godot/drivers/vulkan/rendering_device_vulkan.cpp

/*************************************************************************/
/*  rendering_device_vulkan.cpp                                          */
/*************************************************************************/
/*                       This file is part of:                           */
/*                           GODOT ENGINE                                */
/*                      https://godotengine.org                          */
/*************************************************************************/
/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur.                 */
/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md).   */
/*                                                                       */
/* Permission is hereby granted, free of charge, to any person obtaining */
/* a copy of this software and associated documentation files (the       */
/* "Software"), to deal in the Software without restriction, including   */
/* without limitation the rights to use, copy, modify, merge, publish,   */
/* distribute, sublicense, and/or sell copies of the Software, and to    */
/* permit persons to whom the Software is furnished to do so, subject to */
/* the following conditions:                                             */
/*                                                                       */
/* The above copyright notice and this permission notice shall be        */
/* included in all copies or substantial portions of the Software.       */
/*                                                                       */
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,       */
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF    */
/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY  */
/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,  */
/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE     */
/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.                */
/*************************************************************************/

#include "rendering_device_vulkan.h"

#include "core/config/project_settings.h"
#include "core/io/compression.h"
#include "core/io/file_access.h"
#include "core/io/marshalls.h"
#include "core/os/os.h"
#include "core/templates/hashfuncs.h"
#include "core/version.h"
#include "drivers/vulkan/vulkan_context.h"

#include "thirdparty/misc/smolv.h"
#include "thirdparty/spirv-reflect/spirv_reflect.h"

//#define FORCE_FULL_BARRIER

static const uint32_t SMALL_ALLOCATION_MAX_SIZE = 4096;

// Get the Vulkan object information and possible stage access types (bitwise OR'd with incoming values).
RenderingDeviceVulkan::Buffer *RenderingDeviceVulkan::_get_buffer_from_owner(RID p_buffer, VkPipelineStageFlags &r_stage_mask, VkAccessFlags &r_access_mask, BitField<BarrierMask> p_post_barrier) {
	Buffer *buffer = nullptr;
	if (vertex_buffer_owner.owns(p_buffer)) {
		buffer = vertex_buffer_owner.get_or_null(p_buffer);

		r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT;
		r_access_mask |= VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT;
		if (buffer->usage & VK_BUFFER_USAGE_STORAGE_BUFFER_BIT) {
			if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) {
				r_access_mask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
				r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
			}
			if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) {
				r_access_mask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
				r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
			}
		}
	} else if (index_buffer_owner.owns(p_buffer)) {
		r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT;
		r_access_mask |= VK_ACCESS_INDEX_READ_BIT;
		buffer = index_buffer_owner.get_or_null(p_buffer);
	} else if (uniform_buffer_owner.owns(p_buffer)) {
		if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) {
			r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
		}
		if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) {
			r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
		}
		r_access_mask |= VK_ACCESS_UNIFORM_READ_BIT;
		buffer = uniform_buffer_owner.get_or_null(p_buffer);
	} else if (texture_buffer_owner.owns(p_buffer)) {
		if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) {
			r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
			r_access_mask |= VK_ACCESS_SHADER_READ_BIT;
		}
		if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) {
			r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
			r_access_mask |= VK_ACCESS_SHADER_READ_BIT;
		}

		buffer = &texture_buffer_owner.get_or_null(p_buffer)->buffer;
	} else if (storage_buffer_owner.owns(p_buffer)) {
		buffer = storage_buffer_owner.get_or_null(p_buffer);
		if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) {
			r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
			r_access_mask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
		}
		if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) {
			r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
			r_access_mask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
		}

		if (buffer->usage & VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT) {
			r_stage_mask |= VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT;
			r_access_mask |= VK_ACCESS_INDIRECT_COMMAND_READ_BIT;
		}
	}
	return buffer;
}

static void update_external_dependency_for_store(VkSubpassDependency2KHR &dependency, bool is_sampled, bool is_storage, bool is_depth) {
	// Transitioning from write to read, protect the shaders that may use this next.
	// Allow for copies/image layout transitions.
	dependency.dstStageMask |= VK_PIPELINE_STAGE_TRANSFER_BIT;
	dependency.dstAccessMask |= VK_ACCESS_TRANSFER_READ_BIT;

	if (is_sampled) {
		dependency.dstStageMask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
		dependency.dstAccessMask |= VK_ACCESS_SHADER_READ_BIT;
	} else if (is_storage) {
		dependency.dstStageMask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
		dependency.dstAccessMask |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
	} else {
		dependency.dstStageMask |= VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
		dependency.dstAccessMask |= VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
	}

	if (is_depth) {
		// Depth resources have additional stages that may be interested in them.
		dependency.dstStageMask |= VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
		dependency.dstAccessMask |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
	}
}

void RenderingDeviceVulkan::_add_dependency(RID p_id, RID p_depends_on) {
	if (!dependency_map.has(p_depends_on)) {
		dependency_map[p_depends_on] = HashSet<RID>();
	}

	dependency_map[p_depends_on].insert(p_id);

	if (!reverse_dependency_map.has(p_id)) {
		reverse_dependency_map[p_id] = HashSet<RID>();
	}

	reverse_dependency_map[p_id].insert(p_depends_on);
}

void RenderingDeviceVulkan::_free_dependencies(RID p_id) {
	// Direct dependencies must be freed.

	HashMap<RID, HashSet<RID>>::Iterator E = dependency_map.find(p_id);
	if (E) {
		while (E->value.size()) {
			free(*E->value.begin());
		}
		dependency_map.remove(E);
	}

	// Reverse dependencies must be unreferenced.
	E = reverse_dependency_map.find(p_id);

	if (E) {
		for (const RID &F : E->value) {
			HashMap<RID, HashSet<RID>>::Iterator G = dependency_map.find(F);
			ERR_CONTINUE(!G);
			ERR_CONTINUE(!G->value.has(p_id));
			G->value.erase(p_id);
		}

		reverse_dependency_map.remove(E);
	}
}

const VkFormat RenderingDeviceVulkan::vulkan_formats[RenderingDevice::DATA_FORMAT_MAX] = {
	VK_FORMAT_R4G4_UNORM_PACK8,
	VK_FORMAT_R4G4B4A4_UNORM_PACK16,
	VK_FORMAT_B4G4R4A4_UNORM_PACK16,
	VK_FORMAT_R5G6B5_UNORM_PACK16,
	VK_FORMAT_B5G6R5_UNORM_PACK16,
	VK_FORMAT_R5G5B5A1_UNORM_PACK16,
	VK_FORMAT_B5G5R5A1_UNORM_PACK16,
	VK_FORMAT_A1R5G5B5_UNORM_PACK16,
	VK_FORMAT_R8_UNORM,
	VK_FORMAT_R8_SNORM,
	VK_FORMAT_R8_USCALED,
	VK_FORMAT_R8_SSCALED,
	VK_FORMAT_R8_UINT,
	VK_FORMAT_R8_SINT,
	VK_FORMAT_R8_SRGB,
	VK_FORMAT_R8G8_UNORM,
	VK_FORMAT_R8G8_SNORM,
	VK_FORMAT_R8G8_USCALED,
	VK_FORMAT_R8G8_SSCALED,
	VK_FORMAT_R8G8_UINT,
	VK_FORMAT_R8G8_SINT,
	VK_FORMAT_R8G8_SRGB,
	VK_FORMAT_R8G8B8_UNORM,
	VK_FORMAT_R8G8B8_SNORM,
	VK_FORMAT_R8G8B8_USCALED,
	VK_FORMAT_R8G8B8_SSCALED,
	VK_FORMAT_R8G8B8_UINT,
	VK_FORMAT_R8G8B8_SINT,
	VK_FORMAT_R8G8B8_SRGB,
	VK_FORMAT_B8G8R8_UNORM,
	VK_FORMAT_B8G8R8_SNORM,
	VK_FORMAT_B8G8R8_USCALED,
	VK_FORMAT_B8G8R8_SSCALED,
	VK_FORMAT_B8G8R8_UINT,
	VK_FORMAT_B8G8R8_SINT,
	VK_FORMAT_B8G8R8_SRGB,
	VK_FORMAT_R8G8B8A8_UNORM,
	VK_FORMAT_R8G8B8A8_SNORM,
	VK_FORMAT_R8G8B8A8_USCALED,
	VK_FORMAT_R8G8B8A8_SSCALED,
	VK_FORMAT_R8G8B8A8_UINT,
	VK_FORMAT_R8G8B8A8_SINT,
	VK_FORMAT_R8G8B8A8_SRGB,
	VK_FORMAT_B8G8R8A8_UNORM,
	VK_FORMAT_B8G8R8A8_SNORM,
	VK_FORMAT_B8G8R8A8_USCALED,
	VK_FORMAT_B8G8R8A8_SSCALED,
	VK_FORMAT_B8G8R8A8_UINT,
	VK_FORMAT_B8G8R8A8_SINT,
	VK_FORMAT_B8G8R8A8_SRGB,
	VK_FORMAT_A8B8G8R8_UNORM_PACK32,
	VK_FORMAT_A8B8G8R8_SNORM_PACK32,
	VK_FORMAT_A8B8G8R8_USCALED_PACK32,
	VK_FORMAT_A8B8G8R8_SSCALED_PACK32,
	VK_FORMAT_A8B8G8R8_UINT_PACK32,
	VK_FORMAT_A8B8G8R8_SINT_PACK32,
	VK_FORMAT_A8B8G8R8_SRGB_PACK32,
	VK_FORMAT_A2R10G10B10_UNORM_PACK32,
	VK_FORMAT_A2R10G10B10_SNORM_PACK32,
	VK_FORMAT_A2R10G10B10_USCALED_PACK32,
	VK_FORMAT_A2R10G10B10_SSCALED_PACK32,
	VK_FORMAT_A2R10G10B10_UINT_PACK32,
	VK_FORMAT_A2R10G10B10_SINT_PACK32,
	VK_FORMAT_A2B10G10R10_UNORM_PACK32,
	VK_FORMAT_A2B10G10R10_SNORM_PACK32,
	VK_FORMAT_A2B10G10R10_USCALED_PACK32,
	VK_FORMAT_A2B10G10R10_SSCALED_PACK32,
	VK_FORMAT_A2B10G10R10_UINT_PACK32,
	VK_FORMAT_A2B10G10R10_SINT_PACK32,
	VK_FORMAT_R16_UNORM,
	VK_FORMAT_R16_SNORM,
	VK_FORMAT_R16_USCALED,
	VK_FORMAT_R16_SSCALED,
	VK_FORMAT_R16_UINT,
	VK_FORMAT_R16_SINT,
	VK_FORMAT_R16_SFLOAT,
	VK_FORMAT_R16G16_UNORM,
	VK_FORMAT_R16G16_SNORM,
	VK_FORMAT_R16G16_USCALED,
	VK_FORMAT_R16G16_SSCALED,
	VK_FORMAT_R16G16_UINT,
	VK_FORMAT_R16G16_SINT,
	VK_FORMAT_R16G16_SFLOAT,
	VK_FORMAT_R16G16B16_UNORM,
	VK_FORMAT_R16G16B16_SNORM,
	VK_FORMAT_R16G16B16_USCALED,
	VK_FORMAT_R16G16B16_SSCALED,
	VK_FORMAT_R16G16B16_UINT,
	VK_FORMAT_R16G16B16_SINT,
	VK_FORMAT_R16G16B16_SFLOAT,
	VK_FORMAT_R16G16B16A16_UNORM,
	VK_FORMAT_R16G16B16A16_SNORM,
	VK_FORMAT_R16G16B16A16_USCALED,
	VK_FORMAT_R16G16B16A16_SSCALED,
	VK_FORMAT_R16G16B16A16_UINT,
	VK_FORMAT_R16G16B16A16_SINT,
	VK_FORMAT_R16G16B16A16_SFLOAT,
	VK_FORMAT_R32_UINT,
	VK_FORMAT_R32_SINT,
	VK_FORMAT_R32_SFLOAT,
	VK_FORMAT_R32G32_UINT,
	VK_FORMAT_R32G32_SINT,
	VK_FORMAT_R32G32_SFLOAT,
	VK_FORMAT_R32G32B32_UINT,
	VK_FORMAT_R32G32B32_SINT,
	VK_FORMAT_R32G32B32_SFLOAT,
	VK_FORMAT_R32G32B32A32_UINT,
	VK_FORMAT_R32G32B32A32_SINT,
	VK_FORMAT_R32G32B32A32_SFLOAT,
	VK_FORMAT_R64_UINT,
	VK_FORMAT_R64_SINT,
	VK_FORMAT_R64_SFLOAT,
	VK_FORMAT_R64G64_UINT,
	VK_FORMAT_R64G64_SINT,
	VK_FORMAT_R64G64_SFLOAT,
	VK_FORMAT_R64G64B64_UINT,
	VK_FORMAT_R64G64B64_SINT,
	VK_FORMAT_R64G64B64_SFLOAT,
	VK_FORMAT_R64G64B64A64_UINT,
	VK_FORMAT_R64G64B64A64_SINT,
	VK_FORMAT_R64G64B64A64_SFLOAT,
	VK_FORMAT_B10G11R11_UFLOAT_PACK32,
	VK_FORMAT_E5B9G9R9_UFLOAT_PACK32,
	VK_FORMAT_D16_UNORM,
	VK_FORMAT_X8_D24_UNORM_PACK32,
	VK_FORMAT_D32_SFLOAT,
	VK_FORMAT_S8_UINT,
	VK_FORMAT_D16_UNORM_S8_UINT,
	VK_FORMAT_D24_UNORM_S8_UINT,
	VK_FORMAT_D32_SFLOAT_S8_UINT,
	VK_FORMAT_BC1_RGB_UNORM_BLOCK,
	VK_FORMAT_BC1_RGB_SRGB_BLOCK,
	VK_FORMAT_BC1_RGBA_UNORM_BLOCK,
	VK_FORMAT_BC1_RGBA_SRGB_BLOCK,
	VK_FORMAT_BC2_UNORM_BLOCK,
	VK_FORMAT_BC2_SRGB_BLOCK,
	VK_FORMAT_BC3_UNORM_BLOCK,
	VK_FORMAT_BC3_SRGB_BLOCK,
	VK_FORMAT_BC4_UNORM_BLOCK,
	VK_FORMAT_BC4_SNORM_BLOCK,
	VK_FORMAT_BC5_UNORM_BLOCK,
	VK_FORMAT_BC5_SNORM_BLOCK,
	VK_FORMAT_BC6H_UFLOAT_BLOCK,
	VK_FORMAT_BC6H_SFLOAT_BLOCK,
	VK_FORMAT_BC7_UNORM_BLOCK,
	VK_FORMAT_BC7_SRGB_BLOCK,
	VK_FORMAT_ETC2_R8G8B8_UNORM_BLOCK,
	VK_FORMAT_ETC2_R8G8B8_SRGB_BLOCK,
	VK_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK,
	VK_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK,
	VK_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK,
	VK_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK,
	VK_FORMAT_EAC_R11_UNORM_BLOCK,
	VK_FORMAT_EAC_R11_SNORM_BLOCK,
	VK_FORMAT_EAC_R11G11_UNORM_BLOCK,
	VK_FORMAT_EAC_R11G11_SNORM_BLOCK,
	VK_FORMAT_ASTC_4x4_UNORM_BLOCK,
	VK_FORMAT_ASTC_4x4_SRGB_BLOCK,
	VK_FORMAT_ASTC_5x4_UNORM_BLOCK,
	VK_FORMAT_ASTC_5x4_SRGB_BLOCK,
	VK_FORMAT_ASTC_5x5_UNORM_BLOCK,
	VK_FORMAT_ASTC_5x5_SRGB_BLOCK,
	VK_FORMAT_ASTC_6x5_UNORM_BLOCK,
	VK_FORMAT_ASTC_6x5_SRGB_BLOCK,
	VK_FORMAT_ASTC_6x6_UNORM_BLOCK,
	VK_FORMAT_ASTC_6x6_SRGB_BLOCK,
	VK_FORMAT_ASTC_8x5_UNORM_BLOCK,
	VK_FORMAT_ASTC_8x5_SRGB_BLOCK,
	VK_FORMAT_ASTC_8x6_UNORM_BLOCK,
	VK_FORMAT_ASTC_8x6_SRGB_BLOCK,
	VK_FORMAT_ASTC_8x8_UNORM_BLOCK,
	VK_FORMAT_ASTC_8x8_SRGB_BLOCK,
	VK_FORMAT_ASTC_10x5_UNORM_BLOCK,
	VK_FORMAT_ASTC_10x5_SRGB_BLOCK,
	VK_FORMAT_ASTC_10x6_UNORM_BLOCK,
	VK_FORMAT_ASTC_10x6_SRGB_BLOCK,
	VK_FORMAT_ASTC_10x8_UNORM_BLOCK,
	VK_FORMAT_ASTC_10x8_SRGB_BLOCK,
	VK_FORMAT_ASTC_10x10_UNORM_BLOCK,
	VK_FORMAT_ASTC_10x10_SRGB_BLOCK,
	VK_FORMAT_ASTC_12x10_UNORM_BLOCK,
	VK_FORMAT_ASTC_12x10_SRGB_BLOCK,
	VK_FORMAT_ASTC_12x12_UNORM_BLOCK,
	VK_FORMAT_ASTC_12x12_SRGB_BLOCK,
	VK_FORMAT_G8B8G8R8_422_UNORM,
	VK_FORMAT_B8G8R8G8_422_UNORM,
	VK_FORMAT_G8_B8_R8_3PLANE_420_UNORM,
	VK_FORMAT_G8_B8R8_2PLANE_420_UNORM,
	VK_FORMAT_G8_B8_R8_3PLANE_422_UNORM,
	VK_FORMAT_G8_B8R8_2PLANE_422_UNORM,
	VK_FORMAT_G8_B8_R8_3PLANE_444_UNORM,
	VK_FORMAT_R10X6_UNORM_PACK16,
	VK_FORMAT_R10X6G10X6_UNORM_2PACK16,
	VK_FORMAT_R10X6G10X6B10X6A10X6_UNORM_4PACK16,
	VK_FORMAT_G10X6B10X6G10X6R10X6_422_UNORM_4PACK16,
	VK_FORMAT_B10X6G10X6R10X6G10X6_422_UNORM_4PACK16,
	VK_FORMAT_G10X6_B10X6_R10X6_3PLANE_420_UNORM_3PACK16,
	VK_FORMAT_G10X6_B10X6R10X6_2PLANE_420_UNORM_3PACK16,
	VK_FORMAT_G10X6_B10X6_R10X6_3PLANE_422_UNORM_3PACK16,
	VK_FORMAT_G10X6_B10X6R10X6_2PLANE_422_UNORM_3PACK16,
	VK_FORMAT_G10X6_B10X6_R10X6_3PLANE_444_UNORM_3PACK16,
	VK_FORMAT_R12X4_UNORM_PACK16,
	VK_FORMAT_R12X4G12X4_UNORM_2PACK16,
	VK_FORMAT_R12X4G12X4B12X4A12X4_UNORM_4PACK16,
	VK_FORMAT_G12X4B12X4G12X4R12X4_422_UNORM_4PACK16,
	VK_FORMAT_B12X4G12X4R12X4G12X4_422_UNORM_4PACK16,
	VK_FORMAT_G12X4_B12X4_R12X4_3PLANE_420_UNORM_3PACK16,
	VK_FORMAT_G12X4_B12X4R12X4_2PLANE_420_UNORM_3PACK16,
	VK_FORMAT_G12X4_B12X4_R12X4_3PLANE_422_UNORM_3PACK16,
	VK_FORMAT_G12X4_B12X4R12X4_2PLANE_422_UNORM_3PACK16,
	VK_FORMAT_G12X4_B12X4_R12X4_3PLANE_444_UNORM_3PACK16,
	VK_FORMAT_G16B16G16R16_422_UNORM,
	VK_FORMAT_B16G16R16G16_422_UNORM,
	VK_FORMAT_G16_B16_R16_3PLANE_420_UNORM,
	VK_FORMAT_G16_B16R16_2PLANE_420_UNORM,
	VK_FORMAT_G16_B16_R16_3PLANE_422_UNORM,
	VK_FORMAT_G16_B16R16_2PLANE_422_UNORM,
	VK_FORMAT_G16_B16_R16_3PLANE_444_UNORM,
};

const char *RenderingDeviceVulkan::named_formats[RenderingDevice::DATA_FORMAT_MAX] = {
	"R4G4_Unorm_Pack8",
	"R4G4B4A4_Unorm_Pack16",
	"B4G4R4A4_Unorm_Pack16",
	"R5G6B5_Unorm_Pack16",
	"B5G6R5_Unorm_Pack16",
	"R5G5B5A1_Unorm_Pack16",
	"B5G5R5A1_Unorm_Pack16",
	"A1R5G5B5_Unorm_Pack16",
	"R8_Unorm",
	"R8_Snorm",
	"R8_Uscaled",
	"R8_Sscaled",
	"R8_Uint",
	"R8_Sint",
	"R8_Srgb",
	"R8G8_Unorm",
	"R8G8_Snorm",
	"R8G8_Uscaled",
	"R8G8_Sscaled",
	"R8G8_Uint",
	"R8G8_Sint",
	"R8G8_Srgb",
	"R8G8B8_Unorm",
	"R8G8B8_Snorm",
	"R8G8B8_Uscaled",
	"R8G8B8_Sscaled",
	"R8G8B8_Uint",
	"R8G8B8_Sint",
	"R8G8B8_Srgb",
	"B8G8R8_Unorm",
	"B8G8R8_Snorm",
	"B8G8R8_Uscaled",
	"B8G8R8_Sscaled",
	"B8G8R8_Uint",
	"B8G8R8_Sint",
	"B8G8R8_Srgb",
	"R8G8B8A8_Unorm",
	"R8G8B8A8_Snorm",
	"R8G8B8A8_Uscaled",
	"R8G8B8A8_Sscaled",
	"R8G8B8A8_Uint",
	"R8G8B8A8_Sint",
	"R8G8B8A8_Srgb",
	"B8G8R8A8_Unorm",
	"B8G8R8A8_Snorm",
	"B8G8R8A8_Uscaled",
	"B8G8R8A8_Sscaled",
	"B8G8R8A8_Uint",
	"B8G8R8A8_Sint",
	"B8G8R8A8_Srgb",
	"A8B8G8R8_Unorm_Pack32",
	"A8B8G8R8_Snorm_Pack32",
	"A8B8G8R8_Uscaled_Pack32",
	"A8B8G8R8_Sscaled_Pack32",
	"A8B8G8R8_Uint_Pack32",
	"A8B8G8R8_Sint_Pack32",
	"A8B8G8R8_Srgb_Pack32",
	"A2R10G10B10_Unorm_Pack32",
	"A2R10G10B10_Snorm_Pack32",
	"A2R10G10B10_Uscaled_Pack32",
	"A2R10G10B10_Sscaled_Pack32",
	"A2R10G10B10_Uint_Pack32",
	"A2R10G10B10_Sint_Pack32",
	"A2B10G10R10_Unorm_Pack32",
	"A2B10G10R10_Snorm_Pack32",
	"A2B10G10R10_Uscaled_Pack32",
	"A2B10G10R10_Sscaled_Pack32",
	"A2B10G10R10_Uint_Pack32",
	"A2B10G10R10_Sint_Pack32",
	"R16_Unorm",
	"R16_Snorm",
	"R16_Uscaled",
	"R16_Sscaled",
	"R16_Uint",
	"R16_Sint",
	"R16_Sfloat",
	"R16G16_Unorm",
	"R16G16_Snorm",
	"R16G16_Uscaled",
	"R16G16_Sscaled",
	"R16G16_Uint",
	"R16G16_Sint",
	"R16G16_Sfloat",
	"R16G16B16_Unorm",
	"R16G16B16_Snorm",
	"R16G16B16_Uscaled",
	"R16G16B16_Sscaled",
	"R16G16B16_Uint",
	"R16G16B16_Sint",
	"R16G16B16_Sfloat",
	"R16G16B16A16_Unorm",
	"R16G16B16A16_Snorm",
	"R16G16B16A16_Uscaled",
	"R16G16B16A16_Sscaled",
	"R16G16B16A16_Uint",
	"R16G16B16A16_Sint",
	"R16G16B16A16_Sfloat",
	"R32_Uint",
	"R32_Sint",
	"R32_Sfloat",
	"R32G32_Uint",
	"R32G32_Sint",
	"R32G32_Sfloat",
	"R32G32B32_Uint",
	"R32G32B32_Sint",
	"R32G32B32_Sfloat",
	"R32G32B32A32_Uint",
	"R32G32B32A32_Sint",
	"R32G32B32A32_Sfloat",
	"R64_Uint",
	"R64_Sint",
	"R64_Sfloat",
	"R64G64_Uint",
	"R64G64_Sint",
	"R64G64_Sfloat",
	"R64G64B64_Uint",
	"R64G64B64_Sint",
	"R64G64B64_Sfloat",
	"R64G64B64A64_Uint",
	"R64G64B64A64_Sint",
	"R64G64B64A64_Sfloat",
	"B10G11R11_Ufloat_Pack32",
	"E5B9G9R9_Ufloat_Pack32",
	"D16_Unorm",
	"X8_D24_Unorm_Pack32",
	"D32_Sfloat",
	"S8_Uint",
	"D16_Unorm_S8_Uint",
	"D24_Unorm_S8_Uint",
	"D32_Sfloat_S8_Uint",
	"Bc1_Rgb_Unorm_Block",
	"Bc1_Rgb_Srgb_Block",
	"Bc1_Rgba_Unorm_Block",
	"Bc1_Rgba_Srgb_Block",
	"Bc2_Unorm_Block",
	"Bc2_Srgb_Block",
	"Bc3_Unorm_Block",
	"Bc3_Srgb_Block",
	"Bc4_Unorm_Block",
	"Bc4_Snorm_Block",
	"Bc5_Unorm_Block",
	"Bc5_Snorm_Block",
	"Bc6H_Ufloat_Block",
	"Bc6H_Sfloat_Block",
	"Bc7_Unorm_Block",
	"Bc7_Srgb_Block",
	"Etc2_R8G8B8_Unorm_Block",
	"Etc2_R8G8B8_Srgb_Block",
	"Etc2_R8G8B8A1_Unorm_Block",
	"Etc2_R8G8B8A1_Srgb_Block",
	"Etc2_R8G8B8A8_Unorm_Block",
	"Etc2_R8G8B8A8_Srgb_Block",
	"Eac_R11_Unorm_Block",
	"Eac_R11_Snorm_Block",
	"Eac_R11G11_Unorm_Block",
	"Eac_R11G11_Snorm_Block",
	"Astc_4X4_Unorm_Block",
	"Astc_4X4_Srgb_Block",
	"Astc_5X4_Unorm_Block",
	"Astc_5X4_Srgb_Block",
	"Astc_5X5_Unorm_Block",
	"Astc_5X5_Srgb_Block",
	"Astc_6X5_Unorm_Block",
	"Astc_6X5_Srgb_Block",
	"Astc_6X6_Unorm_Block",
	"Astc_6X6_Srgb_Block",
	"Astc_8X5_Unorm_Block",
	"Astc_8X5_Srgb_Block",
	"Astc_8X6_Unorm_Block",
	"Astc_8X6_Srgb_Block",
	"Astc_8X8_Unorm_Block",
	"Astc_8X8_Srgb_Block",
	"Astc_10X5_Unorm_Block",
	"Astc_10X5_Srgb_Block",
	"Astc_10X6_Unorm_Block",
	"Astc_10X6_Srgb_Block",
	"Astc_10X8_Unorm_Block",
	"Astc_10X8_Srgb_Block",
	"Astc_10X10_Unorm_Block",
	"Astc_10X10_Srgb_Block",
	"Astc_12X10_Unorm_Block",
	"Astc_12X10_Srgb_Block",
	"Astc_12X12_Unorm_Block",
	"Astc_12X12_Srgb_Block",
	"G8B8G8R8_422_Unorm",
	"B8G8R8G8_422_Unorm",
	"G8_B8_R8_3Plane_420_Unorm",
	"G8_B8R8_2Plane_420_Unorm",
	"G8_B8_R8_3Plane_422_Unorm",
	"G8_B8R8_2Plane_422_Unorm",
	"G8_B8_R8_3Plane_444_Unorm",
	"R10X6_Unorm_Pack16",
	"R10X6G10X6_Unorm_2Pack16",
	"R10X6G10X6B10X6A10X6_Unorm_4Pack16",
	"G10X6B10X6G10X6R10X6_422_Unorm_4Pack16",
	"B10X6G10X6R10X6G10X6_422_Unorm_4Pack16",
	"G10X6_B10X6_R10X6_3Plane_420_Unorm_3Pack16",
	"G10X6_B10X6R10X6_2Plane_420_Unorm_3Pack16",
	"G10X6_B10X6_R10X6_3Plane_422_Unorm_3Pack16",
	"G10X6_B10X6R10X6_2Plane_422_Unorm_3Pack16",
	"G10X6_B10X6_R10X6_3Plane_444_Unorm_3Pack16",
	"R12X4_Unorm_Pack16",
	"R12X4G12X4_Unorm_2Pack16",
	"R12X4G12X4B12X4A12X4_Unorm_4Pack16",
	"G12X4B12X4G12X4R12X4_422_Unorm_4Pack16",
	"B12X4G12X4R12X4G12X4_422_Unorm_4Pack16",
	"G12X4_B12X4_R12X4_3Plane_420_Unorm_3Pack16",
	"G12X4_B12X4R12X4_2Plane_420_Unorm_3Pack16",
	"G12X4_B12X4_R12X4_3Plane_422_Unorm_3Pack16",
	"G12X4_B12X4R12X4_2Plane_422_Unorm_3Pack16",
	"G12X4_B12X4_R12X4_3Plane_444_Unorm_3Pack16",
	"G16B16G16R16_422_Unorm",
	"B16G16R16G16_422_Unorm",
	"G16_B16_R16_3Plane_420_Unorm",
	"G16_B16R16_2Plane_420_Unorm",
	"G16_B16_R16_3Plane_422_Unorm",
	"G16_B16R16_2Plane_422_Unorm",
	"G16_B16_R16_3Plane_444_Unorm",
};

int RenderingDeviceVulkan::get_format_vertex_size(DataFormat p_format) {
	switch (p_format) {
		case DATA_FORMAT_R8_UNORM:
		case DATA_FORMAT_R8_SNORM:
		case DATA_FORMAT_R8_UINT:
		case DATA_FORMAT_R8_SINT:
		case DATA_FORMAT_R8G8_UNORM:
		case DATA_FORMAT_R8G8_SNORM:
		case DATA_FORMAT_R8G8_UINT:
		case DATA_FORMAT_R8G8_SINT:
		case DATA_FORMAT_R8G8B8_UNORM:
		case DATA_FORMAT_R8G8B8_SNORM:
		case DATA_FORMAT_R8G8B8_UINT:
		case DATA_FORMAT_R8G8B8_SINT:
		case DATA_FORMAT_B8G8R8_UNORM:
		case DATA_FORMAT_B8G8R8_SNORM:
		case DATA_FORMAT_B8G8R8_UINT:
		case DATA_FORMAT_B8G8R8_SINT:
		case DATA_FORMAT_R8G8B8A8_UNORM:
		case DATA_FORMAT_R8G8B8A8_SNORM:
		case DATA_FORMAT_R8G8B8A8_UINT:
		case DATA_FORMAT_R8G8B8A8_SINT:
		case DATA_FORMAT_B8G8R8A8_UNORM:
		case DATA_FORMAT_B8G8R8A8_SNORM:
		case DATA_FORMAT_B8G8R8A8_UINT:
		case DATA_FORMAT_B8G8R8A8_SINT:
		case DATA_FORMAT_A2B10G10R10_UNORM_PACK32:
			return 4;
		case DATA_FORMAT_R16_UNORM:
		case DATA_FORMAT_R16_SNORM:
		case DATA_FORMAT_R16_UINT:
		case DATA_FORMAT_R16_SINT:
		case DATA_FORMAT_R16_SFLOAT:
			return 4;
		case DATA_FORMAT_R16G16_UNORM:
		case DATA_FORMAT_R16G16_SNORM:
		case DATA_FORMAT_R16G16_UINT:
		case DATA_FORMAT_R16G16_SINT:
		case DATA_FORMAT_R16G16_SFLOAT:
			return 4;
		case DATA_FORMAT_R16G16B16_UNORM:
		case DATA_FORMAT_R16G16B16_SNORM:
		case DATA_FORMAT_R16G16B16_UINT:
		case DATA_FORMAT_R16G16B16_SINT:
		case DATA_FORMAT_R16G16B16_SFLOAT:
			return 8;
		case DATA_FORMAT_R16G16B16A16_UNORM:
		case DATA_FORMAT_R16G16B16A16_SNORM:
		case DATA_FORMAT_R16G16B16A16_UINT:
		case DATA_FORMAT_R16G16B16A16_SINT:
		case DATA_FORMAT_R16G16B16A16_SFLOAT:
			return 8;
		case DATA_FORMAT_R32_UINT:
		case DATA_FORMAT_R32_SINT:
		case DATA_FORMAT_R32_SFLOAT:
			return 4;
		case DATA_FORMAT_R32G32_UINT:
		case DATA_FORMAT_R32G32_SINT:
		case DATA_FORMAT_R32G32_SFLOAT:
			return 8;
		case DATA_FORMAT_R32G32B32_UINT:
		case DATA_FORMAT_R32G32B32_SINT:
		case DATA_FORMAT_R32G32B32_SFLOAT:
			return 12;
		case DATA_FORMAT_R32G32B32A32_UINT:
		case DATA_FORMAT_R32G32B32A32_SINT:
		case DATA_FORMAT_R32G32B32A32_SFLOAT:
			return 16;
		case DATA_FORMAT_R64_UINT:
		case DATA_FORMAT_R64_SINT:
		case DATA_FORMAT_R64_SFLOAT:
			return 8;
		case DATA_FORMAT_R64G64_UINT:
		case DATA_FORMAT_R64G64_SINT:
		case DATA_FORMAT_R64G64_SFLOAT:
			return 16;
		case DATA_FORMAT_R64G64B64_UINT:
		case DATA_FORMAT_R64G64B64_SINT:
		case DATA_FORMAT_R64G64B64_SFLOAT:
			return 24;
		case DATA_FORMAT_R64G64B64A64_UINT:
		case DATA_FORMAT_R64G64B64A64_SINT:
		case DATA_FORMAT_R64G64B64A64_SFLOAT:
			return 32;
		default:
			return 0;
	}
}

uint32_t RenderingDeviceVulkan::get_image_format_pixel_size(DataFormat p_format) {
	switch (p_format) {
		case DATA_FORMAT_R4G4_UNORM_PACK8:
			return 1;
		case DATA_FORMAT_R4G4B4A4_UNORM_PACK16:
		case DATA_FORMAT_B4G4R4A4_UNORM_PACK16:
		case DATA_FORMAT_R5G6B5_UNORM_PACK16:
		case DATA_FORMAT_B5G6R5_UNORM_PACK16:
		case DATA_FORMAT_R5G5B5A1_UNORM_PACK16:
		case DATA_FORMAT_B5G5R5A1_UNORM_PACK16:
		case DATA_FORMAT_A1R5G5B5_UNORM_PACK16:
			return 2;
		case DATA_FORMAT_R8_UNORM:
		case DATA_FORMAT_R8_SNORM:
		case DATA_FORMAT_R8_USCALED:
		case DATA_FORMAT_R8_SSCALED:
		case DATA_FORMAT_R8_UINT:
		case DATA_FORMAT_R8_SINT:
		case DATA_FORMAT_R8_SRGB:
			return 1;
		case DATA_FORMAT_R8G8_UNORM:
		case DATA_FORMAT_R8G8_SNORM:
		case DATA_FORMAT_R8G8_USCALED:
		case DATA_FORMAT_R8G8_SSCALED:
		case DATA_FORMAT_R8G8_UINT:
		case DATA_FORMAT_R8G8_SINT:
		case DATA_FORMAT_R8G8_SRGB:
			return 2;
		case DATA_FORMAT_R8G8B8_UNORM:
		case DATA_FORMAT_R8G8B8_SNORM:
		case DATA_FORMAT_R8G8B8_USCALED:
		case DATA_FORMAT_R8G8B8_SSCALED:
		case DATA_FORMAT_R8G8B8_UINT:
		case DATA_FORMAT_R8G8B8_SINT:
		case DATA_FORMAT_R8G8B8_SRGB:
		case DATA_FORMAT_B8G8R8_UNORM:
		case DATA_FORMAT_B8G8R8_SNORM:
		case DATA_FORMAT_B8G8R8_USCALED:
		case DATA_FORMAT_B8G8R8_SSCALED:
		case DATA_FORMAT_B8G8R8_UINT:
		case DATA_FORMAT_B8G8R8_SINT:
		case DATA_FORMAT_B8G8R8_SRGB:
			return 3;
		case DATA_FORMAT_R8G8B8A8_UNORM:
		case DATA_FORMAT_R8G8B8A8_SNORM:
		case DATA_FORMAT_R8G8B8A8_USCALED:
		case DATA_FORMAT_R8G8B8A8_SSCALED:
		case DATA_FORMAT_R8G8B8A8_UINT:
		case DATA_FORMAT_R8G8B8A8_SINT:
		case DATA_FORMAT_R8G8B8A8_SRGB:
		case DATA_FORMAT_B8G8R8A8_UNORM:
		case DATA_FORMAT_B8G8R8A8_SNORM:
		case DATA_FORMAT_B8G8R8A8_USCALED:
		case DATA_FORMAT_B8G8R8A8_SSCALED:
		case DATA_FORMAT_B8G8R8A8_UINT:
		case DATA_FORMAT_B8G8R8A8_SINT:
		case DATA_FORMAT_B8G8R8A8_SRGB:
			return 4;
		case DATA_FORMAT_A8B8G8R8_UNORM_PACK32:
		case DATA_FORMAT_A8B8G8R8_SNORM_PACK32:
		case DATA_FORMAT_A8B8G8R8_USCALED_PACK32:
		case DATA_FORMAT_A8B8G8R8_SSCALED_PACK32:
		case DATA_FORMAT_A8B8G8R8_UINT_PACK32:
		case DATA_FORMAT_A8B8G8R8_SINT_PACK32:
		case DATA_FORMAT_A8B8G8R8_SRGB_PACK32:
		case DATA_FORMAT_A2R10G10B10_UNORM_PACK32:
		case DATA_FORMAT_A2R10G10B10_SNORM_PACK32:
		case DATA_FORMAT_A2R10G10B10_USCALED_PACK32:
		case DATA_FORMAT_A2R10G10B10_SSCALED_PACK32:
		case DATA_FORMAT_A2R10G10B10_UINT_PACK32:
		case DATA_FORMAT_A2R10G10B10_SINT_PACK32:
		case DATA_FORMAT_A2B10G10R10_UNORM_PACK32:
		case DATA_FORMAT_A2B10G10R10_SNORM_PACK32:
		case DATA_FORMAT_A2B10G10R10_USCALED_PACK32:
		case DATA_FORMAT_A2B10G10R10_SSCALED_PACK32:
		case DATA_FORMAT_A2B10G10R10_UINT_PACK32:
		case DATA_FORMAT_A2B10G10R10_SINT_PACK32:
			return 4;
		case DATA_FORMAT_R16_UNORM:
		case DATA_FORMAT_R16_SNORM:
		case DATA_FORMAT_R16_USCALED:
		case DATA_FORMAT_R16_SSCALED:
		case DATA_FORMAT_R16_UINT:
		case DATA_FORMAT_R16_SINT:
		case DATA_FORMAT_R16_SFLOAT:
			return 2;
		case DATA_FORMAT_R16G16_UNORM:
		case DATA_FORMAT_R16G16_SNORM:
		case DATA_FORMAT_R16G16_USCALED:
		case DATA_FORMAT_R16G16_SSCALED:
		case DATA_FORMAT_R16G16_UINT:
		case DATA_FORMAT_R16G16_SINT:
		case DATA_FORMAT_R16G16_SFLOAT:
			return 4;
		case DATA_FORMAT_R16G16B16_UNORM:
		case DATA_FORMAT_R16G16B16_SNORM:
		case DATA_FORMAT_R16G16B16_USCALED:
		case DATA_FORMAT_R16G16B16_SSCALED:
		case DATA_FORMAT_R16G16B16_UINT:
		case DATA_FORMAT_R16G16B16_SINT:
		case DATA_FORMAT_R16G16B16_SFLOAT:
			return 6;
		case DATA_FORMAT_R16G16B16A16_UNORM:
		case DATA_FORMAT_R16G16B16A16_SNORM:
		case DATA_FORMAT_R16G16B16A16_USCALED:
		case DATA_FORMAT_R16G16B16A16_SSCALED:
		case DATA_FORMAT_R16G16B16A16_UINT:
		case DATA_FORMAT_R16G16B16A16_SINT:
		case DATA_FORMAT_R16G16B16A16_SFLOAT:
			return 8;
		case DATA_FORMAT_R32_UINT:
		case DATA_FORMAT_R32_SINT:
		case DATA_FORMAT_R32_SFLOAT:
			return 4;
		case DATA_FORMAT_R32G32_UINT:
		case DATA_FORMAT_R32G32_SINT:
		case DATA_FORMAT_R32G32_SFLOAT:
			return 8;
		case DATA_FORMAT_R32G32B32_UINT:
		case DATA_FORMAT_R32G32B32_SINT:
		case DATA_FORMAT_R32G32B32_SFLOAT:
			return 12;
		case DATA_FORMAT_R32G32B32A32_UINT:
		case DATA_FORMAT_R32G32B32A32_SINT:
		case DATA_FORMAT_R32G32B32A32_SFLOAT:
			return 16;
		case DATA_FORMAT_R64_UINT:
		case DATA_FORMAT_R64_SINT:
		case DATA_FORMAT_R64_SFLOAT:
			return 8;
		case DATA_FORMAT_R64G64_UINT:
		case DATA_FORMAT_R64G64_SINT:
		case DATA_FORMAT_R64G64_SFLOAT:
			return 16;
		case DATA_FORMAT_R64G64B64_UINT:
		case DATA_FORMAT_R64G64B64_SINT:
		case DATA_FORMAT_R64G64B64_SFLOAT:
			return 24;
		case DATA_FORMAT_R64G64B64A64_UINT:
		case DATA_FORMAT_R64G64B64A64_SINT:
		case DATA_FORMAT_R64G64B64A64_SFLOAT:
			return 32;
		case DATA_FORMAT_B10G11R11_UFLOAT_PACK32:
		case DATA_FORMAT_E5B9G9R9_UFLOAT_PACK32:
			return 4;
		case DATA_FORMAT_D16_UNORM:
			return 2;
		case DATA_FORMAT_X8_D24_UNORM_PACK32:
			return 4;
		case DATA_FORMAT_D32_SFLOAT:
			return 4;
		case DATA_FORMAT_S8_UINT:
			return 1;
		case DATA_FORMAT_D16_UNORM_S8_UINT:
			return 4;
		case DATA_FORMAT_D24_UNORM_S8_UINT:
			return 4;
		case DATA_FORMAT_D32_SFLOAT_S8_UINT:
			return 5; // ?
		case DATA_FORMAT_BC1_RGB_UNORM_BLOCK:
		case DATA_FORMAT_BC1_RGB_SRGB_BLOCK:
		case DATA_FORMAT_BC1_RGBA_UNORM_BLOCK:
		case DATA_FORMAT_BC1_RGBA_SRGB_BLOCK:
		case DATA_FORMAT_BC2_UNORM_BLOCK:
		case DATA_FORMAT_BC2_SRGB_BLOCK:
		case DATA_FORMAT_BC3_UNORM_BLOCK:
		case DATA_FORMAT_BC3_SRGB_BLOCK:
		case DATA_FORMAT_BC4_UNORM_BLOCK:
		case DATA_FORMAT_BC4_SNORM_BLOCK:
		case DATA_FORMAT_BC5_UNORM_BLOCK:
		case DATA_FORMAT_BC5_SNORM_BLOCK:
		case DATA_FORMAT_BC6H_UFLOAT_BLOCK:
		case DATA_FORMAT_BC6H_SFLOAT_BLOCK:
		case DATA_FORMAT_BC7_UNORM_BLOCK:
		case DATA_FORMAT_BC7_SRGB_BLOCK:
			return 1;
		case DATA_FORMAT_ETC2_R8G8B8_UNORM_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8_SRGB_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK:
			return 1;
		case DATA_FORMAT_EAC_R11_UNORM_BLOCK:
		case DATA_FORMAT_EAC_R11_SNORM_BLOCK:
		case DATA_FORMAT_EAC_R11G11_UNORM_BLOCK:
		case DATA_FORMAT_EAC_R11G11_SNORM_BLOCK:
			return 1;
		case DATA_FORMAT_ASTC_4x4_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_4x4_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_5x4_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_5x4_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_5x5_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_5x5_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_6x5_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_6x5_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_6x6_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_6x6_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_8x5_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_8x5_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_8x6_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_8x6_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_8x8_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_8x8_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_10x5_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_10x5_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_10x6_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_10x6_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_10x8_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_10x8_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_10x10_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_10x10_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_12x10_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_12x10_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_12x12_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_12x12_SRGB_BLOCK:
			return 1;
		case DATA_FORMAT_G8B8G8R8_422_UNORM:
		case DATA_FORMAT_B8G8R8G8_422_UNORM:
			return 4;
		case DATA_FORMAT_G8_B8_R8_3PLANE_420_UNORM:
		case DATA_FORMAT_G8_B8R8_2PLANE_420_UNORM:
		case DATA_FORMAT_G8_B8_R8_3PLANE_422_UNORM:
		case DATA_FORMAT_G8_B8R8_2PLANE_422_UNORM:
		case DATA_FORMAT_G8_B8_R8_3PLANE_444_UNORM:
			return 4;
		case DATA_FORMAT_R10X6_UNORM_PACK16:
		case DATA_FORMAT_R10X6G10X6_UNORM_2PACK16:
		case DATA_FORMAT_R10X6G10X6B10X6A10X6_UNORM_4PACK16:
		case DATA_FORMAT_G10X6B10X6G10X6R10X6_422_UNORM_4PACK16:
		case DATA_FORMAT_B10X6G10X6R10X6G10X6_422_UNORM_4PACK16:
		case DATA_FORMAT_G10X6_B10X6_R10X6_3PLANE_420_UNORM_3PACK16:
		case DATA_FORMAT_G10X6_B10X6R10X6_2PLANE_420_UNORM_3PACK16:
		case DATA_FORMAT_G10X6_B10X6_R10X6_3PLANE_422_UNORM_3PACK16:
		case DATA_FORMAT_G10X6_B10X6R10X6_2PLANE_422_UNORM_3PACK16:
		case DATA_FORMAT_G10X6_B10X6_R10X6_3PLANE_444_UNORM_3PACK16:
		case DATA_FORMAT_R12X4_UNORM_PACK16:
		case DATA_FORMAT_R12X4G12X4_UNORM_2PACK16:
		case DATA_FORMAT_R12X4G12X4B12X4A12X4_UNORM_4PACK16:
		case DATA_FORMAT_G12X4B12X4G12X4R12X4_422_UNORM_4PACK16:
		case DATA_FORMAT_B12X4G12X4R12X4G12X4_422_UNORM_4PACK16:
		case DATA_FORMAT_G12X4_B12X4_R12X4_3PLANE_420_UNORM_3PACK16:
		case DATA_FORMAT_G12X4_B12X4R12X4_2PLANE_420_UNORM_3PACK16:
		case DATA_FORMAT_G12X4_B12X4_R12X4_3PLANE_422_UNORM_3PACK16:
		case DATA_FORMAT_G12X4_B12X4R12X4_2PLANE_422_UNORM_3PACK16:
		case DATA_FORMAT_G12X4_B12X4_R12X4_3PLANE_444_UNORM_3PACK16:
			return 2;
		case DATA_FORMAT_G16B16G16R16_422_UNORM:
		case DATA_FORMAT_B16G16R16G16_422_UNORM:
		case DATA_FORMAT_G16_B16_R16_3PLANE_420_UNORM:
		case DATA_FORMAT_G16_B16R16_2PLANE_420_UNORM:
		case DATA_FORMAT_G16_B16_R16_3PLANE_422_UNORM:
		case DATA_FORMAT_G16_B16R16_2PLANE_422_UNORM:
		case DATA_FORMAT_G16_B16_R16_3PLANE_444_UNORM:
			return 8;
		default: {
			ERR_PRINT("Format not handled, bug");
		}
	}

	return 1;
}

// https://www.khronos.org/registry/DataFormat/specs/1.1/dataformat.1.1.pdf

void RenderingDeviceVulkan::get_compressed_image_format_block_dimensions(DataFormat p_format, uint32_t &r_w, uint32_t &r_h) {
	switch (p_format) {
		case DATA_FORMAT_BC1_RGB_UNORM_BLOCK:
		case DATA_FORMAT_BC1_RGB_SRGB_BLOCK:
		case DATA_FORMAT_BC1_RGBA_UNORM_BLOCK:
		case DATA_FORMAT_BC1_RGBA_SRGB_BLOCK:
		case DATA_FORMAT_BC2_UNORM_BLOCK:
		case DATA_FORMAT_BC2_SRGB_BLOCK:
		case DATA_FORMAT_BC3_UNORM_BLOCK:
		case DATA_FORMAT_BC3_SRGB_BLOCK:
		case DATA_FORMAT_BC4_UNORM_BLOCK:
		case DATA_FORMAT_BC4_SNORM_BLOCK:
		case DATA_FORMAT_BC5_UNORM_BLOCK:
		case DATA_FORMAT_BC5_SNORM_BLOCK:
		case DATA_FORMAT_BC6H_UFLOAT_BLOCK:
		case DATA_FORMAT_BC6H_SFLOAT_BLOCK:
		case DATA_FORMAT_BC7_UNORM_BLOCK:
		case DATA_FORMAT_BC7_SRGB_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8_UNORM_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8_SRGB_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK:
		case DATA_FORMAT_EAC_R11_UNORM_BLOCK:
		case DATA_FORMAT_EAC_R11_SNORM_BLOCK:
		case DATA_FORMAT_EAC_R11G11_UNORM_BLOCK:
		case DATA_FORMAT_EAC_R11G11_SNORM_BLOCK:
		case DATA_FORMAT_ASTC_4x4_UNORM_BLOCK: // Again, not sure about astc.
		case DATA_FORMAT_ASTC_4x4_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_5x4_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_5x4_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_5x5_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_5x5_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_6x5_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_6x5_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_6x6_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_6x6_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_8x5_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_8x5_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_8x6_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_8x6_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_8x8_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_8x8_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_10x5_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_10x5_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_10x6_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_10x6_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_10x8_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_10x8_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_10x10_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_10x10_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_12x10_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_12x10_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_12x12_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_12x12_SRGB_BLOCK:
			r_w = 4;
			r_h = 4;
			return;
		default: {
			r_w = 1;
			r_h = 1;
		}
	}
}

uint32_t RenderingDeviceVulkan::get_compressed_image_format_block_byte_size(DataFormat p_format) {
	switch (p_format) {
		case DATA_FORMAT_BC1_RGB_UNORM_BLOCK:
		case DATA_FORMAT_BC1_RGB_SRGB_BLOCK:
		case DATA_FORMAT_BC1_RGBA_UNORM_BLOCK:
		case DATA_FORMAT_BC1_RGBA_SRGB_BLOCK:
			return 8;
		case DATA_FORMAT_BC2_UNORM_BLOCK:
		case DATA_FORMAT_BC2_SRGB_BLOCK:
			return 16;
		case DATA_FORMAT_BC3_UNORM_BLOCK:
		case DATA_FORMAT_BC3_SRGB_BLOCK:
			return 16;
		case DATA_FORMAT_BC4_UNORM_BLOCK:
		case DATA_FORMAT_BC4_SNORM_BLOCK:
			return 8;
		case DATA_FORMAT_BC5_UNORM_BLOCK:
		case DATA_FORMAT_BC5_SNORM_BLOCK:
			return 16;
		case DATA_FORMAT_BC6H_UFLOAT_BLOCK:
		case DATA_FORMAT_BC6H_SFLOAT_BLOCK:
			return 16;
		case DATA_FORMAT_BC7_UNORM_BLOCK:
		case DATA_FORMAT_BC7_SRGB_BLOCK:
			return 16;
		case DATA_FORMAT_ETC2_R8G8B8_UNORM_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8_SRGB_BLOCK:
			return 8;
		case DATA_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK:
			return 8;
		case DATA_FORMAT_ETC2_R8G8B8A8_UNORM_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8A8_SRGB_BLOCK:
			return 16;
		case DATA_FORMAT_EAC_R11_UNORM_BLOCK:
		case DATA_FORMAT_EAC_R11_SNORM_BLOCK:
			return 8;
		case DATA_FORMAT_EAC_R11G11_UNORM_BLOCK:
		case DATA_FORMAT_EAC_R11G11_SNORM_BLOCK:
			return 16;
		case DATA_FORMAT_ASTC_4x4_UNORM_BLOCK: // Again, not sure about astc.
		case DATA_FORMAT_ASTC_4x4_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_5x4_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_5x4_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_5x5_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_5x5_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_6x5_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_6x5_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_6x6_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_6x6_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_8x5_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_8x5_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_8x6_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_8x6_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_8x8_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_8x8_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_10x5_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_10x5_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_10x6_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_10x6_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_10x8_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_10x8_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_10x10_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_10x10_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_12x10_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_12x10_SRGB_BLOCK:
		case DATA_FORMAT_ASTC_12x12_UNORM_BLOCK:
		case DATA_FORMAT_ASTC_12x12_SRGB_BLOCK:
			return 8; // Wrong.
		default: {
		}
	}
	return 1;
}

uint32_t RenderingDeviceVulkan::get_compressed_image_format_pixel_rshift(DataFormat p_format) {
	switch (p_format) {
		case DATA_FORMAT_BC1_RGB_UNORM_BLOCK: // These formats are half byte size, so rshift is 1.
		case DATA_FORMAT_BC1_RGB_SRGB_BLOCK:
		case DATA_FORMAT_BC1_RGBA_UNORM_BLOCK:
		case DATA_FORMAT_BC1_RGBA_SRGB_BLOCK:
		case DATA_FORMAT_BC4_UNORM_BLOCK:
		case DATA_FORMAT_BC4_SNORM_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8_UNORM_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8_SRGB_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8A1_UNORM_BLOCK:
		case DATA_FORMAT_ETC2_R8G8B8A1_SRGB_BLOCK:
		case DATA_FORMAT_EAC_R11_UNORM_BLOCK:
		case DATA_FORMAT_EAC_R11_SNORM_BLOCK:
			return 1;
		default: {
		}
	}

	return 0;
}

bool RenderingDeviceVulkan::format_has_stencil(DataFormat p_format) {
	switch (p_format) {
		case DATA_FORMAT_S8_UINT:
		case DATA_FORMAT_D16_UNORM_S8_UINT:
		case DATA_FORMAT_D24_UNORM_S8_UINT:
		case DATA_FORMAT_D32_SFLOAT_S8_UINT: {
			return true;
		}
		default: {
		}
	}
	return false;
}

uint32_t RenderingDeviceVulkan::get_image_format_required_size(DataFormat p_format, uint32_t p_width, uint32_t p_height, uint32_t p_depth, uint32_t p_mipmaps, uint32_t *r_blockw, uint32_t *r_blockh, uint32_t *r_depth) {
	ERR_FAIL_COND_V(p_mipmaps == 0, 0);
	uint32_t w = p_width;
	uint32_t h = p_height;
	uint32_t d = p_depth;

	uint32_t size = 0;

	uint32_t pixel_size = get_image_format_pixel_size(p_format);
	uint32_t pixel_rshift = get_compressed_image_format_pixel_rshift(p_format);
	uint32_t blockw, blockh;
	get_compressed_image_format_block_dimensions(p_format, blockw, blockh);

	for (uint32_t i = 0; i < p_mipmaps; i++) {
		uint32_t bw = w % blockw != 0 ? w + (blockw - w % blockw) : w;
		uint32_t bh = h % blockh != 0 ? h + (blockh - h % blockh) : h;

		uint32_t s = bw * bh;

		s *= pixel_size;
		s >>= pixel_rshift;
		size += s * d;
		if (r_blockw) {
			*r_blockw = bw;
		}
		if (r_blockh) {
			*r_blockh = bh;
		}
		if (r_depth) {
			*r_depth = d;
		}
		w = MAX(blockw, w >> 1);
		h = MAX(blockh, h >> 1);
		d = MAX(1u, d >> 1);
	}

	return size;
}

uint32_t RenderingDeviceVulkan::get_image_required_mipmaps(uint32_t p_width, uint32_t p_height, uint32_t p_depth) {
	// Formats and block size don't really matter here since they can all go down to 1px (even if block is larger).
	uint32_t w = p_width;
	uint32_t h = p_height;
	uint32_t d = p_depth;

	uint32_t mipmaps = 1;

	while (true) {
		if (w == 1 && h == 1 && d == 1) {
			break;
		}

		w = MAX(1u, w >> 1);
		h = MAX(1u, h >> 1);
		d = MAX(1u, d >> 1);

		mipmaps++;
	}

	return mipmaps;
}

///////////////////////

const VkCompareOp RenderingDeviceVulkan::compare_operators[RenderingDevice::COMPARE_OP_MAX] = {
	VK_COMPARE_OP_NEVER,
	VK_COMPARE_OP_LESS,
	VK_COMPARE_OP_EQUAL,
	VK_COMPARE_OP_LESS_OR_EQUAL,
	VK_COMPARE_OP_GREATER,
	VK_COMPARE_OP_NOT_EQUAL,
	VK_COMPARE_OP_GREATER_OR_EQUAL,
	VK_COMPARE_OP_ALWAYS
};

const VkStencilOp RenderingDeviceVulkan::stencil_operations[RenderingDevice::STENCIL_OP_MAX] = {
	VK_STENCIL_OP_KEEP,
	VK_STENCIL_OP_ZERO,
	VK_STENCIL_OP_REPLACE,
	VK_STENCIL_OP_INCREMENT_AND_CLAMP,
	VK_STENCIL_OP_DECREMENT_AND_CLAMP,
	VK_STENCIL_OP_INVERT,
	VK_STENCIL_OP_INCREMENT_AND_WRAP,
	VK_STENCIL_OP_DECREMENT_AND_WRAP
};

const VkSampleCountFlagBits RenderingDeviceVulkan::rasterization_sample_count[RenderingDevice::TEXTURE_SAMPLES_MAX] = {
	VK_SAMPLE_COUNT_1_BIT,
	VK_SAMPLE_COUNT_2_BIT,
	VK_SAMPLE_COUNT_4_BIT,
	VK_SAMPLE_COUNT_8_BIT,
	VK_SAMPLE_COUNT_16_BIT,
	VK_SAMPLE_COUNT_32_BIT,
	VK_SAMPLE_COUNT_64_BIT,
};

const VkLogicOp RenderingDeviceVulkan::logic_operations[RenderingDevice::LOGIC_OP_MAX] = {
	VK_LOGIC_OP_CLEAR,
	VK_LOGIC_OP_AND,
	VK_LOGIC_OP_AND_REVERSE,
	VK_LOGIC_OP_COPY,
	VK_LOGIC_OP_AND_INVERTED,
	VK_LOGIC_OP_NO_OP,
	VK_LOGIC_OP_XOR,
	VK_LOGIC_OP_OR,
	VK_LOGIC_OP_NOR,
	VK_LOGIC_OP_EQUIVALENT,
	VK_LOGIC_OP_INVERT,
	VK_LOGIC_OP_OR_REVERSE,
	VK_LOGIC_OP_COPY_INVERTED,
	VK_LOGIC_OP_OR_INVERTED,
	VK_LOGIC_OP_NAND,
	VK_LOGIC_OP_SET
};

const VkBlendFactor RenderingDeviceVulkan::blend_factors[RenderingDevice::BLEND_FACTOR_MAX] = {
	VK_BLEND_FACTOR_ZERO,
	VK_BLEND_FACTOR_ONE,
	VK_BLEND_FACTOR_SRC_COLOR,
	VK_BLEND_FACTOR_ONE_MINUS_SRC_COLOR,
	VK_BLEND_FACTOR_DST_COLOR,
	VK_BLEND_FACTOR_ONE_MINUS_DST_COLOR,
	VK_BLEND_FACTOR_SRC_ALPHA,
	VK_BLEND_FACTOR_ONE_MINUS_SRC_ALPHA,
	VK_BLEND_FACTOR_DST_ALPHA,
	VK_BLEND_FACTOR_ONE_MINUS_DST_ALPHA,
	VK_BLEND_FACTOR_CONSTANT_COLOR,
	VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_COLOR,
	VK_BLEND_FACTOR_CONSTANT_ALPHA,
	VK_BLEND_FACTOR_ONE_MINUS_CONSTANT_ALPHA,
	VK_BLEND_FACTOR_SRC_ALPHA_SATURATE,
	VK_BLEND_FACTOR_SRC1_COLOR,
	VK_BLEND_FACTOR_ONE_MINUS_SRC1_COLOR,
	VK_BLEND_FACTOR_SRC1_ALPHA,
	VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA
};
const VkBlendOp RenderingDeviceVulkan::blend_operations[RenderingDevice::BLEND_OP_MAX] = {
	VK_BLEND_OP_ADD,
	VK_BLEND_OP_SUBTRACT,
	VK_BLEND_OP_REVERSE_SUBTRACT,
	VK_BLEND_OP_MIN,
	VK_BLEND_OP_MAX
};

const VkSamplerAddressMode RenderingDeviceVulkan::address_modes[RenderingDevice::SAMPLER_REPEAT_MODE_MAX] = {
	VK_SAMPLER_ADDRESS_MODE_REPEAT,
	VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT,
	VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE,
	VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER,
	VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE
};

const VkBorderColor RenderingDeviceVulkan::sampler_border_colors[RenderingDevice::SAMPLER_BORDER_COLOR_MAX] = {
	VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK,
	VK_BORDER_COLOR_INT_TRANSPARENT_BLACK,
	VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK,
	VK_BORDER_COLOR_INT_OPAQUE_BLACK,
	VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE,
	VK_BORDER_COLOR_INT_OPAQUE_WHITE
};

const VkImageType RenderingDeviceVulkan::vulkan_image_type[RenderingDevice::TEXTURE_TYPE_MAX] = {
	VK_IMAGE_TYPE_1D,
	VK_IMAGE_TYPE_2D,
	VK_IMAGE_TYPE_3D,
	VK_IMAGE_TYPE_2D,
	VK_IMAGE_TYPE_1D,
	VK_IMAGE_TYPE_2D,
	VK_IMAGE_TYPE_2D
};

/***************************/
/**** BUFFER MANAGEMENT ****/
/***************************/

Error RenderingDeviceVulkan::_buffer_allocate(Buffer *p_buffer, uint32_t p_size, uint32_t p_usage, VmaMemoryUsage p_mem_usage, VmaAllocationCreateFlags p_mem_flags) {
	VkBufferCreateInfo bufferInfo;
	bufferInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
	bufferInfo.pNext = nullptr;
	bufferInfo.flags = 0;
	bufferInfo.size = p_size;
	bufferInfo.usage = p_usage;
	bufferInfo.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
	bufferInfo.queueFamilyIndexCount = 0;
	bufferInfo.pQueueFamilyIndices = nullptr;

	VmaAllocationCreateInfo allocInfo;
	allocInfo.flags = p_mem_flags;
	allocInfo.usage = p_mem_usage;
	allocInfo.requiredFlags = 0;
	allocInfo.preferredFlags = 0;
	allocInfo.memoryTypeBits = 0;
	allocInfo.pool = nullptr;
	allocInfo.pUserData = nullptr;
	if (p_size <= SMALL_ALLOCATION_MAX_SIZE) {
		uint32_t mem_type_index = 0;
		vmaFindMemoryTypeIndexForBufferInfo(allocator, &bufferInfo, &allocInfo, &mem_type_index);
		allocInfo.pool = _find_or_create_small_allocs_pool(mem_type_index);
	}

	VkResult err = vmaCreateBuffer(allocator, &bufferInfo, &allocInfo, &p_buffer->buffer, &p_buffer->allocation, nullptr);
	ERR_FAIL_COND_V_MSG(err, ERR_CANT_CREATE, "Can't create buffer of size: " + itos(p_size) + ", error " + itos(err) + ".");
	p_buffer->size = p_size;
	p_buffer->buffer_info.buffer = p_buffer->buffer;
	p_buffer->buffer_info.offset = 0;
	p_buffer->buffer_info.range = p_size;
	p_buffer->usage = p_usage;

	buffer_memory += p_size;

	return OK;
}

Error RenderingDeviceVulkan::_buffer_free(Buffer *p_buffer) {
	ERR_FAIL_COND_V(p_buffer->size == 0, ERR_INVALID_PARAMETER);

	buffer_memory -= p_buffer->size;
	vmaDestroyBuffer(allocator, p_buffer->buffer, p_buffer->allocation);
	p_buffer->buffer = VK_NULL_HANDLE;
	p_buffer->allocation = nullptr;
	p_buffer->size = 0;

	return OK;
}

Error RenderingDeviceVulkan::_insert_staging_block() {
	VkBufferCreateInfo bufferInfo;
	bufferInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
	bufferInfo.pNext = nullptr;
	bufferInfo.flags = 0;
	bufferInfo.size = staging_buffer_block_size;
	bufferInfo.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT;
	bufferInfo.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
	bufferInfo.queueFamilyIndexCount = 0;
	bufferInfo.pQueueFamilyIndices = nullptr;

	VmaAllocationCreateInfo allocInfo;
	allocInfo.flags = VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT;
	allocInfo.usage = VMA_MEMORY_USAGE_AUTO_PREFER_HOST;
	allocInfo.requiredFlags = 0;
	allocInfo.preferredFlags = 0;
	allocInfo.memoryTypeBits = 0;
	allocInfo.pool = nullptr;
	allocInfo.pUserData = nullptr;

	StagingBufferBlock block;

	VkResult err = vmaCreateBuffer(allocator, &bufferInfo, &allocInfo, &block.buffer, &block.allocation, nullptr);
	ERR_FAIL_COND_V_MSG(err, ERR_CANT_CREATE, "vmaCreateBuffer failed with error " + itos(err) + ".");

	block.frame_used = 0;
	block.fill_amount = 0;

	staging_buffer_blocks.insert(staging_buffer_current, block);
	return OK;
}

Error RenderingDeviceVulkan::_staging_buffer_allocate(uint32_t p_amount, uint32_t p_required_align, uint32_t &r_alloc_offset, uint32_t &r_alloc_size, bool p_can_segment) {
	// Determine a block to use.

	r_alloc_size = p_amount;

	while (true) {
		r_alloc_offset = 0;

		// See if we can use current block.
		if (staging_buffer_blocks[staging_buffer_current].frame_used == frames_drawn) {
			// We used this block this frame, let's see if there is still room.

			uint32_t write_from = staging_buffer_blocks[staging_buffer_current].fill_amount;

			{
				uint32_t align_remainder = write_from % p_required_align;
				if (align_remainder != 0) {
					write_from += p_required_align - align_remainder;
				}
			}

			int32_t available_bytes = int32_t(staging_buffer_block_size) - int32_t(write_from);

			if ((int32_t)p_amount < available_bytes) {
				// All is good, we should be ok, all will fit.
				r_alloc_offset = write_from;
			} else if (p_can_segment && available_bytes >= (int32_t)p_required_align) {
				// Ok all won't fit but at least we can fit a chunkie.
				// All is good, update what needs to be written to.
				r_alloc_offset = write_from;
				r_alloc_size = available_bytes - (available_bytes % p_required_align);

			} else {
				// Can't fit it into this buffer.
				// Will need to try next buffer.

				staging_buffer_current = (staging_buffer_current + 1) % staging_buffer_blocks.size();

				// Before doing anything, though, let's check that we didn't manage to fill all blocks.
				// Possible in a single frame.
				if (staging_buffer_blocks[staging_buffer_current].frame_used == frames_drawn) {
					// Guess we did.. ok, let's see if we can insert a new block.
					if ((uint64_t)staging_buffer_blocks.size() * staging_buffer_block_size < staging_buffer_max_size) {
						// We can, so we are safe.
						Error err = _insert_staging_block();
						if (err) {
							return err;
						}
						// Claim for this frame.
						staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn;
					} else {
						// Ok, worst case scenario, all the staging buffers belong to this frame
						// and this frame is not even done.
						// If this is the main thread, it means the user is likely loading a lot of resources at once,.
						// Otherwise, the thread should just be blocked until the next frame (currently unimplemented).

						if (false) { // Separate thread from render.

							//block_until_next_frame()
							continue;
						} else {
							// Flush EVERYTHING including setup commands. IF not immediate, also need to flush the draw commands.
							_flush(true);

							// Clear the whole staging buffer.
							for (int i = 0; i < staging_buffer_blocks.size(); i++) {
								staging_buffer_blocks.write[i].frame_used = 0;
								staging_buffer_blocks.write[i].fill_amount = 0;
							}
							// Claim current.
							staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn;
						}
					}

				} else {
					// Not from current frame, so continue and try again.
					continue;
				}
			}

		} else if (staging_buffer_blocks[staging_buffer_current].frame_used <= frames_drawn - frame_count) {
			// This is an old block, which was already processed, let's reuse.
			staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn;
			staging_buffer_blocks.write[staging_buffer_current].fill_amount = 0;
		} else {
			// This block may still be in use, let's not touch it unless we have to, so.. can we create a new one?
			if ((uint64_t)staging_buffer_blocks.size() * staging_buffer_block_size < staging_buffer_max_size) {
				// We are still allowed to create a new block, so let's do that and insert it for current pos.
				Error err = _insert_staging_block();
				if (err) {
					return err;
				}
				// Claim for this frame.
				staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn;
			} else {
				// Oops, we are out of room and we can't create more.
				// Let's flush older frames.
				// The logic here is that if a game is loading a lot of data from the main thread, it will need to be stalled anyway.
				// If loading from a separate thread, we can block that thread until next frame when more room is made (not currently implemented, though).

				if (false) {
					// Separate thread from render.
					//block_until_next_frame()
					continue; // And try again.
				} else {
					_flush(false);

					for (int i = 0; i < staging_buffer_blocks.size(); i++) {
						// Clear all blocks but the ones from this frame.
						int block_idx = (i + staging_buffer_current) % staging_buffer_blocks.size();
						if (staging_buffer_blocks[block_idx].frame_used == frames_drawn) {
							break; // Ok, we reached something from this frame, abort.
						}

						staging_buffer_blocks.write[block_idx].frame_used = 0;
						staging_buffer_blocks.write[block_idx].fill_amount = 0;
					}

					// Claim for current frame.
					staging_buffer_blocks.write[staging_buffer_current].frame_used = frames_drawn;
				}
			}
		}

		// All was good, break.
		break;
	}

	staging_buffer_used = true;

	return OK;
}

Error RenderingDeviceVulkan::_buffer_update(Buffer *p_buffer, size_t p_offset, const uint8_t *p_data, size_t p_data_size, bool p_use_draw_command_buffer, uint32_t p_required_align) {
	// Submitting may get chunked for various reasons, so convert this to a task.
	size_t to_submit = p_data_size;
	size_t submit_from = 0;

	while (to_submit > 0) {
		uint32_t block_write_offset;
		uint32_t block_write_amount;

		Error err = _staging_buffer_allocate(MIN(to_submit, staging_buffer_block_size), p_required_align, block_write_offset, block_write_amount);
		if (err) {
			return err;
		}

		// Map staging buffer (It's CPU and coherent).

		void *data_ptr = nullptr;
		{
			VkResult vkerr = vmaMapMemory(allocator, staging_buffer_blocks[staging_buffer_current].allocation, &data_ptr);
			ERR_FAIL_COND_V_MSG(vkerr, ERR_CANT_CREATE, "vmaMapMemory failed with error " + itos(vkerr) + ".");
		}

		// Copy to staging buffer.
		memcpy(((uint8_t *)data_ptr) + block_write_offset, p_data + submit_from, block_write_amount);

		// Unmap.
		vmaUnmapMemory(allocator, staging_buffer_blocks[staging_buffer_current].allocation);
		// Insert a command to copy this.

		VkBufferCopy region;
		region.srcOffset = block_write_offset;
		region.dstOffset = submit_from + p_offset;
		region.size = block_write_amount;

		vkCmdCopyBuffer(p_use_draw_command_buffer ? frames[frame].draw_command_buffer : frames[frame].setup_command_buffer, staging_buffer_blocks[staging_buffer_current].buffer, p_buffer->buffer, 1, &region);

		staging_buffer_blocks.write[staging_buffer_current].fill_amount = block_write_offset + block_write_amount;

		to_submit -= block_write_amount;
		submit_from += block_write_amount;
	}

	return OK;
}

void RenderingDeviceVulkan::_memory_barrier(VkPipelineStageFlags p_src_stage_mask, VkPipelineStageFlags p_dst_stage_mask, VkAccessFlags p_src_access, VkAccessFlags p_dst_sccess, bool p_sync_with_draw) {
	VkMemoryBarrier mem_barrier;
	mem_barrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER;
	mem_barrier.pNext = nullptr;
	mem_barrier.srcAccessMask = p_src_access;
	mem_barrier.dstAccessMask = p_dst_sccess;

	if (p_src_stage_mask == 0 || p_dst_stage_mask == 0) {
		return; // No barrier, since this is invalid.
	}
	vkCmdPipelineBarrier(p_sync_with_draw ? frames[frame].draw_command_buffer : frames[frame].setup_command_buffer, p_src_stage_mask, p_dst_stage_mask, 0, 1, &mem_barrier, 0, nullptr, 0, nullptr);
}

void RenderingDeviceVulkan::_full_barrier(bool p_sync_with_draw) {
	// Used for debug.
	_memory_barrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
			VK_ACCESS_INDIRECT_COMMAND_READ_BIT |
					VK_ACCESS_INDEX_READ_BIT |
					VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT |
					VK_ACCESS_UNIFORM_READ_BIT |
					VK_ACCESS_INPUT_ATTACHMENT_READ_BIT |
					VK_ACCESS_SHADER_READ_BIT |
					VK_ACCESS_SHADER_WRITE_BIT |
					VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
					VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
					VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
					VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
					VK_ACCESS_TRANSFER_READ_BIT |
					VK_ACCESS_TRANSFER_WRITE_BIT |
					VK_ACCESS_HOST_READ_BIT |
					VK_ACCESS_HOST_WRITE_BIT,
			VK_ACCESS_INDIRECT_COMMAND_READ_BIT |
					VK_ACCESS_INDEX_READ_BIT |
					VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT |
					VK_ACCESS_UNIFORM_READ_BIT |
					VK_ACCESS_INPUT_ATTACHMENT_READ_BIT |
					VK_ACCESS_SHADER_READ_BIT |
					VK_ACCESS_SHADER_WRITE_BIT |
					VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
					VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
					VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
					VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
					VK_ACCESS_TRANSFER_READ_BIT |
					VK_ACCESS_TRANSFER_WRITE_BIT |
					VK_ACCESS_HOST_READ_BIT |
					VK_ACCESS_HOST_WRITE_BIT,
			p_sync_with_draw);
}

void RenderingDeviceVulkan::_buffer_memory_barrier(VkBuffer buffer, uint64_t p_from, uint64_t p_size, VkPipelineStageFlags p_src_stage_mask, VkPipelineStageFlags p_dst_stage_mask, VkAccessFlags p_src_access, VkAccessFlags p_dst_sccess, bool p_sync_with_draw) {
	VkBufferMemoryBarrier buffer_mem_barrier;
	buffer_mem_barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
	buffer_mem_barrier.pNext = nullptr;
	buffer_mem_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
	buffer_mem_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
	buffer_mem_barrier.srcAccessMask = p_src_access;
	buffer_mem_barrier.dstAccessMask = p_dst_sccess;
	buffer_mem_barrier.buffer = buffer;
	buffer_mem_barrier.offset = p_from;
	buffer_mem_barrier.size = p_size;

	vkCmdPipelineBarrier(p_sync_with_draw ? frames[frame].draw_command_buffer : frames[frame].setup_command_buffer, p_src_stage_mask, p_dst_stage_mask, 0, 0, nullptr, 1, &buffer_mem_barrier, 0, nullptr);
}

/*****************/
/**** TEXTURE ****/
/*****************/

RID RenderingDeviceVulkan::texture_create(const TextureFormat &p_format, const TextureView &p_view, const Vector<Vector<uint8_t>> &p_data) {
	_THREAD_SAFE_METHOD_

	VkImageCreateInfo image_create_info;
	image_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO;
	image_create_info.pNext = nullptr;
	image_create_info.flags = 0;

	// TODO: Check for support via RenderingDevice to enable on mobile when possible.

#ifndef ANDROID_ENABLED

	// vkCreateImage fails with format list on Android (VK_ERROR_OUT_OF_HOST_MEMORY)
	VkImageFormatListCreateInfoKHR format_list_create_info; // Keep out of the if, needed for creation.
	Vector<VkFormat> allowed_formats; // Keep out of the if, needed for creation.
#endif
	if (p_format.shareable_formats.size()) {
		image_create_info.flags |= VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT;

#ifndef ANDROID_ENABLED

		for (int i = 0; i < p_format.shareable_formats.size(); i++) {
			allowed_formats.push_back(vulkan_formats[p_format.shareable_formats[i]]);
		}

		format_list_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_FORMAT_LIST_CREATE_INFO_KHR;
		format_list_create_info.pNext = nullptr;
		format_list_create_info.viewFormatCount = allowed_formats.size();
		format_list_create_info.pViewFormats = allowed_formats.ptr();
		image_create_info.pNext = &format_list_create_info;

		ERR_FAIL_COND_V_MSG(p_format.shareable_formats.find(p_format.format) == -1, RID(),
				"If supplied a list of shareable formats, the current format must be present in the list");
		ERR_FAIL_COND_V_MSG(p_view.format_override != DATA_FORMAT_MAX && p_format.shareable_formats.find(p_view.format_override) == -1, RID(),
				"If supplied a list of shareable formats, the current view format override must be present in the list");
#endif
	}

	if (p_format.texture_type == TEXTURE_TYPE_CUBE || p_format.texture_type == TEXTURE_TYPE_CUBE_ARRAY) {
		image_create_info.flags |= VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT;
	}
	/*if (p_format.type == TEXTURE_TYPE_2D || p_format.type == TEXTURE_TYPE_2D_ARRAY) {
		image_create_info.flags |= VK_IMAGE_CREATE_2D_ARRAY_COMPATIBLE_BIT;
	}*/

	ERR_FAIL_INDEX_V(p_format.texture_type, TEXTURE_TYPE_MAX, RID());

	image_create_info.imageType = vulkan_image_type[p_format.texture_type];

	ERR_FAIL_COND_V_MSG(p_format.width < 1, RID(), "Width must be equal or greater than 1 for all textures");

	image_create_info.format = vulkan_formats[p_format.format];

	image_create_info.extent.width = p_format.width;
	if (image_create_info.imageType == VK_IMAGE_TYPE_3D || image_create_info.imageType == VK_IMAGE_TYPE_2D) {
		ERR_FAIL_COND_V_MSG(p_format.height < 1, RID(), "Height must be equal or greater than 1 for 2D and 3D textures");
		image_create_info.extent.height = p_format.height;
	} else {
		image_create_info.extent.height = 1;
	}

	if (image_create_info.imageType == VK_IMAGE_TYPE_3D) {
		ERR_FAIL_COND_V_MSG(p_format.depth < 1, RID(), "Depth must be equal or greater than 1 for 3D textures");
		image_create_info.extent.depth = p_format.depth;
	} else {
		image_create_info.extent.depth = 1;
	}

	ERR_FAIL_COND_V(p_format.mipmaps < 1, RID());

	image_create_info.mipLevels = p_format.mipmaps;

	if (p_format.texture_type == TEXTURE_TYPE_1D_ARRAY || p_format.texture_type == TEXTURE_TYPE_2D_ARRAY || p_format.texture_type == TEXTURE_TYPE_CUBE_ARRAY || p_format.texture_type == TEXTURE_TYPE_CUBE) {
		ERR_FAIL_COND_V_MSG(p_format.array_layers < 1, RID(),
				"Amount of layers must be equal or greater than 1 for arrays and cubemaps.");
		ERR_FAIL_COND_V_MSG((p_format.texture_type == TEXTURE_TYPE_CUBE_ARRAY || p_format.texture_type == TEXTURE_TYPE_CUBE) && (p_format.array_layers % 6) != 0, RID(),
				"Cubemap and cubemap array textures must provide a layer number that is multiple of 6");
		image_create_info.arrayLayers = p_format.array_layers;
	} else {
		image_create_info.arrayLayers = 1;
	}

	ERR_FAIL_INDEX_V(p_format.samples, TEXTURE_SAMPLES_MAX, RID());

	image_create_info.samples = _ensure_supported_sample_count(p_format.samples);
	image_create_info.tiling = (p_format.usage_bits & TEXTURE_USAGE_CPU_READ_BIT) ? VK_IMAGE_TILING_LINEAR : VK_IMAGE_TILING_OPTIMAL;

	// Usage.
	image_create_info.usage = 0;

	if (p_format.usage_bits & TEXTURE_USAGE_SAMPLING_BIT) {
		image_create_info.usage |= VK_IMAGE_USAGE_SAMPLED_BIT;
	}

	if (p_format.usage_bits & TEXTURE_USAGE_STORAGE_BIT) {
		image_create_info.usage |= VK_IMAGE_USAGE_STORAGE_BIT;
	}

	if (p_format.usage_bits & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
		image_create_info.usage |= VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
	}

	if (p_format.usage_bits & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
		image_create_info.usage |= VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
	}

	if (p_format.usage_bits & TEXTURE_USAGE_INPUT_ATTACHMENT_BIT) {
		image_create_info.usage |= VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT;
	}

	if (p_format.usage_bits & TEXTURE_USAGE_VRS_ATTACHMENT_BIT) {
		image_create_info.usage |= VK_IMAGE_USAGE_FRAGMENT_SHADING_RATE_ATTACHMENT_BIT_KHR;
	}

	if (p_format.usage_bits & TEXTURE_USAGE_CAN_UPDATE_BIT) {
		image_create_info.usage |= VK_IMAGE_USAGE_TRANSFER_DST_BIT;
	}
	if (p_format.usage_bits & TEXTURE_USAGE_CAN_COPY_FROM_BIT) {
		image_create_info.usage |= VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
	}

	if (p_format.usage_bits & TEXTURE_USAGE_CAN_COPY_TO_BIT) {
		image_create_info.usage |= VK_IMAGE_USAGE_TRANSFER_DST_BIT;
	}

	image_create_info.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
	image_create_info.queueFamilyIndexCount = 0;
	image_create_info.pQueueFamilyIndices = nullptr;
	image_create_info.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;

	uint32_t required_mipmaps = get_image_required_mipmaps(image_create_info.extent.width, image_create_info.extent.height, image_create_info.extent.depth);

	ERR_FAIL_COND_V_MSG(required_mipmaps < image_create_info.mipLevels, RID(),
			"Too many mipmaps requested for texture format and dimensions (" + itos(image_create_info.mipLevels) + "), maximum allowed: (" + itos(required_mipmaps) + ").");

	if (p_data.size()) {
		ERR_FAIL_COND_V_MSG(!(p_format.usage_bits & TEXTURE_USAGE_CAN_UPDATE_BIT), RID(),
				"Texture needs the TEXTURE_USAGE_CAN_UPDATE_BIT usage flag in order to be updated at initialization or later");

		int expected_images = image_create_info.arrayLayers;
		ERR_FAIL_COND_V_MSG(p_data.size() != expected_images, RID(),
				"Default supplied data for image format is of invalid length (" + itos(p_data.size()) + "), should be (" + itos(expected_images) + ").");

		for (uint32_t i = 0; i < image_create_info.arrayLayers; i++) {
			uint32_t required_size = get_image_format_required_size(p_format.format, image_create_info.extent.width, image_create_info.extent.height, image_create_info.extent.depth, image_create_info.mipLevels);
			ERR_FAIL_COND_V_MSG((uint32_t)p_data[i].size() != required_size, RID(),
					"Data for slice index " + itos(i) + " (mapped to layer " + itos(i) + ") differs in size (supplied: " + itos(p_data[i].size()) + ") than what is required by the format (" + itos(required_size) + ").");
		}
	}

	{
		// Validate that this image is supported for the intended use.
		VkFormatProperties properties;
		vkGetPhysicalDeviceFormatProperties(context->get_physical_device(), image_create_info.format, &properties);
		VkFormatFeatureFlags flags;

		String format_text = "'" + String(named_formats[p_format.format]) + "'";

		if (p_format.usage_bits & TEXTURE_USAGE_CPU_READ_BIT) {
			flags = properties.linearTilingFeatures;
			format_text += " (with CPU read bit)";
		} else {
			flags = properties.optimalTilingFeatures;
		}

		if (p_format.usage_bits & TEXTURE_USAGE_SAMPLING_BIT && !(flags & VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT)) {
			ERR_FAIL_V_MSG(RID(), "Format " + format_text + " does not support usage as sampling texture.");
		}

		if (p_format.usage_bits & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT && !(flags & VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT)) {
			ERR_FAIL_V_MSG(RID(), "Format " + format_text + " does not support usage as color attachment.");
		}

		if (p_format.usage_bits & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT && !(flags & VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT)) {
			printf("vkformat: %x\n", image_create_info.format);
			ERR_FAIL_V_MSG(RID(), "Format " + format_text + " does not support usage as depth-stencil attachment.");
		}

		if (p_format.usage_bits & TEXTURE_USAGE_STORAGE_BIT && !(flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT)) {
			ERR_FAIL_V_MSG(RID(), "Format " + format_text + " does not support usage as storage image.");
		}

		if (p_format.usage_bits & TEXTURE_USAGE_STORAGE_ATOMIC_BIT && !(flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_ATOMIC_BIT)) {
			ERR_FAIL_V_MSG(RID(), "Format " + format_text + " does not support usage as atomic storage image.");
		}

		// Validation via VK_FORMAT_FEATURE_FRAGMENT_SHADING_RATE_ATTACHMENT_BIT_KHR fails if VRS attachment is not supported.
		if (p_format.usage_bits & TEXTURE_USAGE_VRS_ATTACHMENT_BIT && p_format.format != DATA_FORMAT_R8_UINT) {
			ERR_FAIL_V_MSG(RID(), "Format " + format_text + " does not support usage as VRS attachment.");
		}
	}

	// Some view validation.

	if (p_view.format_override != DATA_FORMAT_MAX) {
		ERR_FAIL_INDEX_V(p_view.format_override, DATA_FORMAT_MAX, RID());
	}
	ERR_FAIL_INDEX_V(p_view.swizzle_r, TEXTURE_SWIZZLE_MAX, RID());
	ERR_FAIL_INDEX_V(p_view.swizzle_g, TEXTURE_SWIZZLE_MAX, RID());
	ERR_FAIL_INDEX_V(p_view.swizzle_b, TEXTURE_SWIZZLE_MAX, RID());
	ERR_FAIL_INDEX_V(p_view.swizzle_a, TEXTURE_SWIZZLE_MAX, RID());

	// Allocate memory.

	uint32_t width, height;
	uint32_t image_size = get_image_format_required_size(p_format.format, p_format.width, p_format.height, p_format.depth, p_format.mipmaps, &width, &height);

	VmaAllocationCreateInfo allocInfo;
	allocInfo.flags = (p_format.usage_bits & TEXTURE_USAGE_CPU_READ_BIT) ? VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT : 0;
	allocInfo.pool = nullptr;
	allocInfo.usage = VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE;
	allocInfo.requiredFlags = 0;
	allocInfo.preferredFlags = 0;
	allocInfo.memoryTypeBits = 0;
	allocInfo.pUserData = nullptr;
	if (image_size <= SMALL_ALLOCATION_MAX_SIZE) {
		uint32_t mem_type_index = 0;
		vmaFindMemoryTypeIndexForImageInfo(allocator, &image_create_info, &allocInfo, &mem_type_index);
		allocInfo.pool = _find_or_create_small_allocs_pool(mem_type_index);
	}

	Texture texture;

	VkResult err = vmaCreateImage(allocator, &image_create_info, &allocInfo, &texture.image, &texture.allocation, &texture.allocation_info);
	ERR_FAIL_COND_V_MSG(err, RID(), "vmaCreateImage failed with error " + itos(err) + ".");
	image_memory += texture.allocation_info.size;
	texture.type = p_format.texture_type;
	texture.format = p_format.format;
	texture.width = image_create_info.extent.width;
	texture.height = image_create_info.extent.height;
	texture.depth = image_create_info.extent.depth;
	texture.layers = image_create_info.arrayLayers;
	texture.mipmaps = image_create_info.mipLevels;
	texture.base_mipmap = 0;
	texture.base_layer = 0;
	texture.is_resolve_buffer = p_format.is_resolve_buffer;
	texture.usage_flags = p_format.usage_bits;
	texture.samples = p_format.samples;
	texture.allowed_shared_formats = p_format.shareable_formats;

	// Set base layout based on usage priority.

	if (p_format.usage_bits & TEXTURE_USAGE_SAMPLING_BIT) {
		// First priority, readable.
		texture.layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;

	} else if (p_format.usage_bits & TEXTURE_USAGE_STORAGE_BIT) {
		// Second priority, storage.

		texture.layout = VK_IMAGE_LAYOUT_GENERAL;

	} else if (p_format.usage_bits & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
		// Third priority, color or depth.

		texture.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;

	} else if (p_format.usage_bits & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
		texture.layout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;

	} else {
		texture.layout = VK_IMAGE_LAYOUT_GENERAL;
	}

	if (p_format.usage_bits & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
		texture.read_aspect_mask = VK_IMAGE_ASPECT_DEPTH_BIT;
		texture.barrier_aspect_mask = VK_IMAGE_ASPECT_DEPTH_BIT;

		if (format_has_stencil(p_format.format)) {
			texture.barrier_aspect_mask |= VK_IMAGE_ASPECT_STENCIL_BIT;
		}
	} else {
		texture.read_aspect_mask = VK_IMAGE_ASPECT_COLOR_BIT;
		texture.barrier_aspect_mask = VK_IMAGE_ASPECT_COLOR_BIT;
	}

	texture.bound = false;

	// Create view.

	VkImageViewCreateInfo image_view_create_info;
	image_view_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO;
	image_view_create_info.pNext = nullptr;
	image_view_create_info.flags = 0;
	image_view_create_info.image = texture.image;

	static const VkImageViewType view_types[TEXTURE_TYPE_MAX] = {
		VK_IMAGE_VIEW_TYPE_1D,
		VK_IMAGE_VIEW_TYPE_2D,
		VK_IMAGE_VIEW_TYPE_3D,
		VK_IMAGE_VIEW_TYPE_CUBE,
		VK_IMAGE_VIEW_TYPE_1D_ARRAY,
		VK_IMAGE_VIEW_TYPE_2D_ARRAY,
		VK_IMAGE_VIEW_TYPE_CUBE_ARRAY,
	};

	image_view_create_info.viewType = view_types[p_format.texture_type];
	if (p_view.format_override == DATA_FORMAT_MAX) {
		image_view_create_info.format = image_create_info.format;
	} else {
		image_view_create_info.format = vulkan_formats[p_view.format_override];
	}

	static const VkComponentSwizzle component_swizzles[TEXTURE_SWIZZLE_MAX] = {
		VK_COMPONENT_SWIZZLE_IDENTITY,
		VK_COMPONENT_SWIZZLE_ZERO,
		VK_COMPONENT_SWIZZLE_ONE,
		VK_COMPONENT_SWIZZLE_R,
		VK_COMPONENT_SWIZZLE_G,
		VK_COMPONENT_SWIZZLE_B,
		VK_COMPONENT_SWIZZLE_A
	};

	image_view_create_info.components.r = component_swizzles[p_view.swizzle_r];
	image_view_create_info.components.g = component_swizzles[p_view.swizzle_g];
	image_view_create_info.components.b = component_swizzles[p_view.swizzle_b];
	image_view_create_info.components.a = component_swizzles[p_view.swizzle_a];

	image_view_create_info.subresourceRange.baseMipLevel = 0;
	image_view_create_info.subresourceRange.levelCount = image_create_info.mipLevels;
	image_view_create_info.subresourceRange.baseArrayLayer = 0;
	image_view_create_info.subresourceRange.layerCount = image_create_info.arrayLayers;
	if (p_format.usage_bits & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
		image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
	} else {
		image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
	}

	err = vkCreateImageView(device, &image_view_create_info, nullptr, &texture.view);

	if (err) {
		vmaDestroyImage(allocator, texture.image, texture.allocation);
		ERR_FAIL_V_MSG(RID(), "vkCreateImageView failed with error " + itos(err) + ".");
	}

	// Barrier to set layout.
	{
		VkImageMemoryBarrier image_memory_barrier;
		image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
		image_memory_barrier.pNext = nullptr;
		image_memory_barrier.srcAccessMask = 0;
		image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
		image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_UNDEFINED;
		image_memory_barrier.newLayout = texture.layout;
		image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.image = texture.image;
		image_memory_barrier.subresourceRange.aspectMask = texture.barrier_aspect_mask;
		image_memory_barrier.subresourceRange.baseMipLevel = 0;
		image_memory_barrier.subresourceRange.levelCount = image_create_info.mipLevels;
		image_memory_barrier.subresourceRange.baseArrayLayer = 0;
		image_memory_barrier.subresourceRange.layerCount = image_create_info.arrayLayers;

		vkCmdPipelineBarrier(frames[frame].setup_command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
	}

	RID id = texture_owner.make_rid(texture);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif

	if (p_data.size()) {
		for (uint32_t i = 0; i < image_create_info.arrayLayers; i++) {
			_texture_update(id, i, p_data[i], RD::BARRIER_MASK_ALL_BARRIERS, true);
		}
	}
	return id;
}

RID RenderingDeviceVulkan::texture_create_shared(const TextureView &p_view, RID p_with_texture) {
	_THREAD_SAFE_METHOD_

	Texture *src_texture = texture_owner.get_or_null(p_with_texture);
	ERR_FAIL_COND_V(!src_texture, RID());

	if (src_texture->owner.is_valid()) { // Ahh this is a share.
		p_with_texture = src_texture->owner;
		src_texture = texture_owner.get_or_null(src_texture->owner);
		ERR_FAIL_COND_V(!src_texture, RID()); // This is a bug.
	}

	// Create view.

	Texture texture = *src_texture;

	VkImageViewCreateInfo image_view_create_info;
	image_view_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO;
	image_view_create_info.pNext = nullptr;
	image_view_create_info.flags = 0;
	image_view_create_info.image = texture.image;

	static const VkImageViewType view_types[TEXTURE_TYPE_MAX] = {
		VK_IMAGE_VIEW_TYPE_1D,
		VK_IMAGE_VIEW_TYPE_2D,
		VK_IMAGE_VIEW_TYPE_3D,
		VK_IMAGE_VIEW_TYPE_CUBE,
		VK_IMAGE_VIEW_TYPE_1D_ARRAY,
		VK_IMAGE_VIEW_TYPE_2D_ARRAY,
		VK_IMAGE_VIEW_TYPE_CUBE_ARRAY,
	};

	image_view_create_info.viewType = view_types[texture.type];
	if (p_view.format_override == DATA_FORMAT_MAX || p_view.format_override == texture.format) {
		image_view_create_info.format = vulkan_formats[texture.format];
	} else {
		ERR_FAIL_INDEX_V(p_view.format_override, DATA_FORMAT_MAX, RID());

		ERR_FAIL_COND_V_MSG(texture.allowed_shared_formats.find(p_view.format_override) == -1, RID(),
				"Format override is not in the list of allowed shareable formats for original texture.");
		image_view_create_info.format = vulkan_formats[p_view.format_override];
	}

	static const VkComponentSwizzle component_swizzles[TEXTURE_SWIZZLE_MAX] = {
		VK_COMPONENT_SWIZZLE_IDENTITY,
		VK_COMPONENT_SWIZZLE_ZERO,
		VK_COMPONENT_SWIZZLE_ONE,
		VK_COMPONENT_SWIZZLE_R,
		VK_COMPONENT_SWIZZLE_G,
		VK_COMPONENT_SWIZZLE_B,
		VK_COMPONENT_SWIZZLE_A
	};

	image_view_create_info.components.r = component_swizzles[p_view.swizzle_r];
	image_view_create_info.components.g = component_swizzles[p_view.swizzle_g];
	image_view_create_info.components.b = component_swizzles[p_view.swizzle_b];
	image_view_create_info.components.a = component_swizzles[p_view.swizzle_a];

	image_view_create_info.subresourceRange.baseMipLevel = 0;
	image_view_create_info.subresourceRange.levelCount = texture.mipmaps;
	image_view_create_info.subresourceRange.layerCount = texture.layers;
	image_view_create_info.subresourceRange.baseArrayLayer = 0;

	if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
		image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
	} else {
		image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
	}

	VkImageViewUsageCreateInfo usage_info;
	usage_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_USAGE_CREATE_INFO;
	usage_info.pNext = nullptr;
	if (p_view.format_override != DATA_FORMAT_MAX) {
		// Need to validate usage with vulkan.

		usage_info.usage = 0;

		if (texture.usage_flags & TEXTURE_USAGE_SAMPLING_BIT) {
			usage_info.usage |= VK_IMAGE_USAGE_SAMPLED_BIT;
		}

		if (texture.usage_flags & TEXTURE_USAGE_STORAGE_BIT) {
			if (texture_is_format_supported_for_usage(p_view.format_override, TEXTURE_USAGE_STORAGE_BIT)) {
				usage_info.usage |= VK_IMAGE_USAGE_STORAGE_BIT;
			}
		}

		if (texture.usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
			if (texture_is_format_supported_for_usage(p_view.format_override, TEXTURE_USAGE_COLOR_ATTACHMENT_BIT)) {
				usage_info.usage |= VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
			}
		}

		if (texture.usage_flags & TEXTURE_USAGE_INPUT_ATTACHMENT_BIT) {
			usage_info.usage |= VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT;
		}

		if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
			usage_info.usage |= VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
		}

		if (texture.usage_flags & TEXTURE_USAGE_CAN_UPDATE_BIT) {
			usage_info.usage |= VK_IMAGE_USAGE_TRANSFER_DST_BIT;
		}
		if (texture.usage_flags & TEXTURE_USAGE_CAN_COPY_FROM_BIT) {
			usage_info.usage |= VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
		}

		if (texture.usage_flags & TEXTURE_USAGE_CAN_COPY_TO_BIT) {
			usage_info.usage |= VK_IMAGE_USAGE_TRANSFER_DST_BIT;
		}

		image_view_create_info.pNext = &usage_info;
	}

	VkResult err = vkCreateImageView(device, &image_view_create_info, nullptr, &texture.view);
	ERR_FAIL_COND_V_MSG(err, RID(), "vkCreateImageView failed with error " + itos(err) + ".");

	texture.owner = p_with_texture;
	RID id = texture_owner.make_rid(texture);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif
	_add_dependency(id, p_with_texture);

	return id;
}

RID RenderingDeviceVulkan::texture_create_from_extension(TextureType p_type, DataFormat p_format, TextureSamples p_samples, uint64_t p_flags, uint64_t p_image, uint64_t p_width, uint64_t p_height, uint64_t p_depth, uint64_t p_layers) {
	_THREAD_SAFE_METHOD_
	// This method creates a texture object using a VkImage created by an extension, module or other external source (OpenXR uses this).
	VkImage image = (VkImage)p_image;

	Texture texture;
	texture.image = image;
	// If we leave texture.allocation as a nullptr, would that be enough to detect we don't "own" the image?
	// Also leave texture.allocation_info alone.
	// We'll set texture.view later on.
	texture.type = p_type;
	texture.format = p_format;
	texture.samples = p_samples;
	texture.width = p_width;
	texture.height = p_height;
	texture.depth = p_depth;
	texture.layers = p_layers;
	texture.mipmaps = 1;
	texture.usage_flags = p_flags;
	texture.base_mipmap = 0;
	texture.base_layer = 0;
	texture.allowed_shared_formats.push_back(RD::DATA_FORMAT_R8G8B8A8_UNORM);
	texture.allowed_shared_formats.push_back(RD::DATA_FORMAT_R8G8B8A8_SRGB);

	// Set base layout based on usage priority.

	if (texture.usage_flags & TEXTURE_USAGE_SAMPLING_BIT) {
		// First priority, readable.
		texture.layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;

	} else if (texture.usage_flags & TEXTURE_USAGE_STORAGE_BIT) {
		// Second priority, storage.

		texture.layout = VK_IMAGE_LAYOUT_GENERAL;

	} else if (texture.usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
		// Third priority, color or depth.

		texture.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;

	} else if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
		texture.layout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;

	} else {
		texture.layout = VK_IMAGE_LAYOUT_GENERAL;
	}

	if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
		texture.read_aspect_mask = VK_IMAGE_ASPECT_DEPTH_BIT;
		texture.barrier_aspect_mask = VK_IMAGE_ASPECT_DEPTH_BIT;

		// if (format_has_stencil(p_format.format)) {
		// 	texture.barrier_aspect_mask |= VK_IMAGE_ASPECT_STENCIL_BIT;
		// }
	} else {
		texture.read_aspect_mask = VK_IMAGE_ASPECT_COLOR_BIT;
		texture.barrier_aspect_mask = VK_IMAGE_ASPECT_COLOR_BIT;
	}

	// Create a view for us to use.

	VkImageViewCreateInfo image_view_create_info;
	image_view_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO;
	image_view_create_info.pNext = nullptr;
	image_view_create_info.flags = 0;
	image_view_create_info.image = texture.image;

	static const VkImageViewType view_types[TEXTURE_TYPE_MAX] = {
		VK_IMAGE_VIEW_TYPE_1D,
		VK_IMAGE_VIEW_TYPE_2D,
		VK_IMAGE_VIEW_TYPE_3D,
		VK_IMAGE_VIEW_TYPE_CUBE,
		VK_IMAGE_VIEW_TYPE_1D_ARRAY,
		VK_IMAGE_VIEW_TYPE_2D_ARRAY,
		VK_IMAGE_VIEW_TYPE_CUBE_ARRAY,
	};

	image_view_create_info.viewType = view_types[texture.type];
	image_view_create_info.format = vulkan_formats[texture.format];

	static const VkComponentSwizzle component_swizzles[TEXTURE_SWIZZLE_MAX] = {
		VK_COMPONENT_SWIZZLE_IDENTITY,
		VK_COMPONENT_SWIZZLE_ZERO,
		VK_COMPONENT_SWIZZLE_ONE,
		VK_COMPONENT_SWIZZLE_R,
		VK_COMPONENT_SWIZZLE_G,
		VK_COMPONENT_SWIZZLE_B,
		VK_COMPONENT_SWIZZLE_A
	};

	// Hardcode for now, maybe make this settable from outside.
	image_view_create_info.components.r = component_swizzles[TEXTURE_SWIZZLE_R];
	image_view_create_info.components.g = component_swizzles[TEXTURE_SWIZZLE_G];
	image_view_create_info.components.b = component_swizzles[TEXTURE_SWIZZLE_B];
	image_view_create_info.components.a = component_swizzles[TEXTURE_SWIZZLE_A];

	image_view_create_info.subresourceRange.baseMipLevel = 0;
	image_view_create_info.subresourceRange.levelCount = texture.mipmaps;
	image_view_create_info.subresourceRange.baseArrayLayer = 0;
	image_view_create_info.subresourceRange.layerCount = texture.layers;
	if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
		image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
	} else {
		image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
	}

	VkResult err = vkCreateImageView(device, &image_view_create_info, nullptr, &texture.view);

	if (err) {
		// vmaDestroyImage(allocator, texture.image, texture.allocation);
		ERR_FAIL_V_MSG(RID(), "vkCreateImageView failed with error " + itos(err) + ".");
	}

	// Barrier to set layout.
	{
		VkImageMemoryBarrier image_memory_barrier;
		image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
		image_memory_barrier.pNext = nullptr;
		image_memory_barrier.srcAccessMask = 0;
		image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
		image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_UNDEFINED;
		image_memory_barrier.newLayout = texture.layout;
		image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.image = texture.image;
		image_memory_barrier.subresourceRange.aspectMask = texture.barrier_aspect_mask;
		image_memory_barrier.subresourceRange.baseMipLevel = 0;
		image_memory_barrier.subresourceRange.levelCount = texture.mipmaps;
		image_memory_barrier.subresourceRange.baseArrayLayer = 0;
		image_memory_barrier.subresourceRange.layerCount = texture.layers;

		vkCmdPipelineBarrier(frames[frame].setup_command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
	}

	RID id = texture_owner.make_rid(texture);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif

	return id;
}

RID RenderingDeviceVulkan::texture_create_shared_from_slice(const TextureView &p_view, RID p_with_texture, uint32_t p_layer, uint32_t p_mipmap, uint32_t p_mipmaps, TextureSliceType p_slice_type) {
	_THREAD_SAFE_METHOD_

	Texture *src_texture = texture_owner.get_or_null(p_with_texture);
	ERR_FAIL_COND_V(!src_texture, RID());

	if (src_texture->owner.is_valid()) { // Ahh this is a share.
		p_with_texture = src_texture->owner;
		src_texture = texture_owner.get_or_null(src_texture->owner);
		ERR_FAIL_COND_V(!src_texture, RID()); // This is a bug.
	}

	ERR_FAIL_COND_V_MSG(p_slice_type == TEXTURE_SLICE_CUBEMAP && (src_texture->type != TEXTURE_TYPE_CUBE && src_texture->type != TEXTURE_TYPE_CUBE_ARRAY), RID(),
			"Can only create a cubemap slice from a cubemap or cubemap array mipmap");

	ERR_FAIL_COND_V_MSG(p_slice_type == TEXTURE_SLICE_3D && src_texture->type != TEXTURE_TYPE_3D, RID(),
			"Can only create a 3D slice from a 3D texture");

	ERR_FAIL_COND_V_MSG(p_slice_type == TEXTURE_SLICE_2D_ARRAY && (src_texture->type != TEXTURE_TYPE_2D_ARRAY), RID(),
			"Can only create an array slice from a 2D array mipmap");

	// Create view.

	ERR_FAIL_UNSIGNED_INDEX_V(p_mipmap, src_texture->mipmaps, RID());
	ERR_FAIL_COND_V(p_mipmap + p_mipmaps > src_texture->mipmaps, RID());
	ERR_FAIL_UNSIGNED_INDEX_V(p_layer, src_texture->layers, RID());

	int slice_layers = 1;
	if (p_slice_type == TEXTURE_SLICE_2D_ARRAY) {
		ERR_FAIL_COND_V_MSG(p_layer != 0, RID(), "layer must be 0 when obtaining a 2D array mipmap slice");
		slice_layers = src_texture->layers;
	} else if (p_slice_type == TEXTURE_SLICE_CUBEMAP) {
		slice_layers = 6;
	}

	Texture texture = *src_texture;
	get_image_format_required_size(texture.format, texture.width, texture.height, texture.depth, p_mipmap + 1, &texture.width, &texture.height);
	texture.mipmaps = p_mipmaps;
	texture.layers = slice_layers;
	texture.base_mipmap = p_mipmap;
	texture.base_layer = p_layer;

	VkImageViewCreateInfo image_view_create_info;
	image_view_create_info.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO;
	image_view_create_info.pNext = nullptr;
	image_view_create_info.flags = 0;
	image_view_create_info.image = texture.image;

	static const VkImageViewType view_types[TEXTURE_TYPE_MAX] = {
		VK_IMAGE_VIEW_TYPE_1D,
		VK_IMAGE_VIEW_TYPE_2D,
		VK_IMAGE_VIEW_TYPE_2D,
		VK_IMAGE_VIEW_TYPE_2D,
		VK_IMAGE_VIEW_TYPE_1D,
		VK_IMAGE_VIEW_TYPE_2D,
		VK_IMAGE_VIEW_TYPE_2D,
	};

	image_view_create_info.viewType = view_types[texture.type];

	if (p_slice_type == TEXTURE_SLICE_CUBEMAP) {
		image_view_create_info.viewType = VK_IMAGE_VIEW_TYPE_CUBE;
	} else if (p_slice_type == TEXTURE_SLICE_3D) {
		image_view_create_info.viewType = VK_IMAGE_VIEW_TYPE_3D;
	} else if (p_slice_type == TEXTURE_SLICE_2D_ARRAY) {
		image_view_create_info.viewType = VK_IMAGE_VIEW_TYPE_2D_ARRAY;
	}

	if (p_slice_type == TEXTURE_SLICE_2D) {
		texture.type = TEXTURE_TYPE_2D;
	} else if (p_slice_type == TEXTURE_SLICE_3D) {
		texture.type = TEXTURE_TYPE_3D;
	}

	if (p_view.format_override == DATA_FORMAT_MAX || p_view.format_override == texture.format) {
		image_view_create_info.format = vulkan_formats[texture.format];
	} else {
		ERR_FAIL_INDEX_V(p_view.format_override, DATA_FORMAT_MAX, RID());

		ERR_FAIL_COND_V_MSG(texture.allowed_shared_formats.find(p_view.format_override) == -1, RID(),
				"Format override is not in the list of allowed shareable formats for original texture.");
		image_view_create_info.format = vulkan_formats[p_view.format_override];
	}

	static const VkComponentSwizzle component_swizzles[TEXTURE_SWIZZLE_MAX] = {
		VK_COMPONENT_SWIZZLE_IDENTITY,
		VK_COMPONENT_SWIZZLE_ZERO,
		VK_COMPONENT_SWIZZLE_ONE,
		VK_COMPONENT_SWIZZLE_R,
		VK_COMPONENT_SWIZZLE_G,
		VK_COMPONENT_SWIZZLE_B,
		VK_COMPONENT_SWIZZLE_A
	};

	image_view_create_info.components.r = component_swizzles[p_view.swizzle_r];
	image_view_create_info.components.g = component_swizzles[p_view.swizzle_g];
	image_view_create_info.components.b = component_swizzles[p_view.swizzle_b];
	image_view_create_info.components.a = component_swizzles[p_view.swizzle_a];

	if (p_slice_type == TEXTURE_SLICE_CUBEMAP) {
		ERR_FAIL_COND_V_MSG(p_layer >= src_texture->layers, RID(),
				"Specified layer is invalid for cubemap");
		ERR_FAIL_COND_V_MSG((p_layer % 6) != 0, RID(),
				"Specified layer must be a multiple of 6.");
	}
	image_view_create_info.subresourceRange.baseMipLevel = p_mipmap;
	image_view_create_info.subresourceRange.levelCount = p_mipmaps;
	image_view_create_info.subresourceRange.layerCount = slice_layers;
	image_view_create_info.subresourceRange.baseArrayLayer = p_layer;

	if (texture.usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
		image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
	} else {
		image_view_create_info.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
	}

	VkResult err = vkCreateImageView(device, &image_view_create_info, nullptr, &texture.view);
	ERR_FAIL_COND_V_MSG(err, RID(), "vkCreateImageView failed with error " + itos(err) + ".");

	texture.owner = p_with_texture;
	RID id = texture_owner.make_rid(texture);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif
	_add_dependency(id, p_with_texture);

	return id;
}

Error RenderingDeviceVulkan::texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, BitField<BarrierMask> p_post_barrier) {
	return _texture_update(p_texture, p_layer, p_data, p_post_barrier, false);
}

static _ALWAYS_INLINE_ void _copy_region(uint8_t const *__restrict p_src, uint8_t *__restrict p_dst, uint32_t p_src_x, uint32_t p_src_y, uint32_t p_src_w, uint32_t p_src_h, uint32_t p_src_full_w, uint32_t p_unit_size) {
	uint32_t src_offset = (p_src_y * p_src_full_w + p_src_x) * p_unit_size;
	uint32_t dst_offset = 0;
	for (uint32_t y = p_src_h; y > 0; y--) {
		uint8_t const *__restrict src = p_src + src_offset;
		uint8_t *__restrict dst = p_dst + dst_offset;
		for (uint32_t x = p_src_w * p_unit_size; x > 0; x--) {
			*dst = *src;
			src++;
			dst++;
		}
		src_offset += p_src_full_w * p_unit_size;
		dst_offset += p_src_w * p_unit_size;
	}
}

Error RenderingDeviceVulkan::_texture_update(RID p_texture, uint32_t p_layer, const Vector<uint8_t> &p_data, BitField<BarrierMask> p_post_barrier, bool p_use_setup_queue) {
	_THREAD_SAFE_METHOD_

	ERR_FAIL_COND_V_MSG((draw_list || compute_list) && !p_use_setup_queue, ERR_INVALID_PARAMETER,
			"Updating textures is forbidden during creation of a draw or compute list");

	Texture *texture = texture_owner.get_or_null(p_texture);
	ERR_FAIL_COND_V(!texture, ERR_INVALID_PARAMETER);

	if (texture->owner != RID()) {
		p_texture = texture->owner;
		texture = texture_owner.get_or_null(texture->owner);
		ERR_FAIL_COND_V(!texture, ERR_BUG); // This is a bug.
	}

	ERR_FAIL_COND_V_MSG(texture->bound, ERR_CANT_ACQUIRE_RESOURCE,
			"Texture can't be updated while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");

	ERR_FAIL_COND_V_MSG(!(texture->usage_flags & TEXTURE_USAGE_CAN_UPDATE_BIT), ERR_INVALID_PARAMETER,
			"Texture requires the TEXTURE_USAGE_CAN_UPDATE_BIT in order to be updatable.");

	uint32_t layer_count = texture->layers;
	if (texture->type == TEXTURE_TYPE_CUBE || texture->type == TEXTURE_TYPE_CUBE_ARRAY) {
		layer_count *= 6;
	}
	ERR_FAIL_COND_V(p_layer >= layer_count, ERR_INVALID_PARAMETER);

	uint32_t width, height;
	uint32_t image_size = get_image_format_required_size(texture->format, texture->width, texture->height, texture->depth, texture->mipmaps, &width, &height);
	uint32_t required_size = image_size;
	uint32_t required_align = get_compressed_image_format_block_byte_size(texture->format);
	if (required_align == 1) {
		required_align = get_image_format_pixel_size(texture->format);
	}
	if ((required_align % 4) != 0) { // Alignment rules are really strange.
		required_align *= 4;
	}

	ERR_FAIL_COND_V_MSG(required_size != (uint32_t)p_data.size(), ERR_INVALID_PARAMETER,
			"Required size for texture update (" + itos(required_size) + ") does not match data supplied size (" + itos(p_data.size()) + ").");

	uint32_t region_size = texture_upload_region_size_px;

	const uint8_t *r = p_data.ptr();

	VkCommandBuffer command_buffer = p_use_setup_queue ? frames[frame].setup_command_buffer : frames[frame].draw_command_buffer;

	// Barrier to transfer.
	{
		VkImageMemoryBarrier image_memory_barrier;
		image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
		image_memory_barrier.pNext = nullptr;
		image_memory_barrier.srcAccessMask = 0;
		image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
		image_memory_barrier.oldLayout = texture->layout;
		image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;

		image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.image = texture->image;
		image_memory_barrier.subresourceRange.aspectMask = texture->barrier_aspect_mask;
		image_memory_barrier.subresourceRange.baseMipLevel = 0;
		image_memory_barrier.subresourceRange.levelCount = texture->mipmaps;
		image_memory_barrier.subresourceRange.baseArrayLayer = p_layer;
		image_memory_barrier.subresourceRange.layerCount = 1;

		vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
	}

	uint32_t mipmap_offset = 0;

	uint32_t logic_width = texture->width;
	uint32_t logic_height = texture->height;

	for (uint32_t mm_i = 0; mm_i < texture->mipmaps; mm_i++) {
		uint32_t depth;
		uint32_t image_total = get_image_format_required_size(texture->format, texture->width, texture->height, texture->depth, mm_i + 1, &width, &height, &depth);

		const uint8_t *read_ptr_mipmap = r + mipmap_offset;
		image_size = image_total - mipmap_offset;

		for (uint32_t z = 0; z < depth; z++) { // For 3D textures, depth may be > 0.

			const uint8_t *read_ptr = read_ptr_mipmap + image_size * z / depth;

			for (uint32_t y = 0; y < height; y += region_size) {
				for (uint32_t x = 0; x < width; x += region_size) {
					uint32_t region_w = MIN(region_size, width - x);
					uint32_t region_h = MIN(region_size, height - y);

					uint32_t region_logic_w = MIN(region_size, logic_width - x);
					uint32_t region_logic_h = MIN(region_size, logic_height - y);

					uint32_t pixel_size = get_image_format_pixel_size(texture->format);
					uint32_t to_allocate = region_w * region_h * pixel_size;
					to_allocate >>= get_compressed_image_format_pixel_rshift(texture->format);

					uint32_t alloc_offset, alloc_size;
					Error err = _staging_buffer_allocate(to_allocate, required_align, alloc_offset, alloc_size, false);
					ERR_FAIL_COND_V(err, ERR_CANT_CREATE);

					uint8_t *write_ptr;

					{ // Map.
						void *data_ptr = nullptr;
						VkResult vkerr = vmaMapMemory(allocator, staging_buffer_blocks[staging_buffer_current].allocation, &data_ptr);
						ERR_FAIL_COND_V_MSG(vkerr, ERR_CANT_CREATE, "vmaMapMemory failed with error " + itos(vkerr) + ".");
						write_ptr = (uint8_t *)data_ptr;
						write_ptr += alloc_offset;
					}

					uint32_t block_w, block_h;
					get_compressed_image_format_block_dimensions(texture->format, block_w, block_h);

					ERR_FAIL_COND_V(region_w % block_w, ERR_BUG);
					ERR_FAIL_COND_V(region_h % block_h, ERR_BUG);

					if (block_w != 1 || block_h != 1) {
						// Compressed image (blocks).
						// Must copy a block region.

						uint32_t block_size = get_compressed_image_format_block_byte_size(texture->format);
						// Re-create current variables in blocky format.
						uint32_t xb = x / block_w;
						uint32_t yb = y / block_h;
						uint32_t wb = width / block_w;
						//uint32_t hb = height / block_h;
						uint32_t region_wb = region_w / block_w;
						uint32_t region_hb = region_h / block_h;
						_copy_region(read_ptr, write_ptr, xb, yb, region_wb, region_hb, wb, block_size);
					} else {
						// Regular image (pixels).
						// Must copy a pixel region.
						_copy_region(read_ptr, write_ptr, x, y, region_w, region_h, width, pixel_size);
					}

					{ // Unmap.
						vmaUnmapMemory(allocator, staging_buffer_blocks[staging_buffer_current].allocation);
					}

					VkBufferImageCopy buffer_image_copy;
					buffer_image_copy.bufferOffset = alloc_offset;
					buffer_image_copy.bufferRowLength = 0; // Tightly packed.
					buffer_image_copy.bufferImageHeight = 0; // Tightly packed.

					buffer_image_copy.imageSubresource.aspectMask = texture->read_aspect_mask;
					buffer_image_copy.imageSubresource.mipLevel = mm_i;
					buffer_image_copy.imageSubresource.baseArrayLayer = p_layer;
					buffer_image_copy.imageSubresource.layerCount = 1;

					buffer_image_copy.imageOffset.x = x;
					buffer_image_copy.imageOffset.y = y;
					buffer_image_copy.imageOffset.z = z;

					buffer_image_copy.imageExtent.width = region_logic_w;
					buffer_image_copy.imageExtent.height = region_logic_h;
					buffer_image_copy.imageExtent.depth = 1;

					vkCmdCopyBufferToImage(command_buffer, staging_buffer_blocks[staging_buffer_current].buffer, texture->image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &buffer_image_copy);

					staging_buffer_blocks.write[staging_buffer_current].fill_amount += alloc_size;
				}
			}
		}

		mipmap_offset = image_total;
		logic_width = MAX(1u, logic_width >> 1);
		logic_height = MAX(1u, logic_height >> 1);
	}

	// Barrier to restore layout.
	{
		uint32_t barrier_flags = 0;
		uint32_t access_flags = 0;
		if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) {
			barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
			access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
		}
		if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) {
			barrier_flags |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
			access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
		}
		if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) {
			barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
			access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT;
		}

		if (barrier_flags == 0) {
			barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
		}

		VkImageMemoryBarrier image_memory_barrier;
		image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
		image_memory_barrier.pNext = nullptr;
		image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
		image_memory_barrier.dstAccessMask = access_flags;
		image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
		image_memory_barrier.newLayout = texture->layout;
		image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.image = texture->image;
		image_memory_barrier.subresourceRange.aspectMask = texture->barrier_aspect_mask;
		image_memory_barrier.subresourceRange.baseMipLevel = 0;
		image_memory_barrier.subresourceRange.levelCount = texture->mipmaps;
		image_memory_barrier.subresourceRange.baseArrayLayer = p_layer;
		image_memory_barrier.subresourceRange.layerCount = 1;

		vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
	}

	if (texture->used_in_frame != frames_drawn) {
		texture->used_in_raster = false;
		texture->used_in_compute = false;
		texture->used_in_frame = frames_drawn;
	}
	texture->used_in_transfer = true;

	return OK;
}

Vector<uint8_t> RenderingDeviceVulkan::_texture_get_data_from_image(Texture *tex, VkImage p_image, VmaAllocation p_allocation, uint32_t p_layer, bool p_2d) {
	uint32_t width, height, depth;
	uint32_t image_size = get_image_format_required_size(tex->format, tex->width, tex->height, p_2d ? 1 : tex->depth, tex->mipmaps, &width, &height, &depth);

	Vector<uint8_t> image_data;
	image_data.resize(image_size);

	void *img_mem;
	vmaMapMemory(allocator, p_allocation, &img_mem);

	uint32_t blockw, blockh;
	get_compressed_image_format_block_dimensions(tex->format, blockw, blockh);
	uint32_t block_size = get_compressed_image_format_block_byte_size(tex->format);
	uint32_t pixel_size = get_image_format_pixel_size(tex->format);

	{
		uint8_t *w = image_data.ptrw();

		uint32_t mipmap_offset = 0;
		for (uint32_t mm_i = 0; mm_i < tex->mipmaps; mm_i++) {
			uint32_t image_total = get_image_format_required_size(tex->format, tex->width, tex->height, p_2d ? 1 : tex->depth, mm_i + 1, &width, &height, &depth);

			uint8_t *write_ptr_mipmap = w + mipmap_offset;
			image_size = image_total - mipmap_offset;

			VkImageSubresource image_sub_resorce;
			image_sub_resorce.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
			image_sub_resorce.arrayLayer = p_layer;
			image_sub_resorce.mipLevel = mm_i;
			VkSubresourceLayout layout;
			vkGetImageSubresourceLayout(device, p_image, &image_sub_resorce, &layout);

			for (uint32_t z = 0; z < depth; z++) {
				uint8_t *write_ptr = write_ptr_mipmap + z * image_size / depth;
				const uint8_t *slice_read_ptr = ((uint8_t *)img_mem) + layout.offset + z * layout.depthPitch;

				if (block_size > 1) {
					// Compressed.
					uint32_t line_width = (block_size * (width / blockw));
					for (uint32_t y = 0; y < height / blockh; y++) {
						const uint8_t *rptr = slice_read_ptr + y * layout.rowPitch;
						uint8_t *wptr = write_ptr + y * line_width;

						memcpy(wptr, rptr, line_width);
					}

				} else {
					// Uncompressed.
					for (uint32_t y = 0; y < height; y++) {
						const uint8_t *rptr = slice_read_ptr + y * layout.rowPitch;
						uint8_t *wptr = write_ptr + y * pixel_size * width;
						memcpy(wptr, rptr, (uint64_t)pixel_size * width);
					}
				}
			}

			mipmap_offset = image_total;
		}
	}

	vmaUnmapMemory(allocator, p_allocation);

	return image_data;
}

Vector<uint8_t> RenderingDeviceVulkan::texture_get_data(RID p_texture, uint32_t p_layer) {
	_THREAD_SAFE_METHOD_

	Texture *tex = texture_owner.get_or_null(p_texture);
	ERR_FAIL_COND_V(!tex, Vector<uint8_t>());

	ERR_FAIL_COND_V_MSG(tex->bound, Vector<uint8_t>(),
			"Texture can't be retrieved while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");
	ERR_FAIL_COND_V_MSG(!(tex->usage_flags & TEXTURE_USAGE_CAN_COPY_FROM_BIT), Vector<uint8_t>(),
			"Texture requires the TEXTURE_USAGE_CAN_COPY_FROM_BIT in order to be retrieved.");

	uint32_t layer_count = tex->layers;
	if (tex->type == TEXTURE_TYPE_CUBE || tex->type == TEXTURE_TYPE_CUBE_ARRAY) {
		layer_count *= 6;
	}
	ERR_FAIL_COND_V(p_layer >= layer_count, Vector<uint8_t>());

	if (tex->usage_flags & TEXTURE_USAGE_CPU_READ_BIT) {
		// Does not need anything fancy, map and read.
		return _texture_get_data_from_image(tex, tex->image, tex->allocation, p_layer);
	} else {
		// Compute total image size.
		uint32_t width, height, depth;
		uint32_t buffer_size = get_image_format_required_size(tex->format, tex->width, tex->height, tex->depth, tex->mipmaps, &width, &height, &depth);

		// Allocate buffer.
		VkCommandBuffer command_buffer = frames[frame].draw_command_buffer; // Makes more sense to retrieve.
		Buffer tmp_buffer;
		_buffer_allocate(&tmp_buffer, buffer_size, VK_BUFFER_USAGE_TRANSFER_DST_BIT, VMA_MEMORY_USAGE_AUTO_PREFER_HOST, VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT);

		{ // Source image barrier.
			VkImageMemoryBarrier image_memory_barrier;
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = 0;
			image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
			image_memory_barrier.oldLayout = tex->layout;
			image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;

			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = tex->image;
			image_memory_barrier.subresourceRange.aspectMask = tex->barrier_aspect_mask;
			image_memory_barrier.subresourceRange.baseMipLevel = 0;
			image_memory_barrier.subresourceRange.levelCount = tex->mipmaps;
			image_memory_barrier.subresourceRange.baseArrayLayer = p_layer;
			image_memory_barrier.subresourceRange.layerCount = 1;

			vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
		}

		uint32_t computed_w = tex->width;
		uint32_t computed_h = tex->height;
		uint32_t computed_d = tex->depth;

		uint32_t prev_size = 0;
		uint32_t offset = 0;
		for (uint32_t i = 0; i < tex->mipmaps; i++) {
			VkBufferImageCopy buffer_image_copy;

			uint32_t image_size = get_image_format_required_size(tex->format, tex->width, tex->height, tex->depth, i + 1);
			uint32_t size = image_size - prev_size;
			prev_size = image_size;

			buffer_image_copy.bufferOffset = offset;
			buffer_image_copy.bufferImageHeight = 0;
			buffer_image_copy.bufferRowLength = 0;
			buffer_image_copy.imageSubresource.aspectMask = tex->read_aspect_mask;
			buffer_image_copy.imageSubresource.baseArrayLayer = p_layer;
			buffer_image_copy.imageSubresource.layerCount = 1;
			buffer_image_copy.imageSubresource.mipLevel = i;
			buffer_image_copy.imageOffset.x = 0;
			buffer_image_copy.imageOffset.y = 0;
			buffer_image_copy.imageOffset.z = 0;
			buffer_image_copy.imageExtent.width = computed_w;
			buffer_image_copy.imageExtent.height = computed_h;
			buffer_image_copy.imageExtent.depth = computed_d;

			vkCmdCopyImageToBuffer(command_buffer, tex->image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, tmp_buffer.buffer, 1, &buffer_image_copy);

			computed_w = MAX(1u, computed_w >> 1);
			computed_h = MAX(1u, computed_h >> 1);
			computed_d = MAX(1u, computed_d >> 1);
			offset += size;
		}

		{ // Restore src.
			VkImageMemoryBarrier image_memory_barrier;
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
			image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
			if (tex->usage_flags & TEXTURE_USAGE_STORAGE_BIT) {
				image_memory_barrier.dstAccessMask |= VK_ACCESS_SHADER_WRITE_BIT;
			}
			image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
			image_memory_barrier.newLayout = tex->layout;
			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = tex->image;
			image_memory_barrier.subresourceRange.aspectMask = tex->barrier_aspect_mask;
			image_memory_barrier.subresourceRange.baseMipLevel = 0;
			image_memory_barrier.subresourceRange.levelCount = tex->mipmaps;
			image_memory_barrier.subresourceRange.baseArrayLayer = p_layer;
			image_memory_barrier.subresourceRange.layerCount = 1;

			vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
		}

		_flush(true);

		void *buffer_mem;
		VkResult vkerr = vmaMapMemory(allocator, tmp_buffer.allocation, &buffer_mem);
		ERR_FAIL_COND_V_MSG(vkerr, Vector<uint8_t>(), "vmaMapMemory failed with error " + itos(vkerr) + ".");

		Vector<uint8_t> buffer_data;
		{
			buffer_data.resize(buffer_size);
			uint8_t *w = buffer_data.ptrw();
			memcpy(w, buffer_mem, buffer_size);
		}

		vmaUnmapMemory(allocator, tmp_buffer.allocation);

		_buffer_free(&tmp_buffer);

		return buffer_data;
	}
}

bool RenderingDeviceVulkan::texture_is_shared(RID p_texture) {
	_THREAD_SAFE_METHOD_

	Texture *tex = texture_owner.get_or_null(p_texture);
	ERR_FAIL_COND_V(!tex, false);
	return tex->owner.is_valid();
}

bool RenderingDeviceVulkan::texture_is_valid(RID p_texture) {
	return texture_owner.owns(p_texture);
}

Size2i RenderingDeviceVulkan::texture_size(RID p_texture) {
	_THREAD_SAFE_METHOD_

	Texture *tex = texture_owner.get_or_null(p_texture);
	ERR_FAIL_COND_V(!tex, Size2i());
	return Size2i(tex->width, tex->height);
}

Error RenderingDeviceVulkan::texture_copy(RID p_from_texture, RID p_to_texture, const Vector3 &p_from, const Vector3 &p_to, const Vector3 &p_size, uint32_t p_src_mipmap, uint32_t p_dst_mipmap, uint32_t p_src_layer, uint32_t p_dst_layer, BitField<BarrierMask> p_post_barrier) {
	_THREAD_SAFE_METHOD_

	Texture *src_tex = texture_owner.get_or_null(p_from_texture);
	ERR_FAIL_COND_V(!src_tex, ERR_INVALID_PARAMETER);

	ERR_FAIL_COND_V_MSG(src_tex->bound, ERR_INVALID_PARAMETER,
			"Source texture can't be copied while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");
	ERR_FAIL_COND_V_MSG(!(src_tex->usage_flags & TEXTURE_USAGE_CAN_COPY_FROM_BIT), ERR_INVALID_PARAMETER,
			"Source texture requires the TEXTURE_USAGE_CAN_COPY_FROM_BIT in order to be retrieved.");

	uint32_t src_layer_count = src_tex->layers;
	uint32_t src_width, src_height, src_depth;
	get_image_format_required_size(src_tex->format, src_tex->width, src_tex->height, src_tex->depth, p_src_mipmap + 1, &src_width, &src_height, &src_depth);
	if (src_tex->type == TEXTURE_TYPE_CUBE || src_tex->type == TEXTURE_TYPE_CUBE_ARRAY) {
		src_layer_count *= 6;
	}

	ERR_FAIL_COND_V(p_from.x < 0 || p_from.x + p_size.x > src_width, ERR_INVALID_PARAMETER);
	ERR_FAIL_COND_V(p_from.y < 0 || p_from.y + p_size.y > src_height, ERR_INVALID_PARAMETER);
	ERR_FAIL_COND_V(p_from.z < 0 || p_from.z + p_size.z > src_depth, ERR_INVALID_PARAMETER);
	ERR_FAIL_COND_V(p_src_mipmap >= src_tex->mipmaps, ERR_INVALID_PARAMETER);
	ERR_FAIL_COND_V(p_src_layer >= src_layer_count, ERR_INVALID_PARAMETER);

	Texture *dst_tex = texture_owner.get_or_null(p_to_texture);
	ERR_FAIL_COND_V(!dst_tex, ERR_INVALID_PARAMETER);

	ERR_FAIL_COND_V_MSG(dst_tex->bound, ERR_INVALID_PARAMETER,
			"Destination texture can't be copied while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");
	ERR_FAIL_COND_V_MSG(!(dst_tex->usage_flags & TEXTURE_USAGE_CAN_COPY_TO_BIT), ERR_INVALID_PARAMETER,
			"Destination texture requires the TEXTURE_USAGE_CAN_COPY_TO_BIT in order to be retrieved.");

	uint32_t dst_layer_count = dst_tex->layers;
	uint32_t dst_width, dst_height, dst_depth;
	get_image_format_required_size(dst_tex->format, dst_tex->width, dst_tex->height, dst_tex->depth, p_dst_mipmap + 1, &dst_width, &dst_height, &dst_depth);
	if (dst_tex->type == TEXTURE_TYPE_CUBE || dst_tex->type == TEXTURE_TYPE_CUBE_ARRAY) {
		dst_layer_count *= 6;
	}

	ERR_FAIL_COND_V(p_to.x < 0 || p_to.x + p_size.x > dst_width, ERR_INVALID_PARAMETER);
	ERR_FAIL_COND_V(p_to.y < 0 || p_to.y + p_size.y > dst_height, ERR_INVALID_PARAMETER);
	ERR_FAIL_COND_V(p_to.z < 0 || p_to.z + p_size.z > dst_depth, ERR_INVALID_PARAMETER);
	ERR_FAIL_COND_V(p_dst_mipmap >= dst_tex->mipmaps, ERR_INVALID_PARAMETER);
	ERR_FAIL_COND_V(p_dst_layer >= dst_layer_count, ERR_INVALID_PARAMETER);

	ERR_FAIL_COND_V_MSG(src_tex->read_aspect_mask != dst_tex->read_aspect_mask, ERR_INVALID_PARAMETER,
			"Source and destination texture must be of the same type (color or depth).");

	VkCommandBuffer command_buffer = frames[frame].draw_command_buffer;

	{
		// PRE Copy the image.

		{ // Source.
			VkImageMemoryBarrier image_memory_barrier;
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = 0;
			image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
			image_memory_barrier.oldLayout = src_tex->layout;
			image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;

			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = src_tex->image;
			image_memory_barrier.subresourceRange.aspectMask = src_tex->barrier_aspect_mask;
			image_memory_barrier.subresourceRange.baseMipLevel = p_src_mipmap;
			image_memory_barrier.subresourceRange.levelCount = 1;
			image_memory_barrier.subresourceRange.baseArrayLayer = p_src_layer;
			image_memory_barrier.subresourceRange.layerCount = 1;

			vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
		}
		{ // Dest.
			VkImageMemoryBarrier image_memory_barrier;
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = 0;
			image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
			image_memory_barrier.oldLayout = dst_tex->layout;
			image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;

			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = dst_tex->image;
			image_memory_barrier.subresourceRange.aspectMask = dst_tex->read_aspect_mask;
			image_memory_barrier.subresourceRange.baseMipLevel = p_dst_mipmap;
			image_memory_barrier.subresourceRange.levelCount = 1;
			image_memory_barrier.subresourceRange.baseArrayLayer = p_dst_layer;
			image_memory_barrier.subresourceRange.layerCount = 1;

			vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
		}

		// COPY.

		{
			VkImageCopy image_copy_region;
			image_copy_region.srcSubresource.aspectMask = src_tex->read_aspect_mask;
			image_copy_region.srcSubresource.baseArrayLayer = p_src_layer;
			image_copy_region.srcSubresource.layerCount = 1;
			image_copy_region.srcSubresource.mipLevel = p_src_mipmap;
			image_copy_region.srcOffset.x = p_from.x;
			image_copy_region.srcOffset.y = p_from.y;
			image_copy_region.srcOffset.z = p_from.z;

			image_copy_region.dstSubresource.aspectMask = dst_tex->read_aspect_mask;
			image_copy_region.dstSubresource.baseArrayLayer = p_dst_layer;
			image_copy_region.dstSubresource.layerCount = 1;
			image_copy_region.dstSubresource.mipLevel = p_dst_mipmap;
			image_copy_region.dstOffset.x = p_to.x;
			image_copy_region.dstOffset.y = p_to.y;
			image_copy_region.dstOffset.z = p_to.z;

			image_copy_region.extent.width = p_size.x;
			image_copy_region.extent.height = p_size.y;
			image_copy_region.extent.depth = p_size.z;

			vkCmdCopyImage(command_buffer, src_tex->image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_tex->image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &image_copy_region);
		}

		// RESTORE LAYOUT for SRC and DST.

		uint32_t barrier_flags = 0;
		uint32_t access_flags = 0;
		if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) {
			barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
			access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
		}
		if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) {
			barrier_flags |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
			access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
		}
		if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) {
			barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
			access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT;
		}

		if (barrier_flags == 0) {
			barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
		}

		{ // Restore src.
			VkImageMemoryBarrier image_memory_barrier;
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
			image_memory_barrier.dstAccessMask = access_flags;
			image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
			image_memory_barrier.newLayout = src_tex->layout;
			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = src_tex->image;
			image_memory_barrier.subresourceRange.aspectMask = src_tex->barrier_aspect_mask;
			image_memory_barrier.subresourceRange.baseMipLevel = p_src_mipmap;
			image_memory_barrier.subresourceRange.levelCount = src_tex->mipmaps;
			image_memory_barrier.subresourceRange.baseArrayLayer = p_src_layer;
			image_memory_barrier.subresourceRange.layerCount = 1;

			vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
		}

		{ // Make dst readable.

			VkImageMemoryBarrier image_memory_barrier;
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
			image_memory_barrier.dstAccessMask = access_flags;
			image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
			image_memory_barrier.newLayout = dst_tex->layout;

			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = dst_tex->image;
			image_memory_barrier.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
			image_memory_barrier.subresourceRange.baseMipLevel = p_src_mipmap;
			image_memory_barrier.subresourceRange.levelCount = 1;
			image_memory_barrier.subresourceRange.baseArrayLayer = p_src_layer;
			image_memory_barrier.subresourceRange.layerCount = 1;

			vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
		}
	}

	if (dst_tex->used_in_frame != frames_drawn) {
		dst_tex->used_in_raster = false;
		dst_tex->used_in_compute = false;
		dst_tex->used_in_frame = frames_drawn;
	}
	dst_tex->used_in_transfer = true;

	return OK;
}

Error RenderingDeviceVulkan::texture_resolve_multisample(RID p_from_texture, RID p_to_texture, BitField<BarrierMask> p_post_barrier) {
	_THREAD_SAFE_METHOD_

	Texture *src_tex = texture_owner.get_or_null(p_from_texture);
	ERR_FAIL_COND_V(!src_tex, ERR_INVALID_PARAMETER);

	ERR_FAIL_COND_V_MSG(src_tex->bound, ERR_INVALID_PARAMETER,
			"Source texture can't be copied while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");
	ERR_FAIL_COND_V_MSG(!(src_tex->usage_flags & TEXTURE_USAGE_CAN_COPY_FROM_BIT), ERR_INVALID_PARAMETER,
			"Source texture requires the TEXTURE_USAGE_CAN_COPY_FROM_BIT in order to be retrieved.");

	ERR_FAIL_COND_V_MSG(src_tex->type != TEXTURE_TYPE_2D, ERR_INVALID_PARAMETER, "Source texture must be 2D (or a slice of a 3D/Cube texture)");
	ERR_FAIL_COND_V_MSG(src_tex->samples == TEXTURE_SAMPLES_1, ERR_INVALID_PARAMETER, "Source texture must be multisampled.");

	Texture *dst_tex = texture_owner.get_or_null(p_to_texture);
	ERR_FAIL_COND_V(!dst_tex, ERR_INVALID_PARAMETER);

	ERR_FAIL_COND_V_MSG(dst_tex->bound, ERR_INVALID_PARAMETER,
			"Destination texture can't be copied while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");
	ERR_FAIL_COND_V_MSG(!(dst_tex->usage_flags & TEXTURE_USAGE_CAN_COPY_TO_BIT), ERR_INVALID_PARAMETER,
			"Destination texture requires the TEXTURE_USAGE_CAN_COPY_TO_BIT in order to be retrieved.");

	ERR_FAIL_COND_V_MSG(dst_tex->type != TEXTURE_TYPE_2D, ERR_INVALID_PARAMETER, "Destination texture must be 2D (or a slice of a 3D/Cube texture).");
	ERR_FAIL_COND_V_MSG(dst_tex->samples != TEXTURE_SAMPLES_1, ERR_INVALID_PARAMETER, "Destination texture must not be multisampled.");

	ERR_FAIL_COND_V_MSG(src_tex->format != dst_tex->format, ERR_INVALID_PARAMETER, "Source and Destination textures must be the same format.");
	ERR_FAIL_COND_V_MSG(src_tex->width != dst_tex->width && src_tex->height != dst_tex->height && src_tex->depth != dst_tex->depth, ERR_INVALID_PARAMETER, "Source and Destination textures must have the same dimensions.");

	ERR_FAIL_COND_V_MSG(src_tex->read_aspect_mask != dst_tex->read_aspect_mask, ERR_INVALID_PARAMETER,
			"Source and destination texture must be of the same type (color or depth).");

	VkCommandBuffer command_buffer = frames[frame].draw_command_buffer;

	{
		// PRE Copy the image.

		{ // Source.
			VkImageMemoryBarrier image_memory_barrier;
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = 0;
			image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
			image_memory_barrier.oldLayout = src_tex->layout;
			image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;

			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = src_tex->image;
			image_memory_barrier.subresourceRange.aspectMask = src_tex->barrier_aspect_mask;
			image_memory_barrier.subresourceRange.baseMipLevel = src_tex->base_mipmap;
			image_memory_barrier.subresourceRange.levelCount = 1;
			image_memory_barrier.subresourceRange.baseArrayLayer = src_tex->base_layer;
			image_memory_barrier.subresourceRange.layerCount = 1;

			vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
		}
		{ // Dest.
			VkImageMemoryBarrier image_memory_barrier;
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = 0;
			image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
			image_memory_barrier.oldLayout = dst_tex->layout;
			image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;

			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = dst_tex->image;
			image_memory_barrier.subresourceRange.aspectMask = dst_tex->read_aspect_mask;
			image_memory_barrier.subresourceRange.baseMipLevel = dst_tex->base_mipmap;
			image_memory_barrier.subresourceRange.levelCount = 1;
			image_memory_barrier.subresourceRange.baseArrayLayer = dst_tex->base_layer;
			image_memory_barrier.subresourceRange.layerCount = 1;

			vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
		}

		// COPY.

		{
			VkImageResolve image_copy_region;
			image_copy_region.srcSubresource.aspectMask = src_tex->read_aspect_mask;
			image_copy_region.srcSubresource.baseArrayLayer = src_tex->base_layer;
			image_copy_region.srcSubresource.layerCount = 1;
			image_copy_region.srcSubresource.mipLevel = src_tex->base_mipmap;
			image_copy_region.srcOffset.x = 0;
			image_copy_region.srcOffset.y = 0;
			image_copy_region.srcOffset.z = 0;

			image_copy_region.dstSubresource.aspectMask = dst_tex->read_aspect_mask;
			image_copy_region.dstSubresource.baseArrayLayer = dst_tex->base_layer;
			image_copy_region.dstSubresource.layerCount = 1;
			image_copy_region.dstSubresource.mipLevel = dst_tex->base_mipmap;
			image_copy_region.dstOffset.x = 0;
			image_copy_region.dstOffset.y = 0;
			image_copy_region.dstOffset.z = 0;

			image_copy_region.extent.width = src_tex->width;
			image_copy_region.extent.height = src_tex->height;
			image_copy_region.extent.depth = src_tex->depth;

			vkCmdResolveImage(command_buffer, src_tex->image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_tex->image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &image_copy_region);
		}

		// RESTORE LAYOUT for SRC and DST.

		uint32_t barrier_flags = 0;
		uint32_t access_flags = 0;
		if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) {
			barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
			access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
		}
		if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) {
			barrier_flags |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
			access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
		}
		if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) {
			barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
			access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT;
		}

		if (barrier_flags == 0) {
			barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
		}

		{ // Restore src.
			VkImageMemoryBarrier image_memory_barrier;
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_READ_BIT;
			image_memory_barrier.dstAccessMask = access_flags;
			image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
			image_memory_barrier.newLayout = src_tex->layout;
			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = src_tex->image;
			image_memory_barrier.subresourceRange.aspectMask = src_tex->barrier_aspect_mask;
			image_memory_barrier.subresourceRange.baseMipLevel = src_tex->base_mipmap;
			image_memory_barrier.subresourceRange.levelCount = 1;
			image_memory_barrier.subresourceRange.baseArrayLayer = src_tex->base_layer;
			image_memory_barrier.subresourceRange.layerCount = 1;

			vkCmdPipelineBarrier(command_buffer, VK_ACCESS_TRANSFER_WRITE_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
		}

		{ // Make dst readable.

			VkImageMemoryBarrier image_memory_barrier;
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
			image_memory_barrier.dstAccessMask = access_flags;
			image_memory_barrier.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
			image_memory_barrier.newLayout = dst_tex->layout;

			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = dst_tex->image;
			image_memory_barrier.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
			image_memory_barrier.subresourceRange.baseMipLevel = dst_tex->base_mipmap;
			image_memory_barrier.subresourceRange.levelCount = 1;
			image_memory_barrier.subresourceRange.baseArrayLayer = dst_tex->base_layer;
			image_memory_barrier.subresourceRange.layerCount = 1;

			vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
		}
	}

	return OK;
}

Error RenderingDeviceVulkan::texture_clear(RID p_texture, const Color &p_color, uint32_t p_base_mipmap, uint32_t p_mipmaps, uint32_t p_base_layer, uint32_t p_layers, BitField<BarrierMask> p_post_barrier) {
	_THREAD_SAFE_METHOD_

	Texture *src_tex = texture_owner.get_or_null(p_texture);
	ERR_FAIL_COND_V(!src_tex, ERR_INVALID_PARAMETER);

	ERR_FAIL_COND_V_MSG(src_tex->bound, ERR_INVALID_PARAMETER,
			"Source texture can't be cleared while a render pass that uses it is being created. Ensure render pass is finalized (and that it was created with RENDER_PASS_CONTENTS_FINISH) to unbind this texture.");

	ERR_FAIL_COND_V(p_layers == 0, ERR_INVALID_PARAMETER);
	ERR_FAIL_COND_V(p_mipmaps == 0, ERR_INVALID_PARAMETER);

	ERR_FAIL_COND_V_MSG(!(src_tex->usage_flags & TEXTURE_USAGE_CAN_COPY_TO_BIT), ERR_INVALID_PARAMETER,
			"Source texture requires the TEXTURE_USAGE_CAN_COPY_TO_BIT in order to be cleared.");

	uint32_t src_layer_count = src_tex->layers;
	if (src_tex->type == TEXTURE_TYPE_CUBE || src_tex->type == TEXTURE_TYPE_CUBE_ARRAY) {
		src_layer_count *= 6;
	}

	ERR_FAIL_COND_V(p_base_mipmap + p_mipmaps > src_tex->mipmaps, ERR_INVALID_PARAMETER);
	ERR_FAIL_COND_V(p_base_layer + p_layers > src_layer_count, ERR_INVALID_PARAMETER);

	VkCommandBuffer command_buffer = frames[frame].draw_command_buffer;

	VkImageLayout clear_layout = (src_tex->layout == VK_IMAGE_LAYOUT_GENERAL) ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;

	// NOTE: Perhaps the valid stages/accesses for a given owner should be a property of the owner. (Here and places like _get_buffer_from_owner.)
	const VkPipelineStageFlags valid_texture_stages = VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
	constexpr VkAccessFlags read_access = VK_ACCESS_SHADER_READ_BIT;
	constexpr VkAccessFlags read_write_access = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
	const VkAccessFlags valid_texture_access = (src_tex->usage_flags & TEXTURE_USAGE_STORAGE_BIT) ? read_write_access : read_access;

	{ // Barrier from previous access with optional layout change (see clear_layout logic above).
		VkImageMemoryBarrier image_memory_barrier;
		image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
		image_memory_barrier.pNext = nullptr;
		image_memory_barrier.srcAccessMask = valid_texture_access;
		image_memory_barrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
		image_memory_barrier.oldLayout = src_tex->layout;
		image_memory_barrier.newLayout = clear_layout;

		image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.image = src_tex->image;
		image_memory_barrier.subresourceRange.aspectMask = src_tex->read_aspect_mask;
		image_memory_barrier.subresourceRange.baseMipLevel = src_tex->base_mipmap + p_base_mipmap;
		image_memory_barrier.subresourceRange.levelCount = p_mipmaps;
		image_memory_barrier.subresourceRange.baseArrayLayer = src_tex->base_layer + p_base_layer;
		image_memory_barrier.subresourceRange.layerCount = p_layers;

		vkCmdPipelineBarrier(command_buffer, valid_texture_stages, VK_PIPELINE_STAGE_TRANSFER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
	}

	VkClearColorValue clear_color;
	clear_color.float32[0] = p_color.r;
	clear_color.float32[1] = p_color.g;
	clear_color.float32[2] = p_color.b;
	clear_color.float32[3] = p_color.a;

	VkImageSubresourceRange range;
	range.aspectMask = src_tex->read_aspect_mask;
	range.baseArrayLayer = src_tex->base_layer + p_base_layer;
	range.layerCount = p_layers;
	range.baseMipLevel = src_tex->base_mipmap + p_base_mipmap;
	range.levelCount = p_mipmaps;

	vkCmdClearColorImage(command_buffer, src_tex->image, clear_layout, &clear_color, 1, &range);

	{ // Barrier to post clear accesses (changing back the layout if needed).

		uint32_t barrier_flags = 0;
		uint32_t access_flags = 0;
		if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) {
			barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
			access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
		}
		if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) {
			barrier_flags |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
			access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
		}
		if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) {
			barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
			access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT;
		}

		if (barrier_flags == 0) {
			barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
		}

		VkImageMemoryBarrier image_memory_barrier;
		image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
		image_memory_barrier.pNext = nullptr;
		image_memory_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
		image_memory_barrier.dstAccessMask = access_flags;
		image_memory_barrier.oldLayout = clear_layout;
		image_memory_barrier.newLayout = src_tex->layout;

		image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.image = src_tex->image;
		image_memory_barrier.subresourceRange.aspectMask = src_tex->read_aspect_mask;
		image_memory_barrier.subresourceRange.baseMipLevel = src_tex->base_mipmap + p_base_mipmap;
		image_memory_barrier.subresourceRange.levelCount = p_mipmaps;
		image_memory_barrier.subresourceRange.baseArrayLayer = src_tex->base_layer + p_base_layer;
		image_memory_barrier.subresourceRange.layerCount = p_layers;

		vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);
	}

	if (src_tex->used_in_frame != frames_drawn) {
		src_tex->used_in_raster = false;
		src_tex->used_in_compute = false;
		src_tex->used_in_frame = frames_drawn;
	}
	src_tex->used_in_transfer = true;

	return OK;
}

bool RenderingDeviceVulkan::texture_is_format_supported_for_usage(DataFormat p_format, uint32_t p_usage) const {
	ERR_FAIL_INDEX_V(p_format, DATA_FORMAT_MAX, false);

	_THREAD_SAFE_METHOD_

	// Validate that this image is supported for the intended use.
	VkFormatProperties properties;
	vkGetPhysicalDeviceFormatProperties(context->get_physical_device(), vulkan_formats[p_format], &properties);
	VkFormatFeatureFlags flags;

	if (p_usage & TEXTURE_USAGE_CPU_READ_BIT) {
		flags = properties.linearTilingFeatures;
	} else {
		flags = properties.optimalTilingFeatures;
	}

	if (p_usage & TEXTURE_USAGE_SAMPLING_BIT && !(flags & VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT)) {
		return false;
	}

	if (p_usage & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT && !(flags & VK_FORMAT_FEATURE_COLOR_ATTACHMENT_BIT)) {
		return false;
	}

	if (p_usage & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT && !(flags & VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT)) {
		return false;
	}

	if (p_usage & TEXTURE_USAGE_STORAGE_BIT && !(flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT)) {
		return false;
	}

	if (p_usage & TEXTURE_USAGE_STORAGE_ATOMIC_BIT && !(flags & VK_FORMAT_FEATURE_STORAGE_IMAGE_ATOMIC_BIT)) {
		return false;
	}

	// Validation via VK_FORMAT_FEATURE_FRAGMENT_SHADING_RATE_ATTACHMENT_BIT_KHR fails if VRS attachment is not supported.
	if (p_usage & TEXTURE_USAGE_VRS_ATTACHMENT_BIT && p_format != DATA_FORMAT_R8_UINT) {
		return false;
	}

	return true;
}

/********************/
/**** ATTACHMENT ****/
/********************/

VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentFormat> &p_attachments, const Vector<FramebufferPass> &p_passes, InitialAction p_initial_action, FinalAction p_final_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, uint32_t p_view_count, Vector<TextureSamples> *r_samples) {
	// Set up dependencies from/to external equivalent to the default (implicit) one, and then amend them.
	const VkPipelineStageFlags default_access_mask = VK_ACCESS_INPUT_ATTACHMENT_READ_BIT |
			VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
			VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
			VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
			VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | // From Section 7.1 of Vulkan API Spec v1.1.148.
			VK_ACCESS_FRAGMENT_SHADING_RATE_ATTACHMENT_READ_BIT_KHR;

	VkPipelineStageFlags reading_stages = VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_TRANSFER_BIT;
	VkSubpassDependency2KHR dependencies[2] = {
		{ VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_2_KHR, nullptr, VK_SUBPASS_EXTERNAL, 0, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT, 0, default_access_mask, 0, 0 },
		{ VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_2_KHR, nullptr, 0, VK_SUBPASS_EXTERNAL, VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, default_access_mask, 0, 0, 0 }
	};
	VkSubpassDependency2KHR &dependency_from_external = dependencies[0];
	VkSubpassDependency2KHR &dependency_to_external = dependencies[1];
	LocalVector<int32_t> attachment_last_pass;
	attachment_last_pass.resize(p_attachments.size());

	if (p_view_count > 1) {
		const VulkanContext::MultiviewCapabilities capabilities = context->get_multiview_capabilities();

		// This only works with multiview!
		ERR_FAIL_COND_V_MSG(!capabilities.is_supported, VK_NULL_HANDLE, "Multiview not supported");

		// Make sure we limit this to the number of views we support.
		ERR_FAIL_COND_V_MSG(p_view_count > capabilities.max_view_count, VK_NULL_HANDLE, "Hardware does not support requested number of views for Multiview render pass");
	}

	// These are only used if we use multiview but we need to define them in scope.
	const uint32_t view_mask = (1 << p_view_count) - 1;
	const uint32_t correlation_mask = (1 << p_view_count) - 1;

	Vector<VkAttachmentDescription2KHR> attachments;
	Vector<int> attachment_remap;

	for (int i = 0; i < p_attachments.size(); i++) {
		if (p_attachments[i].usage_flags == AttachmentFormat::UNUSED_ATTACHMENT) {
			attachment_remap.push_back(VK_ATTACHMENT_UNUSED);
			continue;
		}

		ERR_FAIL_INDEX_V(p_attachments[i].format, DATA_FORMAT_MAX, VK_NULL_HANDLE);
		ERR_FAIL_INDEX_V(p_attachments[i].samples, TEXTURE_SAMPLES_MAX, VK_NULL_HANDLE);
		ERR_FAIL_COND_V_MSG(!(p_attachments[i].usage_flags & (TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | TEXTURE_USAGE_INPUT_ATTACHMENT_BIT | TEXTURE_USAGE_VRS_ATTACHMENT_BIT)),
				VK_NULL_HANDLE, "Texture format for index (" + itos(i) + ") requires an attachment (color, depth-stencil, input or VRS) bit set.");

		VkAttachmentDescription2KHR description = {};
		description.sType = VK_STRUCTURE_TYPE_ATTACHMENT_DESCRIPTION_2_KHR;
		description.pNext = nullptr;
		description.flags = 0;
		description.format = vulkan_formats[p_attachments[i].format];
		description.samples = _ensure_supported_sample_count(p_attachments[i].samples);

		bool is_sampled = p_attachments[i].usage_flags & TEXTURE_USAGE_SAMPLING_BIT;
		bool is_storage = p_attachments[i].usage_flags & TEXTURE_USAGE_STORAGE_BIT;
		bool is_depth = p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;

		// We can setup a framebuffer where we write to our VRS texture to set it up.
		// We make the assumption here that if our texture is actually used as our VRS attachment.
		// It is used as such for each subpass. This is fairly certain seeing the restrictions on subpasses.
		bool is_vrs = p_attachments[i].usage_flags & TEXTURE_USAGE_VRS_ATTACHMENT_BIT && i == p_passes[0].vrs_attachment;

		if (is_vrs) {
			// For VRS we only read, there is no writing to this texture.
			description.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
			description.initialLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
			description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
		} else {
			// For each UNDEFINED, assume the prior use was a *read*, as we'd be discarding the output of a write.
			// Also, each UNDEFINED will do an immediate layout transition (write), s.t. we must ensure execution synchronization vs
			// the read. If this is a performance issue, one could track the actual last accessor of each resource, adding only that
			// stage.

			switch (is_depth ? p_initial_depth_action : p_initial_action) {
				case INITIAL_ACTION_CLEAR_REGION:
				case INITIAL_ACTION_CLEAR: {
					if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR;
						description.initialLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL);
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
					} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR;
						description.initialLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL);
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_CLEAR;
						dependency_from_external.srcStageMask |= reading_stages;
					} else {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there.
						dependency_from_external.srcStageMask |= reading_stages;
					}
				} break;
				case INITIAL_ACTION_KEEP: {
					if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
						description.initialLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL);
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
					} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
						description.initialLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL);
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
						dependency_from_external.srcStageMask |= reading_stages;
					} else {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there.
						dependency_from_external.srcStageMask |= reading_stages;
					}
				} break;
				case INITIAL_ACTION_DROP: {
					if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						description.initialLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL);
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
					} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there.
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						dependency_from_external.srcStageMask |= reading_stages;
					} else {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there.
						dependency_from_external.srcStageMask |= reading_stages;
					}
				} break;
				case INITIAL_ACTION_CLEAR_REGION_CONTINUE:
				case INITIAL_ACTION_CONTINUE: {
					if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
						description.initialLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
					} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
						description.initialLayout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD;
					} else {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there.
						dependency_from_external.srcStageMask |= reading_stages;
					}
				} break;
				default: {
					ERR_FAIL_V(VK_NULL_HANDLE); // Should never reach here.
				}
			}
		}

		bool used_last = false;

		{
			int last_pass = p_passes.size() - 1;

			if (is_depth) {
				// Likely missing depth resolve?
				if (p_passes[last_pass].depth_attachment == i) {
					used_last = true;
				}
			} else if (is_vrs) {
				if (p_passes[last_pass].vrs_attachment == i) {
					used_last = true;
				}
			} else {
				if (p_passes[last_pass].resolve_attachments.size()) {
					// If using resolve attachments, check resolve attachments.
					for (int j = 0; j < p_passes[last_pass].resolve_attachments.size(); j++) {
						if (p_passes[last_pass].resolve_attachments[j] == i) {
							used_last = true;
							break;
						}
					}
				}
				if (!used_last) {
					for (int j = 0; j < p_passes[last_pass].color_attachments.size(); j++) {
						if (p_passes[last_pass].color_attachments[j] == i) {
							used_last = true;
							break;
						}
					}
				}
			}

			if (!used_last) {
				for (int j = 0; j < p_passes[last_pass].preserve_attachments.size(); j++) {
					if (p_passes[last_pass].preserve_attachments[j] == i) {
						used_last = true;
						break;
					}
				}
			}
		}

		FinalAction final_action = p_final_action;
		FinalAction final_depth_action = p_final_depth_action;

		if (!used_last) {
			if (is_depth) {
				final_depth_action = FINAL_ACTION_DISCARD;

			} else {
				final_action = FINAL_ACTION_DISCARD;
			}
		}

		if (is_vrs) {
			// We don't change our VRS texture during this process.

			description.storeOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
			description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
			description.finalLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;

			// TODO: Do we need to update our external dependency?
			// update_external_dependency_for_store(dependency_to_external, is_sampled, is_storage, false);
		} else {
			switch (is_depth ? final_depth_action : final_action) {
				case FINAL_ACTION_READ: {
					if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
						description.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
						description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
						description.finalLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL);
						update_external_dependency_for_store(dependency_to_external, is_sampled, is_storage, false);
					} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
						description.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
						description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_STORE;
						description.finalLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL);
						update_external_dependency_for_store(dependency_to_external, is_sampled, is_storage, true);
					} else {
						description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE;
						description.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there.
						// TODO: What does this mean about the next usage (and thus appropriate dependency masks.
					}
				} break;
				case FINAL_ACTION_DISCARD: {
					if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
						description.storeOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
						description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
						description.finalLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL);
					} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
						description.storeOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
						description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
						description.finalLayout = is_sampled ? VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL : (is_storage ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL);
					} else {
						description.storeOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
						description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
						description.finalLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there.
					}
				} break;
				case FINAL_ACTION_CONTINUE: {
					if (p_attachments[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
						description.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
						description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
						description.finalLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
					} else if (p_attachments[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
						description.storeOp = VK_ATTACHMENT_STORE_OP_STORE;
						description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_STORE;
						description.finalLayout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
					} else {
						description.storeOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
						description.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE;
						description.finalLayout = VK_IMAGE_LAYOUT_UNDEFINED; // Don't care what is there.
					}

				} break;
				default: {
					ERR_FAIL_V(VK_NULL_HANDLE); // Should never reach here.
				}
			}
		}

		attachment_last_pass[i] = -1;
		attachment_remap.push_back(attachments.size());
		attachments.push_back(description);
	}

	LocalVector<VkSubpassDescription2KHR> subpasses;
	LocalVector<LocalVector<VkAttachmentReference2KHR>> color_reference_array;
	LocalVector<LocalVector<VkAttachmentReference2KHR>> input_reference_array;
	LocalVector<LocalVector<VkAttachmentReference2KHR>> resolve_reference_array;
	LocalVector<LocalVector<uint32_t>> preserve_reference_array;
	LocalVector<VkAttachmentReference2KHR> depth_reference_array;
	LocalVector<VkAttachmentReference2KHR> vrs_reference_array;
	LocalVector<VkFragmentShadingRateAttachmentInfoKHR> vrs_attachment_info_array;

	subpasses.resize(p_passes.size());
	color_reference_array.resize(p_passes.size());
	input_reference_array.resize(p_passes.size());
	resolve_reference_array.resize(p_passes.size());
	preserve_reference_array.resize(p_passes.size());
	depth_reference_array.resize(p_passes.size());
	vrs_reference_array.resize(p_passes.size());
	vrs_attachment_info_array.resize(p_passes.size());

	LocalVector<VkSubpassDependency2KHR> subpass_dependencies;

	for (int i = 0; i < p_passes.size(); i++) {
		const FramebufferPass *pass = &p_passes[i];

		LocalVector<VkAttachmentReference2KHR> &color_references = color_reference_array[i];

		TextureSamples texture_samples = TEXTURE_SAMPLES_1;
		bool is_multisample_first = true;
		void *subpass_nextptr = nullptr;

		for (int j = 0; j < pass->color_attachments.size(); j++) {
			int32_t attachment = pass->color_attachments[j];
			VkAttachmentReference2KHR reference;
			reference.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR;
			reference.pNext = nullptr;
			if (attachment == FramebufferPass::ATTACHMENT_UNUSED) {
				reference.attachment = VK_ATTACHMENT_UNUSED;
				reference.layout = VK_IMAGE_LAYOUT_UNDEFINED;
			} else {
				ERR_FAIL_INDEX_V_MSG(attachment, p_attachments.size(), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), color attachment (" + itos(j) + ").");
				ERR_FAIL_COND_V_MSG(!(p_attachments[attachment].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it's marked as depth, but it's not usable as color attachment.");
				ERR_FAIL_COND_V_MSG(attachment_last_pass[attachment] == i, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it already was used for something else before in this pass.");

				if (is_multisample_first) {
					texture_samples = p_attachments[attachment].samples;
					is_multisample_first = false;
				} else {
					ERR_FAIL_COND_V_MSG(texture_samples != p_attachments[attachment].samples, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), if an attachment is marked as multisample, all of them should be multisample and use the same number of samples.");
				}
				reference.attachment = attachment_remap[attachment];
				reference.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
				attachment_last_pass[attachment] = i;
			}
			reference.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
			color_references.push_back(reference);
		}

		LocalVector<VkAttachmentReference2KHR> &input_references = input_reference_array[i];

		for (int j = 0; j < pass->input_attachments.size(); j++) {
			int32_t attachment = pass->input_attachments[j];
			VkAttachmentReference2KHR reference;
			reference.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR;
			reference.pNext = nullptr;
			if (attachment == FramebufferPass::ATTACHMENT_UNUSED) {
				reference.attachment = VK_ATTACHMENT_UNUSED;
				reference.layout = VK_IMAGE_LAYOUT_UNDEFINED;
			} else {
				ERR_FAIL_INDEX_V_MSG(attachment, p_attachments.size(), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), input attachment (" + itos(j) + ").");
				ERR_FAIL_COND_V_MSG(!(p_attachments[attachment].usage_flags & TEXTURE_USAGE_INPUT_ATTACHMENT_BIT), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it isn't marked as an input texture.");
				ERR_FAIL_COND_V_MSG(attachment_last_pass[attachment] == i, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it already was used for something else before in this pass.");
				reference.attachment = attachment_remap[attachment];
				reference.layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
				attachment_last_pass[attachment] = i;
			}
			reference.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
			input_references.push_back(reference);
		}

		LocalVector<VkAttachmentReference2KHR> &resolve_references = resolve_reference_array[i];

		if (pass->resolve_attachments.size() > 0) {
			ERR_FAIL_COND_V_MSG(pass->resolve_attachments.size() != pass->color_attachments.size(), VK_NULL_HANDLE, "The amount of resolve attachments (" + itos(pass->resolve_attachments.size()) + ") must match the number of color attachments (" + itos(pass->color_attachments.size()) + ").");
			ERR_FAIL_COND_V_MSG(texture_samples == TEXTURE_SAMPLES_1, VK_NULL_HANDLE, "Resolve attachments specified, but color attachments are not multisample.");
		}
		for (int j = 0; j < pass->resolve_attachments.size(); j++) {
			int32_t attachment = pass->resolve_attachments[j];
			VkAttachmentReference2KHR reference;
			reference.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR;
			reference.pNext = nullptr;
			if (attachment == FramebufferPass::ATTACHMENT_UNUSED) {
				reference.attachment = VK_ATTACHMENT_UNUSED;
				reference.layout = VK_IMAGE_LAYOUT_UNDEFINED;
			} else {
				ERR_FAIL_INDEX_V_MSG(attachment, p_attachments.size(), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), resolve attachment (" + itos(j) + ").");
				ERR_FAIL_COND_V_MSG(pass->color_attachments[j] == FramebufferPass::ATTACHMENT_UNUSED, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), resolve attachment (" + itos(j) + "), the respective color attachment is marked as unused.");
				ERR_FAIL_COND_V_MSG(!(p_attachments[attachment].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), resolve attachment, it isn't marked as a color texture.");
				ERR_FAIL_COND_V_MSG(attachment_last_pass[attachment] == i, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it already was used for something else before in this pass.");
				bool multisample = p_attachments[attachment].samples > TEXTURE_SAMPLES_1;
				ERR_FAIL_COND_V_MSG(multisample, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), resolve attachments can't be multisample.");
				reference.attachment = attachment_remap[attachment];
				reference.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; // VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL
				attachment_last_pass[attachment] = i;
			}
			reference.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
			resolve_references.push_back(reference);
		}

		VkAttachmentReference2KHR &depth_stencil_reference = depth_reference_array[i];
		depth_stencil_reference.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR;
		depth_stencil_reference.pNext = nullptr;

		if (pass->depth_attachment != FramebufferPass::ATTACHMENT_UNUSED) {
			int32_t attachment = pass->depth_attachment;
			ERR_FAIL_INDEX_V_MSG(attachment, p_attachments.size(), VK_NULL_HANDLE, "Invalid framebuffer depth format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), depth attachment.");
			ERR_FAIL_COND_V_MSG(!(p_attachments[attachment].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT), VK_NULL_HANDLE, "Invalid framebuffer depth format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it's marked as depth, but it's not a depth attachment.");
			ERR_FAIL_COND_V_MSG(attachment_last_pass[attachment] == i, VK_NULL_HANDLE, "Invalid framebuffer depth format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it already was used for something else before in this pass.");
			depth_stencil_reference.attachment = attachment_remap[attachment];
			depth_stencil_reference.layout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
			depth_stencil_reference.aspectMask = VK_IMAGE_ASPECT_NONE;
			attachment_last_pass[attachment] = i;

			if (is_multisample_first) {
				texture_samples = p_attachments[attachment].samples;
				is_multisample_first = false;
			} else {
				ERR_FAIL_COND_V_MSG(texture_samples != p_attachments[attachment].samples, VK_NULL_HANDLE, "Invalid framebuffer depth format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), if an attachment is marked as multisample, all of them should be multisample and use the same number of samples including the depth.");
			}

		} else {
			depth_stencil_reference.attachment = VK_ATTACHMENT_UNUSED;
			depth_stencil_reference.layout = VK_IMAGE_LAYOUT_UNDEFINED;
		}

		if (context->get_vrs_capabilities().attachment_vrs_supported && pass->vrs_attachment != FramebufferPass::ATTACHMENT_UNUSED) {
			int32_t attachment = pass->vrs_attachment;
			ERR_FAIL_INDEX_V_MSG(attachment, p_attachments.size(), VK_NULL_HANDLE, "Invalid framebuffer VRS format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), VRS attachment.");
			ERR_FAIL_COND_V_MSG(!(p_attachments[attachment].usage_flags & TEXTURE_USAGE_VRS_ATTACHMENT_BIT), VK_NULL_HANDLE, "Invalid framebuffer VRS format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it's marked as VRS, but it's not a VRS attachment.");
			ERR_FAIL_COND_V_MSG(attachment_last_pass[attachment] == i, VK_NULL_HANDLE, "Invalid framebuffer VRS attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), it already was used for something else before in this pass.");

			VkAttachmentReference2KHR &vrs_reference = vrs_reference_array[i];
			vrs_reference.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR;
			vrs_reference.pNext = nullptr;
			vrs_reference.attachment = attachment_remap[attachment];
			vrs_reference.layout = VK_IMAGE_LAYOUT_FRAGMENT_SHADING_RATE_ATTACHMENT_OPTIMAL_KHR;
			vrs_reference.aspectMask = VK_IMAGE_ASPECT_NONE;

			Size2i texel_size = context->get_vrs_capabilities().texel_size;

			VkFragmentShadingRateAttachmentInfoKHR &vrs_attachment_info = vrs_attachment_info_array[i];
			vrs_attachment_info.sType = VK_STRUCTURE_TYPE_FRAGMENT_SHADING_RATE_ATTACHMENT_INFO_KHR;
			vrs_attachment_info.pNext = nullptr;
			vrs_attachment_info.pFragmentShadingRateAttachment = &vrs_reference;
			vrs_attachment_info.shadingRateAttachmentTexelSize = { uint32_t(texel_size.x), uint32_t(texel_size.y) };

			attachment_last_pass[attachment] = i;

			subpass_nextptr = &vrs_attachment_info;
		}

		LocalVector<uint32_t> &preserve_references = preserve_reference_array[i];

		for (int j = 0; j < pass->preserve_attachments.size(); j++) {
			int32_t attachment = pass->preserve_attachments[j];

			ERR_FAIL_COND_V_MSG(attachment == FramebufferPass::ATTACHMENT_UNUSED, VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), preserve attachment (" + itos(j) + "). Preserve attachments can't be unused.");

			ERR_FAIL_INDEX_V_MSG(attachment, p_attachments.size(), VK_NULL_HANDLE, "Invalid framebuffer format attachment(" + itos(attachment) + "), in pass (" + itos(i) + "), preserve attachment (" + itos(j) + ").");

			if (attachment_last_pass[attachment] != i) {
				// Preserve can still be used to keep depth or color from being discarded after use.
				attachment_last_pass[attachment] = i;
				preserve_references.push_back(attachment);
			}
		}

		VkSubpassDescription2KHR &subpass = subpasses[i];
		subpass.sType = VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_2_KHR;
		subpass.pNext = subpass_nextptr;
		subpass.flags = 0;
		subpass.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
		if (p_view_count == 1) {
			// VUID-VkSubpassDescription2-multiview-06558: If the multiview feature is not enabled, viewMask must be 0.
			subpass.viewMask = 0;
		} else {
			subpass.viewMask = view_mask;
		}
		subpass.inputAttachmentCount = input_references.size();
		if (input_references.size()) {
			subpass.pInputAttachments = input_references.ptr();
		} else {
			subpass.pInputAttachments = nullptr;
		}
		subpass.colorAttachmentCount = color_references.size();
		if (color_references.size()) {
			subpass.pColorAttachments = color_references.ptr();
		} else {
			subpass.pColorAttachments = nullptr;
		}
		if (depth_stencil_reference.attachment != VK_ATTACHMENT_UNUSED) {
			subpass.pDepthStencilAttachment = &depth_stencil_reference;
		} else {
			subpass.pDepthStencilAttachment = nullptr;
		}

		if (resolve_references.size()) {
			subpass.pResolveAttachments = resolve_references.ptr();
		} else {
			subpass.pResolveAttachments = nullptr;
		}

		subpass.preserveAttachmentCount = preserve_references.size();
		if (preserve_references.size()) {
			subpass.pPreserveAttachments = preserve_references.ptr();
		} else {
			subpass.pPreserveAttachments = nullptr;
		}

		if (r_samples) {
			r_samples->push_back(texture_samples);
		}

		if (i > 0) {
			VkSubpassDependency2KHR dependency;
			dependency.sType = VK_STRUCTURE_TYPE_SUBPASS_DEPENDENCY_2_KHR;
			dependency.pNext = nullptr;
			dependency.srcSubpass = i - 1;
			dependency.dstSubpass = i;
			dependency.srcStageMask = 0;
			dependency.srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
			dependency.dstStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;

			dependency.srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
			dependency.dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT | VK_ACCESS_INPUT_ATTACHMENT_READ_BIT;
			dependency.dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT;
			dependency.viewOffset = 0;
			subpass_dependencies.push_back(dependency);
		}
		/*
		// NOTE: Big Mallet Approach -- any layout transition causes a full barrier.
		if (reference.layout != description.initialLayout) {
			// NOTE: This should be smarter based on the texture's knowledge of its previous role.
			dependency_from_external.srcStageMask |= VK_PIPELINE_STAGE_ALL_COMMANDS_BIT;
			dependency_from_external.srcAccessMask |= VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT;
		}
		if (reference.layout != description.finalLayout) {
			// NOTE: This should be smarter based on the texture's knowledge of its subsequent role.
			dependency_to_external.dstStageMask |= VK_PIPELINE_STAGE_ALL_COMMANDS_BIT;
			dependency_to_external.dstAccessMask |= VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT;
		}
		*/
	}

	VkRenderPassCreateInfo2KHR render_pass_create_info;
	render_pass_create_info.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO_2_KHR;
	render_pass_create_info.pNext = nullptr;
	render_pass_create_info.flags = 0;

	render_pass_create_info.attachmentCount = attachments.size();
	render_pass_create_info.pAttachments = attachments.ptr();
	render_pass_create_info.subpassCount = subpasses.size();
	render_pass_create_info.pSubpasses = subpasses.ptr();
	// Commenting this because it seems it just avoids raster and compute to work at the same time.
	// Other barriers seem to be protecting the render pass fine.
	//	render_pass_create_info.dependencyCount = 2;
	//	render_pass_create_info.pDependencies = dependencies;

	render_pass_create_info.dependencyCount = subpass_dependencies.size();
	if (subpass_dependencies.size()) {
		render_pass_create_info.pDependencies = subpass_dependencies.ptr();
	} else {
		render_pass_create_info.pDependencies = nullptr;
	}

	if (p_view_count == 1) {
		// VUID-VkRenderPassCreateInfo2-viewMask-03057: If the VkSubpassDescription2::viewMask member of all elements of pSubpasses is 0, correlatedViewMaskCount must be 0.
		render_pass_create_info.correlatedViewMaskCount = 0;
		render_pass_create_info.pCorrelatedViewMasks = nullptr;
	} else {
		render_pass_create_info.correlatedViewMaskCount = 1;
		render_pass_create_info.pCorrelatedViewMasks = &correlation_mask;
	}

	Vector<uint32_t> view_masks;
	VkRenderPassMultiviewCreateInfo render_pass_multiview_create_info;

	if ((p_view_count > 1) && !context->supports_renderpass2()) {
		// This is only required when using vkCreateRenderPass, we add it if vkCreateRenderPass2KHR is not supported
		// resulting this in being passed to our vkCreateRenderPass fallback.

		// Set view masks for each subpass.
		for (uint32_t i = 0; i < subpasses.size(); i++) {
			view_masks.push_back(view_mask);
		}

		render_pass_multiview_create_info.sType = VK_STRUCTURE_TYPE_RENDER_PASS_MULTIVIEW_CREATE_INFO;
		render_pass_multiview_create_info.pNext = nullptr;
		render_pass_multiview_create_info.subpassCount = subpasses.size();
		render_pass_multiview_create_info.pViewMasks = view_masks.ptr();
		render_pass_multiview_create_info.dependencyCount = 0;
		render_pass_multiview_create_info.pViewOffsets = nullptr;
		render_pass_multiview_create_info.correlationMaskCount = 1;
		render_pass_multiview_create_info.pCorrelationMasks = &correlation_mask;

		render_pass_create_info.pNext = &render_pass_multiview_create_info;
	}

	VkRenderPass render_pass;
	VkResult res = context->vkCreateRenderPass2KHR(device, &render_pass_create_info, nullptr, &render_pass);
	ERR_FAIL_COND_V_MSG(res, VK_NULL_HANDLE, "vkCreateRenderPass2KHR failed with error " + itos(res) + ".");

	return render_pass;
}

RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_create(const Vector<AttachmentFormat> &p_format, uint32_t p_view_count) {
	FramebufferPass pass;
	for (int i = 0; i < p_format.size(); i++) {
		if (p_format[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
			pass.depth_attachment = i;
		} else {
			pass.color_attachments.push_back(i);
		}
	}

	Vector<FramebufferPass> passes;
	passes.push_back(pass);
	return framebuffer_format_create_multipass(p_format, passes, p_view_count);
}
RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_create_multipass(const Vector<AttachmentFormat> &p_attachments, const Vector<FramebufferPass> &p_passes, uint32_t p_view_count) {
	_THREAD_SAFE_METHOD_

	FramebufferFormatKey key;
	key.attachments = p_attachments;
	key.passes = p_passes;
	key.view_count = p_view_count;

	const RBMap<FramebufferFormatKey, FramebufferFormatID>::Element *E = framebuffer_format_cache.find(key);
	if (E) {
		// Exists, return.
		return E->get();
	}

	Vector<TextureSamples> samples;
	VkRenderPass render_pass = _render_pass_create(p_attachments, p_passes, INITIAL_ACTION_CLEAR, FINAL_ACTION_READ, INITIAL_ACTION_CLEAR, FINAL_ACTION_READ, p_view_count, &samples); // Actions don't matter for this use case.

	if (render_pass == VK_NULL_HANDLE) { // Was likely invalid.
		return INVALID_ID;
	}
	FramebufferFormatID id = FramebufferFormatID(framebuffer_format_cache.size()) | (FramebufferFormatID(ID_TYPE_FRAMEBUFFER_FORMAT) << FramebufferFormatID(ID_BASE_SHIFT));

	E = framebuffer_format_cache.insert(key, id);
	FramebufferFormat fb_format;
	fb_format.E = E;
	fb_format.render_pass = render_pass;
	fb_format.pass_samples = samples;
	fb_format.view_count = p_view_count;
	framebuffer_formats[id] = fb_format;
	return id;
}

RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_format_create_empty(TextureSamples p_samples) {
	FramebufferFormatKey key;
	key.passes.push_back(FramebufferPass());

	const RBMap<FramebufferFormatKey, FramebufferFormatID>::Element *E = framebuffer_format_cache.find(key);
	if (E) {
		// Exists, return.
		return E->get();
	}

	VkSubpassDescription2KHR subpass;
	subpass.sType = VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_2_KHR;
	subpass.pNext = nullptr;
	subpass.flags = 0;
	subpass.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS;
	subpass.viewMask = 0;
	subpass.inputAttachmentCount = 0; // Unsupported for now.
	subpass.pInputAttachments = nullptr;
	subpass.colorAttachmentCount = 0;
	subpass.pColorAttachments = nullptr;
	subpass.pDepthStencilAttachment = nullptr;
	subpass.pResolveAttachments = nullptr;
	subpass.preserveAttachmentCount = 0;
	subpass.pPreserveAttachments = nullptr;

	VkRenderPassCreateInfo2KHR render_pass_create_info;
	render_pass_create_info.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO_2_KHR;
	render_pass_create_info.pNext = nullptr;
	render_pass_create_info.flags = 0;
	render_pass_create_info.attachmentCount = 0;
	render_pass_create_info.pAttachments = nullptr;
	render_pass_create_info.subpassCount = 1;
	render_pass_create_info.pSubpasses = &subpass;
	render_pass_create_info.dependencyCount = 0;
	render_pass_create_info.pDependencies = nullptr;
	render_pass_create_info.correlatedViewMaskCount = 0;
	render_pass_create_info.pCorrelatedViewMasks = nullptr;

	VkRenderPass render_pass;
	VkResult res = context->vkCreateRenderPass2KHR(device, &render_pass_create_info, nullptr, &render_pass);

	ERR_FAIL_COND_V_MSG(res, 0, "vkCreateRenderPass2KHR for empty fb failed with error " + itos(res) + ".");

	if (render_pass == VK_NULL_HANDLE) { // Was likely invalid.
		return INVALID_ID;
	}

	FramebufferFormatID id = FramebufferFormatID(framebuffer_format_cache.size()) | (FramebufferFormatID(ID_TYPE_FRAMEBUFFER_FORMAT) << FramebufferFormatID(ID_BASE_SHIFT));

	E = framebuffer_format_cache.insert(key, id);

	FramebufferFormat fb_format;
	fb_format.E = E;
	fb_format.render_pass = render_pass;
	fb_format.pass_samples.push_back(p_samples);
	framebuffer_formats[id] = fb_format;
	return id;
}

RenderingDevice::TextureSamples RenderingDeviceVulkan::framebuffer_format_get_texture_samples(FramebufferFormatID p_format, uint32_t p_pass) {
	HashMap<FramebufferFormatID, FramebufferFormat>::Iterator E = framebuffer_formats.find(p_format);
	ERR_FAIL_COND_V(!E, TEXTURE_SAMPLES_1);
	ERR_FAIL_COND_V(p_pass >= uint32_t(E->value.pass_samples.size()), TEXTURE_SAMPLES_1);

	return E->value.pass_samples[p_pass];
}

/***********************/
/**** RENDER TARGET ****/
/***********************/

RID RenderingDeviceVulkan::framebuffer_create_empty(const Size2i &p_size, TextureSamples p_samples, FramebufferFormatID p_format_check) {
	_THREAD_SAFE_METHOD_
	Framebuffer framebuffer;
	framebuffer.format_id = framebuffer_format_create_empty(p_samples);
	ERR_FAIL_COND_V(p_format_check != INVALID_FORMAT_ID && framebuffer.format_id != p_format_check, RID());
	framebuffer.size = p_size;
	framebuffer.view_count = 1;

	RID id = framebuffer_owner.make_rid(framebuffer);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif
	return id;
}

RID RenderingDeviceVulkan::framebuffer_create(const Vector<RID> &p_texture_attachments, FramebufferFormatID p_format_check, uint32_t p_view_count) {
	_THREAD_SAFE_METHOD_

	FramebufferPass pass;

	for (int i = 0; i < p_texture_attachments.size(); i++) {
		Texture *texture = texture_owner.get_or_null(p_texture_attachments[i]);

		ERR_FAIL_COND_V_MSG(texture && texture->layers != p_view_count, RID(), "Layers of our texture doesn't match view count for this framebuffer");

		if (texture && texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
			pass.depth_attachment = i;
		} else if (texture && texture->usage_flags & TEXTURE_USAGE_VRS_ATTACHMENT_BIT) {
			pass.vrs_attachment = i;
		} else {
			if (texture && texture->is_resolve_buffer) {
				pass.resolve_attachments.push_back(i);
			} else {
				pass.color_attachments.push_back(texture ? i : FramebufferPass::ATTACHMENT_UNUSED);
			}
		}
	}

	Vector<FramebufferPass> passes;
	passes.push_back(pass);

	return framebuffer_create_multipass(p_texture_attachments, passes, p_format_check, p_view_count);
}

RID RenderingDeviceVulkan::framebuffer_create_multipass(const Vector<RID> &p_texture_attachments, const Vector<FramebufferPass> &p_passes, FramebufferFormatID p_format_check, uint32_t p_view_count) {
	_THREAD_SAFE_METHOD_

	Vector<AttachmentFormat> attachments;
	attachments.resize(p_texture_attachments.size());
	Size2i size;
	bool size_set = false;
	for (int i = 0; i < p_texture_attachments.size(); i++) {
		AttachmentFormat af;
		Texture *texture = texture_owner.get_or_null(p_texture_attachments[i]);
		if (!texture) {
			af.usage_flags = AttachmentFormat::UNUSED_ATTACHMENT;
		} else {
			ERR_FAIL_COND_V_MSG(texture->layers != p_view_count, RID(), "Layers of our texture doesn't match view count for this framebuffer");

			if (!size_set) {
				size.width = texture->width;
				size.height = texture->height;
				size_set = true;
			} else if (texture->usage_flags & TEXTURE_USAGE_VRS_ATTACHMENT_BIT) {
				// If this is not the first attachment we assume this is used as the VRS attachment.
				// In this case this texture will be 1/16th the size of the color attachment.
				// So we skip the size check.
			} else {
				ERR_FAIL_COND_V_MSG((uint32_t)size.width != texture->width || (uint32_t)size.height != texture->height, RID(),
						"All textures in a framebuffer should be the same size.");
			}

			af.format = texture->format;
			af.samples = texture->samples;
			af.usage_flags = texture->usage_flags;
		}
		attachments.write[i] = af;
	}

	ERR_FAIL_COND_V_MSG(!size_set, RID(), "All attachments unused.");

	FramebufferFormatID format_id = framebuffer_format_create_multipass(attachments, p_passes, p_view_count);
	if (format_id == INVALID_ID) {
		return RID();
	}

	ERR_FAIL_COND_V_MSG(p_format_check != INVALID_ID && format_id != p_format_check, RID(),
			"The format used to check this framebuffer differs from the intended framebuffer format.");

	Framebuffer framebuffer;
	framebuffer.format_id = format_id;
	framebuffer.texture_ids = p_texture_attachments;
	framebuffer.size = size;
	framebuffer.view_count = p_view_count;

	RID id = framebuffer_owner.make_rid(framebuffer);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif

	for (int i = 0; i < p_texture_attachments.size(); i++) {
		if (p_texture_attachments[i].is_valid()) {
			_add_dependency(id, p_texture_attachments[i]);
		}
	}

	return id;
}

RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::framebuffer_get_format(RID p_framebuffer) {
	_THREAD_SAFE_METHOD_

	Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_framebuffer);
	ERR_FAIL_COND_V(!framebuffer, INVALID_ID);

	return framebuffer->format_id;
}

bool RenderingDeviceVulkan::framebuffer_is_valid(RID p_framebuffer) const {
	_THREAD_SAFE_METHOD_

	return framebuffer_owner.owns(p_framebuffer);
}

void RenderingDeviceVulkan::framebuffer_set_invalidation_callback(RID p_framebuffer, InvalidationCallback p_callback, void *p_userdata) {
	_THREAD_SAFE_METHOD_

	Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_framebuffer);
	ERR_FAIL_COND(!framebuffer);

	framebuffer->invalidated_callback = p_callback;
	framebuffer->invalidated_callback_userdata = p_userdata;
}

/*****************/
/**** SAMPLER ****/
/*****************/

RID RenderingDeviceVulkan::sampler_create(const SamplerState &p_state) {
	_THREAD_SAFE_METHOD_

	VkSamplerCreateInfo sampler_create_info;
	sampler_create_info.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO;
	sampler_create_info.pNext = nullptr;
	sampler_create_info.flags = 0;
	sampler_create_info.magFilter = p_state.mag_filter == SAMPLER_FILTER_LINEAR ? VK_FILTER_LINEAR : VK_FILTER_NEAREST;
	sampler_create_info.minFilter = p_state.min_filter == SAMPLER_FILTER_LINEAR ? VK_FILTER_LINEAR : VK_FILTER_NEAREST;
	sampler_create_info.mipmapMode = p_state.mip_filter == SAMPLER_FILTER_LINEAR ? VK_SAMPLER_MIPMAP_MODE_LINEAR : VK_SAMPLER_MIPMAP_MODE_NEAREST;

	ERR_FAIL_INDEX_V(p_state.repeat_u, SAMPLER_REPEAT_MODE_MAX, RID());
	sampler_create_info.addressModeU = address_modes[p_state.repeat_u];
	ERR_FAIL_INDEX_V(p_state.repeat_v, SAMPLER_REPEAT_MODE_MAX, RID());
	sampler_create_info.addressModeV = address_modes[p_state.repeat_v];
	ERR_FAIL_INDEX_V(p_state.repeat_w, SAMPLER_REPEAT_MODE_MAX, RID());
	sampler_create_info.addressModeW = address_modes[p_state.repeat_w];

	sampler_create_info.mipLodBias = p_state.lod_bias;
	sampler_create_info.anisotropyEnable = p_state.use_anisotropy;
	sampler_create_info.maxAnisotropy = p_state.anisotropy_max;
	sampler_create_info.compareEnable = p_state.enable_compare;

	ERR_FAIL_INDEX_V(p_state.compare_op, COMPARE_OP_MAX, RID());
	sampler_create_info.compareOp = compare_operators[p_state.compare_op];

	sampler_create_info.minLod = p_state.min_lod;
	sampler_create_info.maxLod = p_state.max_lod;

	ERR_FAIL_INDEX_V(p_state.border_color, SAMPLER_BORDER_COLOR_MAX, RID());
	sampler_create_info.borderColor = sampler_border_colors[p_state.border_color];

	sampler_create_info.unnormalizedCoordinates = p_state.unnormalized_uvw;

	VkSampler sampler;
	VkResult res = vkCreateSampler(device, &sampler_create_info, nullptr, &sampler);
	ERR_FAIL_COND_V_MSG(res, RID(), "vkCreateSampler failed with error " + itos(res) + ".");

	RID id = sampler_owner.make_rid(sampler);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif
	return id;
}

/**********************/
/**** VERTEX ARRAY ****/
/**********************/

RID RenderingDeviceVulkan::vertex_buffer_create(uint32_t p_size_bytes, const Vector<uint8_t> &p_data, bool p_use_as_storage) {
	_THREAD_SAFE_METHOD_

	ERR_FAIL_COND_V(p_data.size() && (uint32_t)p_data.size() != p_size_bytes, RID());
	ERR_FAIL_COND_V_MSG(draw_list != nullptr && p_data.size(), RID(),
			"Creating buffers with data is forbidden during creation of a draw list");
	ERR_FAIL_COND_V_MSG(compute_list != nullptr && p_data.size(), RID(),
			"Creating buffers with data is forbidden during creation of a draw list");

	uint32_t usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT;
	if (p_use_as_storage) {
		usage |= VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;
	}
	Buffer buffer;
	_buffer_allocate(&buffer, p_size_bytes, usage, VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE, 0);
	if (p_data.size()) {
		uint64_t data_size = p_data.size();
		const uint8_t *r = p_data.ptr();
		_buffer_update(&buffer, 0, r, data_size);
		_buffer_memory_barrier(buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT, false);
	}

	RID id = vertex_buffer_owner.make_rid(buffer);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif
	return id;
}

// Internally reference counted, this ID is warranted to be unique for the same description, but needs to be freed as many times as it was allocated.
RenderingDevice::VertexFormatID RenderingDeviceVulkan::vertex_format_create(const Vector<VertexAttribute> &p_vertex_formats) {
	_THREAD_SAFE_METHOD_

	VertexDescriptionKey key;
	key.vertex_formats = p_vertex_formats;

	VertexFormatID *idptr = vertex_format_cache.getptr(key);
	if (idptr) {
		return *idptr;
	}

	// Does not exist, create one and cache it.
	VertexDescriptionCache vdcache;
	vdcache.bindings = memnew_arr(VkVertexInputBindingDescription, p_vertex_formats.size());
	vdcache.attributes = memnew_arr(VkVertexInputAttributeDescription, p_vertex_formats.size());

	HashSet<int> used_locations;
	for (int i = 0; i < p_vertex_formats.size(); i++) {
		ERR_CONTINUE(p_vertex_formats[i].format >= DATA_FORMAT_MAX);
		ERR_FAIL_COND_V(used_locations.has(p_vertex_formats[i].location), INVALID_ID);

		ERR_FAIL_COND_V_MSG(get_format_vertex_size(p_vertex_formats[i].format) == 0, INVALID_ID,
				"Data format for attachment (" + itos(i) + "), '" + named_formats[p_vertex_formats[i].format] + "', is not valid for a vertex array.");

		vdcache.bindings[i].binding = i;
		vdcache.bindings[i].stride = p_vertex_formats[i].stride;
		vdcache.bindings[i].inputRate = p_vertex_formats[i].frequency == VERTEX_FREQUENCY_INSTANCE ? VK_VERTEX_INPUT_RATE_INSTANCE : VK_VERTEX_INPUT_RATE_VERTEX;
		vdcache.attributes[i].binding = i;
		vdcache.attributes[i].location = p_vertex_formats[i].location;
		vdcache.attributes[i].format = vulkan_formats[p_vertex_formats[i].format];
		vdcache.attributes[i].offset = p_vertex_formats[i].offset;
		used_locations.insert(p_vertex_formats[i].location);
	}

	vdcache.create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO;
	vdcache.create_info.pNext = nullptr;
	vdcache.create_info.flags = 0;

	vdcache.create_info.vertexAttributeDescriptionCount = p_vertex_formats.size();
	vdcache.create_info.pVertexAttributeDescriptions = vdcache.attributes;

	vdcache.create_info.vertexBindingDescriptionCount = p_vertex_formats.size();
	vdcache.create_info.pVertexBindingDescriptions = vdcache.bindings;
	vdcache.vertex_formats = p_vertex_formats;

	VertexFormatID id = VertexFormatID(vertex_format_cache.size()) | (VertexFormatID(ID_TYPE_VERTEX_FORMAT) << ID_BASE_SHIFT);
	vertex_format_cache[key] = id;
	vertex_formats[id] = vdcache;
	return id;
}

RID RenderingDeviceVulkan::vertex_array_create(uint32_t p_vertex_count, VertexFormatID p_vertex_format, const Vector<RID> &p_src_buffers) {
	_THREAD_SAFE_METHOD_

	ERR_FAIL_COND_V(!vertex_formats.has(p_vertex_format), RID());
	const VertexDescriptionCache &vd = vertex_formats[p_vertex_format];

	ERR_FAIL_COND_V(vd.vertex_formats.size() != p_src_buffers.size(), RID());

	for (int i = 0; i < p_src_buffers.size(); i++) {
		ERR_FAIL_COND_V(!vertex_buffer_owner.owns(p_src_buffers[i]), RID());
	}

	VertexArray vertex_array;

	vertex_array.vertex_count = p_vertex_count;
	vertex_array.description = p_vertex_format;
	vertex_array.max_instances_allowed = 0xFFFFFFFF; // By default as many as you want.
	for (int i = 0; i < p_src_buffers.size(); i++) {
		Buffer *buffer = vertex_buffer_owner.get_or_null(p_src_buffers[i]);

		// Validate with buffer.
		{
			const VertexAttribute &atf = vd.vertex_formats[i];

			uint32_t element_size = get_format_vertex_size(atf.format);
			ERR_FAIL_COND_V(element_size == 0, RID()); // Should never happens since this was prevalidated.

			if (atf.frequency == VERTEX_FREQUENCY_VERTEX) {
				// Validate size for regular drawing.
				uint64_t total_size = uint64_t(atf.stride) * (p_vertex_count - 1) + atf.offset + element_size;
				ERR_FAIL_COND_V_MSG(total_size > buffer->size, RID(),
						"Attachment (" + itos(i) + ") will read past the end of the buffer.");

			} else {
				// Validate size for instances drawing.
				uint64_t available = buffer->size - atf.offset;
				ERR_FAIL_COND_V_MSG(available < element_size, RID(),
						"Attachment (" + itos(i) + ") uses instancing, but it's just too small.");

				uint32_t instances_allowed = available / atf.stride;
				vertex_array.max_instances_allowed = MIN(instances_allowed, vertex_array.max_instances_allowed);
			}
		}

		vertex_array.buffers.push_back(buffer->buffer);
		vertex_array.offsets.push_back(0); // Offset unused, but passing anyway.
	}

	RID id = vertex_array_owner.make_rid(vertex_array);
	for (int i = 0; i < p_src_buffers.size(); i++) {
		_add_dependency(id, p_src_buffers[i]);
	}

	return id;
}

RID RenderingDeviceVulkan::index_buffer_create(uint32_t p_index_count, IndexBufferFormat p_format, const Vector<uint8_t> &p_data, bool p_use_restart_indices) {
	_THREAD_SAFE_METHOD_
	ERR_FAIL_COND_V_MSG(draw_list != nullptr && p_data.size(), RID(),
			"Creating buffers with data is forbidden during creation of a draw list");
	ERR_FAIL_COND_V_MSG(compute_list != nullptr && p_data.size(), RID(),
			"Creating buffers with data is forbidden during creation of a draw list");

	ERR_FAIL_COND_V(p_index_count == 0, RID());

	IndexBuffer index_buffer;
	index_buffer.index_type = (p_format == INDEX_BUFFER_FORMAT_UINT16) ? VK_INDEX_TYPE_UINT16 : VK_INDEX_TYPE_UINT32;
	index_buffer.supports_restart_indices = p_use_restart_indices;
	index_buffer.index_count = p_index_count;
	uint32_t size_bytes = p_index_count * ((p_format == INDEX_BUFFER_FORMAT_UINT16) ? 2 : 4);
#ifdef DEBUG_ENABLED
	if (p_data.size()) {
		index_buffer.max_index = 0;
		ERR_FAIL_COND_V_MSG((uint32_t)p_data.size() != size_bytes, RID(),
				"Default index buffer initializer array size (" + itos(p_data.size()) + ") does not match format required size (" + itos(size_bytes) + ").");
		const uint8_t *r = p_data.ptr();
		if (p_format == INDEX_BUFFER_FORMAT_UINT16) {
			const uint16_t *index16 = (const uint16_t *)r;
			for (uint32_t i = 0; i < p_index_count; i++) {
				if (p_use_restart_indices && index16[i] == 0xFFFF) {
					continue; // Restart index, ignore.
				}
				index_buffer.max_index = MAX(index16[i], index_buffer.max_index);
			}
		} else {
			const uint32_t *index32 = (const uint32_t *)r;
			for (uint32_t i = 0; i < p_index_count; i++) {
				if (p_use_restart_indices && index32[i] == 0xFFFFFFFF) {
					continue; // Restart index, ignore.
				}
				index_buffer.max_index = MAX(index32[i], index_buffer.max_index);
			}
		}
	} else {
		index_buffer.max_index = 0xFFFFFFFF;
	}
#else
	index_buffer.max_index = 0xFFFFFFFF;
#endif
	_buffer_allocate(&index_buffer, size_bytes, VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_INDEX_BUFFER_BIT, VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE, 0);
	if (p_data.size()) {
		uint64_t data_size = p_data.size();
		const uint8_t *r = p_data.ptr();
		_buffer_update(&index_buffer, 0, r, data_size);
		_buffer_memory_barrier(index_buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_INPUT_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_INDEX_READ_BIT, false);
	}
	RID id = index_buffer_owner.make_rid(index_buffer);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif
	return id;
}

RID RenderingDeviceVulkan::index_array_create(RID p_index_buffer, uint32_t p_index_offset, uint32_t p_index_count) {
	_THREAD_SAFE_METHOD_

	ERR_FAIL_COND_V(!index_buffer_owner.owns(p_index_buffer), RID());

	IndexBuffer *index_buffer = index_buffer_owner.get_or_null(p_index_buffer);

	ERR_FAIL_COND_V(p_index_count == 0, RID());
	ERR_FAIL_COND_V(p_index_offset + p_index_count > index_buffer->index_count, RID());

	IndexArray index_array;
	index_array.max_index = index_buffer->max_index;
	index_array.buffer = index_buffer->buffer;
	index_array.offset = p_index_offset;
	index_array.indices = p_index_count;
	index_array.index_type = index_buffer->index_type;
	index_array.supports_restart_indices = index_buffer->supports_restart_indices;

	RID id = index_array_owner.make_rid(index_array);
	_add_dependency(id, p_index_buffer);
	return id;
}

/****************/
/**** SHADER ****/
/****************/

static const char *shader_stage_names[RenderingDevice::SHADER_STAGE_MAX] = {
	"Vertex",
	"Fragment",
	"TesselationControl",
	"TesselationEvaluation",
	"Compute"
};

static const char *shader_uniform_names[RenderingDevice::UNIFORM_TYPE_MAX] = {
	"Sampler", "CombinedSampler", "Texture", "Image", "TextureBuffer", "SamplerTextureBuffer", "ImageBuffer", "UniformBuffer", "StorageBuffer", "InputAttachment"
};

static VkShaderStageFlagBits shader_stage_masks[RenderingDevice::SHADER_STAGE_MAX] = {
	VK_SHADER_STAGE_VERTEX_BIT,
	VK_SHADER_STAGE_FRAGMENT_BIT,
	VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT,
	VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT,
	VK_SHADER_STAGE_COMPUTE_BIT,
};

String RenderingDeviceVulkan::_shader_uniform_debug(RID p_shader, int p_set) {
	String ret;
	const Shader *shader = shader_owner.get_or_null(p_shader);
	ERR_FAIL_COND_V(!shader, String());
	for (int i = 0; i < shader->sets.size(); i++) {
		if (p_set >= 0 && i != p_set) {
			continue;
		}
		for (int j = 0; j < shader->sets[i].uniform_info.size(); j++) {
			const UniformInfo &ui = shader->sets[i].uniform_info[j];
			if (!ret.is_empty()) {
				ret += "\n";
			}
			ret += "Set: " + itos(i) + " Binding: " + itos(ui.binding) + " Type: " + shader_uniform_names[ui.type] + " Writable: " + (ui.writable ? "Y" : "N") + " Length: " + itos(ui.length);
		}
	}
	return ret;
}
#if 0
bool RenderingDeviceVulkan::_uniform_add_binding(Vector<Vector<VkDescriptorSetLayoutBinding> > &bindings, Vector<Vector<UniformInfo> > &uniform_infos, const glslang::TObjectReflection &reflection, RenderingDevice::ShaderStage p_stage, Shader::PushConstant &push_constant, String *r_error) {
	VkDescriptorSetLayoutBinding layout_binding;
	UniformInfo info;

	switch (reflection.getType()->getBasicType()) {
		case glslang::EbtSampler: {
			//print_line("DEBUG: IsSampler");
			if (reflection.getType()->getSampler().dim == glslang::EsdBuffer) {
				// Texture buffers.
				if (reflection.getType()->getSampler().isCombined()) {
					layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER;
					info.type = UNIFORM_TYPE_SAMPLER_WITH_TEXTURE_BUFFER;
					//print_line("DEBUG: SAMPLER: texel combined");
				} else if (reflection.getType()->getSampler().isTexture()) {
					layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER;
					info.type = UNIFORM_TYPE_TEXTURE_BUFFER;
					//print_line("DEBUG: SAMPLER: texel alone");
				} else if (reflection.getType()->getSampler().isImage()) {
					layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER;
					info.type = UNIFORM_TYPE_IMAGE_BUFFER;
					//print_line("DEBUG: SAMPLER: texel buffer");
				} else {
					if (r_error) {
						*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' is of unsupported buffer type.";
					}
					return false;
				}
			} else if (reflection.getType()->getSampler().isCombined()) {
				layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
				info.type = UNIFORM_TYPE_SAMPLER_WITH_TEXTURE;
				//print_line("DEBUG: SAMPLER: combined");
			} else if (reflection.getType()->getSampler().isPureSampler()) {
				layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLER;
				info.type = UNIFORM_TYPE_SAMPLER;
				//print_line("DEBUG: SAMPLER: sampler");
			} else if (reflection.getType()->getSampler().isTexture()) {
				layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE;
				info.type = UNIFORM_TYPE_TEXTURE;
				//print_line("DEBUG: SAMPLER: image");
			} else if (reflection.getType()->getSampler().isImage()) {
				layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE;
				info.type = UNIFORM_TYPE_IMAGE;
				//print_line("DEBUG: SAMPLER: storage image");
			} else {
				//print_line("DEBUG: sampler unknown");
				if (r_error) {
					*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' is of unsupported sampler type.";
				}
				return false;
			}

			if (reflection.getType()->isArray()) {
				layout_binding.descriptorCount = reflection.getType()->getArraySizes()->getCumulativeSize();
				//print_line("DEBUG: array of size: " + itos(layout_binding.descriptorCount));
			} else {
				layout_binding.descriptorCount = 1;
			}

			info.length = layout_binding.descriptorCount;

		} break;
		/*case glslang::EbtStruct: {
			print_line("DEBUG: Struct");

		} break;*/
		case glslang::EbtBlock: {
			//print_line("DEBUG: Block");
			if (reflection.getType()->getQualifier().storage == glslang::EvqUniform) {
				if (reflection.getType()->getQualifier().layoutPushConstant) {
					uint32_t len = reflection.size;
					if (push_constant.push_constant_size != 0 && push_constant.push_constant_size != len) {
						*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' push constants for different stages should all be the same size.";
						return false;
					}
					push_constant.push_constant_size = len;
					push_constant.push_constants_vk_stage |= shader_stage_masks[p_stage];
					return true;
				}
				//print_line("DEBUG: Uniform buffer");
				layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
				info.type = UNIFORM_TYPE_UNIFORM_BUFFER;
			} else if (reflection.getType()->getQualifier().storage == glslang::EvqBuffer) {
				layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
				info.type = UNIFORM_TYPE_STORAGE_BUFFER;
				//print_line("DEBUG: Storage buffer");
			} else {
				if (r_error) {
					*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' is of unsupported block type: (" + itos(reflection.getType()->getQualifier().storage) + ").";
				}
				return false;
			}

			if (reflection.getType()->isArray()) {
				layout_binding.descriptorCount = reflection.getType()->getArraySizes()->getCumulativeSize();
				//print_line("DEBUG: array of size: " + itos(layout_binding.descriptorCount));
			} else {
				layout_binding.descriptorCount = 1;
			}

			info.length = reflection.size;

		} break;
		/*case glslang::EbtReference: {
		} break;*/
		/*case glslang::EbtAtomicUint: {
		} break;*/
		default: {
			if (reflection.getType()->getQualifier().hasOffset() || reflection.name.find(".") != std::string::npos) {
				// Member of uniform block?
				return true;
			}

			if (r_error) {
				*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' unsupported uniform type.";
			}
			return false;
		}
	}

	if (!reflection.getType()->getQualifier().hasBinding()) {
		if (r_error) {
			*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' lacks a binding number.";
		}
		return false;
	}

	uint32_t set = reflection.getType()->getQualifier().hasSet() ? reflection.getType()->getQualifier().layoutSet : 0;

	if (set >= MAX_UNIFORM_SETS) {
		if (r_error) {
			*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' uses a set (" + itos(set) + ") index larger than what is supported (" + itos(MAX_UNIFORM_SETS) + ").";
		}
		return false;
	}

	if (set >= limits.maxBoundDescriptorSets) {
		if (r_error) {
			*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' uses a set (" + itos(set) + ") index larger than what is supported by the hardware (" + itos(limits.maxBoundDescriptorSets) + ").";
		}
		return false;
	}

	uint32_t binding = reflection.getType()->getQualifier().layoutBinding;

	if (set < (uint32_t)bindings.size()) {
		// Check if this already exists.
		for (int i = 0; i < bindings[set].size(); i++) {
			if (bindings[set][i].binding == binding) {
				// Already exists, verify that it's the same type.
				if (bindings[set][i].descriptorType != layout_binding.descriptorType) {
					if (r_error) {
						*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(binding) + " with different uniform type.";
					}
					return false;
				}

				// Also, verify that it's the same size.
				if (bindings[set][i].descriptorCount != layout_binding.descriptorCount || uniform_infos[set][i].length != info.length) {
					if (r_error) {
						*r_error = "On shader stage '" + String(shader_stage_names[p_stage]) + "', uniform '" + reflection.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(binding) + " with different uniform size.";
					}
					return false;
				}

				// Just append stage mask and return.
				bindings.write[set].write[i].stageFlags |= shader_stage_masks[p_stage];
				uniform_infos.write[set].write[i].stages |= 1 << p_stage;
				return true;
			}
		}
	}
	layout_binding.binding = binding;
	layout_binding.stageFlags = shader_stage_masks[p_stage];
	layout_binding.pImmutableSamplers = nullptr; // No support for this yet.

	info.stages = 1 << p_stage;
	info.binding = binding;

	if (set >= (uint32_t)bindings.size()) {
		bindings.resize(set + 1);
		uniform_infos.resize(set + 1);
	}
#if 0
	print_line("stage: " + String(shader_stage_names[p_stage]) + " set: " + itos(set) + " binding: " + itos(info.binding) + " type:" + shader_uniform_names[info.type] + " length: " + itos(info.length));
#endif
	bindings.write[set].push_back(layout_binding);
	uniform_infos.write[set].push_back(info);

	return true;
}
#endif

// Version 1: initial.
// Version 2: Added shader name.
// Version 3: Added writable.

#define SHADER_BINARY_VERSION 3

String RenderingDeviceVulkan::shader_get_binary_cache_key() const {
	return "Vulkan-SV" + itos(SHADER_BINARY_VERSION) + "-" + String(VERSION_NUMBER) + "-" + String(VERSION_HASH);
}

struct RenderingDeviceVulkanShaderBinaryDataBinding {
	uint32_t type;
	uint32_t binding;
	uint32_t stages;
	uint32_t length; // Size of arrays (in total elements), or ubos (in bytes * total elements).
	uint32_t writable;
};

struct RenderingDeviceVulkanShaderBinarySpecializationConstant {
	uint32_t type;
	uint32_t constant_id;
	union {
		uint32_t int_value;
		float float_value;
		bool bool_value;
	};
	uint32_t stage_flags;
};

struct RenderingDeviceVulkanShaderBinaryData {
	uint32_t vertex_input_mask;
	uint32_t fragment_outputs;
	uint32_t specialization_constant_count;
	uint32_t is_compute;
	uint32_t compute_local_size[3];
	uint32_t set_count;
	uint32_t push_constant_size;
	uint32_t push_constants_vk_stage;
	uint32_t stage_count;
	uint32_t shader_name_len;
};

Vector<uint8_t> RenderingDeviceVulkan::shader_compile_binary_from_spirv(const Vector<ShaderStageSPIRVData> &p_spirv, const String &p_shader_name) {
	RenderingDeviceVulkanShaderBinaryData binary_data;
	binary_data.vertex_input_mask = 0;
	binary_data.fragment_outputs = 0;
	binary_data.specialization_constant_count = 0;
	binary_data.is_compute = 0;
	binary_data.compute_local_size[0] = 0;
	binary_data.compute_local_size[1] = 0;
	binary_data.compute_local_size[2] = 0;
	binary_data.set_count = 0;
	binary_data.push_constant_size = 0;
	binary_data.push_constants_vk_stage = 0;

	Vector<Vector<RenderingDeviceVulkanShaderBinaryDataBinding>> uniform_info; // Set bindings.
	Vector<RenderingDeviceVulkanShaderBinarySpecializationConstant> specialization_constants;

	uint32_t stages_processed = 0;

	for (int i = 0; i < p_spirv.size(); i++) {
		if (p_spirv[i].shader_stage == SHADER_STAGE_COMPUTE) {
			binary_data.is_compute = true;
			ERR_FAIL_COND_V_MSG(p_spirv.size() != 1, Vector<uint8_t>(),
					"Compute shaders can only receive one stage, dedicated to compute.");
		}
		ERR_FAIL_COND_V_MSG(stages_processed & (1 << p_spirv[i].shader_stage), Vector<uint8_t>(),
				"Stage " + String(shader_stage_names[p_spirv[i].shader_stage]) + " submitted more than once.");

		{
			SpvReflectShaderModule module;
			const uint8_t *spirv = p_spirv[i].spir_v.ptr();
			SpvReflectResult result = spvReflectCreateShaderModule(p_spirv[i].spir_v.size(), spirv, &module);
			ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
					"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed parsing shader.");

			if (binary_data.is_compute) {
				binary_data.compute_local_size[0] = module.entry_points->local_size.x;
				binary_data.compute_local_size[1] = module.entry_points->local_size.y;
				binary_data.compute_local_size[2] = module.entry_points->local_size.z;
			}
			uint32_t binding_count = 0;
			result = spvReflectEnumerateDescriptorBindings(&module, &binding_count, nullptr);
			ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
					"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating descriptor bindings.");

			uint32_t stage = p_spirv[i].shader_stage;

			if (binding_count > 0) {
				// Parse bindings.

				Vector<SpvReflectDescriptorBinding *> bindings;
				bindings.resize(binding_count);
				result = spvReflectEnumerateDescriptorBindings(&module, &binding_count, bindings.ptrw());

				ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
						"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed getting descriptor bindings.");

				for (uint32_t j = 0; j < binding_count; j++) {
					const SpvReflectDescriptorBinding &binding = *bindings[j];

					RenderingDeviceVulkanShaderBinaryDataBinding info{};

					bool need_array_dimensions = false;
					bool need_block_size = false;
					bool may_be_writable = false;

					switch (binding.descriptor_type) {
						case SPV_REFLECT_DESCRIPTOR_TYPE_SAMPLER: {
							info.type = UNIFORM_TYPE_SAMPLER;
							need_array_dimensions = true;
						} break;
						case SPV_REFLECT_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER: {
							info.type = UNIFORM_TYPE_SAMPLER_WITH_TEXTURE;
							need_array_dimensions = true;
						} break;
						case SPV_REFLECT_DESCRIPTOR_TYPE_SAMPLED_IMAGE: {
							info.type = UNIFORM_TYPE_TEXTURE;
							need_array_dimensions = true;
						} break;
						case SPV_REFLECT_DESCRIPTOR_TYPE_STORAGE_IMAGE: {
							info.type = UNIFORM_TYPE_IMAGE;
							need_array_dimensions = true;
							may_be_writable = true;
						} break;
						case SPV_REFLECT_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER: {
							info.type = UNIFORM_TYPE_TEXTURE_BUFFER;
							need_array_dimensions = true;
						} break;
						case SPV_REFLECT_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER: {
							info.type = UNIFORM_TYPE_IMAGE_BUFFER;
							need_array_dimensions = true;
							may_be_writable = true;
						} break;
						case SPV_REFLECT_DESCRIPTOR_TYPE_UNIFORM_BUFFER: {
							info.type = UNIFORM_TYPE_UNIFORM_BUFFER;
							need_block_size = true;
						} break;
						case SPV_REFLECT_DESCRIPTOR_TYPE_STORAGE_BUFFER: {
							info.type = UNIFORM_TYPE_STORAGE_BUFFER;
							need_block_size = true;
							may_be_writable = true;
						} break;
						case SPV_REFLECT_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC: {
							ERR_PRINT("Dynamic uniform buffer not supported.");
							continue;
						} break;
						case SPV_REFLECT_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC: {
							ERR_PRINT("Dynamic storage buffer not supported.");
							continue;
						} break;
						case SPV_REFLECT_DESCRIPTOR_TYPE_INPUT_ATTACHMENT: {
							info.type = UNIFORM_TYPE_INPUT_ATTACHMENT;
							need_array_dimensions = true;
						} break;
						case SPV_REFLECT_DESCRIPTOR_TYPE_ACCELERATION_STRUCTURE_KHR: {
							ERR_PRINT("Acceleration structure not supported.");
							continue;
						} break;
					}

					if (need_array_dimensions) {
						if (binding.array.dims_count == 0) {
							info.length = 1;
						} else {
							for (uint32_t k = 0; k < binding.array.dims_count; k++) {
								if (k == 0) {
									info.length = binding.array.dims[0];
								} else {
									info.length *= binding.array.dims[k];
								}
							}
						}

					} else if (need_block_size) {
						info.length = binding.block.size;
					} else {
						info.length = 0;
					}

					if (may_be_writable) {
						info.writable = !(binding.type_description->decoration_flags & SPV_REFLECT_DECORATION_NON_WRITABLE) && !(binding.block.decoration_flags & SPV_REFLECT_DECORATION_NON_WRITABLE);
					} else {
						info.writable = false;
					}

					info.binding = binding.binding;
					uint32_t set = binding.set;

					ERR_FAIL_COND_V_MSG(set >= MAX_UNIFORM_SETS, Vector<uint8_t>(),
							"On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' uses a set (" + itos(set) + ") index larger than what is supported (" + itos(MAX_UNIFORM_SETS) + ").");

					ERR_FAIL_COND_V_MSG(set >= limits.maxBoundDescriptorSets, Vector<uint8_t>(),
							"On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' uses a set (" + itos(set) + ") index larger than what is supported by the hardware (" + itos(limits.maxBoundDescriptorSets) + ").");

					if (set < (uint32_t)uniform_info.size()) {
						// Check if this already exists.
						bool exists = false;
						for (int k = 0; k < uniform_info[set].size(); k++) {
							if (uniform_info[set][k].binding == (uint32_t)info.binding) {
								// Already exists, verify that it's the same type.
								ERR_FAIL_COND_V_MSG(uniform_info[set][k].type != info.type, Vector<uint8_t>(),
										"On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(info.binding) + " with different uniform type.");

								// Also, verify that it's the same size.
								ERR_FAIL_COND_V_MSG(uniform_info[set][k].length != info.length, Vector<uint8_t>(),
										"On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(info.binding) + " with different uniform size.");

								// Also, verify that it has the same writability.
								ERR_FAIL_COND_V_MSG(uniform_info[set][k].writable != info.writable, Vector<uint8_t>(),
										"On shader stage '" + String(shader_stage_names[stage]) + "', uniform '" + binding.name + "' trying to re-use location for set=" + itos(set) + ", binding=" + itos(info.binding) + " with different writability.");

								// Just append stage mask and return.
								uniform_info.write[set].write[k].stages |= 1 << stage;
								exists = true;
								break;
							}
						}

						if (exists) {
							continue; // Merged.
						}
					}

					info.stages = 1 << stage;

					if (set >= (uint32_t)uniform_info.size()) {
						uniform_info.resize(set + 1);
					}

					uniform_info.write[set].push_back(info);
				}
			}

			{
				// Specialization constants.

				uint32_t sc_count = 0;
				result = spvReflectEnumerateSpecializationConstants(&module, &sc_count, nullptr);
				ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
						"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating specialization constants.");

				if (sc_count) {
					Vector<SpvReflectSpecializationConstant *> spec_constants;
					spec_constants.resize(sc_count);

					result = spvReflectEnumerateSpecializationConstants(&module, &sc_count, spec_constants.ptrw());
					ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
							"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed obtaining specialization constants.");

					for (uint32_t j = 0; j < sc_count; j++) {
						int32_t existing = -1;
						RenderingDeviceVulkanShaderBinarySpecializationConstant sconst{};
						SpvReflectSpecializationConstant *spc = spec_constants[j];

						sconst.constant_id = spc->constant_id;
						sconst.int_value = 0.0; // Clear previous value JIC.
						switch (spc->constant_type) {
							case SPV_REFLECT_SPECIALIZATION_CONSTANT_BOOL: {
								sconst.type = PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL;
								sconst.bool_value = spc->default_value.int_bool_value != 0;
							} break;
							case SPV_REFLECT_SPECIALIZATION_CONSTANT_INT: {
								sconst.type = PIPELINE_SPECIALIZATION_CONSTANT_TYPE_INT;
								sconst.int_value = spc->default_value.int_bool_value;
							} break;
							case SPV_REFLECT_SPECIALIZATION_CONSTANT_FLOAT: {
								sconst.type = PIPELINE_SPECIALIZATION_CONSTANT_TYPE_FLOAT;
								sconst.float_value = spc->default_value.float_value;
							} break;
						}
						sconst.stage_flags = 1 << p_spirv[i].shader_stage;

						for (int k = 0; k < specialization_constants.size(); k++) {
							if (specialization_constants[k].constant_id == sconst.constant_id) {
								ERR_FAIL_COND_V_MSG(specialization_constants[k].type != sconst.type, Vector<uint8_t>(), "More than one specialization constant used for id (" + itos(sconst.constant_id) + "), but their types differ.");
								ERR_FAIL_COND_V_MSG(specialization_constants[k].int_value != sconst.int_value, Vector<uint8_t>(), "More than one specialization constant used for id (" + itos(sconst.constant_id) + "), but their default values differ.");
								existing = k;
								break;
							}
						}

						if (existing > 0) {
							specialization_constants.write[existing].stage_flags |= sconst.stage_flags;
						} else {
							specialization_constants.push_back(sconst);
						}
					}
				}
			}

			if (stage == SHADER_STAGE_VERTEX) {
				uint32_t iv_count = 0;
				result = spvReflectEnumerateInputVariables(&module, &iv_count, nullptr);
				ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
						"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating input variables.");

				if (iv_count) {
					Vector<SpvReflectInterfaceVariable *> input_vars;
					input_vars.resize(iv_count);

					result = spvReflectEnumerateInputVariables(&module, &iv_count, input_vars.ptrw());
					ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
							"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed obtaining input variables.");

					for (uint32_t j = 0; j < iv_count; j++) {
						if (input_vars[j] && input_vars[j]->decoration_flags == 0) { // Regular input.
							binary_data.vertex_input_mask |= (1 << uint32_t(input_vars[j]->location));
						}
					}
				}
			}

			if (stage == SHADER_STAGE_FRAGMENT) {
				uint32_t ov_count = 0;
				result = spvReflectEnumerateOutputVariables(&module, &ov_count, nullptr);
				ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
						"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating output variables.");

				if (ov_count) {
					Vector<SpvReflectInterfaceVariable *> output_vars;
					output_vars.resize(ov_count);

					result = spvReflectEnumerateOutputVariables(&module, &ov_count, output_vars.ptrw());
					ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
							"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed obtaining output variables.");

					for (uint32_t j = 0; j < ov_count; j++) {
						const SpvReflectInterfaceVariable *refvar = output_vars[j];
						if (refvar != nullptr && refvar->built_in != SpvBuiltInFragDepth) {
							binary_data.fragment_outputs |= 1 << refvar->location;
						}
					}
				}
			}

			uint32_t pc_count = 0;
			result = spvReflectEnumeratePushConstantBlocks(&module, &pc_count, nullptr);
			ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
					"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed enumerating push constants.");

			if (pc_count) {
				ERR_FAIL_COND_V_MSG(pc_count > 1, Vector<uint8_t>(),
						"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "': Only one push constant is supported, which should be the same across shader stages.");

				Vector<SpvReflectBlockVariable *> pconstants;
				pconstants.resize(pc_count);
				result = spvReflectEnumeratePushConstantBlocks(&module, &pc_count, pconstants.ptrw());
				ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, Vector<uint8_t>(),
						"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "' failed obtaining push constants.");
#if 0
				if (pconstants[0] == nullptr) {
					Ref<FileAccess> f = FileAccess::open("res://popo.spv", FileAccess::WRITE);
					f->store_buffer((const uint8_t *)&SpirV[0], SpirV.size() * sizeof(uint32_t));
				}
#endif

				ERR_FAIL_COND_V_MSG(binary_data.push_constant_size && binary_data.push_constant_size != pconstants[0]->size, Vector<uint8_t>(),
						"Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_spirv[i].shader_stage]) + "': Push constant block must be the same across shader stages.");

				binary_data.push_constant_size = pconstants[0]->size;
				binary_data.push_constants_vk_stage |= shader_stage_masks[stage];

				//print_line("Stage: " + String(shader_stage_names[stage]) + " push constant of size=" + itos(push_constant.push_constant_size));
			}

			// Destroy the reflection data when no longer required.
			spvReflectDestroyShaderModule(&module);
		}

		stages_processed |= (1 << p_spirv[i].shader_stage);
	}

	Vector<Vector<uint8_t>> compressed_stages;
	Vector<uint32_t> smolv_size;
	Vector<uint32_t> zstd_size; // If 0, zstd not used.

	uint32_t stages_binary_size = 0;

	bool strip_debug = false;

	for (int i = 0; i < p_spirv.size(); i++) {
		smolv::ByteArray smolv;
		if (!smolv::Encode(p_spirv[i].spir_v.ptr(), p_spirv[i].spir_v.size(), smolv, strip_debug ? smolv::kEncodeFlagStripDebugInfo : 0)) {
			ERR_FAIL_V_MSG(Vector<uint8_t>(), "Error compressing shader stage :" + String(shader_stage_names[p_spirv[i].shader_stage]));
		} else {
			smolv_size.push_back(smolv.size());
			{ // zstd.
				Vector<uint8_t> zstd;
				zstd.resize(Compression::get_max_compressed_buffer_size(smolv.size(), Compression::MODE_ZSTD));
				int dst_size = Compression::compress(zstd.ptrw(), &smolv[0], smolv.size(), Compression::MODE_ZSTD);

				if (dst_size > 0 && (uint32_t)dst_size < smolv.size()) {
					zstd_size.push_back(dst_size);
					zstd.resize(dst_size);
					compressed_stages.push_back(zstd);
				} else {
					Vector<uint8_t> smv;
					smv.resize(smolv.size());
					memcpy(smv.ptrw(), &smolv[0], smolv.size());
					zstd_size.push_back(0); // Not using zstd.
					compressed_stages.push_back(smv);
				}
			}
		}
		uint32_t s = compressed_stages[i].size();
		if (s % 4 != 0) {
			s += 4 - (s % 4);
		}
		stages_binary_size += s;
	}

	binary_data.specialization_constant_count = specialization_constants.size();
	binary_data.set_count = uniform_info.size();
	binary_data.stage_count = p_spirv.size();

	CharString shader_name_utf = p_shader_name.utf8();

	binary_data.shader_name_len = shader_name_utf.length();

	uint32_t total_size = sizeof(uint32_t) * 3; // Header + version + main datasize;.
	total_size += sizeof(RenderingDeviceVulkanShaderBinaryData);

	total_size += binary_data.shader_name_len;

	if ((binary_data.shader_name_len % 4) != 0) { // Alignment rules are really strange.
		total_size += 4 - (binary_data.shader_name_len % 4);
	}

	for (int i = 0; i < uniform_info.size(); i++) {
		total_size += sizeof(uint32_t);
		total_size += uniform_info[i].size() * sizeof(RenderingDeviceVulkanShaderBinaryDataBinding);
	}

	total_size += sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant) * specialization_constants.size();

	total_size += compressed_stages.size() * sizeof(uint32_t) * 3; // Sizes.
	total_size += stages_binary_size;

	Vector<uint8_t> ret;
	ret.resize(total_size);
	{
		uint32_t offset = 0;
		uint8_t *binptr = ret.ptrw();
		binptr[0] = 'G';
		binptr[1] = 'S';
		binptr[2] = 'B';
		binptr[3] = 'D'; // Godot Shader Binary Data.
		offset += 4;
		encode_uint32(SHADER_BINARY_VERSION, binptr + offset);
		offset += sizeof(uint32_t);
		encode_uint32(sizeof(RenderingDeviceVulkanShaderBinaryData), binptr + offset);
		offset += sizeof(uint32_t);
		memcpy(binptr + offset, &binary_data, sizeof(RenderingDeviceVulkanShaderBinaryData));
		offset += sizeof(RenderingDeviceVulkanShaderBinaryData);
		memcpy(binptr + offset, shader_name_utf.ptr(), binary_data.shader_name_len);
		offset += binary_data.shader_name_len;

		if ((binary_data.shader_name_len % 4) != 0) { // Alignment rules are really strange.
			offset += 4 - (binary_data.shader_name_len % 4);
		}

		for (int i = 0; i < uniform_info.size(); i++) {
			int count = uniform_info[i].size();
			encode_uint32(count, binptr + offset);
			offset += sizeof(uint32_t);
			if (count > 0) {
				memcpy(binptr + offset, uniform_info[i].ptr(), sizeof(RenderingDeviceVulkanShaderBinaryDataBinding) * count);
				offset += sizeof(RenderingDeviceVulkanShaderBinaryDataBinding) * count;
			}
		}

		if (specialization_constants.size()) {
			memcpy(binptr + offset, specialization_constants.ptr(), sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant) * specialization_constants.size());
			offset += sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant) * specialization_constants.size();
		}

		for (int i = 0; i < compressed_stages.size(); i++) {
			encode_uint32(p_spirv[i].shader_stage, binptr + offset);
			offset += sizeof(uint32_t);
			encode_uint32(smolv_size[i], binptr + offset);
			offset += sizeof(uint32_t);
			encode_uint32(zstd_size[i], binptr + offset);
			offset += sizeof(uint32_t);
			memcpy(binptr + offset, compressed_stages[i].ptr(), compressed_stages[i].size());

			uint32_t s = compressed_stages[i].size();

			if (s % 4 != 0) {
				s += 4 - (s % 4);
			}

			offset += s;
		}

		ERR_FAIL_COND_V(offset != (uint32_t)ret.size(), Vector<uint8_t>());
	}

	return ret;
}

RID RenderingDeviceVulkan::shader_create_from_bytecode(const Vector<uint8_t> &p_shader_binary) {
	const uint8_t *binptr = p_shader_binary.ptr();
	uint32_t binsize = p_shader_binary.size();

	uint32_t read_offset = 0;
	// Consistency check.
	ERR_FAIL_COND_V(binsize < sizeof(uint32_t) * 3 + sizeof(RenderingDeviceVulkanShaderBinaryData), RID());
	ERR_FAIL_COND_V(binptr[0] != 'G' || binptr[1] != 'S' || binptr[2] != 'B' || binptr[3] != 'D', RID());

	uint32_t bin_version = decode_uint32(binptr + 4);
	ERR_FAIL_COND_V(bin_version != SHADER_BINARY_VERSION, RID());

	uint32_t bin_data_size = decode_uint32(binptr + 8);

	const RenderingDeviceVulkanShaderBinaryData &binary_data = *(reinterpret_cast<const RenderingDeviceVulkanShaderBinaryData *>(binptr + 12));

	Shader::PushConstant push_constant;
	push_constant.push_constant_size = binary_data.push_constant_size;
	push_constant.push_constants_vk_stage = binary_data.push_constants_vk_stage;

	uint32_t vertex_input_mask = binary_data.vertex_input_mask;

	uint32_t fragment_outputs = binary_data.fragment_outputs;

	bool is_compute = binary_data.is_compute;

	const uint32_t compute_local_size[3] = { binary_data.compute_local_size[0], binary_data.compute_local_size[1], binary_data.compute_local_size[2] };

	read_offset += sizeof(uint32_t) * 3 + bin_data_size;

	String name;

	if (binary_data.shader_name_len) {
		name.parse_utf8((const char *)(binptr + read_offset), binary_data.shader_name_len);
		read_offset += binary_data.shader_name_len;
		if ((binary_data.shader_name_len % 4) != 0) { // Alignment rules are really strange.
			read_offset += 4 - (binary_data.shader_name_len % 4);
		}
	}

	Vector<Vector<VkDescriptorSetLayoutBinding>> set_bindings;
	Vector<Vector<UniformInfo>> uniform_info;

	set_bindings.resize(binary_data.set_count);
	uniform_info.resize(binary_data.set_count);

	for (uint32_t i = 0; i < binary_data.set_count; i++) {
		ERR_FAIL_COND_V(read_offset + sizeof(uint32_t) >= binsize, RID());
		uint32_t set_count = decode_uint32(binptr + read_offset);
		read_offset += sizeof(uint32_t);
		const RenderingDeviceVulkanShaderBinaryDataBinding *set_ptr = reinterpret_cast<const RenderingDeviceVulkanShaderBinaryDataBinding *>(binptr + read_offset);
		uint32_t set_size = set_count * sizeof(RenderingDeviceVulkanShaderBinaryDataBinding);
		ERR_FAIL_COND_V(read_offset + set_size >= binsize, RID());

		for (uint32_t j = 0; j < set_count; j++) {
			UniformInfo info;
			info.type = UniformType(set_ptr[j].type);
			info.writable = set_ptr[j].writable;
			info.length = set_ptr[j].length;
			info.binding = set_ptr[j].binding;
			info.stages = set_ptr[j].stages;

			VkDescriptorSetLayoutBinding layout_binding;
			layout_binding.pImmutableSamplers = nullptr;
			layout_binding.binding = set_ptr[j].binding;
			layout_binding.descriptorCount = 1;
			layout_binding.stageFlags = 0;
			for (uint32_t k = 0; k < SHADER_STAGE_MAX; k++) {
				if (set_ptr[j].stages & (1 << k)) {
					layout_binding.stageFlags |= shader_stage_masks[k];
				}
			}

			switch (info.type) {
				case UNIFORM_TYPE_SAMPLER: {
					layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLER;
					layout_binding.descriptorCount = set_ptr[j].length;
				} break;
				case UNIFORM_TYPE_SAMPLER_WITH_TEXTURE: {
					layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
					layout_binding.descriptorCount = set_ptr[j].length;
				} break;
				case UNIFORM_TYPE_TEXTURE: {
					layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE;
					layout_binding.descriptorCount = set_ptr[j].length;
				} break;
				case UNIFORM_TYPE_IMAGE: {
					layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE;
					layout_binding.descriptorCount = set_ptr[j].length;
				} break;
				case UNIFORM_TYPE_TEXTURE_BUFFER: {
					layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER;
					layout_binding.descriptorCount = set_ptr[j].length;
				} break;
				case UNIFORM_TYPE_IMAGE_BUFFER: {
					layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER;
				} break;
				case UNIFORM_TYPE_UNIFORM_BUFFER: {
					layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
				} break;
				case UNIFORM_TYPE_STORAGE_BUFFER: {
					layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
				} break;
				case UNIFORM_TYPE_INPUT_ATTACHMENT: {
					layout_binding.descriptorType = VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT;
				} break;
				default: {
					ERR_FAIL_V(RID());
				}
			}

			set_bindings.write[i].push_back(layout_binding);
			uniform_info.write[i].push_back(info);
		}

		read_offset += set_size;
	}

	ERR_FAIL_COND_V(read_offset + binary_data.specialization_constant_count * sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant) >= binsize, RID());

	Vector<Shader::SpecializationConstant> specialization_constants;

	for (uint32_t i = 0; i < binary_data.specialization_constant_count; i++) {
		const RenderingDeviceVulkanShaderBinarySpecializationConstant &src_sc = *(reinterpret_cast<const RenderingDeviceVulkanShaderBinarySpecializationConstant *>(binptr + read_offset));
		Shader::SpecializationConstant sc;
		sc.constant.int_value = src_sc.int_value;
		sc.constant.type = PipelineSpecializationConstantType(src_sc.type);
		sc.constant.constant_id = src_sc.constant_id;
		sc.stage_flags = src_sc.stage_flags;
		specialization_constants.push_back(sc);

		read_offset += sizeof(RenderingDeviceVulkanShaderBinarySpecializationConstant);
	}

	Vector<Vector<uint8_t>> stage_spirv_data;
	Vector<ShaderStage> stage_type;

	for (uint32_t i = 0; i < binary_data.stage_count; i++) {
		ERR_FAIL_COND_V(read_offset + sizeof(uint32_t) * 3 >= binsize, RID());
		uint32_t stage = decode_uint32(binptr + read_offset);
		read_offset += sizeof(uint32_t);
		uint32_t smolv_size = decode_uint32(binptr + read_offset);
		read_offset += sizeof(uint32_t);
		uint32_t zstd_size = decode_uint32(binptr + read_offset);
		read_offset += sizeof(uint32_t);

		uint32_t buf_size = (zstd_size > 0) ? zstd_size : smolv_size;

		Vector<uint8_t> smolv;
		const uint8_t *src_smolv = nullptr;

		if (zstd_size > 0) {
			// Decompress to smolv.
			smolv.resize(smolv_size);
			int dec_smolv_size = Compression::decompress(smolv.ptrw(), smolv.size(), binptr + read_offset, zstd_size, Compression::MODE_ZSTD);
			ERR_FAIL_COND_V(dec_smolv_size != (int32_t)smolv_size, RID());
			src_smolv = smolv.ptr();
		} else {
			src_smolv = binptr + read_offset;
		}

		Vector<uint8_t> spirv;
		uint32_t spirv_size = smolv::GetDecodedBufferSize(src_smolv, smolv_size);
		spirv.resize(spirv_size);
		if (!smolv::Decode(src_smolv, smolv_size, spirv.ptrw(), spirv_size)) {
			ERR_FAIL_V_MSG(RID(), "Malformed smolv input uncompressing shader stage:" + String(shader_stage_names[stage]));
		}
		stage_spirv_data.push_back(spirv);
		stage_type.push_back(ShaderStage(stage));

		if (buf_size % 4 != 0) {
			buf_size += 4 - (buf_size % 4);
		}

		ERR_FAIL_COND_V(read_offset + buf_size > binsize, RID());

		read_offset += buf_size;
	}

	ERR_FAIL_COND_V(read_offset != binsize, RID());

	// All good, let's create modules.

	_THREAD_SAFE_METHOD_

	Shader shader;

	shader.vertex_input_mask = vertex_input_mask;
	shader.fragment_output_mask = fragment_outputs;
	shader.push_constant = push_constant;
	shader.is_compute = is_compute;
	shader.compute_local_size[0] = compute_local_size[0];
	shader.compute_local_size[1] = compute_local_size[1];
	shader.compute_local_size[2] = compute_local_size[2];
	shader.specialization_constants = specialization_constants;
	shader.name = name;

	String error_text;

	bool success = true;
	for (int i = 0; i < stage_spirv_data.size(); i++) {
		VkShaderModuleCreateInfo shader_module_create_info;
		shader_module_create_info.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO;
		shader_module_create_info.pNext = nullptr;
		shader_module_create_info.flags = 0;
		shader_module_create_info.codeSize = stage_spirv_data[i].size();
		const uint8_t *r = stage_spirv_data[i].ptr();

		shader_module_create_info.pCode = (const uint32_t *)r;

		VkShaderModule module;
		VkResult res = vkCreateShaderModule(device, &shader_module_create_info, nullptr, &module);
		if (res) {
			success = false;
			error_text = "Error (" + itos(res) + ") creating shader module for stage: " + String(shader_stage_names[stage_type[i]]);
			break;
		}

		const VkShaderStageFlagBits shader_stage_bits[SHADER_STAGE_MAX] = {
			VK_SHADER_STAGE_VERTEX_BIT,
			VK_SHADER_STAGE_FRAGMENT_BIT,
			VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT,
			VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT,
			VK_SHADER_STAGE_COMPUTE_BIT,
		};

		VkPipelineShaderStageCreateInfo shader_stage;
		shader_stage.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
		shader_stage.pNext = nullptr;
		shader_stage.flags = 0;
		shader_stage.stage = shader_stage_bits[stage_type[i]];
		shader_stage.module = module;
		shader_stage.pName = "main";
		shader_stage.pSpecializationInfo = nullptr;

		shader.pipeline_stages.push_back(shader_stage);
	}
	// Proceed to create descriptor sets.

	if (success) {
		for (int i = 0; i < set_bindings.size(); i++) {
			// Empty ones are fine if they were not used according to spec (binding count will be 0).
			VkDescriptorSetLayoutCreateInfo layout_create_info;
			layout_create_info.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO;
			layout_create_info.pNext = nullptr;
			layout_create_info.flags = 0;
			layout_create_info.bindingCount = set_bindings[i].size();
			layout_create_info.pBindings = set_bindings[i].ptr();

			VkDescriptorSetLayout layout;
			VkResult res = vkCreateDescriptorSetLayout(device, &layout_create_info, nullptr, &layout);
			if (res) {
				error_text = "Error (" + itos(res) + ") creating descriptor set layout for set " + itos(i);
				success = false;
				break;
			}

			Shader::Set set;
			set.descriptor_set_layout = layout;
			set.uniform_info = uniform_info[i];
			// Sort and hash.
			set.uniform_info.sort();

			uint32_t format = 0; // No format, default.

			if (set.uniform_info.size()) {
				// Has data, needs an actual format.
				UniformSetFormat usformat;
				usformat.uniform_info = set.uniform_info;
				RBMap<UniformSetFormat, uint32_t>::Element *E = uniform_set_format_cache.find(usformat);
				if (E) {
					format = E->get();
				} else {
					format = uniform_set_format_cache.size() + 1;
					uniform_set_format_cache.insert(usformat, format);
				}
			}

			shader.sets.push_back(set);
			shader.set_formats.push_back(format);
		}
	}

	if (success) {
		// Create pipeline layout.
		VkPipelineLayoutCreateInfo pipeline_layout_create_info;
		pipeline_layout_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO;
		pipeline_layout_create_info.pNext = nullptr;
		pipeline_layout_create_info.flags = 0;
		pipeline_layout_create_info.setLayoutCount = shader.sets.size();

		Vector<VkDescriptorSetLayout> layouts;
		layouts.resize(shader.sets.size());

		for (int i = 0; i < layouts.size(); i++) {
			layouts.write[i] = shader.sets[i].descriptor_set_layout;
		}

		pipeline_layout_create_info.pSetLayouts = layouts.ptr();
		// Needs to be declared in this outer scope, otherwise it may not outlive its assignment
		// to pipeline_layout_create_info.
		VkPushConstantRange push_constant_range;
		if (push_constant.push_constant_size) {
			push_constant_range.stageFlags = push_constant.push_constants_vk_stage;
			push_constant_range.offset = 0;
			push_constant_range.size = push_constant.push_constant_size;

			pipeline_layout_create_info.pushConstantRangeCount = 1;
			pipeline_layout_create_info.pPushConstantRanges = &push_constant_range;
		} else {
			pipeline_layout_create_info.pushConstantRangeCount = 0;
			pipeline_layout_create_info.pPushConstantRanges = nullptr;
		}

		VkResult err = vkCreatePipelineLayout(device, &pipeline_layout_create_info, nullptr, &shader.pipeline_layout);

		if (err) {
			error_text = "Error (" + itos(err) + ") creating pipeline layout.";
			success = false;
		}
	}

	if (!success) {
		// Clean up if failed.
		for (int i = 0; i < shader.pipeline_stages.size(); i++) {
			vkDestroyShaderModule(device, shader.pipeline_stages[i].module, nullptr);
		}

		for (int i = 0; i < shader.sets.size(); i++) {
			vkDestroyDescriptorSetLayout(device, shader.sets[i].descriptor_set_layout, nullptr);
		}

		ERR_FAIL_V_MSG(RID(), error_text);
	}

	RID id = shader_owner.make_rid(shader);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif
	return id;
}

uint32_t RenderingDeviceVulkan::shader_get_vertex_input_attribute_mask(RID p_shader) {
	_THREAD_SAFE_METHOD_

	const Shader *shader = shader_owner.get_or_null(p_shader);
	ERR_FAIL_COND_V(!shader, 0);
	return shader->vertex_input_mask;
}

/******************/
/**** UNIFORMS ****/
/******************/

RID RenderingDeviceVulkan::uniform_buffer_create(uint32_t p_size_bytes, const Vector<uint8_t> &p_data) {
	_THREAD_SAFE_METHOD_

	ERR_FAIL_COND_V(p_data.size() && (uint32_t)p_data.size() != p_size_bytes, RID());
	ERR_FAIL_COND_V_MSG(draw_list != nullptr && p_data.size(), RID(),
			"Creating buffers with data is forbidden during creation of a draw list");
	ERR_FAIL_COND_V_MSG(compute_list != nullptr && p_data.size(), RID(),
			"Creating buffers with data is forbidden during creation of a draw list");

	Buffer buffer;
	Error err = _buffer_allocate(&buffer, p_size_bytes, VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT, VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE, 0);
	ERR_FAIL_COND_V(err != OK, RID());
	if (p_data.size()) {
		uint64_t data_size = p_data.size();
		const uint8_t *r = p_data.ptr();
		_buffer_update(&buffer, 0, r, data_size);
		_buffer_memory_barrier(buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_UNIFORM_READ_BIT, false);
	}
	RID id = uniform_buffer_owner.make_rid(buffer);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif
	return id;
}

RID RenderingDeviceVulkan::storage_buffer_create(uint32_t p_size_bytes, const Vector<uint8_t> &p_data, uint32_t p_usage) {
	_THREAD_SAFE_METHOD_
	ERR_FAIL_COND_V_MSG(draw_list != nullptr && p_data.size(), RID(),
			"Creating buffers with data is forbidden during creation of a draw list");
	ERR_FAIL_COND_V_MSG(compute_list != nullptr && p_data.size(), RID(),
			"Creating buffers with data is forbidden during creation of a draw list");

	ERR_FAIL_COND_V(p_data.size() && (uint32_t)p_data.size() != p_size_bytes, RID());

	Buffer buffer;
	uint32_t flags = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;
	if (p_usage & STORAGE_BUFFER_USAGE_DISPATCH_INDIRECT) {
		flags |= VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT;
	}
	Error err = _buffer_allocate(&buffer, p_size_bytes, flags, VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE, 0);
	ERR_FAIL_COND_V(err != OK, RID());

	if (p_data.size()) {
		uint64_t data_size = p_data.size();
		const uint8_t *r = p_data.ptr();
		_buffer_update(&buffer, 0, r, data_size);
		_buffer_memory_barrier(buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT, false);
	}
	return storage_buffer_owner.make_rid(buffer);
}

RID RenderingDeviceVulkan::texture_buffer_create(uint32_t p_size_elements, DataFormat p_format, const Vector<uint8_t> &p_data) {
	_THREAD_SAFE_METHOD_
	ERR_FAIL_COND_V_MSG(draw_list != nullptr && p_data.size(), RID(),
			"Creating buffers with data is forbidden during creation of a draw list");
	ERR_FAIL_COND_V_MSG(compute_list != nullptr && p_data.size(), RID(),
			"Creating buffers with data is forbidden during creation of a draw list");

	uint32_t element_size = get_format_vertex_size(p_format);
	ERR_FAIL_COND_V_MSG(element_size == 0, RID(), "Format requested is not supported for texture buffers");
	uint64_t size_bytes = uint64_t(element_size) * p_size_elements;

	ERR_FAIL_COND_V(p_data.size() && (uint32_t)p_data.size() != size_bytes, RID());

	TextureBuffer texture_buffer;
	Error err = _buffer_allocate(&texture_buffer.buffer, size_bytes, VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT, VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE, 0);
	ERR_FAIL_COND_V(err != OK, RID());

	if (p_data.size()) {
		uint64_t data_size = p_data.size();
		const uint8_t *r = p_data.ptr();
		_buffer_update(&texture_buffer.buffer, 0, r, data_size);
		_buffer_memory_barrier(texture_buffer.buffer.buffer, 0, data_size, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT, VK_ACCESS_SHADER_READ_BIT, false);
	}

	VkBufferViewCreateInfo view_create_info;
	view_create_info.sType = VK_STRUCTURE_TYPE_BUFFER_VIEW_CREATE_INFO;
	view_create_info.pNext = nullptr;
	view_create_info.flags = 0;
	view_create_info.buffer = texture_buffer.buffer.buffer;
	view_create_info.format = vulkan_formats[p_format];
	view_create_info.offset = 0;
	view_create_info.range = size_bytes;

	texture_buffer.view = VK_NULL_HANDLE;

	VkResult res = vkCreateBufferView(device, &view_create_info, nullptr, &texture_buffer.view);
	if (res) {
		_buffer_free(&texture_buffer.buffer);
		ERR_FAIL_V_MSG(RID(), "Unable to create buffer view, error " + itos(res) + ".");
	}

	// Allocate the view.
	RID id = texture_buffer_owner.make_rid(texture_buffer);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif
	return id;
}

RenderingDeviceVulkan::DescriptorPool *RenderingDeviceVulkan::_descriptor_pool_allocate(const DescriptorPoolKey &p_key) {
	if (!descriptor_pools.has(p_key)) {
		descriptor_pools[p_key] = HashSet<DescriptorPool *>();
	}

	DescriptorPool *pool = nullptr;

	for (DescriptorPool *E : descriptor_pools[p_key]) {
		if (E->usage < max_descriptors_per_pool) {
			pool = E;
			break;
		}
	}

	if (!pool) {
		// Create a new one.
		pool = memnew(DescriptorPool);
		pool->usage = 0;

		VkDescriptorPoolCreateInfo descriptor_pool_create_info;
		descriptor_pool_create_info.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO;
		descriptor_pool_create_info.pNext = nullptr;
		descriptor_pool_create_info.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT; // Can't think how somebody may NOT need this flag.
		descriptor_pool_create_info.maxSets = max_descriptors_per_pool;
		Vector<VkDescriptorPoolSize> sizes;
		// Here comes more vulkan API strangeness.

		if (p_key.uniform_type[UNIFORM_TYPE_SAMPLER]) {
			VkDescriptorPoolSize s;
			s.type = VK_DESCRIPTOR_TYPE_SAMPLER;
			s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_SAMPLER] * max_descriptors_per_pool;
			sizes.push_back(s);
		}
		if (p_key.uniform_type[UNIFORM_TYPE_SAMPLER_WITH_TEXTURE]) {
			VkDescriptorPoolSize s;
			s.type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
			s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_SAMPLER_WITH_TEXTURE] * max_descriptors_per_pool;
			sizes.push_back(s);
		}
		if (p_key.uniform_type[UNIFORM_TYPE_TEXTURE]) {
			VkDescriptorPoolSize s;
			s.type = VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE;
			s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_TEXTURE] * max_descriptors_per_pool;
			sizes.push_back(s);
		}
		if (p_key.uniform_type[UNIFORM_TYPE_IMAGE]) {
			VkDescriptorPoolSize s;
			s.type = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE;
			s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_IMAGE] * max_descriptors_per_pool;
			sizes.push_back(s);
		}
		if (p_key.uniform_type[UNIFORM_TYPE_TEXTURE_BUFFER] || p_key.uniform_type[UNIFORM_TYPE_SAMPLER_WITH_TEXTURE_BUFFER]) {
			VkDescriptorPoolSize s;
			s.type = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER;
			s.descriptorCount = (p_key.uniform_type[UNIFORM_TYPE_TEXTURE_BUFFER] + p_key.uniform_type[UNIFORM_TYPE_SAMPLER_WITH_TEXTURE_BUFFER]) * max_descriptors_per_pool;
			sizes.push_back(s);
		}
		if (p_key.uniform_type[UNIFORM_TYPE_IMAGE_BUFFER]) {
			VkDescriptorPoolSize s;
			s.type = VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER;
			s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_IMAGE_BUFFER] * max_descriptors_per_pool;
			sizes.push_back(s);
		}
		if (p_key.uniform_type[UNIFORM_TYPE_UNIFORM_BUFFER]) {
			VkDescriptorPoolSize s;
			s.type = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
			s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_UNIFORM_BUFFER] * max_descriptors_per_pool;
			sizes.push_back(s);
		}

		if (p_key.uniform_type[UNIFORM_TYPE_STORAGE_BUFFER]) {
			VkDescriptorPoolSize s;
			s.type = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
			s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_STORAGE_BUFFER] * max_descriptors_per_pool;
			sizes.push_back(s);
		}

		if (p_key.uniform_type[UNIFORM_TYPE_INPUT_ATTACHMENT]) {
			VkDescriptorPoolSize s;
			s.type = VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT;
			s.descriptorCount = p_key.uniform_type[UNIFORM_TYPE_INPUT_ATTACHMENT] * max_descriptors_per_pool;
			sizes.push_back(s);
		}

		descriptor_pool_create_info.poolSizeCount = sizes.size();
		descriptor_pool_create_info.pPoolSizes = sizes.ptr();
		VkResult res = vkCreateDescriptorPool(device, &descriptor_pool_create_info, nullptr, &pool->pool);
		if (res) {
			memdelete(pool);
			ERR_FAIL_COND_V_MSG(res, nullptr, "vkCreateDescriptorPool failed with error " + itos(res) + ".");
		}
		descriptor_pools[p_key].insert(pool);
	}

	pool->usage++;

	return pool;
}

void RenderingDeviceVulkan::_descriptor_pool_free(const DescriptorPoolKey &p_key, DescriptorPool *p_pool) {
#ifdef DEBUG_ENABLED
	ERR_FAIL_COND(!descriptor_pools[p_key].has(p_pool));
#endif
	ERR_FAIL_COND(p_pool->usage == 0);
	p_pool->usage--;
	if (p_pool->usage == 0) {
		vkDestroyDescriptorPool(device, p_pool->pool, nullptr);
		descriptor_pools[p_key].erase(p_pool);
		memdelete(p_pool);
		if (descriptor_pools[p_key].is_empty()) {
			descriptor_pools.erase(p_key);
		}
	}
}

RID RenderingDeviceVulkan::uniform_set_create(const Vector<Uniform> &p_uniforms, RID p_shader, uint32_t p_shader_set) {
	_THREAD_SAFE_METHOD_

	ERR_FAIL_COND_V(p_uniforms.size() == 0, RID());

	Shader *shader = shader_owner.get_or_null(p_shader);
	ERR_FAIL_COND_V(!shader, RID());

	ERR_FAIL_COND_V_MSG(p_shader_set >= (uint32_t)shader->sets.size() || shader->sets[p_shader_set].uniform_info.size() == 0, RID(),
			"Desired set (" + itos(p_shader_set) + ") not used by shader.");
	// See that all sets in shader are satisfied.

	const Shader::Set &set = shader->sets[p_shader_set];

	uint32_t uniform_count = p_uniforms.size();
	const Uniform *uniforms = p_uniforms.ptr();

	uint32_t set_uniform_count = set.uniform_info.size();
	const UniformInfo *set_uniforms = set.uniform_info.ptr();

	Vector<VkWriteDescriptorSet> writes;
	DescriptorPoolKey pool_key;

	// To keep them alive until update call.
	List<Vector<VkDescriptorBufferInfo>> buffer_infos;
	List<Vector<VkBufferView>> buffer_views;
	List<Vector<VkDescriptorImageInfo>> image_infos;
	// Used for verification to make sure a uniform set does not use a framebuffer bound texture.
	LocalVector<UniformSet::AttachableTexture> attachable_textures;
	Vector<Texture *> mutable_sampled_textures;
	Vector<Texture *> mutable_storage_textures;

	for (uint32_t i = 0; i < set_uniform_count; i++) {
		const UniformInfo &set_uniform = set_uniforms[i];
		int uniform_idx = -1;
		for (int j = 0; j < (int)uniform_count; j++) {
			if (uniforms[j].binding == set_uniform.binding) {
				uniform_idx = j;
			}
		}
		ERR_FAIL_COND_V_MSG(uniform_idx == -1, RID(),
				"All the shader bindings for the given set must be covered by the uniforms provided. Binding (" + itos(set_uniform.binding) + "), set (" + itos(p_shader_set) + ") was not provided.");

		const Uniform &uniform = uniforms[uniform_idx];

		ERR_FAIL_COND_V_MSG(uniform.uniform_type != set_uniform.type, RID(),
				"Mismatch uniform type for binding (" + itos(set_uniform.binding) + "), set (" + itos(p_shader_set) + "). Expected '" + shader_uniform_names[set_uniform.type] + "', supplied: '" + shader_uniform_names[uniform.uniform_type] + "'.");

		VkWriteDescriptorSet write; // Common header.
		write.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;
		write.pNext = nullptr;
		write.dstSet = VK_NULL_HANDLE; // Will assign afterwards when everything is valid.
		write.dstBinding = set_uniform.binding;
		write.dstArrayElement = 0;
		write.descriptorCount = 0;
		write.descriptorType = VK_DESCRIPTOR_TYPE_MAX_ENUM; // Invalid value.
		write.pImageInfo = nullptr;
		write.pBufferInfo = nullptr;
		write.pTexelBufferView = nullptr;
		uint32_t type_size = 1;

		switch (uniform.uniform_type) {
			case UNIFORM_TYPE_SAMPLER: {
				if (uniform.get_id_count() != (uint32_t)set_uniform.length) {
					if (set_uniform.length > 1) {
						ERR_FAIL_V_MSG(RID(), "Sampler (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") sampler elements, so it should be provided equal number of sampler IDs to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
					} else {
						ERR_FAIL_V_MSG(RID(), "Sampler (binding: " + itos(uniform.binding) + ") should provide one ID referencing a sampler (IDs provided: " + itos(uniform.get_id_count()) + ").");
					}
				}

				Vector<VkDescriptorImageInfo> image_info;

				for (uint32_t j = 0; j < uniform.get_id_count(); j++) {
					VkSampler *sampler = sampler_owner.get_or_null(uniform.get_id(j));
					ERR_FAIL_COND_V_MSG(!sampler, RID(), "Sampler (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") is not a valid sampler.");

					VkDescriptorImageInfo img_info;
					img_info.sampler = *sampler;
					img_info.imageView = VK_NULL_HANDLE;
					img_info.imageLayout = VK_IMAGE_LAYOUT_UNDEFINED;

					image_info.push_back(img_info);
				}

				write.dstArrayElement = 0;
				write.descriptorCount = uniform.get_id_count();
				write.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLER;
				write.pImageInfo = image_infos.push_back(image_info)->get().ptr();
				write.pBufferInfo = nullptr;
				write.pTexelBufferView = nullptr;

				type_size = uniform.get_id_count();

			} break;
			case UNIFORM_TYPE_SAMPLER_WITH_TEXTURE: {
				if (uniform.get_id_count() != (uint32_t)set_uniform.length * 2) {
					if (set_uniform.length > 1) {
						ERR_FAIL_V_MSG(RID(), "SamplerTexture (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") sampler&texture elements, so it should provided twice the amount of IDs (sampler,texture pairs) to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
					} else {
						ERR_FAIL_V_MSG(RID(), "SamplerTexture (binding: " + itos(uniform.binding) + ") should provide two IDs referencing a sampler and then a texture (IDs provided: " + itos(uniform.get_id_count()) + ").");
					}
				}

				Vector<VkDescriptorImageInfo> image_info;

				for (uint32_t j = 0; j < uniform.get_id_count(); j += 2) {
					VkSampler *sampler = sampler_owner.get_or_null(uniform.get_id(j + 0));
					ERR_FAIL_COND_V_MSG(!sampler, RID(), "SamplerBuffer (binding: " + itos(uniform.binding) + ", index " + itos(j + 1) + ") is not a valid sampler.");

					Texture *texture = texture_owner.get_or_null(uniform.get_id(j + 1));
					ERR_FAIL_COND_V_MSG(!texture, RID(), "Texture (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") is not a valid texture.");

					ERR_FAIL_COND_V_MSG(!(texture->usage_flags & TEXTURE_USAGE_SAMPLING_BIT), RID(),
							"Texture (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") needs the TEXTURE_USAGE_SAMPLING_BIT usage flag set in order to be used as uniform.");

					VkDescriptorImageInfo img_info;
					img_info.sampler = *sampler;
					img_info.imageView = texture->view;

					if (texture->usage_flags & (TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | TEXTURE_USAGE_INPUT_ATTACHMENT_BIT)) {
						UniformSet::AttachableTexture attachable_texture;
						attachable_texture.bind = set_uniform.binding;
						attachable_texture.texture = texture->owner.is_valid() ? texture->owner : uniform.get_id(j + 1);
						attachable_textures.push_back(attachable_texture);
					}

					if (texture->usage_flags & TEXTURE_USAGE_STORAGE_BIT) {
						// Can also be used as storage, add to mutable sampled.
						mutable_sampled_textures.push_back(texture);
					}
					if (texture->owner.is_valid()) {
						texture = texture_owner.get_or_null(texture->owner);
						ERR_FAIL_COND_V(!texture, RID()); // Bug, should never happen.
					}

					img_info.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;

					image_info.push_back(img_info);
				}

				write.dstArrayElement = 0;
				write.descriptorCount = uniform.get_id_count() / 2;
				write.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
				write.pImageInfo = image_infos.push_back(image_info)->get().ptr();
				write.pBufferInfo = nullptr;
				write.pTexelBufferView = nullptr;

				type_size = uniform.get_id_count() / 2;

			} break;
			case UNIFORM_TYPE_TEXTURE: {
				if (uniform.get_id_count() != (uint32_t)set_uniform.length) {
					if (set_uniform.length > 1) {
						ERR_FAIL_V_MSG(RID(), "Texture (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") textures, so it should be provided equal number of texture IDs to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
					} else {
						ERR_FAIL_V_MSG(RID(), "Texture (binding: " + itos(uniform.binding) + ") should provide one ID referencing a texture (IDs provided: " + itos(uniform.get_id_count()) + ").");
					}
				}

				Vector<VkDescriptorImageInfo> image_info;

				for (uint32_t j = 0; j < uniform.get_id_count(); j++) {
					Texture *texture = texture_owner.get_or_null(uniform.get_id(j));
					ERR_FAIL_COND_V_MSG(!texture, RID(), "Texture (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") is not a valid texture.");

					ERR_FAIL_COND_V_MSG(!(texture->usage_flags & TEXTURE_USAGE_SAMPLING_BIT), RID(),
							"Texture (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") needs the TEXTURE_USAGE_SAMPLING_BIT usage flag set in order to be used as uniform.");

					VkDescriptorImageInfo img_info;
					img_info.sampler = VK_NULL_HANDLE;
					img_info.imageView = texture->view;

					if (texture->usage_flags & (TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | TEXTURE_USAGE_INPUT_ATTACHMENT_BIT)) {
						UniformSet::AttachableTexture attachable_texture;
						attachable_texture.bind = set_uniform.binding;
						attachable_texture.texture = texture->owner.is_valid() ? texture->owner : uniform.get_id(j);
						attachable_textures.push_back(attachable_texture);
					}

					if (texture->usage_flags & TEXTURE_USAGE_STORAGE_BIT) {
						// Can also be used as storage, add to mutable sampled.
						mutable_sampled_textures.push_back(texture);
					}

					if (texture->owner.is_valid()) {
						texture = texture_owner.get_or_null(texture->owner);
						ERR_FAIL_COND_V(!texture, RID()); // Bug, should never happen.
					}

					img_info.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;

					image_info.push_back(img_info);
				}

				write.dstArrayElement = 0;
				write.descriptorCount = uniform.get_id_count();
				write.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE;
				write.pImageInfo = image_infos.push_back(image_info)->get().ptr();
				write.pBufferInfo = nullptr;
				write.pTexelBufferView = nullptr;

				type_size = uniform.get_id_count();
			} break;
			case UNIFORM_TYPE_IMAGE: {
				if (uniform.get_id_count() != (uint32_t)set_uniform.length) {
					if (set_uniform.length > 1) {
						ERR_FAIL_V_MSG(RID(), "Image (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") textures, so it should be provided equal number of texture IDs to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
					} else {
						ERR_FAIL_V_MSG(RID(), "Image (binding: " + itos(uniform.binding) + ") should provide one ID referencing a texture (IDs provided: " + itos(uniform.get_id_count()) + ").");
					}
				}

				Vector<VkDescriptorImageInfo> image_info;

				for (uint32_t j = 0; j < uniform.get_id_count(); j++) {
					Texture *texture = texture_owner.get_or_null(uniform.get_id(j));

					ERR_FAIL_COND_V_MSG(!texture, RID(),
							"Image (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") is not a valid texture.");

					ERR_FAIL_COND_V_MSG(!(texture->usage_flags & TEXTURE_USAGE_STORAGE_BIT), RID(),
							"Image (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") needs the TEXTURE_USAGE_STORAGE_BIT usage flag set in order to be used as uniform.");

					VkDescriptorImageInfo img_info;
					img_info.sampler = VK_NULL_HANDLE;
					img_info.imageView = texture->view;

					if (texture->usage_flags & TEXTURE_USAGE_SAMPLING_BIT) {
						// Can also be used as storage, add to mutable sampled.
						mutable_storage_textures.push_back(texture);
					}

					if (texture->owner.is_valid()) {
						texture = texture_owner.get_or_null(texture->owner);
						ERR_FAIL_COND_V(!texture, RID()); // Bug, should never happen.
					}

					img_info.imageLayout = VK_IMAGE_LAYOUT_GENERAL;

					image_info.push_back(img_info);
				}

				write.dstArrayElement = 0;
				write.descriptorCount = uniform.get_id_count();
				write.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE;
				write.pImageInfo = image_infos.push_back(image_info)->get().ptr();
				write.pBufferInfo = nullptr;
				write.pTexelBufferView = nullptr;

				type_size = uniform.get_id_count();

			} break;
			case UNIFORM_TYPE_TEXTURE_BUFFER: {
				if (uniform.get_id_count() != (uint32_t)set_uniform.length) {
					if (set_uniform.length > 1) {
						ERR_FAIL_V_MSG(RID(), "Buffer (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") texture buffer elements, so it should be provided equal number of texture buffer IDs to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
					} else {
						ERR_FAIL_V_MSG(RID(), "Buffer (binding: " + itos(uniform.binding) + ") should provide one ID referencing a texture buffer (IDs provided: " + itos(uniform.get_id_count()) + ").");
					}
				}

				Vector<VkDescriptorBufferInfo> buffer_info;
				Vector<VkBufferView> buffer_view;

				for (uint32_t j = 0; j < uniform.get_id_count(); j++) {
					TextureBuffer *buffer = texture_buffer_owner.get_or_null(uniform.get_id(j));
					ERR_FAIL_COND_V_MSG(!buffer, RID(), "Texture Buffer (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") is not a valid texture buffer.");

					buffer_info.push_back(buffer->buffer.buffer_info);
					buffer_view.push_back(buffer->view);
				}

				write.dstArrayElement = 0;
				write.descriptorCount = uniform.get_id_count();
				write.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER;
				write.pImageInfo = nullptr;
				write.pBufferInfo = buffer_infos.push_back(buffer_info)->get().ptr();
				write.pTexelBufferView = buffer_views.push_back(buffer_view)->get().ptr();

				type_size = uniform.get_id_count();

			} break;
			case UNIFORM_TYPE_SAMPLER_WITH_TEXTURE_BUFFER: {
				if (uniform.get_id_count() != (uint32_t)set_uniform.length * 2) {
					if (set_uniform.length > 1) {
						ERR_FAIL_V_MSG(RID(), "SamplerBuffer (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") sampler buffer elements, so it should provided twice the amount of IDs (sampler,buffer pairs) to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
					} else {
						ERR_FAIL_V_MSG(RID(), "SamplerBuffer (binding: " + itos(uniform.binding) + ") should provide two IDs referencing a sampler and then a texture buffer (IDs provided: " + itos(uniform.get_id_count()) + ").");
					}
				}

				Vector<VkDescriptorImageInfo> image_info;
				Vector<VkDescriptorBufferInfo> buffer_info;
				Vector<VkBufferView> buffer_view;

				for (uint32_t j = 0; j < uniform.get_id_count(); j += 2) {
					VkSampler *sampler = sampler_owner.get_or_null(uniform.get_id(j + 0));
					ERR_FAIL_COND_V_MSG(!sampler, RID(), "SamplerBuffer (binding: " + itos(uniform.binding) + ", index " + itos(j + 1) + ") is not a valid sampler.");

					TextureBuffer *buffer = texture_buffer_owner.get_or_null(uniform.get_id(j + 1));

					VkDescriptorImageInfo img_info;
					img_info.sampler = *sampler;
					img_info.imageView = VK_NULL_HANDLE;
					img_info.imageLayout = VK_IMAGE_LAYOUT_UNDEFINED;

					image_info.push_back(img_info);

					ERR_FAIL_COND_V_MSG(!buffer, RID(), "SamplerBuffer (binding: " + itos(uniform.binding) + ", index " + itos(j + 1) + ") is not a valid texture buffer.");

					buffer_info.push_back(buffer->buffer.buffer_info);
					buffer_view.push_back(buffer->view);
				}

				write.dstArrayElement = 0;
				write.descriptorCount = uniform.get_id_count() / 2;
				write.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER;
				write.pImageInfo = image_infos.push_back(image_info)->get().ptr();
				write.pBufferInfo = buffer_infos.push_back(buffer_info)->get().ptr();
				write.pTexelBufferView = buffer_views.push_back(buffer_view)->get().ptr();

				type_size = uniform.get_id_count() / 2;
			} break;
			case UNIFORM_TYPE_IMAGE_BUFFER: {
				// Todo.

			} break;
			case UNIFORM_TYPE_UNIFORM_BUFFER: {
				ERR_FAIL_COND_V_MSG(uniform.get_id_count() != 1, RID(),
						"Uniform buffer supplied (binding: " + itos(uniform.binding) + ") must provide one ID (" + itos(uniform.get_id_count()) + " provided).");

				Buffer *buffer = uniform_buffer_owner.get_or_null(uniform.get_id(0));
				ERR_FAIL_COND_V_MSG(!buffer, RID(), "Uniform buffer supplied (binding: " + itos(uniform.binding) + ") is invalid.");

				ERR_FAIL_COND_V_MSG(buffer->size != (uint32_t)set_uniform.length, RID(),
						"Uniform buffer supplied (binding: " + itos(uniform.binding) + ") size (" + itos(buffer->size) + " does not match size of shader uniform: (" + itos(set_uniform.length) + ").");

				write.dstArrayElement = 0;
				write.descriptorCount = 1;
				write.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
				write.pImageInfo = nullptr;
				write.pBufferInfo = &buffer->buffer_info;
				write.pTexelBufferView = nullptr;

			} break;
			case UNIFORM_TYPE_STORAGE_BUFFER: {
				ERR_FAIL_COND_V_MSG(uniform.get_id_count() != 1, RID(),
						"Storage buffer supplied (binding: " + itos(uniform.binding) + ") must provide one ID (" + itos(uniform.get_id_count()) + " provided).");

				Buffer *buffer = nullptr;

				if (storage_buffer_owner.owns(uniform.get_id(0))) {
					buffer = storage_buffer_owner.get_or_null(uniform.get_id(0));
				} else if (vertex_buffer_owner.owns(uniform.get_id(0))) {
					buffer = vertex_buffer_owner.get_or_null(uniform.get_id(0));

					ERR_FAIL_COND_V_MSG(!(buffer->usage & VK_BUFFER_USAGE_STORAGE_BUFFER_BIT), RID(), "Vertex buffer supplied (binding: " + itos(uniform.binding) + ") was not created with storage flag.");
				}
				ERR_FAIL_COND_V_MSG(!buffer, RID(), "Storage buffer supplied (binding: " + itos(uniform.binding) + ") is invalid.");

				// If 0, then it's sized on link time.
				ERR_FAIL_COND_V_MSG(set_uniform.length > 0 && buffer->size != (uint32_t)set_uniform.length, RID(),
						"Storage buffer supplied (binding: " + itos(uniform.binding) + ") size (" + itos(buffer->size) + " does not match size of shader uniform: (" + itos(set_uniform.length) + ").");

				write.dstArrayElement = 0;
				write.descriptorCount = 1;
				write.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER;
				write.pImageInfo = nullptr;
				write.pBufferInfo = &buffer->buffer_info;
				write.pTexelBufferView = nullptr;
			} break;
			case UNIFORM_TYPE_INPUT_ATTACHMENT: {
				ERR_FAIL_COND_V_MSG(shader->is_compute, RID(), "InputAttachment (binding: " + itos(uniform.binding) + ") supplied for compute shader (this is not allowed).");

				if (uniform.get_id_count() != (uint32_t)set_uniform.length) {
					if (set_uniform.length > 1) {
						ERR_FAIL_V_MSG(RID(), "InputAttachment (binding: " + itos(uniform.binding) + ") is an array of (" + itos(set_uniform.length) + ") textures, so it should be provided equal number of texture IDs to satisfy it (IDs provided: " + itos(uniform.get_id_count()) + ").");
					} else {
						ERR_FAIL_V_MSG(RID(), "InputAttachment (binding: " + itos(uniform.binding) + ") should provide one ID referencing a texture (IDs provided: " + itos(uniform.get_id_count()) + ").");
					}
				}

				Vector<VkDescriptorImageInfo> image_info;

				for (uint32_t j = 0; j < uniform.get_id_count(); j++) {
					Texture *texture = texture_owner.get_or_null(uniform.get_id(j));

					ERR_FAIL_COND_V_MSG(!texture, RID(),
							"InputAttachment (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") is not a valid texture.");

					ERR_FAIL_COND_V_MSG(!(texture->usage_flags & TEXTURE_USAGE_SAMPLING_BIT), RID(),
							"InputAttachment (binding: " + itos(uniform.binding) + ", index " + itos(j) + ") needs the TEXTURE_USAGE_SAMPLING_BIT usage flag set in order to be used as uniform.");

					VkDescriptorImageInfo img_info;
					img_info.sampler = VK_NULL_HANDLE;
					img_info.imageView = texture->view;

					if (texture->owner.is_valid()) {
						texture = texture_owner.get_or_null(texture->owner);
						ERR_FAIL_COND_V(!texture, RID()); // Bug, should never happen.
					}

					img_info.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;

					image_info.push_back(img_info);
				}

				write.dstArrayElement = 0;
				write.descriptorCount = uniform.get_id_count();
				write.descriptorType = VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT;
				write.pImageInfo = image_infos.push_back(image_info)->get().ptr();
				write.pBufferInfo = nullptr;
				write.pTexelBufferView = nullptr;

				type_size = uniform.get_id_count();
			} break;
			default: {
			}
		}

		writes.push_back(write);

		ERR_FAIL_COND_V_MSG(pool_key.uniform_type[set_uniform.type] == MAX_DESCRIPTOR_POOL_ELEMENT, RID(),
				"Uniform set reached the limit of bindings for the same type (" + itos(MAX_DESCRIPTOR_POOL_ELEMENT) + ").");
		pool_key.uniform_type[set_uniform.type] += type_size;
	}

	// Need a descriptor pool.
	DescriptorPool *pool = _descriptor_pool_allocate(pool_key);

	ERR_FAIL_COND_V(!pool, RID());

	VkDescriptorSetAllocateInfo descriptor_set_allocate_info;

	descriptor_set_allocate_info.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO;
	descriptor_set_allocate_info.pNext = nullptr;
	descriptor_set_allocate_info.descriptorPool = pool->pool;
	descriptor_set_allocate_info.descriptorSetCount = 1;
	descriptor_set_allocate_info.pSetLayouts = &shader->sets[p_shader_set].descriptor_set_layout;

	VkDescriptorSet descriptor_set;

	VkResult res = vkAllocateDescriptorSets(device, &descriptor_set_allocate_info, &descriptor_set);
	if (res) {
		_descriptor_pool_free(pool_key, pool); // Meh.
		ERR_FAIL_V_MSG(RID(), "Cannot allocate descriptor sets, error " + itos(res) + ".");
	}

	UniformSet uniform_set;
	uniform_set.pool = pool;
	uniform_set.pool_key = pool_key;
	uniform_set.descriptor_set = descriptor_set;
	uniform_set.format = shader->set_formats[p_shader_set];
	uniform_set.attachable_textures = attachable_textures;
	uniform_set.mutable_sampled_textures = mutable_sampled_textures;
	uniform_set.mutable_storage_textures = mutable_storage_textures;
	uniform_set.shader_set = p_shader_set;
	uniform_set.shader_id = p_shader;

	RID id = uniform_set_owner.make_rid(uniform_set);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif
	// Add dependencies.
	_add_dependency(id, p_shader);
	for (uint32_t i = 0; i < uniform_count; i++) {
		const Uniform &uniform = uniforms[i];
		int id_count = uniform.get_id_count();
		for (int j = 0; j < id_count; j++) {
			_add_dependency(id, uniform.get_id(j));
		}
	}

	// Write the contents.
	if (writes.size()) {
		for (int i = 0; i < writes.size(); i++) {
			writes.write[i].dstSet = descriptor_set;
		}
		vkUpdateDescriptorSets(device, writes.size(), writes.ptr(), 0, nullptr);
	}

	return id;
}

bool RenderingDeviceVulkan::uniform_set_is_valid(RID p_uniform_set) {
	return uniform_set_owner.owns(p_uniform_set);
}

void RenderingDeviceVulkan::uniform_set_set_invalidation_callback(RID p_uniform_set, InvalidationCallback p_callback, void *p_userdata) {
	UniformSet *us = uniform_set_owner.get_or_null(p_uniform_set);
	ERR_FAIL_COND(!us);
	us->invalidated_callback = p_callback;
	us->invalidated_callback_userdata = p_userdata;
}

Error RenderingDeviceVulkan::buffer_update(RID p_buffer, uint32_t p_offset, uint32_t p_size, const void *p_data, BitField<BarrierMask> p_post_barrier) {
	_THREAD_SAFE_METHOD_

	ERR_FAIL_COND_V_MSG(draw_list, ERR_INVALID_PARAMETER,
			"Updating buffers is forbidden during creation of a draw list");
	ERR_FAIL_COND_V_MSG(compute_list, ERR_INVALID_PARAMETER,
			"Updating buffers is forbidden during creation of a compute list");

	VkPipelineStageFlags dst_stage_mask = 0;
	VkAccessFlags dst_access = 0;
	if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) {
		// Protect subsequent updates.
		dst_stage_mask = VK_PIPELINE_STAGE_TRANSFER_BIT;
		dst_access = VK_ACCESS_TRANSFER_WRITE_BIT;
	}
	Buffer *buffer = _get_buffer_from_owner(p_buffer, dst_stage_mask, dst_access, p_post_barrier);
	if (!buffer) {
		ERR_FAIL_V_MSG(ERR_INVALID_PARAMETER, "Buffer argument is not a valid buffer of any type.");
	}

	ERR_FAIL_COND_V_MSG(p_offset + p_size > buffer->size, ERR_INVALID_PARAMETER,
			"Attempted to write buffer (" + itos((p_offset + p_size) - buffer->size) + " bytes) past the end.");

	// No barrier should be needed here.
	// _buffer_memory_barrier(buffer->buffer, p_offset, p_size, dst_stage_mask, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_access, VK_ACCESS_TRANSFER_WRITE_BIT, true);

	Error err = _buffer_update(buffer, p_offset, (uint8_t *)p_data, p_size, p_post_barrier);
	if (err) {
		return err;
	}

#ifdef FORCE_FULL_BARRIER
	_full_barrier(true);
#else
	if (dst_stage_mask == 0) {
		dst_stage_mask = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
	}

	if (p_post_barrier != RD::BARRIER_MASK_NO_BARRIER) {
		_buffer_memory_barrier(buffer->buffer, p_offset, p_size, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_stage_mask, VK_ACCESS_TRANSFER_WRITE_BIT, dst_access, true);
	}

#endif
	return err;
}

Error RenderingDeviceVulkan::buffer_clear(RID p_buffer, uint32_t p_offset, uint32_t p_size, BitField<BarrierMask> p_post_barrier) {
	_THREAD_SAFE_METHOD_

	ERR_FAIL_COND_V_MSG((p_size % 4) != 0, ERR_INVALID_PARAMETER,
			"Size must be a multiple of four");
	ERR_FAIL_COND_V_MSG(draw_list, ERR_INVALID_PARAMETER,
			"Updating buffers in is forbidden during creation of a draw list");
	ERR_FAIL_COND_V_MSG(compute_list, ERR_INVALID_PARAMETER,
			"Updating buffers is forbidden during creation of a compute list");

	VkPipelineStageFlags dst_stage_mask = 0;
	VkAccessFlags dst_access = 0;
	if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) {
		// Protect subsequent updates.
		dst_stage_mask = VK_PIPELINE_STAGE_TRANSFER_BIT;
		dst_access = VK_ACCESS_TRANSFER_WRITE_BIT;
	}

	Buffer *buffer = _get_buffer_from_owner(p_buffer, dst_stage_mask, dst_access, p_post_barrier);
	if (!buffer) {
		ERR_FAIL_V_MSG(ERR_INVALID_PARAMETER, "Buffer argument is not a valid buffer of any type.");
	}

	ERR_FAIL_COND_V_MSG(p_offset + p_size > buffer->size, ERR_INVALID_PARAMETER,
			"Attempted to write buffer (" + itos((p_offset + p_size) - buffer->size) + " bytes) past the end.");

	// Should not be needed.
	// _buffer_memory_barrier(buffer->buffer, p_offset, p_size, dst_stage_mask, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_access, VK_ACCESS_TRANSFER_WRITE_BIT, p_post_barrier);

	vkCmdFillBuffer(frames[frame].draw_command_buffer, buffer->buffer, p_offset, p_size, 0);

#ifdef FORCE_FULL_BARRIER
	_full_barrier(true);
#else
	if (dst_stage_mask == 0) {
		dst_stage_mask = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
	}

	_buffer_memory_barrier(buffer->buffer, p_offset, p_size, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_stage_mask, VK_ACCESS_TRANSFER_WRITE_BIT, dst_access, dst_stage_mask);

#endif
	return OK;
}

Vector<uint8_t> RenderingDeviceVulkan::buffer_get_data(RID p_buffer) {
	_THREAD_SAFE_METHOD_

	// It could be this buffer was just created.
	VkPipelineShaderStageCreateFlags src_stage_mask = VK_PIPELINE_STAGE_TRANSFER_BIT;
	VkAccessFlags src_access_mask = VK_ACCESS_TRANSFER_WRITE_BIT;
	// Get the vulkan buffer and the potential stage/access possible.
	Buffer *buffer = _get_buffer_from_owner(p_buffer, src_stage_mask, src_access_mask, BARRIER_MASK_ALL_BARRIERS);
	if (!buffer) {
		ERR_FAIL_V_MSG(Vector<uint8_t>(), "Buffer is either invalid or this type of buffer can't be retrieved. Only Index and Vertex buffers allow retrieving.");
	}

	// Make sure no one is using the buffer -- the "false" gets us to the same command buffer as below.
	_buffer_memory_barrier(buffer->buffer, 0, buffer->size, src_stage_mask, src_access_mask, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_READ_BIT, false);

	VkCommandBuffer command_buffer = frames[frame].setup_command_buffer;

	Buffer tmp_buffer;
	_buffer_allocate(&tmp_buffer, buffer->size, VK_BUFFER_USAGE_TRANSFER_DST_BIT, VMA_MEMORY_USAGE_AUTO_PREFER_HOST, VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT);
	VkBufferCopy region;
	region.srcOffset = 0;
	region.dstOffset = 0;
	region.size = buffer->size;
	vkCmdCopyBuffer(command_buffer, buffer->buffer, tmp_buffer.buffer, 1, &region); // Dst buffer is in CPU, but I wonder if src buffer needs a barrier for this.
	// Flush everything so memory can be safely mapped.
	_flush(true);

	void *buffer_mem;
	VkResult vkerr = vmaMapMemory(allocator, tmp_buffer.allocation, &buffer_mem);
	ERR_FAIL_COND_V_MSG(vkerr, Vector<uint8_t>(), "vmaMapMemory failed with error " + itos(vkerr) + ".");

	Vector<uint8_t> buffer_data;
	{
		buffer_data.resize(buffer->size);
		uint8_t *w = buffer_data.ptrw();
		memcpy(w, buffer_mem, buffer->size);
	}

	vmaUnmapMemory(allocator, tmp_buffer.allocation);

	_buffer_free(&tmp_buffer);

	return buffer_data;
}

/*************************/
/**** RENDER PIPELINE ****/
/*************************/

RID RenderingDeviceVulkan::render_pipeline_create(RID p_shader, FramebufferFormatID p_framebuffer_format, VertexFormatID p_vertex_format, RenderPrimitive p_render_primitive, const PipelineRasterizationState &p_rasterization_state, const PipelineMultisampleState &p_multisample_state, const PipelineDepthStencilState &p_depth_stencil_state, const PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags, uint32_t p_for_render_pass, const Vector<PipelineSpecializationConstant> &p_specialization_constants) {
	_THREAD_SAFE_METHOD_

	// Needs a shader.
	Shader *shader = shader_owner.get_or_null(p_shader);
	ERR_FAIL_COND_V(!shader, RID());

	ERR_FAIL_COND_V_MSG(shader->is_compute, RID(),
			"Compute shaders can't be used in render pipelines");

	if (p_framebuffer_format == INVALID_ID) {
		// If nothing provided, use an empty one (no attachments).
		p_framebuffer_format = framebuffer_format_create(Vector<AttachmentFormat>());
	}
	ERR_FAIL_COND_V(!framebuffer_formats.has(p_framebuffer_format), RID());
	const FramebufferFormat &fb_format = framebuffer_formats[p_framebuffer_format];

	{ // Validate shader vs framebuffer.

		ERR_FAIL_COND_V_MSG(p_for_render_pass >= uint32_t(fb_format.E->key().passes.size()), RID(), "Render pass requested for pipeline creation (" + itos(p_for_render_pass) + ") is out of bounds");
		const FramebufferPass &pass = fb_format.E->key().passes[p_for_render_pass];
		uint32_t output_mask = 0;
		for (int i = 0; i < pass.color_attachments.size(); i++) {
			if (pass.color_attachments[i] != FramebufferPass::ATTACHMENT_UNUSED) {
				output_mask |= 1 << i;
			}
		}
		ERR_FAIL_COND_V_MSG(shader->fragment_output_mask != output_mask, RID(),
				"Mismatch fragment shader output mask (" + itos(shader->fragment_output_mask) + ") and framebuffer color output mask (" + itos(output_mask) + ") when binding both in render pipeline.");
	}
	// Vertex.
	VkPipelineVertexInputStateCreateInfo pipeline_vertex_input_state_create_info;

	if (p_vertex_format != INVALID_ID) {
		// Uses vertices, else it does not.
		ERR_FAIL_COND_V(!vertex_formats.has(p_vertex_format), RID());
		const VertexDescriptionCache &vd = vertex_formats[p_vertex_format];

		pipeline_vertex_input_state_create_info = vd.create_info;

		// Validate with inputs.
		for (uint32_t i = 0; i < 32; i++) {
			if (!(shader->vertex_input_mask & (1UL << i))) {
				continue;
			}
			bool found = false;
			for (int j = 0; j < vd.vertex_formats.size(); j++) {
				if (vd.vertex_formats[j].location == i) {
					found = true;
				}
			}

			ERR_FAIL_COND_V_MSG(!found, RID(),
					"Shader vertex input location (" + itos(i) + ") not provided in vertex input description for pipeline creation.");
		}

	} else {
		// Does not use vertices.
		pipeline_vertex_input_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO;
		pipeline_vertex_input_state_create_info.pNext = nullptr;
		pipeline_vertex_input_state_create_info.flags = 0;
		pipeline_vertex_input_state_create_info.vertexBindingDescriptionCount = 0;
		pipeline_vertex_input_state_create_info.pVertexBindingDescriptions = nullptr;
		pipeline_vertex_input_state_create_info.vertexAttributeDescriptionCount = 0;
		pipeline_vertex_input_state_create_info.pVertexAttributeDescriptions = nullptr;

		ERR_FAIL_COND_V_MSG(shader->vertex_input_mask != 0, RID(),
				"Shader contains vertex inputs, but no vertex input description was provided for pipeline creation.");
	}
	// Input assembly.

	ERR_FAIL_INDEX_V(p_render_primitive, RENDER_PRIMITIVE_MAX, RID());

	VkPipelineInputAssemblyStateCreateInfo input_assembly_create_info;
	input_assembly_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO;
	input_assembly_create_info.pNext = nullptr;
	input_assembly_create_info.flags = 0;

	static const VkPrimitiveTopology topology_list[RENDER_PRIMITIVE_MAX] = {
		VK_PRIMITIVE_TOPOLOGY_POINT_LIST,
		VK_PRIMITIVE_TOPOLOGY_LINE_LIST,
		VK_PRIMITIVE_TOPOLOGY_LINE_LIST_WITH_ADJACENCY,
		VK_PRIMITIVE_TOPOLOGY_LINE_STRIP,
		VK_PRIMITIVE_TOPOLOGY_LINE_STRIP_WITH_ADJACENCY,
		VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST,
		VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST_WITH_ADJACENCY,
		VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP,
		VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP_WITH_ADJACENCY,
		VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP,
		VK_PRIMITIVE_TOPOLOGY_PATCH_LIST
	};

	input_assembly_create_info.topology = topology_list[p_render_primitive];
	input_assembly_create_info.primitiveRestartEnable = (p_render_primitive == RENDER_PRIMITIVE_TRIANGLE_STRIPS_WITH_RESTART_INDEX);

	// Tessellation.
	VkPipelineTessellationStateCreateInfo tessellation_create_info;
	tessellation_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO;
	tessellation_create_info.pNext = nullptr;
	tessellation_create_info.flags = 0;
	ERR_FAIL_COND_V(limits.maxTessellationPatchSize > 0 && (p_rasterization_state.patch_control_points < 1 || p_rasterization_state.patch_control_points > limits.maxTessellationPatchSize), RID());
	tessellation_create_info.patchControlPoints = p_rasterization_state.patch_control_points;

	VkPipelineViewportStateCreateInfo viewport_state_create_info;
	viewport_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO;
	viewport_state_create_info.pNext = nullptr;
	viewport_state_create_info.flags = 0;
	viewport_state_create_info.viewportCount = 1; // If VR extensions are supported at some point, this will have to be customizable in the framebuffer format.
	viewport_state_create_info.pViewports = nullptr;
	viewport_state_create_info.scissorCount = 1;
	viewport_state_create_info.pScissors = nullptr;

	// Rasterization.
	VkPipelineRasterizationStateCreateInfo rasterization_state_create_info;
	rasterization_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO;
	rasterization_state_create_info.pNext = nullptr;
	rasterization_state_create_info.flags = 0;
	rasterization_state_create_info.depthClampEnable = p_rasterization_state.enable_depth_clamp;
	rasterization_state_create_info.rasterizerDiscardEnable = p_rasterization_state.discard_primitives;
	rasterization_state_create_info.polygonMode = (p_rasterization_state.wireframe ? VK_POLYGON_MODE_LINE : VK_POLYGON_MODE_FILL);
	static const VkCullModeFlags cull_mode[3] = {
		VK_CULL_MODE_NONE,
		VK_CULL_MODE_FRONT_BIT,
		VK_CULL_MODE_BACK_BIT
	};

	ERR_FAIL_INDEX_V(p_rasterization_state.cull_mode, 3, RID());
	rasterization_state_create_info.cullMode = cull_mode[p_rasterization_state.cull_mode];
	rasterization_state_create_info.frontFace = (p_rasterization_state.front_face == POLYGON_FRONT_FACE_CLOCKWISE ? VK_FRONT_FACE_CLOCKWISE : VK_FRONT_FACE_COUNTER_CLOCKWISE);
	rasterization_state_create_info.depthBiasEnable = p_rasterization_state.depth_bias_enabled;
	rasterization_state_create_info.depthBiasConstantFactor = p_rasterization_state.depth_bias_constant_factor;
	rasterization_state_create_info.depthBiasClamp = p_rasterization_state.depth_bias_clamp;
	rasterization_state_create_info.depthBiasSlopeFactor = p_rasterization_state.depth_bias_slope_factor;
	rasterization_state_create_info.lineWidth = p_rasterization_state.line_width;

	// Multisample.
	VkPipelineMultisampleStateCreateInfo multisample_state_create_info;
	multisample_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO;
	multisample_state_create_info.pNext = nullptr;
	multisample_state_create_info.flags = 0;

	multisample_state_create_info.rasterizationSamples = _ensure_supported_sample_count(p_multisample_state.sample_count);
	multisample_state_create_info.sampleShadingEnable = p_multisample_state.enable_sample_shading;
	multisample_state_create_info.minSampleShading = p_multisample_state.min_sample_shading;
	Vector<VkSampleMask> sample_mask;
	if (p_multisample_state.sample_mask.size()) {
		// Use sample mask.
		const int rasterization_sample_mask_expected_size[TEXTURE_SAMPLES_MAX] = {
			1, 2, 4, 8, 16, 32, 64
		};
		ERR_FAIL_COND_V(rasterization_sample_mask_expected_size[p_multisample_state.sample_count] != p_multisample_state.sample_mask.size(), RID());
		sample_mask.resize(p_multisample_state.sample_mask.size());
		for (int i = 0; i < p_multisample_state.sample_mask.size(); i++) {
			VkSampleMask mask = p_multisample_state.sample_mask[i];
			sample_mask.push_back(mask);
		}
		multisample_state_create_info.pSampleMask = sample_mask.ptr();
	} else {
		multisample_state_create_info.pSampleMask = nullptr;
	}

	multisample_state_create_info.alphaToCoverageEnable = p_multisample_state.enable_alpha_to_coverage;
	multisample_state_create_info.alphaToOneEnable = p_multisample_state.enable_alpha_to_one;

	// Depth stencil.

	VkPipelineDepthStencilStateCreateInfo depth_stencil_state_create_info;
	depth_stencil_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO;
	depth_stencil_state_create_info.pNext = nullptr;
	depth_stencil_state_create_info.flags = 0;
	depth_stencil_state_create_info.depthTestEnable = p_depth_stencil_state.enable_depth_test;
	depth_stencil_state_create_info.depthWriteEnable = p_depth_stencil_state.enable_depth_write;
	ERR_FAIL_INDEX_V(p_depth_stencil_state.depth_compare_operator, COMPARE_OP_MAX, RID());
	depth_stencil_state_create_info.depthCompareOp = compare_operators[p_depth_stencil_state.depth_compare_operator];
	depth_stencil_state_create_info.depthBoundsTestEnable = p_depth_stencil_state.enable_depth_range;
	depth_stencil_state_create_info.stencilTestEnable = p_depth_stencil_state.enable_stencil;

	ERR_FAIL_INDEX_V(p_depth_stencil_state.front_op.fail, STENCIL_OP_MAX, RID());
	depth_stencil_state_create_info.front.failOp = stencil_operations[p_depth_stencil_state.front_op.fail];
	ERR_FAIL_INDEX_V(p_depth_stencil_state.front_op.pass, STENCIL_OP_MAX, RID());
	depth_stencil_state_create_info.front.passOp = stencil_operations[p_depth_stencil_state.front_op.pass];
	ERR_FAIL_INDEX_V(p_depth_stencil_state.front_op.depth_fail, STENCIL_OP_MAX, RID());
	depth_stencil_state_create_info.front.depthFailOp = stencil_operations[p_depth_stencil_state.front_op.depth_fail];
	ERR_FAIL_INDEX_V(p_depth_stencil_state.front_op.compare, COMPARE_OP_MAX, RID());
	depth_stencil_state_create_info.front.compareOp = compare_operators[p_depth_stencil_state.front_op.compare];
	depth_stencil_state_create_info.front.compareMask = p_depth_stencil_state.front_op.compare_mask;
	depth_stencil_state_create_info.front.writeMask = p_depth_stencil_state.front_op.write_mask;
	depth_stencil_state_create_info.front.reference = p_depth_stencil_state.front_op.reference;

	ERR_FAIL_INDEX_V(p_depth_stencil_state.back_op.fail, STENCIL_OP_MAX, RID());
	depth_stencil_state_create_info.back.failOp = stencil_operations[p_depth_stencil_state.back_op.fail];
	ERR_FAIL_INDEX_V(p_depth_stencil_state.back_op.pass, STENCIL_OP_MAX, RID());
	depth_stencil_state_create_info.back.passOp = stencil_operations[p_depth_stencil_state.back_op.pass];
	ERR_FAIL_INDEX_V(p_depth_stencil_state.back_op.depth_fail, STENCIL_OP_MAX, RID());
	depth_stencil_state_create_info.back.depthFailOp = stencil_operations[p_depth_stencil_state.back_op.depth_fail];
	ERR_FAIL_INDEX_V(p_depth_stencil_state.back_op.compare, COMPARE_OP_MAX, RID());
	depth_stencil_state_create_info.back.compareOp = compare_operators[p_depth_stencil_state.back_op.compare];
	depth_stencil_state_create_info.back.compareMask = p_depth_stencil_state.back_op.compare_mask;
	depth_stencil_state_create_info.back.writeMask = p_depth_stencil_state.back_op.write_mask;
	depth_stencil_state_create_info.back.reference = p_depth_stencil_state.back_op.reference;

	depth_stencil_state_create_info.minDepthBounds = p_depth_stencil_state.depth_range_min;
	depth_stencil_state_create_info.maxDepthBounds = p_depth_stencil_state.depth_range_max;

	// Blend state.
	VkPipelineColorBlendStateCreateInfo color_blend_state_create_info;
	color_blend_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO;
	color_blend_state_create_info.pNext = nullptr;
	color_blend_state_create_info.flags = 0;
	color_blend_state_create_info.logicOpEnable = p_blend_state.enable_logic_op;
	ERR_FAIL_INDEX_V(p_blend_state.logic_op, LOGIC_OP_MAX, RID());
	color_blend_state_create_info.logicOp = logic_operations[p_blend_state.logic_op];

	Vector<VkPipelineColorBlendAttachmentState> attachment_states;
	{
		const FramebufferPass &pass = fb_format.E->key().passes[p_for_render_pass];
		attachment_states.resize(pass.color_attachments.size());
		ERR_FAIL_COND_V(p_blend_state.attachments.size() < pass.color_attachments.size(), RID());
		for (int i = 0; i < pass.color_attachments.size(); i++) {
			VkPipelineColorBlendAttachmentState state;
			if (pass.color_attachments[i] == FramebufferPass::ATTACHMENT_UNUSED) {
				state.blendEnable = false;

				state.srcColorBlendFactor = VK_BLEND_FACTOR_ZERO;
				state.dstColorBlendFactor = VK_BLEND_FACTOR_ZERO;
				state.colorBlendOp = VK_BLEND_OP_ADD;

				state.srcAlphaBlendFactor = VK_BLEND_FACTOR_ZERO;
				state.dstAlphaBlendFactor = VK_BLEND_FACTOR_ZERO;
				state.alphaBlendOp = VK_BLEND_OP_ADD;

				state.colorWriteMask = 0;
			} else {
				state.blendEnable = p_blend_state.attachments[i].enable_blend;

				ERR_FAIL_INDEX_V(p_blend_state.attachments[i].src_color_blend_factor, BLEND_FACTOR_MAX, RID());
				state.srcColorBlendFactor = blend_factors[p_blend_state.attachments[i].src_color_blend_factor];
				ERR_FAIL_INDEX_V(p_blend_state.attachments[i].dst_color_blend_factor, BLEND_FACTOR_MAX, RID());
				state.dstColorBlendFactor = blend_factors[p_blend_state.attachments[i].dst_color_blend_factor];
				ERR_FAIL_INDEX_V(p_blend_state.attachments[i].color_blend_op, BLEND_OP_MAX, RID());
				state.colorBlendOp = blend_operations[p_blend_state.attachments[i].color_blend_op];

				ERR_FAIL_INDEX_V(p_blend_state.attachments[i].src_alpha_blend_factor, BLEND_FACTOR_MAX, RID());
				state.srcAlphaBlendFactor = blend_factors[p_blend_state.attachments[i].src_alpha_blend_factor];
				ERR_FAIL_INDEX_V(p_blend_state.attachments[i].dst_alpha_blend_factor, BLEND_FACTOR_MAX, RID());
				state.dstAlphaBlendFactor = blend_factors[p_blend_state.attachments[i].dst_alpha_blend_factor];
				ERR_FAIL_INDEX_V(p_blend_state.attachments[i].alpha_blend_op, BLEND_OP_MAX, RID());
				state.alphaBlendOp = blend_operations[p_blend_state.attachments[i].alpha_blend_op];

				state.colorWriteMask = 0;
				if (p_blend_state.attachments[i].write_r) {
					state.colorWriteMask |= VK_COLOR_COMPONENT_R_BIT;
				}
				if (p_blend_state.attachments[i].write_g) {
					state.colorWriteMask |= VK_COLOR_COMPONENT_G_BIT;
				}
				if (p_blend_state.attachments[i].write_b) {
					state.colorWriteMask |= VK_COLOR_COMPONENT_B_BIT;
				}
				if (p_blend_state.attachments[i].write_a) {
					state.colorWriteMask |= VK_COLOR_COMPONENT_A_BIT;
				}
			}
			attachment_states.write[i] = state;
		}
	}

	color_blend_state_create_info.attachmentCount = attachment_states.size();
	color_blend_state_create_info.pAttachments = attachment_states.ptr();

	color_blend_state_create_info.blendConstants[0] = p_blend_state.blend_constant.r;
	color_blend_state_create_info.blendConstants[1] = p_blend_state.blend_constant.g;
	color_blend_state_create_info.blendConstants[2] = p_blend_state.blend_constant.b;
	color_blend_state_create_info.blendConstants[3] = p_blend_state.blend_constant.a;

	// Dynamic state.

	VkPipelineDynamicStateCreateInfo dynamic_state_create_info;
	dynamic_state_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO;
	dynamic_state_create_info.pNext = nullptr;
	dynamic_state_create_info.flags = 0;
	Vector<VkDynamicState> dynamic_states; // Vulkan is weird.

	dynamic_states.push_back(VK_DYNAMIC_STATE_VIEWPORT); // Viewport and scissor are always dynamic.
	dynamic_states.push_back(VK_DYNAMIC_STATE_SCISSOR);

	if (p_dynamic_state_flags & DYNAMIC_STATE_LINE_WIDTH) {
		dynamic_states.push_back(VK_DYNAMIC_STATE_LINE_WIDTH);
	}

	if (p_dynamic_state_flags & DYNAMIC_STATE_DEPTH_BIAS) {
		dynamic_states.push_back(VK_DYNAMIC_STATE_DEPTH_BIAS);
	}

	if (p_dynamic_state_flags & DYNAMIC_STATE_BLEND_CONSTANTS) {
		dynamic_states.push_back(VK_DYNAMIC_STATE_BLEND_CONSTANTS);
	}

	if (p_dynamic_state_flags & DYNAMIC_STATE_DEPTH_BOUNDS) {
		dynamic_states.push_back(VK_DYNAMIC_STATE_DEPTH_BOUNDS);
	}

	if (p_dynamic_state_flags & DYNAMIC_STATE_STENCIL_COMPARE_MASK) {
		dynamic_states.push_back(VK_DYNAMIC_STATE_STENCIL_COMPARE_MASK);
	}

	if (p_dynamic_state_flags & DYNAMIC_STATE_STENCIL_WRITE_MASK) {
		dynamic_states.push_back(VK_DYNAMIC_STATE_STENCIL_WRITE_MASK);
	}

	if (p_dynamic_state_flags & DYNAMIC_STATE_STENCIL_REFERENCE) {
		dynamic_states.push_back(VK_DYNAMIC_STATE_STENCIL_REFERENCE);
	}

	dynamic_state_create_info.dynamicStateCount = dynamic_states.size();
	dynamic_state_create_info.pDynamicStates = dynamic_states.ptr();

	void *graphics_pipeline_nextptr = nullptr;

	VkPipelineFragmentShadingRateStateCreateInfoKHR vrs_create_info;
	if (context->get_vrs_capabilities().attachment_vrs_supported) {
		// If VRS is used, this defines how the different VRS types are combined.
		// combinerOps[0] decides how we use the output of pipeline and primitive (drawcall) VRS.
		// combinerOps[1] decides how we use the output of combinerOps[0] and our attachment VRS.

		vrs_create_info.sType = VK_STRUCTURE_TYPE_PIPELINE_FRAGMENT_SHADING_RATE_STATE_CREATE_INFO_KHR;
		vrs_create_info.pNext = nullptr;
		vrs_create_info.fragmentSize = { 4, 4 };
		vrs_create_info.combinerOps[0] = VK_FRAGMENT_SHADING_RATE_COMBINER_OP_KEEP_KHR; // We don't use pipeline/primitive VRS so this really doesn't matter.
		vrs_create_info.combinerOps[1] = VK_FRAGMENT_SHADING_RATE_COMBINER_OP_REPLACE_KHR; // Always use the outcome of attachment VRS if enabled.

		graphics_pipeline_nextptr = &vrs_create_info;
	}

	// Finally, pipeline create info.
	VkGraphicsPipelineCreateInfo graphics_pipeline_create_info;

	graphics_pipeline_create_info.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO;
	graphics_pipeline_create_info.pNext = graphics_pipeline_nextptr;
	graphics_pipeline_create_info.flags = 0;

	Vector<VkPipelineShaderStageCreateInfo> pipeline_stages = shader->pipeline_stages;
	Vector<VkSpecializationInfo> specialization_info;
	Vector<Vector<VkSpecializationMapEntry>> specialization_map_entries;
	Vector<uint32_t> specialization_constant_data;

	if (shader->specialization_constants.size()) {
		specialization_constant_data.resize(shader->specialization_constants.size());
		uint32_t *data_ptr = specialization_constant_data.ptrw();
		specialization_info.resize(pipeline_stages.size());
		specialization_map_entries.resize(pipeline_stages.size());
		for (int i = 0; i < shader->specialization_constants.size(); i++) {
			// See if overridden.
			const Shader::SpecializationConstant &sc = shader->specialization_constants[i];
			data_ptr[i] = sc.constant.int_value; // Just copy the 32 bits.

			for (int j = 0; j < p_specialization_constants.size(); j++) {
				const PipelineSpecializationConstant &psc = p_specialization_constants[j];
				if (psc.constant_id == sc.constant.constant_id) {
					ERR_FAIL_COND_V_MSG(psc.type != sc.constant.type, RID(), "Specialization constant provided for id (" + itos(sc.constant.constant_id) + ") is of the wrong type.");
					data_ptr[i] = psc.int_value;
					break;
				}
			}

			VkSpecializationMapEntry entry;

			entry.constantID = sc.constant.constant_id;
			entry.offset = i * sizeof(uint32_t);
			entry.size = sizeof(uint32_t);

			for (int j = 0; j < SHADER_STAGE_MAX; j++) {
				if (sc.stage_flags & (1 << j)) {
					VkShaderStageFlagBits stage = shader_stage_masks[j];
					for (int k = 0; k < pipeline_stages.size(); k++) {
						if (pipeline_stages[k].stage == stage) {
							specialization_map_entries.write[k].push_back(entry);
						}
					}
				}
			}
		}

		for (int i = 0; i < pipeline_stages.size(); i++) {
			if (specialization_map_entries[i].size()) {
				specialization_info.write[i].dataSize = specialization_constant_data.size() * sizeof(uint32_t);
				specialization_info.write[i].pData = data_ptr;
				specialization_info.write[i].mapEntryCount = specialization_map_entries[i].size();
				specialization_info.write[i].pMapEntries = specialization_map_entries[i].ptr();
				pipeline_stages.write[i].pSpecializationInfo = specialization_info.ptr() + i;
			}
		}
	}

	graphics_pipeline_create_info.stageCount = pipeline_stages.size();
	graphics_pipeline_create_info.pStages = pipeline_stages.ptr();

	graphics_pipeline_create_info.pVertexInputState = &pipeline_vertex_input_state_create_info;
	graphics_pipeline_create_info.pInputAssemblyState = &input_assembly_create_info;
	graphics_pipeline_create_info.pTessellationState = &tessellation_create_info;
	graphics_pipeline_create_info.pViewportState = &viewport_state_create_info;
	graphics_pipeline_create_info.pRasterizationState = &rasterization_state_create_info;
	graphics_pipeline_create_info.pMultisampleState = &multisample_state_create_info;
	graphics_pipeline_create_info.pDepthStencilState = &depth_stencil_state_create_info;
	graphics_pipeline_create_info.pColorBlendState = &color_blend_state_create_info;
	graphics_pipeline_create_info.pDynamicState = &dynamic_state_create_info;
	graphics_pipeline_create_info.layout = shader->pipeline_layout;
	graphics_pipeline_create_info.renderPass = fb_format.render_pass;

	graphics_pipeline_create_info.subpass = p_for_render_pass;
	graphics_pipeline_create_info.basePipelineHandle = VK_NULL_HANDLE;
	graphics_pipeline_create_info.basePipelineIndex = 0;

	RenderPipeline pipeline;
	VkResult err = vkCreateGraphicsPipelines(device, VK_NULL_HANDLE, 1, &graphics_pipeline_create_info, nullptr, &pipeline.pipeline);
	ERR_FAIL_COND_V_MSG(err, RID(), "vkCreateGraphicsPipelines failed with error " + itos(err) + " for shader '" + shader->name + "'.");

	pipeline.set_formats = shader->set_formats;
	pipeline.push_constant_stages = shader->push_constant.push_constants_vk_stage;
	pipeline.pipeline_layout = shader->pipeline_layout;
	pipeline.shader = p_shader;
	pipeline.push_constant_size = shader->push_constant.push_constant_size;

#ifdef DEBUG_ENABLED
	pipeline.validation.dynamic_state = p_dynamic_state_flags;
	pipeline.validation.framebuffer_format = p_framebuffer_format;
	pipeline.validation.render_pass = p_for_render_pass;
	pipeline.validation.vertex_format = p_vertex_format;
	pipeline.validation.uses_restart_indices = input_assembly_create_info.primitiveRestartEnable;

	static const uint32_t primitive_divisor[RENDER_PRIMITIVE_MAX] = {
		1, 2, 1, 1, 1, 3, 1, 1, 1, 1, 1
	};
	pipeline.validation.primitive_divisor = primitive_divisor[p_render_primitive];
	static const uint32_t primitive_minimum[RENDER_PRIMITIVE_MAX] = {
		1,
		2,
		2,
		2,
		2,
		3,
		3,
		3,
		3,
		3,
		1,
	};
	pipeline.validation.primitive_minimum = primitive_minimum[p_render_primitive];
#endif
	// Create ID to associate with this pipeline.
	RID id = render_pipeline_owner.make_rid(pipeline);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif
	// Now add all the dependencies.
	_add_dependency(id, p_shader);
	return id;
}

bool RenderingDeviceVulkan::render_pipeline_is_valid(RID p_pipeline) {
	_THREAD_SAFE_METHOD_
	return render_pipeline_owner.owns(p_pipeline);
}

/**************************/
/**** COMPUTE PIPELINE ****/
/**************************/

RID RenderingDeviceVulkan::compute_pipeline_create(RID p_shader, const Vector<PipelineSpecializationConstant> &p_specialization_constants) {
	_THREAD_SAFE_METHOD_

	// Needs a shader.
	Shader *shader = shader_owner.get_or_null(p_shader);
	ERR_FAIL_COND_V(!shader, RID());

	ERR_FAIL_COND_V_MSG(!shader->is_compute, RID(),
			"Non-compute shaders can't be used in compute pipelines");

	// Finally, pipeline create info.
	VkComputePipelineCreateInfo compute_pipeline_create_info;

	compute_pipeline_create_info.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO;
	compute_pipeline_create_info.pNext = nullptr;
	compute_pipeline_create_info.flags = 0;

	compute_pipeline_create_info.stage = shader->pipeline_stages[0];
	compute_pipeline_create_info.layout = shader->pipeline_layout;
	compute_pipeline_create_info.basePipelineHandle = VK_NULL_HANDLE;
	compute_pipeline_create_info.basePipelineIndex = 0;

	VkSpecializationInfo specialization_info;
	Vector<VkSpecializationMapEntry> specialization_map_entries;
	Vector<uint32_t> specialization_constant_data;

	if (shader->specialization_constants.size()) {
		specialization_constant_data.resize(shader->specialization_constants.size());
		uint32_t *data_ptr = specialization_constant_data.ptrw();
		for (int i = 0; i < shader->specialization_constants.size(); i++) {
			// See if overridden.
			const Shader::SpecializationConstant &sc = shader->specialization_constants[i];
			data_ptr[i] = sc.constant.int_value; // Just copy the 32 bits.

			for (int j = 0; j < p_specialization_constants.size(); j++) {
				const PipelineSpecializationConstant &psc = p_specialization_constants[j];
				if (psc.constant_id == sc.constant.constant_id) {
					ERR_FAIL_COND_V_MSG(psc.type != sc.constant.type, RID(), "Specialization constant provided for id (" + itos(sc.constant.constant_id) + ") is of the wrong type.");
					data_ptr[i] = psc.int_value;
					break;
				}
			}

			VkSpecializationMapEntry entry;

			entry.constantID = sc.constant.constant_id;
			entry.offset = i * sizeof(uint32_t);
			entry.size = sizeof(uint32_t);

			specialization_map_entries.push_back(entry);
		}

		specialization_info.dataSize = specialization_constant_data.size() * sizeof(uint32_t);
		specialization_info.pData = data_ptr;
		specialization_info.mapEntryCount = specialization_map_entries.size();
		specialization_info.pMapEntries = specialization_map_entries.ptr();

		compute_pipeline_create_info.stage.pSpecializationInfo = &specialization_info;
	}

	ComputePipeline pipeline;
	VkResult err = vkCreateComputePipelines(device, VK_NULL_HANDLE, 1, &compute_pipeline_create_info, nullptr, &pipeline.pipeline);
	ERR_FAIL_COND_V_MSG(err, RID(), "vkCreateComputePipelines failed with error " + itos(err) + ".");

	pipeline.set_formats = shader->set_formats;
	pipeline.push_constant_stages = shader->push_constant.push_constants_vk_stage;
	pipeline.pipeline_layout = shader->pipeline_layout;
	pipeline.shader = p_shader;
	pipeline.push_constant_size = shader->push_constant.push_constant_size;
	pipeline.local_group_size[0] = shader->compute_local_size[0];
	pipeline.local_group_size[1] = shader->compute_local_size[1];
	pipeline.local_group_size[2] = shader->compute_local_size[2];

	// Create ID to associate with this pipeline.
	RID id = compute_pipeline_owner.make_rid(pipeline);
#ifdef DEV_ENABLED
	set_resource_name(id, "RID:" + itos(id.get_id()));
#endif
	// Now add all the dependencies.
	_add_dependency(id, p_shader);
	return id;
}

bool RenderingDeviceVulkan::compute_pipeline_is_valid(RID p_pipeline) {
	return compute_pipeline_owner.owns(p_pipeline);
}

/****************/
/**** SCREEN ****/
/****************/

int RenderingDeviceVulkan::screen_get_width(DisplayServer::WindowID p_screen) const {
	_THREAD_SAFE_METHOD_
	ERR_FAIL_COND_V_MSG(local_device.is_valid(), -1, "Local devices have no screen");
	return context->window_get_width(p_screen);
}

int RenderingDeviceVulkan::screen_get_height(DisplayServer::WindowID p_screen) const {
	_THREAD_SAFE_METHOD_
	ERR_FAIL_COND_V_MSG(local_device.is_valid(), -1, "Local devices have no screen");

	return context->window_get_height(p_screen);
}

RenderingDevice::FramebufferFormatID RenderingDeviceVulkan::screen_get_framebuffer_format() const {
	_THREAD_SAFE_METHOD_
	ERR_FAIL_COND_V_MSG(local_device.is_valid(), INVALID_ID, "Local devices have no screen");

	// Very hacky, but not used often per frame so I guess ok.
	VkFormat vkformat = context->get_screen_format();
	DataFormat format = DATA_FORMAT_MAX;
	for (int i = 0; i < DATA_FORMAT_MAX; i++) {
		if (vkformat == vulkan_formats[i]) {
			format = DataFormat(i);
			break;
		}
	}

	ERR_FAIL_COND_V(format == DATA_FORMAT_MAX, INVALID_ID);

	AttachmentFormat attachment;
	attachment.format = format;
	attachment.samples = TEXTURE_SAMPLES_1;
	attachment.usage_flags = TEXTURE_USAGE_COLOR_ATTACHMENT_BIT;
	Vector<AttachmentFormat> screen_attachment;
	screen_attachment.push_back(attachment);
	return const_cast<RenderingDeviceVulkan *>(this)->framebuffer_format_create(screen_attachment);
}

/*******************/
/**** DRAW LIST ****/
/*******************/

RenderingDevice::DrawListID RenderingDeviceVulkan::draw_list_begin_for_screen(DisplayServer::WindowID p_screen, const Color &p_clear_color) {
	_THREAD_SAFE_METHOD_
	ERR_FAIL_COND_V_MSG(local_device.is_valid(), INVALID_ID, "Local devices have no screen");

	ERR_FAIL_COND_V_MSG(draw_list != nullptr, INVALID_ID, "Only one draw list can be active at the same time.");
	ERR_FAIL_COND_V_MSG(compute_list != nullptr, INVALID_ID, "Only one draw/compute list can be active at the same time.");

	VkCommandBuffer command_buffer = frames[frame].draw_command_buffer;

	if (!context->window_is_valid_swapchain(p_screen)) {
		return INVALID_ID;
	}

	Size2i size = Size2i(context->window_get_width(p_screen), context->window_get_height(p_screen));

	_draw_list_allocate(Rect2i(Vector2i(), size), 0, 0);
#ifdef DEBUG_ENABLED
	draw_list_framebuffer_format = screen_get_framebuffer_format();
#endif
	draw_list_subpass_count = 1;

	VkRenderPassBeginInfo render_pass_begin;
	render_pass_begin.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO;
	render_pass_begin.pNext = nullptr;
	render_pass_begin.renderPass = context->window_get_render_pass(p_screen);
	render_pass_begin.framebuffer = context->window_get_framebuffer(p_screen);

	render_pass_begin.renderArea.extent.width = size.width;
	render_pass_begin.renderArea.extent.height = size.height;
	render_pass_begin.renderArea.offset.x = 0;
	render_pass_begin.renderArea.offset.y = 0;

	render_pass_begin.clearValueCount = 1;

	VkClearValue clear_value;
	clear_value.color.float32[0] = p_clear_color.r;
	clear_value.color.float32[1] = p_clear_color.g;
	clear_value.color.float32[2] = p_clear_color.b;
	clear_value.color.float32[3] = p_clear_color.a;

	render_pass_begin.pClearValues = &clear_value;

	vkCmdBeginRenderPass(command_buffer, &render_pass_begin, VK_SUBPASS_CONTENTS_INLINE);

	uint32_t size_x = screen_get_width(p_screen);
	uint32_t size_y = screen_get_height(p_screen);

	VkViewport viewport;
	viewport.x = 0;
	viewport.y = 0;
	viewport.width = size_x;
	viewport.height = size_y;
	viewport.minDepth = 0;
	viewport.maxDepth = 1.0;

	vkCmdSetViewport(command_buffer, 0, 1, &viewport);

	VkRect2D scissor;
	scissor.offset.x = 0;
	scissor.offset.y = 0;
	scissor.extent.width = size_x;
	scissor.extent.height = size_y;

	vkCmdSetScissor(command_buffer, 0, 1, &scissor);

	return int64_t(ID_TYPE_DRAW_LIST) << ID_BASE_SHIFT;
}

Error RenderingDeviceVulkan::_draw_list_setup_framebuffer(Framebuffer *p_framebuffer, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, VkFramebuffer *r_framebuffer, VkRenderPass *r_render_pass, uint32_t *r_subpass_count) {
	Framebuffer::VersionKey vk;
	vk.initial_color_action = p_initial_color_action;
	vk.final_color_action = p_final_color_action;
	vk.initial_depth_action = p_initial_depth_action;
	vk.final_depth_action = p_final_depth_action;
	vk.view_count = p_framebuffer->view_count;

	if (!p_framebuffer->framebuffers.has(vk)) {
		// Need to create this version.
		Framebuffer::Version version;

		version.render_pass = _render_pass_create(framebuffer_formats[p_framebuffer->format_id].E->key().attachments, framebuffer_formats[p_framebuffer->format_id].E->key().passes, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_framebuffer->view_count);

		VkFramebufferCreateInfo framebuffer_create_info;
		framebuffer_create_info.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO;
		framebuffer_create_info.pNext = nullptr;
		framebuffer_create_info.flags = 0;
		framebuffer_create_info.renderPass = version.render_pass;
		Vector<VkImageView> attachments;
		for (int i = 0; i < p_framebuffer->texture_ids.size(); i++) {
			Texture *texture = texture_owner.get_or_null(p_framebuffer->texture_ids[i]);
			if (texture) {
				attachments.push_back(texture->view);
				if (!(texture->usage_flags & TEXTURE_USAGE_VRS_ATTACHMENT_BIT)) { // VRS attachment will be a different size.
					ERR_FAIL_COND_V(texture->width != p_framebuffer->size.width, ERR_BUG);
					ERR_FAIL_COND_V(texture->height != p_framebuffer->size.height, ERR_BUG);
				}
			}
		}
		framebuffer_create_info.attachmentCount = attachments.size();
		framebuffer_create_info.pAttachments = attachments.ptr();
		framebuffer_create_info.width = p_framebuffer->size.width;
		framebuffer_create_info.height = p_framebuffer->size.height;
		framebuffer_create_info.layers = 1;

		VkResult err = vkCreateFramebuffer(device, &framebuffer_create_info, nullptr, &version.framebuffer);
		ERR_FAIL_COND_V_MSG(err, ERR_CANT_CREATE, "vkCreateFramebuffer failed with error " + itos(err) + ".");

		version.subpass_count = framebuffer_formats[p_framebuffer->format_id].E->key().passes.size();

		p_framebuffer->framebuffers.insert(vk, version);
	}
	const Framebuffer::Version &version = p_framebuffer->framebuffers[vk];
	*r_framebuffer = version.framebuffer;
	*r_render_pass = version.render_pass;
	*r_subpass_count = version.subpass_count;

	return OK;
}

Error RenderingDeviceVulkan::_draw_list_render_pass_begin(Framebuffer *framebuffer, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, const Vector<Color> &p_clear_colors, float p_clear_depth, uint32_t p_clear_stencil, Point2i viewport_offset, Point2i viewport_size, VkFramebuffer vkframebuffer, VkRenderPass render_pass, VkCommandBuffer command_buffer, VkSubpassContents subpass_contents, const Vector<RID> &p_storage_textures) {
	VkRenderPassBeginInfo render_pass_begin;
	render_pass_begin.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO;
	render_pass_begin.pNext = nullptr;
	render_pass_begin.renderPass = render_pass;
	render_pass_begin.framebuffer = vkframebuffer;
	/*
	 * Given how API works, it makes sense to always fully operate on the whole framebuffer.
	 * This allows better continue operations for operations like shadowmapping.
	render_pass_begin.renderArea.extent.width = viewport_size.width;
	render_pass_begin.renderArea.extent.height = viewport_size.height;
	render_pass_begin.renderArea.offset.x = viewport_offset.x;
	render_pass_begin.renderArea.offset.y = viewport_offset.y;
	*/
	render_pass_begin.renderArea.extent.width = framebuffer->size.width;
	render_pass_begin.renderArea.extent.height = framebuffer->size.height;
	render_pass_begin.renderArea.offset.x = 0;
	render_pass_begin.renderArea.offset.y = 0;

	Vector<VkClearValue> clear_values;
	clear_values.resize(framebuffer->texture_ids.size());
	int clear_values_count = 0;
	{
		int color_index = 0;
		for (int i = 0; i < framebuffer->texture_ids.size(); i++) {
			VkClearValue clear_value;

			Texture *texture = texture_owner.get_or_null(framebuffer->texture_ids[i]);
			if (!texture) {
				color_index++;
				continue;
			}

			if (color_index < p_clear_colors.size() && texture->usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
				ERR_FAIL_INDEX_V(color_index, p_clear_colors.size(), ERR_BUG); // A bug.
				Color clear_color = p_clear_colors[color_index];
				clear_value.color.float32[0] = clear_color.r;
				clear_value.color.float32[1] = clear_color.g;
				clear_value.color.float32[2] = clear_color.b;
				clear_value.color.float32[3] = clear_color.a;
				color_index++;
			} else if (texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
				clear_value.depthStencil.depth = p_clear_depth;
				clear_value.depthStencil.stencil = p_clear_stencil;
			} else {
				clear_value.color.float32[0] = 0;
				clear_value.color.float32[1] = 0;
				clear_value.color.float32[2] = 0;
				clear_value.color.float32[3] = 0;
			}
			clear_values.write[clear_values_count++] = clear_value;
		}
	}

	render_pass_begin.clearValueCount = clear_values_count;
	render_pass_begin.pClearValues = clear_values.ptr();

	for (int i = 0; i < p_storage_textures.size(); i++) {
		Texture *texture = texture_owner.get_or_null(p_storage_textures[i]);
		if (!texture) {
			continue;
		}
		ERR_CONTINUE_MSG(!(texture->usage_flags & TEXTURE_USAGE_STORAGE_BIT), "Supplied storage texture " + itos(i) + " for draw list is not set to be used for storage.");

		if (texture->usage_flags & TEXTURE_USAGE_SAMPLING_BIT) {
			// Must change layout to general.
			VkImageMemoryBarrier image_memory_barrier;
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
			image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
			image_memory_barrier.oldLayout = texture->layout;
			image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_GENERAL;

			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = texture->image;
			image_memory_barrier.subresourceRange.aspectMask = texture->read_aspect_mask;
			image_memory_barrier.subresourceRange.baseMipLevel = texture->base_mipmap;
			image_memory_barrier.subresourceRange.levelCount = texture->mipmaps;
			image_memory_barrier.subresourceRange.baseArrayLayer = texture->base_layer;
			image_memory_barrier.subresourceRange.layerCount = texture->layers;

			vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier);

			texture->layout = VK_IMAGE_LAYOUT_GENERAL;

			draw_list_storage_textures.push_back(p_storage_textures[i]);
		}
	}

	vkCmdBeginRenderPass(command_buffer, &render_pass_begin, subpass_contents);

	// Mark textures as bound.
	draw_list_bound_textures.clear();
	draw_list_unbind_color_textures = p_final_color_action != FINAL_ACTION_CONTINUE;
	draw_list_unbind_depth_textures = p_final_depth_action != FINAL_ACTION_CONTINUE;

	for (int i = 0; i < framebuffer->texture_ids.size(); i++) {
		Texture *texture = texture_owner.get_or_null(framebuffer->texture_ids[i]);
		if (!texture) {
			continue;
		}
		texture->bound = true;
		draw_list_bound_textures.push_back(framebuffer->texture_ids[i]);
	}

	return OK;
}

void RenderingDeviceVulkan::_draw_list_insert_clear_region(DrawList *p_draw_list, Framebuffer *p_framebuffer, Point2i p_viewport_offset, Point2i p_viewport_size, bool p_clear_color, const Vector<Color> &p_clear_colors, bool p_clear_depth, float p_depth, uint32_t p_stencil) {
	Vector<VkClearAttachment> clear_attachments;
	int color_index = 0;
	int texture_index = 0;
	for (int i = 0; i < p_framebuffer->texture_ids.size(); i++) {
		Texture *texture = texture_owner.get_or_null(p_framebuffer->texture_ids[i]);

		if (!texture) {
			texture_index++;
			continue;
		}

		VkClearAttachment clear_at = {};
		if (p_clear_color && texture->usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) {
			Color clear_color = p_clear_colors[texture_index++];
			clear_at.clearValue.color.float32[0] = clear_color.r;
			clear_at.clearValue.color.float32[1] = clear_color.g;
			clear_at.clearValue.color.float32[2] = clear_color.b;
			clear_at.clearValue.color.float32[3] = clear_color.a;
			clear_at.colorAttachment = color_index++;
			clear_at.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
		} else if (p_clear_depth && texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) {
			clear_at.clearValue.depthStencil.depth = p_depth;
			clear_at.clearValue.depthStencil.stencil = p_stencil;
			clear_at.colorAttachment = 0;
			clear_at.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
			if (format_has_stencil(texture->format)) {
				clear_at.aspectMask |= VK_IMAGE_ASPECT_STENCIL_BIT;
			}
		} else {
			ERR_CONTINUE(true);
		}
		clear_attachments.push_back(clear_at);
	}

	VkClearRect cr;
	cr.baseArrayLayer = 0;
	cr.layerCount = 1;
	cr.rect.offset.x = p_viewport_offset.x;
	cr.rect.offset.y = p_viewport_offset.y;
	cr.rect.extent.width = p_viewport_size.width;
	cr.rect.extent.height = p_viewport_size.height;

	vkCmdClearAttachments(p_draw_list->command_buffer, clear_attachments.size(), clear_attachments.ptr(), 1, &cr);
}

RenderingDevice::DrawListID RenderingDeviceVulkan::draw_list_begin(RID p_framebuffer, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values, float p_clear_depth, uint32_t p_clear_stencil, const Rect2 &p_region, const Vector<RID> &p_storage_textures) {
	_THREAD_SAFE_METHOD_

	ERR_FAIL_COND_V_MSG(draw_list != nullptr, INVALID_ID, "Only one draw list can be active at the same time.");
	ERR_FAIL_COND_V_MSG(compute_list != nullptr && !compute_list->state.allow_draw_overlap, INVALID_ID, "Only one draw/compute list can be active at the same time.");

	Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_framebuffer);
	ERR_FAIL_COND_V(!framebuffer, INVALID_ID);

	Point2i viewport_offset;
	Point2i viewport_size = framebuffer->size;
	bool needs_clear_color = false;
	bool needs_clear_depth = false;

	if (p_region != Rect2() && p_region != Rect2(Vector2(), viewport_size)) { // Check custom region.
		Rect2i viewport(viewport_offset, viewport_size);
		Rect2i regioni = p_region;
		if (!(regioni.position.x >= viewport.position.x) && (regioni.position.y >= viewport.position.y) &&
				((regioni.position.x + regioni.size.x) <= (viewport.position.x + viewport.size.x)) &&
				((regioni.position.y + regioni.size.y) <= (viewport.position.y + viewport.size.y))) {
			ERR_FAIL_V_MSG(INVALID_ID, "When supplying a custom region, it must be contained within the framebuffer rectangle");
		}

		viewport_offset = regioni.position;
		viewport_size = regioni.size;
		if (p_initial_color_action == INITIAL_ACTION_CLEAR_REGION_CONTINUE) {
			needs_clear_color = true;
			p_initial_color_action = INITIAL_ACTION_CONTINUE;
		}
		if (p_initial_depth_action == INITIAL_ACTION_CLEAR_REGION_CONTINUE) {
			needs_clear_depth = true;
			p_initial_depth_action = INITIAL_ACTION_CONTINUE;
		}
		if (p_initial_color_action == INITIAL_ACTION_CLEAR_REGION) {
			needs_clear_color = true;
			p_initial_color_action = INITIAL_ACTION_KEEP;
		}
		if (p_initial_depth_action == INITIAL_ACTION_CLEAR_REGION) {
			needs_clear_depth = true;
			p_initial_depth_action = INITIAL_ACTION_KEEP;
		}
	}

	if (p_initial_color_action == INITIAL_ACTION_CLEAR || needs_clear_color) { // Check clear values.
		int color_count = 0;
		for (int i = 0; i < framebuffer->texture_ids.size(); i++) {
			Texture *texture = texture_owner.get_or_null(framebuffer->texture_ids[i]);
			// We only check for our VRS usage bit if this is not the first texture id.
			// If it is the first we're likely populating our VRS texture.
			// Bit dirty but...
			if (!texture || (!(texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) && !(i != 0 && texture->usage_flags & TEXTURE_USAGE_VRS_ATTACHMENT_BIT))) {
				if (!texture || !texture->is_resolve_buffer) {
					color_count++;
				}
			}
		}
		ERR_FAIL_COND_V_MSG(p_clear_color_values.size() != color_count, INVALID_ID, "Clear color values supplied (" + itos(p_clear_color_values.size()) + ") differ from the amount required for framebuffer color attachments (" + itos(color_count) + ").");
	}

	VkFramebuffer vkframebuffer;
	VkRenderPass render_pass;

	Error err = _draw_list_setup_framebuffer(framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, &vkframebuffer, &render_pass, &draw_list_subpass_count);
	ERR_FAIL_COND_V(err != OK, INVALID_ID);

	VkCommandBuffer command_buffer = frames[frame].draw_command_buffer;
	err = _draw_list_render_pass_begin(framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, viewport_offset, viewport_size, vkframebuffer, render_pass, command_buffer, VK_SUBPASS_CONTENTS_INLINE, p_storage_textures);

	if (err != OK) {
		return INVALID_ID;
	}

	draw_list_render_pass = render_pass;
	draw_list_vkframebuffer = vkframebuffer;

	_draw_list_allocate(Rect2i(viewport_offset, viewport_size), 0, 0);
#ifdef DEBUG_ENABLED
	draw_list_framebuffer_format = framebuffer->format_id;
#endif
	draw_list_current_subpass = 0;

	if (needs_clear_color || needs_clear_depth) {
		_draw_list_insert_clear_region(draw_list, framebuffer, viewport_offset, viewport_size, needs_clear_color, p_clear_color_values, needs_clear_depth, p_clear_depth, p_clear_stencil);
	}

	VkViewport viewport;
	viewport.x = viewport_offset.x;
	viewport.y = viewport_offset.y;
	viewport.width = viewport_size.width;
	viewport.height = viewport_size.height;
	viewport.minDepth = 0;
	viewport.maxDepth = 1.0;

	vkCmdSetViewport(command_buffer, 0, 1, &viewport);

	VkRect2D scissor;
	scissor.offset.x = viewport_offset.x;
	scissor.offset.y = viewport_offset.y;
	scissor.extent.width = viewport_size.width;
	scissor.extent.height = viewport_size.height;

	vkCmdSetScissor(command_buffer, 0, 1, &scissor);

	return int64_t(ID_TYPE_DRAW_LIST) << ID_BASE_SHIFT;
}

Error RenderingDeviceVulkan::draw_list_begin_split(RID p_framebuffer, uint32_t p_splits, DrawListID *r_split_ids, InitialAction p_initial_color_action, FinalAction p_final_color_action, InitialAction p_initial_depth_action, FinalAction p_final_depth_action, const Vector<Color> &p_clear_color_values, float p_clear_depth, uint32_t p_clear_stencil, const Rect2 &p_region, const Vector<RID> &p_storage_textures) {
	_THREAD_SAFE_METHOD_

	ERR_FAIL_COND_V_MSG(draw_list != nullptr, ERR_BUSY, "Only one draw list can be active at the same time.");
	ERR_FAIL_COND_V_MSG(compute_list != nullptr && !compute_list->state.allow_draw_overlap, ERR_BUSY, "Only one draw/compute list can be active at the same time.");

	ERR_FAIL_COND_V(p_splits < 1, ERR_INVALID_DECLARATION);

	Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_framebuffer);
	ERR_FAIL_COND_V(!framebuffer, ERR_INVALID_DECLARATION);

	Point2i viewport_offset;
	Point2i viewport_size = framebuffer->size;

	bool needs_clear_color = false;
	bool needs_clear_depth = false;

	if (p_region != Rect2() && p_region != Rect2(Vector2(), viewport_size)) { // Check custom region.
		Rect2i viewport(viewport_offset, viewport_size);
		Rect2i regioni = p_region;
		if (!(regioni.position.x >= viewport.position.x) && (regioni.position.y >= viewport.position.y) &&
				((regioni.position.x + regioni.size.x) <= (viewport.position.x + viewport.size.x)) &&
				((regioni.position.y + regioni.size.y) <= (viewport.position.y + viewport.size.y))) {
			ERR_FAIL_V_MSG(ERR_INVALID_PARAMETER, "When supplying a custom region, it must be contained within the framebuffer rectangle");
		}

		viewport_offset = regioni.position;
		viewport_size = regioni.size;

		if (p_initial_color_action == INITIAL_ACTION_CLEAR_REGION) {
			needs_clear_color = true;
			p_initial_color_action = INITIAL_ACTION_KEEP;
		}
		if (p_initial_depth_action == INITIAL_ACTION_CLEAR_REGION) {
			needs_clear_depth = true;
			p_initial_depth_action = INITIAL_ACTION_KEEP;
		}
	}

	if (p_initial_color_action == INITIAL_ACTION_CLEAR || needs_clear_color) { // Check clear values.

		int color_count = 0;
		for (int i = 0; i < framebuffer->texture_ids.size(); i++) {
			Texture *texture = texture_owner.get_or_null(framebuffer->texture_ids[i]);

			if (!texture || !(texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT)) {
				color_count++;
			}
		}

		ERR_FAIL_COND_V_MSG(p_clear_color_values.size() != color_count, ERR_INVALID_PARAMETER,
				"Clear color values supplied (" + itos(p_clear_color_values.size()) + ") differ from the amount required for framebuffer (" + itos(color_count) + ").");
	}

	VkFramebuffer vkframebuffer;
	VkRenderPass render_pass;

	Error err = _draw_list_setup_framebuffer(framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, &vkframebuffer, &render_pass, &draw_list_subpass_count);
	ERR_FAIL_COND_V(err != OK, ERR_CANT_CREATE);

	VkCommandBuffer frame_command_buffer = frames[frame].draw_command_buffer;
	err = _draw_list_render_pass_begin(framebuffer, p_initial_color_action, p_final_color_action, p_initial_depth_action, p_final_depth_action, p_clear_color_values, p_clear_depth, p_clear_stencil, viewport_offset, viewport_size, vkframebuffer, render_pass, frame_command_buffer, VK_SUBPASS_CONTENTS_SECONDARY_COMMAND_BUFFERS, p_storage_textures);

	if (err != OK) {
		return ERR_CANT_CREATE;
	}

	draw_list_current_subpass = 0;

#ifdef DEBUG_ENABLED
	draw_list_framebuffer_format = framebuffer->format_id;
#endif
	draw_list_render_pass = render_pass;
	draw_list_vkframebuffer = vkframebuffer;

	err = _draw_list_allocate(Rect2i(viewport_offset, viewport_size), p_splits, 0);
	if (err != OK) {
		return err;
	}

	if (needs_clear_color || needs_clear_depth) {
		_draw_list_insert_clear_region(&draw_list[0], framebuffer, viewport_offset, viewport_size, needs_clear_color, p_clear_color_values, needs_clear_depth, p_clear_depth, p_clear_stencil);
	}

	for (uint32_t i = 0; i < p_splits; i++) {
		VkViewport viewport;
		viewport.x = viewport_offset.x;
		viewport.y = viewport_offset.y;
		viewport.width = viewport_size.width;
		viewport.height = viewport_size.height;
		viewport.minDepth = 0;
		viewport.maxDepth = 1.0;

		vkCmdSetViewport(draw_list[i].command_buffer, 0, 1, &viewport);

		VkRect2D scissor;
		scissor.offset.x = viewport_offset.x;
		scissor.offset.y = viewport_offset.y;
		scissor.extent.width = viewport_size.width;
		scissor.extent.height = viewport_size.height;

		vkCmdSetScissor(draw_list[i].command_buffer, 0, 1, &scissor);
		r_split_ids[i] = (int64_t(ID_TYPE_SPLIT_DRAW_LIST) << ID_BASE_SHIFT) + i;
	}

	return OK;
}

RenderingDeviceVulkan::DrawList *RenderingDeviceVulkan::_get_draw_list_ptr(DrawListID p_id) {
	if (p_id < 0) {
		return nullptr;
	}

	if (!draw_list) {
		return nullptr;
	} else if (p_id == (int64_t(ID_TYPE_DRAW_LIST) << ID_BASE_SHIFT)) {
		if (draw_list_split) {
			return nullptr;
		}
		return draw_list;
	} else if (p_id >> DrawListID(ID_BASE_SHIFT) == ID_TYPE_SPLIT_DRAW_LIST) {
		if (!draw_list_split) {
			return nullptr;
		}

		uint64_t index = p_id & ((DrawListID(1) << DrawListID(ID_BASE_SHIFT)) - 1); // Mask.

		if (index >= draw_list_count) {
			return nullptr;
		}

		return &draw_list[index];
	} else {
		return nullptr;
	}
}

void RenderingDeviceVulkan::draw_list_set_blend_constants(DrawListID p_list, const Color &p_color) {
	DrawList *dl = _get_draw_list_ptr(p_list);
	ERR_FAIL_COND(!dl);
#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
#endif

	vkCmdSetBlendConstants(dl->command_buffer, p_color.components);
}

void RenderingDeviceVulkan::draw_list_bind_render_pipeline(DrawListID p_list, RID p_render_pipeline) {
	DrawList *dl = _get_draw_list_ptr(p_list);
	ERR_FAIL_COND(!dl);
#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
#endif

	const RenderPipeline *pipeline = render_pipeline_owner.get_or_null(p_render_pipeline);
	ERR_FAIL_COND(!pipeline);
#ifdef DEBUG_ENABLED
	ERR_FAIL_COND(pipeline->validation.framebuffer_format != draw_list_framebuffer_format && pipeline->validation.render_pass != draw_list_current_subpass);
#endif

	if (p_render_pipeline == dl->state.pipeline) {
		return; // Redundant state, return.
	}

	dl->state.pipeline = p_render_pipeline;
	dl->state.pipeline_layout = pipeline->pipeline_layout;

	vkCmdBindPipeline(dl->command_buffer, VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline->pipeline);

	if (dl->state.pipeline_shader != pipeline->shader) {
		// Shader changed, so descriptor sets may become incompatible.

		// Go through ALL sets, and unbind them (and all those above) if the format is different.

		uint32_t pcount = pipeline->set_formats.size(); // Formats count in this pipeline.
		dl->state.set_count = MAX(dl->state.set_count, pcount);
		const uint32_t *pformats = pipeline->set_formats.ptr(); // Pipeline set formats.

		bool sets_valid = true; // Once invalid, all above become invalid.
		for (uint32_t i = 0; i < pcount; i++) {
			// If a part of the format is different, invalidate it (and the rest).
			if (!sets_valid || dl->state.sets[i].pipeline_expected_format != pformats[i]) {
				dl->state.sets[i].bound = false;
				dl->state.sets[i].pipeline_expected_format = pformats[i];
				sets_valid = false;
			}
		}

		for (uint32_t i = pcount; i < dl->state.set_count; i++) {
			// Unbind the ones above (not used) if exist.
			dl->state.sets[i].bound = false;
		}

		dl->state.set_count = pcount; // Update set count.

		if (pipeline->push_constant_size) {
			dl->state.pipeline_push_constant_stages = pipeline->push_constant_stages;
#ifdef DEBUG_ENABLED
			dl->validation.pipeline_push_constant_supplied = false;
#endif
		}

		dl->state.pipeline_shader = pipeline->shader;
	}

#ifdef DEBUG_ENABLED
	// Update render pass pipeline info.
	dl->validation.pipeline_active = true;
	dl->validation.pipeline_dynamic_state = pipeline->validation.dynamic_state;
	dl->validation.pipeline_vertex_format = pipeline->validation.vertex_format;
	dl->validation.pipeline_uses_restart_indices = pipeline->validation.uses_restart_indices;
	dl->validation.pipeline_primitive_divisor = pipeline->validation.primitive_divisor;
	dl->validation.pipeline_primitive_minimum = pipeline->validation.primitive_minimum;
	dl->validation.pipeline_push_constant_size = pipeline->push_constant_size;
#endif
}

void RenderingDeviceVulkan::draw_list_bind_uniform_set(DrawListID p_list, RID p_uniform_set, uint32_t p_index) {
#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(p_index >= limits.maxBoundDescriptorSets || p_index >= MAX_UNIFORM_SETS,
			"Attempting to bind a descriptor set (" + itos(p_index) + ") greater than what the hardware supports (" + itos(limits.maxBoundDescriptorSets) + ").");
#endif
	DrawList *dl = _get_draw_list_ptr(p_list);
	ERR_FAIL_COND(!dl);

#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
#endif

	const UniformSet *uniform_set = uniform_set_owner.get_or_null(p_uniform_set);
	ERR_FAIL_COND(!uniform_set);

	if (p_index > dl->state.set_count) {
		dl->state.set_count = p_index;
	}

	dl->state.sets[p_index].descriptor_set = uniform_set->descriptor_set; // Update set pointer.
	dl->state.sets[p_index].bound = false; // Needs rebind.
	dl->state.sets[p_index].uniform_set_format = uniform_set->format;
	dl->state.sets[p_index].uniform_set = p_uniform_set;

	uint32_t mst_count = uniform_set->mutable_storage_textures.size();
	if (mst_count) {
		Texture **mst_textures = const_cast<UniformSet *>(uniform_set)->mutable_storage_textures.ptrw();
		for (uint32_t i = 0; i < mst_count; i++) {
			if (mst_textures[i]->used_in_frame != frames_drawn) {
				mst_textures[i]->used_in_frame = frames_drawn;
				mst_textures[i]->used_in_transfer = false;
				mst_textures[i]->used_in_compute = false;
			}
			mst_textures[i]->used_in_raster = true;
		}
	}

#ifdef DEBUG_ENABLED
	{ // Validate that textures bound are not attached as framebuffer bindings.
		uint32_t attachable_count = uniform_set->attachable_textures.size();
		const UniformSet::AttachableTexture *attachable_ptr = uniform_set->attachable_textures.ptr();
		uint32_t bound_count = draw_list_bound_textures.size();
		const RID *bound_ptr = draw_list_bound_textures.ptr();
		for (uint32_t i = 0; i < attachable_count; i++) {
			for (uint32_t j = 0; j < bound_count; j++) {
				ERR_FAIL_COND_MSG(attachable_ptr[i].texture == bound_ptr[j],
						"Attempted to use the same texture in framebuffer attachment and a uniform (set: " + itos(p_index) + ", binding: " + itos(attachable_ptr[i].bind) + "), this is not allowed.");
			}
		}
	}
#endif
}

void RenderingDeviceVulkan::draw_list_bind_vertex_array(DrawListID p_list, RID p_vertex_array) {
	DrawList *dl = _get_draw_list_ptr(p_list);
	ERR_FAIL_COND(!dl);
#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
#endif

	const VertexArray *vertex_array = vertex_array_owner.get_or_null(p_vertex_array);
	ERR_FAIL_COND(!vertex_array);

	if (dl->state.vertex_array == p_vertex_array) {
		return; // Already set.
	}

	dl->state.vertex_array = p_vertex_array;

#ifdef DEBUG_ENABLED
	dl->validation.vertex_format = vertex_array->description;
	dl->validation.vertex_max_instances_allowed = vertex_array->max_instances_allowed;
#endif
	dl->validation.vertex_array_size = vertex_array->vertex_count;
	vkCmdBindVertexBuffers(dl->command_buffer, 0, vertex_array->buffers.size(), vertex_array->buffers.ptr(), vertex_array->offsets.ptr());
}

void RenderingDeviceVulkan::draw_list_bind_index_array(DrawListID p_list, RID p_index_array) {
	DrawList *dl = _get_draw_list_ptr(p_list);
	ERR_FAIL_COND(!dl);
#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
#endif

	const IndexArray *index_array = index_array_owner.get_or_null(p_index_array);
	ERR_FAIL_COND(!index_array);

	if (dl->state.index_array == p_index_array) {
		return; // Already set.
	}

	dl->state.index_array = p_index_array;
#ifdef DEBUG_ENABLED
	dl->validation.index_array_max_index = index_array->max_index;
#endif
	dl->validation.index_array_size = index_array->indices;
	dl->validation.index_array_offset = index_array->offset;

	vkCmdBindIndexBuffer(dl->command_buffer, index_array->buffer, 0, index_array->index_type);
}

void RenderingDeviceVulkan::draw_list_set_line_width(DrawListID p_list, float p_width) {
	DrawList *dl = _get_draw_list_ptr(p_list);
	ERR_FAIL_COND(!dl);
#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
#endif

	vkCmdSetLineWidth(dl->command_buffer, p_width);
}

void RenderingDeviceVulkan::draw_list_set_push_constant(DrawListID p_list, const void *p_data, uint32_t p_data_size) {
	DrawList *dl = _get_draw_list_ptr(p_list);
	ERR_FAIL_COND(!dl);

#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
#endif

#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(p_data_size != dl->validation.pipeline_push_constant_size,
			"This render pipeline requires (" + itos(dl->validation.pipeline_push_constant_size) + ") bytes of push constant data, supplied: (" + itos(p_data_size) + ")");
#endif
	vkCmdPushConstants(dl->command_buffer, dl->state.pipeline_layout, dl->state.pipeline_push_constant_stages, 0, p_data_size, p_data);
#ifdef DEBUG_ENABLED
	dl->validation.pipeline_push_constant_supplied = true;
#endif
}

void RenderingDeviceVulkan::draw_list_draw(DrawListID p_list, bool p_use_indices, uint32_t p_instances, uint32_t p_procedural_vertices) {
	DrawList *dl = _get_draw_list_ptr(p_list);
	ERR_FAIL_COND(!dl);
#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
#endif

#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!dl->validation.pipeline_active,
			"No render pipeline was set before attempting to draw.");
	if (dl->validation.pipeline_vertex_format != INVALID_ID) {
		// Pipeline uses vertices, validate format.
		ERR_FAIL_COND_MSG(dl->validation.vertex_format == INVALID_ID,
				"No vertex array was bound, and render pipeline expects vertices.");
		// Make sure format is right.
		ERR_FAIL_COND_MSG(dl->validation.pipeline_vertex_format != dl->validation.vertex_format,
				"The vertex format used to create the pipeline does not match the vertex format bound.");
		// Make sure number of instances is valid.
		ERR_FAIL_COND_MSG(p_instances > dl->validation.vertex_max_instances_allowed,
				"Number of instances requested (" + itos(p_instances) + " is larger than the maximum number supported by the bound vertex array (" + itos(dl->validation.vertex_max_instances_allowed) + ").");
	}

	if (dl->validation.pipeline_push_constant_size > 0) {
		// Using push constants, check that they were supplied.
		ERR_FAIL_COND_MSG(!dl->validation.pipeline_push_constant_supplied,
				"The shader in this pipeline requires a push constant to be set before drawing, but it's not present.");
	}

#endif

	// Bind descriptor sets.

	for (uint32_t i = 0; i < dl->state.set_count; i++) {
		if (dl->state.sets[i].pipeline_expected_format == 0) {
			continue; // Nothing expected by this pipeline.
		}
#ifdef DEBUG_ENABLED
		if (dl->state.sets[i].pipeline_expected_format != dl->state.sets[i].uniform_set_format) {
			if (dl->state.sets[i].uniform_set_format == 0) {
				ERR_FAIL_MSG("Uniforms were never supplied for set (" + itos(i) + ") at the time of drawing, which are required by the pipeline");
			} else if (uniform_set_owner.owns(dl->state.sets[i].uniform_set)) {
				UniformSet *us = uniform_set_owner.get_or_null(dl->state.sets[i].uniform_set);
				ERR_FAIL_MSG("Uniforms supplied for set (" + itos(i) + "):\n" + _shader_uniform_debug(us->shader_id, us->shader_set) + "\nare not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n" + _shader_uniform_debug(dl->state.pipeline_shader));
			} else {
				ERR_FAIL_MSG("Uniforms supplied for set (" + itos(i) + ", which was was just freed) are not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n" + _shader_uniform_debug(dl->state.pipeline_shader));
			}
		}
#endif
		if (!dl->state.sets[i].bound) {
			// All good, see if this requires re-binding.
			vkCmdBindDescriptorSets(dl->command_buffer, VK_PIPELINE_BIND_POINT_GRAPHICS, dl->state.pipeline_layout, i, 1, &dl->state.sets[i].descriptor_set, 0, nullptr);
			dl->state.sets[i].bound = true;
		}
	}

	if (p_use_indices) {
#ifdef DEBUG_ENABLED
		ERR_FAIL_COND_MSG(p_procedural_vertices > 0,
				"Procedural vertices can't be used together with indices.");

		ERR_FAIL_COND_MSG(!dl->validation.index_array_size,
				"Draw command requested indices, but no index buffer was set.");

		ERR_FAIL_COND_MSG(dl->validation.pipeline_uses_restart_indices != dl->validation.index_buffer_uses_restart_indices,
				"The usage of restart indices in index buffer does not match the render primitive in the pipeline.");
#endif
		uint32_t to_draw = dl->validation.index_array_size;

#ifdef DEBUG_ENABLED
		ERR_FAIL_COND_MSG(to_draw < dl->validation.pipeline_primitive_minimum,
				"Too few indices (" + itos(to_draw) + ") for the render primitive set in the render pipeline (" + itos(dl->validation.pipeline_primitive_minimum) + ").");

		ERR_FAIL_COND_MSG((to_draw % dl->validation.pipeline_primitive_divisor) != 0,
				"Index amount (" + itos(to_draw) + ") must be a multiple of the amount of indices required by the render primitive (" + itos(dl->validation.pipeline_primitive_divisor) + ").");
#endif
		vkCmdDrawIndexed(dl->command_buffer, to_draw, p_instances, dl->validation.index_array_offset, 0, 0);
	} else {
		uint32_t to_draw;

		if (p_procedural_vertices > 0) {
#ifdef DEBUG_ENABLED
			ERR_FAIL_COND_MSG(dl->validation.pipeline_vertex_format != INVALID_ID,
					"Procedural vertices requested, but pipeline expects a vertex array.");
#endif
			to_draw = p_procedural_vertices;
		} else {
#ifdef DEBUG_ENABLED
			ERR_FAIL_COND_MSG(dl->validation.pipeline_vertex_format == INVALID_ID,
					"Draw command lacks indices, but pipeline format does not use vertices.");
#endif
			to_draw = dl->validation.vertex_array_size;
		}

#ifdef DEBUG_ENABLED
		ERR_FAIL_COND_MSG(to_draw < dl->validation.pipeline_primitive_minimum,
				"Too few vertices (" + itos(to_draw) + ") for the render primitive set in the render pipeline (" + itos(dl->validation.pipeline_primitive_minimum) + ").");

		ERR_FAIL_COND_MSG((to_draw % dl->validation.pipeline_primitive_divisor) != 0,
				"Vertex amount (" + itos(to_draw) + ") must be a multiple of the amount of vertices required by the render primitive (" + itos(dl->validation.pipeline_primitive_divisor) + ").");
#endif

		vkCmdDraw(dl->command_buffer, to_draw, p_instances, 0, 0);
	}
}

void RenderingDeviceVulkan::draw_list_enable_scissor(DrawListID p_list, const Rect2 &p_rect) {
	DrawList *dl = _get_draw_list_ptr(p_list);

	ERR_FAIL_COND(!dl);
#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
#endif
	Rect2i rect = p_rect;
	rect.position += dl->viewport.position;

	rect = dl->viewport.intersection(rect);

	if (rect.get_area() == 0) {
		return;
	}
	VkRect2D scissor;
	scissor.offset.x = rect.position.x;
	scissor.offset.y = rect.position.y;
	scissor.extent.width = rect.size.width;
	scissor.extent.height = rect.size.height;

	vkCmdSetScissor(dl->command_buffer, 0, 1, &scissor);
}

void RenderingDeviceVulkan::draw_list_disable_scissor(DrawListID p_list) {
	DrawList *dl = _get_draw_list_ptr(p_list);
	ERR_FAIL_COND(!dl);
#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!dl->validation.active, "Submitted Draw Lists can no longer be modified.");
#endif

	VkRect2D scissor;
	scissor.offset.x = dl->viewport.position.x;
	scissor.offset.y = dl->viewport.position.y;
	scissor.extent.width = dl->viewport.size.width;
	scissor.extent.height = dl->viewport.size.height;
	vkCmdSetScissor(dl->command_buffer, 0, 1, &scissor);
}

uint32_t RenderingDeviceVulkan::draw_list_get_current_pass() {
	return draw_list_current_subpass;
}

RenderingDevice::DrawListID RenderingDeviceVulkan::draw_list_switch_to_next_pass() {
	ERR_FAIL_COND_V(draw_list == nullptr, INVALID_ID);
	ERR_FAIL_COND_V(draw_list_current_subpass >= draw_list_subpass_count - 1, INVALID_FORMAT_ID);

	draw_list_current_subpass++;

	Rect2i viewport;
	_draw_list_free(&viewport);

	vkCmdNextSubpass(frames[frame].draw_command_buffer, VK_SUBPASS_CONTENTS_INLINE);

	_draw_list_allocate(viewport, 0, draw_list_current_subpass);

	return int64_t(ID_TYPE_DRAW_LIST) << ID_BASE_SHIFT;
}
Error RenderingDeviceVulkan::draw_list_switch_to_next_pass_split(uint32_t p_splits, DrawListID *r_split_ids) {
	ERR_FAIL_COND_V(draw_list == nullptr, ERR_INVALID_PARAMETER);
	ERR_FAIL_COND_V(draw_list_current_subpass >= draw_list_subpass_count - 1, ERR_INVALID_PARAMETER);

	draw_list_current_subpass++;

	Rect2i viewport;
	_draw_list_free(&viewport);

	vkCmdNextSubpass(frames[frame].draw_command_buffer, VK_SUBPASS_CONTENTS_INLINE);

	_draw_list_allocate(viewport, p_splits, draw_list_current_subpass);

	for (uint32_t i = 0; i < p_splits; i++) {
		r_split_ids[i] = (int64_t(ID_TYPE_SPLIT_DRAW_LIST) << ID_BASE_SHIFT) + i;
	}

	return OK;
}

Error RenderingDeviceVulkan::_draw_list_allocate(const Rect2i &p_viewport, uint32_t p_splits, uint32_t p_subpass) {
	// Lock while draw_list is active.
	_THREAD_SAFE_LOCK_

	if (p_splits == 0) {
		draw_list = memnew(DrawList);
		draw_list->command_buffer = frames[frame].draw_command_buffer;
		draw_list->viewport = p_viewport;
		draw_list_count = 0;
		draw_list_split = false;
	} else {
		if (p_splits > (uint32_t)split_draw_list_allocators.size()) {
			uint32_t from = split_draw_list_allocators.size();
			split_draw_list_allocators.resize(p_splits);
			for (uint32_t i = from; i < p_splits; i++) {
				VkCommandPoolCreateInfo cmd_pool_info;
				cmd_pool_info.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO;
				cmd_pool_info.pNext = nullptr;
				cmd_pool_info.queueFamilyIndex = context->get_graphics_queue_family_index();
				cmd_pool_info.flags = VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT;

				VkResult res = vkCreateCommandPool(device, &cmd_pool_info, nullptr, &split_draw_list_allocators.write[i].command_pool);
				ERR_FAIL_COND_V_MSG(res, ERR_CANT_CREATE, "vkCreateCommandPool failed with error " + itos(res) + ".");

				for (int j = 0; j < frame_count; j++) {
					VkCommandBuffer command_buffer;

					VkCommandBufferAllocateInfo cmdbuf;
					// No command buffer exists, create it.
					cmdbuf.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
					cmdbuf.pNext = nullptr;
					cmdbuf.commandPool = split_draw_list_allocators[i].command_pool;
					cmdbuf.level = VK_COMMAND_BUFFER_LEVEL_SECONDARY;
					cmdbuf.commandBufferCount = 1;

					VkResult err = vkAllocateCommandBuffers(device, &cmdbuf, &command_buffer);
					ERR_FAIL_COND_V_MSG(err, ERR_CANT_CREATE, "vkAllocateCommandBuffers failed with error " + itos(err) + ".");

					split_draw_list_allocators.write[i].command_buffers.push_back(command_buffer);
				}
			}
		}
		draw_list = memnew_arr(DrawList, p_splits);
		draw_list_count = p_splits;
		draw_list_split = true;

		for (uint32_t i = 0; i < p_splits; i++) {
			// Take a command buffer and initialize it.
			VkCommandBuffer command_buffer = split_draw_list_allocators[i].command_buffers[frame];

			VkCommandBufferInheritanceInfo inheritance_info;
			inheritance_info.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_INHERITANCE_INFO;
			inheritance_info.pNext = nullptr;
			inheritance_info.renderPass = draw_list_render_pass;
			inheritance_info.subpass = p_subpass;
			inheritance_info.framebuffer = draw_list_vkframebuffer;
			inheritance_info.occlusionQueryEnable = false;
			inheritance_info.queryFlags = 0; // ?
			inheritance_info.pipelineStatistics = 0;

			VkCommandBufferBeginInfo cmdbuf_begin;
			cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
			cmdbuf_begin.pNext = nullptr;
			cmdbuf_begin.flags = VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT | VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT;
			cmdbuf_begin.pInheritanceInfo = &inheritance_info;

			VkResult res = vkResetCommandBuffer(command_buffer, 0);
			if (res) {
				memdelete_arr(draw_list);
				draw_list = nullptr;
				ERR_FAIL_V_MSG(ERR_CANT_CREATE, "vkResetCommandBuffer failed with error " + itos(res) + ".");
			}

			res = vkBeginCommandBuffer(command_buffer, &cmdbuf_begin);
			if (res) {
				memdelete_arr(draw_list);
				draw_list = nullptr;
				ERR_FAIL_V_MSG(ERR_CANT_CREATE, "vkBeginCommandBuffer failed with error " + itos(res) + ".");
			}

			draw_list[i].command_buffer = command_buffer;
			draw_list[i].viewport = p_viewport;
		}
	}

	return OK;
}

void RenderingDeviceVulkan::_draw_list_free(Rect2i *r_last_viewport) {
	if (draw_list_split) {
		// Send all command buffers.
		VkCommandBuffer *command_buffers = (VkCommandBuffer *)alloca(sizeof(VkCommandBuffer) * draw_list_count);
		for (uint32_t i = 0; i < draw_list_count; i++) {
			vkEndCommandBuffer(draw_list[i].command_buffer);
			command_buffers[i] = draw_list[i].command_buffer;
			if (r_last_viewport) {
				if (i == 0 || draw_list[i].viewport_set) {
					*r_last_viewport = draw_list[i].viewport;
				}
			}
		}

		vkCmdExecuteCommands(frames[frame].draw_command_buffer, draw_list_count, command_buffers);
		memdelete_arr(draw_list);
		draw_list = nullptr;

	} else {
		if (r_last_viewport) {
			*r_last_viewport = draw_list->viewport;
		}
		// Just end the list.
		memdelete(draw_list);
		draw_list = nullptr;
	}

	// Draw_list is no longer active.
	_THREAD_SAFE_UNLOCK_
}

void RenderingDeviceVulkan::draw_list_end(BitField<BarrierMask> p_post_barrier) {
	_THREAD_SAFE_METHOD_

	ERR_FAIL_COND_MSG(!draw_list, "Immediate draw list is already inactive.");

	_draw_list_free();

	vkCmdEndRenderPass(frames[frame].draw_command_buffer);

	for (int i = 0; i < draw_list_bound_textures.size(); i++) {
		Texture *texture = texture_owner.get_or_null(draw_list_bound_textures[i]);
		ERR_CONTINUE(!texture); // Wtf.
		if (draw_list_unbind_color_textures && (texture->usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT)) {
			texture->bound = false;
		}
		if (draw_list_unbind_depth_textures && (texture->usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT)) {
			texture->bound = false;
		}
	}

	uint32_t barrier_flags = 0;
	uint32_t access_flags = 0;
	if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) {
		barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
		access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
	}
	if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) {
		barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT /*| VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT*/;
		access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT /*| VK_ACCESS_INDIRECT_COMMAND_READ_BIT*/;
	}
	if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) {
		barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
		access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT;
	}

	if (barrier_flags == 0) {
		barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
	}

	draw_list_bound_textures.clear();

	VkImageMemoryBarrier *image_barriers = nullptr;

	uint32_t image_barrier_count = draw_list_storage_textures.size();

	if (image_barrier_count) {
		image_barriers = (VkImageMemoryBarrier *)alloca(sizeof(VkImageMemoryBarrier) * draw_list_storage_textures.size());
	}

	uint32_t src_stage = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
	uint32_t src_access = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;

	if (image_barrier_count) {
		src_stage |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
		src_access |= VK_ACCESS_SHADER_WRITE_BIT;
	}

	for (uint32_t i = 0; i < image_barrier_count; i++) {
		Texture *texture = texture_owner.get_or_null(draw_list_storage_textures[i]);

		VkImageMemoryBarrier &image_memory_barrier = image_barriers[i];
		image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
		image_memory_barrier.pNext = nullptr;
		image_memory_barrier.srcAccessMask = src_access;
		image_memory_barrier.dstAccessMask = access_flags;
		image_memory_barrier.oldLayout = texture->layout;
		image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;

		image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.image = texture->image;
		image_memory_barrier.subresourceRange.aspectMask = texture->read_aspect_mask;
		image_memory_barrier.subresourceRange.baseMipLevel = texture->base_mipmap;
		image_memory_barrier.subresourceRange.levelCount = texture->mipmaps;
		image_memory_barrier.subresourceRange.baseArrayLayer = texture->base_layer;
		image_memory_barrier.subresourceRange.layerCount = texture->layers;

		texture->layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
	}

	draw_list_storage_textures.clear();

	// To ensure proper synchronization, we must make sure rendering is done before:
	// * Some buffer is copied.
	// * Another render pass happens (since we may be done).

	VkMemoryBarrier mem_barrier;
	mem_barrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER;
	mem_barrier.pNext = nullptr;
	mem_barrier.srcAccessMask = src_access;
	mem_barrier.dstAccessMask = access_flags;

	if (image_barrier_count > 0 || p_post_barrier != BARRIER_MASK_NO_BARRIER) {
		vkCmdPipelineBarrier(frames[frame].draw_command_buffer, src_stage, barrier_flags, 0, 1, &mem_barrier, 0, nullptr, image_barrier_count, image_barriers);
	}

#ifdef FORCE_FULL_BARRIER
	_full_barrier(true);
#endif
}

/***********************/
/**** COMPUTE LISTS ****/
/***********************/

RenderingDevice::ComputeListID RenderingDeviceVulkan::compute_list_begin(bool p_allow_draw_overlap) {
	ERR_FAIL_COND_V_MSG(!p_allow_draw_overlap && draw_list != nullptr, INVALID_ID, "Only one draw list can be active at the same time.");
	ERR_FAIL_COND_V_MSG(compute_list != nullptr, INVALID_ID, "Only one draw/compute list can be active at the same time.");

	// Lock while compute_list is active.
	_THREAD_SAFE_LOCK_

	compute_list = memnew(ComputeList);
	compute_list->command_buffer = frames[frame].draw_command_buffer;
	compute_list->state.allow_draw_overlap = p_allow_draw_overlap;

	return ID_TYPE_COMPUTE_LIST;
}

void RenderingDeviceVulkan::compute_list_bind_compute_pipeline(ComputeListID p_list, RID p_compute_pipeline) {
	ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST);
	ERR_FAIL_COND(!compute_list);

	ComputeList *cl = compute_list;

	const ComputePipeline *pipeline = compute_pipeline_owner.get_or_null(p_compute_pipeline);
	ERR_FAIL_COND(!pipeline);

	if (p_compute_pipeline == cl->state.pipeline) {
		return; // Redundant state, return.
	}

	cl->state.pipeline = p_compute_pipeline;
	cl->state.pipeline_layout = pipeline->pipeline_layout;

	vkCmdBindPipeline(cl->command_buffer, VK_PIPELINE_BIND_POINT_COMPUTE, pipeline->pipeline);

	if (cl->state.pipeline_shader != pipeline->shader) {
		// Shader changed, so descriptor sets may become incompatible.

		// Go through ALL sets, and unbind them (and all those above) if the format is different.

		uint32_t pcount = pipeline->set_formats.size(); // Formats count in this pipeline.
		cl->state.set_count = MAX(cl->state.set_count, pcount);
		const uint32_t *pformats = pipeline->set_formats.ptr(); // Pipeline set formats.

		bool sets_valid = true; // Once invalid, all above become invalid.
		for (uint32_t i = 0; i < pcount; i++) {
			// If a part of the format is different, invalidate it (and the rest).
			if (!sets_valid || cl->state.sets[i].pipeline_expected_format != pformats[i]) {
				cl->state.sets[i].bound = false;
				cl->state.sets[i].pipeline_expected_format = pformats[i];
				sets_valid = false;
			}
		}

		for (uint32_t i = pcount; i < cl->state.set_count; i++) {
			// Unbind the ones above (not used) if exist.
			cl->state.sets[i].bound = false;
		}

		cl->state.set_count = pcount; // Update set count.

		if (pipeline->push_constant_size) {
			cl->state.pipeline_push_constant_stages = pipeline->push_constant_stages;
#ifdef DEBUG_ENABLED
			cl->validation.pipeline_push_constant_supplied = false;
#endif
		}

		cl->state.pipeline_shader = pipeline->shader;
		cl->state.local_group_size[0] = pipeline->local_group_size[0];
		cl->state.local_group_size[1] = pipeline->local_group_size[1];
		cl->state.local_group_size[2] = pipeline->local_group_size[2];
	}

#ifdef DEBUG_ENABLED
	// Update compute pass pipeline info.
	cl->validation.pipeline_active = true;
	cl->validation.pipeline_push_constant_size = pipeline->push_constant_size;
#endif
}

void RenderingDeviceVulkan::compute_list_bind_uniform_set(ComputeListID p_list, RID p_uniform_set, uint32_t p_index) {
	ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST);
	ERR_FAIL_COND(!compute_list);

	ComputeList *cl = compute_list;

#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(p_index >= limits.maxBoundDescriptorSets || p_index >= MAX_UNIFORM_SETS,
			"Attempting to bind a descriptor set (" + itos(p_index) + ") greater than what the hardware supports (" + itos(limits.maxBoundDescriptorSets) + ").");
#endif

#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!cl->validation.active, "Submitted Compute Lists can no longer be modified.");
#endif

	UniformSet *uniform_set = uniform_set_owner.get_or_null(p_uniform_set);
	ERR_FAIL_COND(!uniform_set);

	if (p_index > cl->state.set_count) {
		cl->state.set_count = p_index;
	}

	cl->state.sets[p_index].descriptor_set = uniform_set->descriptor_set; // Update set pointer.
	cl->state.sets[p_index].bound = false; // Needs rebind.
	cl->state.sets[p_index].uniform_set_format = uniform_set->format;
	cl->state.sets[p_index].uniform_set = p_uniform_set;

	uint32_t textures_to_sampled_count = uniform_set->mutable_sampled_textures.size();
	uint32_t textures_to_storage_count = uniform_set->mutable_storage_textures.size();

	Texture **textures_to_sampled = uniform_set->mutable_sampled_textures.ptrw();

	VkImageMemoryBarrier *texture_barriers = nullptr;

	if (textures_to_sampled_count + textures_to_storage_count) {
		texture_barriers = (VkImageMemoryBarrier *)alloca(sizeof(VkImageMemoryBarrier) * (textures_to_sampled_count + textures_to_storage_count));
	}
	uint32_t texture_barrier_count = 0;

	uint32_t src_stage_flags = 0;

	for (uint32_t i = 0; i < textures_to_sampled_count; i++) {
		if (textures_to_sampled[i]->layout != VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL) {
			src_stage_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;

			VkImageMemoryBarrier &image_memory_barrier = texture_barriers[texture_barrier_count++];
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
			image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
			image_memory_barrier.oldLayout = textures_to_sampled[i]->layout;
			image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;

			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = textures_to_sampled[i]->image;
			image_memory_barrier.subresourceRange.aspectMask = textures_to_sampled[i]->read_aspect_mask;
			image_memory_barrier.subresourceRange.baseMipLevel = textures_to_sampled[i]->base_mipmap;
			image_memory_barrier.subresourceRange.levelCount = textures_to_sampled[i]->mipmaps;
			image_memory_barrier.subresourceRange.baseArrayLayer = textures_to_sampled[i]->base_layer;
			image_memory_barrier.subresourceRange.layerCount = textures_to_sampled[i]->layers;

			textures_to_sampled[i]->layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;

			cl->state.textures_to_sampled_layout.erase(textures_to_sampled[i]);
		}

		if (textures_to_sampled[i]->used_in_frame != frames_drawn) {
			textures_to_sampled[i]->used_in_frame = frames_drawn;
			textures_to_sampled[i]->used_in_transfer = false;
			textures_to_sampled[i]->used_in_raster = false;
		}
		textures_to_sampled[i]->used_in_compute = true;
	}

	Texture **textures_to_storage = uniform_set->mutable_storage_textures.ptrw();

	for (uint32_t i = 0; i < textures_to_storage_count; i++) {
		if (textures_to_storage[i]->layout != VK_IMAGE_LAYOUT_GENERAL) {
			uint32_t src_access_flags = 0;

			if (textures_to_storage[i]->used_in_frame == frames_drawn) {
				if (textures_to_storage[i]->used_in_compute) {
					src_stage_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
					src_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
				}
				if (textures_to_storage[i]->used_in_raster) {
					src_stage_flags |= VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT;
					src_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
				}
				if (textures_to_storage[i]->used_in_transfer) {
					src_stage_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
					src_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT;
				}

				textures_to_storage[i]->used_in_compute = false;
				textures_to_storage[i]->used_in_raster = false;
				textures_to_storage[i]->used_in_transfer = false;

			} else {
				src_access_flags = 0;
				textures_to_storage[i]->used_in_compute = false;
				textures_to_storage[i]->used_in_raster = false;
				textures_to_storage[i]->used_in_transfer = false;
				textures_to_storage[i]->used_in_frame = frames_drawn;
			}

			VkImageMemoryBarrier &image_memory_barrier = texture_barriers[texture_barrier_count++];
			image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
			image_memory_barrier.pNext = nullptr;
			image_memory_barrier.srcAccessMask = src_access_flags;
			image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
			image_memory_barrier.oldLayout = textures_to_storage[i]->layout;
			image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_GENERAL;

			image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
			image_memory_barrier.image = textures_to_storage[i]->image;
			image_memory_barrier.subresourceRange.aspectMask = textures_to_storage[i]->read_aspect_mask;
			image_memory_barrier.subresourceRange.baseMipLevel = textures_to_storage[i]->base_mipmap;
			image_memory_barrier.subresourceRange.levelCount = textures_to_storage[i]->mipmaps;
			image_memory_barrier.subresourceRange.baseArrayLayer = textures_to_storage[i]->base_layer;
			image_memory_barrier.subresourceRange.layerCount = textures_to_storage[i]->layers;

			textures_to_storage[i]->layout = VK_IMAGE_LAYOUT_GENERAL;

			cl->state.textures_to_sampled_layout.insert(textures_to_storage[i]); // Needs to go back to sampled layout afterwards.
		}
	}

	if (texture_barrier_count) {
		if (src_stage_flags == 0) {
			src_stage_flags = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
		}

		vkCmdPipelineBarrier(cl->command_buffer, src_stage_flags, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, 0, nullptr, texture_barrier_count, texture_barriers);
	}

#if 0
	{ // Validate that textures bound are not attached as framebuffer bindings.
		uint32_t attachable_count = uniform_set->attachable_textures.size();
		const RID *attachable_ptr = uniform_set->attachable_textures.ptr();
		uint32_t bound_count = draw_list_bound_textures.size();
		const RID *bound_ptr = draw_list_bound_textures.ptr();
		for (uint32_t i = 0; i < attachable_count; i++) {
			for (uint32_t j = 0; j < bound_count; j++) {
				ERR_FAIL_COND_MSG(attachable_ptr[i] == bound_ptr[j],
						"Attempted to use the same texture in framebuffer attachment and a uniform set, this is not allowed.");
			}
		}
	}
#endif
}

void RenderingDeviceVulkan::compute_list_set_push_constant(ComputeListID p_list, const void *p_data, uint32_t p_data_size) {
	ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST);
	ERR_FAIL_COND(!compute_list);

	ComputeList *cl = compute_list;

#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!cl->validation.active, "Submitted Compute Lists can no longer be modified.");
#endif

#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(p_data_size != cl->validation.pipeline_push_constant_size,
			"This compute pipeline requires (" + itos(cl->validation.pipeline_push_constant_size) + ") bytes of push constant data, supplied: (" + itos(p_data_size) + ")");
#endif
	vkCmdPushConstants(cl->command_buffer, cl->state.pipeline_layout, cl->state.pipeline_push_constant_stages, 0, p_data_size, p_data);
#ifdef DEBUG_ENABLED
	cl->validation.pipeline_push_constant_supplied = true;
#endif
}

void RenderingDeviceVulkan::compute_list_dispatch(ComputeListID p_list, uint32_t p_x_groups, uint32_t p_y_groups, uint32_t p_z_groups) {
	ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST);
	ERR_FAIL_COND(!compute_list);

	ComputeList *cl = compute_list;

#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(p_x_groups == 0, "Dispatch amount of X compute groups (" + itos(p_x_groups) + ") is zero.");
	ERR_FAIL_COND_MSG(p_z_groups == 0, "Dispatch amount of Z compute groups (" + itos(p_z_groups) + ") is zero.");
	ERR_FAIL_COND_MSG(p_y_groups == 0, "Dispatch amount of Y compute groups (" + itos(p_y_groups) + ") is zero.");
	ERR_FAIL_COND_MSG(p_x_groups > limits.maxComputeWorkGroupCount[0],
			"Dispatch amount of X compute groups (" + itos(p_x_groups) + ") is larger than device limit (" + itos(limits.maxComputeWorkGroupCount[0]) + ")");
	ERR_FAIL_COND_MSG(p_y_groups > limits.maxComputeWorkGroupCount[1],
			"Dispatch amount of Y compute groups (" + itos(p_y_groups) + ") is larger than device limit (" + itos(limits.maxComputeWorkGroupCount[1]) + ")");
	ERR_FAIL_COND_MSG(p_z_groups > limits.maxComputeWorkGroupCount[2],
			"Dispatch amount of Z compute groups (" + itos(p_z_groups) + ") is larger than device limit (" + itos(limits.maxComputeWorkGroupCount[2]) + ")");

	ERR_FAIL_COND_MSG(!cl->validation.active, "Submitted Compute Lists can no longer be modified.");
#endif

#ifdef DEBUG_ENABLED

	ERR_FAIL_COND_MSG(!cl->validation.pipeline_active, "No compute pipeline was set before attempting to draw.");

	if (cl->validation.pipeline_push_constant_size > 0) {
		// Using push constants, check that they were supplied.
		ERR_FAIL_COND_MSG(!cl->validation.pipeline_push_constant_supplied,
				"The shader in this pipeline requires a push constant to be set before drawing, but it's not present.");
	}

#endif

	// Bind descriptor sets.

	for (uint32_t i = 0; i < cl->state.set_count; i++) {
		if (cl->state.sets[i].pipeline_expected_format == 0) {
			continue; // Nothing expected by this pipeline.
		}
#ifdef DEBUG_ENABLED
		if (cl->state.sets[i].pipeline_expected_format != cl->state.sets[i].uniform_set_format) {
			if (cl->state.sets[i].uniform_set_format == 0) {
				ERR_FAIL_MSG("Uniforms were never supplied for set (" + itos(i) + ") at the time of drawing, which are required by the pipeline");
			} else if (uniform_set_owner.owns(cl->state.sets[i].uniform_set)) {
				UniformSet *us = uniform_set_owner.get_or_null(cl->state.sets[i].uniform_set);
				ERR_FAIL_MSG("Uniforms supplied for set (" + itos(i) + "):\n" + _shader_uniform_debug(us->shader_id, us->shader_set) + "\nare not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n" + _shader_uniform_debug(cl->state.pipeline_shader));
			} else {
				ERR_FAIL_MSG("Uniforms supplied for set (" + itos(i) + ", which was was just freed) are not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n" + _shader_uniform_debug(cl->state.pipeline_shader));
			}
		}
#endif
		if (!cl->state.sets[i].bound) {
			// All good, see if this requires re-binding.
			vkCmdBindDescriptorSets(cl->command_buffer, VK_PIPELINE_BIND_POINT_COMPUTE, cl->state.pipeline_layout, i, 1, &cl->state.sets[i].descriptor_set, 0, nullptr);
			cl->state.sets[i].bound = true;
		}
	}

	vkCmdDispatch(cl->command_buffer, p_x_groups, p_y_groups, p_z_groups);
}

void RenderingDeviceVulkan::compute_list_dispatch_threads(ComputeListID p_list, uint32_t p_x_threads, uint32_t p_y_threads, uint32_t p_z_threads) {
	ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST);
	ERR_FAIL_COND(!compute_list);

#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(p_x_threads == 0, "Dispatch amount of X compute threads (" + itos(p_x_threads) + ") is zero.");
	ERR_FAIL_COND_MSG(p_y_threads == 0, "Dispatch amount of Y compute threads (" + itos(p_y_threads) + ") is zero.");
	ERR_FAIL_COND_MSG(p_z_threads == 0, "Dispatch amount of Z compute threads (" + itos(p_z_threads) + ") is zero.");
#endif

	ComputeList *cl = compute_list;

#ifdef DEBUG_ENABLED

	ERR_FAIL_COND_MSG(!cl->validation.pipeline_active, "No compute pipeline was set before attempting to draw.");

	if (cl->validation.pipeline_push_constant_size > 0) {
		// Using push constants, check that they were supplied.
		ERR_FAIL_COND_MSG(!cl->validation.pipeline_push_constant_supplied,
				"The shader in this pipeline requires a push constant to be set before drawing, but it's not present.");
	}

#endif

	compute_list_dispatch(p_list, (p_x_threads - 1) / cl->state.local_group_size[0] + 1, (p_y_threads - 1) / cl->state.local_group_size[1] + 1, (p_z_threads - 1) / cl->state.local_group_size[2] + 1);
}

void RenderingDeviceVulkan::compute_list_dispatch_indirect(ComputeListID p_list, RID p_buffer, uint32_t p_offset) {
	ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST);
	ERR_FAIL_COND(!compute_list);

	ComputeList *cl = compute_list;
	Buffer *buffer = storage_buffer_owner.get_or_null(p_buffer);
	ERR_FAIL_COND(!buffer);

	ERR_FAIL_COND_MSG(!(buffer->usage & VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT), "Buffer provided was not created to do indirect dispatch.");

	ERR_FAIL_COND_MSG(p_offset + 12 > buffer->size, "Offset provided (+12) is past the end of buffer.");

#ifdef DEBUG_ENABLED
	ERR_FAIL_COND_MSG(!cl->validation.active, "Submitted Compute Lists can no longer be modified.");
#endif

#ifdef DEBUG_ENABLED

	ERR_FAIL_COND_MSG(!cl->validation.pipeline_active, "No compute pipeline was set before attempting to draw.");

	if (cl->validation.pipeline_push_constant_size > 0) {
		// Using push constants, check that they were supplied.
		ERR_FAIL_COND_MSG(!cl->validation.pipeline_push_constant_supplied,
				"The shader in this pipeline requires a push constant to be set before drawing, but it's not present.");
	}

#endif

	// Bind descriptor sets.

	for (uint32_t i = 0; i < cl->state.set_count; i++) {
		if (cl->state.sets[i].pipeline_expected_format == 0) {
			continue; // Nothing expected by this pipeline.
		}
#ifdef DEBUG_ENABLED
		if (cl->state.sets[i].pipeline_expected_format != cl->state.sets[i].uniform_set_format) {
			if (cl->state.sets[i].uniform_set_format == 0) {
				ERR_FAIL_MSG("Uniforms were never supplied for set (" + itos(i) + ") at the time of drawing, which are required by the pipeline");
			} else if (uniform_set_owner.owns(cl->state.sets[i].uniform_set)) {
				UniformSet *us = uniform_set_owner.get_or_null(cl->state.sets[i].uniform_set);
				ERR_FAIL_MSG("Uniforms supplied for set (" + itos(i) + "):\n" + _shader_uniform_debug(us->shader_id, us->shader_set) + "\nare not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n" + _shader_uniform_debug(cl->state.pipeline_shader));
			} else {
				ERR_FAIL_MSG("Uniforms supplied for set (" + itos(i) + ", which was was just freed) are not the same format as required by the pipeline shader. Pipeline shader requires the following bindings:\n" + _shader_uniform_debug(cl->state.pipeline_shader));
			}
		}
#endif
		if (!cl->state.sets[i].bound) {
			// All good, see if this requires re-binding.
			vkCmdBindDescriptorSets(cl->command_buffer, VK_PIPELINE_BIND_POINT_COMPUTE, cl->state.pipeline_layout, i, 1, &cl->state.sets[i].descriptor_set, 0, nullptr);
			cl->state.sets[i].bound = true;
		}
	}

	vkCmdDispatchIndirect(cl->command_buffer, buffer->buffer, p_offset);
}

void RenderingDeviceVulkan::compute_list_add_barrier(ComputeListID p_list) {
#ifdef FORCE_FULL_BARRIER
	_full_barrier(true);
#else
	_memory_barrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_ACCESS_SHADER_WRITE_BIT, VK_ACCESS_SHADER_READ_BIT, true);
#endif
}

void RenderingDeviceVulkan::compute_list_end(BitField<BarrierMask> p_post_barrier) {
	ERR_FAIL_COND(!compute_list);

	uint32_t barrier_flags = 0;
	uint32_t access_flags = 0;
	if (p_post_barrier.has_flag(BARRIER_MASK_COMPUTE)) {
		barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
		access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
	}
	if (p_post_barrier.has_flag(BARRIER_MASK_RASTER)) {
		barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT;
		access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_INDIRECT_COMMAND_READ_BIT;
	}
	if (p_post_barrier.has_flag(BARRIER_MASK_TRANSFER)) {
		barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
		access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT;
	}

	if (barrier_flags == 0) {
		barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
	}

	VkImageMemoryBarrier *image_barriers = nullptr;

	uint32_t image_barrier_count = compute_list->state.textures_to_sampled_layout.size();

	if (image_barrier_count) {
		image_barriers = (VkImageMemoryBarrier *)alloca(sizeof(VkImageMemoryBarrier) * image_barrier_count);
	}

	uint32_t barrier_idx = 0;

	for (Texture *E : compute_list->state.textures_to_sampled_layout) {
		VkImageMemoryBarrier &image_memory_barrier = image_barriers[barrier_idx++];
		image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
		image_memory_barrier.pNext = nullptr;
		image_memory_barrier.srcAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
		image_memory_barrier.dstAccessMask = access_flags;
		image_memory_barrier.oldLayout = E->layout;
		image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;

		image_memory_barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
		image_memory_barrier.image = E->image;
		image_memory_barrier.subresourceRange.aspectMask = E->read_aspect_mask;
		image_memory_barrier.subresourceRange.baseMipLevel = E->base_mipmap;
		image_memory_barrier.subresourceRange.levelCount = E->mipmaps;
		image_memory_barrier.subresourceRange.baseArrayLayer = E->base_layer;
		image_memory_barrier.subresourceRange.layerCount = E->layers;

		E->layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;

		if (E->used_in_frame != frames_drawn) {
			E->used_in_transfer = false;
			E->used_in_raster = false;
			E->used_in_compute = false;
			E->used_in_frame = frames_drawn;
		}
	}

	VkMemoryBarrier mem_barrier;
	mem_barrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER;
	mem_barrier.pNext = nullptr;
	mem_barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT;
	mem_barrier.dstAccessMask = access_flags;

	if (image_barrier_count > 0 || p_post_barrier != BARRIER_MASK_NO_BARRIER) {
		vkCmdPipelineBarrier(compute_list->command_buffer, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, barrier_flags, 0, 1, &mem_barrier, 0, nullptr, image_barrier_count, image_barriers);
	}

#ifdef FORCE_FULL_BARRIER
	_full_barrier(true);
#endif

	memdelete(compute_list);
	compute_list = nullptr;

	// Compute_list is no longer active.
	_THREAD_SAFE_UNLOCK_
}

void RenderingDeviceVulkan::barrier(BitField<BarrierMask> p_from, BitField<BarrierMask> p_to) {
	uint32_t src_barrier_flags = 0;
	uint32_t src_access_flags = 0;

	if (p_from == 0) {
		src_barrier_flags = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
	} else {
		if (p_from.has_flag(BARRIER_MASK_COMPUTE)) {
			src_barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
			src_access_flags |= VK_ACCESS_SHADER_WRITE_BIT;
		}
		if (p_from.has_flag(BARRIER_MASK_RASTER)) {
			src_barrier_flags |= VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
			src_access_flags |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
		}
		if (p_from.has_flag(BARRIER_MASK_TRANSFER)) {
			src_barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
			src_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT;
		}
	}

	uint32_t dst_barrier_flags = 0;
	uint32_t dst_access_flags = 0;

	if (p_to == 0) {
		dst_barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT;
	} else {
		if (p_to.has_flag(BARRIER_MASK_COMPUTE)) {
			dst_barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
			dst_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
		}
		if (p_to.has_flag(BARRIER_MASK_RASTER)) {
			dst_barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT;
			dst_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_INDIRECT_COMMAND_READ_BIT;
		}
		if (p_to.has_flag(BARRIER_MASK_TRANSFER)) {
			dst_barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT;
			dst_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT;
		}
	}

	_memory_barrier(src_barrier_flags, dst_barrier_flags, src_access_flags, dst_access_flags, true);
}

void RenderingDeviceVulkan::full_barrier() {
#ifndef DEBUG_ENABLED
	ERR_PRINT("Full barrier is debug-only, should not be used in production");
#endif
	_full_barrier(true);
}

#if 0
void RenderingDeviceVulkan::draw_list_render_secondary_to_framebuffer(ID p_framebuffer, ID *p_draw_lists, uint32_t p_draw_list_count, InitialAction p_initial_action, FinalAction p_final_action, const Vector<Variant> &p_clear_colors) {
	VkCommandBuffer frame_cmdbuf = frames[frame].frame_buffer;
	ERR_FAIL_COND(!frame_cmdbuf);

	VkRenderPassBeginInfo render_pass_begin;
	render_pass_begin.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO;
	render_pass_begin.pNext = nullptr;
	render_pass_begin.renderPass = context->get_render_pass();
	render_pass_begin.framebuffer = context->get_frame_framebuffer(frame);

	render_pass_begin.renderArea.extent.width = context->get_screen_width(p_screen);
	render_pass_begin.renderArea.extent.height = context->get_screen_height(p_screen);
	render_pass_begin.renderArea.offset.x = 0;
	render_pass_begin.renderArea.offset.y = 0;

	render_pass_begin.clearValueCount = 1;

	VkClearValue clear_value;
	clear_value.color.float32[0] = p_clear_color.r;
	clear_value.color.float32[1] = p_clear_color.g;
	clear_value.color.float32[2] = p_clear_color.b;
	clear_value.color.float32[3] = p_clear_color.a;

	render_pass_begin.pClearValues = &clear_value;

	vkCmdBeginRenderPass(frame_cmdbuf, &render_pass_begin, VK_SUBPASS_CONTENTS_SECONDARY_COMMAND_BUFFERS);

	ID screen_format = screen_get_framebuffer_format();
	{
		VkCommandBuffer *command_buffers = (VkCommandBuffer *)alloca(sizeof(VkCommandBuffer) * p_draw_list_count);
		uint32_t command_buffer_count = 0;

		for (uint32_t i = 0; i < p_draw_list_count; i++) {
			DrawList *dl = _get_draw_list_ptr(p_draw_lists[i]);
			ERR_CONTINUE_MSG(!dl, "Draw list index (" + itos(i) + ") is not a valid draw list ID.");
			ERR_CONTINUE_MSG(dl->validation.framebuffer_format != p_format_check,
					"Draw list index (" + itos(i) + ") is created with a framebuffer format incompatible with this render pass.");

			if (dl->validation.active) {
				// Needs to be closed, so close it.
				vkEndCommandBuffer(dl->command_buffer);
				dl->validation.active = false;
			}

			command_buffers[command_buffer_count++] = dl->command_buffer;
		}

		print_line("to draw: " + itos(command_buffer_count));
		vkCmdExecuteCommands(p_primary, command_buffer_count, command_buffers);
	}

	vkCmdEndRenderPass(frame_cmdbuf);
}
#endif

void RenderingDeviceVulkan::_free_internal(RID p_id) {
#ifdef DEV_ENABLED
	String resource_name;
	if (resource_names.has(p_id)) {
		resource_name = resource_names[p_id];
		resource_names.erase(p_id);
	}
#endif

	// Push everything so it's disposed of next time this frame index is processed (means, it's safe to do it).
	if (texture_owner.owns(p_id)) {
		Texture *texture = texture_owner.get_or_null(p_id);
		frames[frame].textures_to_dispose_of.push_back(*texture);
		texture_owner.free(p_id);
	} else if (framebuffer_owner.owns(p_id)) {
		Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_id);
		frames[frame].framebuffers_to_dispose_of.push_back(*framebuffer);

		if (framebuffer->invalidated_callback != nullptr) {
			framebuffer->invalidated_callback(framebuffer->invalidated_callback_userdata);
		}

		framebuffer_owner.free(p_id);
	} else if (sampler_owner.owns(p_id)) {
		VkSampler *sampler = sampler_owner.get_or_null(p_id);
		frames[frame].samplers_to_dispose_of.push_back(*sampler);
		sampler_owner.free(p_id);
	} else if (vertex_buffer_owner.owns(p_id)) {
		Buffer *vertex_buffer = vertex_buffer_owner.get_or_null(p_id);
		frames[frame].buffers_to_dispose_of.push_back(*vertex_buffer);
		vertex_buffer_owner.free(p_id);
	} else if (vertex_array_owner.owns(p_id)) {
		vertex_array_owner.free(p_id);
	} else if (index_buffer_owner.owns(p_id)) {
		IndexBuffer *index_buffer = index_buffer_owner.get_or_null(p_id);
		Buffer b;
		b.allocation = index_buffer->allocation;
		b.buffer = index_buffer->buffer;
		b.size = index_buffer->size;
		b.buffer_info = {};
		frames[frame].buffers_to_dispose_of.push_back(b);
		index_buffer_owner.free(p_id);
	} else if (index_array_owner.owns(p_id)) {
		index_array_owner.free(p_id);
	} else if (shader_owner.owns(p_id)) {
		Shader *shader = shader_owner.get_or_null(p_id);
		frames[frame].shaders_to_dispose_of.push_back(*shader);
		shader_owner.free(p_id);
	} else if (uniform_buffer_owner.owns(p_id)) {
		Buffer *uniform_buffer = uniform_buffer_owner.get_or_null(p_id);
		frames[frame].buffers_to_dispose_of.push_back(*uniform_buffer);
		uniform_buffer_owner.free(p_id);
	} else if (texture_buffer_owner.owns(p_id)) {
		TextureBuffer *texture_buffer = texture_buffer_owner.get_or_null(p_id);
		frames[frame].buffers_to_dispose_of.push_back(texture_buffer->buffer);
		frames[frame].buffer_views_to_dispose_of.push_back(texture_buffer->view);
		texture_buffer_owner.free(p_id);
	} else if (storage_buffer_owner.owns(p_id)) {
		Buffer *storage_buffer = storage_buffer_owner.get_or_null(p_id);
		frames[frame].buffers_to_dispose_of.push_back(*storage_buffer);
		storage_buffer_owner.free(p_id);
	} else if (uniform_set_owner.owns(p_id)) {
		UniformSet *uniform_set = uniform_set_owner.get_or_null(p_id);
		frames[frame].uniform_sets_to_dispose_of.push_back(*uniform_set);
		uniform_set_owner.free(p_id);

		if (uniform_set->invalidated_callback != nullptr) {
			uniform_set->invalidated_callback(uniform_set->invalidated_callback_userdata);
		}
	} else if (render_pipeline_owner.owns(p_id)) {
		RenderPipeline *pipeline = render_pipeline_owner.get_or_null(p_id);
		frames[frame].render_pipelines_to_dispose_of.push_back(*pipeline);
		render_pipeline_owner.free(p_id);
	} else if (compute_pipeline_owner.owns(p_id)) {
		ComputePipeline *pipeline = compute_pipeline_owner.get_or_null(p_id);
		frames[frame].compute_pipelines_to_dispose_of.push_back(*pipeline);
		compute_pipeline_owner.free(p_id);
	} else {
#ifdef DEV_ENABLED
		ERR_PRINT("Attempted to free invalid ID: " + itos(p_id.get_id()) + " " + resource_name);
#else
		ERR_PRINT("Attempted to free invalid ID: " + itos(p_id.get_id()));
#endif
	}
}

void RenderingDeviceVulkan::free(RID p_id) {
	_THREAD_SAFE_METHOD_

	_free_dependencies(p_id); // Recursively erase dependencies first, to avoid potential API problems.
	_free_internal(p_id);
}

// The full list of resources that can be named is in the VkObjectType enum.
// We just expose the resources that are owned and can be accessed easily.
void RenderingDeviceVulkan::set_resource_name(RID p_id, const String p_name) {
	if (texture_owner.owns(p_id)) {
		Texture *texture = texture_owner.get_or_null(p_id);
		if (texture->owner.is_null()) {
			// Don't set the source texture's name when calling on a texture view.
			context->set_object_name(VK_OBJECT_TYPE_IMAGE, uint64_t(texture->image), p_name);
		}
		context->set_object_name(VK_OBJECT_TYPE_IMAGE_VIEW, uint64_t(texture->view), p_name + " View");
	} else if (framebuffer_owner.owns(p_id)) {
		//Framebuffer *framebuffer = framebuffer_owner.get_or_null(p_id);
		// Not implemented for now as the relationship between Framebuffer and RenderPass is very complex.
	} else if (sampler_owner.owns(p_id)) {
		VkSampler *sampler = sampler_owner.get_or_null(p_id);
		context->set_object_name(VK_OBJECT_TYPE_SAMPLER, uint64_t(*sampler), p_name);
	} else if (vertex_buffer_owner.owns(p_id)) {
		Buffer *vertex_buffer = vertex_buffer_owner.get_or_null(p_id);
		context->set_object_name(VK_OBJECT_TYPE_BUFFER, uint64_t(vertex_buffer->buffer), p_name);
	} else if (index_buffer_owner.owns(p_id)) {
		IndexBuffer *index_buffer = index_buffer_owner.get_or_null(p_id);
		context->set_object_name(VK_OBJECT_TYPE_BUFFER, uint64_t(index_buffer->buffer), p_name);
	} else if (shader_owner.owns(p_id)) {
		Shader *shader = shader_owner.get_or_null(p_id);
		context->set_object_name(VK_OBJECT_TYPE_PIPELINE_LAYOUT, uint64_t(shader->pipeline_layout), p_name + " Pipeline Layout");
		for (int i = 0; i < shader->sets.size(); i++) {
			context->set_object_name(VK_OBJECT_TYPE_DESCRIPTOR_SET_LAYOUT, uint64_t(shader->sets[i].descriptor_set_layout), p_name);
		}
	} else if (uniform_buffer_owner.owns(p_id)) {
		Buffer *uniform_buffer = uniform_buffer_owner.get_or_null(p_id);
		context->set_object_name(VK_OBJECT_TYPE_BUFFER, uint64_t(uniform_buffer->buffer), p_name);
	} else if (texture_buffer_owner.owns(p_id)) {
		TextureBuffer *texture_buffer = texture_buffer_owner.get_or_null(p_id);
		context->set_object_name(VK_OBJECT_TYPE_BUFFER, uint64_t(texture_buffer->buffer.buffer), p_name);
		context->set_object_name(VK_OBJECT_TYPE_BUFFER_VIEW, uint64_t(texture_buffer->view), p_name + " View");
	} else if (storage_buffer_owner.owns(p_id)) {
		Buffer *storage_buffer = storage_buffer_owner.get_or_null(p_id);
		context->set_object_name(VK_OBJECT_TYPE_BUFFER, uint64_t(storage_buffer->buffer), p_name);
	} else if (uniform_set_owner.owns(p_id)) {
		UniformSet *uniform_set = uniform_set_owner.get_or_null(p_id);
		context->set_object_name(VK_OBJECT_TYPE_DESCRIPTOR_SET, uint64_t(uniform_set->descriptor_set), p_name);
	} else if (render_pipeline_owner.owns(p_id)) {
		RenderPipeline *pipeline = render_pipeline_owner.get_or_null(p_id);
		context->set_object_name(VK_OBJECT_TYPE_PIPELINE, uint64_t(pipeline->pipeline), p_name);
		context->set_object_name(VK_OBJECT_TYPE_PIPELINE_LAYOUT, uint64_t(pipeline->pipeline_layout), p_name + " Layout");
	} else if (compute_pipeline_owner.owns(p_id)) {
		ComputePipeline *pipeline = compute_pipeline_owner.get_or_null(p_id);
		context->set_object_name(VK_OBJECT_TYPE_PIPELINE, uint64_t(pipeline->pipeline), p_name);
		context->set_object_name(VK_OBJECT_TYPE_PIPELINE_LAYOUT, uint64_t(pipeline->pipeline_layout), p_name + " Layout");
	} else {
		ERR_PRINT("Attempted to name invalid ID: " + itos(p_id.get_id()));
		return;
	}
#ifdef DEV_ENABLED
	resource_names[p_id] = p_name;
#endif
}

void RenderingDeviceVulkan::draw_command_begin_label(String p_label_name, const Color p_color) {
	context->command_begin_label(frames[frame].draw_command_buffer, p_label_name, p_color);
}

void RenderingDeviceVulkan::draw_command_insert_label(String p_label_name, const Color p_color) {
	context->command_insert_label(frames[frame].draw_command_buffer, p_label_name, p_color);
}

void RenderingDeviceVulkan::draw_command_end_label() {
	context->command_end_label(frames[frame].draw_command_buffer);
}

String RenderingDeviceVulkan::get_device_vendor_name() const {
	return context->get_device_vendor_name();
}

String RenderingDeviceVulkan::get_device_name() const {
	return context->get_device_name();
}

RenderingDevice::DeviceType RenderingDeviceVulkan::get_device_type() const {
	return context->get_device_type();
}

String RenderingDeviceVulkan::get_device_api_version() const {
	return context->get_device_api_version();
}

String RenderingDeviceVulkan::get_device_pipeline_cache_uuid() const {
	return context->get_device_pipeline_cache_uuid();
}

void RenderingDeviceVulkan::_finalize_command_bufers() {
	if (draw_list) {
		ERR_PRINT("Found open draw list at the end of the frame, this should never happen (further drawing will likely not work).");
	}

	if (compute_list) {
		ERR_PRINT("Found open compute list at the end of the frame, this should never happen (further compute will likely not work).");
	}

	{ // Complete the setup buffer (that needs to be processed before anything else).
		vkEndCommandBuffer(frames[frame].setup_command_buffer);
		vkEndCommandBuffer(frames[frame].draw_command_buffer);
	}
}

void RenderingDeviceVulkan::_begin_frame() {
	// Erase pending resources.
	_free_pending_resources(frame);

	// Create setup command buffer and set as the setup buffer.

	{
		VkCommandBufferBeginInfo cmdbuf_begin;
		cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
		cmdbuf_begin.pNext = nullptr;
		cmdbuf_begin.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
		cmdbuf_begin.pInheritanceInfo = nullptr;

		VkResult err = vkResetCommandBuffer(frames[frame].setup_command_buffer, 0);
		ERR_FAIL_COND_MSG(err, "vkResetCommandBuffer failed with error " + itos(err) + ".");

		err = vkBeginCommandBuffer(frames[frame].setup_command_buffer, &cmdbuf_begin);
		ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
		err = vkBeginCommandBuffer(frames[frame].draw_command_buffer, &cmdbuf_begin);
		ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");

		if (local_device.is_null()) {
			context->append_command_buffer(frames[frame].draw_command_buffer);
			context->set_setup_buffer(frames[frame].setup_command_buffer); // Append now so it's added before everything else.
		}
	}

	// Advance current frame.
	frames_drawn++;
	// Advance staging buffer if used.
	if (staging_buffer_used) {
		staging_buffer_current = (staging_buffer_current + 1) % staging_buffer_blocks.size();
		staging_buffer_used = false;
	}

	if (frames[frame].timestamp_count) {
		vkGetQueryPoolResults(device, frames[frame].timestamp_pool, 0, frames[frame].timestamp_count, sizeof(uint64_t) * max_timestamp_query_elements, frames[frame].timestamp_result_values.ptr(), sizeof(uint64_t), VK_QUERY_RESULT_64_BIT);
		vkCmdResetQueryPool(frames[frame].setup_command_buffer, frames[frame].timestamp_pool, 0, frames[frame].timestamp_count);
		SWAP(frames[frame].timestamp_names, frames[frame].timestamp_result_names);
		SWAP(frames[frame].timestamp_cpu_values, frames[frame].timestamp_cpu_result_values);
	}

	frames[frame].timestamp_result_count = frames[frame].timestamp_count;
	frames[frame].timestamp_count = 0;
	frames[frame].index = Engine::get_singleton()->get_frames_drawn();
}

VkSampleCountFlagBits RenderingDeviceVulkan::_ensure_supported_sample_count(TextureSamples p_requested_sample_count) const {
	VkSampleCountFlags sample_count_flags = limits.framebufferColorSampleCounts & limits.framebufferDepthSampleCounts;

	if (sample_count_flags & rasterization_sample_count[p_requested_sample_count]) {
		// The requested sample count is supported.
		return rasterization_sample_count[p_requested_sample_count];
	} else {
		// Find the closest lower supported sample count.
		VkSampleCountFlagBits sample_count = rasterization_sample_count[p_requested_sample_count];
		while (sample_count > VK_SAMPLE_COUNT_1_BIT) {
			if (sample_count_flags & sample_count) {
				return sample_count;
			}
			sample_count = (VkSampleCountFlagBits)(sample_count >> 1);
		}
	}
	return VK_SAMPLE_COUNT_1_BIT;
}

void RenderingDeviceVulkan::swap_buffers() {
	ERR_FAIL_COND_MSG(local_device.is_valid(), "Local devices can't swap buffers.");
	_THREAD_SAFE_METHOD_

	_finalize_command_bufers();

	screen_prepared = false;
	// Swap buffers.
	context->swap_buffers();

	frame = (frame + 1) % frame_count;

	_begin_frame();
}

void RenderingDeviceVulkan::submit() {
	ERR_FAIL_COND_MSG(local_device.is_null(), "Only local devices can submit and sync.");
	ERR_FAIL_COND_MSG(local_device_processing, "device already submitted, call sync to wait until done.");

	_finalize_command_bufers();

	VkCommandBuffer command_buffers[2] = { frames[frame].setup_command_buffer, frames[frame].draw_command_buffer };
	context->local_device_push_command_buffers(local_device, command_buffers, 2);
	local_device_processing = true;
}

void RenderingDeviceVulkan::sync() {
	ERR_FAIL_COND_MSG(local_device.is_null(), "Only local devices can submit and sync.");
	ERR_FAIL_COND_MSG(!local_device_processing, "sync can only be called after a submit");

	context->local_device_sync(local_device);
	_begin_frame();
	local_device_processing = false;
}

VmaPool RenderingDeviceVulkan::_find_or_create_small_allocs_pool(uint32_t p_mem_type_index) {
	if (small_allocs_pools.has(p_mem_type_index)) {
		return small_allocs_pools[p_mem_type_index];
	}

	print_verbose("Creating VMA small objects pool for memory type index " + itos(p_mem_type_index));

	VmaPoolCreateInfo pci;
	pci.memoryTypeIndex = p_mem_type_index;
	pci.flags = 0;
	pci.blockSize = 0;
	pci.minBlockCount = 0;
	pci.maxBlockCount = SIZE_MAX;
	pci.priority = 0.5f;
	pci.minAllocationAlignment = 0;
	pci.pMemoryAllocateNext = nullptr;
	VmaPool pool = VK_NULL_HANDLE;
	VkResult res = vmaCreatePool(allocator, &pci, &pool);
	small_allocs_pools[p_mem_type_index] = pool; // Don't try to create it again if failed the first time.
	ERR_FAIL_COND_V_MSG(res, pool, "vmaCreatePool failed with error " + itos(res) + ".");

	return pool;
}

void RenderingDeviceVulkan::_free_pending_resources(int p_frame) {
	// Free in dependency usage order, so nothing weird happens.
	// Pipelines.
	while (frames[p_frame].render_pipelines_to_dispose_of.front()) {
		RenderPipeline *pipeline = &frames[p_frame].render_pipelines_to_dispose_of.front()->get();

		vkDestroyPipeline(device, pipeline->pipeline, nullptr);

		frames[p_frame].render_pipelines_to_dispose_of.pop_front();
	}

	while (frames[p_frame].compute_pipelines_to_dispose_of.front()) {
		ComputePipeline *pipeline = &frames[p_frame].compute_pipelines_to_dispose_of.front()->get();

		vkDestroyPipeline(device, pipeline->pipeline, nullptr);

		frames[p_frame].compute_pipelines_to_dispose_of.pop_front();
	}

	// Uniform sets.
	while (frames[p_frame].uniform_sets_to_dispose_of.front()) {
		UniformSet *uniform_set = &frames[p_frame].uniform_sets_to_dispose_of.front()->get();

		vkFreeDescriptorSets(device, uniform_set->pool->pool, 1, &uniform_set->descriptor_set);
		_descriptor_pool_free(uniform_set->pool_key, uniform_set->pool);

		frames[p_frame].uniform_sets_to_dispose_of.pop_front();
	}

	// Buffer views.
	while (frames[p_frame].buffer_views_to_dispose_of.front()) {
		VkBufferView buffer_view = frames[p_frame].buffer_views_to_dispose_of.front()->get();

		vkDestroyBufferView(device, buffer_view, nullptr);

		frames[p_frame].buffer_views_to_dispose_of.pop_front();
	}

	// Shaders.
	while (frames[p_frame].shaders_to_dispose_of.front()) {
		Shader *shader = &frames[p_frame].shaders_to_dispose_of.front()->get();

		// Descriptor set layout for each set.
		for (int i = 0; i < shader->sets.size(); i++) {
			vkDestroyDescriptorSetLayout(device, shader->sets[i].descriptor_set_layout, nullptr);
		}

		// Pipeline layout.
		vkDestroyPipelineLayout(device, shader->pipeline_layout, nullptr);

		// Shaders themselves.
		for (int i = 0; i < shader->pipeline_stages.size(); i++) {
			vkDestroyShaderModule(device, shader->pipeline_stages[i].module, nullptr);
		}

		frames[p_frame].shaders_to_dispose_of.pop_front();
	}

	// Samplers.
	while (frames[p_frame].samplers_to_dispose_of.front()) {
		VkSampler sampler = frames[p_frame].samplers_to_dispose_of.front()->get();

		vkDestroySampler(device, sampler, nullptr);

		frames[p_frame].samplers_to_dispose_of.pop_front();
	}

	// Framebuffers.
	while (frames[p_frame].framebuffers_to_dispose_of.front()) {
		Framebuffer *framebuffer = &frames[p_frame].framebuffers_to_dispose_of.front()->get();

		for (const KeyValue<Framebuffer::VersionKey, Framebuffer::Version> &E : framebuffer->framebuffers) {
			// First framebuffer, then render pass because it depends on it.
			vkDestroyFramebuffer(device, E.value.framebuffer, nullptr);
			vkDestroyRenderPass(device, E.value.render_pass, nullptr);
		}

		frames[p_frame].framebuffers_to_dispose_of.pop_front();
	}

	// Textures.
	while (frames[p_frame].textures_to_dispose_of.front()) {
		Texture *texture = &frames[p_frame].textures_to_dispose_of.front()->get();

		if (texture->bound) {
			WARN_PRINT("Deleted a texture while it was bound.");
		}
		vkDestroyImageView(device, texture->view, nullptr);
		if (texture->owner.is_null()) {
			// Actually owns the image and the allocation too.
			image_memory -= texture->allocation_info.size;
			vmaDestroyImage(allocator, texture->image, texture->allocation);
		}
		frames[p_frame].textures_to_dispose_of.pop_front();
	}

	// Buffers.
	while (frames[p_frame].buffers_to_dispose_of.front()) {
		_buffer_free(&frames[p_frame].buffers_to_dispose_of.front()->get());

		frames[p_frame].buffers_to_dispose_of.pop_front();
	}
}

void RenderingDeviceVulkan::prepare_screen_for_drawing() {
	_THREAD_SAFE_METHOD_
	context->prepare_buffers();
	screen_prepared = true;
}

uint32_t RenderingDeviceVulkan::get_frame_delay() const {
	return frame_count;
}

uint64_t RenderingDeviceVulkan::get_memory_usage(MemoryType p_type) const {
	if (p_type == MEMORY_BUFFERS) {
		return buffer_memory;
	} else if (p_type == MEMORY_TEXTURES) {
		return image_memory;
	} else {
		VmaTotalStatistics stats;
		vmaCalculateStatistics(allocator, &stats);
		return stats.total.statistics.allocationBytes;
	}
}

void RenderingDeviceVulkan::_flush(bool p_current_frame) {
	if (local_device.is_valid() && !p_current_frame) {
		return; // Flushing previous frames has no effect with local device.
	}
	// Not doing this crashes RADV (undefined behavior).
	if (p_current_frame) {
		vkEndCommandBuffer(frames[frame].setup_command_buffer);
		vkEndCommandBuffer(frames[frame].draw_command_buffer);
	}

	if (local_device.is_valid()) {
		VkCommandBuffer command_buffers[2] = { frames[frame].setup_command_buffer, frames[frame].draw_command_buffer };
		context->local_device_push_command_buffers(local_device, command_buffers, 2);
		context->local_device_sync(local_device);

		VkCommandBufferBeginInfo cmdbuf_begin;
		cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
		cmdbuf_begin.pNext = nullptr;
		cmdbuf_begin.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
		cmdbuf_begin.pInheritanceInfo = nullptr;

		VkResult err = vkBeginCommandBuffer(frames[frame].setup_command_buffer, &cmdbuf_begin);
		ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
		err = vkBeginCommandBuffer(frames[frame].draw_command_buffer, &cmdbuf_begin);
		ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");

	} else {
		context->flush(p_current_frame, p_current_frame);
		// Re-create the setup command.
		if (p_current_frame) {
			VkCommandBufferBeginInfo cmdbuf_begin;
			cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
			cmdbuf_begin.pNext = nullptr;
			cmdbuf_begin.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
			cmdbuf_begin.pInheritanceInfo = nullptr;

			VkResult err = vkBeginCommandBuffer(frames[frame].setup_command_buffer, &cmdbuf_begin);
			ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
			context->set_setup_buffer(frames[frame].setup_command_buffer); // Append now so it's added before everything else.
		}

		if (p_current_frame) {
			VkCommandBufferBeginInfo cmdbuf_begin;
			cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
			cmdbuf_begin.pNext = nullptr;
			cmdbuf_begin.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
			cmdbuf_begin.pInheritanceInfo = nullptr;

			VkResult err = vkBeginCommandBuffer(frames[frame].draw_command_buffer, &cmdbuf_begin);
			ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
			context->append_command_buffer(frames[frame].draw_command_buffer);
		}
	}
}

void RenderingDeviceVulkan::initialize(VulkanContext *p_context, bool p_local_device) {
	// Get our device capabilities.
	{
		device_capabilities.version_major = p_context->get_vulkan_major();
		device_capabilities.version_minor = p_context->get_vulkan_minor();
	}

	context = p_context;
	device = p_context->get_device();
	if (p_local_device) {
		frame_count = 1;
		local_device = p_context->local_device_create();
		device = p_context->local_device_get_vk_device(local_device);
	} else {
		frame_count = p_context->get_swapchain_image_count() + 1; // Always need one extra to ensure it's unused at any time, without having to use a fence for this.
	}
	limits = p_context->get_device_limits();
	max_timestamp_query_elements = 256;

	{ // Initialize allocator.

		VmaAllocatorCreateInfo allocatorInfo;
		memset(&allocatorInfo, 0, sizeof(VmaAllocatorCreateInfo));
		allocatorInfo.physicalDevice = p_context->get_physical_device();
		allocatorInfo.device = device;
		allocatorInfo.instance = p_context->get_instance();
		vmaCreateAllocator(&allocatorInfo, &allocator);
	}

	frames.resize(frame_count);
	frame = 0;
	// Create setup and frame buffers.
	for (int i = 0; i < frame_count; i++) {
		frames[i].index = 0;

		{ // Create command pool, one per frame is recommended.
			VkCommandPoolCreateInfo cmd_pool_info;
			cmd_pool_info.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO;
			cmd_pool_info.pNext = nullptr;
			cmd_pool_info.queueFamilyIndex = p_context->get_graphics_queue_family_index();
			cmd_pool_info.flags = VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT;

			VkResult res = vkCreateCommandPool(device, &cmd_pool_info, nullptr, &frames[i].command_pool);
			ERR_FAIL_COND_MSG(res, "vkCreateCommandPool failed with error " + itos(res) + ".");
		}

		{ // Create command buffers.

			VkCommandBufferAllocateInfo cmdbuf;
			// No command buffer exists, create it.
			cmdbuf.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO;
			cmdbuf.pNext = nullptr;
			cmdbuf.commandPool = frames[i].command_pool;
			cmdbuf.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY;
			cmdbuf.commandBufferCount = 1;

			VkResult err = vkAllocateCommandBuffers(device, &cmdbuf, &frames[i].setup_command_buffer);
			ERR_CONTINUE_MSG(err, "vkAllocateCommandBuffers failed with error " + itos(err) + ".");

			err = vkAllocateCommandBuffers(device, &cmdbuf, &frames[i].draw_command_buffer);
			ERR_CONTINUE_MSG(err, "vkAllocateCommandBuffers failed with error " + itos(err) + ".");
		}

		{
			// Create query pool.
			VkQueryPoolCreateInfo query_pool_create_info;
			query_pool_create_info.sType = VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO;
			query_pool_create_info.flags = 0;
			query_pool_create_info.pNext = nullptr;
			query_pool_create_info.queryType = VK_QUERY_TYPE_TIMESTAMP;
			query_pool_create_info.queryCount = max_timestamp_query_elements;
			query_pool_create_info.pipelineStatistics = 0;

			vkCreateQueryPool(device, &query_pool_create_info, nullptr, &frames[i].timestamp_pool);

			frames[i].timestamp_names.resize(max_timestamp_query_elements);
			frames[i].timestamp_cpu_values.resize(max_timestamp_query_elements);
			frames[i].timestamp_count = 0;
			frames[i].timestamp_result_names.resize(max_timestamp_query_elements);
			frames[i].timestamp_cpu_result_values.resize(max_timestamp_query_elements);
			frames[i].timestamp_result_values.resize(max_timestamp_query_elements);
			frames[i].timestamp_result_count = 0;
		}
	}

	{
		// Begin the first command buffer for the first frame, so
		// setting up things can be done in the meantime until swap_buffers(), which is called before advance.
		VkCommandBufferBeginInfo cmdbuf_begin;
		cmdbuf_begin.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO;
		cmdbuf_begin.pNext = nullptr;
		cmdbuf_begin.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT;
		cmdbuf_begin.pInheritanceInfo = nullptr;

		VkResult err = vkBeginCommandBuffer(frames[0].setup_command_buffer, &cmdbuf_begin);
		ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");

		err = vkBeginCommandBuffer(frames[0].draw_command_buffer, &cmdbuf_begin);
		ERR_FAIL_COND_MSG(err, "vkBeginCommandBuffer failed with error " + itos(err) + ".");
		if (local_device.is_null()) {
			context->set_setup_buffer(frames[0].setup_command_buffer); // Append now so it's added before everything else.
			context->append_command_buffer(frames[0].draw_command_buffer);
		}
	}

	staging_buffer_block_size = GLOBAL_GET("rendering/rendering_device/staging_buffer/block_size_kb");
	staging_buffer_block_size = MAX(4u, staging_buffer_block_size);
	staging_buffer_block_size *= 1024; // Kb -> bytes.
	staging_buffer_max_size = GLOBAL_GET("rendering/rendering_device/staging_buffer/max_size_mb");
	staging_buffer_max_size = MAX(1u, staging_buffer_max_size);
	staging_buffer_max_size *= 1024 * 1024;

	if (staging_buffer_max_size < staging_buffer_block_size * 4) {
		// Validate enough blocks.
		staging_buffer_max_size = staging_buffer_block_size * 4;
	}
	texture_upload_region_size_px = GLOBAL_GET("rendering/rendering_device/staging_buffer/texture_upload_region_size_px");
	texture_upload_region_size_px = nearest_power_of_2_templated(texture_upload_region_size_px);

	frames_drawn = frame_count; // Start from frame count, so everything else is immediately old.

	// Ensure current staging block is valid and at least one per frame exists.
	staging_buffer_current = 0;
	staging_buffer_used = false;

	for (int i = 0; i < frame_count; i++) {
		// Staging was never used, create a block.
		Error err = _insert_staging_block();
		ERR_CONTINUE(err != OK);
	}

	max_descriptors_per_pool = GLOBAL_GET("rendering/rendering_device/vulkan/max_descriptors_per_pool");

	// Check to make sure DescriptorPoolKey is good.
	static_assert(sizeof(uint64_t) * 3 >= UNIFORM_TYPE_MAX * sizeof(uint16_t));

	draw_list = nullptr;
	draw_list_count = 0;
	draw_list_split = false;

	compute_list = nullptr;
}

template <class T>
void RenderingDeviceVulkan::_free_rids(T &p_owner, const char *p_type) {
	List<RID> owned;
	p_owner.get_owned_list(&owned);
	if (owned.size()) {
		if (owned.size() == 1) {
			WARN_PRINT(vformat("1 RID of type \"%s\" was leaked.", p_type));
		} else {
			WARN_PRINT(vformat("%d RIDs of type \"%s\" were leaked.", owned.size(), p_type));
		}
		for (const RID &E : owned) {
#ifdef DEV_ENABLED
			if (resource_names.has(E)) {
				print_line(String(" - ") + resource_names[E]);
			}
#endif
			free(E);
		}
	}
}

void RenderingDeviceVulkan::capture_timestamp(const String &p_name) {
	ERR_FAIL_COND_MSG(draw_list != nullptr, "Capturing timestamps during draw list creation is not allowed. Offending timestamp was: " + p_name);
	ERR_FAIL_COND(frames[frame].timestamp_count >= max_timestamp_query_elements);

	// This should be optional for profiling, else it will slow things down.
	{
		VkMemoryBarrier memoryBarrier;

		memoryBarrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER;
		memoryBarrier.pNext = nullptr;
		memoryBarrier.srcAccessMask = VK_ACCESS_INDIRECT_COMMAND_READ_BIT |
				VK_ACCESS_INDEX_READ_BIT |
				VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT |
				VK_ACCESS_UNIFORM_READ_BIT |
				VK_ACCESS_INPUT_ATTACHMENT_READ_BIT |
				VK_ACCESS_SHADER_READ_BIT |
				VK_ACCESS_SHADER_WRITE_BIT |
				VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
				VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
				VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
				VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
				VK_ACCESS_TRANSFER_READ_BIT |
				VK_ACCESS_TRANSFER_WRITE_BIT |
				VK_ACCESS_HOST_READ_BIT |
				VK_ACCESS_HOST_WRITE_BIT;
		memoryBarrier.dstAccessMask = VK_ACCESS_INDIRECT_COMMAND_READ_BIT |
				VK_ACCESS_INDEX_READ_BIT |
				VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT |
				VK_ACCESS_UNIFORM_READ_BIT |
				VK_ACCESS_INPUT_ATTACHMENT_READ_BIT |
				VK_ACCESS_SHADER_READ_BIT |
				VK_ACCESS_SHADER_WRITE_BIT |
				VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
				VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
				VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
				VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
				VK_ACCESS_TRANSFER_READ_BIT |
				VK_ACCESS_TRANSFER_WRITE_BIT |
				VK_ACCESS_HOST_READ_BIT |
				VK_ACCESS_HOST_WRITE_BIT;

		vkCmdPipelineBarrier(frames[frame].draw_command_buffer, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, 1, &memoryBarrier, 0, nullptr, 0, nullptr);
	}

	vkCmdWriteTimestamp(frames[frame].draw_command_buffer, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, frames[frame].timestamp_pool, frames[frame].timestamp_count);
	frames[frame].timestamp_names[frames[frame].timestamp_count] = p_name;
	frames[frame].timestamp_cpu_values[frames[frame].timestamp_count] = OS::get_singleton()->get_ticks_usec();
	frames[frame].timestamp_count++;
}

uint64_t RenderingDeviceVulkan::get_driver_resource(DriverResource p_resource, RID p_rid, uint64_t p_index) {
	_THREAD_SAFE_METHOD_

	switch (p_resource) {
		case DRIVER_RESOURCE_VULKAN_DEVICE: {
			return (uint64_t)context->get_device();
		} break;
		case DRIVER_RESOURCE_VULKAN_PHYSICAL_DEVICE: {
			return (uint64_t)context->get_physical_device();
		} break;
		case DRIVER_RESOURCE_VULKAN_INSTANCE: {
			return (uint64_t)context->get_instance();
		} break;
		case DRIVER_RESOURCE_VULKAN_QUEUE: {
			return (uint64_t)context->get_graphics_queue();
		} break;
		case DRIVER_RESOURCE_VULKAN_QUEUE_FAMILY_INDEX: {
			return context->get_graphics_queue_family_index();
		} break;
		case DRIVER_RESOURCE_VULKAN_IMAGE: {
			Texture *tex = texture_owner.get_or_null(p_rid);
			ERR_FAIL_NULL_V(tex, 0);

			return (uint64_t)tex->image;
		} break;
		case DRIVER_RESOURCE_VULKAN_IMAGE_VIEW: {
			Texture *tex = texture_owner.get_or_null(p_rid);
			ERR_FAIL_NULL_V(tex, 0);

			return (uint64_t)tex->view;
		} break;
		case DRIVER_RESOURCE_VULKAN_IMAGE_NATIVE_TEXTURE_FORMAT: {
			Texture *tex = texture_owner.get_or_null(p_rid);
			ERR_FAIL_NULL_V(tex, 0);

			return vulkan_formats[tex->format];
		} break;
		case DRIVER_RESOURCE_VULKAN_SAMPLER: {
			VkSampler *sampler = sampler_owner.get_or_null(p_rid);
			ERR_FAIL_NULL_V(sampler, 0);

			return uint64_t(*sampler);
		} break;
		case DRIVER_RESOURCE_VULKAN_DESCRIPTOR_SET: {
			UniformSet *uniform_set = uniform_set_owner.get_or_null(p_rid);
			ERR_FAIL_NULL_V(uniform_set, 0);

			return uint64_t(uniform_set->descriptor_set);
		} break;
		case DRIVER_RESOURCE_VULKAN_BUFFER: {
			Buffer *buffer = nullptr;
			if (vertex_buffer_owner.owns(p_rid)) {
				buffer = vertex_buffer_owner.get_or_null(p_rid);
			} else if (index_buffer_owner.owns(p_rid)) {
				buffer = index_buffer_owner.get_or_null(p_rid);
			} else if (uniform_buffer_owner.owns(p_rid)) {
				buffer = uniform_buffer_owner.get_or_null(p_rid);
			} else if (texture_buffer_owner.owns(p_rid)) {
				buffer = &texture_buffer_owner.get_or_null(p_rid)->buffer;
			} else if (storage_buffer_owner.owns(p_rid)) {
				buffer = storage_buffer_owner.get_or_null(p_rid);
			}

			ERR_FAIL_NULL_V(buffer, 0);

			return uint64_t(buffer->buffer);
		} break;
		case DRIVER_RESOURCE_VULKAN_COMPUTE_PIPELINE: {
			ComputePipeline *compute_pipeline = compute_pipeline_owner.get_or_null(p_rid);
			ERR_FAIL_NULL_V(compute_pipeline, 0);

			return uint64_t(compute_pipeline->pipeline);
		} break;
		case DRIVER_RESOURCE_VULKAN_RENDER_PIPELINE: {
			RenderPipeline *render_pipeline = render_pipeline_owner.get_or_null(p_rid);
			ERR_FAIL_NULL_V(render_pipeline, 0);

			return uint64_t(render_pipeline->pipeline);
		} break;
		default: {
			// Not supported for this driver.
			return 0;
		} break;
	}
}

uint32_t RenderingDeviceVulkan::get_captured_timestamps_count() const {
	return frames[frame].timestamp_result_count;
}

uint64_t RenderingDeviceVulkan::get_captured_timestamps_frame() const {
	return frames[frame].index;
}

static void mult64to128(uint64_t u, uint64_t v, uint64_t &h, uint64_t &l) {
	uint64_t u1 = (u & 0xffffffff);
	uint64_t v1 = (v & 0xffffffff);
	uint64_t t = (u1 * v1);
	uint64_t w3 = (t & 0xffffffff);
	uint64_t k = (t >> 32);

	u >>= 32;
	t = (u * v1) + k;
	k = (t & 0xffffffff);
	uint64_t w1 = (t >> 32);

	v >>= 32;
	t = (u1 * v) + k;
	k = (t >> 32);

	h = (u * v) + w1 + k;
	l = (t << 32) + w3;
}

uint64_t RenderingDeviceVulkan::get_captured_timestamp_gpu_time(uint32_t p_index) const {
	ERR_FAIL_UNSIGNED_INDEX_V(p_index, frames[frame].timestamp_result_count, 0);

	// This sucks because timestampPeriod multiplier is a float, while the timestamp is 64 bits nanosecs.
	// So, in cases like nvidia which give you enormous numbers and 1 as multiplier, multiplying is next to impossible.
	// Need to do 128 bits fixed point multiplication to get the right value.

	uint64_t shift_bits = 16;

	uint64_t h, l;

	mult64to128(frames[frame].timestamp_result_values[p_index], uint64_t(double(limits.timestampPeriod) * double(1 << shift_bits)), h, l);
	l >>= shift_bits;
	l |= h << (64 - shift_bits);

	return l;
}

uint64_t RenderingDeviceVulkan::get_captured_timestamp_cpu_time(uint32_t p_index) const {
	ERR_FAIL_UNSIGNED_INDEX_V(p_index, frames[frame].timestamp_result_count, 0);
	return frames[frame].timestamp_cpu_result_values[p_index];
}

String RenderingDeviceVulkan::get_captured_timestamp_name(uint32_t p_index) const {
	ERR_FAIL_UNSIGNED_INDEX_V(p_index, frames[frame].timestamp_result_count, String());
	return frames[frame].timestamp_result_names[p_index];
}

uint64_t RenderingDeviceVulkan::limit_get(Limit p_limit) const {
	switch (p_limit) {
		case LIMIT_MAX_BOUND_UNIFORM_SETS:
			return limits.maxBoundDescriptorSets;
		case LIMIT_MAX_FRAMEBUFFER_COLOR_ATTACHMENTS:
			return limits.maxColorAttachments;
		case LIMIT_MAX_TEXTURES_PER_UNIFORM_SET:
			return limits.maxDescriptorSetSampledImages;
		case LIMIT_MAX_SAMPLERS_PER_UNIFORM_SET:
			return limits.maxDescriptorSetSamplers;
		case LIMIT_MAX_STORAGE_BUFFERS_PER_UNIFORM_SET:
			return limits.maxDescriptorSetStorageBuffers;
		case LIMIT_MAX_STORAGE_IMAGES_PER_UNIFORM_SET:
			return limits.maxDescriptorSetStorageImages;
		case LIMIT_MAX_UNIFORM_BUFFERS_PER_UNIFORM_SET:
			return limits.maxDescriptorSetUniformBuffers;
		case LIMIT_MAX_DRAW_INDEXED_INDEX:
			return limits.maxDrawIndexedIndexValue;
		case LIMIT_MAX_FRAMEBUFFER_HEIGHT:
			return limits.maxFramebufferHeight;
		case LIMIT_MAX_FRAMEBUFFER_WIDTH:
			return limits.maxFramebufferWidth;
		case LIMIT_MAX_TEXTURE_ARRAY_LAYERS:
			return limits.maxImageArrayLayers;
		case LIMIT_MAX_TEXTURE_SIZE_1D:
			return limits.maxImageDimension1D;
		case LIMIT_MAX_TEXTURE_SIZE_2D:
			return limits.maxImageDimension2D;
		case LIMIT_MAX_TEXTURE_SIZE_3D:
			return limits.maxImageDimension3D;
		case LIMIT_MAX_TEXTURE_SIZE_CUBE:
			return limits.maxImageDimensionCube;
		case LIMIT_MAX_TEXTURES_PER_SHADER_STAGE:
			return limits.maxPerStageDescriptorSampledImages;
		case LIMIT_MAX_SAMPLERS_PER_SHADER_STAGE:
			return limits.maxPerStageDescriptorSamplers;
		case LIMIT_MAX_STORAGE_BUFFERS_PER_SHADER_STAGE:
			return limits.maxPerStageDescriptorStorageBuffers;
		case LIMIT_MAX_STORAGE_IMAGES_PER_SHADER_STAGE:
			return limits.maxPerStageDescriptorStorageImages;
		case LIMIT_MAX_UNIFORM_BUFFERS_PER_SHADER_STAGE:
			return limits.maxPerStageDescriptorUniformBuffers;
		case LIMIT_MAX_PUSH_CONSTANT_SIZE:
			return limits.maxPushConstantsSize;
		case LIMIT_MAX_UNIFORM_BUFFER_SIZE:
			return limits.maxUniformBufferRange;
		case LIMIT_MAX_VERTEX_INPUT_ATTRIBUTE_OFFSET:
			return limits.maxVertexInputAttributeOffset;
		case LIMIT_MAX_VERTEX_INPUT_ATTRIBUTES:
			return limits.maxVertexInputAttributes;
		case LIMIT_MAX_VERTEX_INPUT_BINDINGS:
			return limits.maxVertexInputBindings;
		case LIMIT_MAX_VERTEX_INPUT_BINDING_STRIDE:
			return limits.maxVertexInputBindingStride;
		case LIMIT_MIN_UNIFORM_BUFFER_OFFSET_ALIGNMENT:
			return limits.minUniformBufferOffsetAlignment;
		case LIMIT_MAX_COMPUTE_WORKGROUP_COUNT_X:
			return limits.maxComputeWorkGroupCount[0];
		case LIMIT_MAX_COMPUTE_WORKGROUP_COUNT_Y:
			return limits.maxComputeWorkGroupCount[1];
		case LIMIT_MAX_COMPUTE_WORKGROUP_COUNT_Z:
			return limits.maxComputeWorkGroupCount[2];
		case LIMIT_MAX_COMPUTE_WORKGROUP_INVOCATIONS:
			return limits.maxComputeWorkGroupInvocations;
		case LIMIT_MAX_COMPUTE_WORKGROUP_SIZE_X:
			return limits.maxComputeWorkGroupSize[0];
		case LIMIT_MAX_COMPUTE_WORKGROUP_SIZE_Y:
			return limits.maxComputeWorkGroupSize[1];
		case LIMIT_MAX_COMPUTE_WORKGROUP_SIZE_Z:
			return limits.maxComputeWorkGroupSize[2];
		case LIMIT_MAX_VIEWPORT_DIMENSIONS_X:
			return limits.maxViewportDimensions[0];
		case LIMIT_MAX_VIEWPORT_DIMENSIONS_Y:
			return limits.maxViewportDimensions[1];
		case LIMIT_SUBGROUP_SIZE: {
			VulkanContext::SubgroupCapabilities subgroup_capabilities = context->get_subgroup_capabilities();
			return subgroup_capabilities.size;
		}
		case LIMIT_SUBGROUP_IN_SHADERS: {
			VulkanContext::SubgroupCapabilities subgroup_capabilities = context->get_subgroup_capabilities();
			return subgroup_capabilities.supported_stages_flags_rd();
		}
		case LIMIT_SUBGROUP_OPERATIONS: {
			VulkanContext::SubgroupCapabilities subgroup_capabilities = context->get_subgroup_capabilities();
			return subgroup_capabilities.supported_operations_flags_rd();
		}
		case LIMIT_VRS_TEXEL_WIDTH: {
			return context->get_vrs_capabilities().texel_size.x;
		}
		case LIMIT_VRS_TEXEL_HEIGHT: {
			return context->get_vrs_capabilities().texel_size.y;
		}
		default:
			ERR_FAIL_V(0);
	}

	return 0;
}

void RenderingDeviceVulkan::finalize() {
	// Free all resources.

	_flush(false);

	_free_rids(render_pipeline_owner, "Pipeline");
	_free_rids(compute_pipeline_owner, "Compute");
	_free_rids(uniform_set_owner, "UniformSet");
	_free_rids(texture_buffer_owner, "TextureBuffer");
	_free_rids(storage_buffer_owner, "StorageBuffer");
	_free_rids(uniform_buffer_owner, "UniformBuffer");
	_free_rids(shader_owner, "Shader");
	_free_rids(index_array_owner, "IndexArray");
	_free_rids(index_buffer_owner, "IndexBuffer");
	_free_rids(vertex_array_owner, "VertexArray");
	_free_rids(vertex_buffer_owner, "VertexBuffer");
	_free_rids(framebuffer_owner, "Framebuffer");
	_free_rids(sampler_owner, "Sampler");
	{
		// For textures it's a bit more difficult because they may be shared.
		List<RID> owned;
		texture_owner.get_owned_list(&owned);
		if (owned.size()) {
			if (owned.size() == 1) {
				WARN_PRINT("1 RID of type \"Texture\" was leaked.");
			} else {
				WARN_PRINT(vformat("%d RIDs of type \"Texture\" were leaked.", owned.size()));
			}
			// Free shared first.
			for (List<RID>::Element *E = owned.front(); E;) {
				List<RID>::Element *N = E->next();
				if (texture_is_shared(E->get())) {
#ifdef DEV_ENABLED
					if (resource_names.has(E->get())) {
						print_line(String(" - ") + resource_names[E->get()]);
					}
#endif
					free(E->get());
					owned.erase(E);
				}
				E = N;
			}
			// Free non shared second, this will avoid an error trying to free unexisting textures due to dependencies.
			for (const RID &E : owned) {
#ifdef DEV_ENABLED
				if (resource_names.has(E)) {
					print_line(String(" - ") + resource_names[E]);
				}
#endif
				free(E);
			}
		}
	}

	// Free everything pending.
	for (int i = 0; i < frame_count; i++) {
		int f = (frame + i) % frame_count;
		_free_pending_resources(f);
		vkDestroyCommandPool(device, frames[i].command_pool, nullptr);
		vkDestroyQueryPool(device, frames[i].timestamp_pool, nullptr);
	}

	for (int i = 0; i < split_draw_list_allocators.size(); i++) {
		vkDestroyCommandPool(device, split_draw_list_allocators[i].command_pool, nullptr);
	}

	frames.clear();

	for (int i = 0; i < staging_buffer_blocks.size(); i++) {
		vmaDestroyBuffer(allocator, staging_buffer_blocks[i].buffer, staging_buffer_blocks[i].allocation);
	}
	while (small_allocs_pools.size()) {
		HashMap<uint32_t, VmaPool>::Iterator E = small_allocs_pools.begin();
		vmaDestroyPool(allocator, E->value);
		small_allocs_pools.remove(E);
	}
	vmaDestroyAllocator(allocator);

	while (vertex_formats.size()) {
		HashMap<VertexFormatID, VertexDescriptionCache>::Iterator temp = vertex_formats.begin();
		memdelete_arr(temp->value.bindings);
		memdelete_arr(temp->value.attributes);
		vertex_formats.remove(temp);
	}

	for (KeyValue<FramebufferFormatID, FramebufferFormat> &E : framebuffer_formats) {
		vkDestroyRenderPass(device, E.value.render_pass, nullptr);
	}
	framebuffer_formats.clear();

	// All these should be clear at this point.
	ERR_FAIL_COND(descriptor_pools.size());
	ERR_FAIL_COND(dependency_map.size());
	ERR_FAIL_COND(reverse_dependency_map.size());
}

RenderingDevice *RenderingDeviceVulkan::create_local_device() {
	RenderingDeviceVulkan *rd = memnew(RenderingDeviceVulkan);
	rd->initialize(context, true);
	return rd;
}

bool RenderingDeviceVulkan::has_feature(const Features p_feature) const {
	switch (p_feature) {
		case SUPPORTS_MULTIVIEW: {
			VulkanContext::MultiviewCapabilities multiview_capabilies = context->get_multiview_capabilities();
			return multiview_capabilies.is_supported && multiview_capabilies.max_view_count > 1;
		} break;
		case SUPPORTS_FSR_HALF_FLOAT: {
			return context->get_shader_capabilities().shader_float16_is_supported && context->get_storage_buffer_capabilities().storage_buffer_16_bit_access_is_supported;
		} break;
		case SUPPORTS_ATTACHMENT_VRS: {
			VulkanContext::VRSCapabilities vrs_capabilities = context->get_vrs_capabilities();
			return vrs_capabilities.attachment_vrs_supported;
		} break;
		default: {
			return false;
		}
	}
}

RenderingDeviceVulkan::RenderingDeviceVulkan() {
	device_capabilities.device_family = DEVICE_VULKAN;
}

RenderingDeviceVulkan::~RenderingDeviceVulkan() {
	if (local_device.is_valid()) {
		finalize();
		context->local_device_free(local_device);
	}
}