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

v3.7.4

Browse Source
This commit is contained in:
Jay D Dee
2017-11-28 16:32:04 -05:00
parent 6d1361c87f
commit 4b57ac0eb9
70 changed files with 10549 additions and 2852 deletions
Download Patch File Download Diff File Expand all files Collapse all files

178
algo/nist5/nist5-4way.c Normal file
View File

@@ -0,0 +1,178 @@
#include "nist5-gate.h"
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#if defined(NIST5_4WAY)
#include "algo/blake/blake-hash-4way.h"
#include "algo/skein/skein-hash-4way.h"
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/groestl/aes_ni/hash-groestl.h"
// no improvement with midstate
//static __thread blake512_4way_context ctx_mid;
void nist5hash_4way( void *output, const void *input )
{
uint64_t hash0[8] __attribute__ ((aligned (64)));
uint64_t hash1[8] __attribute__ ((aligned (64)));
uint64_t hash2[8] __attribute__ ((aligned (64)));
uint64_t hash3[8] __attribute__ ((aligned (64)));
uint64_t vhash[8*4] __attribute__ ((aligned (64)));
blake512_4way_context ctx_blake;
hashState_groestl ctx_groestl;
jh512_4way_context ctx_jh;
skein512_4way_context ctx_skein;
keccak512_4way_context ctx_keccak;
// memcpy( &ctx_blake, &ctx_mid, sizeof(ctx_mid) );
// blake512_4way( &ctx_blake, input + (64<<2), 16 );
blake512_4way_init( &ctx_blake );
blake512_4way( &ctx_blake, input, 80 );
blake512_4way_close( &ctx_blake, vhash );
m256_deinterleave_4x64x( hash0, hash1, hash2, hash3, vhash, 512 );
init_groestl( &ctx_groestl, 64 );
update_and_final_groestl( &ctx_groestl, (char*)hash0,
(const char*)hash0, 512 );
init_groestl( &ctx_groestl, 64 );
update_and_final_groestl( &ctx_groestl, (char*)hash1,
(const char*)hash1, 512 );
init_groestl( &ctx_groestl, 64 );
update_and_final_groestl( &ctx_groestl, (char*)hash2,
(const char*)hash2, 512 );
init_groestl( &ctx_groestl, 64 );
update_and_final_groestl( &ctx_groestl, (char*)hash3,
(const char*)hash3, 512 );
m256_interleave_4x64x( vhash, hash0, hash1, hash2, hash3, 512 );
jh512_4way_init( &ctx_jh );
jh512_4way( &ctx_jh, vhash, 64 );
jh512_4way_close( &ctx_jh, vhash );
keccak512_4way_init( &ctx_keccak );
keccak512_4way( &ctx_keccak, vhash, 64 );
keccak512_4way_close( &ctx_keccak, vhash );
skein512_4way_init( &ctx_skein );
skein512_4way( &ctx_skein, vhash, 64 );
skein512_4way_close( &ctx_skein, vhash );
m256_deinterleave_4x64x( hash0, hash1, hash2, hash3, vhash, 512 );
memcpy( output, hash0, 32 );
memcpy( output+32, hash1, 32 );
memcpy( output+64, hash2, 32 );
memcpy( output+96, hash3, 32 );
}
int scanhash_nist5_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done)
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t endiandata[20] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
uint32_t *nonces = work->nonces;
bool *found = work->nfound;
int num_found = 0;
uint32_t *noncep0 = vdata + 73; // 9*8 + 1
uint32_t *noncep1 = vdata + 75;
uint32_t *noncep2 = vdata + 77;
uint32_t *noncep3 = vdata + 79;
uint64_t htmax[] = { 0,
0xF,
0xFF,
0xFFF,
0xFFFF,
0x10000000 };
uint32_t masks[] = { 0xFFFFFFFF,
0xFFFFFFF0,
0xFFFFFF00,
0xFFFFF000,
0xFFFF0000,
0 };
// we need bigendian data...
swab32_array( endiandata, pdata, 20 );
uint64_t *edata = (uint64_t*)endiandata;
m256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
// precalc midstate
// blake512_4way_init( &ctx_mid );
// blake512_4way( &ctx_mid, vdata, 64 );
for ( int m=0; m < 6; m++ )
{
if (Htarg <= htmax[m])
{
uint32_t mask = masks[m];
do {
found[0] = found[1] = found[2] = found[3] = false;
be32enc( noncep0, n );
be32enc( noncep1, n+1 );
be32enc( noncep2, n+2 );
be32enc( noncep3, n+3 );
nist5hash_4way( hash, vdata );
pdata[19] = n;
if ( ( !(hash[7] & mask) )
&& fulltest( hash, ptarget ) )
{
found[0] = true;
num_found++;
nonces[0] = n;
work_set_target_ratio( work, hash );
}
if ( ( !((hash+8)[7] & mask) )
&& fulltest( hash+8, ptarget ) )
{
found[1] = true;
num_found++;
nonces[1] = n+1;
work_set_target_ratio( work, hash+8 );
}
if ( ( !((hash+16)[7] & mask) )
&& fulltest( hash+16, ptarget ) )
{
found[2] = true;
num_found++;
nonces[2] = n+2;
work_set_target_ratio( work, hash+16 );
}
if ( ( !((hash+24)[7] & mask) )
&& fulltest( hash+24, ptarget ) )
{
found[3] = true;
num_found++;
nonces[3] = n+3;
work_set_target_ratio( work, hash+24 );
}
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

17
algo/nist5/nist5-gate.c Normal file
View File

@@ -0,0 +1,17 @@
#include "nist5-gate.h"
bool register_nist5_algo( algo_gate_t* gate )
{
#if defined (NIST5_4WAY)
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT;
gate->scanhash = (void*)&scanhash_nist5_4way;
gate->hash = (void*)&nist5hash_4way;
#else
gate->optimizations = SSE2_OPT | AES_OPT;
init_nist5_ctx();
gate->scanhash = (void*)&scanhash_nist5;
gate->hash = (void*)&nist5hash;
#endif
return true;
};

26
algo/nist5/nist5-gate.h Normal file
View File

@@ -0,0 +1,26 @@
#ifndef __NIST5_GATE_H__
#define __NIST5_GATE_H__
#include "algo-gate-api.h"
#include <stdint.h>
#if defined(FOUR_WAY) && defined(__AVX2__) && !defined(NO_AES_NI)
#define NIST5_4WAY
#endif
#if defined(NIST5_4WAY)
void nist5hash_4way( void *state, const void *input );
int scanhash_nist5_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
#else
void nist5hash( void *state, const void *input );
int scanhash_nist5( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
#endif
#endif

159
algo/nist5/nist5.c Normal file
View File

@@ -0,0 +1,159 @@
#include "nist5-gate.h"
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "algo/blake/sph_blake.h"
#include "algo/groestl/sph_groestl.h"
#include "algo/skein/sph_skein.h"
#include "algo/jh/sph_jh.h"
#include "algo/keccak/sph_keccak.h"
#include "algo/blake/sse2/blake.c"
#include "algo/keccak/sse2/keccak.c"
#include "algo/skein/sse2/skein.c"
#include "algo/jh/sse2/jh_sse2_opt64.h"
#ifndef NO_AES_NI
#include "algo/groestl/aes_ni/hash-groestl.h"
#endif
typedef struct {
#ifdef NO_AES_NI
sph_groestl512_context groestl;
#else
hashState_groestl groestl;
#endif
} nist5_ctx_holder;
nist5_ctx_holder nist5_ctx;
void init_nist5_ctx()
{
#ifdef NO_AES_NI
sph_groestl512_init( &nist5_ctx.groestl );
#else
init_groestl( &nist5_ctx.groestl, 64 );
#endif
}
void nist5hash(void *output, const void *input)
{
size_t hashptr;
unsigned char hashbuf[128];
sph_u64 hashctA;
sph_u64 hashctB;
unsigned char hash[128] __attribute__ ((aligned (64))) ;
#define hashA hash
#define hashB hash+64
nist5_ctx_holder ctx __attribute__ ((aligned (64)));
memcpy( &ctx, &nist5_ctx, sizeof(nist5_ctx) );
DECL_BLK;
BLK_I;
BLK_W;
BLK_C;
#ifdef NO_AES_NI
sph_groestl512 (&ctx.groestl, hash, 64);
sph_groestl512_close(&ctx.groestl, hash);
#else
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, 512 );
#endif
DECL_JH;
JH_H;
DECL_KEC;
KEC_I;
KEC_U;
KEC_C;
DECL_SKN;
SKN_I;
SKN_U;
SKN_C;
memcpy(output, hash, 32);
}
int scanhash_nist5(int thr_id, struct work *work,
uint32_t max_nonce, uint64_t *hashes_done)
{
uint32_t endiandata[20] __attribute__((aligned(64)));
uint32_t hash64[8] __attribute__((aligned(32)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19] - 1;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
uint64_t htmax[] = {
0,
0xF,
0xFF,
0xFFF,
0xFFFF,
0x10000000
};
uint32_t masks[] = {
0xFFFFFFFF,
0xFFFFFFF0,
0xFFFFFF00,
0xFFFFF000,
0xFFFF0000,
0
};
// we need bigendian data...
swab32_array( endiandata, pdata, 20 );
#ifdef DEBUG_ALGO
printf("[%d] Htarg=%X\n", thr_id, Htarg);
#endif
for (int m=0; m < 6; m++) {
if (Htarg <= htmax[m]) {
uint32_t mask = masks[m];
do {
pdata[19] = ++n;
be32enc(&endiandata[19], n);
nist5hash(hash64, endiandata);
#ifndef DEBUG_ALGO
if ((!(hash64[7] & mask)) && fulltest(hash64, ptarget)) {
*hashes_done = n - first_nonce + 1;
return true;
}
#else
if (!(n % 0x1000) && !thr_id) printf(".");
if (!(hash64[7] & mask)) {
printf("[%d]",thr_id);
if (fulltest(hash64, ptarget)) {
*hashes_done = n - first_nonce + 1;
return true;
}
}
#endif
} while (n < max_nonce && !work_restart[thr_id].restart);
// see blake.c if else to understand the loop on htmax => mask
break;
}
}
*hashes_done = n - first_nonce + 1;
pdata[19] = n;
return 0;
}
/*
bool register_nist5_algo( algo_gate_t* gate )
{
gate->optimizations = SSE2_OPT | AES_OPT;
init_nist5_ctx();
gate->scanhash = (void*)&scanhash_nist5;
gate->hash = (void*)&nist5hash;
return true;
};
*/