popov/cpuminer-opt-gpu
1
0
Fork 0
mirror of https://github.com/JayDDee/cpuminer-opt.git synced 2025-09-17 23:44:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

v3.9.9.1

Browse Source
This commit is contained in:
Jay D Dee
2019-10-24 14:11:26 -04:00
parent 72330eb5a7
commit 86b889e1b0
65 changed files with 547 additions and 638 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
simd-utils/simd-256.h
View File

@@ -367,38 +367,49 @@ static inline void memcpy_256( __m256i *dst, const __m256i *src, const int n )
//
// AVX512 has bit rotate for 256 bit vectors with 64 or 32 bit elements
// compiler doesn't like when a variable is used for the last arg of
// _mm_rol_epi32, must be "8 bit immediate".
// _mm_rol_epi32, must be "8 bit immediate". Therefore use rol_var where
// necessary.
#define mm256_ror_var_64( v, c ) \
_mm256_or_si256( _mm256_srli_epi64( v, c ), \
_mm256_slli_epi64( v, 64-(c) ) )
#define mm256_rol_var_64( v, c ) \
_mm256_or_si256( _mm256_slli_epi64( v, c ), \
_mm256_srli_epi64( v, 64-(c) ) )
#define mm256_ror_var_32( v, c ) \
_mm256_or_si256( _mm256_srli_epi32( v, c ), \
_mm256_slli_epi32( v, 32-(c) ) )
#define mm256_rol_var_32( v, c ) \
_mm256_or_si256( _mm256_slli_epi32( v, c ), \
_mm256_srli_epi32( v, 32-(c) ) )
/*
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
#define mm256_ror_64( v, c ) _mm256_ror_epi64( v, c )
#define mm256_rol_64( v, c ) _mm256_rol_epi64( v, c )
#define mm256_ror_32( v, c ) _mm256_ror_epi32( v, c )
#define mm256_rol_32( v, c ) _mm256_rol_epi32( v, c )
// AVX512, control must be 8 bit immediate.
#define mm256_ror_64 _mm256_ror_epi64
#define mm256_rol_64 _mm256_rol_epi64
#define mm256_ror_32 _mm256_ror_epi32
#define mm256_rol_32 _mm256_rol_epi32
#else
*/
#define mm256_ror_64( v, c ) \
_mm256_or_si256( _mm256_srli_epi64( v, c ), \
_mm256_slli_epi64( v, 64-(c) ) )
// No AVX512, use fallback.
#define mm256_rol_64( v, c ) \
_mm256_or_si256( _mm256_slli_epi64( v, c ), \
_mm256_srli_epi64( v, 64-(c) ) )
#define mm256_ror_32( v, c ) \
_mm256_or_si256( _mm256_srli_epi32( v, c ), \
_mm256_slli_epi32( v, 32-(c) ) )
#define mm256_rol_32( v, c ) \
_mm256_or_si256( _mm256_slli_epi32( v, c ), \
_mm256_srli_epi32( v, 32-(c) ) )
#define mm256_ror_64 mm256_ror_var_64
#define mm256_rol_64 mm256_rol_var_64
#define mm256_ror_32 mm256_ror_var_32
#define mm256_rol_32 mm256_rol_var_32
// #endif // AVX512 else
#define mm256_ror_16( v, c ) \
_mm256_or_si256( _mm256_srli_epi16( v, c ), \
_mm256_slli_epi16( v, 16-(c) ) )