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

v3.21.3 Unreleased

Browse Source
This commit is contained in:
Jay D Dee
2023-03-13 03:20:13 -04:00
parent b339450898
commit c6bc9d67fb
49 changed files with 1126 additions and 1111 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
algo/x16/hex.c
View File

@@ -25,7 +25,7 @@ static void hex_getAlgoString(const uint32_t* prevblock, char *output)
static __thread x16r_context_overlay hex_ctx;
int hex_hash( void* output, const void* input, const int thrid )
int hex_hash( void* output, const void* input, int thrid )
{
uint32_t _ALIGN(128) hash[16];
x16r_context_overlay ctx;

4
algo/x16/minotaur.c
View File

@@ -72,7 +72,7 @@ struct TortureGarden
// Get a 64-byte hash for given 64-byte input, using given TortureGarden contexts and given algo index
static int get_hash( void *output, const void *input, TortureGarden *garden,
unsigned int algo, const int thr_id )
unsigned int algo, int thr_id )
{
unsigned char hash[64] __attribute__ ((aligned (64)));
int rc = 1;
@@ -233,7 +233,7 @@ bool initialize_torture_garden()
}
// Produce a 32-byte hash from 80-byte input data
int minotaur_hash( void *output, const void *input, const int thr_id )
int minotaur_hash( void *output, const void *input, int thr_id )
{
unsigned char hash[64] __attribute__ ((aligned (64)));
int rc = 1;

68
algo/x16/x16r-4way.c
View File

@@ -19,7 +19,7 @@
// Perform midstate prehash of hash functions with block size <= 72 bytes,
// 76 bytes for hash functions that operate on 32 bit data.
void x16r_8way_do_prehash( void *vdata, const void *pdata )
void x16r_8way_prehash( void *vdata, void *pdata )
{
uint32_t vdata2[20*8] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
@@ -106,18 +106,11 @@ void x16r_8way_do_prehash( void *vdata, const void *pdata )
}
}
int x16r_8way_prehash( struct work *work )
{
x16r_gate_get_hash_order( work, x16r_hash_order );
x16r_8way_do_prehash( x16r_8way_vdata, work->data );
return 1;
}
// Perform the full x16r hash and returns 512 bit intermediate hash.
// Called by wrapper hash function to optionally continue hashing and
// convert to final hash.
int x16r_8way_hash_generic( void* output, const void* input, const int thrid )
int x16r_8way_hash_generic( void* output, const void* input, int thrid )
{
uint32_t vhash[20*8] __attribute__ ((aligned (128)));
uint32_t hash0[20] __attribute__ ((aligned (16)));
@@ -478,7 +471,7 @@ int x16r_8way_hash_generic( void* output, const void* input, const int thrid )
// x16-r,-s,-rt wrapper called directly by scanhash to repackage 512 bit
// hash to 256 bit final hash.
int x16r_8way_hash( void* output, const void* input, const int thrid )
int x16r_8way_hash( void* output, const void* input, int thrid )
{
uint8_t hash[64*8] __attribute__ ((aligned (128)));
if ( !x16r_8way_hash_generic( hash, input, thrid ) )
@@ -502,6 +495,7 @@ int scanhash_x16r_8way( struct work *work, uint32_t max_nonce,
{
uint32_t hash[16*8] __attribute__ ((aligned (128)));
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t bedata1[2];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
@@ -514,16 +508,27 @@ int scanhash_x16r_8way( struct work *work, uint32_t max_nonce,
if ( bench ) ptarget[7] = 0x0cff;
pthread_rwlock_rdlock( &g_work_lock );
memcpy( vdata, x16r_8way_vdata, sizeof vdata );
pthread_rwlock_unlock( &g_work_lock );
bedata1[0] = bswap_32( pdata[1] );
bedata1[1] = bswap_32( pdata[2] );
static __thread uint32_t s_ntime = UINT32_MAX;
const uint32_t ntime = bswap_32( pdata[17] );
if ( s_ntime != ntime )
{
x16_r_s_getAlgoString( (const uint8_t*)bedata1, x16r_hash_order );
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_INFO, "Hash order %s Ntime %08x", x16r_hash_order, ntime );
}
x16r_8way_prehash( vdata, pdata );
*noncev = mm512_intrlv_blend_32( _mm512_set_epi32(
n+7, 0, n+6, 0, n+5, 0, n+4, 0,
n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
do
{
if( algo_gate.hash( hash, vdata, thr_id ) );
if( x16r_8way_hash( hash, vdata, thr_id ) );
for ( int i = 0; i < 8; i++ )
if ( unlikely( valid_hash( hash + (i<<3), ptarget ) && !bench ) )
{
@@ -541,7 +546,7 @@ int scanhash_x16r_8way( struct work *work, uint32_t max_nonce,
#elif defined (X16R_4WAY)
void x16r_4way_do_prehash( void *vdata, const void *pdata )
void x16r_4way_prehash( void *vdata, void *pdata )
{
uint32_t vdata2[20*4] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
@@ -622,14 +627,7 @@ void x16r_4way_do_prehash( void *vdata, const void *pdata )
}
}
int x16r_4way_prehash( struct work *work )
{
x16r_gate_get_hash_order( work, x16r_hash_order );
x16r_4way_do_prehash( x16r_4way_vdata, work->data );
return 1;
}
int x16r_4way_hash_generic( void* output, const void* input, const int thrid )
int x16r_4way_hash_generic( void* output, const void* input, int thrid )
{
uint32_t vhash[20*4] __attribute__ ((aligned (128)));
uint32_t hash0[20] __attribute__ ((aligned (32)));
@@ -637,14 +635,13 @@ int x16r_4way_hash_generic( void* output, const void* input, const int thrid )
uint32_t hash2[20] __attribute__ ((aligned (32)));
uint32_t hash3[20] __attribute__ ((aligned (32)));
x16r_4way_context_overlay ctx;
memcpy( &ctx, &x16r_ctx, sizeof(ctx) );
void *in0 = (void*) hash0;
void *in1 = (void*) hash1;
void *in2 = (void*) hash2;
void *in3 = (void*) hash3;
int size = 80;
memcpy( &ctx, &x16r_ctx, sizeof(ctx) );
dintrlv_4x64( hash0, hash1, hash2, hash3, input, 640 );
for ( int i = 0; i < 16; i++ )
@@ -908,7 +905,7 @@ int x16r_4way_hash_generic( void* output, const void* input, const int thrid )
return 1;
}
int x16r_4way_hash( void* output, const void* input, const int thrid )
int x16r_4way_hash( void* output, const void* input, int thrid )
{
uint8_t hash[64*4] __attribute__ ((aligned (64)));
if ( !x16r_4way_hash_generic( hash, input, thrid ) )
@@ -927,6 +924,7 @@ int scanhash_x16r_4way( struct work *work, uint32_t max_nonce,
{
uint32_t hash[16*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t bedata1[2];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
@@ -939,15 +937,25 @@ int scanhash_x16r_4way( struct work *work, uint32_t max_nonce,
if ( bench ) ptarget[7] = 0x0cff;
pthread_rwlock_rdlock( &g_work_lock );
memcpy( vdata, x16r_4way_vdata, sizeof vdata );
pthread_rwlock_unlock( &g_work_lock );
bedata1[0] = bswap_32( pdata[1] );
bedata1[1] = bswap_32( pdata[2] );
static __thread uint32_t s_ntime = UINT32_MAX;
const uint32_t ntime = bswap_32( pdata[17] );
if ( s_ntime != ntime )
{
x16_r_s_getAlgoString( (const uint8_t*)bedata1, x16r_hash_order );
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_INFO, "Hash order %s Ntime %08x", x16r_hash_order, ntime );
}
x16r_4way_prehash( vdata, pdata );
*noncev = mm256_intrlv_blend_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
do
{
if ( algo_gate.hash( hash, vdata, thr_id ) );
if ( x16r_4way_hash( hash, vdata, thr_id ) );
for ( int i = 0; i < 4; i++ )
if ( unlikely( valid_hash( hash + (i<<3), ptarget ) && !bench ) )
{

139
algo/x16/x16r-gate.c
View File

@@ -1,44 +1,26 @@
#include "x16r-gate.h"
#include "algo/sha/sha256d.h"
char x16r_hash_order[ X16R_HASH_FUNC_COUNT + 1 ] = {0};
__thread char x16r_hash_order[ X16R_HASH_FUNC_COUNT + 1 ] = { 0 };
void (*x16r_gate_get_hash_order) ( const struct work *, char * ) = NULL;
void (*x16_r_s_getAlgoString) ( const uint8_t*, char* ) = NULL;
#if defined (X16R_8WAY)
x16r_8way_context_overlay x16r_ctx;
uint32_t x16r_8way_vdata[24*8] __attribute__ ((aligned (64)));
__thread x16r_8way_context_overlay x16r_ctx;
#elif defined (X16R_4WAY)
x16r_4way_context_overlay x16r_ctx;
uint32_t x16r_4way_vdata[24*4] __attribute__ ((aligned (64)));
__thread x16r_4way_context_overlay x16r_ctx;
#endif
#if defined (X16RV2_8WAY)
__thread x16r_context_overlay x16_ctx;
x16rv2_8way_context_overlay x16rv2_ctx;
#elif defined (X16RV2_4WAY)
x16rv2_4way_context_overlay x16rv2_ctx;
#endif
x16r_context_overlay x16_ctx;
uint32_t x16r_edata[24] __attribute__ ((aligned (32)));
void x16r_get_hash_order( const struct work *work, char *hash_order )
void x16r_getAlgoString( const uint8_t* prevblock, char *output )
{
char *sptr = hash_order;
const uint32_t *pdata = work->data;
uint8_t prevblock[16];
((uint32_t*)prevblock)[0] = bswap_32( pdata[1] );
((uint32_t*)prevblock)[1] = bswap_32( pdata[2] );
char *sptr = output;
for ( int j = 0; j < X16R_HASH_FUNC_COUNT; j++ )
{
uint8_t b = (15 - j) >> 1; // 16 first ascii hex chars (lsb in uint256)
@@ -50,51 +32,38 @@ void x16r_get_hash_order( const struct work *work, char *hash_order )
sptr++;
}
*sptr = '\0';
if ( !opt_quiet )
applog( LOG_INFO, "Hash order %s", x16r_hash_order );
}
void x16s_get_hash_order( const struct work *work, char *hash_order )
void x16s_getAlgoString( const uint8_t* prevblock, char *output )
{
const uint32_t *pdata = work->data;
uint8_t prevblock[16];
((uint32_t*)prevblock)[0] = bswap_32( pdata[1] );
((uint32_t*)prevblock)[1] = bswap_32( pdata[2] );
strcpy( hash_order, "0123456789ABCDEF" );
strcpy( output, "0123456789ABCDEF" );
for ( int i = 0; i < 16; i++ )
{
uint8_t b = (15 - i) >> 1; // 16 ascii hex chars, reversed
uint8_t algoDigit = (i & 1) ? prevblock[b] & 0xF : prevblock[b] >> 4;
int offset = algoDigit;
// insert the nth character at the front
char oldVal = hash_order[ offset ];
char oldVal = output[offset];
for( int j = offset; j-- > 0; )
hash_order[ j+1 ] = hash_order[ j ];
hash_order[ 0 ] = oldVal;
output[j+1] = output[j];
output[0] = oldVal;
}
if ( !opt_quiet )
applog( LOG_INFO, "Hash order %s", x16r_hash_order );
}
bool register_x16r_algo( algo_gate_t* gate )
{
#if defined (X16R_8WAY)
gate->scanhash = (void*)&scanhash_x16r_8way;
gate->prehash = (void*)&x16r_8way_prehash;
gate->hash = (void*)&x16r_8way_hash;
#elif defined (X16R_4WAY)
gate->scanhash = (void*)&scanhash_x16r_4way;
gate->prehash = (void*)&x16r_4way_prehash;
gate->hash = (void*)&x16r_4way_hash;
#else
gate->scanhash = (void*)&scanhash_x16r;
gate->prehash = (void*)&x16r_prehash;
gate->hash = (void*)&x16r_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
x16r_gate_get_hash_order = (void*)&x16r_get_hash_order;
x16_r_s_getAlgoString = (void*)&x16r_getAlgoString;
opt_target_factor = 256.0;
return true;
};
@@ -102,20 +71,17 @@ bool register_x16r_algo( algo_gate_t* gate )
bool register_x16rv2_algo( algo_gate_t* gate )
{
#if defined (X16RV2_8WAY)
gate->scanhash = (void*)&scanhash_x16r_8way;
gate->prehash = (void*)&x16rv2_8way_prehash;
gate->scanhash = (void*)&scanhash_x16rv2_8way;
gate->hash = (void*)&x16rv2_8way_hash;
#elif defined (X16RV2_4WAY)
gate->scanhash = (void*)&scanhash_x16r_4way;
gate->prehash = (void*)&x16rv2_4way_prehash;
gate->scanhash = (void*)&scanhash_x16rv2_4way;
gate->hash = (void*)&x16rv2_4way_hash;
#else
gate->scanhash = (void*)&scanhash_x16r;
gate->prehash = (void*)&x16rv2_prehash;
gate->scanhash = (void*)&scanhash_x16rv2;
gate->hash = (void*)&x16rv2_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
x16r_gate_get_hash_order = (void*)&x16r_get_hash_order;
x16_r_s_getAlgoString = (void*)&x16r_getAlgoString;
opt_target_factor = 256.0;
return true;
};
@@ -124,19 +90,16 @@ bool register_x16s_algo( algo_gate_t* gate )
{
#if defined (X16R_8WAY)
gate->scanhash = (void*)&scanhash_x16r_8way;
gate->prehash = (void*)&x16r_8way_prehash;
gate->hash = (void*)&x16r_8way_hash;
#elif defined (X16R_4WAY)
gate->scanhash = (void*)&scanhash_x16r_4way;
gate->prehash = (void*)&x16r_4way_prehash;
gate->hash = (void*)&x16r_4way_hash;
#else
gate->scanhash = (void*)&scanhash_x16r;
gate->prehash = (void*)&x16r_prehash;
gate->hash = (void*)&x16r_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
x16r_gate_get_hash_order = (void*)&x16s_get_hash_order;
x16_r_s_getAlgoString = (void*)&x16s_getAlgoString;
opt_target_factor = 256.0;
return true;
};
@@ -145,33 +108,30 @@ bool register_x16s_algo( algo_gate_t* gate )
//
// X16RT
void x16rt_get_hash_order( const struct work * work, char * hash_order )
{
uint32_t _ALIGN(64) timehash[8*8];
const uint32_t ntime = bswap_32( work->data[17] );
const int32_t masked_ntime = ntime & 0xffffff80;
uint8_t* data = (uint8_t*)timehash;
char *sptr = hash_order;
sha256d( (unsigned char*)timehash, (const unsigned char*)( &masked_ntime ),
sizeof( masked_ntime ) );
void x16rt_getTimeHash( const uint32_t timeStamp, void* timeHash )
{
int32_t maskedTime = timeStamp & 0xffffff80;
sha256d( (unsigned char*)timeHash, (const unsigned char*)( &maskedTime ),
sizeof( maskedTime ) );
}
for ( uint8_t j = 0; j < X16R_HASH_FUNC_COUNT; j++ )
{
void x16rt_getAlgoString( const uint32_t *timeHash, char *output)
{
char *sptr = output;
uint8_t* data = (uint8_t*)timeHash;
for (uint8_t j = 0; j < X16R_HASH_FUNC_COUNT; j++) {
uint8_t b = (15 - j) >> 1; // 16 ascii hex chars, reversed
uint8_t algoDigit = (j & 1) ? data[b] & 0xF : data[b] >> 4;
if ( algoDigit >= 10 )
sprintf( sptr, "%c", 'A' + (algoDigit - 10) );
if (algoDigit >= 10)
sprintf(sptr, "%c", 'A' + (algoDigit - 10));
else
sprintf( sptr, "%u", (uint32_t) algoDigit );
sprintf(sptr, "%u", (uint32_t) algoDigit);
sptr++;
}
*sptr = '\0';
if ( !opt_quiet )
applog( LOG_INFO, "Hash order %s, ntime %08x, time hash %08x",
hash_order, ntime, timehash );
}
void veil_build_extraheader( struct work* g_work, struct stratum_ctx* sctx )
@@ -262,19 +222,15 @@ void veil_build_extraheader( struct work* g_work, struct stratum_ctx* sctx )
bool register_x16rt_algo( algo_gate_t* gate )
{
#if defined (X16R_8WAY)
gate->scanhash = (void*)&scanhash_x16r_8way;
gate->prehash = (void*)&x16r_8way_prehash;
gate->scanhash = (void*)&scanhash_x16rt_8way;
gate->hash = (void*)&x16r_8way_hash;
#elif defined (X16R_4WAY)
gate->scanhash = (void*)&scanhash_x16r_4way;
gate->prehash = (void*)&x16r_4way_prehash;
gate->scanhash = (void*)&scanhash_x16rt_4way;
gate->hash = (void*)&x16r_4way_hash;
#else
gate->scanhash = (void*)&scanhash_x16r;
gate->prehash = (void*)&x16r_prehash;
gate->scanhash = (void*)&scanhash_x16rt;
gate->hash = (void*)&x16r_hash;
#endif
x16r_gate_get_hash_order = (void*)&x16rt_get_hash_order;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
opt_target_factor = 256.0;
return true;
@@ -283,20 +239,16 @@ bool register_x16rt_algo( algo_gate_t* gate )
bool register_x16rt_veil_algo( algo_gate_t* gate )
{
#if defined (X16R_8WAY)
gate->scanhash = (void*)&scanhash_x16r_8way;
gate->prehash = (void*)&x16r_8way_prehash;
gate->scanhash = (void*)&scanhash_x16rt_8way;
gate->hash = (void*)&x16r_8way_hash;
#elif defined (X16R_4WAY)
gate->scanhash = (void*)&scanhash_x16r_4way;
gate->prehash = (void*)&x16r_4way_prehash;
gate->scanhash = (void*)&scanhash_x16rt_4way;
gate->hash = (void*)&x16r_4way_hash;
#else
gate->scanhash = (void*)&scanhash_x16r;
gate->prehash = (void*)&x16r_prehash;
gate->scanhash = (void*)&scanhash_x16rt;
gate->hash = (void*)&x16r_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
x16r_gate_get_hash_order = (void*)&x16rt_get_hash_order;
gate->build_extraheader = (void*)&veil_build_extraheader;
opt_target_factor = 256.0;
return true;
@@ -323,23 +275,20 @@ bool register_hex_algo( algo_gate_t* gate )
bool register_x21s_algo( algo_gate_t* gate )
{
#if defined (X16R_8WAY)
gate->scanhash = (void*)&scanhash_x16r_8way;
gate->prehash = (void*)&x16r_8way_prehash;
gate->scanhash = (void*)&scanhash_x21s_8way;
gate->hash = (void*)&x21s_8way_hash;
gate->miner_thread_init = (void*)&x21s_8way_thread_init;
#elif defined (X16R_4WAY)
gate->scanhash = (void*)&scanhash_x16r_4way;
gate->prehash = (void*)&x16r_4way_prehash;
gate->scanhash = (void*)&scanhash_x21s_4way;
gate->hash = (void*)&x21s_4way_hash;
gate->miner_thread_init = (void*)&x21s_4way_thread_init;
#else
gate->scanhash = (void*)&scanhash_x16r;
gate->prehash = (void*)&x16r_prehash;
gate->scanhash = (void*)&scanhash_x21s;
gate->hash = (void*)&x21s_hash;
gate->miner_thread_init = (void*)&x21s_thread_init;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
x16r_gate_get_hash_order = (void*)&x16s_get_hash_order;
x16_r_s_getAlgoString = (void*)&x16s_getAlgoString;
opt_target_factor = 256.0;
return true;
};

163
algo/x16/x16r-gate.h
View File

@@ -21,7 +21,6 @@
#include "algo/shabal/sph_shabal.h"
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/sha/sph_sha2.h"
#include "algo/tiger/sph_tiger.h"
#if defined(__AES__)
#include "algo/echo/aes_ni/hash_api.h"
@@ -58,11 +57,13 @@
#define X16R_8WAY 1
#define X16RV2_8WAY 1
#define X16RT_8WAY 1
#define X21S_8WAY 1
#elif defined(__AVX2__) && defined(__AES__)
#define X16RV2_4WAY 1
#define X16RT_4WAY 1
#define X21S_4WAY 1
#define X16R_4WAY 1
@@ -88,29 +89,23 @@ enum x16r_Algo {
X16R_HASH_FUNC_COUNT
};
extern __thread char x16r_hash_order[ X16R_HASH_FUNC_COUNT + 1 ];
//extern __thread char x16r_hash_order[ X16R_HASH_FUNC_COUNT + 1 ];
extern char x16r_hash_order[ X16R_HASH_FUNC_COUNT + 1 ];
extern void (*x16r_gate_get_hash_order) ( const struct work *, char * );
// x16r, x16rv2
void x16r_get_hash_order( const struct work *, char * );
// x16s, x21s
void x16s_get_hash_order( const struct work *, char * );
// x16rt
void x16rt_get_hash_order( const struct work *, char * );
extern void (*x16_r_s_getAlgoString) ( const uint8_t*, char* );
void x16r_getAlgoString( const uint8_t *prevblock, char *output );
void x16s_getAlgoString( const uint8_t *prevblock, char *output );
void x16rt_getAlgoString( const uint32_t *timeHash, char *output );
void x16rt_getTimeHash( const uint32_t timeStamp, void* timeHash );
bool register_x16r_algo( algo_gate_t* gate );
bool register_x16rv2_algo( algo_gate_t* gate );
bool register_x16s_algo( algo_gate_t* gate );
bool register_x16rt_algo( algo_gate_t* gate );
bool register_hex_algo( algo_gate_t* gate );
bool register_x21s_algo( algo_gate_t* gate );
bool register_hex__algo( algo_gate_t* gate );
bool register_x21s__algo( algo_gate_t* gate );
// x16r, x16s, x16rt
// x16r, x16s
#if defined(X16R_8WAY)
union _x16r_8way_context_overlay
@@ -141,15 +136,15 @@ union _x16r_8way_context_overlay
typedef union _x16r_8way_context_overlay x16r_8way_context_overlay;
extern x16r_8way_context_overlay x16r_ctx;
extern uint32_t x16r_8way_vdata[24*8] __attribute__ ((aligned (64)));
extern __thread x16r_8way_context_overlay x16r_ctx;
void x16r_8way_do_prehash( void *, const void * );
int x16r_8way_prehash( struct work * );
int x16r_8way_hash_generic( void *, const void *, const int );
int x16r_8way_hash( void *, const void *, const int );
void x16r_8way_prehash( void *, void * );
int x16r_8way_hash_generic( void *, const void *, int );
int x16r_8way_hash( void *, const void *, int );
int scanhash_x16r_8way( struct work *, uint32_t ,
uint64_t *, struct thr_info * );
extern __thread x16r_8way_context_overlay x16r_ctx;
#elif defined(X16R_4WAY)
@@ -182,15 +177,14 @@ union _x16r_4way_context_overlay
typedef union _x16r_4way_context_overlay x16r_4way_context_overlay;
extern x16r_4way_context_overlay x16r_ctx;
extern uint32_t x16r_4way_vdata[24*4] __attribute__ ((aligned (64)));
extern __thread x16r_4way_context_overlay x16r_ctx;
void x16r_4way_do_prehash( void *, const void * );
int x16r_4way_prehash( struct work * );
int x16r_4way_hash_generic( void *, const void *, const int );
int x16r_4way_hash( void *, const void *, const int );
void x16r_4way_prehash( void *, void * );
int x16r_4way_hash_generic( void *, const void *, int );
int x16r_4way_hash( void *, const void *, int );
int scanhash_x16r_4way( struct work *, uint32_t,
uint64_t *, struct thr_info * );
extern __thread x16r_4way_context_overlay x16r_ctx;
#endif
@@ -223,113 +217,80 @@ union _x16r_context_overlay
typedef union _x16r_context_overlay x16r_context_overlay;
extern x16r_context_overlay x16_ctx;
extern uint32_t x16r_edata[24] __attribute__ ((aligned (32)));
extern __thread x16r_context_overlay x16_ctx;
void x16r_do_prehash( const void * );
int x16r_prehash( const struct work * );
int x16r_hash_generic( void *, const void *, const int );
int x16r_hash( void *, const void *, const int );
void x16r_prehash( void *, void * );
int x16r_hash_generic( void *, const void *, int );
int x16r_hash( void *, const void *, int );
int scanhash_x16r( struct work *, uint32_t, uint64_t *, struct thr_info * );
// x16Rv2
#if defined(X16RV2_8WAY)
union _x16rv2_8way_context_overlay
{
blake512_8way_context blake;
bmw512_8way_context bmw;
skein512_8way_context skein;
jh512_8way_context jh;
keccak512_8way_context keccak;
luffa_4way_context luffa;
cubehashParam cube;
simd_4way_context simd;
hamsi512_8way_context hamsi;
hashState_fugue fugue;
shabal512_8way_context shabal;
sph_whirlpool_context whirlpool;
sha512_8way_context sha512;
sph_tiger_context tiger;
#if defined(__VAES__)
groestl512_4way_context groestl;
shavite512_4way_context shavite;
echo_4way_context echo;
#else
hashState_groestl groestl;
shavite512_context shavite;
hashState_echo echo;
#endif
} __attribute__ ((aligned (64)));
typedef union _x16rv2_8way_context_overlay x16rv2_8way_context_overlay;
extern x16rv2_8way_context_overlay x16rv2_ctx;
int x16rv2_8way_prehash( struct work * );
int x16rv2_8way_hash( void *state, const void *input, const int thrid );
//int scanhash_x16rv2_8way( struct work *work, uint32_t max_nonce,
// uint64_t *hashes_done, struct thr_info *mythr );
int x16rv2_8way_hash( void *state, const void *input, int thrid );
int scanhash_x16rv2_8way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
#elif defined(X16RV2_4WAY)
union _x16rv2_4way_context_overlay
{
blake512_4way_context blake;
bmw512_4way_context bmw;
#if defined(__VAES__)
groestl512_2way_context groestl;
shavite512_2way_context shavite;
echo_2way_context echo;
int x16rv2_4way_hash( void *state, const void *input, int thrid );
int scanhash_x16rv2_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
#else
hashState_groestl groestl;
shavite512_context shavite;
hashState_echo echo;
int x16rv2_hash( void *state, const void *input, int thr_id );
int scanhash_x16rv2( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
#endif
skein512_4way_context skein;
jh512_4way_context jh;
keccak512_4way_context keccak;
luffa_2way_context luffa;
cubehashParam cube;
simd_2way_context simd;
hamsi512_4way_context hamsi;
hashState_fugue fugue;
shabal512_4way_context shabal;
sph_whirlpool_context whirlpool;
sha512_4way_context sha512;
sph_tiger_context tiger;
};
typedef union _x16rv2_4way_context_overlay x16rv2_4way_context_overlay;
extern x16rv2_4way_context_overlay x16rv2_ctx;
// x16rt, veil
#if defined(X16R_8WAY)
int x16rv2_4way_hash( void *state, const void *input, const int thrid );
int x16rv2_4way_prehash( struct work * );
//void x16rt_8way_hash( void *state, const void *input );
int scanhash_x16rt_8way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
#elif defined(X16R_4WAY)
//void x16rt_4way_hash( void *state, const void *input );
int scanhash_x16rt_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
#else
int x16rv2_hash( void *state, const void *input, const int thr_id );
int x16rv2_prehash( const struct work * );
//void x16rt_hash( void *state, const void *input );
int scanhash_x16rt( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
#endif
// x21s
#if defined(X16R_8WAY)
int x21s_8way_hash( void *state, const void *input, const int thrid );
int x21s_8way_hash( void *state, const void *input, int thrid );
int scanhash_x21s_8way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
bool x21s_8way_thread_init();
#elif defined(X16R_4WAY)
int x21s_4way_hash( void *state, const void *input, const int thrid );
int x21s_4way_hash( void *state, const void *input, int thrid );
int scanhash_x21s_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
bool x21s_4way_thread_init();
#else
int x21s_hash( void *state, const void *input, const int thr_id );
int x21s_hash( void *state, const void *input, int thr_id );
int scanhash_x21s( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
bool x21s_thread_init();
#endif
//void hex_hash( void *state, const void *input );
int scanhash_hex( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );

36
algo/x16/x16r.c
View File

@@ -10,7 +10,7 @@
#include <stdlib.h>
#include <string.h>
void x16r_do_prehash( const void *edata )
void x16r_prehash( void *edata, void *pdata )
{
const char elem = x16r_hash_order[0];
const uint8_t algo = elem >= 'A' ? elem - 'A' + 10 : elem - '0';
@@ -48,7 +48,7 @@ void x16r_do_prehash( const void *edata )
}
}
int x16r_hash_generic( void* output, const void* input, const int thrid )
int x16r_hash_generic( void* output, const void* input, int thrid )
{
uint32_t _ALIGN(128) hash[16];
x16r_context_overlay ctx;
@@ -192,15 +192,7 @@ int x16r_hash_generic( void* output, const void* input, const int thrid )
return true;
}
int x16r_prehash( const struct work *work )
{
mm128_bswap32_80( x16r_edata, work->data );
x16r_gate_get_hash_order( work, x16r_hash_order );
x16r_do_prehash( x16r_edata );
return 1;
}
int x16r_hash( void* output, const void* input, const int thrid )
int x16r_hash( void* output, const void* input, int thrid )
{
uint8_t hash[64] __attribute__ ((aligned (64)));
if ( !x16r_hash_generic( hash, input, thrid ) )
@@ -213,8 +205,8 @@ int x16r_hash( void* output, const void* input, const int thrid )
int scanhash_x16r( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t _ALIGN(32) hash32[8];
uint32_t _ALIGN(32) edata[20];
uint32_t _ALIGN(128) hash32[8];
uint32_t _ALIGN(128) edata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
@@ -224,14 +216,24 @@ int scanhash_x16r( struct work *work, uint32_t max_nonce,
const bool bench = opt_benchmark;
if ( bench ) ptarget[7] = 0x0cff;
pthread_rwlock_rdlock( &g_work_lock );
memcpy( edata, x16r_edata, sizeof edata );
pthread_rwlock_unlock( &g_work_lock );
mm128_bswap32_80( edata, pdata );
static __thread uint32_t s_ntime = UINT32_MAX;
if ( s_ntime != pdata[17] )
{
uint32_t ntime = swab32(pdata[17]);
x16_r_s_getAlgoString( (const uint8_t*)(&edata[1]), x16r_hash_order );
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_DEBUG, "hash order %s (%08x)", x16r_hash_order, ntime );
}
x16r_prehash( edata, pdata );
do
{
edata[19] = nonce;
if ( algo_gate.hash( hash32, edata, thr_id ) )
if ( x16r_hash( hash32, edata, thr_id ) )
if ( unlikely( valid_hash( hash32, ptarget ) && !bench ) )
{
pdata[19] = bswap_32( nonce );

113
algo/x16/x16rt-4way.c Normal file
View File

@@ -0,0 +1,113 @@
#include "x16r-gate.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#if defined (X16R_8WAY)
int scanhash_x16rt_8way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t hash[16*8] __attribute__ ((aligned (128)));
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t _ALIGN(64) timeHash[8*8];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const uint32_t last_nonce = max_nonce - 8;
uint32_t n = first_nonce;
__m512i *noncev = (__m512i*)vdata + 9; // aligned
const int thr_id = mythr->id;
volatile uint8_t *restart = &(work_restart[thr_id].restart);
const bool bench = opt_benchmark;
if ( bench ) ptarget[7] = 0x0cff;
static __thread uint32_t s_ntime = UINT32_MAX;
uint32_t masked_ntime = bswap_32( pdata[17] ) & 0xffffff80;
if ( s_ntime != masked_ntime )
{
x16rt_getTimeHash( masked_ntime, &timeHash );
x16rt_getAlgoString( &timeHash[0], x16r_hash_order );
s_ntime = masked_ntime;
if ( !thr_id )
applog( LOG_INFO, "Hash order %s, Ntime %08x, time hash %08x",
x16r_hash_order, bswap_32( pdata[17] ), timeHash );
}
x16r_8way_prehash( vdata, pdata );
*noncev = mm512_intrlv_blend_32( _mm512_set_epi32(
n+7, 0, n+6, 0, n+5, 0, n+4, 0,
n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
do
{
if ( x16r_8way_hash( hash, vdata, thr_id ) )
for ( int i = 0; i < 8; i++ )
if ( unlikely( valid_hash( hash + (i<<3), ptarget ) && !bench ) )
{
pdata[19] = bswap_32( n+i );
submit_solution( work, hash+(i<<3), mythr );
}
*noncev = _mm512_add_epi32( *noncev,
m512_const1_64( 0x0000000800000000 ) );
n += 8;
} while ( likely( ( n < last_nonce ) && !(*restart) ) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}
#elif defined (X16R_4WAY)
int scanhash_x16rt_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t hash[4*16] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t _ALIGN(64) timeHash[4*8];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const uint32_t last_nonce = max_nonce - 4;
uint32_t n = first_nonce;
const int thr_id = mythr->id;
__m256i *noncev = (__m256i*)vdata + 9; // aligned
volatile uint8_t *restart = &(work_restart[thr_id].restart);
const bool bench = opt_benchmark;
if ( bench ) ptarget[7] = 0x0cff;
static __thread uint32_t s_ntime = UINT32_MAX;
uint32_t masked_ntime = bswap_32( pdata[17] ) & 0xffffff80;
if ( s_ntime != masked_ntime )
{
x16rt_getTimeHash( masked_ntime, &timeHash );
x16rt_getAlgoString( &timeHash[0], x16r_hash_order );
s_ntime = masked_ntime;
if ( !thr_id )
applog( LOG_INFO, "Hash order %s, Ntime %08x, time hash %08x",
x16r_hash_order, bswap_32( pdata[17] ), timeHash );
}
x16r_4way_prehash( vdata, pdata );
*noncev = mm256_intrlv_blend_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
do
{
if ( x16r_4way_hash( hash, vdata, thr_id ) )
for ( int i = 0; i < 4; i++ )
if ( unlikely( valid_hash( hash + (i<<3), ptarget ) && !bench ) )
{
pdata[19] = bswap_32( n+i );
submit_solution( work, hash+(i<<3), mythr );
}
*noncev = _mm256_add_epi32( *noncev,
m256_const1_64( 0x0000000400000000 ) );
n += 4;
} while ( ( n < last_nonce ) && !(*restart) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}
#endif

53
algo/x16/x16rt.c Normal file
View File

@@ -0,0 +1,53 @@
#include "x16r-gate.h"
#if !defined(X16R_8WAY) && !defined(X16R_4WAY)
int scanhash_x16rt( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t _ALIGN(128) hash32[8];
uint32_t _ALIGN(128) edata[20];
uint32_t _ALIGN(64) timeHash[8];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const int thr_id = mythr->id;
uint32_t nonce = first_nonce;
volatile uint8_t *restart = &(work_restart[thr_id].restart);
const bool bench = opt_benchmark;
if ( bench ) ptarget[7] = 0x0cff;
mm128_bswap32_80( edata, pdata );
static __thread uint32_t s_ntime = UINT32_MAX;
uint32_t masked_ntime = swab32( pdata[17] ) & 0xffffff80;
if ( s_ntime != masked_ntime )
{
x16rt_getTimeHash( masked_ntime, &timeHash );
x16rt_getAlgoString( &timeHash[0], x16r_hash_order );
s_ntime = masked_ntime;
if ( opt_debug && !thr_id )
applog( LOG_INFO, "hash order: %s time: (%08x) time hash: (%08x)",
x16r_hash_order, swab32( pdata[17] ), timeHash );
}
x16r_prehash( edata, pdata );
do
{
edata[19] = nonce;
if ( x16r_hash( hash32, edata, thr_id ) )
if ( valid_hash( hash32, ptarget ) && !bench )
{
pdata[19] = bswap_32( nonce );
submit_solution( work, hash32, mythr );
}
nonce++;
} while ( nonce < max_nonce && !(*restart) );
pdata[19] = nonce;
*hashes_done = pdata[19] - first_nonce;
return 0;
}
#endif // !defined(X16R_8WAY) && !defined(X16R_4WAY)

336
algo/x16/x16rv2-4way.c
View File

@@ -12,73 +12,37 @@
#if defined (X16RV2_8WAY)
void x16rv2_8way_do_prehash( void *vdata, void *pdata )
union _x16rv2_8way_context_overlay
{
uint32_t vdata32[20*8] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
blake512_8way_context blake;
bmw512_8way_context bmw;
skein512_8way_context skein;
jh512_8way_context jh;
keccak512_8way_context keccak;
luffa_4way_context luffa;
cubehashParam cube;
simd_4way_context simd;
hamsi512_8way_context hamsi;
hashState_fugue fugue;
shabal512_8way_context shabal;
sph_whirlpool_context whirlpool;
sha512_8way_context sha512;
sph_tiger_context tiger;
#if defined(__VAES__)
groestl512_4way_context groestl;
shavite512_4way_context shavite;
echo_4way_context echo;
#else
hashState_groestl groestl;
shavite512_context shavite;
hashState_echo echo;
#endif
} __attribute__ ((aligned (64)));
const char elem = x16r_hash_order[0];
const uint8_t algo = elem >= 'A' ? elem - 'A' + 10 : elem - '0';
typedef union _x16rv2_8way_context_overlay x16rv2_8way_context_overlay;
static __thread x16rv2_8way_context_overlay x16rv2_ctx;
switch ( algo )
{
case JH:
mm512_bswap32_intrlv80_8x64( vdata, pdata );
jh512_8way_init( &x16rv2_ctx.jh );
jh512_8way_update( &x16rv2_ctx.jh, vdata, 64 );
break;
case KECCAK:
case LUFFA:
case SHA_512:
mm128_bswap32_80( edata, pdata );
sph_tiger_init( &x16rv2_ctx.tiger );
sph_tiger( &x16rv2_ctx.tiger, edata, 64 );
intrlv_8x64( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
break;
case SKEIN:
mm512_bswap32_intrlv80_8x64( vdata, pdata );
skein512_8way_init( &x16rv2_ctx.skein );
skein512_8way_update( &x16rv2_ctx.skein, vdata, 64 );
break;
case CUBEHASH:
mm128_bswap32_80( edata, pdata );
cubehashInit( &x16rv2_ctx.cube, 512, 16, 32 );
cubehashUpdate( &x16rv2_ctx.cube, (const byte*)edata, 64 );
intrlv_8x64( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
break;
case HAMSI:
mm512_bswap32_intrlv80_8x64( vdata, pdata );
hamsi512_8way_init( &x16rv2_ctx.hamsi );
hamsi512_8way_update( &x16rv2_ctx.hamsi, vdata, 64 );
break;
case SHABAL:
mm256_bswap32_intrlv80_8x32( vdata32, pdata );
shabal512_8way_init( &x16rv2_ctx.shabal );
shabal512_8way_update( &x16rv2_ctx.shabal, vdata32, 64 );
rintrlv_8x32_8x64( vdata, vdata32, 640 );
break;
case WHIRLPOOL:
mm128_bswap32_80( edata, pdata );
sph_whirlpool_init( &x16rv2_ctx.whirlpool );
sph_whirlpool( &x16rv2_ctx.whirlpool, edata, 64 );
intrlv_8x64( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
break;
default:
mm512_bswap32_intrlv80_8x64( vdata, pdata );
}
}
int x16rv2_8way_prehash( struct work *work )
{
x16r_gate_get_hash_order( work, x16r_hash_order );
x16rv2_8way_do_prehash( x16r_8way_vdata, work->data );
return 1;
}
int x16rv2_8way_hash( void* output, const void* input, const int thrid )
int x16rv2_8way_hash( void* output, const void* input, int thrid )
{
uint32_t vhash[24*8] __attribute__ ((aligned (128)));
uint32_t hash0[24] __attribute__ ((aligned (32)));
@@ -593,28 +557,50 @@ int x16rv2_8way_hash( void* output, const void* input, const int thrid )
return 1;
}
#elif defined (X16RV2_4WAY)
// Pad the 24 bytes tiger hash to 64 bytes
inline void padtiger512( uint32_t* hash )
int scanhash_x16rv2_8way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
for ( int i = 6; i < 16; i++ ) hash[i] = 0;
}
void x16rv2_4way_do_prehash( void *vdata, void *pdata )
{
uint32_t vdata32[20*4] __attribute__ ((aligned (64)));
uint32_t hash[16*8] __attribute__ ((aligned (128)));
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t vdata2[20*8] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t bedata1[2] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const uint32_t last_nonce = max_nonce - 8;
uint32_t n = first_nonce;
__m512i *noncev = (__m512i*)vdata + 9; // aligned
const int thr_id = mythr->id;
volatile uint8_t *restart = &(work_restart[thr_id].restart);
const bool bench = opt_benchmark;
if ( bench ) ptarget[7] = 0x0cff;
mm512_bswap32_intrlv80_8x64( vdata, pdata );
bedata1[0] = bswap_32( pdata[1] );
bedata1[1] = bswap_32( pdata[2] );
static __thread uint32_t s_ntime = UINT32_MAX;
const uint32_t ntime = bswap_32( pdata[17] );
if ( s_ntime != ntime )
{
x16_r_s_getAlgoString( (const uint8_t*)bedata1, x16r_hash_order );
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_INFO, "hash order %s (%08x)", x16r_hash_order, ntime );
}
// Do midstate prehash on hash functions with block size <= 64 bytes.
const char elem = x16r_hash_order[0];
const uint8_t algo = elem >= 'A' ? elem - 'A' + 10 : elem - '0';
switch ( algo )
{
case JH:
mm256_bswap32_intrlv80_4x64( vdata, pdata );
jh512_4way_init( &x16rv2_ctx.jh );
jh512_4way_update( &x16rv2_ctx.jh, vdata, 64 );
mm512_bswap32_intrlv80_8x64( vdata, pdata );
jh512_8way_init( &x16rv2_ctx.jh );
jh512_8way_update( &x16rv2_ctx.jh, vdata, 64 );
break;
case KECCAK:
case LUFFA:
@@ -622,45 +608,100 @@ void x16rv2_4way_do_prehash( void *vdata, void *pdata )
mm128_bswap32_80( edata, pdata );
sph_tiger_init( &x16rv2_ctx.tiger );
sph_tiger( &x16rv2_ctx.tiger, edata, 64 );
intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
intrlv_8x64( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
break;
case SKEIN:
mm256_bswap32_intrlv80_4x64( vdata, pdata );
skein512_4way_prehash64( &x16r_ctx.skein, vdata );
mm512_bswap32_intrlv80_8x64( vdata, pdata );
skein512_8way_init( &x16rv2_ctx.skein );
skein512_8way_update( &x16rv2_ctx.skein, vdata, 64 );
break;
case CUBEHASH:
mm128_bswap32_80( edata, pdata );
cubehashInit( &x16rv2_ctx.cube, 512, 16, 32 );
cubehashUpdate( &x16rv2_ctx.cube, (const byte*)edata, 64 );
intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
intrlv_8x64( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
break;
case HAMSI:
mm256_bswap32_intrlv80_4x64( vdata, pdata );
hamsi512_4way_init( &x16rv2_ctx.hamsi );
hamsi512_4way_update( &x16rv2_ctx.hamsi, vdata, 64 );
mm512_bswap32_intrlv80_8x64( vdata, pdata );
hamsi512_8way_init( &x16rv2_ctx.hamsi );
hamsi512_8way_update( &x16rv2_ctx.hamsi, vdata, 64 );
break;
case SHABAL:
mm128_bswap32_intrlv80_4x32( vdata32, pdata );
shabal512_4way_init( &x16rv2_ctx.shabal );
shabal512_4way_update( &x16rv2_ctx.shabal, vdata32, 64 );
rintrlv_4x32_4x64( vdata, vdata32, 640 );
mm256_bswap32_intrlv80_8x32( vdata2, pdata );
shabal512_8way_init( &x16rv2_ctx.shabal );
shabal512_8way_update( &x16rv2_ctx.shabal, vdata2, 64 );
rintrlv_8x32_8x64( vdata, vdata2, 640 );
break;
case WHIRLPOOL:
mm128_bswap32_80( edata, pdata );
sph_whirlpool_init( &x16rv2_ctx.whirlpool );
sph_whirlpool( &x16rv2_ctx.whirlpool, edata, 64 );
intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
intrlv_8x64( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
break;
default:
mm256_bswap32_intrlv80_4x64( vdata, pdata );
mm512_bswap32_intrlv80_8x64( vdata, pdata );
}
}
*noncev = mm512_intrlv_blend_32( _mm512_set_epi32(
n+7, 0, n+6, 0, n+5, 0, n+4, 0,
n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
do
{
if ( x16rv2_8way_hash( hash, vdata, thr_id ) )
for ( int i = 0; i < 8; i++ )
if ( unlikely( valid_hash( hash + (i<<3), ptarget ) && !bench ) )
{
pdata[19] = bswap_32( n+i );
submit_solution( work, hash+(i<<3), mythr );
}
*noncev = _mm512_add_epi32( *noncev,
m512_const1_64( 0x0000000800000000 ) );
n += 8;
} while ( likely( ( n < last_nonce ) && !(*restart) ) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}
int x16rv2_4way_prehash( struct work *work )
#elif defined (X16RV2_4WAY)
union _x16rv2_4way_context_overlay
{
x16r_gate_get_hash_order( work, x16r_hash_order );
x16rv2_4way_do_prehash( x16r_4way_vdata, work->data );
return 1;
blake512_4way_context blake;
bmw512_4way_context bmw;
#if defined(__VAES__)
groestl512_2way_context groestl;
shavite512_2way_context shavite;
echo_2way_context echo;
#else
hashState_groestl groestl;
shavite512_context shavite;
hashState_echo echo;
#endif
skein512_4way_context skein;
jh512_4way_context jh;
keccak512_4way_context keccak;
luffa_2way_context luffa;
cubehashParam cube;
simd_2way_context simd;
hamsi512_4way_context hamsi;
hashState_fugue fugue;
shabal512_4way_context shabal;
sph_whirlpool_context whirlpool;
sha512_4way_context sha512;
sph_tiger_context tiger;
};
typedef union _x16rv2_4way_context_overlay x16rv2_4way_context_overlay;
static __thread x16rv2_4way_context_overlay x16rv2_ctx;
// Pad the 24 bytes tiger hash to 64 bytes
inline void padtiger512( uint32_t* hash )
{
for ( int i = 6; i < 16; i++ ) hash[i] = 0;
}
int x16rv2_4way_hash( void* output, const void* input, int thrid )
@@ -1007,4 +1048,107 @@ int x16rv2_4way_hash( void* output, const void* input, int thrid )
return 1;
}
int scanhash_x16rv2_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t hash[4*16] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t vdata32[20*4] __attribute__ ((aligned (64)));
uint32_t edata[20];
uint32_t bedata1[2];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const uint32_t last_nonce = max_nonce - 4;
uint32_t n = first_nonce;
const int thr_id = mythr->id;
__m256i *noncev = (__m256i*)vdata + 9;
volatile uint8_t *restart = &(work_restart[thr_id].restart);
const bool bench = opt_benchmark;
if ( bench ) ptarget[7] = 0x0fff;
bedata1[0] = bswap_32( pdata[1] );
bedata1[1] = bswap_32( pdata[2] );
static __thread uint32_t s_ntime = UINT32_MAX;
const uint32_t ntime = bswap_32(pdata[17]);
if ( s_ntime != ntime )
{
x16_r_s_getAlgoString( (const uint8_t*)bedata1, x16r_hash_order );
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_INFO, "hash order %s (%08x)", x16r_hash_order, ntime );
}
// Do midstate prehash on hash functions with block size <= 64 bytes.
const char elem = x16r_hash_order[0];
const uint8_t algo = elem >= 'A' ? elem - 'A' + 10 : elem - '0';
switch ( algo )
{
case JH:
mm256_bswap32_intrlv80_4x64( vdata, pdata );
jh512_4way_init( &x16rv2_ctx.jh );
jh512_4way_update( &x16rv2_ctx.jh, vdata, 64 );
break;
case KECCAK:
case LUFFA:
case SHA_512:
mm128_bswap32_80( edata, pdata );
sph_tiger_init( &x16rv2_ctx.tiger );
sph_tiger( &x16rv2_ctx.tiger, edata, 64 );
intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
break;
case SKEIN:
mm256_bswap32_intrlv80_4x64( vdata, pdata );
skein512_4way_prehash64( &x16r_ctx.skein, vdata );
break;
case CUBEHASH:
mm128_bswap32_80( edata, pdata );
cubehashInit( &x16rv2_ctx.cube, 512, 16, 32 );
cubehashUpdate( &x16rv2_ctx.cube, (const byte*)edata, 64 );
intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
break;
case HAMSI:
mm256_bswap32_intrlv80_4x64( vdata, pdata );
hamsi512_4way_init( &x16rv2_ctx.hamsi );
hamsi512_4way_update( &x16rv2_ctx.hamsi, vdata, 64 );
break;
case SHABAL:
mm128_bswap32_intrlv80_4x32( vdata32, pdata );
shabal512_4way_init( &x16rv2_ctx.shabal );
shabal512_4way_update( &x16rv2_ctx.shabal, vdata32, 64 );
rintrlv_4x32_4x64( vdata, vdata32, 640 );
break;
case WHIRLPOOL:
mm128_bswap32_80( edata, pdata );
sph_whirlpool_init( &x16rv2_ctx.whirlpool );
sph_whirlpool( &x16rv2_ctx.whirlpool, edata, 64 );
intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
break;
default:
mm256_bswap32_intrlv80_4x64( vdata, pdata );
}
*noncev = mm256_intrlv_blend_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
do
{
if ( x16rv2_4way_hash( hash, vdata, thr_id ) )
for ( int i = 0; i < 4; i++ )
if ( unlikely( valid_hash( hash + (i<<3), ptarget ) && !bench ) )
{
pdata[19] = bswap_32( n+i );
submit_solution( work, hash+(i<<3), mythr );
}
*noncev = _mm256_add_epi32( *noncev,
m256_const1_64( 0x0000000400000000 ) );
n += 4;
} while ( likely( ( n < last_nonce ) && !(*restart) ) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}
#endif

59
algo/x16/x16rv2.c
View File

@@ -43,16 +43,9 @@ inline void padtiger512(uint32_t* hash) {
for (int i = (24/4); i < (64/4); i++) hash[i] = 0;
}
// no prehash
int x16rv2_prehash( const struct work *work )
int x16rv2_hash( void* output, const void* input, int thrid )
{
x16r_gate_get_hash_order( work, x16r_hash_order );
return 1;
}
int x16rv2_hash( void* output, const void* input, const int thrid )
{
uint32_t _ALIGN(32) hash[16];
uint32_t _ALIGN(128) hash[16];
x16rv2_context_overlay ctx;
void *in = (void*) input;
int size = 80;
@@ -177,4 +170,52 @@ int x16rv2_hash( void* output, const void* input, const int thrid )
return 1;
}
int scanhash_x16rv2( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t _ALIGN(128) hash32[8];
uint32_t _ALIGN(128) edata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const int thr_id = mythr->id;
uint32_t nonce = first_nonce;
volatile uint8_t *restart = &(work_restart[thr_id].restart);
const bool bench = opt_benchmark;
casti_m128i( edata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
casti_m128i( edata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
casti_m128i( edata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
casti_m128i( edata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
casti_m128i( edata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
static __thread uint32_t s_ntime = UINT32_MAX;
if ( s_ntime != pdata[17] )
{
uint32_t ntime = swab32(pdata[17]);
x16_r_s_getAlgoString( (const uint8_t*) (&edata[1]), x16r_hash_order );
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_DEBUG, "hash order %s (%08x)",
x16r_hash_order, ntime );
}
if ( bench ) ptarget[7] = 0x0cff;
do
{
edata[19] = nonce;
if ( x16rv2_hash( hash32, edata, thr_id ) )
if ( unlikely( valid_hash( hash32, ptarget ) && !bench ) )
{
pdata[19] = bswap_32( nonce );
submit_solution( work, hash32, mythr );
}
nonce++;
} while ( nonce < max_nonce && !(*restart) );
pdata[19] = nonce;
*hashes_done = pdata[19] - first_nonce;
return 0;
}
#endif

116
algo/x16/x21s-4way.c
View File

@@ -30,7 +30,7 @@ union _x21s_8way_context_overlay
typedef union _x21s_8way_context_overlay x21s_8way_context_overlay;
int x21s_8way_hash( void* output, const void* input, const int thrid )
int x21s_8way_hash( void* output, const void* input, int thrid )
{
uint32_t vhash[16*8] __attribute__ ((aligned (128)));
uint8_t shash[64*8] __attribute__ ((aligned (64)));
@@ -129,6 +129,66 @@ int x21s_8way_hash( void* output, const void* input, const int thrid )
return 1;
}
int scanhash_x21s_8way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t hash[16*8] __attribute__ ((aligned (128)));
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &hash[7<<3];
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t bedata1[2] __attribute__((aligned(64)));
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;
const uint32_t last_nonce = max_nonce - 16;
const int thr_id = mythr->id;
__m512i *noncev = (__m512i*)vdata + 9; // aligned
volatile uint8_t *restart = &(work_restart[thr_id].restart);
const bool bench = opt_benchmark;
if ( bench ) ptarget[7] = 0x0cff;
bedata1[0] = bswap_32( pdata[1] );
bedata1[1] = bswap_32( pdata[2] );
static __thread uint32_t s_ntime = UINT32_MAX;
uint32_t ntime = bswap_32( pdata[17] );
if ( s_ntime != ntime )
{
x16_r_s_getAlgoString( (const uint8_t*)bedata1, x16r_hash_order );
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_INFO, "hash order %s (%08x)", x16r_hash_order, ntime );
}
x16r_8way_prehash( vdata, pdata );
*noncev = mm512_intrlv_blend_32( _mm512_set_epi32(
n+7, 0, n+6, 0, n+5, 0, n+4, 0,
n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
do
{
if ( x21s_8way_hash( hash, vdata, thr_id ) )
for ( int lane = 0; lane < 8; lane++ )
if ( unlikely( hash7[lane] <= Htarg ) )
{
extr_lane_8x32( lane_hash, hash, lane, 256 );
if ( likely( valid_hash( lane_hash, ptarget ) && !bench ) )
{
pdata[19] = bswap_32( n + lane );
submit_solution( work, lane_hash, mythr );
}
}
*noncev = _mm512_add_epi32( *noncev,
m512_const1_64( 0x0000000800000000 ) );
n += 8;
} while ( likely( ( n < last_nonce ) && !(*restart) ) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}
bool x21s_8way_thread_init()
{
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * 4; // nCols
@@ -155,7 +215,7 @@ union _x21s_4way_context_overlay
typedef union _x21s_4way_context_overlay x21s_4way_context_overlay;
int x21s_4way_hash( void* output, const void* input, const int thrid )
int x21s_4way_hash( void* output, const void* input, int thrid )
{
uint32_t vhash[16*4] __attribute__ ((aligned (64)));
uint8_t shash[64*4] __attribute__ ((aligned (64)));
@@ -231,6 +291,58 @@ int x21s_4way_hash( void* output, const void* input, const int thrid )
return 1;
}
int scanhash_x21s_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t hash[16*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t bedata1[2] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const uint32_t last_nonce = max_nonce - 4;
uint32_t n = first_nonce;
const int thr_id = mythr->id;
const bool bench = opt_benchmark;
__m256i *noncev = (__m256i*)vdata + 9; // aligned
volatile uint8_t *restart = &(work_restart[thr_id].restart);
if ( bench ) ptarget[7] = 0x0cff;
bedata1[0] = bswap_32( pdata[1] );
bedata1[1] = bswap_32( pdata[2] );
static __thread uint32_t s_ntime = UINT32_MAX;
uint32_t ntime = bswap_32( pdata[17] );
if ( s_ntime != ntime )
{
x16_r_s_getAlgoString( (const uint8_t*)bedata1, x16r_hash_order );
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_DEBUG, "hash order %s (%08x)", x16r_hash_order, ntime );
}
x16r_4way_prehash( vdata, pdata );
*noncev = mm256_intrlv_blend_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
do
{
if ( x21s_4way_hash( hash, vdata, thr_id ) )
for ( int i = 0; i < 4; i++ )
if ( unlikely( valid_hash( hash + (i<<3), ptarget ) && !bench ) )
{
pdata[19] = bswap_32( n+i );
submit_solution( work, hash+(i<<3), mythr );
}
*noncev = _mm256_add_epi32( *noncev,
m256_const1_64( 0x0000000400000000 ) );
n += 4;
} while ( likely( ( n < last_nonce ) && !(*restart) ) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}
bool x21s_4way_thread_init()
{
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * 4; // nCols

46
algo/x16/x21s.c
View File

@@ -27,7 +27,7 @@ union _x21s_context_overlay
};
typedef union _x21s_context_overlay x21s_context_overlay;
int x21s_hash( void* output, const void* input, const int thrid )
int x21s_hash( void* output, const void* input, int thrid )
{
uint32_t _ALIGN(128) hash[16];
x21s_context_overlay ctx;
@@ -57,6 +57,50 @@ int x21s_hash( void* output, const void* input, const int thrid )
return 1;
}
int scanhash_x21s( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t _ALIGN(128) hash32[8];
uint32_t _ALIGN(128) edata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const int thr_id = mythr->id;
uint32_t nonce = first_nonce;
volatile uint8_t *restart = &(work_restart[thr_id].restart);
const bool bench = opt_benchmark;
if ( bench ) ptarget[7] = 0x0cff;
mm128_bswap32_80( edata, pdata );
static __thread uint32_t s_ntime = UINT32_MAX;
if ( s_ntime != pdata[17] )
{
uint32_t ntime = swab32(pdata[17]);
x16_r_s_getAlgoString( (const uint8_t*)(&edata[1]), x16r_hash_order );
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_INFO, "hash order %s (%08x)", x16r_hash_order, ntime );
}
x16r_prehash( edata, pdata );
do
{
edata[19] = nonce;
if ( x21s_hash( hash32, edata, thr_id ) )
if ( unlikely( valid_hash( hash32, ptarget ) && !bench ) )
{
pdata[19] = bswap_32( nonce );
submit_solution( work, hash32, mythr );
}
nonce++;
} while ( nonce < max_nonce && !(*restart) );
pdata[19] = nonce;
*hashes_done = pdata[19] - first_nonce;
return 0;
}
bool x21s_thread_init()
{
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * 4; // nCols