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

v3.9.6.2

Browse Source
This commit is contained in:
Jay D Dee
2019-07-30 10:16:43 -04:00
parent a51f59086b
commit 9d49e0be7a
66 changed files with 1949 additions and 1470 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
algo/sha/sha2-hash-4way.h → algo/sha/sha-hash-4way.h
View File

@@ -55,32 +55,13 @@ typedef struct {
__m128i buf[64>>2];
__m128i val[8];
uint32_t count_high, count_low;
bool initialized;
} sha256_4way_context;
void sha256_4way_init( sha256_4way_context *sc );
void sha256_4way( sha256_4way_context *sc, const void *data, size_t len );
void sha256_4way_close( sha256_4way_context *sc, void *dst );
/*
// SHA-256 7 way hybrid
// Combines SSE, MMX and scalar data to do 8 + 2 + 1 parallel.
typedef struct {
__m128i bufx[64>>2];
__m128i valx[8];
__m64 bufy[64>>2];
__m64 valy[8];
uint32_t bufz[64>>2];
uint32_t valz[8];
uint32_t count_high, count_low;
} sha256_7way_context;
void sha256_7way_init( sha256_7way_context *ctx );
void sha256_7way( sha256_7way_context *ctx, const void *datax,
void *datay, void *dataz, size_t len );
void sha256_7way_close( sha256_7way_context *ctx, void *dstx, void *dstyx,
void *dstz );
*/
#if defined (__AVX2__)
// SHA-256 8 way
@@ -89,6 +70,7 @@ typedef struct {
__m256i buf[64>>2];
__m256i val[8];
uint32_t count_high, count_low;
bool initialized;
} sha256_8way_context;
void sha256_8way_init( sha256_8way_context *sc );
@@ -103,6 +85,7 @@ typedef struct {
__m256i buf[128>>3];
__m256i val[8];
uint64_t count;
bool initialized;
} sha512_4way_context;
void sha512_4way_init( sha512_4way_context *sc);

53
algo/sha/sha2.c
View File

@@ -12,6 +12,7 @@
#include <string.h>
#include <inttypes.h>
#include <openssl/sha.h>
#if defined(USE_ASM) && defined(__arm__) && defined(__APCS_32__)
#define EXTERN_SHA256
@@ -197,7 +198,17 @@ static void sha256d_80_swap(uint32_t *hash, const uint32_t *data)
extern void sha256d(unsigned char *hash, const unsigned char *data, int len)
{
uint32_t S[16], T[16];
#if defined(__SHA__)
SHA256_CTX ctx;
SHA256_Init( &ctx );
SHA256_Update( &ctx, data, len );
SHA256_Final( (unsigned char*)hash, &ctx );
SHA256_Init( &ctx );
SHA256_Update( &ctx, hash, 32 );
SHA256_Final( (unsigned char*)hash, &ctx );
#else
uint32_t S[16], T[16];
int i, r;
sha256_init(S);
@@ -218,6 +229,7 @@ extern void sha256d(unsigned char *hash, const unsigned char *data, int len)
sha256_transform(T, S, 0);
for (i = 0; i < 8; i++)
be32enc((uint32_t *)hash + i, T[i]);
#endif
}
static inline void sha256d_preextend(uint32_t *W)
@@ -635,9 +647,46 @@ int scanhash_sha256d( struct work *work,
return 0;
}
int scanhash_SHA256d( struct work *work, const uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t _ALIGN(128) hash[8];
uint32_t _ALIGN(64) data[20];
uint32_t *pdata = work->data;
const uint32_t *ptarget = work->target;
uint32_t n = pdata[19] - 1;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
int thr_id = mythr->id; // thr_id arg is deprecated
memcpy( data, pdata, 80 );
do {
data[19] = ++n;
sha256d( (unsigned char*)hash, (const unsigned char*)data, 80 );
if ( unlikely( swab32( hash[7] ) <= Htarg ) )
{
pdata[19] = n;
sha256d_80_swap(hash, pdata);
if ( fulltest( hash, ptarget ) && !opt_benchmark )
submit_solution( work, hash, mythr );
}
} while ( likely( n < max_nonce && !work_restart[thr_id].restart ) );
*hashes_done = n - first_nonce + 1;
pdata[19] = n;
return 0;
}
bool register_sha256d_algo( algo_gate_t* gate )
{
gate->scanhash = (void*)&scanhash_sha256d;
#if defined(__SHA__)
gate->optimizations = SHA_OPT;
gate->scanhash = (void*)&scanhash_SHA256d;
#else
gate->optimizations = SSE2_OPT | AVX2_OPT;
gate->scanhash = (void*)&scanhash_sha256d;
#endif
gate->hash = (void*)&sha256d;
return true;
};

365
algo/sha/sha2-hash-4way.c → algo/sha/sha256-hash-4way.c
View File

@@ -34,19 +34,18 @@
#include <stddef.h>
#include <string.h>
#include "sha2-hash-4way.h"
#include <stdio.h>
#include "sha-hash-4way.h"
// SHA-256 32 bit
/*
static const sph_u32 H256[8] = {
SPH_C32(0x6A09E667), SPH_C32(0xBB67AE85),
SPH_C32(0x3C6EF372), SPH_C32(0xA54FF53A),
SPH_C32(0x510E527F), SPH_C32(0x9B05688C),
SPH_C32(0x1F83D9AB), SPH_C32(0x5BE0CD19)
};
*/
static const sph_u32 K256[64] = {
SPH_C32(0x428A2F98), SPH_C32(0x71374491),
@@ -113,16 +112,17 @@ static const sph_u32 K256[64] = {
#define SHA2s_4WAY_STEP(A, B, C, D, E, F, G, H, i, j) \
do { \
register __m128i T1, T2; \
__m128i T1, T2; \
__m128i K = _mm_set1_epi32( K256[( (j)+(i) )] ); \
T1 = _mm_add_epi32( H, mm128_add4_32( BSG2_1(E), CHs(E, F, G), \
_mm_set1_epi32( K256[( (j)+(i) )] ), W[i] ) ); \
K, W[i] ) ); \
T2 = _mm_add_epi32( BSG2_0(A), MAJs(A, B, C) ); \
D = _mm_add_epi32( D, T1 ); \
H = _mm_add_epi32( T1, T2 ); \
} while (0)
static void
sha256_4way_round( __m128i *in, __m128i r[8] )
sha256_4way_round( sha256_4way_context *ctx, __m128i *in, __m128i r[8] )
{
register __m128i A, B, C, D, E, F, G, H;
__m128i W[16];
@@ -130,14 +130,28 @@ sha256_4way_round( __m128i *in, __m128i r[8] )
mm128_block_bswap_32( W, in );
mm128_block_bswap_32( W+8, in+8 );
A = r[0];
B = r[1];
C = r[2];
D = r[3];
E = r[4];
F = r[5];
G = r[6];
H = r[7];
if ( ctx->initialized )
{
A = r[0];
B = r[1];
C = r[2];
D = r[3];
E = r[4];
F = r[5];
G = r[6];
H = r[7];
}
else
{
A = m128_const1_64( 0x6A09E6676A09E667 );
B = m128_const1_64( 0xBB67AE85BB67AE85 );
C = m128_const1_64( 0x3C6EF3723C6EF372 );
D = m128_const1_64( 0xA54FF53AA54FF53A );
E = m128_const1_64( 0x510E527F510E527F );
F = m128_const1_64( 0x9B05688C9B05688C );
G = m128_const1_64( 0x1F83D9AB1F83D9AB );
H = m128_const1_64( 0x5BE0CD195BE0CD19 );
}
SHA2s_4WAY_STEP( A, B, C, D, E, F, G, H, 0, 0 );
SHA2s_4WAY_STEP( H, A, B, C, D, E, F, G, 1, 0 );
@@ -193,19 +207,36 @@ sha256_4way_round( __m128i *in, __m128i r[8] )
SHA2s_4WAY_STEP( B, C, D, E, F, G, H, A, 15, j );
}
r[0] = _mm_add_epi32( r[0], A );
r[1] = _mm_add_epi32( r[1], B );
r[2] = _mm_add_epi32( r[2], C );
r[3] = _mm_add_epi32( r[3], D );
r[4] = _mm_add_epi32( r[4], E );
r[5] = _mm_add_epi32( r[5], F );
r[6] = _mm_add_epi32( r[6], G );
r[7] = _mm_add_epi32( r[7], H );
if ( ctx->initialized )
{
r[0] = _mm_add_epi32( r[0], A );
r[1] = _mm_add_epi32( r[1], B );
r[2] = _mm_add_epi32( r[2], C );
r[3] = _mm_add_epi32( r[3], D );
r[4] = _mm_add_epi32( r[4], E );
r[5] = _mm_add_epi32( r[5], F );
r[6] = _mm_add_epi32( r[6], G );
r[7] = _mm_add_epi32( r[7], H );
}
else
{
ctx->initialized = true;
r[0] = _mm_add_epi32( A, m128_const1_64( 0x6A09E6676A09E667 ) );
r[1] = _mm_add_epi32( B, m128_const1_64( 0xBB67AE85BB67AE85 ) );
r[2] = _mm_add_epi32( C, m128_const1_64( 0x3C6EF3723C6EF372 ) );
r[3] = _mm_add_epi32( D, m128_const1_64( 0xA54FF53AA54FF53A ) );
r[4] = _mm_add_epi32( E, m128_const1_64( 0x510E527F510E527F ) );
r[5] = _mm_add_epi32( F, m128_const1_64( 0x9B05688C9B05688C ) );
r[6] = _mm_add_epi32( G, m128_const1_64( 0x1F83D9AB1F83D9AB ) );
r[7] = _mm_add_epi32( H, m128_const1_64( 0x5BE0CD195BE0CD19 ) );
}
}
void sha256_4way_init( sha256_4way_context *sc )
{
sc->initialized = false;
sc->count_high = sc->count_low = 0;
/*
sc->val[0] = _mm_set1_epi32( H256[0] );
sc->val[1] = _mm_set1_epi32( H256[1] );
sc->val[2] = _mm_set1_epi32( H256[2] );
@@ -214,6 +245,7 @@ void sha256_4way_init( sha256_4way_context *sc )
sc->val[5] = _mm_set1_epi32( H256[5] );
sc->val[6] = _mm_set1_epi32( H256[6] );
sc->val[7] = _mm_set1_epi32( H256[7] );
*/
}
void sha256_4way( sha256_4way_context *sc, const void *data, size_t len )
@@ -237,7 +269,7 @@ void sha256_4way( sha256_4way_context *sc, const void *data, size_t len )
len -= clen;
if ( ptr == buf_size )
{
sha256_4way_round( sc->buf, sc->val );
sha256_4way_round( sc, sc->buf, sc->val );
ptr = 0;
}
clow = sc->count_low;
@@ -256,13 +288,13 @@ void sha256_4way_close( sha256_4way_context *sc, void *dst )
const int pad = buf_size - 8;
ptr = (unsigned)sc->count_low & (buf_size - 1U);
sc->buf[ ptr>>2 ] = _mm_set1_epi32( 0x80 );
sc->buf[ ptr>>2 ] = m128_const1_64( 0x0000008000000080 );
ptr += 4;
if ( ptr > pad )
{
memset_zero_128( sc->buf + (ptr>>2), (buf_size - ptr) >> 2 );
sha256_4way_round( sc->buf, sc->val );
sha256_4way_round( sc, sc->buf, sc->val );
memset_zero_128( sc->buf, pad >> 2 );
}
else
@@ -276,7 +308,7 @@ void sha256_4way_close( sha256_4way_context *sc, void *dst )
mm128_bswap_32( _mm_set1_epi32( high ) );
sc->buf[ ( pad+4 ) >> 2 ] =
mm128_bswap_32( _mm_set1_epi32( low ) );
sha256_4way_round( sc->buf, sc->val );
sha256_4way_round( sc, sc->buf, sc->val );
mm128_block_bswap_32( dst, sc->val );
}
@@ -313,16 +345,17 @@ void sha256_4way_close( sha256_4way_context *sc, void *dst )
#define SHA2s_8WAY_STEP(A, B, C, D, E, F, G, H, i, j) \
do { \
register __m256i T1, T2; \
T1 = _mm256_add_epi32( H, mm256_add4_32( BSG2_1x(E), CHx(E, F, G), \
_mm256_set1_epi32( K256[( (j)+(i) )] ), W[i] ) ); \
__m256i T1, T2; \
__m256i K = _mm256_set1_epi32( K256[( (j)+(i) )] ); \
T1 = _mm256_add_epi32( H, mm256_add4_32( BSG2_1x(E), CHx(E, F, G), \
K, W[i] ) ); \
T2 = _mm256_add_epi32( BSG2_0x(A), MAJx(A, B, C) ); \
D = _mm256_add_epi32( D, T1 ); \
H = _mm256_add_epi32( T1, T2 ); \
} while (0)
static void
sha256_8way_round( __m256i *in, __m256i r[8] )
sha256_8way_round( sha256_8way_context *ctx, __m256i *in, __m256i r[8] )
{
register __m256i A, B, C, D, E, F, G, H;
__m256i W[16];
@@ -330,14 +363,28 @@ sha256_8way_round( __m256i *in, __m256i r[8] )
mm256_block_bswap_32( W , in );
mm256_block_bswap_32( W+8, in+8 );
A = r[0];
B = r[1];
C = r[2];
D = r[3];
E = r[4];
F = r[5];
G = r[6];
H = r[7];
if ( ctx->initialized )
{
A = r[0];
B = r[1];
C = r[2];
D = r[3];
E = r[4];
F = r[5];
G = r[6];
H = r[7];
}
else
{
A = m256_const1_64( 0x6A09E6676A09E667 );
B = m256_const1_64( 0xBB67AE85BB67AE85 );
C = m256_const1_64( 0x3C6EF3723C6EF372 );
D = m256_const1_64( 0xA54FF53AA54FF53A );
E = m256_const1_64( 0x510E527F510E527F );
F = m256_const1_64( 0x9B05688C9B05688C );
G = m256_const1_64( 0x1F83D9AB1F83D9AB );
H = m256_const1_64( 0x5BE0CD195BE0CD19 );
}
SHA2s_8WAY_STEP( A, B, C, D, E, F, G, H, 0, 0 );
SHA2s_8WAY_STEP( H, A, B, C, D, E, F, G, 1, 0 );
@@ -393,20 +440,36 @@ sha256_8way_round( __m256i *in, __m256i r[8] )
SHA2s_8WAY_STEP( B, C, D, E, F, G, H, A, 15, j );
}
r[0] = _mm256_add_epi32( r[0], A );
r[1] = _mm256_add_epi32( r[1], B );
r[2] = _mm256_add_epi32( r[2], C );
r[3] = _mm256_add_epi32( r[3], D );
r[4] = _mm256_add_epi32( r[4], E );
r[5] = _mm256_add_epi32( r[5], F );
r[6] = _mm256_add_epi32( r[6], G );
r[7] = _mm256_add_epi32( r[7], H );
if ( ctx->initialized )
{
r[0] = _mm256_add_epi32( r[0], A );
r[1] = _mm256_add_epi32( r[1], B );
r[2] = _mm256_add_epi32( r[2], C );
r[3] = _mm256_add_epi32( r[3], D );
r[4] = _mm256_add_epi32( r[4], E );
r[5] = _mm256_add_epi32( r[5], F );
r[6] = _mm256_add_epi32( r[6], G );
r[7] = _mm256_add_epi32( r[7], H );
}
else
{
ctx->initialized = true;
r[0] = _mm256_add_epi32( A, m256_const1_64( 0x6A09E6676A09E667 ) );
r[1] = _mm256_add_epi32( B, m256_const1_64( 0xBB67AE85BB67AE85 ) );
r[2] = _mm256_add_epi32( C, m256_const1_64( 0x3C6EF3723C6EF372 ) );
r[3] = _mm256_add_epi32( D, m256_const1_64( 0xA54FF53AA54FF53A ) );
r[4] = _mm256_add_epi32( E, m256_const1_64( 0x510E527F510E527F ) );
r[5] = _mm256_add_epi32( F, m256_const1_64( 0x9B05688C9B05688C ) );
r[6] = _mm256_add_epi32( G, m256_const1_64( 0x1F83D9AB1F83D9AB ) );
r[7] = _mm256_add_epi32( H, m256_const1_64( 0x5BE0CD195BE0CD19 ) );
}
}
void sha256_8way_init( sha256_8way_context *sc )
{
sc->initialized = false;
sc->count_high = sc->count_low = 0;
/*
sc->val[0] = _mm256_set1_epi32( H256[0] );
sc->val[1] = _mm256_set1_epi32( H256[1] );
sc->val[2] = _mm256_set1_epi32( H256[2] );
@@ -415,6 +478,7 @@ void sha256_8way_init( sha256_8way_context *sc )
sc->val[5] = _mm256_set1_epi32( H256[5] );
sc->val[6] = _mm256_set1_epi32( H256[6] );
sc->val[7] = _mm256_set1_epi32( H256[7] );
*/
}
void sha256_8way( sha256_8way_context *sc, const void *data, size_t len )
@@ -438,7 +502,7 @@ void sha256_8way( sha256_8way_context *sc, const void *data, size_t len )
len -= clen;
if ( ptr == buf_size )
{
sha256_8way_round( sc->buf, sc->val );
sha256_8way_round( sc, sc->buf, sc->val );
ptr = 0;
}
clow = sc->count_low;
@@ -457,13 +521,13 @@ void sha256_8way_close( sha256_8way_context *sc, void *dst )
const int pad = buf_size - 8;
ptr = (unsigned)sc->count_low & (buf_size - 1U);
sc->buf[ ptr>>2 ] = _mm256_set1_epi32( 0x80 );
sc->buf[ ptr>>2 ] = m256_const1_64( 0x0000008000000080 );
ptr += 4;
if ( ptr > pad )
{
memset_zero_256( sc->buf + (ptr>>2), (buf_size - ptr) >> 2 );
sha256_8way_round( sc->buf, sc->val );
sha256_8way_round( sc, sc->buf, sc->val );
memset_zero_256( sc->buf, pad >> 2 );
}
else
@@ -478,207 +542,10 @@ void sha256_8way_close( sha256_8way_context *sc, void *dst )
sc->buf[ ( pad+4 ) >> 2 ] =
mm256_bswap_32( _mm256_set1_epi32( low ) );
sha256_8way_round( sc->buf, sc->val );
sha256_8way_round( sc, sc->buf, sc->val );
mm256_block_bswap_32( dst, sc->val );
}
// SHA-512 4 way 64 bit
static const sph_u64 H512[8] = {
SPH_C64(0x6A09E667F3BCC908), SPH_C64(0xBB67AE8584CAA73B),
SPH_C64(0x3C6EF372FE94F82B), SPH_C64(0xA54FF53A5F1D36F1),
SPH_C64(0x510E527FADE682D1), SPH_C64(0x9B05688C2B3E6C1F),
SPH_C64(0x1F83D9ABFB41BD6B), SPH_C64(0x5BE0CD19137E2179)
};
static const sph_u64 K512[80] = {
SPH_C64(0x428A2F98D728AE22), SPH_C64(0x7137449123EF65CD),
SPH_C64(0xB5C0FBCFEC4D3B2F), SPH_C64(0xE9B5DBA58189DBBC),
SPH_C64(0x3956C25BF348B538), SPH_C64(0x59F111F1B605D019),
SPH_C64(0x923F82A4AF194F9B), SPH_C64(0xAB1C5ED5DA6D8118),
SPH_C64(0xD807AA98A3030242), SPH_C64(0x12835B0145706FBE),
SPH_C64(0x243185BE4EE4B28C), SPH_C64(0x550C7DC3D5FFB4E2),
SPH_C64(0x72BE5D74F27B896F), SPH_C64(0x80DEB1FE3B1696B1),
SPH_C64(0x9BDC06A725C71235), SPH_C64(0xC19BF174CF692694),
SPH_C64(0xE49B69C19EF14AD2), SPH_C64(0xEFBE4786384F25E3),
SPH_C64(0x0FC19DC68B8CD5B5), SPH_C64(0x240CA1CC77AC9C65),
SPH_C64(0x2DE92C6F592B0275), SPH_C64(0x4A7484AA6EA6E483),
SPH_C64(0x5CB0A9DCBD41FBD4), SPH_C64(0x76F988DA831153B5),
SPH_C64(0x983E5152EE66DFAB), SPH_C64(0xA831C66D2DB43210),
SPH_C64(0xB00327C898FB213F), SPH_C64(0xBF597FC7BEEF0EE4),
SPH_C64(0xC6E00BF33DA88FC2), SPH_C64(0xD5A79147930AA725),
SPH_C64(0x06CA6351E003826F), SPH_C64(0x142929670A0E6E70),
SPH_C64(0x27B70A8546D22FFC), SPH_C64(0x2E1B21385C26C926),
SPH_C64(0x4D2C6DFC5AC42AED), SPH_C64(0x53380D139D95B3DF),
SPH_C64(0x650A73548BAF63DE), SPH_C64(0x766A0ABB3C77B2A8),
SPH_C64(0x81C2C92E47EDAEE6), SPH_C64(0x92722C851482353B),
SPH_C64(0xA2BFE8A14CF10364), SPH_C64(0xA81A664BBC423001),
SPH_C64(0xC24B8B70D0F89791), SPH_C64(0xC76C51A30654BE30),
SPH_C64(0xD192E819D6EF5218), SPH_C64(0xD69906245565A910),
SPH_C64(0xF40E35855771202A), SPH_C64(0x106AA07032BBD1B8),
SPH_C64(0x19A4C116B8D2D0C8), SPH_C64(0x1E376C085141AB53),
SPH_C64(0x2748774CDF8EEB99), SPH_C64(0x34B0BCB5E19B48A8),
SPH_C64(0x391C0CB3C5C95A63), SPH_C64(0x4ED8AA4AE3418ACB),
SPH_C64(0x5B9CCA4F7763E373), SPH_C64(0x682E6FF3D6B2B8A3),
SPH_C64(0x748F82EE5DEFB2FC), SPH_C64(0x78A5636F43172F60),
SPH_C64(0x84C87814A1F0AB72), SPH_C64(0x8CC702081A6439EC),
SPH_C64(0x90BEFFFA23631E28), SPH_C64(0xA4506CEBDE82BDE9),
SPH_C64(0xBEF9A3F7B2C67915), SPH_C64(0xC67178F2E372532B),
SPH_C64(0xCA273ECEEA26619C), SPH_C64(0xD186B8C721C0C207),
SPH_C64(0xEADA7DD6CDE0EB1E), SPH_C64(0xF57D4F7FEE6ED178),
SPH_C64(0x06F067AA72176FBA), SPH_C64(0x0A637DC5A2C898A6),
SPH_C64(0x113F9804BEF90DAE), SPH_C64(0x1B710B35131C471B),
SPH_C64(0x28DB77F523047D84), SPH_C64(0x32CAAB7B40C72493),
SPH_C64(0x3C9EBE0A15C9BEBC), SPH_C64(0x431D67C49C100D4C),
SPH_C64(0x4CC5D4BECB3E42B6), SPH_C64(0x597F299CFC657E2A),
SPH_C64(0x5FCB6FAB3AD6FAEC), SPH_C64(0x6C44198C4A475817)
};
#define CH(X, Y, Z) \
_mm256_xor_si256( _mm256_and_si256( _mm256_xor_si256( Y, Z ), X ), Z )
#define MAJ(X, Y, Z) \
_mm256_or_si256( _mm256_and_si256( X, Y ), \
_mm256_and_si256( _mm256_or_si256( X, Y ), Z ) )
#define BSG5_0(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_ror_64(x, 28), mm256_ror_64(x, 34) ), mm256_ror_64(x, 39) )
#define BSG5_1(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_ror_64(x, 14), mm256_ror_64(x, 18) ), mm256_ror_64(x, 41) )
#define SSG5_0(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_ror_64(x, 1), mm256_ror_64(x, 8) ), _mm256_srli_epi64(x, 7) )
#define SSG5_1(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_ror_64(x, 19), mm256_ror_64(x, 61) ), _mm256_srli_epi64(x, 6) )
#define SHA3_4WAY_STEP(A, B, C, D, E, F, G, H, i) \
do { \
register __m256i T1, T2; \
T1 = _mm256_add_epi64( H, mm256_add4_64( BSG5_1(E), CH(E, F, G), \
_mm256_set1_epi64x( K512[i] ), W[i] ) ); \
T2 = _mm256_add_epi64( BSG5_0(A), MAJ(A, B, C) ); \
D = _mm256_add_epi64( D, T1 ); \
H = _mm256_add_epi64( T1, T2 ); \
} while (0)
static void
sha512_4way_round( __m256i *in, __m256i r[8] )
{
int i;
register __m256i A, B, C, D, E, F, G, H;
__m256i W[80];
mm256_block_bswap_64( W , in );
mm256_block_bswap_64( W+8, in+8 );
for ( i = 16; i < 80; i++ )
W[i] = mm256_add4_64( SSG5_1( W[ i- 2 ] ), W[ i- 7 ],
SSG5_0( W[ i-15 ] ), W[ i-16 ] );
A = r[0];
B = r[1];
C = r[2];
D = r[3];
E = r[4];
F = r[5];
G = r[6];
H = r[7];
for ( i = 0; i < 80; i += 8 )
{
SHA3_4WAY_STEP( A, B, C, D, E, F, G, H, i + 0 );
SHA3_4WAY_STEP( H, A, B, C, D, E, F, G, i + 1 );
SHA3_4WAY_STEP( G, H, A, B, C, D, E, F, i + 2 );
SHA3_4WAY_STEP( F, G, H, A, B, C, D, E, i + 3 );
SHA3_4WAY_STEP( E, F, G, H, A, B, C, D, i + 4 );
SHA3_4WAY_STEP( D, E, F, G, H, A, B, C, i + 5 );
SHA3_4WAY_STEP( C, D, E, F, G, H, A, B, i + 6 );
SHA3_4WAY_STEP( B, C, D, E, F, G, H, A, i + 7 );
}
r[0] = _mm256_add_epi64( r[0], A );
r[1] = _mm256_add_epi64( r[1], B );
r[2] = _mm256_add_epi64( r[2], C );
r[3] = _mm256_add_epi64( r[3], D );
r[4] = _mm256_add_epi64( r[4], E );
r[5] = _mm256_add_epi64( r[5], F );
r[6] = _mm256_add_epi64( r[6], G );
r[7] = _mm256_add_epi64( r[7], H );
}
void sha512_4way_init( sha512_4way_context *sc )
{
sc->count = 0;
sc->val[0] = _mm256_set1_epi64x( H512[0] );
sc->val[1] = _mm256_set1_epi64x( H512[1] );
sc->val[2] = _mm256_set1_epi64x( H512[2] );
sc->val[3] = _mm256_set1_epi64x( H512[3] );
sc->val[4] = _mm256_set1_epi64x( H512[4] );
sc->val[5] = _mm256_set1_epi64x( H512[5] );
sc->val[6] = _mm256_set1_epi64x( H512[6] );
sc->val[7] = _mm256_set1_epi64x( H512[7] );
}
void sha512_4way( sha512_4way_context *sc, const void *data, size_t len )
{
__m256i *vdata = (__m256i*)data;
size_t ptr;
const int buf_size = 128;
ptr = (unsigned)sc->count & (buf_size - 1U);
while ( len > 0 )
{
size_t clen;
clen = buf_size - ptr;
if ( clen > len )
clen = len;
memcpy_256( sc->buf + (ptr>>3), vdata, clen>>3 );
vdata = vdata + (clen>>3);
ptr += clen;
len -= clen;
if ( ptr == buf_size )
{
sha512_4way_round( sc->buf, sc->val );
ptr = 0;
}
sc->count += clen;
}
}
void sha512_4way_close( sha512_4way_context *sc, void *dst )
{
unsigned ptr;
const int buf_size = 128;
const int pad = buf_size - 16;
ptr = (unsigned)sc->count & (buf_size - 1U);
sc->buf[ ptr>>3 ] = m256_const1_64( 0x80 );
ptr += 8;
if ( ptr > pad )
{
memset_zero_256( sc->buf + (ptr>>3), (buf_size - ptr) >> 3 );
sha512_4way_round( sc->buf, sc->val );
memset_zero_256( sc->buf, pad >> 3 );
}
else
memset_zero_256( sc->buf + (ptr>>3), (pad - ptr) >> 3 );
sc->buf[ pad >> 3 ] =
mm256_bswap_64( _mm256_set1_epi64x( sc->count >> 61 ) );
sc->buf[ ( pad+8 ) >> 3 ] =
mm256_bswap_64( _mm256_set1_epi64x( sc->count << 3 ) );
sha512_4way_round( sc->buf, sc->val );
mm256_block_bswap_64( dst, sc->val );
}
#endif // __AVX2__
#endif // __SSE2__

2
algo/sha/sha256q-4way.c
View File

@@ -3,7 +3,7 @@
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "sha2-hash-4way.h"
#include "sha-hash-4way.h"
#if defined(SHA256T_8WAY)

18
algo/sha/sha256t-4way.c
View File

@@ -3,7 +3,7 @@
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "sha2-hash-4way.h"
#include "sha-hash-4way.h"
#if defined(SHA256T_11WAY)
@@ -158,7 +158,7 @@ void sha256t_8way_hash( void* output, const void* input )
sha256_8way_close( &ctx, output );
}
int scanhash_sha256t_8way( struct work *work, uint32_t max_nonce,
int scanhash_sha256t_8way( struct work *work, const uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
@@ -166,12 +166,12 @@ int scanhash_sha256t_8way( struct work *work, uint32_t max_nonce,
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *hash7 = &(hash[7<<3]);
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t *ptarget = work->target;
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
__m256i *noncev = (__m256i*)vdata + 19; // aligned
int thr_id = mythr->id; // thr_id arg is deprecated
const int thr_id = mythr->id;
const uint64_t htmax[] = { 0,
0xF,
@@ -194,7 +194,7 @@ int scanhash_sha256t_8way( struct work *work, uint32_t max_nonce,
for ( int m = 0; m < 6; m++ ) if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
const uint32_t mask = masks[m];
do
{
*noncev = mm256_bswap_32( _mm256_set_epi32(
@@ -244,7 +244,7 @@ void sha256t_4way_hash( void* output, const void* input )
sha256_4way_close( &ctx, output );
}
int scanhash_sha256t_4way( struct work *work, uint32_t max_nonce,
int scanhash_sha256t_4way( struct work *work, const uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
@@ -252,12 +252,12 @@ int scanhash_sha256t_4way( struct work *work, uint32_t max_nonce,
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t *ptarget = work->target;
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
__m128i *noncev = (__m128i*)vdata + 19; // aligned
int thr_id = mythr->id; // thr_id arg is deprecated
const int thr_id = mythr->id;
const uint64_t htmax[] = { 0,
0xF,
@@ -278,7 +278,7 @@ int scanhash_sha256t_4way( struct work *work, uint32_t max_nonce,
for ( int m = 0; m < 6; m++ ) if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
const uint32_t mask = masks[m];
do {
*noncev = mm128_bswap_32( _mm_set_epi32( n+3,n+2,n+1,n ) );
pdata[19] = n;

320
algo/sha/sha512-hash-4way.c Normal file
View File

@@ -0,0 +1,320 @@
/* $Id: sha2big.c 216 2010-06-08 09:46:57Z tp $ */
/*
* SHA-384 / SHA-512 implementation.
*
* ==========================(LICENSE BEGIN)============================
*
* Copyright (c) 2007-2010 Projet RNRT SAPHIR
*
* Permission is hereby granted, free of charge, to any person obtaining
* a copy of this software and associated documentation files (the
* "Software"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sublicense, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
*
* The above copyright notice and this permission notice shall be
* included in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* ===========================(LICENSE END)=============================
*
* @author Thomas Pornin <thomas.pornin@cryptolog.com>
*/
#if defined(__AVX2__)
#include <stddef.h>
#include <string.h>
#include "sha-hash-4way.h"
// SHA-512 4 way 64 bit
/*
static const sph_u64 H512[8] = {
SPH_C64(0x6A09E667F3BCC908), SPH_C64(0xBB67AE8584CAA73B),
SPH_C64(0x3C6EF372FE94F82B), SPH_C64(0xA54FF53A5F1D36F1),
SPH_C64(0x510E527FADE682D1), SPH_C64(0x9B05688C2B3E6C1F),
SPH_C64(0x1F83D9ABFB41BD6B), SPH_C64(0x5BE0CD19137E2179)
};
*/
static const sph_u64 K512[80] = {
SPH_C64(0x428A2F98D728AE22), SPH_C64(0x7137449123EF65CD),
SPH_C64(0xB5C0FBCFEC4D3B2F), SPH_C64(0xE9B5DBA58189DBBC),
SPH_C64(0x3956C25BF348B538), SPH_C64(0x59F111F1B605D019),
SPH_C64(0x923F82A4AF194F9B), SPH_C64(0xAB1C5ED5DA6D8118),
SPH_C64(0xD807AA98A3030242), SPH_C64(0x12835B0145706FBE),
SPH_C64(0x243185BE4EE4B28C), SPH_C64(0x550C7DC3D5FFB4E2),
SPH_C64(0x72BE5D74F27B896F), SPH_C64(0x80DEB1FE3B1696B1),
SPH_C64(0x9BDC06A725C71235), SPH_C64(0xC19BF174CF692694),
SPH_C64(0xE49B69C19EF14AD2), SPH_C64(0xEFBE4786384F25E3),
SPH_C64(0x0FC19DC68B8CD5B5), SPH_C64(0x240CA1CC77AC9C65),
SPH_C64(0x2DE92C6F592B0275), SPH_C64(0x4A7484AA6EA6E483),
SPH_C64(0x5CB0A9DCBD41FBD4), SPH_C64(0x76F988DA831153B5),
SPH_C64(0x983E5152EE66DFAB), SPH_C64(0xA831C66D2DB43210),
SPH_C64(0xB00327C898FB213F), SPH_C64(0xBF597FC7BEEF0EE4),
SPH_C64(0xC6E00BF33DA88FC2), SPH_C64(0xD5A79147930AA725),
SPH_C64(0x06CA6351E003826F), SPH_C64(0x142929670A0E6E70),
SPH_C64(0x27B70A8546D22FFC), SPH_C64(0x2E1B21385C26C926),
SPH_C64(0x4D2C6DFC5AC42AED), SPH_C64(0x53380D139D95B3DF),
SPH_C64(0x650A73548BAF63DE), SPH_C64(0x766A0ABB3C77B2A8),
SPH_C64(0x81C2C92E47EDAEE6), SPH_C64(0x92722C851482353B),
SPH_C64(0xA2BFE8A14CF10364), SPH_C64(0xA81A664BBC423001),
SPH_C64(0xC24B8B70D0F89791), SPH_C64(0xC76C51A30654BE30),
SPH_C64(0xD192E819D6EF5218), SPH_C64(0xD69906245565A910),
SPH_C64(0xF40E35855771202A), SPH_C64(0x106AA07032BBD1B8),
SPH_C64(0x19A4C116B8D2D0C8), SPH_C64(0x1E376C085141AB53),
SPH_C64(0x2748774CDF8EEB99), SPH_C64(0x34B0BCB5E19B48A8),
SPH_C64(0x391C0CB3C5C95A63), SPH_C64(0x4ED8AA4AE3418ACB),
SPH_C64(0x5B9CCA4F7763E373), SPH_C64(0x682E6FF3D6B2B8A3),
SPH_C64(0x748F82EE5DEFB2FC), SPH_C64(0x78A5636F43172F60),
SPH_C64(0x84C87814A1F0AB72), SPH_C64(0x8CC702081A6439EC),
SPH_C64(0x90BEFFFA23631E28), SPH_C64(0xA4506CEBDE82BDE9),
SPH_C64(0xBEF9A3F7B2C67915), SPH_C64(0xC67178F2E372532B),
SPH_C64(0xCA273ECEEA26619C), SPH_C64(0xD186B8C721C0C207),
SPH_C64(0xEADA7DD6CDE0EB1E), SPH_C64(0xF57D4F7FEE6ED178),
SPH_C64(0x06F067AA72176FBA), SPH_C64(0x0A637DC5A2C898A6),
SPH_C64(0x113F9804BEF90DAE), SPH_C64(0x1B710B35131C471B),
SPH_C64(0x28DB77F523047D84), SPH_C64(0x32CAAB7B40C72493),
SPH_C64(0x3C9EBE0A15C9BEBC), SPH_C64(0x431D67C49C100D4C),
SPH_C64(0x4CC5D4BECB3E42B6), SPH_C64(0x597F299CFC657E2A),
SPH_C64(0x5FCB6FAB3AD6FAEC), SPH_C64(0x6C44198C4A475817)
};
#define CH(X, Y, Z) \
_mm256_xor_si256( _mm256_and_si256( _mm256_xor_si256( Y, Z ), X ), Z )
#define MAJ(X, Y, Z) \
_mm256_or_si256( _mm256_and_si256( X, Y ), \
_mm256_and_si256( _mm256_or_si256( X, Y ), Z ) )
#define BSG5_0(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_ror_64(x, 28), mm256_ror_64(x, 34) ), mm256_ror_64(x, 39) )
#define BSG5_1(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_ror_64(x, 14), mm256_ror_64(x, 18) ), mm256_ror_64(x, 41) )
#define SSG5_0(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_ror_64(x, 1), mm256_ror_64(x, 8) ), _mm256_srli_epi64(x, 7) )
#define SSG5_1(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_ror_64(x, 19), mm256_ror_64(x, 61) ), _mm256_srli_epi64(x, 6) )
// Interleave SSG0 & SSG1 for better throughput.
// return ssg0(w0) + ssg1(w1)
static inline __m256i ssg512_add( __m256i w0, __m256i w1 )
{
__m256i w0a, w1a, w0b, w1b;
w0a = mm256_ror_64( w0, 1 );
w1a = mm256_ror_64( w1,19 );
w0b = mm256_ror_64( w0, 8 );
w1b = mm256_ror_64( w1,61 );
w0a = _mm256_xor_si256( w0a, w0b );
w1a = _mm256_xor_si256( w1a, w1b );
w0b = _mm256_srli_epi64( w0, 7 );
w1b = _mm256_srli_epi64( w1, 6 );
w0a = _mm256_xor_si256( w0a, w0b );
w1a = _mm256_xor_si256( w1a, w1b );
return _mm256_add_epi64( w0a, w1a );
}
#define SSG512x2_0( w0, w1, i ) do \
{ \
__m256i X0a, X1a, X0b, X1b; \
X0a = mm256_ror_64( W[i-15], 1 ); \
X1a = mm256_ror_64( W[i-14], 1 ); \
X0b = mm256_ror_64( W[i-15], 8 ); \
X1b = mm256_ror_64( W[i-14], 8 ); \
X0a = _mm256_xor_si256( X0a, X0b ); \
X1a = _mm256_xor_si256( X1a, X1b ); \
X0b = _mm256_srli_epi64( W[i-15], 7 ); \
X1b = _mm256_srli_epi64( W[i-14], 7 ); \
w0 = _mm256_xor_si256( X0a, X0b ); \
w1 = _mm256_xor_si256( X1a, X1b ); \
} while(0)
#define SSG512x2_1( w0, w1, i ) do \
{ \
__m256i X0a, X1a, X0b, X1b; \
X0a = mm256_ror_64( W[i-2],19 ); \
X1a = mm256_ror_64( W[i-1],19 ); \
X0b = mm256_ror_64( W[i-2],61 ); \
X1b = mm256_ror_64( W[i-1],61 ); \
X0a = _mm256_xor_si256( X0a, X0b ); \
X1a = _mm256_xor_si256( X1a, X1b ); \
X0b = _mm256_srli_epi64( W[i-2], 6 ); \
X1b = _mm256_srli_epi64( W[i-1], 6 ); \
w0 = _mm256_xor_si256( X0a, X0b ); \
w1 = _mm256_xor_si256( X1a, X1b ); \
} while(0)
#define SHA3_4WAY_STEP(A, B, C, D, E, F, G, H, i) \
do { \
__m256i T1, T2; \
__m256i K = _mm256_set1_epi64x( K512[ i ] ); \
T1 = _mm256_add_epi64( H, mm256_add4_64( BSG5_1(E), CH(E, F, G), \
K, W[i] ) ); \
T2 = _mm256_add_epi64( BSG5_0(A), MAJ(A, B, C) ); \
D = _mm256_add_epi64( D, T1 ); \
H = _mm256_add_epi64( T1, T2 ); \
} while (0)
static void
sha512_4way_round( sha512_4way_context *ctx, __m256i *in, __m256i r[8] )
{
int i;
register __m256i A, B, C, D, E, F, G, H;
__m256i W[80];
mm256_block_bswap_64( W , in );
mm256_block_bswap_64( W+8, in+8 );
for ( i = 16; i < 80; i++ )
W[i] = _mm256_add_epi64( ssg512_add( W[i-15], W[i-2] ),
_mm256_add_epi64( W[ i- 7 ], W[ i-16 ] ) );
if ( ctx->initialized )
{
A = r[0];
B = r[1];
C = r[2];
D = r[3];
E = r[4];
F = r[5];
G = r[6];
H = r[7];
}
else
{
A = m256_const1_64( 0x6A09E667F3BCC908 );
B = m256_const1_64( 0xBB67AE8584CAA73B );
C = m256_const1_64( 0x3C6EF372FE94F82B );
D = m256_const1_64( 0xA54FF53A5F1D36F1 );
E = m256_const1_64( 0x510E527FADE682D1 );
F = m256_const1_64( 0x9B05688C2B3E6C1F );
G = m256_const1_64( 0x1F83D9ABFB41BD6B );
H = m256_const1_64( 0x5BE0CD19137E2179 );
}
for ( i = 0; i < 80; i += 8 )
{
SHA3_4WAY_STEP( A, B, C, D, E, F, G, H, i + 0 );
SHA3_4WAY_STEP( H, A, B, C, D, E, F, G, i + 1 );
SHA3_4WAY_STEP( G, H, A, B, C, D, E, F, i + 2 );
SHA3_4WAY_STEP( F, G, H, A, B, C, D, E, i + 3 );
SHA3_4WAY_STEP( E, F, G, H, A, B, C, D, i + 4 );
SHA3_4WAY_STEP( D, E, F, G, H, A, B, C, i + 5 );
SHA3_4WAY_STEP( C, D, E, F, G, H, A, B, i + 6 );
SHA3_4WAY_STEP( B, C, D, E, F, G, H, A, i + 7 );
}
if ( ctx->initialized )
{
r[0] = _mm256_add_epi64( r[0], A );
r[1] = _mm256_add_epi64( r[1], B );
r[2] = _mm256_add_epi64( r[2], C );
r[3] = _mm256_add_epi64( r[3], D );
r[4] = _mm256_add_epi64( r[4], E );
r[5] = _mm256_add_epi64( r[5], F );
r[6] = _mm256_add_epi64( r[6], G );
r[7] = _mm256_add_epi64( r[7], H );
}
else
{
ctx->initialized = true;
r[0] = _mm256_add_epi64( A, m256_const1_64( 0x6A09E667F3BCC908 ) );
r[1] = _mm256_add_epi64( B, m256_const1_64( 0xBB67AE8584CAA73B ) );
r[2] = _mm256_add_epi64( C, m256_const1_64( 0x3C6EF372FE94F82B ) );
r[3] = _mm256_add_epi64( D, m256_const1_64( 0xA54FF53A5F1D36F1 ) );
r[4] = _mm256_add_epi64( E, m256_const1_64( 0x510E527FADE682D1 ) );
r[5] = _mm256_add_epi64( F, m256_const1_64( 0x9B05688C2B3E6C1F ) );
r[6] = _mm256_add_epi64( G, m256_const1_64( 0x1F83D9ABFB41BD6B ) );
r[7] = _mm256_add_epi64( H, m256_const1_64( 0x5BE0CD19137E2179 ) );
}
}
void sha512_4way_init( sha512_4way_context *sc )
{
sc->initialized = false;
sc->count = 0;
/*
sc->val[0] = _mm256_set1_epi64x( H512[0] );
sc->val[1] = _mm256_set1_epi64x( H512[1] );
sc->val[2] = _mm256_set1_epi64x( H512[2] );
sc->val[3] = _mm256_set1_epi64x( H512[3] );
sc->val[4] = _mm256_set1_epi64x( H512[4] );
sc->val[5] = _mm256_set1_epi64x( H512[5] );
sc->val[6] = _mm256_set1_epi64x( H512[6] );
sc->val[7] = _mm256_set1_epi64x( H512[7] );
*/
}
void sha512_4way( sha512_4way_context *sc, const void *data, size_t len )
{
__m256i *vdata = (__m256i*)data;
size_t ptr;
const int buf_size = 128;
ptr = (unsigned)sc->count & (buf_size - 1U);
while ( len > 0 )
{
size_t clen;
clen = buf_size - ptr;
if ( clen > len )
clen = len;
memcpy_256( sc->buf + (ptr>>3), vdata, clen>>3 );
vdata = vdata + (clen>>3);
ptr += clen;
len -= clen;
if ( ptr == buf_size )
{
sha512_4way_round( sc, sc->buf, sc->val );
ptr = 0;
}
sc->count += clen;
}
}
void sha512_4way_close( sha512_4way_context *sc, void *dst )
{
unsigned ptr;
const int buf_size = 128;
const int pad = buf_size - 16;
ptr = (unsigned)sc->count & (buf_size - 1U);
sc->buf[ ptr>>3 ] = m256_const1_64( 0x80 );
ptr += 8;
if ( ptr > pad )
{
memset_zero_256( sc->buf + (ptr>>3), (buf_size - ptr) >> 3 );
sha512_4way_round( sc, sc->buf, sc->val );
memset_zero_256( sc->buf, pad >> 3 );
}
else
memset_zero_256( sc->buf + (ptr>>3), (pad - ptr) >> 3 );
sc->buf[ pad >> 3 ] =
mm256_bswap_64( _mm256_set1_epi64x( sc->count >> 61 ) );
sc->buf[ ( pad+8 ) >> 3 ] =
mm256_bswap_64( _mm256_set1_epi64x( sc->count << 3 ) );
sha512_4way_round( sc, sc->buf, sc->val );
mm256_block_bswap_64( dst, sc->val );
}
#endif // __AVX2__