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

v23.12

Browse Source
This commit is contained in:
Jay D Dee
2023-11-20 11:51:57 -05:00
parent f3fde95f27
commit fc696dbbe5
36 changed files with 3597 additions and 3152 deletions
Download Patch File Download Diff File Expand all files Collapse all files

60
algo/x16/hex.c
View File

@@ -23,13 +23,12 @@ static void hex_getAlgoString(const uint32_t* prevblock, char *output)
*sptr = '\0';
}
static __thread x16r_context_overlay hex_ctx;
int hex_hash( void* output, const void* input, int thrid )
{
uint32_t _ALIGN(128) hash[16];
x16r_context_overlay ctx;
memcpy( &ctx, &hex_ctx, sizeof(ctx) );
memcpy( &ctx, &x16r_ref_ctx, sizeof(ctx) );
void *in = (void*) input;
int size = 80;
@@ -52,7 +51,7 @@ int hex_hash( void* output, const void* input, int thrid )
break;
case GROESTL:
#if defined(__AES__)
groestl512_full( &ctx.groestl, (char*)hash, (char*)in, size<<3 );
groestl512_full( &ctx.groestl, hash, in, size<<3 );
#else
sph_groestl512_init( &ctx.groestl );
sph_groestl512( &ctx.groestl, in, size );
@@ -87,7 +86,7 @@ int hex_hash( void* output, const void* input, int thrid )
case LUFFA:
if ( i == 0 )
{
update_and_final_luffa( &ctx.luffa, hash, (const void*)in+64, 16 );
update_and_final_luffa( &ctx.luffa, hash, in+64, 16 );
}
else
{
@@ -97,7 +96,7 @@ int hex_hash( void* output, const void* input, int thrid )
break;
case CUBEHASH:
if ( i == 0 )
cubehashUpdateDigest( &ctx.cube, hash, (const void*)in+64, 16 );
cubehashUpdateDigest( &ctx.cube, hash, in+64, 16 );
else
{
cubehashInit( &ctx.cube, 512, 16, 32 );
@@ -108,26 +107,15 @@ int hex_hash( void* output, const void* input, int thrid )
shavite512_full( &ctx.shavite, hash, in, size );
break;
case SIMD:
#if defined(__aarch64__)
sph_simd512_init( &ctx.simd );
sph_simd512(&ctx.simd, (const void*) hash, 64);
sph_simd512_close(&ctx.simd, hash);
#else
simd_full( &ctx.simd, (BitSequence *)hash,
(const BitSequence*)in, size<<3 );
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence*)in, size<<3 );
#endif
simd512_ctx( &ctx.simd, hash, in, size<<3 );
break;
case ECHO:
#if defined(__AES__)
echo_full( &ctx.echo, (BitSequence *)hash, 512,
(const BitSequence *)in, size );
#if defined(__AES__) || defined(__ARM_FEATURE_AES)
echo_full( &ctx.echo, hash, 512, in, size );
#else
sph_echo512_init( &ctx.echo );
sph_echo512( &ctx.echo, in, size );
sph_echo512_close( &ctx.echo, hash );
sph_echo512_init( &ctx.echo );
sph_echo512( &ctx.echo, in, size );
sph_echo512_close( &ctx.echo, hash );
#endif
break;
case HAMSI:
@@ -216,32 +204,32 @@ int scanhash_hex( struct work *work, uint32_t max_nonce,
switch ( algo )
{
case JH:
sph_jh512_init( &hex_ctx.jh );
sph_jh512( &hex_ctx.jh, edata, 64 );
sph_jh512_init( &x16r_ref_ctx.jh );
sph_jh512( &x16r_ref_ctx.jh, edata, 64 );
break;
case SKEIN:
sph_skein512_init( &hex_ctx.skein );
sph_skein512( &hex_ctx.skein, edata, 64 );
sph_skein512_init( &x16r_ref_ctx.skein );
sph_skein512( &x16r_ref_ctx.skein, edata, 64 );
break;
case LUFFA:
init_luffa( &hex_ctx.luffa, 512 );
update_luffa( &hex_ctx.luffa, edata, 64 );
init_luffa( &x16r_ref_ctx.luffa, 512 );
update_luffa( &x16r_ref_ctx.luffa, edata, 64 );
break;
case CUBEHASH:
cubehashInit( &hex_ctx.cube, 512, 16, 32 );
cubehashUpdate( &hex_ctx.cube, edata, 64 );
cubehashInit( &x16r_ref_ctx.cube, 512, 16, 32 );
cubehashUpdate( &x16r_ref_ctx.cube, edata, 64 );
break;
case HAMSI:
sph_hamsi512_init( &hex_ctx.hamsi );
sph_hamsi512( &hex_ctx.hamsi, edata, 64 );
sph_hamsi512_init( &x16r_ref_ctx.hamsi );
sph_hamsi512( &x16r_ref_ctx.hamsi, edata, 64 );
break;
case SHABAL:
sph_shabal512_init( &hex_ctx.shabal );
sph_shabal512( &hex_ctx.shabal, edata, 64 );
sph_shabal512_init( &x16r_ref_ctx.shabal );
sph_shabal512( &x16r_ref_ctx.shabal, edata, 64 );
break;
case WHIRLPOOL:
sph_whirlpool_init( &hex_ctx.whirlpool );
sph_whirlpool( &hex_ctx.whirlpool, edata, 64 );
sph_whirlpool_init( &x16r_ref_ctx.whirlpool );
sph_whirlpool( &x16r_ref_ctx.whirlpool, edata, 64 );
break;
}

16
algo/x16/minotaur.c
View File

@@ -14,9 +14,9 @@
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/simd-hash-2way.h"
#if defined(__aarch64__)
#include "algo/simd/sph_simd.h"
#endif
//#if defined(__aarch64__)
// #include "algo/simd/sph_simd.h"
//#endif
#include "algo/hamsi/sph_hamsi.h"
#include "algo/shabal/sph_shabal.h"
#include "algo/whirlpool/sph_whirlpool.h"
@@ -24,10 +24,14 @@
#include "algo/yespower/yespower.h"
#if defined(__AES__) || defined(__ARM_FEATURE_AES)
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/groestl/aes_ni/hash-groestl.h"
#else
#include "algo/echo/sph_echo.h"
#endif
#if defined(__AES__) // || defined(__ARM_FEATURE_AES)
#include "algo/groestl/aes_ni/hash-groestl.h"
#else
#include "algo/groestl/sph_groestl.h"
#endif
#include "algo/echo/sph_echo.h"
#include "algo/groestl/sph_groestl.h"
#if defined(__AES__)
#include "algo/fugue/fugue-aesni.h"
#else

401
algo/x16/x16r-4way.c
View File

@@ -971,4 +971,405 @@ int scanhash_x16r_4way( struct work *work, uint32_t max_nonce,
return 0;
}
#elif defined (X16R_2WAY)
void x16r_2x64_prehash( void *vdata, void *pdata )
{
uint32_t edata[20] __attribute__ ((aligned (64)));
const char elem = x16r_hash_order[0];
const uint8_t algo = elem >= 'A' ? elem - 'A' + 10 : elem - '0';
switch ( algo )
{
case JH:
v128_bswap32_intrlv80_2x64( vdata, pdata );
jh512_2x64_init( &x16r_ctx.jh );
jh512_2x64_update( &x16r_ctx.jh, vdata, 64 );
break;
case KECCAK:
v128_bswap32_intrlv80_2x64( vdata, pdata );
keccak512_2x64_init( &x16r_ctx.keccak );
keccak512_2x64_update( &x16r_ctx.keccak, vdata, 72 );
break;
case SKEIN:
v128_bswap32_intrlv80_2x64( vdata, pdata );
skein512_2x64_prehash64( &x16r_ctx.skein, vdata );
break;
case LUFFA:
{
v128_bswap32_80( edata, pdata );
init_luffa( &x16r_ctx.luffa, 512 );
update_luffa( &x16r_ctx.luffa, edata, 64 );
intrlv_2x64( vdata, edata, edata, 640 );
}
break;
case CUBEHASH:
{
v128_bswap32_80( edata, pdata );
cubehashInit( &x16r_ctx.cube, 512, 16, 32 );
cubehashUpdate( &x16r_ctx.cube, edata, 64 );
intrlv_2x64( vdata, edata, edata, 640 );
}
break;
case HAMSI:
#if defined(__SSE4_2__) || defined(__ARM_NEON)
v128_bswap32_intrlv80_2x64( vdata, pdata );
hamsi512_2x64_init( &x16r_ctx.hamsi );
hamsi512_2x64_update( &x16r_ctx.hamsi, vdata, 72 );
#else
v128_bswap32_80( edata, pdata );
sph_hamsi512_init( &x16r_ctx.hamsi );
sph_hamsi512( &x16r_ctx.hamsi, edata, 72 );
intrlv_2x64( vdata, edata, edata, 640 );
#endif
break;
case FUGUE:
v128_bswap32_80( edata, pdata );
#if defined(__AES__)
fugue512_init( &x16r_ctx.fugue );
fugue512_update( &x16r_ctx.fugue, edata, 76 );
#else
sph_fugue512_init( &x16r_ctx.fugue );
sph_fugue512( &x16r_ctx.fugue, edata, 76 );
#endif
intrlv_2x64( vdata, edata, edata, 640 );
break;
case SHABAL:
v128_bswap32_80( edata, pdata );
sph_shabal512_init( &x16r_ctx.shabal );
sph_shabal512( &x16r_ctx.shabal, edata, 64);
intrlv_2x64( vdata, edata, edata, 640 );
break;
case WHIRLPOOL:
v128_bswap32_80( edata, pdata );
sph_whirlpool_init( &x16r_ctx.whirlpool );
sph_whirlpool( &x16r_ctx.whirlpool, edata, 64 );
intrlv_2x64( vdata, edata, edata, 640 );
break;
default:
v128_bswap32_intrlv80_2x64( vdata, pdata );
}
}
int x16r_2x64_hash_generic( void* output, const void* input, int thrid )
{
uint32_t vhash[20*2] __attribute__ ((aligned (64)));
uint32_t hash0[20] __attribute__ ((aligned (32)));
uint32_t hash1[20] __attribute__ ((aligned (32)));
x16r_2x64_context_overlay ctx;
memcpy( &ctx, &x16r_ctx, sizeof(ctx) );
void *in0 = (void*) hash0;
void *in1 = (void*) hash1;
int size = 80;
dintrlv_2x64( hash0, hash1, input, 640 );
for ( int i = 0; i < 16; i++ )
{
const char elem = x16r_hash_order[i];
const uint8_t algo = elem >= 'A' ? elem - 'A' + 10 : elem - '0';
switch ( algo )
{
case BLAKE:
if ( i == 0 )
blake512_2x64_full( &ctx.blake, vhash, input, size );
else
{
intrlv_2x64( vhash, in0, in1, size<<3 );
blake512_2x64_full( &ctx.blake, vhash, vhash, size );
}
dintrlv_2x64( hash0, hash1, vhash, 512 );
break;
case BMW:
bmw512_2x64_init( &ctx.bmw );
if ( i == 0 )
bmw512_2x64_update( &ctx.bmw, input, size );
else
{
intrlv_2x64( vhash, in0, in1, size<<3 );
bmw512_2x64_update( &ctx.bmw, vhash, size );
}
bmw512_2x64_close( &ctx.bmw, vhash );
dintrlv_2x64( hash0, hash1, vhash, 512 );
break;
case GROESTL:
#if defined(__AES__) // || defined(__ARM_FEATURE_AES)
groestl512_full( &ctx.groestl, hash0, in0, size<<3 );
groestl512_full( &ctx.groestl, hash1, in1, size<<3 );
#else
sph_groestl512_init( &ctx.groestl );
sph_groestl512( &ctx.groestl, in0, size );
sph_groestl512_close( &ctx.groestl, hash0 );
sph_groestl512_init( &ctx.groestl );
sph_groestl512( &ctx.groestl, in1, size );
sph_groestl512_close( &ctx.groestl, hash1 );
#endif
break;
case JH:
if ( i == 0 )
jh512_2x64_update( &ctx.jh, input + (64*2), 16 );
else
{
intrlv_2x64( vhash, in0, in1, size<<3 );
jh512_2x64_init( &ctx.jh );
jh512_2x64_update( &ctx.jh, vhash, size );
}
jh512_2x64_close( &ctx.jh, vhash );
dintrlv_2x64( hash0, hash1, vhash, 512 );
break;
case KECCAK:
if ( i == 0 )
keccak512_2x64_update( &ctx.keccak, input + (72*2), 8 );
else
{
intrlv_2x64( vhash, in0, in1, size<<3 );
keccak512_2x64_init( &ctx.keccak );
keccak512_2x64_update( &ctx.keccak, vhash, size );
}
keccak512_2x64_close( &ctx.keccak, vhash );
dintrlv_2x64( hash0, hash1, vhash, 512 );
break;
case SKEIN:
if ( i == 0 )
skein512_2x64_final16( &ctx.skein, vhash, input + (64*2) );
else
{
intrlv_2x64( vhash, in0, in1, size<<3 );
skein512_2x64_full( &ctx.skein, vhash, vhash, size );
}
dintrlv_2x64( hash0, hash1, vhash, 512 );
break;
case LUFFA:
if ( i == 0 )
{
update_and_final_luffa( &ctx.luffa, hash0, in0 + 64, 16 );
memcpy( &ctx, &x16r_ctx, sizeof(ctx) );
update_and_final_luffa( &ctx.luffa, hash1, in1 + 64, 16 );
}
else
{
luffa_full( &ctx.luffa, hash0, 512, hash0, size );
luffa_full( &ctx.luffa, hash1, 512, hash1, size );
}
break;
case CUBEHASH:
if ( i == 0 )
{
cubehashUpdateDigest( &ctx.cube, hash0, in0 + 64, 16 );
memcpy( &ctx, &x16r_ctx, sizeof(ctx) );
cubehashUpdateDigest( &ctx.cube, hash1, in1 + 64, 16 );
}
else
{
cubehash_full( &ctx.cube, hash0, 512, hash0, size );
cubehash_full( &ctx.cube, hash1, 512, hash1, size );
}
break;
case SHAVITE:
shavite512_full( &ctx.shavite, hash0, in0, size );
shavite512_full( &ctx.shavite, hash1, in1, size );
break;
case SIMD:
simd512_ctx( &ctx.simd, hash0, in0, size );
simd512_ctx( &ctx.simd, hash1, in1, size );
break;
case ECHO:
#if defined(__AES__)
echo_full( &ctx.echo, hash0, 512, in0, size );
echo_full( &ctx.echo, hash1, 512, in1, size );
#else
sph_echo512_init( &ctx.echo );
sph_echo512( &ctx.echo, in0, size );
sph_echo512_close( &ctx.echo, hash0 );
sph_echo512_init( &ctx.echo );
sph_echo512( &ctx.echo, in1, size );
sph_echo512_close( &ctx.echo, hash1 );
#endif
break;
case HAMSI:
#if defined(__SSE4_2__) || defined(__ARM_NEON)
if ( i == 0 )
hamsi512_2x64_update( &ctx.hamsi, input + (72*2), 8 );
else
{
intrlv_2x64( vhash, hash0, hash1, size<<3 );
hamsi512_2x64_init( &ctx.hamsi );
hamsi512_2x64_update( &ctx.hamsi, vhash, size );
}
hamsi512_2x64_close( &ctx.hamsi, vhash );
dintrlv_2x64( hash0, hash1, vhash, 512 );
#else
if ( i == 0 )
{
sph_hamsi512( &ctx.hamsi, in0 + 72, 8 );
sph_hamsi512_close( &ctx.hamsi, hash0 );
memcpy( &ctx, &x16r_ctx, sizeof(ctx) );
sph_hamsi512( &ctx.hamsi, in1 + 72, 8 );
sph_hamsi512_close( &ctx.hamsi, hash1 );
}
else
{
sph_hamsi512_init( &ctx.hamsi );
sph_hamsi512( &ctx.hamsi, hash0, size );
sph_hamsi512_close( &ctx.hamsi, hash0 );
sph_hamsi512_init( &ctx.hamsi );
sph_hamsi512( &ctx.hamsi, hash1, size );
sph_hamsi512_close( &ctx.hamsi, hash1 );
}
#endif
break;
case FUGUE:
#if defined(__AES__)
if ( i == 0 )
{
fugue512_update( &ctx.fugue, in0 + 76, 4 );
fugue512_final( &ctx.fugue, hash0 );
memcpy( &ctx, &x16r_ctx, sizeof(hashState_fugue) );
fugue512_update( &ctx.fugue, in1 + 76, 4 );
fugue512_final( &ctx.fugue, hash1 );
}
else
{
fugue512_full( &ctx.fugue, hash0, hash0, size );
fugue512_full( &ctx.fugue, hash1, hash1, size );
}
#else
if ( i == 0 )
{
sph_fugue512( &ctx.fugue, in0 + 76, 4 );
sph_fugue512_close( &ctx.fugue, hash0 );
memcpy( &ctx, &x16r_ctx, sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, in1 + 76, 4 );
sph_fugue512_close( &ctx.fugue, hash1 );
}
else
{
sph_fugue512_full( &ctx.fugue, hash0, hash0, size );
sph_fugue512_full( &ctx.fugue, hash1, hash1, size );
}
#endif
break;
case SHABAL:
if ( i == 0 )
{
sph_shabal512( &ctx.shabal, in0 + 64, 16 );
sph_shabal512_close( &ctx.shabal, hash0 );
memcpy( &ctx, &x16r_ctx, sizeof(ctx) );
sph_shabal512( &ctx.shabal, in1 + 64, 16 );
sph_shabal512_close( &ctx.shabal, hash1 );
}
else
{
sph_shabal512_init( &ctx.shabal );
sph_shabal512( &ctx.shabal, hash0, size );
sph_shabal512_close( &ctx.shabal, hash0 );
sph_shabal512_init( &ctx.shabal );
sph_shabal512( &ctx.shabal, hash1, size );
sph_shabal512_close( &ctx.shabal, hash1 );
}
break;
case WHIRLPOOL:
if ( i == 0 )
{
sph_whirlpool( &ctx.whirlpool, in0 + 64, 16 );
sph_whirlpool_close( &ctx.whirlpool, hash0 );
memcpy( &ctx, &x16r_ctx, sizeof(ctx) );
sph_whirlpool( &ctx.whirlpool, in1 + 64, 16 );
sph_whirlpool_close( &ctx.whirlpool, hash1 );
memcpy( &ctx, &x16r_ctx, sizeof(ctx) );
}
else
{
sph_whirlpool512_full( &ctx.whirlpool, hash0, hash0, size );
sph_whirlpool512_full( &ctx.whirlpool, hash1, hash1, size );
}
break;
case SHA_512:
sha512_2x64_init( &ctx.sha512 );
if ( i == 0 )
sha512_2x64_update( &ctx.sha512, input, size );
else
{
intrlv_2x64( vhash, in0, in1, size<<3 );
sha512_2x64_init( &ctx.sha512 );
sha512_2x64_update( &ctx.sha512, vhash, size );
}
sha512_2x64_close( &ctx.sha512, vhash );
dintrlv_2x64( hash0, hash1, vhash, 512 );
break;
}
if ( work_restart[thrid].restart ) return 0;
size = 64;
}
memcpy( output, hash0, 64 );
memcpy( output+64, hash1, 64 );
return 1;
}
int x16r_2x64_hash( void* output, const void* input, int thrid )
{
uint8_t hash[64*2] __attribute__ ((aligned (64)));
if ( !x16r_2x64_hash_generic( hash, input, thrid ) )
return 0;
memcpy( output, hash, 32 );
memcpy( output+32, hash+64, 32 );
return 1;
}
int scanhash_x16r_2x64( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t hash[16*2] __attribute__ ((aligned (64)));
uint32_t vdata[20*2] __attribute__ ((aligned (64)));
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 - 2;
uint32_t n = first_nonce;
v128_t *noncev = (v128_t*)vdata + 9;
const int thr_id = mythr->id;
const bool bench = opt_benchmark;
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;
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_2x64_prehash( vdata, pdata );
*noncev = v128_intrlv_blend_32( v128_set32( n+1, 0, n, 0 ), *noncev );
do
{
if ( x16r_2x64_hash( hash, vdata, thr_id ) );
for ( int i = 0; i < 2; i++ )
if ( unlikely( valid_hash( hash + (i<<3), ptarget ) && !bench ) )
{
pdata[19] = bswap_32( n+i );
submit_solution( work, hash+(i<<3), mythr );
}
*noncev = v128_add32( *noncev, v128_64( 0x0000000200000000 ) );
n += 2;
} while ( likely( ( n < last_nonce ) && !(*restart) ) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}
#endif

59
algo/x16/x16r-gate.c
View File

@@ -13,10 +13,13 @@ __thread x16r_8way_context_overlay x16r_ctx;
__thread x16r_4way_context_overlay x16r_ctx;
#elif defined (X16R_2WAY)
__thread x16r_2x64_context_overlay x16r_ctx;
#endif
__thread x16r_context_overlay x16_ctx;
__thread x16r_context_overlay x16r_ref_ctx;
void x16r_getAlgoString( const uint8_t* prevblock, char *output )
{
@@ -58,11 +61,15 @@ bool register_x16r_algo( algo_gate_t* gate )
#elif defined (X16R_4WAY)
gate->scanhash = (void*)&scanhash_x16r_4way;
gate->hash = (void*)&x16r_4way_hash;
#elif defined (X16R_2WAY)
gate->scanhash = (void*)&scanhash_x16r_2x64;
gate->hash = (void*)&x16r_2x64_hash;
#else
gate->scanhash = (void*)&scanhash_x16r;
gate->hash = (void*)&x16r_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT
| NEON_OPT;
x16_r_s_getAlgoString = (void*)&x16r_getAlgoString;
opt_target_factor = 256.0;
return true;
@@ -76,11 +83,15 @@ bool register_x16rv2_algo( algo_gate_t* gate )
#elif defined (X16RV2_4WAY)
gate->scanhash = (void*)&scanhash_x16rv2_4way;
gate->hash = (void*)&x16rv2_4way_hash;
#elif defined (X16RV2_2WAY)
gate->scanhash = (void*)&scanhash_x16rv2_2x64;
gate->hash = (void*)&x16rv2_2x64_hash;
#else
gate->scanhash = (void*)&scanhash_x16rv2;
gate->hash = (void*)&x16rv2_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT
| NEON_OPT;
x16_r_s_getAlgoString = (void*)&x16r_getAlgoString;
opt_target_factor = 256.0;
return true;
@@ -94,11 +105,15 @@ bool register_x16s_algo( algo_gate_t* gate )
#elif defined (X16R_4WAY)
gate->scanhash = (void*)&scanhash_x16r_4way;
gate->hash = (void*)&x16r_4way_hash;
#elif defined (X16R_2WAY)
gate->scanhash = (void*)&scanhash_x16r_2x64;
gate->hash = (void*)&x16r_2x64_hash;
#else
gate->scanhash = (void*)&scanhash_x16r;
gate->hash = (void*)&x16r_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT
| NEON_OPT;
x16_r_s_getAlgoString = (void*)&x16s_getAlgoString;
opt_target_factor = 256.0;
return true;
@@ -108,7 +123,6 @@ bool register_x16s_algo( algo_gate_t* gate )
//
// X16RT
void x16rt_getTimeHash( const uint32_t timeStamp, void* timeHash )
{
int32_t maskedTime = timeStamp & 0xffffff80;
@@ -221,34 +235,42 @@ void veil_build_extraheader( struct work* g_work, struct stratum_ctx* sctx )
bool register_x16rt_algo( algo_gate_t* gate )
{
#if defined (X16R_8WAY)
#if defined (X16RT_8WAY)
gate->scanhash = (void*)&scanhash_x16rt_8way;
gate->hash = (void*)&x16r_8way_hash;
#elif defined (X16R_4WAY)
#elif defined (X16RT_4WAY)
gate->scanhash = (void*)&scanhash_x16rt_4way;
gate->hash = (void*)&x16r_4way_hash;
#elif defined (X16RT_2WAY)
gate->scanhash = (void*)&scanhash_x16rt_2x64;
gate->hash = (void*)&x16r_2x64_hash;
#else
gate->scanhash = (void*)&scanhash_x16rt;
gate->hash = (void*)&x16r_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT
| NEON_OPT;
opt_target_factor = 256.0;
return true;
};
bool register_x16rt_veil_algo( algo_gate_t* gate )
{
#if defined (X16R_8WAY)
#if defined (X16RT_8WAY)
gate->scanhash = (void*)&scanhash_x16rt_8way;
gate->hash = (void*)&x16r_8way_hash;
#elif defined (X16R_4WAY)
#elif defined (X16RT_4WAY)
gate->scanhash = (void*)&scanhash_x16rt_4way;
gate->hash = (void*)&x16r_4way_hash;
#elif defined (X16RT_2WAY)
gate->scanhash = (void*)&scanhash_x16rt_2x64;
gate->hash = (void*)&x16r_2x64_hash;
#else
gate->scanhash = (void*)&scanhash_x16rt;
gate->hash = (void*)&x16r_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT
| NEON_OPT;
gate->build_extraheader = (void*)&veil_build_extraheader;
opt_target_factor = 256.0;
return true;
@@ -262,7 +284,7 @@ bool register_hex_algo( algo_gate_t* gate )
{
gate->scanhash = (void*)&scanhash_hex;
gate->hash = (void*)&x16r_hash;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | NEON_OPT;
gate->gen_merkle_root = (void*)&sha256_gen_merkle_root;
opt_target_factor = 128.0;
return true;
@@ -274,20 +296,25 @@ bool register_hex_algo( algo_gate_t* gate )
bool register_x21s_algo( algo_gate_t* gate )
{
#if defined (X16R_8WAY)
#if defined (X21S_8WAY)
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)
#elif defined (X21S_4WAY)
gate->scanhash = (void*)&scanhash_x21s_4way;
gate->hash = (void*)&x21s_4way_hash;
gate->miner_thread_init = (void*)&x21s_4way_thread_init;
#elif defined (X21S_2WAY)
gate->scanhash = (void*)&scanhash_x21s_2x64;
gate->hash = (void*)&x21s_2x64_hash;
gate->miner_thread_init = (void*)&x21s_2x64_thread_init;
#else
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;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT
| NEON_OPT;
x16_r_s_getAlgoString = (void*)&x16s_getAlgoString;
opt_target_factor = 256.0;
return true;

155
algo/x16/x16r-gate.h
View File

@@ -7,13 +7,15 @@
#include <unistd.h>
#include "algo/blake/blake512-hash.h"
#include "algo/bmw/sph_bmw.h"
#include "algo/groestl/sph_groestl.h"
#include "algo/jh/sph_jh.h"
#include "algo/groestl/sph_groestl.h"
#include "algo/keccak/sph_keccak.h"
#include "algo/skein/sph_skein.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/sph_simd.h"
#include "algo/simd/nist.h"
#include "algo/echo/sph_echo.h"
#include "algo/hamsi/sph_hamsi.h"
#include "algo/fugue/sph_fugue.h"
@@ -21,13 +23,13 @@
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/sha/sha512-hash.h"
#if defined(__AES__)
#if defined(__AES__) || defined(__ARM_FEATURE_AES)
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/groestl/aes_ni/hash-groestl.h"
#include "algo/fugue/fugue-aesni.h"
#endif
#if defined (__AVX2__)
//#if defined (__AVX2__)
#include "algo/bmw/bmw-hash-4way.h"
#include "algo/groestl/aes_ni/hash-groestl.h"
#include "algo/skein/skein-hash-4way.h"
@@ -39,7 +41,7 @@
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/hamsi/hamsi-hash-4way.h"
#include "algo/shabal/shabal-hash-4way.h"
#endif
//#endif
#if defined(__VAES__)
#include "algo/groestl/groestl512-hash-4way.h"
@@ -48,28 +50,41 @@
#include "algo/echo/echo-hash-4way.h"
#endif
#if defined(__aarch64__)
#include "algo/simd/sph_simd.h"
#else
#include "algo/simd/nist.h"
// X16R, X16S
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
#define X16R_8WAY 1
#elif defined(__AVX2__) && defined(__AES__)
#define X16R_4WAY 1
#elif defined(__SSE2__) || defined(__ARM_NEON)
#define X16R_2WAY 1
#endif
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
#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
#elif defined(__SSE2__) || defined(__ARM_NEON)
#define X16RV2_2WAY 1
#endif
// X16RT, VEIL
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
#define X16RT_8WAY 1
#elif defined(__AVX2__) && defined(__AES__)
#define X16RT_4WAY 1
#elif defined(__SSE2__) || defined(__ARM_NEON)
#define X16RT_2WAY 1
#endif
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
#define X21S_8WAY 1
#elif defined(__AVX2__) && defined(__AES__)
#define X21S_4WAY 1
#elif defined(__SSE2__) || defined(__ARM_NEON)
#define X21S_2WAY 1
#endif
enum x16r_Algo {
BLAKE = 0,
BMW,
@@ -167,7 +182,6 @@ union _x16r_4way_context_overlay
keccak512_4way_context keccak;
luffa_2way_context luffa;
cube_2way_context cube;
hashState_luffa luffa1;
simd_2way_context simd;
hamsi512_4way_context hamsi;
hashState_fugue fugue;
@@ -187,34 +201,84 @@ int scanhash_x16r_4way( struct work *, uint32_t,
uint64_t *, struct thr_info * );
extern __thread x16r_4way_context_overlay x16r_ctx;
#elif defined(X16R_2WAY)
union _x16r_2x64_context_overlay
{
blake512_2x64_context blake;
bmw512_2x64_context bmw;
#if defined(__AES__) // || defined(__ARM_FEATURE_AES)
hashState_groestl groestl;
#else
sph_groestl512_context groestl;
#endif
skein512_2x64_context skein;
jh512_2x64_context jh;
keccak512_2x64_context keccak;
hashState_luffa luffa;
cubehashParam cube;
shavite512_context shavite;
simd512_context simd;
#if defined(__AES__) || defined(__ARM_FEATURE_AES)
hashState_echo echo;
#else
sph_echo512_context echo;
#endif
#if defined(__SSE4_2__) || defined(__ARM_NEON)
hamsi_2x64_context hamsi;
#else
sph_hamsi512_context hamsi;
#endif
#if defined(__AES__)
hashState_fugue fugue;
#else
sph_fugue512_context fugue;
#endif
sph_shabal512_context shabal;
sph_whirlpool_context whirlpool;
sha512_2x64_context sha512;
} __attribute__ ((aligned (64)));
typedef union _x16r_2x64_context_overlay x16r_2x64_context_overlay;
void x16r_2x64_prehash( void *, void * );
int x16r_2x64_hash_generic( void *, const void *, int );
int x16r_2x64_hash( void *, const void *, int );
int scanhash_x16r_2x64( struct work *, uint32_t,
uint64_t *, struct thr_info * );
extern __thread x16r_2x64_context_overlay x16r_ctx;
#endif
// need a reference, add hooks for SSE2.
// needed for hex
union _x16r_context_overlay
{
#if defined(__AES__)
hashState_echo echo;
hashState_groestl groestl;
hashState_fugue fugue;
#else
sph_groestl512_context groestl;
sph_echo512_context echo;
sph_fugue512_context fugue;
#endif
blake512_context blake;
sph_bmw512_context bmw;
#if defined(__AES__) // || defined(__ARM_FEATURE_AES)
hashState_groestl groestl;
#else
sph_groestl512_context groestl;
#endif
sph_skein512_context skein;
sph_jh512_context jh;
sph_keccak512_context keccak;
hashState_luffa luffa;
cubehashParam cube;
shavite512_context shavite;
#if defined(__aarch64__)
sph_simd512_context simd;
simd512_context simd;
#if defined(__AES__) || defined(__ARM_FEATURE_AES)
hashState_echo echo;
#else
hashState_sd simd;
sph_echo512_context echo;
#endif
sph_hamsi512_context hamsi;
#if defined(__AES__)
hashState_fugue fugue;
#else
sph_fugue512_context fugue;
#endif
sph_shabal512_context shabal;
sph_whirlpool_context whirlpool;
sph_sha512_context sha512;
@@ -222,7 +286,7 @@ union _x16r_context_overlay
typedef union _x16r_context_overlay x16r_context_overlay;
extern __thread x16r_context_overlay x16_ctx;
extern __thread x16r_context_overlay x16r_ref_ctx;
void x16r_prehash( void *, void * );
int x16r_hash_generic( void *, const void *, int );
@@ -242,6 +306,12 @@ 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 );
#elif defined(X16RV2_2WAY)
int x16rv2_2x64_hash( void *state, const void *input, int thrid );
int scanhash_x16rv2_2x64( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
#else
int x16rv2_hash( void *state, const void *input, int thr_id );
@@ -251,18 +321,24 @@ int scanhash_x16rv2( struct work *work, uint32_t max_nonce,
#endif
// x16rt, veil
#if defined(X16R_8WAY)
#if defined(X16RT_8WAY)
//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)
#elif defined(X16RT_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 );
#elif defined(X16RT_2WAY)
//void x16rt_4way_hash( void *state, const void *input );
int scanhash_x16rt_2x64( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
#else
//void x16rt_hash( void *state, const void *input );
@@ -272,20 +348,27 @@ int scanhash_x16rt( struct work *work, uint32_t max_nonce,
#endif
// x21s
#if defined(X16R_8WAY)
#if defined(X21S_8WAY)
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)
#elif defined(X21S_4WAY)
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();
#elif defined(X21S_2WAY)
int x21s_2x64_hash( void *state, const void *input, int thrid );
int scanhash_x21s_2x64( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
bool x21s_2x64_thread_init();
#else
int x21s_hash( void *state, const void *input, int thr_id );

89
algo/x16/x16r.c
View File

@@ -18,32 +18,36 @@ void x16r_prehash( void *edata, void *pdata )
switch ( algo )
{
case JH:
sph_jh512_init( &x16_ctx.jh );
sph_jh512( &x16_ctx.jh, edata, 64 );
sph_jh512_init( &x16r_ref_ctx.jh );
sph_jh512( &x16r_ref_ctx.jh, edata, 64 );
break;
case SKEIN:
sph_skein512_init( &x16_ctx.skein );
sph_skein512( &x16_ctx.skein, edata, 64 );
sph_skein512_init( &x16r_ref_ctx.skein );
sph_skein512( &x16r_ref_ctx.skein, edata, 64 );
break;
case KECCAK:
sph_keccak512_init( &x16r_ref_ctx.keccak );
sph_keccak512( &x16r_ref_ctx.keccak, edata, 72 );
break;
case LUFFA:
init_luffa( &x16_ctx.luffa, 512 );
update_luffa( &x16_ctx.luffa, edata, 64 );
init_luffa( &x16r_ref_ctx.luffa, 512 );
update_luffa( &x16r_ref_ctx.luffa, edata, 64 );
break;
case CUBEHASH:
cubehashInit( &x16_ctx.cube, 512, 16, 32 );
cubehashUpdate( &x16_ctx.cube, edata, 64 );
cubehashInit( &x16r_ref_ctx.cube, 512, 16, 32 );
cubehashUpdate( &x16r_ref_ctx.cube, edata, 64 );
break;
case HAMSI:
sph_hamsi512_init( &x16_ctx.hamsi );
sph_hamsi512( &x16_ctx.hamsi, edata, 64 );
break;
sph_hamsi512_init( &x16r_ref_ctx.hamsi );
sph_hamsi512( &x16r_ref_ctx.hamsi, edata, 72 );
break;
case SHABAL:
sph_shabal512_init( &x16_ctx.shabal );
sph_shabal512( &x16_ctx.shabal, edata, 64 );
sph_shabal512_init( &x16r_ref_ctx.shabal );
sph_shabal512( &x16r_ref_ctx.shabal, edata, 64 );
break;
case WHIRLPOOL:
sph_whirlpool_init( &x16_ctx.whirlpool );
sph_whirlpool( &x16_ctx.whirlpool, edata, 64 );
sph_whirlpool_init( &x16r_ref_ctx.whirlpool );
sph_whirlpool( &x16r_ref_ctx.whirlpool, edata, 64 );
break;
}
}
@@ -52,7 +56,7 @@ int x16r_hash_generic( void* output, const void* input, int thrid )
{
uint32_t _ALIGN(128) hash[16];
x16r_context_overlay ctx;
memcpy( &ctx, &x16_ctx, sizeof(ctx) );
memcpy( &ctx, &x16r_ref_ctx, sizeof(ctx) );
void *in = (void*) input;
int size = 80;
@@ -70,36 +74,41 @@ int x16r_hash_generic( void* output, const void* input, int thrid )
break;
case BMW:
sph_bmw512_init( &ctx.bmw );
sph_bmw512(&ctx.bmw, in, size);
sph_bmw512_close(&ctx.bmw, hash);
sph_bmw512( &ctx.bmw, in, size );
sph_bmw512_close( &ctx.bmw, hash );
break;
case GROESTL:
#if defined(__AES__)
groestl512_full( &ctx.groestl, (char*)hash, (char*)in, size<<3 );
#if defined(__AES__) // || defined(__ARM_FEATURE_AES)
groestl512_full( &ctx.groestl, hash, in, size<<3 );
#else
sph_groestl512_init( &ctx.groestl );
sph_groestl512( &ctx.groestl, in, size );
sph_groestl512_close(&ctx.groestl, hash);
sph_groestl512_close( &ctx.groestl, hash );
#endif
break;
case JH:
if ( i == 0 )
sph_jh512(&ctx.jh, in+64, 16 );
sph_jh512( &ctx.jh, in+64, 16 );
else
{
sph_jh512_init( &ctx.jh );
sph_jh512(&ctx.jh, in, size );
sph_jh512( &ctx.jh, in, size );
}
sph_jh512_close(&ctx.jh, hash );
sph_jh512_close( &ctx.jh, hash );
break;
case KECCAK:
sph_keccak512_init( &ctx.keccak );
sph_keccak512( &ctx.keccak, in, size );
if ( i == 0 )
sph_keccak512( &ctx.keccak, in+72, 8 );
else
{
sph_keccak512_init( &ctx.keccak );
sph_keccak512( &ctx.keccak, in, size );
}
sph_keccak512_close( &ctx.keccak, hash );
break;
case SKEIN:
if ( i == 0 )
sph_skein512(&ctx.skein, in+64, 16 );
sph_skein512( &ctx.skein, in+64, 16 );
else
{
sph_skein512_init( &ctx.skein );
@@ -109,13 +118,13 @@ int x16r_hash_generic( void* output, const void* input, int thrid )
break;
case LUFFA:
if ( i == 0 )
update_and_final_luffa( &ctx.luffa, hash, (const void*)in+64, 16 );
update_and_final_luffa( &ctx.luffa, hash, in+64, 16 );
else
luffa_full( &ctx.luffa, hash, 512, in, size );
break;
case CUBEHASH:
if ( i == 0 )
cubehashUpdateDigest( &ctx.cube, hash, (const void*)in+64, 16 );
cubehashUpdateDigest( &ctx.cube, hash, in+64, 16 );
else
cubehash_full( &ctx.cube, hash, 512, in, size );
break;
@@ -123,19 +132,13 @@ int x16r_hash_generic( void* output, const void* input, int thrid )
shavite512_full( &ctx.shavite, hash, in, size );
break;
case SIMD:
#if defined(__aarch64__)
sph_simd512_init( &ctx.simd );
sph_simd512(&ctx.simd, (const void*) hash, 64);
sph_simd512_close(&ctx.simd, hash);
#else
simd_full( &ctx.simd, (BitSequence *)hash,
(const BitSequence*)in, size<<3 );
#endif
sph_simd512( &ctx.simd, hash, size );
sph_simd512_close( &ctx.simd, hash );
break;
case ECHO:
#if defined(__AES__)
echo_full( &ctx.echo, (BitSequence*)hash, 512,
(const BitSequence*)in, size );
echo_full( &ctx.echo, hash, 512, in, size );
#else
sph_echo512_init( &ctx.echo );
sph_echo512( &ctx.echo, in, size );
@@ -144,7 +147,7 @@ int x16r_hash_generic( void* output, const void* input, int thrid )
break;
case HAMSI:
if ( i == 0 )
sph_hamsi512( &ctx.hamsi, in+64, 16 );
sph_hamsi512( &ctx.hamsi, in+72, 8 );
else
{
sph_hamsi512_init( &ctx.hamsi );
@@ -153,12 +156,8 @@ int x16r_hash_generic( void* output, const void* input, int thrid )
sph_hamsi512_close( &ctx.hamsi, hash );
break;
case FUGUE:
#if defined(__AES__)
fugue512_full( &ctx.fugue, hash, in, size );
#else
sph_fugue512_full( &ctx.fugue, hash, in, size );
#endif
break;
sph_fugue512_full( &ctx.fugue, hash, in, size );
break;
case SHABAL:
if ( i == 0 )
sph_shabal512( &ctx.shabal, in+64, 16 );

55
algo/x16/x16rt-4way.c
View File

@@ -3,7 +3,7 @@
#include <stdlib.h>
#include <string.h>
#if defined (X16R_8WAY)
#if defined (X16RT_8WAY)
int scanhash_x16rt_8way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
@@ -57,7 +57,7 @@ int scanhash_x16rt_8way( struct work *work, uint32_t max_nonce,
return 0;
}
#elif defined (X16R_4WAY)
#elif defined (X16RT_4WAY)
int scanhash_x16rt_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
@@ -110,4 +110,55 @@ int scanhash_x16rt_4way( struct work *work, uint32_t max_nonce,
return 0;
}
#elif defined (X16RT_2WAY)
int scanhash_x16rt_2x64( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t hash[2*16] __attribute__ ((aligned (64)));
uint32_t vdata[24*2] __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 - 2;
uint32_t n = first_nonce;
const int thr_id = mythr->id;
v128_t *noncev = (v128_t*)vdata + 9;
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_2x64_prehash( vdata, pdata );
*noncev = v128_intrlv_blend_32( v128_set32( n+1, 0, n, 0 ), *noncev );
do
{
if ( x16r_2x64_hash( hash, vdata, thr_id ) )
for ( int i = 0; i < 2; i++ )
if ( unlikely( valid_hash( hash + (i<<3), ptarget ) && !bench ) )
{
pdata[19] = bswap_32( n+i );
submit_solution( work, hash+(i<<3), mythr );
}
*noncev = v128_add32( *noncev, v128_64( 0x0000000200000000 ) );
n += 2;
} while ( ( n < last_nonce ) && !(*restart) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}
#endif

2
algo/x16/x16rt.c
View File

@@ -1,6 +1,6 @@
#include "x16r-gate.h"
#if !defined(X16R_8WAY) && !defined(X16R_4WAY)
#if !defined(X16RT_8WAY) && !defined(X16RT_4WAY)
int scanhash_x16rt( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )

548
algo/x16/x16rv2-4way.c
View File

@@ -395,7 +395,7 @@ int x16rv2_8way_hash( void* output, const void* input, int thrid )
break;
case HAMSI:
if ( i == 0 )
hamsi512_8way_update( &ctx.hamsi, input + (64<<3), 16 );
hamsi512_8way_update( &ctx.hamsi, input + (72<<3), 8 );
else
{
intrlv_8x64( vhash, in0, in1, in2, in3, in4, in5, in6, in7,
@@ -409,14 +409,43 @@ int x16rv2_8way_hash( void* output, const void* input, int thrid )
hash7, vhash );
break;
case FUGUE:
fugue512_full( &ctx.fugue, hash0, in0, size );
fugue512_full( &ctx.fugue, hash1, in1, size );
fugue512_full( &ctx.fugue, hash2, in2, size );
fugue512_full( &ctx.fugue, hash3, in3, size );
fugue512_full( &ctx.fugue, hash4, in4, size );
fugue512_full( &ctx.fugue, hash5, in5, size );
fugue512_full( &ctx.fugue, hash6, in6, size );
fugue512_full( &ctx.fugue, hash7, in7, size );
if ( i == 0 )
{
fugue512_update( &ctx.fugue, in0 + 76, 4 );
fugue512_final( &ctx.fugue, hash0 );
memcpy( &ctx, &x16rv2_ctx, sizeof(hashState_fugue) );
fugue512_update( &ctx.fugue, in1 + 76, 4 );
fugue512_final( &ctx.fugue, hash1 );
memcpy( &ctx, &x16rv2_ctx, sizeof(hashState_fugue) );
fugue512_update( &ctx.fugue, in2 + 76, 4 );
fugue512_final( &ctx.fugue, hash2 );
memcpy( &ctx, &x16rv2_ctx, sizeof(hashState_fugue) );
fugue512_update( &ctx.fugue, in3 + 76, 4 );
fugue512_final( &ctx.fugue, hash3 );
memcpy( &ctx, &x16rv2_ctx, sizeof(hashState_fugue) );
fugue512_update( &ctx.fugue, in4 + 76, 4 );
fugue512_final( &ctx.fugue, hash4 );
memcpy( &ctx, &x16rv2_ctx, sizeof(hashState_fugue) );
fugue512_update( &ctx.fugue, in5 + 76, 4 );
fugue512_final( &ctx.fugue, hash5 );
memcpy( &ctx, &x16rv2_ctx, sizeof(hashState_fugue) );
fugue512_update( &ctx.fugue, in6 + 76, 4 );
fugue512_final( &ctx.fugue, hash6 );
memcpy( &ctx, &x16rv2_ctx, sizeof(hashState_fugue) );
fugue512_update( &ctx.fugue, in7 + 76, 4 );
fugue512_final( &ctx.fugue, hash7 );
}
else
{
fugue512_full( &ctx.fugue, hash0, hash0, size );
fugue512_full( &ctx.fugue, hash1, hash1, size );
fugue512_full( &ctx.fugue, hash2, hash2, size );
fugue512_full( &ctx.fugue, hash3, hash3, size );
fugue512_full( &ctx.fugue, hash4, hash4, size );
fugue512_full( &ctx.fugue, hash5, hash5, size );
fugue512_full( &ctx.fugue, hash6, hash6, size );
fugue512_full( &ctx.fugue, hash7, hash7, size );
}
break;
case SHABAL:
intrlv_8x32( vhash, in0, in1, in2, in3, in4, in5, in6, in7,
@@ -588,7 +617,7 @@ int scanhash_x16rv2_8way( struct work *work, uint32_t max_nonce,
{
x16_r_s_getAlgoString( (const uint8_t*)bedata1, x16r_hash_order );
s_ntime = ntime;
if ( opt_debug && !thr_id )
if ( !opt_quiet && !thr_id )
applog( LOG_INFO, "hash order %s (%08x)", x16r_hash_order, ntime );
}
@@ -626,7 +655,14 @@ int scanhash_x16rv2_8way( struct work *work, uint32_t max_nonce,
case HAMSI:
mm512_bswap32_intrlv80_8x64( vdata, pdata );
hamsi512_8way_init( &x16rv2_ctx.hamsi );
hamsi512_8way_update( &x16rv2_ctx.hamsi, vdata, 64 );
hamsi512_8way_update( &x16rv2_ctx.hamsi, vdata, 72 );
break;
case FUGUE:
v128_bswap32_80( edata, pdata );
fugue512_init( &x16rv2_ctx.fugue );
fugue512_update( &x16rv2_ctx.fugue, edata, 76 );
intrlv_8x64( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
break;
case SHABAL:
mm256_bswap32_intrlv80_8x32( vdata2, pdata );
@@ -824,8 +860,8 @@ int x16rv2_4way_hash( void* output, const void* input, int thrid )
intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
skein512_4way_init( &ctx.skein );
skein512_4way_update( &ctx.skein, vhash, size );
skein512_4way_close( &ctx.skein, vhash );
}
skein512_4way_close( &ctx.skein, vhash );
dintrlv_4x64_512( hash0, hash1, hash2, hash3, vhash );
break;
case LUFFA:
@@ -945,7 +981,7 @@ int x16rv2_4way_hash( void* output, const void* input, int thrid )
break;
case HAMSI:
if ( i == 0 )
hamsi512_4way_update( &ctx.hamsi, input + (64<<2), 16 );
hamsi512_4way_update( &ctx.hamsi, input + (72<<2), 8 );
else
{
intrlv_4x64( vhash, in0, in1, in2, in3, size<<3 );
@@ -956,10 +992,27 @@ int x16rv2_4way_hash( void* output, const void* input, int thrid )
dintrlv_4x64_512( hash0, hash1, hash2, hash3, vhash );
break;
case FUGUE:
fugue512_full( &ctx.fugue, hash0, in0, size );
fugue512_full( &ctx.fugue, hash1, in1, size );
fugue512_full( &ctx.fugue, hash2, in2, size );
fugue512_full( &ctx.fugue, hash3, in3, size );
if ( i == 0 )
{
fugue512_update( &ctx.fugue, in0 + 76, 4 );
fugue512_final( &ctx.fugue, hash0 );
memcpy( &ctx, &x16rv2_ctx, sizeof(hashState_fugue) );
fugue512_update( &ctx.fugue, in1 + 76, 4 );
fugue512_final( &ctx.fugue, hash1 );
memcpy( &ctx, &x16rv2_ctx, sizeof(hashState_fugue) );
fugue512_update( &ctx.fugue, in2 + 76, 4 );
fugue512_final( &ctx.fugue, hash2 );
memcpy( &ctx, &x16rv2_ctx, sizeof(hashState_fugue) );
fugue512_update( &ctx.fugue, in3 + 76, 4 );
fugue512_final( &ctx.fugue, hash3 );
}
else
{
fugue512_full( &ctx.fugue, hash0, hash0, size );
fugue512_full( &ctx.fugue, hash1, hash1, size );
fugue512_full( &ctx.fugue, hash2, hash2, size );
fugue512_full( &ctx.fugue, hash3, hash3, size );
}
break;
case SHABAL:
intrlv_4x32( vhash, in0, in1, in2, in3, size<<3 );
@@ -1077,7 +1130,7 @@ int scanhash_x16rv2_4way( struct work *work, uint32_t max_nonce,
{
x16_r_s_getAlgoString( (const uint8_t*)bedata1, x16r_hash_order );
s_ntime = ntime;
if ( opt_debug && !thr_id )
if ( !opt_quiet && !thr_id )
applog( LOG_INFO, "hash order %s (%08x)", x16r_hash_order, ntime );
}
@@ -1101,7 +1154,7 @@ int scanhash_x16rv2_4way( struct work *work, uint32_t max_nonce,
break;
case SKEIN:
mm256_bswap32_intrlv80_4x64( vdata, pdata );
skein512_4way_prehash64( &x16r_ctx.skein, vdata );
skein512_4way_prehash64( &x16rv2_ctx.skein, vdata );
break;
case CUBEHASH:
v128_bswap32_80( edata, pdata );
@@ -1112,7 +1165,13 @@ int scanhash_x16rv2_4way( struct work *work, uint32_t max_nonce,
case HAMSI:
mm256_bswap32_intrlv80_4x64( vdata, pdata );
hamsi512_4way_init( &x16rv2_ctx.hamsi );
hamsi512_4way_update( &x16rv2_ctx.hamsi, vdata, 64 );
hamsi512_4way_update( &x16rv2_ctx.hamsi, vdata, 72 );
break;
case FUGUE:
v128_bswap32_80( edata, pdata );
fugue512_init( &x16rv2_ctx.fugue );
fugue512_update( &x16rv2_ctx.fugue, edata, 76 );
intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
break;
case SHABAL:
v128_bswap32_intrlv80_4x32( vdata32, pdata );
@@ -1151,4 +1210,453 @@ int scanhash_x16rv2_4way( struct work *work, uint32_t max_nonce,
return 0;
}
#elif defined (X16RV2_2WAY)
union _x16rv2_2x64_context_overlay
{
blake512_2x64_context blake;
bmw512_2x64_context bmw;
#if defined(__AES__) // || defined(__ARM_FEATURE_AES)
hashState_groestl groestl;
#else
sph_groestl512_context groestl;
#endif
skein512_2x64_context skein;
jh512_2x64_context jh;
keccak512_2x64_context keccak;
hashState_luffa luffa;
cubehashParam cube;
shavite512_context shavite;
simd512_context simd;
#if defined(__AES__) || defined(__ARM_FEATURE_AES)
hashState_echo echo;
#else
sph_echo512_context echo;
#endif
#if defined(__SSE4_2__) || defined(__ARM_NEON)
hamsi_2x64_context hamsi;
#else
sph_hamsi512_context hamsi;
#endif
#if defined(__AES__)
hashState_fugue fugue;
#else
sph_fugue512_context fugue;
#endif
sph_shabal512_context shabal;
sph_whirlpool_context whirlpool;
sha512_2x64_context sha512;
sph_tiger_context tiger;
} __attribute__ ((aligned (64)));
typedef union _x16rv2_2x64_context_overlay x16rv2_2x64_context_overlay;
static __thread x16rv2_2x64_context_overlay x16rv2_ctx;
// Pad the 24 bytes tiger hash to 64 bytes
static inline void padtiger512( uint32_t* hash )
{
for ( int i = 6; i < 16; i++ ) hash[i] = 0;
}
int x16rv2_2x64_hash( void* output, const void* input, int thrid )
{
uint32_t vhash[20*2] __attribute__ ((aligned (64)));
uint32_t hash0[20] __attribute__ ((aligned (32)));
uint32_t hash1[20] __attribute__ ((aligned (32)));
x16rv2_2x64_context_overlay ctx;
memcpy( &ctx, &x16rv2_ctx, sizeof(ctx) );
void *in0 = (void*) hash0;
void *in1 = (void*) hash1;
int size = 80;
dintrlv_2x64( hash0, hash1, input, 640 );
for ( int i = 0; i < 16; i++ )
{
const char elem = x16r_hash_order[i];
const uint8_t algo = elem >= 'A' ? elem - 'A' + 10 : elem - '0';
switch ( algo )
{
case BLAKE:
if ( i == 0 )
blake512_2x64_full( &ctx.blake, vhash, input, size );
else
{
intrlv_2x64( vhash, in0, in1, size<<3 );
blake512_2x64_full( &ctx.blake, vhash, vhash, size );
}
dintrlv_2x64( hash0, hash1, vhash, 512 );
break;
case BMW:
bmw512_2x64_init( &ctx.bmw );
if ( i == 0 )
bmw512_2x64_update( &ctx.bmw, input, size );
else
{
intrlv_2x64( vhash, in0, in1, size<<3 );
bmw512_2x64_update( &ctx.bmw, vhash, size );
}
bmw512_2x64_close( &ctx.bmw, vhash );
dintrlv_2x64( hash0, hash1, vhash, 512 );
break;
case GROESTL:
#if defined(__AES__)
groestl512_full( &ctx.groestl, hash0, in0, size<<3 );
groestl512_full( &ctx.groestl, hash1, in1, size<<3 );
#else
sph_groestl512_init( &ctx.groestl );
sph_groestl512( &ctx.groestl, in0, size );
sph_groestl512_close( &ctx.groestl, hash0 );
sph_groestl512_init( &ctx.groestl );
sph_groestl512( &ctx.groestl, in1, size );
sph_groestl512_close( &ctx.groestl, hash1 );
#endif
break;
case JH:
if ( i == 0 )
jh512_2x64_update( &ctx.jh, input + (64<<1), 16 );
else
{
intrlv_2x64( vhash, in0, in1, size<<3 );
jh512_2x64_init( &ctx.jh );
jh512_2x64_update( &ctx.jh, vhash, size );
}
jh512_2x64_close( &ctx.jh, vhash );
dintrlv_2x64( hash0, hash1, vhash, 512 );
break;
case KECCAK:
if ( i == 0 )
{
sph_tiger( &ctx.tiger, in0 + 64, 16 );
sph_tiger_close( &ctx.tiger, hash0 );
memcpy( &ctx, &x16rv2_ctx, sizeof(ctx) );
sph_tiger( &ctx.tiger, in1 + 64, 16 );
sph_tiger_close( &ctx.tiger, hash1 );
}
else
{
sph_tiger_init( &ctx.tiger );
sph_tiger( &ctx.tiger, in0, size );
sph_tiger_close( &ctx.tiger, hash0 );
sph_tiger_init( &ctx.tiger );
sph_tiger( &ctx.tiger, in1, size );
sph_tiger_close( &ctx.tiger, hash1 );
}
for ( int i = (24/4); i < (64/4); i++ )
hash0[i] = hash1[i] = 0;
intrlv_2x64( vhash, hash0, hash1, 512 );
keccak512_2x64_init( &ctx.keccak );
keccak512_2x64_update( &ctx.keccak, vhash, 64 );
keccak512_2x64_close( &ctx.keccak, vhash );
dintrlv_2x64( hash0, hash1, vhash, 512 );
break;
case SKEIN:
if ( i == 0 )
skein512_2x64_final16( &ctx.skein, vhash, input + (64*2) );
else
{
intrlv_2x64( vhash, in0, in1, size<<3 );
skein512_2x64_full( &ctx.skein, vhash, vhash, size );
}
dintrlv_2x64( hash0, hash1, vhash, 512 );
break;
case LUFFA:
if ( i == 0 )
{
sph_tiger( &ctx.tiger, in0 + 64, 16 );
sph_tiger_close( &ctx.tiger, hash0 );
memcpy( &ctx, &x16rv2_ctx, sizeof(ctx) );
sph_tiger( &ctx.tiger, in1 + 64, 16 );
sph_tiger_close( &ctx.tiger, hash1 );
}
else
{
sph_tiger_init( &ctx.tiger );
sph_tiger( &ctx.tiger, in0, size );
sph_tiger_close( &ctx.tiger, hash0 );
sph_tiger_init( &ctx.tiger );
sph_tiger( &ctx.tiger, in1, size );
sph_tiger_close( &ctx.tiger, hash1 );
}
for ( int i = (24/4); i < (64/4); i++ )
hash0[i] = hash1[i] = 0;
luffa_full( &ctx.luffa, hash0, 512, hash0, 64 );
luffa_full( &ctx.luffa, hash1, 512, hash1, 64 );
break;
case CUBEHASH:
if ( i == 0 )
{
cubehashUpdateDigest( &ctx.cube, hash0, in0 + 64, 16 );
memcpy( &ctx, &x16rv2_ctx, sizeof(ctx) );
cubehashUpdateDigest( &ctx.cube, hash1, in1 + 64, 16 );
}
else
{
cubehash_full( &ctx.cube, hash0, 512, hash0, size );
cubehash_full( &ctx.cube, hash1, 512, hash1, size );
}
break;
case SHAVITE:
shavite512_full( &ctx.shavite, hash0, in0, size );
shavite512_full( &ctx.shavite, hash1, in1, size );
break;
case SIMD:
simd512_ctx( &ctx.simd, hash0, in0, size );
simd512_ctx( &ctx.simd, hash1, in1, size );
break;
case ECHO:
#if defined(__AES__)
echo_full( &ctx.echo, hash0, 512, in0, size );
echo_full( &ctx.echo, hash1, 512, in1, size );
#else
sph_echo512_init( &ctx.echo );
sph_echo512( &ctx.echo, in0, size );
sph_echo512_close( &ctx.echo, hash0 );
sph_echo512_init( &ctx.echo );
sph_echo512( &ctx.echo, in1, size );
sph_echo512_close( &ctx.echo, hash1 );
#endif
break;
case HAMSI:
#if defined(__SSE4_2__) || defined(__ARM_NEON)
if ( i == 0 )
hamsi512_2x64_update( &ctx.hamsi, input + (72*2), 8 );
else
{
intrlv_2x64( vhash, hash0, hash1, size<<3 );
hamsi512_2x64_init( &ctx.hamsi );
hamsi512_2x64_update( &ctx.hamsi, vhash, size );
}
hamsi512_2x64_close( &ctx.hamsi, vhash );
dintrlv_2x64( hash0, hash1, vhash, 512 );
#else
if ( i == 0 )
{
sph_hamsi512( &ctx.hamsi, in0 + 72, 8 );
sph_hamsi512_close( &ctx.hamsi, hash0 );
memcpy( &ctx, &x16rv2_ctx, sizeof(ctx) );
sph_hamsi512( &ctx.hamsi, in1 + 72, 8 );
sph_hamsi512_close( &ctx.hamsi, hash1 );
}
else
{
sph_hamsi512_init( &ctx.hamsi );
sph_hamsi512( &ctx.hamsi, hash0, size );
sph_hamsi512_close( &ctx.hamsi, hash0 );
sph_hamsi512_init( &ctx.hamsi );
sph_hamsi512( &ctx.hamsi, hash1, size );
sph_hamsi512_close( &ctx.hamsi, hash1 );
}
#endif
break;
case FUGUE:
#if defined(__AES__)
if ( i == 0 )
{
fugue512_update( &ctx.fugue, in0 + 76, 4 );
fugue512_final( &ctx.fugue, hash0 );
memcpy( &ctx, &x16rv2_ctx, sizeof(hashState_fugue) );
fugue512_update( &ctx.fugue, in1 + 76, 4 );
fugue512_final( &ctx.fugue, hash1 );
}
else
{
fugue512_full( &ctx.fugue, hash0, hash0, size );
fugue512_full( &ctx.fugue, hash1, hash1, size );
}
#else
if ( i == 0 )
{
sph_fugue512( &ctx.fugue, in0 + 76, 4 );
sph_fugue512_close( &ctx.fugue, hash0 );
memcpy( &ctx, &x16rv2_ctx, sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, in1 + 76, 4 );
sph_fugue512_close( &ctx.fugue, hash1 );
}
else
{
sph_fugue512_full( &ctx.fugue, hash0, hash0, size );
sph_fugue512_full( &ctx.fugue, hash1, hash1, size );
}
#endif
break;
case SHABAL:
if ( i == 0 )
{
sph_shabal512( &ctx.shabal, in0 + 64, 16 );
sph_shabal512_close( &ctx.shabal, hash0 );
memcpy( &ctx, &x16rv2_ctx, sizeof(ctx) );
sph_shabal512( &ctx.shabal, in1 + 64, 16 );
sph_shabal512_close( &ctx.shabal, hash1 );
}
else
{
sph_shabal512_init( &ctx.shabal );
sph_shabal512( &ctx.shabal, hash0, size );
sph_shabal512_close( &ctx.shabal, hash0 );
sph_shabal512_init( &ctx.shabal );
sph_shabal512( &ctx.shabal, hash1, size );
sph_shabal512_close( &ctx.shabal, hash1 );
}
break;
case WHIRLPOOL:
sph_whirlpool512_full( &ctx.whirlpool, hash0, in0, size );
sph_whirlpool512_full( &ctx.whirlpool, hash1, in1, size );
break;
case SHA_512:
if ( i == 0 )
{
sph_tiger( &ctx.tiger, in0 + 64, 16 );
sph_tiger_close( &ctx.tiger, hash0 );
memcpy( &ctx, &x16rv2_ctx, sizeof(ctx) );
sph_tiger( &ctx.tiger, in1 + 64, 16 );
sph_tiger_close( &ctx.tiger, hash1 );
}
else
{
sph_tiger_init( &ctx.tiger );
sph_tiger( &ctx.tiger, in0, size );
sph_tiger_close( &ctx.tiger, hash0 );
sph_tiger_init( &ctx.tiger );
sph_tiger( &ctx.tiger, in1, size );
sph_tiger_close( &ctx.tiger, hash1 );
}
for ( int i = (24/4); i < (64/4); i++ )
hash0[i] = hash1[i] = 0;
intrlv_2x64( vhash, hash0, hash1, 512 );
sha512_2x64_init( &ctx.sha512 );
sha512_2x64_update( &ctx.sha512, vhash, 64 );
sha512_2x64_close( &ctx.sha512, vhash );
dintrlv_2x64( hash0, hash1, vhash, 512 );
break;
}
if ( work_restart[thrid].restart ) return 0;
size = 64;
}
memcpy( output, hash0, 32 );
memcpy( output+32, hash1, 32 );
return 1;
}
int scanhash_x16rv2_2x64( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t hash[2*16] __attribute__ ((aligned (64)));
uint32_t vdata[24*2] __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 - 2;
uint32_t n = first_nonce;
const int thr_id = mythr->id;
v128_t *noncev = (v128_t*)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_quiet && !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:
v128_bswap32_intrlv80_2x64( vdata, pdata );
jh512_2x64_init( &x16rv2_ctx.jh );
jh512_2x64_update( &x16rv2_ctx.jh, vdata, 64 );
break;
case KECCAK:
case LUFFA:
case SHA_512:
v128_bswap32_80( edata, pdata );
sph_tiger_init( &x16rv2_ctx.tiger );
sph_tiger( &x16rv2_ctx.tiger, edata, 64 );
intrlv_2x64( vdata, edata, edata, 640 );
break;
case SKEIN:
v128_bswap32_intrlv80_2x64( vdata, pdata );
skein512_2x64_prehash64( &x16rv2_ctx.skein, vdata );
break;
case CUBEHASH:
v128_bswap32_80( edata, pdata );
cubehashInit( &x16rv2_ctx.cube, 512, 16, 32 );
cubehashUpdate( &x16rv2_ctx.cube, edata, 64 );
intrlv_2x64( vdata, edata, edata, 640 );
break;
case HAMSI:
#if defined(__SSE4_2__) || defined(__ARM_NEON)
v128_bswap32_intrlv80_2x64( vdata, pdata );
hamsi512_2x64_init( &x16rv2_ctx.hamsi );
hamsi512_2x64_update( &x16rv2_ctx.hamsi, vdata, 72 );
#else
v128_bswap32_80( edata, pdata );
sph_hamsi512_init( &x16rv2_ctx.hamsi );
sph_hamsi512( &x16rv2_ctx.hamsi, edata, 72 );
intrlv_2x64( vdata, edata, edata, 640 );
#endif
break;
case FUGUE:
v128_bswap32_80( edata, pdata );
#if defined(__AES__)
fugue512_init( &x16rv2_ctx.fugue );
fugue512_update( &x16rv2_ctx.fugue, edata, 76 );
#else
sph_fugue512_init( &x16rv2_ctx.fugue );
sph_fugue512( &x16rv2_ctx.fugue, edata, 76 );
#endif
intrlv_2x64( vdata, edata, edata, 640 );
break;
case SHABAL:
v128_bswap32_80( edata, pdata );
sph_shabal512_init( &x16rv2_ctx.shabal );
sph_shabal512( &x16rv2_ctx.shabal, edata, 64);
intrlv_2x64( vdata, edata, edata, 640 );
break;
default:
v128_bswap32_intrlv80_2x64( vdata, pdata );
}
*noncev = v128_intrlv_blend_32( v128_set32( n+1, 0, n, 0 ), *noncev );
do
{
if ( x16rv2_2x64_hash( hash, vdata, thr_id ) )
for ( int i = 0; i < 2; i++ )
if ( unlikely( valid_hash( hash + (i<<3), ptarget ) && !bench ) )
{
pdata[19] = bswap_32( n+i );
submit_solution( work, hash+(i<<3), mythr );
}
*noncev = v128_add32( *noncev, v128_64( 0x0000000200000000 ) );
n += 2;
} while ( likely( ( n < last_nonce ) && !(*restart) ) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}
#endif

35
algo/x16/x16rv2.c
View File

@@ -6,21 +6,15 @@
*/
#include "x16r-gate.h"
#if !defined(X16R_8WAY) && !defined(X16R_4WAY)
#if !defined(X16RV2_8WAY) && !defined(X16RV2_4WAY) && !defined(X16RV2_2WAY)
#include "algo/tiger/sph_tiger.h"
union _x16rv2_context_overlay
{
#if defined(__AES__)
hashState_echo echo;
hashState_groestl groestl;
hashState_fugue fugue;
#else
sph_groestl512_context groestl;
sph_echo512_context echo;
sph_fugue512_context fugue;
#endif
blake512_context blake;
sph_bmw512_context bmw;
sph_skein512_context skein;
@@ -29,11 +23,7 @@ union _x16rv2_context_overlay
hashState_luffa luffa;
cubehashParam cube;
shavite512_context shavite;
#if defined(__aarch64__)
sph_simd512_context simd;
#else
hashState_sd simd;
#endif
sph_hamsi512_context hamsi;
sph_shabal512_context shabal;
sph_whirlpool_context whirlpool;
@@ -72,15 +62,9 @@ int x16rv2_hash( void* output, const void* input, int thrid )
sph_bmw512_close(&ctx.bmw, hash);
break;
case GROESTL:
#if defined(__AES__)
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)in, size<<3 );
#else
sph_groestl512_init( &ctx.groestl );
sph_groestl512( &ctx.groestl, in, size );
sph_groestl512_close(&ctx.groestl, hash);
#endif
break;
case SKEIN:
sph_skein512_init( &ctx.skein );
@@ -117,25 +101,14 @@ int x16rv2_hash( void* output, const void* input, int thrid )
shavite512_full( &ctx.shavite, hash, in, size );
break;
case SIMD:
#if defined(__aarch64__)
sph_simd512_init( &ctx.simd );
sph_simd512(&ctx.simd, (const void*) hash, 64);
sph_simd512(&ctx.simd, hash, 64);
sph_simd512_close(&ctx.simd, hash);
#else
simd_full( &ctx.simd, (BitSequence *)hash,
(const BitSequence*)in, size<<3 );
#endif
break;
case ECHO:
#if defined(__AES__)
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash,
(const BitSequence*)in, size<<3 );
#else
sph_echo512_init( &ctx.echo );
sph_echo512( &ctx.echo, in, size );
sph_echo512_close( &ctx.echo, hash );
#endif
break;
case HAMSI:
sph_hamsi512_init( &ctx.hamsi );
@@ -143,11 +116,7 @@ int x16rv2_hash( void* output, const void* input, int thrid )
sph_hamsi512_close( &ctx.hamsi, hash );
break;
case FUGUE:
#if defined(__AES__)
fugue512_full( &ctx.fugue, hash, in, size );
#else
sph_fugue512_full( &ctx.fugue, hash, in, size );
#endif
break;
case SHABAL:
sph_shabal512_init( &ctx.shabal );

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

@@ -9,6 +9,7 @@
#include <stdlib.h>
#include <string.h>
#include "algo/haval/haval-hash-4way.h"
#include "algo/haval/sph-haval.h"
#include "algo/tiger/sph_tiger.h"
#include "algo/gost/sph_gost.h"
#include "algo/lyra2/lyra2.h"
@@ -351,4 +352,119 @@ bool x21s_4way_thread_init()
return x21s_4way_matrix;
}
#elif defined (X21S_2WAY)
static __thread uint64_t* x21s_2x64_matrix;
union _x21s_2x64_context_overlay
{
sph_haval256_5_context haval;
sph_tiger_context tiger;
sph_gost512_context gost;
} __attribute__ ((aligned (64)));
typedef union _x21s_2x64_context_overlay x21s_2x64_context_overlay;
int x21s_2x64_hash( void* output, const void* input, int thrid )
{
uint8_t shash[64*2] __attribute__ ((aligned (64)));
x21s_2x64_context_overlay ctx;
uint32_t *hash0 = (uint32_t*) shash;
uint32_t *hash1 = (uint32_t*)( shash+64 );
if ( !x16r_2x64_hash_generic( shash, input, thrid ) )
return 0;
sph_haval256_5_init( &ctx.haval );
sph_haval256_5( &ctx.haval, hash0, 64 );
sph_haval256_5_close( &ctx.haval, hash0 );
sph_haval256_5_init( &ctx.haval );
sph_haval256_5( &ctx.haval, hash1, 64 );
sph_haval256_5_close( &ctx.haval, hash1 );
sph_tiger_init( &ctx.tiger );
sph_tiger ( &ctx.tiger, (const void*) hash0, 64 );
sph_tiger_close( &ctx.tiger, (void*) hash0 );
sph_tiger_init( &ctx.tiger );
sph_tiger ( &ctx.tiger, (const void*) hash1, 64 );
sph_tiger_close( &ctx.tiger, (void*) hash1 );
LYRA2REV2( x21s_2x64_matrix, (void*) hash0, 32, (const void*) hash0, 32,
(const void*) hash0, 32, 1, 4, 4 );
LYRA2REV2( x21s_2x64_matrix, (void*) hash1, 32, (const void*) hash1, 32,
(const void*) hash1, 32, 1, 4, 4 );
sph_gost512_init( &ctx.gost );
sph_gost512 ( &ctx.gost, (const void*) hash0, 64 );
sph_gost512_close( &ctx.gost, (void*) hash0 );
sph_gost512_init( &ctx.gost );
sph_gost512 ( &ctx.gost, (const void*) hash1, 64 );
sph_gost512_close( &ctx.gost, (void*) hash1 );
sha256_full( output, hash0, 64 );
sha256_full( output+32, hash1, 64 );
return 1;
}
int scanhash_x21s_2x64( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t hash[16*2] __attribute__ ((aligned (64)));
uint32_t vdata[20*2] __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 - 2;
uint32_t n = first_nonce;
const int thr_id = mythr->id;
const bool bench = opt_benchmark;
v128_t *noncev = (v128_t*)vdata + 9;
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_2x64_prehash( vdata, pdata );
*noncev = v128_intrlv_blend_32( v128_set32( n+1, 0, n, 0 ), *noncev );
do
{
if ( x21s_2x64_hash( hash, vdata, thr_id ) )
for ( int i = 0; i < 2; i++ )
if ( unlikely( valid_hash( hash + (i<<3), ptarget ) && !bench ) )
{
pdata[19] = bswap_32( n+i );
submit_solution( work, hash+(i<<3), mythr );
}
*noncev = v128_add32( *noncev, v128_64( 0x0000000200000000 ) );
n += 2;
} while ( likely( ( n < last_nonce ) && !(*restart) ) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}
bool x21s_2x64_thread_init()
{
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * 4; // nCols
const int64_t ROW_LEN_BYTES = ROW_LEN_INT64 * 8;
const int size = (int64_t)ROW_LEN_BYTES * 4; // nRows;
x21s_2x64_matrix = mm_malloc( size, 64 );
return x21s_2x64_matrix;
}
#endif

2
algo/x16/x21s.c
View File

@@ -15,7 +15,7 @@
#include "algo/gost/sph_gost.h"
#include "algo/lyra2/lyra2.h"
#if !defined(X16R_8WAY) && !defined(X16R_4WAY)
#if !defined(X21S_8WAY) && !defined(X21S_4WAY)
static __thread uint64_t* x21s_matrix;