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

v3.8.5

Browse Source
This commit is contained in:
Jay D Dee
2018-03-27 20:20:05 -04:00
parent 3363d61524
commit f449c6725f
105 changed files with 4560 additions and 1846 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
algo/sha/sha2-hash-4way.c
View File

@@ -30,7 +30,7 @@
* @author Thomas Pornin <thomas.pornin@cryptolog.com>
*/
#if defined(__AVX__)
#if defined(__SSE4_2__)
#include <stddef.h>
#include <string.h>
@@ -98,19 +98,19 @@ static const sph_u32 K256[64] = {
#define BSG2_0(x) \
_mm_xor_si128( _mm_xor_si128( \
mm_rotr_32(x, 2), mm_rotr_32(x, 13) ), mm_rotr_32( x, 22) )
mm_ror_32(x, 2), mm_ror_32(x, 13) ), mm_ror_32( x, 22) )
#define BSG2_1(x) \
_mm_xor_si128( _mm_xor_si128( \
mm_rotr_32(x, 6), mm_rotr_32(x, 11) ), mm_rotr_32( x, 25) )
mm_ror_32(x, 6), mm_ror_32(x, 11) ), mm_ror_32( x, 25) )
#define SSG2_0(x) \
_mm_xor_si128( _mm_xor_si128( \
mm_rotr_32(x, 7), mm_rotr_32(x, 18) ), _mm_srli_epi32(x, 3) )
mm_ror_32(x, 7), mm_ror_32(x, 18) ), _mm_srli_epi32(x, 3) )
#define SSG2_1(x) \
_mm_xor_si128( _mm_xor_si128( \
mm_rotr_32(x, 17), mm_rotr_32(x, 19) ), _mm_srli_epi32(x, 10) )
mm_ror_32(x, 17), mm_ror_32(x, 19) ), _mm_srli_epi32(x, 10) )
#define SHA2s_4WAY_STEP(A, B, C, D, E, F, G, H, i, j) \
do { \
@@ -311,19 +311,19 @@ void sha256_4way_close( sha256_4way_context *sc, void *dst )
#define BSG2_0x(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_rotr_32(x, 2), mm256_rotr_32(x, 13) ), mm256_rotr_32( x, 22) )
mm256_ror_32(x, 2), mm256_ror_32(x, 13) ), mm256_ror_32( x, 22) )
#define BSG2_1x(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_rotr_32(x, 6), mm256_rotr_32(x, 11) ), mm256_rotr_32( x, 25) )
mm256_ror_32(x, 6), mm256_ror_32(x, 11) ), mm256_ror_32( x, 25) )
#define SSG2_0x(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_rotr_32(x, 7), mm256_rotr_32(x, 18) ), _mm256_srli_epi32(x, 3) )
mm256_ror_32(x, 7), mm256_ror_32(x, 18) ), _mm256_srli_epi32(x, 3) )
#define SSG2_1x(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_rotr_32(x, 17), mm256_rotr_32(x, 19) ), _mm256_srli_epi32(x, 10) )
mm256_ror_32(x, 17), mm256_ror_32(x, 19) ), _mm256_srli_epi32(x, 10) )
#define SHA2x_MEXP( a, b, c, d ) \
_mm256_add_epi32( _mm256_add_epi32( _mm256_add_epi32( \
@@ -644,19 +644,19 @@ static const sph_u64 K512[80] = {
#define BSG5_0(x) \
_mm256_xor_si256( _mm256_xor_si256( \
mm256_rotr_64(x, 28), mm256_rotr_64(x, 34) ), mm256_rotr_64(x, 39) )
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_rotr_64(x, 14), mm256_rotr_64(x, 18) ), mm256_rotr_64(x, 41) )
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_rotr_64(x, 1), mm256_rotr_64(x, 8) ), _mm256_srli_epi64(x, 7) )
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_rotr_64(x, 19), mm256_rotr_64(x, 61) ), _mm256_srli_epi64(x, 6) )
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 { \
@@ -781,4 +781,4 @@ void sha512_4way_close( sha512_4way_context *sc, void *dst )
}
#endif // __AVX2__
#endif // __AVX__
#endif // __SSE4_2__

2
algo/sha/sha2-hash-4way.h
View File

@@ -44,7 +44,7 @@
#include "sph_types.h"
#include "avxdefs.h"
#if defined(__AVX__)
#if defined(__SSE4_2__)
//#define SPH_SIZE_sha256 256

198
algo/sha/sha256t-4way.c Normal file
View File

@@ -0,0 +1,198 @@
#include "sha256t-gate.h"
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "sha2-hash-4way.h"
//#include <openssl/sha.h>
#if defined(SHA256T_8WAY)
static __thread sha256_8way_context sha256_ctx8 __attribute__ ((aligned (64)));
void sha256t_8way_hash( void* output, const void* input )
{
uint32_t vhash[8*8] __attribute__ ((aligned (64)));
sha256_8way_context ctx;
memcpy( &ctx, &sha256_ctx8, sizeof ctx );
sha256_8way( &ctx, input + (64<<3), 16 );
sha256_8way_close( &ctx, vhash );
sha256_8way_init( &ctx );
sha256_8way( &ctx, vhash, 32 );
sha256_8way_close( &ctx, vhash );
sha256_8way_init( &ctx );
sha256_8way( &ctx, vhash, 32 );
sha256_8way_close( &ctx, vhash );
mm256_deinterleave_8x32( output, output+ 32, output+ 64, output+ 96,
output+128, output+160, output+192, output+224,
vhash, 256 );
}
int scanhash_sha256t_8way( int thr_id, struct work *work,
uint32_t max_nonce, uint64_t *hashes_done )
{
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t hash[8*8] __attribute__ ((aligned (32)));
uint32_t edata[20] __attribute__ ((aligned (32)));;
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
uint32_t *nonces = work->nonces;
int num_found = 0;
uint32_t *noncep = vdata + 152; // 19*8
const uint64_t htmax[] = { 0,
0xF,
0xFF,
0xFFF,
0xFFFF,
0x10000000 };
const uint32_t masks[] = { 0xFFFFFFFF,
0xFFFFFFF0,
0xFFFFFF00,
0xFFFFF000,
0xFFFF0000,
0 };
for ( int k = 0; k < 19; k++ )
be32enc( &edata[k], pdata[k] );
mm256_interleave_8x32( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
sha256_8way_init( &sha256_ctx8 );
sha256_8way( &sha256_ctx8, vdata, 64 );
for ( int m = 0; m < 6; m++ ) if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
do {
be32enc( noncep, n );
be32enc( noncep +1, n+1 );
be32enc( noncep +2, n+2 );
be32enc( noncep +3, n+3 );
be32enc( noncep +4, n+4 );
be32enc( noncep +5, n+5 );
be32enc( noncep +6, n+6 );
be32enc( noncep +7, n+7 );
pdata[19] = n;
sha256t_8way_hash( hash, vdata );
for ( int i = 0; i < 8; i++ )
if ( ( !( ( hash+(i<<3) )[7] & mask ) )
&& fulltest( hash+(i<<3), ptarget ) )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
}
n += 8;
} while ( (num_found == 0) && (n < max_nonce)
&& !work_restart[thr_id].restart );
break;
}
*hashes_done = n - first_nonce + 1;
return num_found;
}
#elif defined(SHA256T_4WAY)
static __thread sha256_4way_context sha256_ctx4 __attribute__ ((aligned (64)));
void sha256t_4way_hash( void* output, const void* input )
{
uint32_t vhash[8*4] __attribute__ ((aligned (64)));
sha256_4way_context ctx;
memcpy( &ctx, &sha256_ctx4, sizeof ctx );
sha256_4way( &ctx, input + (64<<2), 16 );
sha256_4way_close( &ctx, vhash );
sha256_4way_init( &ctx );
sha256_4way( &ctx, vhash, 32 );
sha256_4way_close( &ctx, vhash );
sha256_4way_init( &ctx );
sha256_4way( &ctx, vhash, 32 );
sha256_4way_close( &ctx, vhash );
mm_deinterleave_4x32( output, output+ 32, output+ 64, output+ 96,
vhash, 256 );
}
int scanhash_sha256t_4way( int thr_id, struct work *work,
uint32_t max_nonce, uint64_t *hashes_done )
{
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t hash[8*4] __attribute__ ((aligned (32)));
uint32_t edata[20] __attribute__ ((aligned (32)));;
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
uint32_t *nonces = work->nonces;
int num_found = 0;
uint32_t *noncep = vdata + 76; // 19*4
const uint64_t htmax[] = { 0,
0xF,
0xFF,
0xFFF,
0xFFFF,
0x10000000 };
const uint32_t masks[] = { 0xFFFFFFFF,
0xFFFFFFF0,
0xFFFFFF00,
0xFFFFF000,
0xFFFF0000,
0 };
for ( int k = 0; k < 19; k++ )
be32enc( &edata[k], pdata[k] );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
sha256_4way_init( &sha256_ctx4 );
sha256_4way( &sha256_ctx4, vdata, 64 );
for ( int m = 0; m < 6; m++ ) if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
do {
be32enc( noncep, n );
be32enc( noncep +1, n+1 );
be32enc( noncep +2, n+2 );
be32enc( noncep +3, n+3 );
pdata[19] = n;
sha256t_4way_hash( hash, vdata );
for ( int i = 0; i < 4; i++ )
if ( ( !( ( hash+(i<<3) )[7] & mask ) )
&& fulltest( hash+(i<<3), ptarget ) )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
}
n += 4;
} while ( (num_found == 0) && (n < max_nonce)
&& !work_restart[thr_id].restart );
break;
}
*hashes_done = n - first_nonce + 1;
return num_found;
}
#endif

26
algo/sha/sha256t-gate.c Normal file
View File

@@ -0,0 +1,26 @@
#include "sha256t-gate.h"
bool register_sha256t_algo( algo_gate_t* gate )
{
#if defined(SHA256T_8WAY)
gate->scanhash = (void*)&scanhash_sha256t_8way;
gate->hash = (void*)&sha256t_8way_hash;
#elif defined(SHA256T_4WAY)
gate->scanhash = (void*)&scanhash_sha256t_4way;
gate->hash = (void*)&sha256t_4way_hash;
#else
gate->scanhash = (void*)&scanhash_sha256t;
gate->hash = (void*)&sha256t_hash;
/*
#ifndef USE_SPH_SHA
SHA256_Init( &sha256t_ctx );
#else
sph_sha256_init( &sha256t_ctx );
#endif
*/
#endif
gate->optimizations = SSE42_OPT | AVX2_OPT | SHA_OPT;
gate->get_max64 = (void*)&get_max64_0x3ffff;
return true;
}

36
algo/sha/sha256t-gate.h Normal file
View File

@@ -0,0 +1,36 @@
#ifndef __SHA256T_GATE_H__
#define __SHA256T_GATE_H__ 1
#include <stdint.h>
#include "algo-gate-api.h"
#if defined(__SSE4_2__)
#define SHA256T_4WAY
#endif
#if defined(__AVX2__)
#define SHA256T_8WAY
#endif
bool register_blake2s_algo( algo_gate_t* gate );
#if defined(SHA256T_8WAY)
void sha256t_8way_hash( void *output, const void *input );
int scanhash_sha256t_8way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
#elif defined (SHA256T_4WAY)
void sha256t_4way_hash( void *output, const void *input );
int scanhash_sha256t_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
#else
void sha256t_hash( void *output, const void *input );
int scanhash_sha256t( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
#endif
#endif

53
algo/sha/sha256t.c
View File

@@ -1,5 +1,4 @@
#include "algo-gate-api.h"
#include "sha256t-gate.h"
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
@@ -7,25 +6,26 @@
#include "sph_sha2.h"
#include <openssl/sha.h>
#if !defined(SHA256T_4WAY)
#ifndef USE_SPH_SHA
static SHA256_CTX sha256t_ctx __attribute__ ((aligned (64)));
static __thread SHA256_CTX sha256t_mid __attribute__ ((aligned (64)));
static __thread SHA256_CTX sha256t_ctx __attribute__ ((aligned (64)));
#else
static sph_sha256_context sha256t_ctx __attribute__ ((aligned (64)));
static __thread sph_sha256_context sha256t_mid __attribute__ ((aligned (64)));
static __thread sph_sha256_context sha256t_ctx __attribute__ ((aligned (64)));
#endif
void sha256t_midstate( const void* input )
{
memcpy( &sha256t_mid, &sha256t_ctx, sizeof sha256t_mid );
#ifndef USE_SPH_SHA
SHA256_Update( &sha256t_mid, input, 64 );
SHA256_Init( &sha256t_ctx );
SHA256_Update( &sha256t_ctx, input, 64 );
#else
sph_sha256( &sha256t_mid, input, 64 );
sph_sha256_init( &sha256t_ctx );
sph_sha256( &sha256t_ctx, input, 64 );
#endif
}
void sha256t_hash(void* output, const void* input, uint32_t len)
void sha256t_hash( void* output, const void* input )
{
uint32_t _ALIGN(64) hashA[16];
const int midlen = 64; // bytes
@@ -33,16 +33,16 @@ void sha256t_hash(void* output, const void* input, uint32_t len)
#ifndef USE_SPH_SHA
SHA256_CTX ctx_sha256 __attribute__ ((aligned (64)));
memcpy( &ctx_sha256, &sha256t_mid, sizeof sha256t_mid );
memcpy( &ctx_sha256, &sha256t_ctx, sizeof sha256t_ctx );
SHA256_Update( &ctx_sha256, input + midlen, tail );
SHA256_Final( (unsigned char*)hashA, &ctx_sha256 );
memcpy( &ctx_sha256, &sha256t_ctx, sizeof sha256t_ctx );
SHA256_Init( &ctx_sha256 );
SHA256_Update( &ctx_sha256, hashA, 32 );
SHA256_Final( (unsigned char*)hashA, &ctx_sha256 );
memcpy( &ctx_sha256, &sha256t_ctx, sizeof sha256t_ctx );
SHA256_Init( &ctx_sha256 );
SHA256_Update( &ctx_sha256, hashA, 32 );
SHA256_Final( (unsigned char*)hashA, &ctx_sha256 );
#else
@@ -52,11 +52,11 @@ void sha256t_hash(void* output, const void* input, uint32_t len)
sph_sha256( &ctx_sha256, input + midlen, tail );
sph_sha256_close( &ctx_sha256, hashA );
memcpy( &ctx_sha256, &sha256t_ctx, sizeof sha256t_ctx );
sph_sha256_init( &ctx_sha256 );
sph_sha256( &ctx_sha256, hashA, 32 );
sph_sha256_close( &ctx_sha256, hashA );
memcpy( &ctx_sha256, &sha256t_ctx, sizeof sha256t_ctx );
sph_sha256_init( &ctx_sha256 );
sph_sha256( &ctx_sha256, hashA, 32 );
sph_sha256_close( &ctx_sha256, hashA );
#endif
@@ -68,8 +68,6 @@ int scanhash_sha256t(int thr_id, struct work *work,
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t len = 80;
uint32_t n = pdata[19] - 1;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
@@ -113,7 +111,7 @@ int scanhash_sha256t(int thr_id, struct work *work,
do {
pdata[19] = ++n;
be32enc(&endiandata[19], n);
sha256t_hash(hash64, endiandata, len);
sha256t_hash( hash64, endiandata );
#ifndef DEBUG_ALGO
if ((!(hash64[7] & mask)) && fulltest(hash64, ptarget)) {
*hashes_done = n - first_nonce + 1;
@@ -139,23 +137,4 @@ int scanhash_sha256t(int thr_id, struct work *work,
pdata[19] = n;
return 0;
}
void sha256t_set_target( struct work* work, double job_diff )
{
work_set_target( work, job_diff / (256.0 * opt_diff_factor) );
}
bool register_sha256t_algo( algo_gate_t* gate )
{
#ifndef USE_SPH_SHA
SHA256_Init( &sha256t_ctx );
#else
sph_sha256_init( &sha256t_ctx );
#endif
gate->optimizations = SSE2_OPT | AVX_OPT | AVX2_OPT | SHA_OPT;
gate->scanhash = (void*)&scanhash_sha256t;
gate->hash = (void*)&sha256t_hash;
// gate->set_target = (void*)&sha256t_set_target;
gate->get_max64 = (void*)&get_max64_0x3ffff;
return true;
}