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

v3.22.3

Browse Source
This commit is contained in:
Jay D Dee
2023-06-14 11:07:40 -04:00
parent de564ccbde
commit 57a6b7b58b
31 changed files with 3724 additions and 3345 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
algo/shavite/shavite-hash-2way.c
View File

@@ -18,14 +18,6 @@ static const uint32_t IV512[] =
0xE275EADE, 0x502D9FCD, 0xB9357178, 0x022A4B9A
};
/*
#define mm256_ror2x256hi_1x32( a, b ) \
_mm256_blend_epi32( mm256_shuflr128_32( a ), \
mm256_shuflr128_32( b ), 0x88 )
*/
//#define mm256_ror2x256hi_1x32( a, b ) _mm256_alignr_epi8( b, a, 4 )
#if defined(__VAES__)
#define mm256_aesenc_2x128( x, k ) \
@@ -34,8 +26,9 @@ static const uint32_t IV512[] =
#else
#define mm256_aesenc_2x128( x, k ) \
mm256_concat_128( _mm_aesenc_si128( mm128_extr_hi128_256( x ), k ), \
_mm_aesenc_si128( mm128_extr_lo128_256( x ), k ) )
_mm256_inserti128_si256( _mm256_castsi128_si256( \
_mm_aesenc_si128( _mm256_castsi256_si128( x ), k ) ), \
_mm_aesenc_si128( _mm256_extracti128_si256( x, 1 ), k ), 1 )
#endif
@@ -257,10 +250,10 @@ void shavite512_2way_init( shavite512_2way_context *ctx )
__m256i *h = (__m256i*)ctx->h;
__m128i *iv = (__m128i*)IV512;
h[0] = m256_const1_128( iv[0] );
h[1] = m256_const1_128( iv[1] );
h[2] = m256_const1_128( iv[2] );
h[3] = m256_const1_128( iv[3] );
h[0] = mm256_bcast_m128( iv[0] );
h[1] = mm256_bcast_m128( iv[1] );
h[2] = mm256_bcast_m128( iv[2] );
h[3] = mm256_bcast_m128( iv[3] );
ctx->ptr = 0;
ctx->count0 = 0;
@@ -320,7 +313,7 @@ void shavite512_2way_close( shavite512_2way_context *ctx, void *dst )
uint32_t vp = ctx->ptr>>5;
// Terminating byte then zero pad
casti_m256i( buf, vp++ ) = m256_const1_i128( 0x0000000000000080 );
casti_m256i( buf, vp++ ) = mm256_bcast128lo_64( 0x0000000000000080 );
// Zero pad full vectors up to count
for ( ; vp < 6; vp++ )
@@ -334,9 +327,9 @@ void shavite512_2way_close( shavite512_2way_context *ctx, void *dst )
count.u32[2] = ctx->count2;
count.u32[3] = ctx->count3;
casti_m256i( buf, 6 ) = m256_const1_128(
casti_m256i( buf, 6 ) = mm256_bcast_m128(
_mm_insert_epi16( m128_zero, count.u16[0], 7 ) );
casti_m256i( buf, 7 ) = m256_const1_128( _mm_set_epi16(
casti_m256i( buf, 7 ) = mm256_bcast_m128( _mm_set_epi16(
0x0200, count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1] ) );
@@ -400,19 +393,19 @@ void shavite512_2way_update_close( shavite512_2way_context *ctx, void *dst,
if ( vp == 0 ) // empty buf, xevan.
{
casti_m256i( buf, 0 ) = m256_const1_i128( 0x0000000000000080 );
casti_m256i( buf, 0 ) = mm256_bcast128lo_64( 0x0000000000000080 );
memset_zero_256( (__m256i*)buf + 1, 5 );
ctx->count0 = ctx->count1 = ctx->count2 = ctx->count3 = 0;
}
else // half full buf, everyone else.
{
casti_m256i( buf, vp++ ) = m256_const1_i128( 0x0000000000000080 );
casti_m256i( buf, vp++ ) = mm256_bcast128lo_64( 0x0000000000000080 );
memset_zero_256( (__m256i*)buf + vp, 6 - vp );
}
casti_m256i( buf, 6 ) = m256_const1_128(
casti_m256i( buf, 6 ) = mm256_bcast_m128(
_mm_insert_epi16( m128_zero, count.u16[0], 7 ) );
casti_m256i( buf, 7 ) = m256_const1_128( _mm_set_epi16(
casti_m256i( buf, 7 ) = mm256_bcast_m128( _mm_set_epi16(
0x0200, count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1] ) );
@@ -430,10 +423,10 @@ void shavite512_2way_full( shavite512_2way_context *ctx, void *dst,
__m256i *h = (__m256i*)ctx->h;
__m128i *iv = (__m128i*)IV512;
h[0] = m256_const1_128( iv[0] );
h[1] = m256_const1_128( iv[1] );
h[2] = m256_const1_128( iv[2] );
h[3] = m256_const1_128( iv[3] );
h[0] = mm256_bcast_m128( iv[0] );
h[1] = mm256_bcast_m128( iv[1] );
h[2] = mm256_bcast_m128( iv[2] );
h[3] = mm256_bcast_m128( iv[3] );
ctx->ptr =
ctx->count0 =
@@ -490,19 +483,19 @@ void shavite512_2way_full( shavite512_2way_context *ctx, void *dst,
if ( vp == 0 ) // empty buf, xevan.
{
casti_m256i( buf, 0 ) = m256_const1_i128( 0x0000000000000080 );
casti_m256i( buf, 0 ) = mm256_bcast128lo_64( 0x0000000000000080 );
memset_zero_256( (__m256i*)buf + 1, 5 );
ctx->count0 = ctx->count1 = ctx->count2 = ctx->count3 = 0;
}
else // half full buf, everyone else.
{
casti_m256i( buf, vp++ ) = m256_const1_i128( 0x0000000000000080 );
casti_m256i( buf, vp++ ) = mm256_bcast128lo_64( 0x0000000000000080 );
memset_zero_256( (__m256i*)buf + vp, 6 - vp );
}
casti_m256i( buf, 6 ) = m256_const1_128(
casti_m256i( buf, 6 ) = mm256_bcast_m128(
_mm_insert_epi16( m128_zero, count.u16[0], 7 ) );
casti_m256i( buf, 7 ) = m256_const1_128( _mm_set_epi16(
casti_m256i( buf, 7 ) = mm256_bcast_m128( _mm_set_epi16(
0x0200, count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1] ) );

38
algo/shavite/shavite-hash-4way.c
View File

@@ -227,10 +227,10 @@ void shavite512_4way_init( shavite512_4way_context *ctx )
__m512i *h = (__m512i*)ctx->h;
__m128i *iv = (__m128i*)IV512;
h[0] = m512_const1_128( iv[0] );
h[1] = m512_const1_128( iv[1] );
h[2] = m512_const1_128( iv[2] );
h[3] = m512_const1_128( iv[3] );
h[0] = mm512_bcast_m128( iv[0] );
h[1] = mm512_bcast_m128( iv[1] );
h[2] = mm512_bcast_m128( iv[2] );
h[3] = mm512_bcast_m128( iv[3] );
ctx->ptr = 0;
ctx->count0 = 0;
@@ -290,7 +290,7 @@ void shavite512_4way_close( shavite512_4way_context *ctx, void *dst )
uint32_t vp = ctx->ptr>>6;
// Terminating byte then zero pad
casti_m512i( buf, vp++ ) = m512_const1_i128( 0x0000000000000080 );
casti_m512i( buf, vp++ ) = mm512_bcast128lo_64( 0x0000000000000080 );
// Zero pad full vectors up to count
for ( ; vp < 6; vp++ )
@@ -304,9 +304,9 @@ void shavite512_4way_close( shavite512_4way_context *ctx, void *dst )
count.u32[2] = ctx->count2;
count.u32[3] = ctx->count3;
casti_m512i( buf, 6 ) = m512_const1_128(
casti_m512i( buf, 6 ) = mm512_bcast_m128(
_mm_insert_epi16( m128_zero, count.u16[0], 7 ) );
casti_m512i( buf, 7 ) = m512_const1_128( _mm_set_epi16(
casti_m512i( buf, 7 ) = mm512_bcast_m128( _mm_set_epi16(
0x0200, count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1] ) );
@@ -370,19 +370,19 @@ void shavite512_4way_update_close( shavite512_4way_context *ctx, void *dst,
if ( vp == 0 ) // empty buf, xevan.
{
casti_m512i( buf, 0 ) = m512_const1_i128( 0x0000000000000080 );
casti_m512i( buf, 0 ) = mm512_bcast128lo_64( 0x0000000000000080 );
memset_zero_512( (__m512i*)buf + 1, 5 );
ctx->count0 = ctx->count1 = ctx->count2 = ctx->count3 = 0;
}
else // half full buf, everyone else.
{
casti_m512i( buf, vp++ ) = m512_const1_i128( 0x0000000000000080 );
casti_m512i( buf, vp++ ) = mm512_bcast128lo_64( 0x0000000000000080 );
memset_zero_512( (__m512i*)buf + vp, 6 - vp );
}
casti_m512i( buf, 6 ) = m512_const1_128(
casti_m512i( buf, 6 ) = mm512_bcast_m128(
_mm_insert_epi16( m128_zero, count.u16[0], 7 ) );
casti_m512i( buf, 7 ) = m512_const1_128( _mm_set_epi16(
casti_m512i( buf, 7 ) = mm512_bcast_m128( _mm_set_epi16(
0x0200, count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1] ) );
@@ -401,10 +401,10 @@ void shavite512_4way_full( shavite512_4way_context *ctx, void *dst,
__m512i *h = (__m512i*)ctx->h;
__m128i *iv = (__m128i*)IV512;
h[0] = m512_const1_128( iv[0] );
h[1] = m512_const1_128( iv[1] );
h[2] = m512_const1_128( iv[2] );
h[3] = m512_const1_128( iv[3] );
h[0] = mm512_bcast_m128( iv[0] );
h[1] = mm512_bcast_m128( iv[1] );
h[2] = mm512_bcast_m128( iv[2] );
h[3] = mm512_bcast_m128( iv[3] );
ctx->ptr =
ctx->count0 =
@@ -461,19 +461,19 @@ void shavite512_4way_full( shavite512_4way_context *ctx, void *dst,
if ( vp == 0 ) // empty buf, xevan.
{
casti_m512i( buf, 0 ) = m512_const1_i128( 0x0000000000000080 );
casti_m512i( buf, 0 ) = mm512_bcast128lo_64( 0x0000000000000080 );
memset_zero_512( (__m512i*)buf + 1, 5 );
ctx->count0 = ctx->count1 = ctx->count2 = ctx->count3 = 0;
}
else // half full buf, everyone else.
{
casti_m512i( buf, vp++ ) = m512_const1_i128( 0x0000000000000080 );
casti_m512i( buf, vp++ ) = mm512_bcast128lo_64( 0x0000000000000080 );
memset_zero_512( (__m512i*)buf + vp, 6 - vp );
}
casti_m512i( buf, 6 ) = m512_const1_128(
casti_m512i( buf, 6 ) = mm512_bcast_m128(
_mm_insert_epi16( m128_zero, count.u16[0], 7 ) );
casti_m512i( buf, 7 ) = m512_const1_128( _mm_set_epi16(
casti_m512i( buf, 7 ) = mm512_bcast_m128( _mm_set_epi16(
0x0200, count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1] ) );