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

v3.9.5

Browse Source
This commit is contained in:
Jay D Dee
2019-06-26 14:16:01 -04:00
parent d6e8d7a46e
commit 0d48d573ce
174 changed files with 1352 additions and 1556 deletions
Download Patch File Download Diff File Expand all files Collapse all files

62
algo/jh/jha-4way.c
View File

@@ -46,7 +46,7 @@ void jha_hash_4way( void *out, const void *input )
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256(
vh[0], _mm256_set1_epi64x( 1 ) ), m256_zero );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
init_groestl( &ctx_groestl, 64 );
update_and_final_groestl( &ctx_groestl, (char*)hash0,
(char*)hash0, 512 );
@@ -59,7 +59,7 @@ void jha_hash_4way( void *out, const void *input )
init_groestl( &ctx_groestl, 64 );
update_and_final_groestl( &ctx_groestl, (char*)hash3,
(char*)hash3, 512 );
mm256_interleave_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
skein512_4way_init( &ctx_skein );
skein512_4way( &ctx_skein, vhash, 64 );
@@ -77,26 +77,26 @@ void jha_hash_4way( void *out, const void *input )
jh512_4way_close( &ctx_jh, vhashB );
for ( int i = 0; i < 8; i++ )
vh[i] = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );
casti_m256i( out, i ) = _mm256_blendv_epi8( vhA[i], vhB[i], vh_mask );
}
mm256_deinterleave_4x64( out, out+32, out+64, out+96, vhash, 256 );
// mm256_dintrlv_4x64( out, out+32, out+64, out+96, vhash, 256 );
}
int scanhash_jha_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
int scanhash_jha_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[8*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t endiandata[20] __attribute__((aligned(64)));
uint32_t *hash7 = &(hash[25]);
uint32_t lane_hash[8];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
uint32_t n = pdata[19];
uint32_t *nonces = work->nonces;
int num_found = 0;
uint32_t *noncep = vdata + 73; // 9*8 + 1
__m256i *noncev = (__m256i*)vdata + 9; // aligned
int thr_id = mythr->id; // thr_id arg is deprecated
uint64_t htmax[] = {
0,
@@ -115,11 +115,12 @@ int scanhash_jha_4way( int thr_id, struct work *work, uint32_t max_nonce,
0
};
for ( int i=0; i < 19; i++ )
be32enc( &endiandata[i], pdata[i] );
// for ( int i=0; i < 19; i++ )
// be32enc( &endiandata[i], pdata[i] );
uint64_t *edata = (uint64_t*)endiandata;
mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
// uint64_t *edata = (uint64_t*)endiandata;
// mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
mm256_bswap_intrlv80_4x64( vdata, pdata );
for ( int m = 0; m < 6; m++ )
{
@@ -127,29 +128,36 @@ int scanhash_jha_4way( int thr_id, struct work *work, uint32_t max_nonce,
{
uint32_t mask = masks[m];
do {
be32enc( noncep, n );
be32enc( noncep+2, n+1 );
be32enc( noncep+4, n+2 );
be32enc( noncep+6, n+3 );
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
// be32enc( noncep, n );
// be32enc( noncep+2, n+1 );
// be32enc( noncep+4, n+2 );
// be32enc( noncep+6, n+3 );
jha_hash_4way( hash, vdata );
pdata[19] = n;
for ( int i = 0; i < 4; i++ )
if ( ( !( (hash+(i<<3))[7] & mask ) == 0 )
&& fulltest( hash+(i<<3), ptarget ) )
// for ( int i = 0; i < 4; i++ )
// if ( ( !( (hash+(i<<3))[7] & mask ) == 0 )
// && fulltest( hash+(i<<3), ptarget ) )
for ( int i = 0; i < 4; i++ ) if ( !( (hash7[i] & mask ) == 0 ) )
{
pdata[19] = n;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
mm256_extract_lane_4x64( lane_hash, hash, i, 256 );
if ( fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
{
pdata[19] = n+i;
submit_lane_solution( work, lane_hash, mythr, i );
// nonces[ num_found++ ] = n+i;
// work_set_target_ratio( work, hash+(i<<3) );
}
}
n += 4;
} while ( ( num_found == 0 ) && ( n < max_nonce )
&& !work_restart[thr_id].restart );
} while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
break;
}
}
*hashes_done = n - first_nonce + 1;
return num_found;
return 0;
}
#endif