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

v3.19.6

Browse Source
This commit is contained in:
Jay D Dee
2022-02-21 23:14:24 -05:00
parent 90137b391e
commit 5b678d2481
16 changed files with 1142 additions and 442 deletions
Download Patch File Download Diff File Expand all files Collapse all files

37
algo/shavite/shavite-hash-2way.c
View File

@@ -18,10 +18,13 @@ 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__)
@@ -127,24 +130,24 @@ c512_2way( shavite512_2way_context *ctx, const void *msg )
// round 2, 6, 10
k00 = _mm256_xor_si256( k00, mm256_ror2x256hi_1x32( k12, k13 ) );
k00 = _mm256_xor_si256( k00, _mm256_alignr_epi8( k13, k12, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p3, k00 ), zero );
k01 = _mm256_xor_si256( k01, mm256_ror2x256hi_1x32( k13, k00 ) );
k01 = _mm256_xor_si256( k01, _mm256_alignr_epi8( k00, k13, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ), zero );
k02 = _mm256_xor_si256( k02, mm256_ror2x256hi_1x32( k00, k01 ) );
k02 = _mm256_xor_si256( k02, _mm256_alignr_epi8( k01, k00, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ), zero );
k03 = _mm256_xor_si256( k03, mm256_ror2x256hi_1x32( k01, k02 ) );
k03 = _mm256_xor_si256( k03, _mm256_alignr_epi8( k02, k01, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ), zero );
p2 = _mm256_xor_si256( p2, x );
k10 = _mm256_xor_si256( k10, mm256_ror2x256hi_1x32( k02, k03 ) );
k10 = _mm256_xor_si256( k10, _mm256_alignr_epi8( k03, k02, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p1, k10 ), zero );
k11 = _mm256_xor_si256( k11, mm256_ror2x256hi_1x32( k03, k10 ) );
k11 = _mm256_xor_si256( k11, _mm256_alignr_epi8( k10, k03, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ), zero );
k12 = _mm256_xor_si256( k12, mm256_ror2x256hi_1x32( k10, k11 ) );
k12 = _mm256_xor_si256( k12, _mm256_alignr_epi8( k11, k10, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ), zero );
k13 = _mm256_xor_si256( k13, mm256_ror2x256hi_1x32( k11, k12 ) );
k13 = _mm256_xor_si256( k13, _mm256_alignr_epi8( k12, k11, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ), zero );
p0 = _mm256_xor_si256( p0, x );
@@ -183,24 +186,24 @@ c512_2way( shavite512_2way_context *ctx, const void *msg )
// round 4, 8, 12
k00 = _mm256_xor_si256( k00, mm256_ror2x256hi_1x32( k12, k13 ) );
k00 = _mm256_xor_si256( k00, _mm256_alignr_epi8( k13, k12, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p1, k00 ), zero );
k01 = _mm256_xor_si256( k01, mm256_ror2x256hi_1x32( k13, k00 ) );
k01 = _mm256_xor_si256( k01, _mm256_alignr_epi8( k00, k13, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ), zero );
k02 = _mm256_xor_si256( k02, mm256_ror2x256hi_1x32( k00, k01 ) );
k02 = _mm256_xor_si256( k02, _mm256_alignr_epi8( k01, k00, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ), zero );
k03 = _mm256_xor_si256( k03, mm256_ror2x256hi_1x32( k01, k02 ) );
k03 = _mm256_xor_si256( k03, _mm256_alignr_epi8( k02, k01, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ), zero );
p0 = _mm256_xor_si256( p0, x );
k10 = _mm256_xor_si256( k10, mm256_ror2x256hi_1x32( k02, k03 ) );
k10 = _mm256_xor_si256( k10, _mm256_alignr_epi8( k03, k02, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p3, k10 ), zero );
k11 = _mm256_xor_si256( k11, mm256_ror2x256hi_1x32( k03, k10 ) );
k11 = _mm256_xor_si256( k11, _mm256_alignr_epi8( k10, k03, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ), zero );
k12 = _mm256_xor_si256( k12, mm256_ror2x256hi_1x32( k10, k11 ) );
k12 = _mm256_xor_si256( k12, _mm256_alignr_epi8( k11, k10, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ), zero );
k13 = _mm256_xor_si256( k13, mm256_ror2x256hi_1x32( k11, k12 ) );
k13 = _mm256_xor_si256( k13, _mm256_alignr_epi8( k12, k11, 4 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ), zero );
p2 = _mm256_xor_si256( p2, x );

36
algo/shavite/shavite-hash-4way.c
View File

@@ -11,10 +11,6 @@ static const uint32_t IV512[] =
0xE275EADE, 0x502D9FCD, 0xB9357178, 0x022A4B9A
};
#define mm512_ror2x512hi_1x32( a, b ) \
_mm512_mask_blend_epi32( 0x8888, mm512_shuflr128_32( a ), \
mm512_shuflr128_32( b ) )
static void
c512_4way( shavite512_4way_context *ctx, const void *msg )
{
@@ -106,24 +102,24 @@ c512_4way( shavite512_4way_context *ctx, const void *msg )
// round 2, 6, 10
K0 = _mm512_xor_si512( K0, mm512_ror2x512hi_1x32( K6, K7 ) );
K0 = _mm512_xor_si512( K0, _mm512_alignr_epi8( K7, K6, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( P3, K0 ), m512_zero );
K1 = _mm512_xor_si512( K1, mm512_ror2x512hi_1x32( K7, K0 ) );
K1 = _mm512_xor_si512( K1, _mm512_alignr_epi8( K0, K7, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( X, K1 ), m512_zero );
K2 = _mm512_xor_si512( K2, mm512_ror2x512hi_1x32( K0, K1 ) );
K2 = _mm512_xor_si512( K2, _mm512_alignr_epi8( K1, K0, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( X, K2 ), m512_zero );
K3 = _mm512_xor_si512( K3, mm512_ror2x512hi_1x32( K1, K2 ) );
K3 = _mm512_xor_si512( K3, _mm512_alignr_epi8( K2, K1, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( X, K3 ), m512_zero );
P2 = _mm512_xor_si512( P2, X );
K4 = _mm512_xor_si512( K4, mm512_ror2x512hi_1x32( K2, K3 ) );
K4 = _mm512_xor_si512( K4, _mm512_alignr_epi8( K3, K2, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( P1, K4 ), m512_zero );
K5 = _mm512_xor_si512( K5, mm512_ror2x512hi_1x32( K3, K4 ) );
K5 = _mm512_xor_si512( K5, _mm512_alignr_epi8( K4, K3, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( X, K5 ), m512_zero );
K6 = _mm512_xor_si512( K6, mm512_ror2x512hi_1x32( K4, K5 ) );
K6 = _mm512_xor_si512( K6, _mm512_alignr_epi8( K5, K4, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( X, K6 ), m512_zero );
K7 = _mm512_xor_si512( K7, mm512_ror2x512hi_1x32( K5, K6 ) );
K7 = _mm512_xor_si512( K7, _mm512_alignr_epi8( K6, K5, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( X, K7 ), m512_zero );
P0 = _mm512_xor_si512( P0, X );
@@ -162,24 +158,24 @@ c512_4way( shavite512_4way_context *ctx, const void *msg )
// round 4, 8, 12
K0 = _mm512_xor_si512( K0, mm512_ror2x512hi_1x32( K6, K7 ) );
K0 = _mm512_xor_si512( K0, _mm512_alignr_epi8( K7, K6, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( P1, K0 ), m512_zero );
K1 = _mm512_xor_si512( K1, mm512_ror2x512hi_1x32( K7, K0 ) );
K1 = _mm512_xor_si512( K1, _mm512_alignr_epi8( K0, K7, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( X, K1 ), m512_zero );
K2 = _mm512_xor_si512( K2, mm512_ror2x512hi_1x32( K0, K1 ) );
K2 = _mm512_xor_si512( K2, _mm512_alignr_epi8( K1, K0, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( X, K2 ), m512_zero );
K3 = _mm512_xor_si512( K3, mm512_ror2x512hi_1x32( K1, K2 ) );
K3 = _mm512_xor_si512( K3, _mm512_alignr_epi8( K2, K1, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( X, K3 ), m512_zero );
P0 = _mm512_xor_si512( P0, X );
K4 = _mm512_xor_si512( K4, mm512_ror2x512hi_1x32( K2, K3 ) );
K4 = _mm512_xor_si512( K4, _mm512_alignr_epi8( K3, K2, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( P3, K4 ), m512_zero );
K5 = _mm512_xor_si512( K5, mm512_ror2x512hi_1x32( K3, K4 ) );
K5 = _mm512_xor_si512( K5, _mm512_alignr_epi8( K4, K3, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( X, K5 ), m512_zero );
K6 = _mm512_xor_si512( K6, mm512_ror2x512hi_1x32( K4, K5 ) );
K6 = _mm512_xor_si512( K6, _mm512_alignr_epi8( K5, K4, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( X, K6 ), m512_zero );
K7 = _mm512_xor_si512( K7, mm512_ror2x512hi_1x32( K5, K6 ) );
K7 = _mm512_xor_si512( K7, _mm512_alignr_epi8( K6, K5, 4 ) );
X = _mm512_aesenc_epi128( _mm512_xor_si512( X, K7 ), m512_zero );
P2 = _mm512_xor_si512( P2, X );

56
algo/shavite/sph-shavite-aesni.c
View File

@@ -59,30 +59,6 @@ static const sph_u32 IV512[] = {
C32(0xE275EADE), C32(0x502D9FCD), C32(0xB9357178), C32(0x022A4B9A)
};
// Partially rotate elements in two 128 bit vectors a & b as one 256 bit vector
// and return the rotated 128 bit vector a.
// a[3:0] = { b[0], a[3], a[2], a[1] }
#if defined(__SSSE3__)
#define mm128_ror256hi_1x32( a, b ) _mm_alignr_epi8( b, a, 4 )
#else // SSE2
#define mm128_ror256hi_1x32( a, b ) \
_mm_or_si128( _mm_srli_si128( a, 4 ), \
_mm_slli_si128( b, 12 ) )
#endif
/*
#if defined(__AVX2__)
// 2 way version of above
// a[7:0] = { b[4], a[7], a[6], a[5], b[0], a[3], a[2], a[1] }
#define mm256_ror2x256hi_1x32( a, b ) \
_mm256_blend_epi32( mm256_ror256_1x32( a ), \
mm256_rol256_3x32( b ), 0x88 )
#endif
*/
static void
c512( sph_shavite_big_context *sc, const void *msg )
@@ -190,31 +166,31 @@ c512( sph_shavite_big_context *sc, const void *msg )
// round 2, 6, 10
k00 = _mm_xor_si128( k00, mm128_ror256hi_1x32( k12, k13 ) );
k00 = _mm_xor_si128( k00, _mm_alignr_epi8( k13, k12, 4 ) );
x = _mm_xor_si128( p3, k00 );
x = _mm_aesenc_si128( x, zero );
k01 = _mm_xor_si128( k01, mm128_ror256hi_1x32( k13, k00 ) );
k01 = _mm_xor_si128( k01, _mm_alignr_epi8( k00, k13, 4 ) );
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, zero );
k02 = _mm_xor_si128( k02, mm128_ror256hi_1x32( k00, k01 ) );
k02 = _mm_xor_si128( k02, _mm_alignr_epi8( k01, k00, 4 ) );
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, zero );
k03 = _mm_xor_si128( k03, mm128_ror256hi_1x32( k01, k02 ) );
k03 = _mm_xor_si128( k03, _mm_alignr_epi8( k02, k01, 4 ) );
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, zero );
p2 = _mm_xor_si128( p2, x );
k10 = _mm_xor_si128( k10, mm128_ror256hi_1x32( k02, k03 ) );
k10 = _mm_xor_si128( k10, _mm_alignr_epi8( k03, k02, 4 ) );
x = _mm_xor_si128( p1, k10 );
x = _mm_aesenc_si128( x, zero );
k11 = _mm_xor_si128( k11, mm128_ror256hi_1x32( k03, k10 ) );
k11 = _mm_xor_si128( k11, _mm_alignr_epi8( k10, k03, 4 ) );
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, zero );
k12 = _mm_xor_si128( k12, mm128_ror256hi_1x32( k10, k11 ) );
k12 = _mm_xor_si128( k12, _mm_alignr_epi8( k11, k10, 4 ) );
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, zero );
k13 = _mm_xor_si128( k13, mm128_ror256hi_1x32( k11, k12 ) );
k13 = _mm_xor_si128( k13, _mm_alignr_epi8( k12, k11, 4 ) );
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, zero );
@@ -262,31 +238,31 @@ c512( sph_shavite_big_context *sc, const void *msg )
// round 4, 8, 12
k00 = _mm_xor_si128( k00, mm128_ror256hi_1x32( k12, k13 ) );
k00 = _mm_xor_si128( k00, _mm_alignr_epi8( k13, k12, 4 ) );
x = _mm_xor_si128( p1, k00 );
x = _mm_aesenc_si128( x, zero );
k01 = _mm_xor_si128( k01, mm128_ror256hi_1x32( k13, k00 ) );
k01 = _mm_xor_si128( k01, _mm_alignr_epi8( k00, k13, 4 ) );
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, zero );
k02 = _mm_xor_si128( k02, mm128_ror256hi_1x32( k00, k01 ) );
k02 = _mm_xor_si128( k02, _mm_alignr_epi8( k01, k00, 4 ) );
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, zero );
k03 = _mm_xor_si128( k03, mm128_ror256hi_1x32( k01, k02 ) );
k03 = _mm_xor_si128( k03, _mm_alignr_epi8( k02, k01, 4 ) );
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, zero );
p0 = _mm_xor_si128( p0, x );
k10 = _mm_xor_si128( k10, mm128_ror256hi_1x32( k02, k03 ) );
k10 = _mm_xor_si128( k10, _mm_alignr_epi8( k03, k02, 4 ) );
x = _mm_xor_si128( p3, k10 );
x = _mm_aesenc_si128( x, zero );
k11 = _mm_xor_si128( k11, mm128_ror256hi_1x32( k03, k10 ) );
k11 = _mm_xor_si128( k11, _mm_alignr_epi8( k10, k03, 4 ) );
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, zero );
k12 = _mm_xor_si128( k12, mm128_ror256hi_1x32( k10, k11 ) );
k12 = _mm_xor_si128( k12, _mm_alignr_epi8( k11, k10, 4 ) );
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, zero );
k13 = _mm_xor_si128( k13, mm128_ror256hi_1x32( k11, k12 ) );
k13 = _mm_xor_si128( k13, _mm_alignr_epi8( k12, k11, 4 ) );
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, zero );

2
algo/shavite/sph_shavite.c
View File

@@ -35,7 +35,7 @@
#include "sph_shavite.h"
#if !defined(__AES__)
#if !(defined(__AES__) && defined(__SSSE3__))
#ifdef __cplusplus
extern "C"{

2
algo/shavite/sph_shavite.h
View File

@@ -263,7 +263,7 @@ void sph_shavite384_addbits_and_close(
void *cc, unsigned ub, unsigned n, void *dst);
//Don't call these directly from application code, use the macros below.
#ifdef __AES__
#if defined(__AES__) && defined(__SSSE3__)
void sph_shavite512_aesni_init(void *cc);
void sph_shavite512_aesni(void *cc, const void *data, size_t len);