Emulation/ppsspp
1
0
Fork 0
mirror of https://github.com/hrydgard/ppsspp.git synced 2025-04-02 11:01:50 -04:00
Code Issues Projects Releases Packages Wiki Activity

Use _mm_movemask_ps for AnyMask

Browse source 
Probably very minor speed improvement, but it's rather neat.
This commit is contained in:
fp64 2023-06-17 01:05:02 -04:00
parent 4a4cd3d977
commit ab85c46161
1 changed files with 2 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

7
GPU/Software/Rasterizer.cpp
View file

@ -852,11 +852,8 @@ static inline bool AnyMask(const Vec4<int> &mask) {
return AnyMaskSSE4(mask.ivec); return AnyMaskSSE4(mask.ivec);
} }
// In other words: !(mask.x < 0 && mask.y < 0 && mask.z < 0 && mask.w < 0) // Source: https://fgiesen.wordpress.com/2013/02/10/optimizing-the-basic-rasterizer/#comment-6676
__m128i low2 = _mm_and_si128(mask.ivec, _mm_shuffle_epi32(mask.ivec, _MM_SHUFFLE(3, 2, 3, 2))); return _mm_movemask_ps(_mm_castsi128_ps(mask.ivec))!=15;
__m128i low1 = _mm_and_si128(low2, _mm_shuffle_epi32(low2, _MM_SHUFFLE(1, 1, 1, 1)));
// Now we only need to check one sign bit.
return _mm_cvtsi128_si32(low1) >= 0;
#elif PPSSPP_ARCH(ARM64_NEON) #elif PPSSPP_ARCH(ARM64_NEON)
int64x2_t sig = vreinterpretq_s64_s32(vshrq_n_s32(mask.ivec, 31)); int64x2_t sig = vreinterpretq_s64_s32(vshrq_n_s32(mask.ivec, 31));
return vgetq_lane_s64(sig, 0) != -1 || vgetq_lane_s64(sig, 1) != -1; return vgetq_lane_s64(sig, 0) != -1 || vgetq_lane_s64(sig, 1) != -1;