Skip to content

Commit

Permalink
v3.17.1
Browse files Browse the repository at this point in the history
  • Loading branch information
@JayDDee
JayDDee committed Jul 26, 2021
1 parent 92b3733 commit 9b905fc
Show file tree
Hide file tree
Showing 33 changed files with 889 additions and 565 deletions.
6 changes: 6 additions & 0 deletions RELEASE_NOTES
View file
Original file line number Diff line number Diff line change
Expand Up @@ -65,6 +65,12 @@ If not what makes it happen or not happen?
Change Log
----------

v3.17.1

Fixed Windows build for AES+SSE4.2 (Westmere), was missing AES.
More ternary logic optimizations for AVX512, AVX512+VAES, and AVX512+AES.
Fixed my-gr algo for VAES.

v3.17.0

AVX512 optimized using ternary logic instructions.
Expand Down
1 change: 1 addition & 0 deletions algo/blake/blake-hash-4way.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -180,6 +180,7 @@ void blake512_8way_update( void *cc, const void *data, size_t len );
void blake512_8way_close( void *cc, void *dst );
void blake512_8way_full( blake_8way_big_context *sc, void * dst,
const void *data, size_t len );
void blake512_8way_hash_le80( void *hash, const void *data );

#endif // AVX512
#endif // AVX2
Expand Down
32 changes: 16 additions & 16 deletions algo/blake/blake256-hash-4way.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -669,14 +669,14 @@ do { \
ROUND_S_8WAY(2); \
ROUND_S_8WAY(3); \
} \
H0 = _mm256_xor_si256( _mm256_xor_si256( V8, V0 ), H0 ); \
H1 = _mm256_xor_si256( _mm256_xor_si256( V9, V1 ), H1 ); \
H2 = _mm256_xor_si256( _mm256_xor_si256( VA, V2 ), H2 ); \
H3 = _mm256_xor_si256( _mm256_xor_si256( VB, V3 ), H3 ); \
H4 = _mm256_xor_si256( _mm256_xor_si256( VC, V4 ), H4 ); \
H5 = _mm256_xor_si256( _mm256_xor_si256( VD, V5 ), H5 ); \
H6 = _mm256_xor_si256( _mm256_xor_si256( VE, V6 ), H6 ); \
H7 = _mm256_xor_si256( _mm256_xor_si256( VF, V7 ), H7 ); \
H0 = mm256_xor3( V8, V0, H0 ); \
H1 = mm256_xor3( V9, V1, H1 ); \
H2 = mm256_xor3( VA, V2, H2 ); \
H3 = mm256_xor3( VB, V3, H3 ); \
H4 = mm256_xor3( VC, V4, H4 ); \
H5 = mm256_xor3( VD, V5, H5 ); \
H6 = mm256_xor3( VE, V6, H6 ); \
H7 = mm256_xor3( VF, V7, H7 ); \
} while (0)


Expand Down Expand Up @@ -808,14 +808,14 @@ do { \
ROUND_S_16WAY(2); \
ROUND_S_16WAY(3); \
} \
H0 = _mm512_xor_si512( _mm512_xor_si512( V8, V0 ), H0 ); \
H1 = _mm512_xor_si512( _mm512_xor_si512( V9, V1 ), H1 ); \
H2 = _mm512_xor_si512( _mm512_xor_si512( VA, V2 ), H2 ); \
H3 = _mm512_xor_si512( _mm512_xor_si512( VB, V3 ), H3 ); \
H4 = _mm512_xor_si512( _mm512_xor_si512( VC, V4 ), H4 ); \
H5 = _mm512_xor_si512( _mm512_xor_si512( VD, V5 ), H5 ); \
H6 = _mm512_xor_si512( _mm512_xor_si512( VE, V6 ), H6 ); \
H7 = _mm512_xor_si512( _mm512_xor_si512( VF, V7 ), H7 ); \
H0 = mm512_xor3( V8, V0, H0 ); \
H1 = mm512_xor3( V9, V1, H1 ); \
H2 = mm512_xor3( VA, V2, H2 ); \
H3 = mm512_xor3( VB, V3, H3 ); \
H4 = mm512_xor3( VC, V4, H4 ); \
H5 = mm512_xor3( VD, V5, H5 ); \
H6 = mm512_xor3( VE, V6, H6 ); \
H7 = mm512_xor3( VF, V7, H7 ); \
} while (0)

#endif
Expand Down
16 changes: 8 additions & 8 deletions algo/blake/blake2b-hash-4way.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -122,14 +122,14 @@ static void blake2b_8way_compress( blake2b_8way_ctx *ctx, int last )
B2B8W_G( 3, 4, 9, 14, m[ sigma[i][14] ], m[ sigma[i][15] ] );
}

ctx->h[0] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[0], v[0] ), v[ 8] );
ctx->h[1] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[1], v[1] ), v[ 9] );
ctx->h[2] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[2], v[2] ), v[10] );
ctx->h[3] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[3], v[3] ), v[11] );
ctx->h[4] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[4], v[4] ), v[12] );
ctx->h[5] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[5], v[5] ), v[13] );
ctx->h[6] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[6], v[6] ), v[14] );
ctx->h[7] = _mm512_xor_si512( _mm512_xor_si512( ctx->h[7], v[7] ), v[15] );
ctx->h[0] = mm512_xor3( ctx->h[0], v[0], v[ 8] );
ctx->h[1] = mm512_xor3( ctx->h[1], v[1], v[ 9] );
ctx->h[2] = mm512_xor3( ctx->h[2], v[2], v[10] );
ctx->h[3] = mm512_xor3( ctx->h[3], v[3], v[11] );
ctx->h[4] = mm512_xor3( ctx->h[4], v[4], v[12] );
ctx->h[5] = mm512_xor3( ctx->h[5], v[5], v[13] );
ctx->h[6] = mm512_xor3( ctx->h[6], v[6], v[14] );
ctx->h[7] = mm512_xor3( ctx->h[7], v[7], v[15] );
}

int blake2b_8way_init( blake2b_8way_ctx *ctx )
Expand Down
3 changes: 0 additions & 3 deletions algo/blake/blake2s-gate.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,6 @@
#include <stdint.h>
#include "algo-gate-api.h"

//#if defined(__SSE4_2__)
#if defined(__SSE2__)
#define BLAKE2S_4WAY
#endif
Expand All @@ -27,8 +26,6 @@ int scanhash_blake2s_16way( struct work *work, uint32_t max_nonce,

#elif defined (BLAKE2S_8WAY)

//#if defined(BLAKE2S_8WAY)

void blake2s_8way_hash( void *state, const void *input );
int scanhash_blake2s_8way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
Expand Down
4 changes: 2 additions & 2 deletions algo/blake/blake2s-hash-4way.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -368,7 +368,7 @@ do { \
ROUND8W( 9 );

for( size_t i = 0; i < 8; ++i )
S->h[i] = _mm256_xor_si256( _mm256_xor_si256( S->h[i], v[i] ), v[i + 8] );
S->h[i] = mm256_xor3( S->h[i], v[i], v[i + 8] );

#undef G8W
#undef ROUND8W
Expand Down Expand Up @@ -566,7 +566,7 @@ do { \
ROUND16W( 9 );

for( size_t i = 0; i < 8; ++i )
S->h[i] = _mm512_xor_si512( _mm512_xor_si512( S->h[i], v[i] ), v[i + 8] );
S->h[i] = mm512_xor3( S->h[i], v[i], v[i + 8] );

#undef G16W
#undef ROUND16W
Expand Down
128 changes: 48 additions & 80 deletions algo/blake/blake512-hash-4way.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -293,10 +293,6 @@ static const sph_u64 CB[16] = {
H5 = (state)->H[5]; \
H6 = (state)->H[6]; \
H7 = (state)->H[7]; \
S0 = (state)->S[0]; \
S1 = (state)->S[1]; \
S2 = (state)->S[2]; \
S3 = (state)->S[3]; \
T0 = (state)->T0; \
T1 = (state)->T1; \
} while (0)
Expand All @@ -310,10 +306,6 @@ static const sph_u64 CB[16] = {
(state)->H[5] = H5; \
(state)->H[6] = H6; \
(state)->H[7] = H7; \
(state)->S[0] = S0; \
(state)->S[1] = S1; \
(state)->S[2] = S2; \
(state)->S[3] = S3; \
(state)->T0 = T0; \
(state)->T1 = T1; \
} while (0)
Expand Down Expand Up @@ -348,7 +340,6 @@ static const sph_u64 CB[16] = {

#define DECL_STATE64_8WAY \
__m512i H0, H1, H2, H3, H4, H5, H6, H7; \
__m512i S0, S1, S2, S3; \
uint64_t T0, T1;

#define COMPRESS64_8WAY( buf ) do \
Expand All @@ -366,10 +357,10 @@ static const sph_u64 CB[16] = {
V5 = H5; \
V6 = H6; \
V7 = H7; \
V8 = _mm512_xor_si512( S0, m512_const1_64( CB0 ) ); \
V9 = _mm512_xor_si512( S1, m512_const1_64( CB1 ) ); \
VA = _mm512_xor_si512( S2, m512_const1_64( CB2 ) ); \
VB = _mm512_xor_si512( S3, m512_const1_64( CB3 ) ); \
V8 = m512_const1_64( CB0 ); \
V9 = m512_const1_64( CB1 ); \
VA = m512_const1_64( CB2 ); \
VB = m512_const1_64( CB3 ); \
VC = _mm512_xor_si512( _mm512_set1_epi64( T0 ), \
m512_const1_64( CB4 ) ); \
VD = _mm512_xor_si512( _mm512_set1_epi64( T0 ), \
Expand Down Expand Up @@ -414,14 +405,14 @@ static const sph_u64 CB[16] = {
ROUND_B_8WAY(3); \
ROUND_B_8WAY(4); \
ROUND_B_8WAY(5); \
H0 = mm512_xor4( V8, V0, S0, H0 ); \
H1 = mm512_xor4( V9, V1, S1, H1 ); \
H2 = mm512_xor4( VA, V2, S2, H2 ); \
H3 = mm512_xor4( VB, V3, S3, H3 ); \
H4 = mm512_xor4( VC, V4, S0, H4 ); \
H5 = mm512_xor4( VD, V5, S1, H5 ); \
H6 = mm512_xor4( VE, V6, S2, H6 ); \
H7 = mm512_xor4( VF, V7, S3, H7 ); \
H0 = mm512_xor3( V8, V0, H0 ); \
H1 = mm512_xor3( V9, V1, H1 ); \
H2 = mm512_xor3( VA, V2, H2 ); \
H3 = mm512_xor3( VB, V3, H3 ); \
H4 = mm512_xor3( VC, V4, H4 ); \
H5 = mm512_xor3( VD, V5, H5 ); \
H6 = mm512_xor3( VE, V6, H6 ); \
H7 = mm512_xor3( VF, V7, H7 ); \
} while (0)

void blake512_8way_compress( blake_8way_big_context *sc )
Expand All @@ -440,10 +431,10 @@ void blake512_8way_compress( blake_8way_big_context *sc )
V5 = sc->H[5];
V6 = sc->H[6];
V7 = sc->H[7];
V8 = _mm512_xor_si512( sc->S[0], m512_const1_64( CB0 ) );
V9 = _mm512_xor_si512( sc->S[1], m512_const1_64( CB1 ) );
VA = _mm512_xor_si512( sc->S[2], m512_const1_64( CB2 ) );
VB = _mm512_xor_si512( sc->S[3], m512_const1_64( CB3 ) );
V8 = m512_const1_64( CB0 );
V9 = m512_const1_64( CB1 );
VA = m512_const1_64( CB2 );
VB = m512_const1_64( CB3 );
VC = _mm512_xor_si512( _mm512_set1_epi64( sc->T0 ),
m512_const1_64( CB4 ) );
VD = _mm512_xor_si512( _mm512_set1_epi64( sc->T0 ),
Expand Down Expand Up @@ -492,19 +483,18 @@ void blake512_8way_compress( blake_8way_big_context *sc )
ROUND_B_8WAY(4);
ROUND_B_8WAY(5);

sc->H[0] = mm512_xor4( V8, V0, sc->S[0], sc->H[0] );
sc->H[1] = mm512_xor4( V9, V1, sc->S[1], sc->H[1] );
sc->H[2] = mm512_xor4( VA, V2, sc->S[2], sc->H[2] );
sc->H[3] = mm512_xor4( VB, V3, sc->S[3], sc->H[3] );
sc->H[4] = mm512_xor4( VC, V4, sc->S[0], sc->H[4] );
sc->H[5] = mm512_xor4( VD, V5, sc->S[1], sc->H[5] );
sc->H[6] = mm512_xor4( VE, V6, sc->S[2], sc->H[6] );
sc->H[7] = mm512_xor4( VF, V7, sc->S[3], sc->H[7] );
sc->H[0] = mm512_xor3( V8, V0, sc->H[0] );
sc->H[1] = mm512_xor3( V9, V1, sc->H[1] );
sc->H[2] = mm512_xor3( VA, V2, sc->H[2] );
sc->H[3] = mm512_xor3( VB, V3, sc->H[3] );
sc->H[4] = mm512_xor3( VC, V4, sc->H[4] );
sc->H[5] = mm512_xor3( VD, V5, sc->H[5] );
sc->H[6] = mm512_xor3( VE, V6, sc->H[6] );
sc->H[7] = mm512_xor3( VF, V7, sc->H[7] );
}

void blake512_8way_init( blake_8way_big_context *sc )
{
__m512i zero = m512_zero;
casti_m512i( sc->H, 0 ) = m512_const1_64( 0x6A09E667F3BCC908 );
casti_m512i( sc->H, 1 ) = m512_const1_64( 0xBB67AE8584CAA73B );
casti_m512i( sc->H, 2 ) = m512_const1_64( 0x3C6EF372FE94F82B );
Expand All @@ -514,11 +504,6 @@ void blake512_8way_init( blake_8way_big_context *sc )
casti_m512i( sc->H, 6 ) = m512_const1_64( 0x1F83D9ABFB41BD6B );
casti_m512i( sc->H, 7 ) = m512_const1_64( 0x5BE0CD19137E2179 );

casti_m512i( sc->S, 0 ) = zero;
casti_m512i( sc->S, 1 ) = zero;
casti_m512i( sc->S, 2 ) = zero;
casti_m512i( sc->S, 3 ) = zero;

sc->T0 = sc->T1 = 0;
sc->ptr = 0;
}
Expand Down Expand Up @@ -641,11 +626,6 @@ void blake512_8way_full( blake_8way_big_context *sc, void * dst,
casti_m512i( sc->H, 6 ) = m512_const1_64( 0x1F83D9ABFB41BD6B );
casti_m512i( sc->H, 7 ) = m512_const1_64( 0x5BE0CD19137E2179 );

casti_m512i( sc->S, 0 ) = m512_zero;
casti_m512i( sc->S, 1 ) = m512_zero;
casti_m512i( sc->S, 2 ) = m512_zero;
casti_m512i( sc->S, 3 ) = m512_zero;

sc->T0 = sc->T1 = 0;
sc->ptr = 0;

Expand Down Expand Up @@ -740,7 +720,6 @@ blake512_8way_close(void *cc, void *dst)

#define DECL_STATE64_4WAY \
__m256i H0, H1, H2, H3, H4, H5, H6, H7; \
__m256i S0, S1, S2, S3; \
uint64_t T0, T1;

#define COMPRESS64_4WAY do \
Expand All @@ -758,10 +737,10 @@ blake512_8way_close(void *cc, void *dst)
V5 = H5; \
V6 = H6; \
V7 = H7; \
V8 = _mm256_xor_si256( S0, m256_const1_64( CB0 ) ); \
V9 = _mm256_xor_si256( S1, m256_const1_64( CB1 ) ); \
VA = _mm256_xor_si256( S2, m256_const1_64( CB2 ) ); \
VB = _mm256_xor_si256( S3, m256_const1_64( CB3 ) ); \
V8 = m256_const1_64( CB0 ); \
V9 = m256_const1_64( CB1 ); \
VA = m256_const1_64( CB2 ); \
VB = m256_const1_64( CB3 ); \
VC = _mm256_xor_si256( _mm256_set1_epi64x( T0 ), \
m256_const1_64( CB4 ) ); \
VD = _mm256_xor_si256( _mm256_set1_epi64x( T0 ), \
Expand Down Expand Up @@ -804,14 +783,14 @@ blake512_8way_close(void *cc, void *dst)
ROUND_B_4WAY(3); \
ROUND_B_4WAY(4); \
ROUND_B_4WAY(5); \
H0 = mm256_xor4( V8, V0, S0, H0 ); \
H1 = mm256_xor4( V9, V1, S1, H1 ); \
H2 = mm256_xor4( VA, V2, S2, H2 ); \
H3 = mm256_xor4( VB, V3, S3, H3 ); \
H4 = mm256_xor4( VC, V4, S0, H4 ); \
H5 = mm256_xor4( VD, V5, S1, H5 ); \
H6 = mm256_xor4( VE, V6, S2, H6 ); \
H7 = mm256_xor4( VF, V7, S3, H7 ); \
H0 = mm256_xor3( V8, V0, H0 ); \
H1 = mm256_xor3( V9, V1, H1 ); \
H2 = mm256_xor3( VA, V2, H2 ); \
H3 = mm256_xor3( VB, V3, H3 ); \
H4 = mm256_xor3( VC, V4, H4 ); \
H5 = mm256_xor3( VD, V5, H5 ); \
H6 = mm256_xor3( VE, V6, H6 ); \
H7 = mm256_xor3( VF, V7, H7 ); \
} while (0)


Expand All @@ -831,10 +810,10 @@ void blake512_4way_compress( blake_4way_big_context *sc )
V5 = sc->H[5];
V6 = sc->H[6];
V7 = sc->H[7];
V8 = _mm256_xor_si256( sc->S[0], m256_const1_64( CB0 ) );
V9 = _mm256_xor_si256( sc->S[1], m256_const1_64( CB1 ) );
VA = _mm256_xor_si256( sc->S[2], m256_const1_64( CB2 ) );
VB = _mm256_xor_si256( sc->S[3], m256_const1_64( CB3 ) );
V8 = m256_const1_64( CB0 );
V9 = m256_const1_64( CB1 );
VA = m256_const1_64( CB2 );
VB = m256_const1_64( CB3 );
VC = _mm256_xor_si256( _mm256_set1_epi64x( sc->T0 ),
m256_const1_64( CB4 ) );
VD = _mm256_xor_si256( _mm256_set1_epi64x( sc->T0 ),
Expand Down Expand Up @@ -880,19 +859,18 @@ void blake512_4way_compress( blake_4way_big_context *sc )
ROUND_B_4WAY(4);
ROUND_B_4WAY(5);

sc->H[0] = mm256_xor4( V8, V0, sc->S[0], sc->H[0] );
sc->H[1] = mm256_xor4( V9, V1, sc->S[1], sc->H[1] );
sc->H[2] = mm256_xor4( VA, V2, sc->S[2], sc->H[2] );
sc->H[3] = mm256_xor4( VB, V3, sc->S[3], sc->H[3] );
sc->H[4] = mm256_xor4( VC, V4, sc->S[0], sc->H[4] );
sc->H[5] = mm256_xor4( VD, V5, sc->S[1], sc->H[5] );
sc->H[6] = mm256_xor4( VE, V6, sc->S[2], sc->H[6] );
sc->H[7] = mm256_xor4( VF, V7, sc->S[3], sc->H[7] );
sc->H[0] = mm256_xor3( V8, V0, sc->H[0] );
sc->H[1] = mm256_xor3( V9, V1, sc->H[1] );
sc->H[2] = mm256_xor3( VA, V2, sc->H[2] );
sc->H[3] = mm256_xor3( VB, V3, sc->H[3] );
sc->H[4] = mm256_xor3( VC, V4, sc->H[4] );
sc->H[5] = mm256_xor3( VD, V5, sc->H[5] );
sc->H[6] = mm256_xor3( VE, V6, sc->H[6] );
sc->H[7] = mm256_xor3( VF, V7, sc->H[7] );
}

void blake512_4way_init( blake_4way_big_context *sc )
{
__m256i zero = m256_zero;
casti_m256i( sc->H, 0 ) = m256_const1_64( 0x6A09E667F3BCC908 );
casti_m256i( sc->H, 1 ) = m256_const1_64( 0xBB67AE8584CAA73B );
casti_m256i( sc->H, 2 ) = m256_const1_64( 0x3C6EF372FE94F82B );
Expand All @@ -902,11 +880,6 @@ void blake512_4way_init( blake_4way_big_context *sc )
casti_m256i( sc->H, 6 ) = m256_const1_64( 0x1F83D9ABFB41BD6B );
casti_m256i( sc->H, 7 ) = m256_const1_64( 0x5BE0CD19137E2179 );

casti_m256i( sc->S, 0 ) = zero;
casti_m256i( sc->S, 1 ) = zero;
casti_m256i( sc->S, 2 ) = zero;
casti_m256i( sc->S, 3 ) = zero;

sc->T0 = sc->T1 = 0;
sc->ptr = 0;
}
Expand Down Expand Up @@ -1026,11 +999,6 @@ void blake512_4way_full( blake_4way_big_context *sc, void * dst,
casti_m256i( sc->H, 6 ) = m256_const1_64( 0x1F83D9ABFB41BD6B );
casti_m256i( sc->H, 7 ) = m256_const1_64( 0x5BE0CD19137E2179 );

casti_m256i( sc->S, 0 ) = m256_zero;
casti_m256i( sc->S, 1 ) = m256_zero;
casti_m256i( sc->S, 2 ) = m256_zero;
casti_m256i( sc->S, 3 ) = m256_zero;

sc->T0 = sc->T1 = 0;
sc->ptr = 0;

Expand Down
Loading

0 comments on commit 9b905fc

Please sign in to comment.