 David Piuva
|
8821d66ac7
Updated laneCountF_32Bit into laneCountF in an outdated comment in simd.h.
|
5 months ago |
 David Piuva
|
275b7dbabd
Implemented non-immediate uniform offset bit shifts for AVX2.
|
10 months ago |
|
David Piuva
|
6f9822216d
Implemented non-immediate uniform offset bit shifts for SSE2.
|
10 months ago |
|
David Piuva
|
5019c97069
Cleaned up after debugging.
|
10 months ago |
|
David Piuva
|
b22011f774
Explicitly defining copy, assignment and move operators to supress the compiler bug.
|
10 months ago |
|
David Piuva
|
fb7fc94991
Explicitly aligning SIMD vectors and supressing as many temps as possible to go around that g++ violates strict alignment requirements in Intel's ABI.
|
10 months ago |
|
David Piuva
|
fe0d26b2e7
Splitting up the texture test to see what goes wrong.
|
10 months ago |
|
David Piuva
|
f2c9daf609
Found the alignment bug.
|
10 months ago |
|
David Piuva
|
0f9393b517
Added more safety checks to SIMD functions.
|
10 months ago |
|
David Piuva
|
321fddf72b
Getting cache line width using system APIs.
|
10 months ago |
|
David Piuva
|
f8ea6b14ce
Enabling all available SIMD extensions when testing.
|
10 months ago |
|
David Piuva
|
a0d397d196
Fixing errors from the Github test server.
|
10 months ago |
|
David Piuva
|
85666a9cd1
Made member functions in simd.h into global functions to work with templates.
|
10 months ago |
|
David Piuva
|
8f988fc8e9
Found a bug in use of AVX2 by compiling for it without hardware support.
|
10 months ago |
|
David Piuva
|
5a2eee2f5f
Replaced std::shared_ptr with dsr::Handle and introduced textures as separate types.
|
10 months ago |
|
David Piuva
|
0bbabf38c5
Created an arena allocator for getting heap memory.
|
11 months ago |
|
David Piuva
|
906b9fef5a
Replacing most C library dependencies with equivalent dependencies in C++.
|
1 year ago |
|
David Piuva
|
df69699082
Fixed typos after testing on Raspberry Pi OS, Windows 7, Manjaro, ARMv7 with NEON, Intel Core i5 with AVX2, AMD A4 with AVX...
|
2 years ago |
|
David Piuva
|
2c72251b3f
Updated comments about direct memory access in SIMD vectors after automatic renaming.
|
2 years ago |
|
David Piuva
|
4d27bde0d4
Automatically aligning float images to the F vector size to allow processing float images using only AVX.
|
2 years ago |
|
David Piuva
|
a66c48dfab
Introduced the F vector length to take advantage of AVX even when AVX2 is not available, because it might be useful for sound engines where only floats are processed.
|
2 years ago |
|
David Piuva
|
b0e5fd88b9
Implemented gradient constructors for SIMD vectors.
|
2 years ago |
|
David Piuva
|
6c53ada255
Made a separate macro for the default vector size, in case that I later come up with a safe, easy and fast way of exploiting AVX without AVX2.
|
2 years ago |
|
David Piuva
|
9440208038
Implemented 256-bit SIMD vectors for AVX2. And some general clean-up.
|
2 years ago |
|
David Piuva
|
cb3806314d
Added saturated subtraction to simd.h
|
3 years ago |
|
David Piuva
|
e119193805
Added emulation of unary negation for convenience in simd.h.
|
3 years ago |
|
David Piuva
|
305389cef6
Added exclusive or to simd.h.
|
3 years ago |
|
David Piuva
|
b0afd333ff
Passed regression tests on ARM.
|
3 years ago |
|
David Piuva
|
b31b47b098
Made gather functions for AVX2 with built-in fallback implementations.
|
4 years ago |
|
David Piuva
|
dd8e3a8b4c
Updated copyright year.
|
4 years ago |
