride * 4;
+    src3 += src_stride * 4;
+    dst += 4 * 4;  // advance 4 columns
+    dst1 += 4 * 4;
+    dst2 += 4 * 4;
+    dst3 += 4 * 4;
+  }
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/rotate_gcc.cc

@@ -0,0 +1,503 @@
+/*
+ *  Copyright 2015 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/rotate_row.h"
+#include "libyuv/row.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+// This module is for GCC x86 and x64.
+#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__))
+
+// Transpose 8x8. 32 or 64 bit, but not NaCL for 64 bit.
+#if defined(HAS_TRANSPOSEWX8_SSSE3)
+void TransposeWx8_SSSE3(const uint8_t* src,
+                        int src_stride,
+                        uint8_t* dst,
+                        int dst_stride,
+                        int width) {
+  asm volatile (
+      // Read in the data from the source pointer.
+      // First round of bit swap.
+      LABELALIGN
+      "1:                                        \n"
+      "movq        (%0),%%xmm0                   \n"
+      "movq        (%0,%3),%%xmm1                \n"
+      "lea         (%0,%3,2),%0                  \n"
+      "punpcklbw   %%xmm1,%%xmm0                 \n"
+      "movq        (%0),%%xmm2                   \n"
+      "movdqa      %%xmm0,%%xmm1                 \n"
+      "palignr     $0x8,%%xmm1,%%xmm1            \n"
+      "movq        (%0,%3),%%xmm3                \n"
+      "lea         (%0,%3,2),%0                  \n"
+      "punpcklbw   %%xmm3,%%xmm2                 \n"
+      "movdqa      %%xmm2,%%xmm3                 \n"
+      "movq        (%0),%%xmm4                   \n"
+      "palignr     $0x8,%%xmm3,%%xmm3            \n"
+      "movq        (%0,%3),%%xmm5                \n"
+      "lea         (%0,%3,2),%0                  \n"
+      "punpcklbw   %%xmm5,%%xmm4                 \n"
+      "movdqa      %%xmm4,%%xmm5                 \n"
+      "movq        (%0),%%xmm6                   \n"
+      "palignr     $0x8,%%xmm5,%%xmm5            \n"
+      "movq        (%0,%3),%%xmm7                \n"
+      "lea         (%0,%3,2),%0                  \n"
+      "punpcklbw   %%xmm7,%%xmm6                 \n"
+      "neg         %3                            \n"
+      "movdqa      %%xmm6,%%xmm7                 \n"
+      "lea         0x8(%0,%3,8),%0               \n"
+      "palignr     $0x8,%%xmm7,%%xmm7            \n"
+      "neg         %3                            \n"
+      // Second round of bit swap.
+      "punpcklwd   %%xmm2,%%xmm0                 \n"
+      "punpcklwd   %%xmm3,%%xmm1                 \n"
+      "movdqa      %%xmm0,%%xmm2                 \n"
+      "movdqa      %%xmm1,%%xmm3                 \n"
+      "palignr     $0x8,%%xmm2,%%xmm2            \n"
+      "palignr     $0x8,%%xmm3,%%xmm3            \n"
+      "punpcklwd   %%xmm6,%%xmm4                 \n"
+      "punpcklwd   %%xmm7,%%xmm5                 \n"
+      "movdqa      %%xmm4,%%xmm6                 \n"
+      "movdqa      %%xmm5,%%xmm7                 \n"
+      "palignr     $0x8,%%xmm6,%%xmm6            \n"
+      "palignr     $0x8,%%xmm7,%%xmm7            \n"
+      // Third round of bit swap.
+      // Write to the destination pointer.
+      "punpckldq   %%xmm4,%%xmm0                 \n"
+      "movq        %%xmm0,(%1)                   \n"
+      "movdqa      %%xmm0,%%xmm4                 \n"
+      "palignr     $0x8,%%xmm4,%%xmm4            \n"
+      "movq        %%xmm4,(%1,%4)                \n"
+      "lea         (%1,%4,2),%1                  \n"
+      "punpckldq   %%xmm6,%%xmm2                 \n"
+      "movdqa      %%xmm2,%%xmm6                 \n"
+      "movq        %%xmm2,(%1)                   \n"
+      "palignr     $0x8,%%xmm6,%%xmm6            \n"
+      "punpckldq   %%xmm5,%%xmm1                 \n"
+      "movq        %%xmm6,(%1,%4)                \n"
+      "lea         (%1,%4,2),%1                  \n"
+      "movdqa      %%xmm1,%%xmm5                 \n"
+      "movq        %%xmm1,(%1)                   \n"
+      "palignr     $0x8,%%xmm5,%%xmm5            \n"
+      "movq        %%xmm5,(%1,%4)                \n"
+      "lea         (%1,%4,2),%1                  \n"
+      "punpckldq   %%xmm7,%%xmm3                 \n"
+      "movq        %%xmm3,(%1)                   \n"
+      "movdqa      %%xmm3,%%xmm7                 \n"
+      "palignr     $0x8,%%xmm7,%%xmm7            \n"
+      "sub         $0x8,%2                       \n"
+      "movq        %%xmm7,(%1,%4)                \n"
+      "lea         (%1,%4,2),%1                  \n"
+      "jg          1b                            \n"
+      : "+r"(src),                    // %0
+        "+r"(dst),                    // %1
+        "+r"(width)                   // %2
+      : "r"((intptr_t)(src_stride)),  // %3
+        "r"((intptr_t)(dst_stride))   // %4
+      : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6",
+        "xmm7");
+}
+#endif  // defined(HAS_TRANSPOSEWX8_SSSE3)
+
+// Transpose 16x8. 64 bit
+#if defined(HAS_TRANSPOSEWX8_FAST_SSSE3)
+void TransposeWx8_Fast_SSSE3(const uint8_t* src,
+                             int src_stride,
+                             uint8_t* dst,
+                             int dst_stride,
+                             int width) {
+  asm volatile (
+      // Read in the data from the source pointer.
+      // First round of bit swap.
+      LABELALIGN
+      "1:                                        \n"
+      "movdqu      (%0),%%xmm0                   \n"
+      "movdqu      (%0,%3),%%xmm1                \n"
+      "lea         (%0,%3,2),%0                  \n"
+      "movdqa      %%xmm0,%%xmm8                 \n"
+      "punpcklbw   %%xmm1,%%xmm0                 \n"
+      "punpckhbw   %%xmm1,%%xmm8                 \n"
+      "movdqu      (%0),%%xmm2                   \n"
+      "movdqa      %%xmm0,%%xmm1                 \n"
+      "movdqa      %%xmm8,%%xmm9                 \n"
+      "palignr     $0x8,%%xmm1,%%xmm1            \n"
+      "palignr     $0x8,%%xmm9,%%xmm9            \n"
+      "movdqu      (%0,%3),%%xmm3                \n"
+      "lea         (%0,%3,2),%0                  \n"
+      "movdqa      %%xmm2,%%xmm10                \n"
+      "punpcklbw   %%xmm3,%%xmm2                 \n"
+      "punpckhbw   %%xmm3,%%xmm10                \n"
+      "movdqa      %%xmm2,%%xmm3                 \n"
+      "movdqa      %%xmm10,%%xmm11               \n"
+      "movdqu      (%0),%%xmm4                   \n"
+      "palignr     $0x8,%%xmm3,%%xmm3            \n"
+      "palignr     $0x8,%%xmm11,%%xmm11          \n"
+      "movdqu      (%0,%3),%%xmm5                \n"
+      "lea         (%0,%3,2),%0                  \n"
+      "movdqa      %%xmm4,%%xmm12                \n"
+      "punpcklbw   %%xmm5,%%xmm4                 \n"
+      "punpckhbw   %%xmm5,%%xmm12                \n"
+      "movdqa      %%xmm4,%%xmm5                 \n"
+      "movdqa      %%xmm12,%%xmm13               \n"
+      "movdqu      (%0),%%xmm6                   \n"
+      "palignr     $0x8,%%xmm5,%%xmm5            \n"
+      "palignr     $0x8,%%xmm13,%%xmm13          \n"
+      "movdqu      (%0,%3),%%xmm7                \n"
+      "lea         (%0,%3,2),%0                  \n"
+      "movdqa      %%xmm6,%%xmm14                \n"
+      "punpcklbw   %%xmm7,%%xmm6                 \n"
+      "punpckhbw   %%xmm7,%%xmm14                \n"
+      "neg         %3                            \n"
+      "movdqa      %%xmm6,%%xmm7                 \n"
+      "movdqa      %%xmm14,%%xmm15               \n"
+      "lea         0x10(%0,%3,8),%0              \n"
+      "palignr     $0x8,%%xmm7,%%xmm7            \n"
+      "palignr     $0x8,%%xmm15,%%xmm15          \n"
+      "neg         %3                            \n"
+      // Second round of bit swap.
+      "punpcklwd   %%xmm2,%%xmm0                 \n"
+      "punpcklwd   %%xmm3,%%xmm1                 \n"
+      "movdqa      %%xmm0,%%xmm2                 \n"
+      "movdqa      %%xmm1,%%xmm3                 \n"
+      "palignr     $0x8,%%xmm2,%%xmm2            \n"
+      "palignr     $0x8,%%xmm3,%%xmm3            \n"
+      "punpcklwd   %%xmm6,%%xmm4                 \n"
+      "punpcklwd   %%xmm7,%%xmm5                 \n"
+      "movdqa      %%xmm4,%%xmm6                 \n"
+      "movdqa      %%xmm5,%%xmm7                 \n"
+      "palignr     $0x8,%%xmm6,%%xmm6            \n"
+      "palignr     $0x8,%%xmm7,%%xmm7            \n"
+      "punpcklwd   %%xmm10,%%xmm8                \n"
+      "punpcklwd   %%xmm11,%%xmm9                \n"
+      "movdqa      %%xmm8,%%xmm10                \n"
+      "movdqa      %%xmm9,%%xmm11                \n"
+      "palignr     $0x8,%%xmm10,%%xmm10          \n"
+      "palignr     $0x8,%%xmm11,%%xmm11          \n"
+      "punpcklwd   %%xmm14,%%xmm12               \n"
+      "punpcklwd   %%xmm15,%%xmm13               \n"
+      "movdqa      %%xmm12,%%xmm14               \n"
+      "movdqa      %%xmm13,%%xmm15               \n"
+      "palignr     $0x8,%%xmm14,%%xmm14          \n"
+      "palignr     $0x8,%%xmm15,%%xmm15          \n"
+      // Third round of bit swap.
+      // Write to the destination pointer.
+      "punpckldq   %%xmm4,%%xmm0                 \n"
+      "movq        %%xmm0,(%1)                   \n"
+      "movdqa      %%xmm0,%%xmm4                 \n"
+      "palignr     $0x8,%%xmm4,%%xmm4            \n"
+      "movq        %%xmm4,(%1,%4)                \n"
+      "lea         (%1,%4,2),%1                  \n"
+      "punpckldq   %%xmm6,%%xmm2                 \n"
+      "movdqa      %%xmm2,%%xmm6                 \n"
+      "movq        %%xmm2,(%1)                   \n"
+      "palignr     $0x8,%%xmm6,%%xmm6            \n"
+      "punpckldq   %%xmm5,%%xmm1                 \n"
+      "movq        %%xmm6,(%1,%4)                \n"
+      "lea         (%1,%4,2),%1                  \n"
+      "movdqa      %%xmm1,%%xmm5                 \n"
+      "movq        %%xmm1,(%1)                   \n"
+      "palignr     $0x8,%%xmm5,%%xmm5            \n"
+      "movq        %%xmm5,(%1,%4)                \n"
+      "lea         (%1,%4,2),%1                  \n"
+      "punpckldq   %%xmm7,%%xmm3                 \n"
+      "movq        %%xmm3,(%1)                   \n"
+      "movdqa      %%xmm3,%%xmm7                 \n"
+      "palignr     $0x8,%%xmm7,%%xmm7            \n"
+      "movq        %%xmm7,(%1,%4)                \n"
+      "lea         (%1,%4,2),%1                  \n"
+      "punpckldq   %%xmm12,%%xmm8                \n"
+      "movq        %%xmm8,(%1)                   \n"
+      "movdqa      %%xmm8,%%xmm12                \n"
+      "palignr     $0x8,%%xmm12,%%xmm12          \n"
+      "movq        %%xmm12,(%1,%4)               \n"
+      "lea         (%1,%4,2),%1                  \n"
+      "punpckldq   %%xmm14,%%xmm10               \n"
+      "movdqa      %%xmm10,%%xmm14               \n"
+      "movq        %%xmm10,(%1)                  \n"
+      "palignr     $0x8,%%xmm14,%%xmm14          \n"
+      "punpckldq   %%xmm13,%%xmm9                \n"
+      "movq        %%xmm14,(%1,%4)               \n"
+      "lea         (%1,%4,2),%1                  \n"
+      "movdqa      %%xmm9,%%xmm13                \n"
+      "movq        %%xmm9,(%1)                   \n"
+      "palignr     $0x8,%%xmm13,%%xmm13          \n"
+      "movq        %%xmm13,(%1,%4)               \n"
+      "lea         (%1,%4,2),%1                  \n"
+      "punpckldq   %%xmm15,%%xmm11               \n"
+      "movq        %%xmm11,(%1)                  \n"
+      "movdqa      %%xmm11,%%xmm15               \n"
+      "palignr     $0x8,%%xmm15,%%xmm15          \n"
+      "sub         $0x10,%2                      \n"
+      "movq        %%xmm15,(%1,%4)               \n"
+      "lea         (%1,%4,2),%1                  \n"
+      "jg          1b                            \n"
+      : "+r"(src),                    // %0
+        "+r"(dst),                    // %1
+        "+r"(width)                   // %2
+      : "r"((intptr_t)(src_stride)),  // %3
+        "r"((intptr_t)(dst_stride))   // %4
+      : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6",
+        "xmm7", "xmm8", "xmm9", "xmm10", "xmm11", "xmm12", "xmm13", "xmm14",
+        "xmm15");
+}
+#endif  // defined(HAS_TRANSPOSEWX8_FAST_SSSE3)
+
+// Transpose UV 8x8.  64 bit.
+#if defined(HAS_TRANSPOSEUVWX8_SSE2)
+void TransposeUVWx8_SSE2(const uint8_t* src,
+                         int src_stride,
+                         uint8_t* dst_a,
+                         int dst_stride_a,
+                         uint8_t* dst_b,
+                         int dst_stride_b,
+                         int width) {
+  asm volatile (
+      // Read in the data from the source pointer.
+      // First round of bit swap.
+      LABELALIGN
+      "1:                                        \n"
+      "movdqu      (%0),%%xmm0                   \n"
+      "movdqu      (%0,%4),%%xmm1                \n"
+      "lea         (%0,%4,2),%0                  \n"
+      "movdqa      %%xmm0,%%xmm8                 \n"
+      "punpcklbw   %%xmm1,%%xmm0                 \n"
+      "punpckhbw   %%xmm1,%%xmm8                 \n"
+      "movdqa      %%xmm8,%%xmm1                 \n"
+      "movdqu      (%0),%%xmm2                   \n"
+      "movdqu      (%0,%4),%%xmm3                \n"
+      "lea         (%0,%4,2),%0                  \n"
+      "movdqa      %%xmm2,%%xmm8                 \n"
+      "punpcklbw   %%xmm3,%%xmm2                 \n"
+      "punpckhbw   %%xmm3,%%xmm8                 \n"
+      "movdqa      %%xmm8,%%xmm3                 \n"
+      "movdqu      (%0),%%xmm4                   \n"
+      "movdqu      (%0,%4),%%xmm5                \n"
+      "lea         (%0,%4,2),%0                  \n"
+      "movdqa      %%xmm4,%%xmm8                 \n"
+      "punpcklbw   %%xmm5,%%xmm4                 \n"
+      "punpckhbw   %%xmm5,%%xmm8                 \n"
+      "movdqa      %%xmm8,%%xmm5                 \n"
+      "movdqu      (%0),%%xmm6                   \n"
+      "movdqu      (%0,%4),%%xmm7                \n"
+      "lea         (%0,%4,2),%0                  \n"
+      "movdqa      %%xmm6,%%xmm8                 \n"
+      "punpcklbw   %%xmm7,%%xmm6                 \n"
+      "neg         %4                            \n"
+      "lea         0x10(%0,%4,8),%0              \n"
+      "punpckhbw   %%xmm7,%%xmm8                 \n"
+      "movdqa      %%xmm8,%%xmm7                 \n"
+      "neg         %4                            \n"
+      // Second round of bit swap.
+      "movdqa      %%xmm0,%%xmm8                 \n"
+      "movdqa      %%xmm1,%%xmm9                 \n"
+      "punpckhwd   %%xmm2,%%xmm8                 \n"
+      "punpckhwd   %%xmm3,%%xmm9                 \n"
+      "punpcklwd   %%xmm2,%%xmm0                 \n"
+      "punpcklwd   %%xmm3,%%xmm1                 \n"
+      "movdqa      %%xmm8,%%xmm2                 \n"
+      "movdqa      %%xmm9,%%xmm3                 \n"
+      "movdqa      %%xmm4,%%xmm8                 \n"
+      "movdqa      %%xmm5,%%xmm9                 \n"
+      "punpckhwd   %%xmm6,%%xmm8                 \n"
+      "punpckhwd   %%xmm7,%%xmm9                 \n"
+      "punpcklwd   %%xmm6,%%xmm4                 \n"
+      "punpcklwd   %%xmm7,%%xmm5                 \n"
+      "movdqa      %%xmm8,%%xmm6                 \n"
+      "movdqa      %%xmm9,%%xmm7                 \n"
+      // Third round of bit swap.
+      // Write to the destination pointer.
+      "movdqa      %%xmm0,%%xmm8                 \n"
+      "punpckldq   %%xmm4,%%xmm0                 \n"
+      "movlpd      %%xmm0,(%1)                   \n"  // Write back U channel
+      "movhpd      %%xmm0,(%2)                   \n"  // Write back V channel
+      "punpckhdq   %%xmm4,%%xmm8                 \n"
+      "movlpd      %%xmm8,(%1,%5)                \n"
+      "lea         (%1,%5,2),%1                  \n"
+      "movhpd      %%xmm8,(%2,%6)                \n"
+      "lea         (%2,%6,2),%2                  \n"
+      "movdqa      %%xmm2,%%xmm8                 \n"
+      "punpckldq   %%xmm6,%%xmm2                 \n"
+      "movlpd      %%xmm2,(%1)                   \n"
+      "movhpd      %%xmm2,(%2)                   \n"
+      "punpckhdq   %%xmm6,%%xmm8                 \n"
+      "movlpd      %%xmm8,(%1,%5)                \n"
+      "lea         (%1,%5,2),%1                  \n"
+      "movhpd      %%xmm8,(%2,%6)                \n"
+      "lea         (%2,%6,2),%2                  \n"
+      "movdqa      %%xmm1,%%xmm8                 \n"
+      "punpckldq   %%xmm5,%%xmm1                 \n"
+      "movlpd      %%xmm1,(%1)                   \n"
+      "movhpd      %%xmm1,(%2)                   \n"
+      "punpckhdq   %%xmm5,%%xmm8                 \n"
+      "movlpd      %%xmm8,(%1,%5)                \n"
+      "lea         (%1,%5,2),%1                  \n"
+      "movhpd      %%xmm8,(%2,%6)                \n"
+      "lea         (%2,%6,2),%2                  \n"
+      "movdqa      %%xmm3,%%xmm8                 \n"
+      "punpckldq   %%xmm7,%%xmm3                 \n"
+      "movlpd      %%xmm3,(%1)                   \n"
+      "movhpd      %%xmm3,(%2)                   \n"
+      "punpckhdq   %%xmm7,%%xmm8                 \n"
+      "sub         $0x8,%3                       \n"
+      "movlpd      %%xmm8,(%1,%5)                \n"
+      "lea         (%1,%5,2),%1                  \n"
+      "movhpd      %%xmm8,(%2,%6)                \n"
+      "lea         (%2,%6,2),%2                  \n"
+      "jg          1b                            \n"
+      : "+r"(src),                      // %0
+        "+r"(dst_a),                    // %1
+        "+r"(dst_b),                    // %2
+        "+r"(width)                     // %3
+      : "r"((intptr_t)(src_stride)),    // %4
+        "r"((intptr_t)(dst_stride_a)),  // %5
+        "r"((intptr_t)(dst_stride_b))   // %6
+      : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6",
+        "xmm7", "xmm8", "xmm9");
+}
+#endif  // defined(HAS_TRANSPOSEUVWX8_SSE2)
+
+#if defined(HAS_TRANSPOSE4X4_32_SSE2)
+// 4 values, little endian view
+// a b c d
+// e f g h
+// i j k l
+// m n o p
+
+// transpose 2x2
+// a e b f   from row 0, 1
+// i m j n   from row 2, 3
+// c g d h   from row 0, 1
+// k o l p   from row 2, 3
+
+// transpose 4x4
+// a e i m   from row 0, 1
+// b f j n   from row 0, 1
+// c g k o   from row 2, 3
+// d h l p   from row 2, 3
+
+// Transpose 32 bit values (ARGB)
+void Transpose4x4_32_SSE2(const uint8_t* src,
+                          int src_stride,
+                          uint8_t* dst,
+                          int dst_stride,
+                          int width) {
+  asm volatile (
+      // Main loop transpose 4x4.  Read a column, write a row.
+      "1:                                        \n"
+      "movdqu      (%0),%%xmm0                   \n"  // a b c d
+      "movdqu      (%0,%3),%%xmm1                \n"  // e f g h
+      "lea         (%0,%3,2),%0                  \n"  // src += stride * 2
+      "movdqu      (%0),%%xmm2                   \n"  // i j k l
+      "movdqu      (%0,%3),%%xmm3                \n"  // m n o p
+      "lea         (%0,%3,2),%0                  \n"  // src += stride * 2
+
+      // Transpose 2x2
+      "movdqa      %%xmm0,%%xmm4                 \n"
+      "movdqa      %%xmm2,%%xmm5                 \n"
+      "movdqa      %%xmm0,%%xmm6                 \n"
+      "movdqa      %%xmm2,%%xmm7                 \n"
+      "punpckldq   %%xmm1,%%xmm4                 \n"  // a e b f   from row 0, 1
+      "punpckldq   %%xmm3,%%xmm5                 \n"  // i m j n   from row 2, 3
+      "punpckhdq   %%xmm1,%%xmm6                 \n"  // c g d h   from row 0, 1
+      "punpckhdq   %%xmm3,%%xmm7                 \n"  // k o l p   from row 2, 3
+
+      // Transpose 4x4
+      "movdqa      %%xmm4,%%xmm0                 \n"
+      "movdqa      %%xmm4,%%xmm1                 \n"
+      "movdqa      %%xmm6,%%xmm2                 \n"
+      "movdqa      %%xmm6,%%xmm3                 \n"
+      "punpcklqdq  %%xmm5,%%xmm0                 \n"  // a e i m   from row 0, 1
+      "punpckhqdq  %%xmm5,%%xmm1                 \n"  // b f j n   from row 0, 1
+      "punpcklqdq  %%xmm7,%%xmm2                 \n"  // c g k o   from row 2, 3
+      "punpckhqdq  %%xmm7,%%xmm3                 \n"  // d h l p   from row 2, 3
+
+      "movdqu      %%xmm0,(%1)                   \n"
+      "lea         16(%1,%4),%1                  \n"  // dst += stride + 16
+      "movdqu      %%xmm1,-16(%1)                \n"
+      "movdqu      %%xmm2,-16(%1,%4)             \n"
+      "movdqu      %%xmm3,-16(%1,%4,2)           \n"
+      "sub         %4,%1                         \n"
+      "sub         $0x4,%2                       \n"
+      "jg          1b                            \n"
+      : "+r"(src),                     // %0
+        "+r"(dst),                     // %1
+        "+rm"(width)                   // %2
+      : "r"((ptrdiff_t)(src_stride)),  // %3
+        "r"((ptrdiff_t)(dst_stride))   // %4
+      : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6",
+        "xmm7");
+}
+#endif  // defined(HAS_TRANSPOSE4X4_32_SSE2)
+
+#if defined(HAS_TRANSPOSE4X4_32_AVX2)
+
+// Transpose 32 bit values (ARGB)
+void Transpose4x4_32_AVX2(const uint8_t* src,
+                          int src_stride,
+                          uint8_t* dst,
+                          int dst_stride,
+                          int width) {
+  asm volatile (
+      // Main loop transpose 2 blocks of 4x4.  Read a column, write a row.
+      "1:                                        \n"
+      "vmovdqu     (%0),%%xmm0                   \n"  // a b c d
+      "vmovdqu     (%0,%3),%%xmm1                \n"  // e f g h
+      "lea         (%0,%3,2),%0                  \n"  // src += stride * 2
+      "vmovdqu     (%0),%%xmm2                   \n"  // i j k l
+      "vmovdqu     (%0,%3),%%xmm3                \n"  // m n o p
+      "lea         (%0,%3,2),%0                  \n"  // src += stride * 2
+
+      "vinserti128 $1,(%0),%%ymm0,%%ymm0         \n"  // a b c d
+      "vinserti128 $1,(%0,%3),%%ymm1,%%ymm1      \n"  // e f g h
+      "lea         (%0,%3,2),%0                  \n"  // src += stride * 2
+      "vinserti128 $1,(%0),%%ymm2,%%ymm2         \n"  // i j k l
+      "vinserti128 $1,(%0,%3),%%ymm3,%%ymm3      \n"  // m n o p
+      "lea         (%0,%3,2),%0                  \n"  // src += stride * 2
+
+      // Transpose 2x2
+      "vpunpckldq  %%ymm1,%%ymm0,%%ymm4          \n"  // a e b f   from row 0, 1
+      "vpunpckldq  %%ymm3,%%ymm2,%%ymm5          \n"  // i m j n   from row 2, 3
+      "vpunpckhdq  %%ymm1,%%ymm0,%%ymm6          \n"  // c g d h   from row 0, 1
+      "vpunpckhdq  %%ymm3,%%ymm2,%%ymm7          \n"  // k o l p   from row 2, 3
+
+      // Transpose 4x4
+      "vpunpcklqdq %%ymm5,%%ymm4,%%ymm0          \n"  // a e i m   from row 0, 1
+      "vpunpckhqdq %%ymm5,%%ymm4,%%ymm1          \n"  // b f j n   from row 0, 1
+      "vpunpcklqdq %%ymm7,%%ymm6,%%ymm2          \n"  // c g k o   from row 2, 3
+      "vpunpckhqdq %%ymm7,%%ymm6,%%ymm3          \n"  // d h l p   from row 2, 3
+
+      "vmovdqu     %%ymm0,(%1)                   \n"
+      "lea         32(%1,%4),%1                  \n"  // dst += stride + 32
+      "vmovdqu     %%ymm1,-32(%1)                \n"
+      "vmovdqu     %%ymm2,-32(%1,%4)             \n"
+      "vmovdqu     %%ymm3,-32(%1,%4,2)           \n"
+      "sub         %4,%1                         \n"
+      "sub         $0x8,%2                       \n"
+      "jg          1b                            \n"
+      "vzeroupper                                \n"
+      : "+r"(src),                     // %0
+        "+r"(dst),                     // %1
+        "+rm"(width)                   // %2
+      : "r"((ptrdiff_t)(src_stride)),  // %3
+        "r"((ptrdiff_t)(dst_stride))   // %4
+      : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6",
+        "xmm7");
+}
+#endif  // defined(HAS_TRANSPOSE4X4_32_AVX2)
+
+#endif  // defined(__x86_64__) || defined(__i386__)
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif

yuv.mod/libyuv/source/rotate_lsx.cc

@@ -0,0 +1,233 @@
+/*
+ *  Copyright 2022 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Copyright (c) 2022 Loongson Technology Corporation Limited
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/rotate_row.h"
+
+#if !defined(LIBYUV_DISABLE_LSX) && defined(__loongarch_sx)
+#include "libyuv/loongson_intrinsics.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+#define ILVLH_B(in0, in1, in2, in3, out0, out1, out2, out3)   \
+  {                                                           \
+    DUP2_ARG2(__lsx_vilvl_b, in1, in0, in3, in2, out0, out2); \
+    DUP2_ARG2(__lsx_vilvh_b, in1, in0, in3, in2, out1, out3); \
+  }
+
+#define ILVLH_H(in0, in1, in2, in3, out0, out1, out2, out3)   \
+  {                                                           \
+    DUP2_ARG2(__lsx_vilvl_h, in1, in0, in3, in2, out0, out2); \
+    DUP2_ARG2(__lsx_vilvh_h, in1, in0, in3, in2, out1, out3); \
+  }
+
+#define ILVLH_W(in0, in1, in2, in3, out0, out1, out2, out3)   \
+  {                                                           \
+    DUP2_ARG2(__lsx_vilvl_w, in1, in0, in3, in2, out0, out2); \
+    DUP2_ARG2(__lsx_vilvh_w, in1, in0, in3, in2, out1, out3); \
+  }
+
+#define ILVLH_D(in0, in1, in2, in3, out0, out1, out2, out3)   \
+  {                                                           \
+    DUP2_ARG2(__lsx_vilvl_d, in1, in0, in3, in2, out0, out2); \
+    DUP2_ARG2(__lsx_vilvh_d, in1, in0, in3, in2, out1, out3); \
+  }
+
+#define LSX_ST_4(_dst0, _dst1, _dst2, _dst3, _dst, _stride, _stride2, \
+                 _stride3, _stride4)                                  \
+  {                                                                   \
+    __lsx_vst(_dst0, _dst, 0);                                        \
+    __lsx_vstx(_dst1, _dst, _stride);                                 \
+    __lsx_vstx(_dst2, _dst, _stride2);                                \
+    __lsx_vstx(_dst3, _dst, _stride3);                                \
+    _dst += _stride4;                                                 \
+  }
+
+#define LSX_ST_2(_dst0, _dst1, _dst, _stride, _stride2) \
+  {                                                     \
+    __lsx_vst(_dst0, _dst, 0);                          \
+    __lsx_vstx(_dst1, _dst, _stride);                   \
+    _dst += _stride2;                                   \
+  }
+
+void TransposeUVWx16_C(const uint8_t* src,
+                       int src_stride,
+                       uint8_t* dst_a,
+                       int dst_stride_a,
+                       uint8_t* dst_b,
+                       int dst_stride_b,
+                       int width) {
+  TransposeUVWx8_C(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
+                   width);
+  TransposeUVWx8_C((src + 8 * src_stride), src_stride, (dst_a + 8),
+                   dst_stride_a, (dst_b + 8), dst_stride_b, width);
+}
+
+void TransposeWx16_LSX(const uint8_t* src,
+                       int src_stride,
+                       uint8_t* dst,
+                       int dst_stride,
+                       int width) {
+  int x;
+  int len = width / 16;
+  uint8_t* s;
+  int src_stride2 = src_stride << 1;
+  int src_stride3 = src_stride + src_stride2;
+  int src_stride4 = src_stride2 << 1;
+  int dst_stride2 = dst_stride << 1;
+  int dst_stride3 = dst_stride + dst_stride2;
+  int dst_stride4 = dst_stride2 << 1;
+  __m128i src0, src1, src2, src3, dst0, dst1, dst2, dst3;
+  __m128i tmp0, tmp1, tmp2, tmp3;
+  __m128i reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7;
+  __m128i res0, res1, res2, res3, res4, res5, res6, res7, res8, res9;
+
+  for (x = 0; x < len; x++) {
+    s = (uint8_t*)src;
+    src0 = __lsx_vld(s, 0);
+    src1 = __lsx_vldx(s, src_stride);
+    src2 = __lsx_vldx(s, src_stride2);
+    src3 = __lsx_vldx(s, src_stride3);
+    s += src_stride4;
+    ILVLH_B(src0, src1, src2, src3, tmp0, tmp1, tmp2, tmp3);
+    ILVLH_H(tmp0, tmp2, tmp1, tmp3, reg0, reg1, reg2, reg3);
+    src0 = __lsx_vld(s, 0);
+    src1 = __lsx_vldx(s, src_stride);
+    src2 = __lsx_vldx(s, src_stride2);
+    src3 = __lsx_vldx(s, src_stride3);
+    s += src_stride4;
+    ILVLH_B(src0, src1, src2, src3, tmp0, tmp1, tmp2, tmp3);
+    ILVLH_H(tmp0, tmp2, tmp1, tmp3, reg4, reg5, reg6, reg7);
+    ILVLH_W(reg0, reg4, reg1, reg5, res0, res1, res2, res3);
+    ILVLH_W(reg2, reg6, reg3, reg7, res4, res5, res6, res7);
+    src0 = __lsx_vld(s, 0);
+    src1 = __lsx_vldx(s, src_stride);
+    src2 = __lsx_vldx(s, src_stride2);
+    src3 = __lsx_vldx(s, src_stride3);
+    s += src_stride4;
+    ILVLH_B(src0, src1, src2, src3, tmp0, tmp1, tmp2, tmp3);
+    ILVLH_H(tmp0, tmp2, tmp1, tmp3, reg0, reg1, reg2, reg3);
+    src0 = __lsx_vld(s, 0);
+    src1 = __lsx_vldx(s, src_stride);
+    src2 = __lsx_vldx(s, src_stride2);
+    src3 = __lsx_vldx(s, src_stride3);
+    s += src_stride4;
+    ILVLH_B(src0, src1, src2, src3, tmp0, tmp1, tmp2, tmp3);
+    ILVLH_H(tmp0, tmp2, tmp1, tmp3, reg4, reg5, reg6, reg7);
+    res8 = __lsx_vilvl_w(reg4, reg0);
+    res9 = __lsx_vilvh_w(reg4, reg0);
+    ILVLH_D(res0, res8, res1, res9, dst0, dst1, dst2, dst3);
+    LSX_ST_4(dst0, dst1, dst2, dst3, dst, dst_stride, dst_stride2, dst_stride3,
+             dst_stride4);
+    res8 = __lsx_vilvl_w(reg5, reg1);
+    res9 = __lsx_vilvh_w(reg5, reg1);
+    ILVLH_D(res2, res8, res3, res9, dst0, dst1, dst2, dst3);
+    LSX_ST_4(dst0, dst1, dst2, dst3, dst, dst_stride, dst_stride2, dst_stride3,
+             dst_stride4);
+    res8 = __lsx_vilvl_w(reg6, reg2);
+    res9 = __lsx_vilvh_w(reg6, reg2);
+    ILVLH_D(res4, res8, res5, res9, dst0, dst1, dst2, dst3);
+    LSX_ST_4(dst0, dst1, dst2, dst3, dst, dst_stride, dst_stride2, dst_stride3,
+             dst_stride4);
+    res8 = __lsx_vilvl_w(reg7, reg3);
+    res9 = __lsx_vilvh_w(reg7, reg3);
+    ILVLH_D(res6, res8, res7, res9, dst0, dst1, dst2, dst3);
+    LSX_ST_4(dst0, dst1, dst2, dst3, dst, dst_stride, dst_stride2, dst_stride3,
+             dst_stride4);
+    src += 16;
+  }
+}
+
+void TransposeUVWx16_LSX(const uint8_t* src,
+                         int src_stride,
+                         uint8_t* dst_a,
+                         int dst_stride_a,
+                         uint8_t* dst_b,
+                         int dst_stride_b,
+                         int width) {
+  int x;
+  int len = width / 8;
+  uint8_t* s;
+  int src_stride2 = src_stride << 1;
+  int src_stride3 = src_stride + src_stride2;
+  int src_stride4 = src_stride2 << 1;
+  int dst_stride_a2 = dst_stride_a << 1;
+  int dst_stride_b2 = dst_stride_b << 1;
+  __m128i src0, src1, src2, src3, dst0, dst1, dst2, dst3;
+  __m128i tmp0, tmp1, tmp2, tmp3;
+  __m128i reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7;
+  __m128i res0, res1, res2, res3, res4, res5, res6, res7, res8, res9;
+
+  for (x = 0; x < len; x++) {
+    s = (uint8_t*)src;
+    src0 = __lsx_vld(s, 0);
+    src1 = __lsx_vldx(s, src_stride);
+    src2 = __lsx_vldx(s, src_stride2);
+    src3 = __lsx_vldx(s, src_stride3);
+    s += src_stride4;
+    ILVLH_B(src0, src1, src2, src3, tmp0, tmp1, tmp2, tmp3);
+    ILVLH_H(tmp0, tmp2, tmp1, tmp3, reg0, reg1, reg2, reg3);
+    src0 = __lsx_vld(s, 0);
+    src1 = __lsx_vldx(s, src_stride);
+    src2 = __lsx_vldx(s, src_stride2);
+    src3 = __lsx_vldx(s, src_stride3);
+    s += src_stride4;
+    ILVLH_B(src0, src1, src2, src3, tmp0, tmp1, tmp2, tmp3);
+    ILVLH_H(tmp0, tmp2, tmp1, tmp3, reg4, reg5, reg6, reg7);
+    ILVLH_W(reg0, reg4, reg1, reg5, res0, res1, res2, res3);
+    ILVLH_W(reg2, reg6, reg3, reg7, res4, res5, res6, res7);
+    src0 = __lsx_vld(s, 0);
+    src1 = __lsx_vldx(s, src_stride);
+    src2 = __lsx_vldx(s, src_stride2);
+    src3 = __lsx_vldx(s, src_stride3);
+    s += src_stride4;
+    ILVLH_B(src0, src1, src2, src3, tmp0, tmp1, tmp2, tmp3);
+    ILVLH_H(tmp0, tmp2, tmp1, tmp3, reg0, reg1, reg2, reg3);
+    src0 = __lsx_vld(s, 0);
+    src1 = __lsx_vldx(s, src_stride);
+    src2 = __lsx_vldx(s, src_stride2);
+    src3 = __lsx_vldx(s, src_stride3);
+    s += src_stride4;
+    ILVLH_B(src0, src1, src2, src3, tmp0, tmp1, tmp2, tmp3);
+    ILVLH_H(tmp0, tmp2, tmp1, tmp3, reg4, reg5, reg6, reg7);
+    res8 = __lsx_vilvl_w(reg4, reg0);
+    res9 = __lsx_vilvh_w(reg4, reg0);
+    ILVLH_D(res0, res8, res1, res9, dst0, dst1, dst2, dst3);
+    LSX_ST_2(dst0, dst2, dst_a, dst_stride_a, dst_stride_a2);
+    LSX_ST_2(dst1, dst3, dst_b, dst_stride_b, dst_stride_b2);
+    res8 = __lsx_vilvl_w(reg5, reg1);
+    res9 = __lsx_vilvh_w(reg5, reg1);
+    ILVLH_D(res2, res8, res3, res9, dst0, dst1, dst2, dst3);
+    LSX_ST_2(dst0, dst2, dst_a, dst_stride_a, dst_stride_a2);
+    LSX_ST_2(dst1, dst3, dst_b, dst_stride_b, dst_stride_b2);
+    res8 = __lsx_vilvl_w(reg6, reg2);
+    res9 = __lsx_vilvh_w(reg6, reg2);
+    ILVLH_D(res4, res8, res5, res9, dst0, dst1, dst2, dst3);
+    LSX_ST_2(dst0, dst2, dst_a, dst_stride_a, dst_stride_a2);
+    LSX_ST_2(dst1, dst3, dst_b, dst_stride_b, dst_stride_b2);
+    res8 = __lsx_vilvl_w(reg7, reg3);
+    res9 = __lsx_vilvh_w(reg7, reg3);
+    ILVLH_D(res6, res8, res7, res9, dst0, dst1, dst2, dst3);
+    LSX_ST_2(dst0, dst2, dst_a, dst_stride_a, dst_stride_a2);
+    LSX_ST_2(dst1, dst3, dst_b, dst_stride_b, dst_stride_b2);
+    src += 16;
+  }
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // !defined(LIBYUV_DISABLE_LSX) && defined(__loongarch_sx)

yuv.mod/libyuv/source/rotate_msa.cc

@@ -0,0 +1,240 @@
+/*
+ *  Copyright 2016 The LibYuv Project Authors. All rights reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS. All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "libyuv/rotate_row.h"
+
+// This module is for GCC MSA
+#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
+#include "libyuv/macros_msa.h"
+
+#ifdef __cplusplus
+namespace libyuv {
+extern "C" {
+#endif
+
+#define ILVRL_B(in0, in1, in2, in3, out0, out1, out2, out3) \
+  {                                                         \
+    out0 = (v16u8)__msa_ilvr_b((v16i8)in1, (v16i8)in0);     \
+    out1 = (v16u8)__msa_ilvl_b((v16i8)in1, (v16i8)in0);     \
+    out2 = (v16u8)__msa_ilvr_b((v16i8)in3, (v16i8)in2);     \
+    out3 = (v16u8)__msa_ilvl_b((v16i8)in3, (v16i8)in2);     \
+  }
+
+#define ILVRL_H(in0, in1, in2, in3, out0, out1, out2, out3) \
+  {                                                         \
+    out0 = (v16u8)__msa_ilvr_h((v8i16)in1, (v8i16)in0);     \
+    out1 = (v16u8)__msa_ilvl_h((v8i16)in1, (v8i16)in0);     \
+    out2 = (v16u8)__msa_ilvr_h((v8i16)in3, (v8i16)in2);     \
+    out3 = (v16u8)__msa_ilvl_h((v8i16)in3, (v8i16)in2);     \
+  }
+
+#define ILVRL_W(in0, in1, in2, in3, out0, out1, out2, out3) \
+  {                                                         \
+    out0 = (v16u8)__msa_ilvr_w((v4i32)in1, (v4i32)in0);     \
+    out1 = (v16u8)__msa_ilvl_w((v4i32)in1, (v4i32)in0);     \
+    out2 = (v16u8)__msa_ilvr_w((v4i32)in3, (v4i32)in2);     \
+    out3 = (v16u8)__msa_ilvl_w((v4i32)in3, (v4i32)in2);     \
+  }
+
+#define ILVRL_D(in0, in1, in2, in3, out0, out1, out2, out3) \
+  {                                                         \
+    out0 = (v16u8)__msa_ilvr_d((v2i64)in1, (v2i64)in0);     \
+    out1 = (v16u8)__msa_ilvl_d((v2i64)in1, (v2i64)in0);     \
+    out2 = (v16u8)__msa_ilvr_d((v2i64)in3, (v2i64)in2);     \
+    out3 = (v16u8)__msa_ilvl_d((v2i64)in3, (v2i64)in2);     \
+  }
+
+void TransposeUVWx16_C(const uint8_t* src,
+                       int src_stride,
+                       uint8_t* dst_a,
+                       int dst_stride_a,
+                       uint8_t* dst_b,
+                       int dst_stride_b,
+                       int width) {
+  TransposeUVWx8_C(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
+                   width);
+  TransposeUVWx8_C((src + 8 * src_stride), src_stride, (dst_a + 8),
+                   dst_stride_a, (dst_b + 8), dst_stride_b, width);
+}
+
+void TransposeWx16_MSA(const uint8_t* src,
+                       int src_stride,
+                       uint8_t* dst,
+                       int dst_stride,
+                       int width) {
+  int x;
+  const uint8_t* s;
+  v16u8 src0, src1, src2, src3, dst0, dst1, dst2, dst3, vec0, vec1, vec2, vec3;
+  v16u8 reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7;
+  v16u8 res0, res1, res2, res3, res4, res5, res6, res7, res8, res9;
+
+  for (x = 0; x < width; x += 16) {
+    s = src;
+    src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
+    ILVRL_H(vec0, vec2, vec1, vec3, reg0, reg1, reg2, reg3);
+    src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
+    ILVRL_H(vec0, vec2, vec1, vec3, reg4, reg5, reg6, reg7);
+    ILVRL_W(reg0, reg4, reg1, reg5, res0, res1, res2, res3);
+    ILVRL_W(reg2, reg6, reg3, reg7, res4, res5, res6, res7);
+    src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
+    ILVRL_H(vec0, vec2, vec1, vec3, reg0, reg1, reg2, reg3);
+    src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
+    ILVRL_H(vec0, vec2, vec1, vec3, reg4, reg5, reg6, reg7);
+    res8 = (v16u8)__msa_ilvr_w((v4i32)reg4, (v4i32)reg0);
+    res9 = (v16u8)__msa_ilvl_w((v4i32)reg4, (v4i32)reg0);
+    ILVRL_D(res0, res8, res1, res9, dst0, dst1, dst2, dst3);
+    ST_UB4(dst0, dst1, dst2, dst3, dst, dst_stride);
+    dst += dst_stride * 4;
+    res8 = (v16u8)__msa_ilvr_w((v4i32)reg5, (v4i32)reg1);
+    res9 = (v16u8)__msa_ilvl_w((v4i32)reg5, (v4i32)reg1);
+    ILVRL_D(res2, res8, res3, res9, dst0, dst1, dst2, dst3);
+    ST_UB4(dst0, dst1, dst2, dst3, dst, dst_stride);
+    dst += dst_stride * 4;
+    res8 = (v16u8)__msa_ilvr_w((v4i32)reg6, (v4i32)reg2);
+    res9 = (v16u8)__msa_ilvl_w((v4i32)reg6, (v4i32)reg2);
+    ILVRL_D(res4, res8, res5, res9, dst0, dst1, dst2, dst3);
+    ST_UB4(dst0, dst1, dst2, dst3, dst, dst_stride);
+    dst += dst_stride * 4;
+    res8 = (v16u8)__msa_ilvr_w((v4i32)reg7, (v4i32)reg3);
+    res9 = (v16u8)__msa_ilvl_w((v4i32)reg7, (v4i32)reg3);
+    ILVRL_D(res6, res8, res7, res9, dst0, dst1, dst2, dst3);
+    ST_UB4(dst0, dst1, dst2, dst3, dst, dst_stride);
+    src += 16;
+    dst += dst_stride * 4;
+  }
+}
+
+void TransposeUVWx16_MSA(const uint8_t* src,
+                         int src_stride,
+                         uint8_t* dst_a,
+                         int dst_stride_a,
+                         uint8_t* dst_b,
+                         int dst_stride_b,
+                         int width) {
+  int x;
+  const uint8_t* s;
+  v16u8 src0, src1, src2, src3, dst0, dst1, dst2, dst3, vec0, vec1, vec2, vec3;
+  v16u8 reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7;
+  v16u8 res0, res1, res2, res3, res4, res5, res6, res7, res8, res9;
+
+  for (x = 0; x < width; x += 8) {
+    s = src;
+    src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
+    ILVRL_H(vec0, vec2, vec1, vec3, reg0, reg1, reg2, reg3);
+    src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
+    ILVRL_H(vec0, vec2, vec1, vec3, reg4, reg5, reg6, reg7);
+    ILVRL_W(reg0, reg4, reg1, reg5, res0, res1, res2, res3);
+    ILVRL_W(reg2, reg6, reg3, reg7, res4, res5, res6, res7);
+    src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
+    ILVRL_H(vec0, vec2, vec1, vec3, reg0, reg1, reg2, reg3);
+    src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
+    s += src_stride;
+    ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
+    ILVRL_H(vec0, vec2, vec1, vec3, reg4, reg5, reg6, reg7);
+    res8 = (v16u8)__msa_ilvr_w((v4i32)reg4, (v4i32)reg0);
+    res9 = (v16u8)__msa_ilvl_w((v4i32)reg4, (v4i32)reg0);
+    ILVRL_D(res0, res8, res1, res9, dst0, dst1, dst2, dst3);
+    ST_UB2(dst0, dst2, dst_a, dst_stride_a);
+    ST_UB2(dst1, dst3, dst_b, dst_stride_b);
+    dst_a += dst_stride_a * 2;
+    dst_b += dst_stride_b * 2;
+    res8 = (v16u8)__msa_ilvr_w((v4i32)reg5, (v4i32)reg1);
+    res9 = (v16u8)__msa_ilvl_w((v4i32)reg5, (v4i32)reg1);
+    ILVRL_D(res2, res8, res3, res9, dst0, dst1, dst2, dst3);
+    ST_UB2(dst0, dst2, dst_a, dst_stride_a);
+    ST_UB2(dst1, dst3, dst_b, dst_stride_b);
+    dst_a += dst_stride_a * 2;
+    dst_b += dst_stride_b * 2;
+    res8 = (v16u8)__msa_ilvr_w((v4i32)reg6, (v4i32)reg2);
+    res9 = (v16u8)__msa_ilvl_w((v4i32)reg6, (v4i32)reg2);
+    ILVRL_D(res4, res8, res5, res9, dst0, dst1, dst2, dst3);
+    ST_UB2(dst0, dst2, dst_a, dst_stride_a);
+    ST_UB2(dst1, dst3, dst_b, dst_stride_b);
+    dst_a += dst_stride_a * 2;
+    dst_b += dst_stride_b * 2;
+    res8 = (v16u8)__msa_ilvr_w((v4i32)reg7, (v4i32)reg3);
+    res9 = (v16u8)__msa_ilvl_w((v4i32)reg7, (v4i32)reg3);
+    ILVRL_D(res6, res8, res7, res9, dst0, dst1, dst2, dst3);
+    ST_UB2(dst0, dst2, dst_a, dst_stride_a);
+    ST_UB2(dst1, dst3, dst_b, dst_stride_b);
+    src += 16;
+    dst_a += dst_stride_a * 2;
+    dst_b += dst_stride_b * 2;
+  }
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+}  // namespace libyuv
+#endif
+
+#endif  // !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)