mirror/tracy
1
0
Fork 0
mirror of https://github.com/wolfpld/tracy.git synced 2024-11-14 12:04:35 +00:00
Code Issues Packages Projects Releases Wiki Activity

Replace two packs with one shuffle.

Browse Source
This commit is contained in:
Bartosz Taudul 2019-07-13 20:01:12 +02:00
parent fc83f97ad3
commit f65373ece7
1 changed files with 3 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
client/TracyDxt1.cpp
View File

@ -616,11 +616,10 @@ static tracy_force_inline void ProcessRGB_AVX( const uint8_t* src, char*& dst )
__m256i mmb = _mm256_srli_epi64( _mm256_and_si256( mm2, mmmb ), 43 ); __m256i mmb = _mm256_srli_epi64( _mm256_and_si256( mm2, mmmb ), 43 );
__m256i mm3 = _mm256_or_si256( mmr, mmg ); __m256i mm3 = _mm256_or_si256( mmr, mmg );
__m256i mm4 = _mm256_or_si256( mm3, mmb ); __m256i mm4 = _mm256_or_si256( mm3, mmb );
__m256i mm5 = _mm256_packus_epi32( mm4, mm4 ); __m256i mm5 = _mm256_shuffle_epi8( mm4, _mm256_set1_epi32( 0x09080100 ) );
__m256i mm6 = _mm256_packus_epi32( mm5, mm5 );
uint32_t minmax0 = _mm256_cvtsi256_si32( mm6 ); uint32_t minmax0 = _mm256_cvtsi256_si32( mm5 );
uint32_t minmax1 = _mm256_extract_epi32( mm6, 4 ); uint32_t minmax1 = _mm256_extract_epi32( mm5, 4 );
uint32_t vp0 = _mm256_cvtsi256_si32( p ); uint32_t vp0 = _mm256_cvtsi256_si32( p );
uint32_t vp1 = _mm256_extract_epi32( p, 4 ); uint32_t vp1 = _mm256_extract_epi32( p, 4 );