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

v3.9.6

Browse Source
This commit is contained in:
Jay D Dee
2019-07-17 17:54:38 -04:00
parent e2d5762ef2
commit 6f49ba09b7
34 changed files with 1930 additions and 382 deletions
Download Patch File Download Diff File Expand all files Collapse all files

83
algo/cubehash/cube-hash-2way.c
View File

@@ -7,6 +7,7 @@
// 2x128
/*
// The result of hashing 10 rounds of initial data which consists of params
// zero padded.
static const uint64_t IV256[] =
@@ -24,13 +25,14 @@ static const uint64_t IV512[] =
0x148FE485FCD398D9, 0xB64445321B017BEF, 0x2FF5781C6A536159, 0x0DBADEA991FA7934,
0xA5A70E75D65C8A2B, 0xBC796576B1C62456, 0xE7989AF11921C8F7, 0xD43E3B447795D246
};
*/
static void transform_2way( cube_2way_context *sp )
{
int r;
const int rounds = sp->rounds;
__m256i x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3;
__m256i x0, x1, x2, x3, x4, x5, x6, x7, y0, y1;
x0 = _mm256_load_si256( (__m256i*)sp->h );
x1 = _mm256_load_si256( (__m256i*)sp->h + 1 );
@@ -47,18 +49,12 @@ static void transform_2way( cube_2way_context *sp )
x5 = _mm256_add_epi32( x1, x5 );
x6 = _mm256_add_epi32( x2, x6 );
x7 = _mm256_add_epi32( x3, x7 );
y0 = x2;
y1 = x3;
y2 = x0;
y3 = x1;
x0 = _mm256_xor_si256( _mm256_slli_epi32( y0, 7 ),
_mm256_srli_epi32( y0, 25 ) );
x1 = _mm256_xor_si256( _mm256_slli_epi32( y1, 7 ),
_mm256_srli_epi32( y1, 25 ) );
x2 = _mm256_xor_si256( _mm256_slli_epi32( y2, 7 ),
_mm256_srli_epi32( y2, 25 ) );
x3 = _mm256_xor_si256( _mm256_slli_epi32( y3, 7 ),
_mm256_srli_epi32( y3, 25 ) );
y0 = x0;
y1 = x1;
x0 = mm256_rol_32( x2, 7 );
x1 = mm256_rol_32( x3, 7 );
x2 = mm256_rol_32( y0, 7 );
x3 = mm256_rol_32( y1, 7 );
x0 = _mm256_xor_si256( x0, x4 );
x1 = _mm256_xor_si256( x1, x5 );
x2 = _mm256_xor_si256( x2, x6 );
@@ -71,18 +67,12 @@ static void transform_2way( cube_2way_context *sp )
x5 = _mm256_add_epi32( x1, x5 );
x6 = _mm256_add_epi32( x2, x6 );
x7 = _mm256_add_epi32( x3, x7 );
y0 = x1;
y1 = x0;
y2 = x3;
y3 = x2;
x0 = _mm256_xor_si256( _mm256_slli_epi32( y0, 11 ),
_mm256_srli_epi32( y0, 21 ) );
x1 = _mm256_xor_si256( _mm256_slli_epi32( y1, 11 ),
_mm256_srli_epi32( y1, 21 ) );
x2 = _mm256_xor_si256( _mm256_slli_epi32( y2, 11 ),
_mm256_srli_epi32( y2, 21 ) );
x3 = _mm256_xor_si256( _mm256_slli_epi32( y3, 11 ),
_mm256_srli_epi32( y3, 21 ) );
y0 = x0;
y1 = x2;
x0 = mm256_rol_32( x1, 11 );
x1 = mm256_rol_32( y0, 11 );
x2 = mm256_rol_32( x3, 11 );
x3 = mm256_rol_32( y1, 11 );
x0 = _mm256_xor_si256( x0, x4 );
x1 = _mm256_xor_si256( x1, x5 );
x2 = _mm256_xor_si256( x2, x6 );
@@ -107,23 +97,40 @@ static void transform_2way( cube_2way_context *sp )
int cube_2way_init( cube_2way_context *sp, int hashbitlen, int rounds,
int blockbytes )
{
const uint64_t* iv = hashbitlen == 512 ? IV512 : IV256;
__m128i* h = (__m128i*)sp->h;
sp->hashlen = hashbitlen/128;
sp->blocksize = blockbytes/16;
sp->rounds = rounds;
sp->pos = 0;
__m256i* h = (__m256i*)sp->h;
h[0] = _mm256_set_epi64x( iv[ 1], iv[ 0], iv[ 1], iv[ 0] );
h[1] = _mm256_set_epi64x( iv[ 3], iv[ 2], iv[ 3], iv[ 2] );
h[2] = _mm256_set_epi64x( iv[ 5], iv[ 4], iv[ 5], iv[ 4] );
h[3] = _mm256_set_epi64x( iv[ 7], iv[ 6], iv[ 7], iv[ 6] );
h[4] = _mm256_set_epi64x( iv[ 9], iv[ 8], iv[ 9], iv[ 8] );
h[5] = _mm256_set_epi64x( iv[11], iv[10], iv[11], iv[10] );
h[6] = _mm256_set_epi64x( iv[13], iv[12], iv[13], iv[12] );
h[7] = _mm256_set_epi64x( iv[15], iv[14], iv[15], iv[14] );
if ( hashbitlen == 512 )
{
h[ 0] = m128_const_64( 0x4167D83E2D538B8B, 0x50F494D42AEA2A61 );
h[ 2] = m128_const_64( 0x50AC5695CC39968E, 0xC701CF8C3FEE2313 );
h[ 4] = m128_const_64( 0x825B453797CF0BEF, 0xA647A8B34D42C787 );
h[ 6] = m128_const_64( 0xA23911AED0E5CD33, 0xF22090C4EEF864D2 );
h[ 8] = m128_const_64( 0xB64445321B017BEF, 0x148FE485FCD398D9 );
h[10] = m128_const_64( 0x0DBADEA991FA7934, 0x2FF5781C6A536159 );
h[12] = m128_const_64( 0xBC796576B1C62456, 0xA5A70E75D65C8A2B );
h[14] = m128_const_64( 0xD43E3B447795D246, 0xE7989AF11921C8F7 );
h[1] = h[ 0]; h[ 3] = h[ 2]; h[ 5] = h[ 4]; h[ 7] = h[ 6];
h[9] = h[ 8]; h[11] = h[10]; h[13] = h[12]; h[15] = h[14];
}
else
{
h[ 0] = m128_const_64( 0x35481EAE63117E71, 0xCCD6F29FEA2BD4B4 );
h[ 2] = m128_const_64( 0xF4CC12BE7E624131, 0xE5D94E6322512D5B );
h[ 4] = m128_const_64( 0x3361DA8CD0720C35, 0x42AF2070C2D0B696 );
h[ 6] = m128_const_64( 0x40E5FBAB4680AC00, 0x8EF8AD8328CCECA4 );
h[ 8] = m128_const_64( 0xF0B266796C859D41, 0x6107FBD5D89041C3 );
h[10] = m128_const_64( 0x93CB628565C892FD, 0x5FA2560309392549 );
h[12] = m128_const_64( 0x85254725774ABFDD, 0x9E4B4E602AF2B5AE );
h[14] = m128_const_64( 0xD6032C0A9CDAF8AF, 0x4AB6AAD615815AEB );
h[1] = h[ 0]; h[ 3] = h[ 2]; h[ 5] = h[ 4]; h[ 7] = h[ 6];
h[9] = h[ 8]; h[11] = h[10]; h[13] = h[12]; h[15] = h[14];
}
return 0;
}
@@ -165,7 +172,7 @@ int cube_2way_close( cube_2way_context *sp, void *output )
for ( i = 0; i < 10; ++i ) transform_2way( sp );
for ( i = 0; i < sp->hashlen; i++ ) hash[i] = sp->h[i];
memcpy( hash, sp->h, sp->hashlen<<5 );
return 0;
}
@@ -198,7 +205,7 @@ int cube_2way_update_close( cube_2way_context *sp, void *output,
for ( i = 0; i < 10; ++i ) transform_2way( sp );
for ( i = 0; i < sp->hashlen; i++ ) hash[i] = sp->h[i];
memcpy( hash, sp->h, sp->hashlen<<5 );
return 0;
}

107
algo/cubehash/cubehash_sse2.c
View File

@@ -16,24 +16,6 @@
#include "simd-utils.h"
#include <stdio.h>
// The result of hashing 10 rounds of initial data which is params and
// mostly zeros.
static const uint64_t IV256[] =
{
0xCCD6F29FEA2BD4B4, 0x35481EAE63117E71, 0xE5D94E6322512D5B, 0xF4CC12BE7E624131,
0x42AF2070C2D0B696, 0x3361DA8CD0720C35, 0x8EF8AD8328CCECA4, 0x40E5FBAB4680AC00,
0x6107FBD5D89041C3, 0xF0B266796C859D41, 0x5FA2560309392549, 0x93CB628565C892FD,
0x9E4B4E602AF2B5AE, 0x85254725774ABFDD, 0x4AB6AAD615815AEB, 0xD6032C0A9CDAF8AF
};
static const uint64_t IV512[] =
{
0x50F494D42AEA2A61, 0x4167D83E2D538B8B, 0xC701CF8C3FEE2313, 0x50AC5695CC39968E,
0xA647A8B34D42C787, 0x825B453797CF0BEF, 0xF22090C4EEF864D2, 0xA23911AED0E5CD33,
0x148FE485FCD398D9, 0xB64445321B017BEF, 0x2FF5781C6A536159, 0x0DBADEA991FA7934,
0xA5A70E75D65C8A2B, 0xBC796576B1C62456, 0xE7989AF11921C8F7, 0xD43E3B447795D246
};
static void transform( cubehashParam *sp )
{
int r;
@@ -53,26 +35,22 @@ static void transform( cubehashParam *sp )
x2 = _mm256_add_epi32( x0, x2 );
x3 = _mm256_add_epi32( x1, x3 );
y0 = x0;
x0 = _mm256_xor_si256( _mm256_slli_epi32( x1, 7 ),
_mm256_srli_epi32( x1, 25 ) );
x1 = _mm256_xor_si256( _mm256_slli_epi32( y0, 7 ),
_mm256_srli_epi32( y0, 25 ) );
x0 = mm256_rol_32( x1, 7 );
x1 = mm256_rol_32( y0, 7 );
x0 = _mm256_xor_si256( x0, x2 );
x1 = _mm256_xor_si256( x1, x3 );
x2 = _mm256_shuffle_epi32( x2, 0x4e );
x3 = _mm256_shuffle_epi32( x3, 0x4e );
x2 = mm256_swap64_128( x2 );
x3 = mm256_swap64_128( x3 );
x2 = _mm256_add_epi32( x0, x2 );
x3 = _mm256_add_epi32( x1, x3 );
y0 = _mm256_permute4x64_epi64( x0, 0x4e );
y1 = _mm256_permute4x64_epi64( x1, 0x4e );
x0 = _mm256_xor_si256( _mm256_slli_epi32( y0, 11 ),
_mm256_srli_epi32( y0, 21 ) );
x1 = _mm256_xor_si256( _mm256_slli_epi32( y1, 11 ),
_mm256_srli_epi32( y1, 21 ) );
y0 = mm256_swap_128( x0 );
y1 = mm256_swap_128( x1 );
x0 = mm256_rol_32( y0, 11 );
x1 = mm256_rol_32( y1, 11 );
x0 = _mm256_xor_si256( x0, x2 );
x1 = _mm256_xor_si256( x1, x3 );
x2 = _mm256_shuffle_epi32( x2, 0xb1 );
x3 = _mm256_shuffle_epi32( x3, 0xb1 );
x2 = mm256_swap32_64( x2 );
x3 = mm256_swap32_64( x3 );
}
_mm256_store_si256( (__m256i*)sp->x, x0 );
@@ -147,37 +125,58 @@ static void transform( cubehashParam *sp )
#endif
} // transform
/*
// The result of hashing 10 rounds of initial data which is params and
// mostly zeros.
static const uint64_t IV256[] =
{
0xCCD6F29FEA2BD4B4, 0x35481EAE63117E71, 0xE5D94E6322512D5B, 0xF4CC12BE7E624131,
0x42AF2070C2D0B696, 0x3361DA8CD0720C35, 0x8EF8AD8328CCECA4, 0x40E5FBAB4680AC00,
0x6107FBD5D89041C3, 0xF0B266796C859D41, 0x5FA2560309392549, 0x93CB628565C892FD,
0x9E4B4E602AF2B5AE, 0x85254725774ABFDD, 0x4AB6AAD615815AEB, 0xD6032C0A9CDAF8AF
};
static const uint64_t IV512[] =
{
0x50F494D42AEA2A61, 0x4167D83E2D538B8B, 0xC701CF8C3FEE2313, 0x50AC5695CC39968E,
0xA647A8B34D42C787, 0x825B453797CF0BEF, 0xF22090C4EEF864D2, 0xA23911AED0E5CD33,
0x148FE485FCD398D9, 0xB64445321B017BEF, 0x2FF5781C6A536159, 0x0DBADEA991FA7934,
0xA5A70E75D65C8A2B, 0xBC796576B1C62456, 0xE7989AF11921C8F7, 0xD43E3B447795D246
};
*/
int cubehashInit(cubehashParam *sp, int hashbitlen, int rounds, int blockbytes)
{
const uint64_t* iv = hashbitlen == 512 ? IV512 : IV256;
__m128i *x = (__m128i*)sp->x;
sp->hashlen = hashbitlen/128;
sp->blocksize = blockbytes/16;
sp->rounds = rounds;
sp->pos = 0;
#if defined(__AVX2__)
__m256i* x = (__m256i*)sp->x;
if ( hashbitlen == 512 )
{
x[0] = _mm256_set_epi64x( iv[ 3], iv[ 2], iv[ 1], iv[ 0] );
x[1] = _mm256_set_epi64x( iv[ 7], iv[ 6], iv[ 5], iv[ 4] );
x[2] = _mm256_set_epi64x( iv[11], iv[10], iv[ 9], iv[ 8] );
x[3] = _mm256_set_epi64x( iv[15], iv[14], iv[13], iv[12] );
x[0] = m128_const_64( 0x4167D83E2D538B8B, 0x50F494D42AEA2A61 );
x[1] = m128_const_64( 0x50AC5695CC39968E, 0xC701CF8C3FEE2313 );
x[2] = m128_const_64( 0x825B453797CF0BEF, 0xA647A8B34D42C787 );
x[3] = m128_const_64( 0xA23911AED0E5CD33, 0xF22090C4EEF864D2 );
x[4] = m128_const_64( 0xB64445321B017BEF, 0x148FE485FCD398D9 );
x[5] = m128_const_64( 0x0DBADEA991FA7934, 0x2FF5781C6A536159 );
x[6] = m128_const_64( 0xBC796576B1C62456, 0xA5A70E75D65C8A2B );
x[7] = m128_const_64( 0xD43E3B447795D246, 0xE7989AF11921C8F7 );
}
else
{
x[0] = m128_const_64( 0x35481EAE63117E71, 0xCCD6F29FEA2BD4B4 );
x[1] = m128_const_64( 0xF4CC12BE7E624131, 0xE5D94E6322512D5B );
x[2] = m128_const_64( 0x3361DA8CD0720C35, 0x42AF2070C2D0B696 );
x[3] = m128_const_64( 0x40E5FBAB4680AC00, 0x8EF8AD8328CCECA4 );
x[4] = m128_const_64( 0xF0B266796C859D41, 0x6107FBD5D89041C3 );
x[5] = m128_const_64( 0x93CB628565C892FD, 0x5FA2560309392549 );
x[6] = m128_const_64( 0x85254725774ABFDD, 0x9E4B4E602AF2B5AE );
x[7] = m128_const_64( 0xD6032C0A9CDAF8AF, 0x4AB6AAD615815AEB );
}
#else
__m128i* x = (__m128i*)sp->x;
x[0] = _mm_set_epi64x( iv[ 1], iv[ 0] );
x[1] = _mm_set_epi64x( iv[ 3], iv[ 2] );
x[2] = _mm_set_epi64x( iv[ 5], iv[ 4] );
x[3] = _mm_set_epi64x( iv[ 7], iv[ 6] );
x[4] = _mm_set_epi64x( iv[ 9], iv[ 8] );
x[5] = _mm_set_epi64x( iv[11], iv[10] );
x[6] = _mm_set_epi64x( iv[13], iv[12] );
x[7] = _mm_set_epi64x( iv[15], iv[14] );
#endif
return SUCCESS;
}