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

v3.12.1

Browse Source
This commit is contained in:
Jay D Dee
2020-02-07 20:18:20 -05:00
parent fc97ef174a
commit dc2f8d81d3
10 changed files with 110 additions and 109 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
algo/lyra2/allium.c
View File

@@ -76,37 +76,34 @@ int scanhash_allium( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t _ALIGN(128) hash[8];
uint32_t _ALIGN(128) endiandata[20];
uint32_t _ALIGN(128) edata[20];
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 nonce = first_nonce;
int thr_id = mythr->id; // thr_id arg is deprecated
const int thr_id = mythr->id;
if ( opt_benchmark )
ptarget[7] = 0x3ffff;
for ( int i = 0; i < 19; i++ )
be32enc( &endiandata[i], pdata[i] );
edata[i] = bswap_32( pdata[i] );
sph_blake256_init( &allium_ctx.blake );
sph_blake256( &allium_ctx.blake, endiandata, 64 );
sph_blake256( &allium_ctx.blake, edata, 64 );
do {
be32enc( &endiandata[19], nonce );
allium_hash( hash, endiandata );
if ( hash[7] <= Htarg )
if ( fulltest( hash, ptarget ) && !opt_benchmark )
edata[19] = nonce;
allium_hash( hash, edata );
if ( valid_hash( hash, ptarget ) && !opt_benchmark )
{
pdata[19] = nonce;
pdata[19] = bswap_32( nonce );
submit_solution( work, hash, mythr );
}
nonce++;
} while ( nonce < max_nonce && !work_restart[thr_id].restart );
pdata[19] = nonce;
*hashes_done = pdata[19] - first_nonce + 1;
*hashes_done = pdata[19] - first_nonce;
return 0;
}

36
algo/lyra2/lyra2rev3-4way.c
View File

@@ -130,15 +130,15 @@ int scanhash_lyra2rev3_16way( struct work *work, const uint32_t max_nonce,
{
uint32_t hash[8*16] __attribute__ ((aligned (128)));
uint32_t vdata[20*16] __attribute__ ((aligned (64)));
uint32_t *hash7 = &hash[7<<4];
uint32_t *hash32 = &hash[7*16];
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
const uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
const uint32_t last_nonce = max_nonce - 16;
const uint32_t Htarg = ptarget[7];
__m512i *noncev = (__m512i*)vdata + 19; // aligned
const uint32_t targ32 = ptarget[7];
__m512i *noncev = (__m512i*)vdata + 19;
const int thr_id = mythr->id;
if ( opt_benchmark ) ( (uint32_t*)ptarget )[7] = 0x0000ff;
@@ -159,7 +159,7 @@ int scanhash_lyra2rev3_16way( struct work *work, const uint32_t max_nonce,
pdata[19] = n;
for ( int lane = 0; lane < 16; lane++ )
if ( unlikely( hash7[lane] <= Htarg ) )
if ( unlikely( hash32[lane] <= targ32 ) )
{
extr_lane_16x32( lane_hash, hash, lane, 256 );
if ( likely( fulltest( lane_hash, ptarget ) && !opt_benchmark ) )
@@ -252,15 +252,15 @@ int scanhash_lyra2rev3_8way( struct work *work, const uint32_t max_nonce,
{
uint32_t hash[8*8] __attribute__ ((aligned (64)));
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &hash[7<<3];
uint32_t *hash32 = &hash[7*8];
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
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;
const uint32_t Htarg = ptarget[7];
__m256i *noncev = (__m256i*)vdata + 19; // aligned
const uint32_t targ32 = ptarget[7];
__m256i *noncev = (__m256i*)vdata + 19;
const int thr_id = mythr->id;
const bool bench = opt_benchmark;
@@ -277,7 +277,7 @@ int scanhash_lyra2rev3_8way( struct work *work, const uint32_t max_nonce,
pdata[19] = n;
for ( int lane = 0; lane < 8; lane++ )
if ( unlikely( hash7[lane] <= Htarg ) )
if ( unlikely( hash32[lane] <= targ32 ) )
{
extr_lane_8x32( lane_hash, hash, lane, 256 );
if ( likely( valid_hash( lane_hash, ptarget ) && !bench ) )
@@ -357,40 +357,38 @@ int scanhash_lyra2rev3_4way( struct work *work, const uint32_t max_nonce,
{
uint32_t hash[8*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t *hash32 = &(hash[7*4]);
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *pdata = work->data;
const uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
const uint32_t Htarg = ptarget[7];
__m128i *noncev = (__m128i*)vdata + 19; // aligned
const int thr_id = mythr->id; // thr_id arg is deprecated
const uint32_t targ32 = ptarget[7];
__m128i *noncev = (__m128i*)vdata + 19;
const int thr_id = mythr->id;
if ( opt_benchmark )
( (uint32_t*)ptarget )[7] = 0x0000ff;
mm128_bswap32_intrlv80_4x32( vdata, pdata );
*noncev = _mm_set_epi32( n+3, n+2, n+1, n );
blake256_4way_init( &l2v3_4way_ctx.blake );
blake256_4way_update( &l2v3_4way_ctx.blake, vdata, 64 );
do
{
*noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
lyra2rev3_4way_hash( hash, vdata );
pdata[19] = n;
for ( int lane = 0; lane < 4; lane++ ) if ( hash7[lane] <= Htarg )
for ( int lane = 0; lane < 4; lane++ ) if ( hash32[lane] <= targ32 )
{
extr_lane_4x32( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
if ( valid_hash( lane_hash, ptarget ) && !opt_benchmark )
{
pdata[19] = n + lane;
pdata[19] = bswap_32( n + lane );
submit_lane_solution( work, lane_hash, mythr, lane );
}
}
*noncev = _mm_add_epi32( *noncev, m128_const1_32( 4 ) );
n += 4;
} while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
*hashes_done = n - first_nonce + 1;

10
algo/quark/anime-4way.c
View File

@@ -90,7 +90,7 @@ void anime_4way_hash( void *state, const void *input )
intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
if ( mm256_anybits0( vh_mask ) )
if ( mm256_anybits1( vh_mask ) )
{
skein512_4way_update( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhashB );
@@ -116,13 +116,13 @@ void anime_4way_hash( void *state, const void *input )
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ), zero );
if ( mm256_anybits1( vh_mask ) )
if ( mm256_anybits0( vh_mask ) )
{
blake512_4way_init( &ctx.blake );
blake512_4way_update( &ctx.blake, vhash, 64 );
blake512_4way_close( &ctx.blake, vhashA );
}
if ( mm256_anybits0( vh_mask ) )
if ( mm256_anybits1( vh_mask ) )
{
bmw512_4way_init( &ctx.bmw );
bmw512_4way_update( &ctx.bmw, vhash, 64 );
@@ -140,13 +140,13 @@ void anime_4way_hash( void *state, const void *input )
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ), zero );
if ( mm256_anybits1( vh_mask ) )
if ( mm256_anybits0( vh_mask ) )
{
keccak512_4way_init( &ctx.keccak );
keccak512_4way_update( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhashA );
}
if ( mm256_anybits0( vh_mask ) )
if ( mm256_anybits1( vh_mask ) )
{
jh512_4way_init( &ctx.jh );
jh512_4way_update( &ctx.jh, vhash, 64 );

45
algo/x16/x16rv2.c
View File

@@ -34,7 +34,6 @@
#endif
static __thread uint32_t s_ntime = UINT32_MAX;
static __thread char hashOrder[X16R_HASH_FUNC_COUNT + 1] = { 0 };
union _x16rv2_context_overlay
{
@@ -74,16 +73,10 @@ void x16rv2_hash( void* output, const void* input )
x16rv2_context_overlay ctx;
void *in = (void*) input;
int size = 80;
/*
if ( s_ntime == UINT32_MAX )
{
const uint8_t* in8 = (uint8_t*) input;
x16_r_s_getAlgoString( &in8[4], hashOrder );
}
*/
for ( int i = 0; i < 16; i++ )
{
const char elem = hashOrder[i];
const char elem = x16r_hash_order[i];
const uint8_t algo = elem >= 'A' ? elem - 'A' + 10 : elem - '0';
switch ( algo )
@@ -203,42 +196,42 @@ 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) endiandata[20];
uint32_t _ALIGN(128) edata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
int thr_id = mythr->id; // thr_id arg is deprecated
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( endiandata, 0 ) = mm128_bswap_32( casti_m128i( pdata, 0 ) );
casti_m128i( endiandata, 1 ) = mm128_bswap_32( casti_m128i( pdata, 1 ) );
casti_m128i( endiandata, 2 ) = mm128_bswap_32( casti_m128i( pdata, 2 ) );
casti_m128i( endiandata, 3 ) = mm128_bswap_32( casti_m128i( pdata, 3 ) );
casti_m128i( endiandata, 4 ) = mm128_bswap_32( casti_m128i( pdata, 4 ) );
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*) (&endiandata[1]), hashOrder );
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)", hashOrder, ntime );
applog( LOG_DEBUG, "hash order %s (%08x)",
x16r_hash_order, ntime );
}
if ( opt_benchmark )
ptarget[7] = 0x0cff;
if ( bench ) ptarget[7] = 0x0cff;
do
{
be32enc( &endiandata[19], nonce );
x16rv2_hash( hash32, endiandata );
edata[19] = nonce;
x16rv2_hash( hash32, edata );
if ( hash32[7] <= Htarg )
if (fulltest( hash32, ptarget ) && !opt_benchmark )
if ( unlikely( valid_hash( hash32, ptarget ) && !bench ) )
{
pdata[19] = nonce;
pdata[19] = bswap_32( nonce );
submit_solution( work, hash32, mythr );
}
nonce++;

22
algo/x17/x17-4way.c
View File

@@ -287,10 +287,10 @@ void x17_8way_hash( void *state, const void *input )
int scanhash_x17_8way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[8*16] __attribute__ ((aligned (128)));
uint32_t vdata[24*8] __attribute__ ((aligned (64)));
uint32_t hash[8*8] __attribute__ ((aligned (128)));
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[7<<3]);
uint32_t *hash32 = &(hash[7*8]);
uint32_t *pdata = work->data;
const uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
@@ -298,7 +298,7 @@ int scanhash_x17_8way( struct work *work, uint32_t max_nonce,
__m512i *noncev = (__m512i*)vdata + 9; // aligned
uint32_t n = first_nonce;
const int thr_id = mythr->id;
const uint32_t Htarg = ptarget[7];
const uint32_t targ32 = ptarget[7];
const bool bench = opt_benchmark;
mm512_bswap32_intrlv80_8x64( vdata, pdata );
@@ -310,7 +310,7 @@ int scanhash_x17_8way( struct work *work, uint32_t max_nonce,
x17_8way_hash( hash, vdata );
for ( int lane = 0; lane < 8; lane++ )
if ( unlikely( ( hash7[ lane ] <= Htarg ) && !bench ) )
if ( unlikely( ( hash32[ lane ] <= targ32 ) && !bench ) )
{
extr_lane_8x32( lane_hash, hash, lane, 256 );
if ( likely( valid_hash( lane_hash, ptarget ) ) )
@@ -474,18 +474,18 @@ void x17_4way_hash( void *state, const void *input )
int scanhash_x17_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 hash[8*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t *hash32 = &(hash[ 7*4 ]);
uint32_t *pdata = work->data;
const uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const uint32_t last_nonce = max_nonce -4;
__m256i *noncev = (__m256i*)vdata + 9; // aligned
const uint32_t last_nonce = max_nonce - 4;
__m256i *noncev = (__m256i*)vdata + 9;
uint32_t n = first_nonce;
const int thr_id = mythr->id;
const uint32_t Htarg = ptarget[7];
const uint32_t targ32 = ptarget[7];
const bool bench = opt_benchmark;
mm256_bswap32_intrlv80_4x64( vdata, pdata );
@@ -496,7 +496,7 @@ int scanhash_x17_4way( struct work *work, uint32_t max_nonce,
x17_4way_hash( hash, vdata );
for ( int lane = 0; lane < 4; lane++ )
if ( unlikely( hash7[ lane ] <= Htarg && !bench ) )
if ( unlikely( hash32[ lane ] <= targ32 && !bench ) )
{
extr_lane_4x32( lane_hash, hash, lane, 256 );
if ( valid_hash( lane_hash, ptarget ) )