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

v3.23.0

Browse Source
This commit is contained in:
Jay D Dee
2023-08-30 20:15:48 -04:00
parent 57a6b7b58b
commit 4378d2f841
72 changed files with 10184 additions and 2182 deletions
Download Patch File Download Diff File Expand all files Collapse all files

157
simd-utils/intrlv.h
View File

@@ -731,6 +731,67 @@ static inline void extr_lane_8x32( void *d, const void *s,
#if defined(__AVX2__)
#if defined(__AVX512VL__) && defined(__AVX512VBMI__)
//TODO Enable for AVX10_256 AVX10_512
// Combine byte swap & broadcast in one permute
static inline void mm256_bswap32_intrlv80_8x32( void *d, const void *src )
{
const __m256i c0 = _mm256_set1_epi32( 0x00010203 );
const __m256i c1 = _mm256_set1_epi32( 0x04050607 );
const __m256i c2 = _mm256_set1_epi32( 0x08090a0b );
const __m256i c3 = _mm256_set1_epi32( 0x0c0d0e0f );
const __m128i s0 = casti_m128i( src,0 );
const __m128i s1 = casti_m128i( src,1 );
const __m128i s2 = casti_m128i( src,2 );
const __m128i s3 = casti_m128i( src,3 );
const __m128i s4 = casti_m128i( src,4 );
casti_m256i( d, 0 ) = _mm256_permutexvar_epi8( c0,
_mm256_castsi128_si256( s0 ) );
casti_m256i( d, 1 ) = _mm256_permutexvar_epi8( c1,
_mm256_castsi128_si256( s0 ) );
casti_m256i( d, 2 ) = _mm256_permutexvar_epi8( c2,
_mm256_castsi128_si256( s0 ) );
casti_m256i( d, 3 ) = _mm256_permutexvar_epi8( c3,
_mm256_castsi128_si256( s0 ) );
casti_m256i( d, 4 ) = _mm256_permutexvar_epi8( c0,
_mm256_castsi128_si256( s1 ) );
casti_m256i( d, 5 ) = _mm256_permutexvar_epi8( c1,
_mm256_castsi128_si256( s1 ) );
casti_m256i( d, 6 ) = _mm256_permutexvar_epi8( c2,
_mm256_castsi128_si256( s1 ) );
casti_m256i( d, 7 ) = _mm256_permutexvar_epi8( c3,
_mm256_castsi128_si256( s1 ) );
casti_m256i( d, 8 ) = _mm256_permutexvar_epi8( c0,
_mm256_castsi128_si256( s2 ) );
casti_m256i( d, 9 ) = _mm256_permutexvar_epi8( c1,
_mm256_castsi128_si256( s2 ) );
casti_m256i( d,10 ) = _mm256_permutexvar_epi8( c2,
_mm256_castsi128_si256( s2 ) );
casti_m256i( d,11 ) = _mm256_permutexvar_epi8( c3,
_mm256_castsi128_si256( s2 ) );
casti_m256i( d,12 ) = _mm256_permutexvar_epi8( c0,
_mm256_castsi128_si256( s3 ) );
casti_m256i( d,13 ) = _mm256_permutexvar_epi8( c1,
_mm256_castsi128_si256( s3 ) );
casti_m256i( d,14 ) = _mm256_permutexvar_epi8( c2,
_mm256_castsi128_si256( s3 ) );
casti_m256i( d,15 ) = _mm256_permutexvar_epi8( c3,
_mm256_castsi128_si256( s3 ) );
casti_m256i( d,16 ) = _mm256_permutexvar_epi8( c0,
_mm256_castsi128_si256( s4 ) );
casti_m256i( d,17 ) = _mm256_permutexvar_epi8( c1,
_mm256_castsi128_si256( s4 ) );
casti_m256i( d,18 ) = _mm256_permutexvar_epi8( c2,
_mm256_castsi128_si256( s4 ) );
casti_m256i( d,19 ) = _mm256_permutexvar_epi8( c3,
_mm256_castsi128_si256( s4 ) );
}
#else
static inline void mm256_bswap32_intrlv80_8x32( void *d, const void *src )
{
const __m128i bswap_shuf = _mm_set_epi64x( 0x0c0d0e0f08090a0b,
@@ -792,6 +853,7 @@ static inline void mm256_bswap32_intrlv80_8x32( void *d, const void *src )
_mm256_castsi128_si256( s4 ), c3 );
}
#endif // AVX512VBMI else
#endif // AVX2
// 16x32
@@ -1173,10 +1235,12 @@ static inline void extr_lane_16x32( void *d, const void *s,
((uint32_t*)d)[15] = ((const uint32_t*)s)[ lane+240 ];
}
#if defined(__AVX512F__) && defined(__AVX512VL__)
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
#if defined(__AVX512VBMI__)
// TODO Enable for AVX10_512
// Combine byte swap & broadcast in one permute
static inline void mm512_bswap32_intrlv80_16x32( void *d, const void *src )
{
@@ -1496,10 +1560,48 @@ static inline void mm256_intrlv80_4x64( void *d, const void *src )
_mm256_castsi128_si256( s4 ), 0x55 );
}
#if defined(__AVX512VL__) && defined(__AVX512VBMI__)
//TODO Enable for AVX10_256 AVX10_512
static inline void mm256_bswap32_intrlv80_4x64( void *d, const void *src )
{
const __m256i c0 = _mm256_set1_epi64x( 0x0405060700010203 );
const __m256i c1 = _mm256_set1_epi64x( 0x0c0d0e0f08090a0b );
const __m128i s0 = casti_m128i( src,0 );
const __m128i s1 = casti_m128i( src,1 );
const __m128i s2 = casti_m128i( src,2 );
const __m128i s3 = casti_m128i( src,3 );
const __m128i s4 = casti_m128i( src,4 );
casti_m256i( d,0 ) = _mm256_permutexvar_epi8( c0,
_mm256_castsi128_si256( s0 ) );
casti_m256i( d,1 ) = _mm256_permutexvar_epi8( c1,
_mm256_castsi128_si256( s0 ) );
casti_m256i( d,2 ) = _mm256_permutexvar_epi8( c0,
_mm256_castsi128_si256( s1 ) );
casti_m256i( d,3 ) = _mm256_permutexvar_epi8( c1,
_mm256_castsi128_si256( s1 ) );
casti_m256i( d,4 ) = _mm256_permutexvar_epi8( c0,
_mm256_castsi128_si256( s2 ) );
casti_m256i( d,5 ) = _mm256_permutexvar_epi8( c1,
_mm256_castsi128_si256( s2 ) );
casti_m256i( d,6 ) = _mm256_permutexvar_epi8( c0,
_mm256_castsi128_si256( s3 ) );
casti_m256i( d,7 ) = _mm256_permutexvar_epi8( c1,
_mm256_castsi128_si256( s3 ) );
casti_m256i( d,8 ) = _mm256_permutexvar_epi8( c0,
_mm256_castsi128_si256( s4 ) );
casti_m256i( d,9 ) = _mm256_permutexvar_epi8( c1,
_mm256_castsi128_si256( s4 ) );
}
#else
static inline void mm256_bswap32_intrlv80_4x64( void *d, const void *src )
{
const __m256i bswap_shuf = mm256_bcast_m128(
_mm_set_epi64x( 0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
_mm_set_epi64x( 0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
__m256i s0 = casti_m256i( src,0 );
__m256i s1 = casti_m256i( src,1 );
__m128i s4 = casti_m128i( src,4 );
@@ -1524,6 +1626,8 @@ static inline void mm256_bswap32_intrlv80_4x64( void *d, const void *src )
_mm256_castsi128_si256( s4 ), 0x55 );
}
#endif
#endif // AVX2
// 8x64 (AVX512)
@@ -1846,6 +1950,8 @@ static inline void extr_lane_8x64( void *dst, const void *src, const int lane,
#if defined(__AVX512F__) && defined(__AVX512VL__)
//TODO Enable for AVX10_512
// broadcast to all lanes
static inline void mm512_intrlv80_8x64( void *dst, const void *src )
{
@@ -2089,10 +2195,36 @@ static inline void dintrlv_4x128_512( void *dst0, void *dst1, void *dst2,
d0[3] = s[12]; d1[3] = s[13]; d2[3] = s[14]; d3[3] = s[15];
}
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
static inline void mm512_bswap32_intrlv80_4x128( void *d, void *src )
#if defined(__AVX512VBMI__)
//TODO Enable for AVX10_512
static inline void mm512_bswap32_intrlv80_4x128( void *d, const void *src )
{
const __m512i bswap_shuf = mm512_bcast_m128(
_mm_set_epi64x( 0x0c0d0e0f08090a0b, 0x0405060700010203 ) );
const __m128i s0 = casti_m128i( src,0 );
const __m128i s1 = casti_m128i( src,1 );
const __m128i s2 = casti_m128i( src,2 );
const __m128i s3 = casti_m128i( src,3 );
const __m128i s4 = casti_m128i( src,4 );
casti_m512i( d,0 ) = _mm512_permutexvar_epi8( _mm512_castsi128_si512( s0 ),
bswap_shuf );
casti_m512i( d,1 ) = _mm512_permutexvar_epi8( _mm512_castsi128_si512( s1 ),
bswap_shuf );
casti_m512i( d,2 ) = _mm512_permutexvar_epi8( _mm512_castsi128_si512( s2 ),
bswap_shuf );
casti_m512i( d,3 ) = _mm512_permutexvar_epi8( _mm512_castsi128_si512( s3 ),
bswap_shuf );
casti_m512i( d,4 ) = _mm512_permutexvar_epi8( _mm512_castsi128_si512( s4 ),
bswap_shuf );
}
#else
static inline void mm512_bswap32_intrlv80_4x128( void *d, const void *src )
{
const __m128i bswap_shuf = _mm_set_epi64x( 0x0c0d0e0f08090a0b,
0x0405060700010203 );
@@ -2108,14 +2240,15 @@ static inline void mm512_bswap32_intrlv80_4x128( void *d, void *src )
s3 = _mm_shuffle_epi8( s3, bswap_shuf );
s4 = _mm_shuffle_epi8( s4, bswap_shuf );
casti_m512i( d, 0 ) = mm512_bcast_m128( s0 );
casti_m512i( d, 1 ) = mm512_bcast_m128( s1 );
casti_m512i( d, 2 ) = mm512_bcast_m128( s2 );
casti_m512i( d, 3 ) = mm512_bcast_m128( s3 );
casti_m512i( d, 4 ) = mm512_bcast_m128( s4 );
}
casti_m512i( d,0 ) = mm512_bcast_m128( s0 );
casti_m512i( d,1 ) = mm512_bcast_m128( s1 );
casti_m512i( d,2 ) = mm512_bcast_m128( s2 );
casti_m512i( d,3 ) = mm512_bcast_m128( s3 );
casti_m512i( d,4 ) = mm512_bcast_m128( s4 );
}
#endif
#endif // AVX512VBMI ELSE
#endif // AVX512
// 2x256 (AVX512)
@@ -2955,6 +3088,8 @@ do { \
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
//TODO Enable for AVX10_512
/*
#define mm512_intrlv_blend_128( hi, lo ) \
_mm512_mask_blend_epi32( 0x0f0f, hi, lo )