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

v3.10.1

Browse Source
This commit is contained in:
Jay D Dee
2019-12-05 19:09:23 -05:00
parent 40039386a0
commit 73430b13b1
52 changed files with 4515 additions and 874 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
algo/blake/blake-hash-4way.h
View File

@@ -118,20 +118,42 @@ void blake256r8_8way_close(void *cc, void *dst);
// Blake-512 4 way
typedef struct {
__m256i buf[16] __attribute__ ((aligned (64)));
__m256i buf[16];
__m256i H[8];
__m256i S[4];
size_t ptr;
sph_u64 T0, T1;
} blake_4way_big_context;
} blake_4way_big_context __attribute__ ((aligned (128)));
typedef blake_4way_big_context blake512_4way_context;
void blake512_4way_init(void *cc);
void blake512_4way(void *cc, const void *data, size_t len);
void blake512_4way_close(void *cc, void *dst);
void blake512_4way_addbits_and_close(
void *cc, unsigned ub, unsigned n, void *dst);
void blake512_4way_init( void *cc );
void blake512_4way_update( void *cc, const void *data, size_t len );
#define blake512_4way blake512_4way_update
void blake512_4way_close( void *cc, void *dst );
void blake512_4way_addbits_and_close( void *cc, unsigned ub, unsigned n,
void *dst );
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
typedef struct {
__m512i buf[16];
__m512i H[8];
__m512i S[4];
size_t ptr;
sph_u64 T0, T1;
} blake_8way_big_context __attribute__ ((aligned (128)));
typedef blake_8way_big_context blake512_8way_context;
void blake512_8way_init( void *cc );
void blake512_8way_update( void *cc, const void *data, size_t len );
void blake512_8way_close( void *cc, void *dst );
void blake512_8way_addbits_and_close( void *cc, unsigned ub, unsigned n,
void *dst );
#endif // AVX512
#endif // AVX2

2
algo/blake/blake2b-4way.c
View File

@@ -17,7 +17,7 @@ int scanhash_blake2b_8way( struct work *work, uint32_t max_nonce,
uint32_t vdata[20*8] __attribute__ ((aligned (64)));;
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
blake2b_8way_ctx ctx __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[25]); // 3*8+1
uint32_t *hash7 = &(hash[49]); // 3*16+1
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
int thr_id = mythr->id;

444
algo/blake/blake512-hash-4way.c
View File

@@ -54,7 +54,6 @@ extern "C"{
#pragma warning (disable: 4146)
#endif
// Blake-512
static const sph_u64 IV512[8] = {
@@ -64,6 +63,7 @@ static const sph_u64 IV512[8] = {
SPH_C64(0x1F83D9ABFB41BD6B), SPH_C64(0x5BE0CD19137E2179)
};
static const sph_u64 salt_zero_big[4] = { 0, 0, 0, 0 };
#if SPH_COMPACT_BLAKE_32 || SPH_COMPACT_BLAKE_64
@@ -264,8 +264,6 @@ static const unsigned sigma[16][16] = {
#define Mx_(n) Mx__(n)
#define Mx__(n) M ## n
// Blake-512 4 way
#define CBx(r, i) CBx_(Z ## r ## i)
#define CBx_(n) CBx__(n)
#define CBx__(n) CB ## n
@@ -287,6 +285,7 @@ static const unsigned sigma[16][16] = {
#define CBE SPH_C64(0x0801F2E2858EFC16)
#define CBF SPH_C64(0x636920D871574E69)
/*
#if SPH_COMPACT_BLAKE_64
// not used
static const sph_u64 CB[16] = {
@@ -301,7 +300,301 @@ static const sph_u64 CB[16] = {
};
#endif
*/
#define READ_STATE64(state) do { \
H0 = (state)->H[0]; \
H1 = (state)->H[1]; \
H2 = (state)->H[2]; \
H3 = (state)->H[3]; \
H4 = (state)->H[4]; \
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)
#define WRITE_STATE64(state) do { \
(state)->H[0] = H0; \
(state)->H[1] = H1; \
(state)->H[2] = H2; \
(state)->H[3] = H3; \
(state)->H[4] = H4; \
(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)
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
// Blake-512 8 way
#define GB_8WAY(m0, m1, c0, c1, a, b, c, d) do { \
a = _mm512_add_epi64( _mm512_add_epi64( _mm512_xor_si512( \
_mm512_set1_epi64( c1 ), m0 ), b ), a ); \
d = mm512_ror_64( _mm512_xor_si512( d, a ), 32 ); \
c = _mm512_add_epi64( c, d ); \
b = mm512_ror_64( _mm512_xor_si512( b, c ), 25 ); \
a = _mm512_add_epi64( _mm512_add_epi64( _mm512_xor_si512( \
_mm512_set1_epi64( c0 ), m1 ), b ), a ); \
d = mm512_ror_64( _mm512_xor_si512( d, a ), 16 ); \
c = _mm512_add_epi64( c, d ); \
b = mm512_ror_64( _mm512_xor_si512( b, c ), 11 ); \
} while (0)
#define ROUND_B_8WAY(r) do { \
GB_8WAY(Mx(r, 0), Mx(r, 1), CBx(r, 0), CBx(r, 1), V0, V4, V8, VC); \
GB_8WAY(Mx(r, 2), Mx(r, 3), CBx(r, 2), CBx(r, 3), V1, V5, V9, VD); \
GB_8WAY(Mx(r, 4), Mx(r, 5), CBx(r, 4), CBx(r, 5), V2, V6, VA, VE); \
GB_8WAY(Mx(r, 6), Mx(r, 7), CBx(r, 6), CBx(r, 7), V3, V7, VB, VF); \
GB_8WAY(Mx(r, 8), Mx(r, 9), CBx(r, 8), CBx(r, 9), V0, V5, VA, VF); \
GB_8WAY(Mx(r, A), Mx(r, B), CBx(r, A), CBx(r, B), V1, V6, VB, VC); \
GB_8WAY(Mx(r, C), Mx(r, D), CBx(r, C), CBx(r, D), V2, V7, V8, VD); \
GB_8WAY(Mx(r, E), Mx(r, F), CBx(r, E), CBx(r, F), V3, V4, V9, VE); \
} while (0)
#define DECL_STATE64_8WAY \
__m512i H0, H1, H2, H3, H4, H5, H6, H7; \
__m512i S0, S1, S2, S3; \
sph_u64 T0, T1;
#define COMPRESS64_8WAY do \
{ \
__m512i M0, M1, M2, M3, M4, M5, M6, M7; \
__m512i M8, M9, MA, MB, MC, MD, ME, MF; \
__m512i V0, V1, V2, V3, V4, V5, V6, V7; \
__m512i V8, V9, VA, VB, VC, VD, VE, VF; \
__m512i shuf_bswap64; \
V0 = H0; \
V1 = H1; \
V2 = H2; \
V3 = H3; \
V4 = H4; \
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 ) ); \
VC = _mm512_xor_si512( _mm512_set1_epi64( T0 ), \
m512_const1_64( CB4 ) ); \
VD = _mm512_xor_si512( _mm512_set1_epi64( T0 ), \
m512_const1_64( CB5 ) ); \
VE = _mm512_xor_si512( _mm512_set1_epi64( T1 ), \
m512_const1_64( CB6 ) ); \
VF = _mm512_xor_si512( _mm512_set1_epi64( T1 ), \
m512_const1_64( CB7 ) ); \
shuf_bswap64 = m512_const_64( 0x38393a3b3c3d3e3f, 0x3031323334353637, \
0x28292a2b2c2d2e2f, 0x2021222324252627, \
0x18191a1b1c1d1e1f, 0x1011121314151617, \
0x08090a0b0c0d0e0f, 0x0001020304050607 ); \
M0 = _mm512_shuffle_epi8( *(buf+ 0), shuf_bswap64 ); \
M1 = _mm512_shuffle_epi8( *(buf+ 1), shuf_bswap64 ); \
M2 = _mm512_shuffle_epi8( *(buf+ 2), shuf_bswap64 ); \
M3 = _mm512_shuffle_epi8( *(buf+ 3), shuf_bswap64 ); \
M4 = _mm512_shuffle_epi8( *(buf+ 4), shuf_bswap64 ); \
M5 = _mm512_shuffle_epi8( *(buf+ 5), shuf_bswap64 ); \
M6 = _mm512_shuffle_epi8( *(buf+ 6), shuf_bswap64 ); \
M7 = _mm512_shuffle_epi8( *(buf+ 7), shuf_bswap64 ); \
M8 = _mm512_shuffle_epi8( *(buf+ 8), shuf_bswap64 ); \
M9 = _mm512_shuffle_epi8( *(buf+ 9), shuf_bswap64 ); \
MA = _mm512_shuffle_epi8( *(buf+10), shuf_bswap64 ); \
MB = _mm512_shuffle_epi8( *(buf+11), shuf_bswap64 ); \
MC = _mm512_shuffle_epi8( *(buf+12), shuf_bswap64 ); \
MD = _mm512_shuffle_epi8( *(buf+13), shuf_bswap64 ); \
ME = _mm512_shuffle_epi8( *(buf+14), shuf_bswap64 ); \
MF = _mm512_shuffle_epi8( *(buf+15), shuf_bswap64 ); \
ROUND_B_8WAY(0); \
ROUND_B_8WAY(1); \
ROUND_B_8WAY(2); \
ROUND_B_8WAY(3); \
ROUND_B_8WAY(4); \
ROUND_B_8WAY(5); \
ROUND_B_8WAY(6); \
ROUND_B_8WAY(7); \
ROUND_B_8WAY(8); \
ROUND_B_8WAY(9); \
ROUND_B_8WAY(0); \
ROUND_B_8WAY(1); \
ROUND_B_8WAY(2); \
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 ); \
} while (0)
static void
blake64_8way_init( blake_8way_big_context *sc, const sph_u64 *iv,
const sph_u64 *salt )
{
__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 );
casti_m512i( sc->H, 3 ) = m512_const1_64( 0xA54FF53A5F1D36F1 );
casti_m512i( sc->H, 4 ) = m512_const1_64( 0x510E527FADE682D1 );
casti_m512i( sc->H, 5 ) = m512_const1_64( 0x9B05688C2B3E6C1F );
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;
}
static void
blake64_8way( blake_8way_big_context *sc, const void *data, size_t len )
{
__m512i *vdata = (__m512i*)data;
__m512i *buf;
size_t ptr;
DECL_STATE64_8WAY
const int buf_size = 128; // sizeof/8
buf = sc->buf;
ptr = sc->ptr;
if ( len < (buf_size - ptr) )
{
memcpy_512( buf + (ptr>>3), vdata, len>>3 );
ptr += len;
sc->ptr = ptr;
return;
}
READ_STATE64(sc);
while ( len > 0 )
{
size_t clen;
clen = buf_size - ptr;
if ( clen > len )
clen = len;
memcpy_512( buf + (ptr>>3), vdata, clen>>3 );
ptr += clen;
vdata = vdata + (clen>>3);
len -= clen;
if ( ptr == buf_size )
{
if ( ( T0 = SPH_T64(T0 + 1024) ) < 1024 )
T1 = SPH_T64(T1 + 1);
COMPRESS64_8WAY;
ptr = 0;
}
}
WRITE_STATE64(sc);
sc->ptr = ptr;
}
static void
blake64_8way_close( blake_8way_big_context *sc,
unsigned ub, unsigned n, void *dst, size_t out_size_w64)
{
__m512i buf[16];
size_t ptr;
unsigned bit_len;
uint64_t z, zz;
sph_u64 th, tl;
ptr = sc->ptr;
bit_len = ((unsigned)ptr << 3);
z = 0x80 >> n;
zz = ((ub & -z) | z) & 0xFF;
buf[ptr>>3] = _mm512_set1_epi64( zz );
tl = sc->T0 + bit_len;
th = sc->T1;
if (ptr == 0 )
{
sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL);
sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFFULL);
}
else if ( sc->T0 == 0 )
{
sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL) + bit_len;
sc->T1 = SPH_T64(sc->T1 - 1);
}
else
{
sc->T0 -= 1024 - bit_len;
}
if ( ptr <= 104 )
{
memset_zero_512( buf + (ptr>>3) + 1, (104-ptr) >> 3 );
if ( out_size_w64 == 8 )
buf[(104>>3)] = _mm512_or_si512( buf[(104>>3)],
m512_const1_64( 0x0100000000000000ULL ) );
*(buf+(112>>3)) = _mm512_set1_epi64( bswap_64( th ) );
*(buf+(120>>3)) = _mm512_set1_epi64( bswap_64( tl ) );
blake64_8way( sc, buf + (ptr>>3), 128 - ptr );
}
else
{
memset_zero_512( buf + (ptr>>3) + 1, (120 - ptr) >> 3 );
blake64_8way( sc, buf + (ptr>>3), 128 - ptr );
sc->T0 = SPH_C64(0xFFFFFFFFFFFFFC00ULL);
sc->T1 = SPH_C64(0xFFFFFFFFFFFFFFFFULL);
memset_zero_512( buf, 112>>3 );
if ( out_size_w64 == 8 )
buf[104>>3] = m512_const1_64( 0x0100000000000000ULL );
*(buf+(112>>3)) = _mm512_set1_epi64( bswap_64( th ) );
*(buf+(120>>3)) = _mm512_set1_epi64( bswap_64( tl ) );
blake64_8way( sc, buf, 128 );
}
mm512_block_bswap_64( (__m512i*)dst, sc->H );
}
void
blake512_8way_init(void *cc)
{
blake64_8way_init(cc, IV512, salt_zero_big);
}
void
blake512_8way_update(void *cc, const void *data, size_t len)
{
blake64_8way(cc, data, len);
}
void
blake512_8way_close(void *cc, void *dst)
{
blake512_8way_addbits_and_close(cc, 0, 0, dst);
}
void
blake512_8way_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
{
blake64_8way_close(cc, ub, n, dst, 8);
}
#endif // AVX512
// Blake-512 4 way
@@ -318,29 +611,6 @@ static const sph_u64 CB[16] = {
b = mm256_ror_64( _mm256_xor_si256( b, c ), 11 ); \
} while (0)
#if SPH_COMPACT_BLAKE_64
// not used
#define ROUND_B_4WAY(r) do { \
GB_4WAY(M[sigma[r][0x0]], M[sigma[r][0x1]], \
CB[sigma[r][0x0]], CB[sigma[r][0x1]], V0, V4, V8, VC); \
GB_4WAY(M[sigma[r][0x2]], M[sigma[r][0x3]], \
CB[sigma[r][0x2]], CB[sigma[r][0x3]], V1, V5, V9, VD); \
GB_4WAY(M[sigma[r][0x4]], M[sigma[r][0x5]], \
CB[sigma[r][0x4]], CB[sigma[r][0x5]], V2, V6, VA, VE); \
GB_4WAY(M[sigma[r][0x6]], M[sigma[r][0x7]], \
CB[sigma[r][0x6]], CB[sigma[r][0x7]], V3, V7, VB, VF); \
GB_4WAY(M[sigma[r][0x8]], M[sigma[r][0x9]], \
CB[sigma[r][0x8]], CB[sigma[r][0x9]], V0, V5, VA, VF); \
GB_4WAY(M[sigma[r][0xA]], M[sigma[r][0xB]], \
CB[sigma[r][0xA]], CB[sigma[r][0xB]], V1, V6, VB, VC); \
GB_4WAY(M[sigma[r][0xC]], M[sigma[r][0xD]], \
CB[sigma[r][0xC]], CB[sigma[r][0xD]], V2, V7, V8, VD); \
GB_4WAY(M[sigma[r][0xE]], M[sigma[r][0xF]], \
CB[sigma[r][0xE]], CB[sigma[r][0xF]], V3, V4, V9, VE); \
} while (0)
#else
//current_impl
#define ROUND_B_4WAY(r) do { \
GB_4WAY(Mx(r, 0), Mx(r, 1), CBx(r, 0), CBx(r, 1), V0, V4, V8, VC); \
GB_4WAY(Mx(r, 2), Mx(r, 3), CBx(r, 2), CBx(r, 3), V1, V5, V9, VD); \
@@ -352,120 +622,11 @@ static const sph_u64 CB[16] = {
GB_4WAY(Mx(r, E), Mx(r, F), CBx(r, E), CBx(r, F), V3, V4, V9, VE); \
} while (0)
#endif
// Blake-512 4 way
#define DECL_STATE64_4WAY \
__m256i H0, H1, H2, H3, H4, H5, H6, H7; \
__m256i S0, S1, S2, S3; \
sph_u64 T0, T1;
#define READ_STATE64_4WAY(state) do { \
H0 = (state)->H[0]; \
H1 = (state)->H[1]; \
H2 = (state)->H[2]; \
H3 = (state)->H[3]; \
H4 = (state)->H[4]; \
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)
#define WRITE_STATE64_4WAY(state) do { \
(state)->H[0] = H0; \
(state)->H[1] = H1; \
(state)->H[2] = H2; \
(state)->H[3] = H3; \
(state)->H[4] = H4; \
(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)
#if SPH_COMPACT_BLAKE_64
// not used
#define COMPRESS64_4WAY do { \
__m256i M[16]; \
__m256i V0, V1, V2, V3, V4, V5, V6, V7; \
__m256i V8, V9, VA, VB, VC, VD, VE, VF; \
const __m256i shuff_bswap64 = m256_const2_64( 0x08090a0b0c0d0e0f, \
0x0001020304050607 ) \
unsigned r; \
V0 = H0; \
V1 = H1; \
V2 = H2; \
V3 = H3; \
V4 = H4; \
V5 = H5; \
V6 = H6; \
V7 = H7; \
V8 = _mm256_xor_si256( S0, _mm256_set1_epi64x( CB0 ) ); \
V9 = _mm256_xor_si256( S1, _mm256_set1_epi64x( CB1 ) ); \
VA = _mm256_xor_si256( S2, _mm256_set1_epi64x( CB2 ) ); \
VB = _mm256_xor_si256( S3, _mm256_set1_epi64x( CB3 ) ); \
VC = _mm256_xor_si256( _mm256_set1_epi64x( T0 ), \
_mm256_set1_epi64x( CB4 ) ); \
VD = _mm256_xor_si256( _mm256_set1_epi64x( T0 ), \
_mm256_set1_epi64x( CB5 ) ); \
VE = _mm256_xor_si256( _mm256_set1_epi64x( T1 ), \
_mm256_set1_epi64x( CB6 ) ); \
VF = _mm256_xor_si256( _mm256_set1_epi64x( T1 ), \
_mm256_set1_epi64x( CB7, CB7, CB7, CB7 ) ); \
M[0x0] = _mm256_shuffle_epi8( *(buf+ 0), shuff_bswap64 ); \
M[0x1] = _mm256_shuffle_epi8( *(buf+ 1), shuff_bswap64 ); \
M[0x2] = _mm256_shuffle_epi8( *(buf+ 2), shuff_bswap64 ); \
M[0x3] = _mm256_shuffle_epi8( *(buf+ 3), shuff_bswap64 ); \
M[0x4] = _mm256_shuffle_epi8( *(buf+ 4), shuff_bswap64 ); \
M[0x5] = _mm256_shuffle_epi8( *(buf+ 5), shuff_bswap64 ); \
M[0x6] = _mm256_shuffle_epi8( *(buf+ 6), shuff_bswap64 ); \
M[0x7] = _mm256_shuffle_epi8( *(buf+ 7), shuff_bswap64 ); \
M[0x8] = _mm256_shuffle_epi8( *(buf+ 8), shuff_bswap64 ); \
M[0x9] = _mm256_shuffle_epi8( *(buf+ 9), shuff_bswap64 ); \
M[0xA] = _mm256_shuffle_epi8( *(buf+10), shuff_bswap64 ); \
M[0xB] = _mm256_shuffle_epi8( *(buf+11), shuff_bswap64 ); \
M[0xC] = _mm256_shuffle_epi8( *(buf+12), shuff_bswap64 ); \
M[0xD] = _mm256_shuffle_epi8( *(buf+13), shuff_bswap64 ); \
M[0xE] = _mm256_shuffle_epi8( *(buf+14), shuff_bswap64 ); \
M[0xF] = _mm256_shuffle_epi8( *(buf+15), shuff_bswap64 ); \
for (r = 0; r < 16; r ++) \
ROUND_B_4WAY(r); \
H0 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S0, V0 ), V8 ), H0 ); \
H1 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S1, V1 ), V9 ), H1 ); \
H2 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S2, V2 ), VA ), H2 ); \
H3 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S3, V3 ), VB ), H3 ); \
H4 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S0, V4 ), VC ), H4 ); \
H5 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S1, V5 ), VD ), H5 ); \
H6 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S2, V6 ), VE ), H6 ); \
H7 = _mm256_xor_si256( _mm256_xor_si256( \
_mm256_xor_si256( S3, V7 ), VF ), H7 ); \
} while (0)
#else
//current impl
#define COMPRESS64_4WAY do \
{ \
__m256i M0, M1, M2, M3, M4, M5, M6, M7; \
@@ -493,7 +654,8 @@ static const sph_u64 CB[16] = {
m256_const1_64( CB6 ) ); \
VF = _mm256_xor_si256( _mm256_set1_epi64x( T1 ), \
m256_const1_64( CB7 ) ); \
shuf_bswap64 = m256_const2_64( 0x08090a0b0c0d0e0f, 0x0001020304050607 ); \
shuf_bswap64 = m256_const_64( 0x18191a1b1c1d1e1f, 0x1011121314151617, \
0x08090a0b0c0d0e0f, 0x0001020304050607 ); \
M0 = _mm256_shuffle_epi8( *(buf+ 0), shuf_bswap64 ); \
M1 = _mm256_shuffle_epi8( *(buf+ 1), shuf_bswap64 ); \
M2 = _mm256_shuffle_epi8( *(buf+ 2), shuf_bswap64 ); \
@@ -536,9 +698,7 @@ static const sph_u64 CB[16] = {
H7 = mm256_xor4( VF, V7, S3, H7 ); \
} while (0)
#endif
static const sph_u64 salt_zero_big[4] = { 0, 0, 0, 0 };
//static const sph_u64 salt_zero_big[4] = { 0, 0, 0, 0 };
static void
blake64_4way_init( blake_4way_big_context *sc, const sph_u64 *iv,
@@ -583,7 +743,7 @@ blake64_4way( blake_4way_big_context *sc, const void *data, size_t len)
return;
}
READ_STATE64_4WAY(sc);
READ_STATE64(sc);
while ( len > 0 )
{
size_t clen;
@@ -603,7 +763,7 @@ blake64_4way( blake_4way_big_context *sc, const void *data, size_t len)
ptr = 0;
}
}
WRITE_STATE64_4WAY(sc);
WRITE_STATE64(sc);
sc->ptr = ptr;
}
@@ -674,7 +834,7 @@ blake512_4way_init(void *cc)
}
void
blake512_4way(void *cc, const void *data, size_t len)
blake512_4way_update(void *cc, const void *data, size_t len)
{
blake64_4way(cc, data, len);
}