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

v3.9.0

Browse Source
This commit is contained in:
Jay D Dee
2019-05-19 13:39:45 -04:00
parent bfd1c002f9
commit e1aead3c76
139 changed files with 10907 additions and 4218 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
algo/blake/blake-4way.c
View File

@@ -15,7 +15,7 @@ void blakehash_4way(void *state, const void *input)
memcpy( &ctx, &blake_4w_ctx, sizeof ctx );
blake256r14_4way( &ctx, input + (64<<2), 16 );
blake256r14_4way_close( &ctx, vhash );
mm_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
mm128_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
}
int scanhash_blake_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -37,7 +37,7 @@ int scanhash_blake_4way( int thr_id, struct work *work, uint32_t max_nonce,
// we need big endian data...
swab32_array( edata, pdata, 20 );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
blake256r14_4way_init( &blake_4w_ctx );
blake256r14_4way( &blake_4w_ctx, vdata, 64 );

82
algo/blake/blake-hash-4way.c
View File

@@ -363,14 +363,14 @@ static const sph_u64 CB[16] = {
do { \
a = _mm_add_epi32( _mm_add_epi32( _mm_xor_si128( \
_mm_set_epi32( c1, c1, c1, c1 ), m0 ), b ), a ); \
d = mm_ror_32( _mm_xor_si128( d, a ), 16 ); \
d = mm128_ror_32( _mm_xor_si128( d, a ), 16 ); \
c = _mm_add_epi32( c, d ); \
b = mm_ror_32( _mm_xor_si128( b, c ), 12 ); \
b = mm128_ror_32( _mm_xor_si128( b, c ), 12 ); \
a = _mm_add_epi32( _mm_add_epi32( _mm_xor_si128( \
_mm_set_epi32( c0, c0, c0, c0 ), m1 ), b ), a ); \
d = mm_ror_32( _mm_xor_si128( d, a ), 8 ); \
d = mm128_ror_32( _mm_xor_si128( d, a ), 8 ); \
c = _mm_add_epi32( c, d ); \
b = mm_ror_32( _mm_xor_si128( b, c ), 7 ); \
b = mm128_ror_32( _mm_xor_si128( b, c ), 7 ); \
} while (0)
#if SPH_COMPACT_BLAKE_32
@@ -562,22 +562,22 @@ do { \
, _mm_set_epi32( CS6, CS6, CS6, CS6 ) ); \
VF = _mm_xor_si128( _mm_set_epi32( T1, T1, T1, T1 ), \
_mm_set_epi32( CS7, CS7, CS7, CS7 ) ); \
M[0x0] = mm_bswap_32( *(buf + 0) ); \
M[0x1] = mm_bswap_32( *(buf + 1) ); \
M[0x2] = mm_bswap_32( *(buf + 2) ); \
M[0x3] = mm_bswap_32( *(buf + 3) ); \
M[0x4] = mm_bswap_32( *(buf + 4) ); \
M[0x5] = mm_bswap_32( *(buf + 5) ); \
M[0x6] = mm_bswap_32( *(buf + 6) ); \
M[0x7] = mm_bswap_32( *(buf + 7) ); \
M[0x8] = mm_bswap_32( *(buf + 8) ); \
M[0x9] = mm_bswap_32( *(buf + 9) ); \
M[0xA] = mm_bswap_32( *(buf + 10) ); \
M[0xB] = mm_bswap_32( *(buf + 11) ); \
M[0xC] = mm_bswap_32( *(buf + 12) ); \
M[0xD] = mm_bswap_32( *(buf + 13) ); \
M[0xE] = mm_bswap_32( *(buf + 14) ); \
M[0xF] = mm_bswap_32( *(buf + 15) ); \
M[0x0] = mm128_bswap_32( *(buf + 0) ); \
M[0x1] = mm128_bswap_32( *(buf + 1) ); \
M[0x2] = mm128_bswap_32( *(buf + 2) ); \
M[0x3] = mm128_bswap_32( *(buf + 3) ); \
M[0x4] = mm128_bswap_32( *(buf + 4) ); \
M[0x5] = mm128_bswap_32( *(buf + 5) ); \
M[0x6] = mm128_bswap_32( *(buf + 6) ); \
M[0x7] = mm128_bswap_32( *(buf + 7) ); \
M[0x8] = mm128_bswap_32( *(buf + 8) ); \
M[0x9] = mm128_bswap_32( *(buf + 9) ); \
M[0xA] = mm128_bswap_32( *(buf + 10) ); \
M[0xB] = mm128_bswap_32( *(buf + 11) ); \
M[0xC] = mm128_bswap_32( *(buf + 12) ); \
M[0xD] = mm128_bswap_32( *(buf + 13) ); \
M[0xE] = mm128_bswap_32( *(buf + 14) ); \
M[0xF] = mm128_bswap_32( *(buf + 15) ); \
for (r = 0; r < rounds; r ++) \
ROUND_S_4WAY(r); \
H0 = _mm_xor_si128( _mm_xor_si128( \
@@ -624,22 +624,22 @@ do { \
VD = _mm_xor_si128( _mm_set1_epi32( T0 ), _mm_set1_epi32( CS5 ) ); \
VE = _mm_xor_si128( _mm_set1_epi32( T1 ), _mm_set1_epi32( CS6 ) ); \
VF = _mm_xor_si128( _mm_set1_epi32( T1 ), _mm_set1_epi32( CS7 ) ); \
M0 = mm_bswap_32( * buf ); \
M1 = mm_bswap_32( *(buf+1) ); \
M2 = mm_bswap_32( *(buf+2) ); \
M3 = mm_bswap_32( *(buf+3) ); \
M4 = mm_bswap_32( *(buf+4) ); \
M5 = mm_bswap_32( *(buf+5) ); \
M6 = mm_bswap_32( *(buf+6) ); \
M7 = mm_bswap_32( *(buf+7) ); \
M8 = mm_bswap_32( *(buf+8) ); \
M9 = mm_bswap_32( *(buf+9) ); \
MA = mm_bswap_32( *(buf+10) ); \
MB = mm_bswap_32( *(buf+11) ); \
MC = mm_bswap_32( *(buf+12) ); \
MD = mm_bswap_32( *(buf+13) ); \
ME = mm_bswap_32( *(buf+14) ); \
MF = mm_bswap_32( *(buf+15) ); \
M0 = mm128_bswap_32( * buf ); \
M1 = mm128_bswap_32( *(buf+1) ); \
M2 = mm128_bswap_32( *(buf+2) ); \
M3 = mm128_bswap_32( *(buf+3) ); \
M4 = mm128_bswap_32( *(buf+4) ); \
M5 = mm128_bswap_32( *(buf+5) ); \
M6 = mm128_bswap_32( *(buf+6) ); \
M7 = mm128_bswap_32( *(buf+7) ); \
M8 = mm128_bswap_32( *(buf+8) ); \
M9 = mm128_bswap_32( *(buf+9) ); \
MA = mm128_bswap_32( *(buf+10) ); \
MB = mm128_bswap_32( *(buf+11) ); \
MC = mm128_bswap_32( *(buf+12) ); \
MD = mm128_bswap_32( *(buf+13) ); \
ME = mm128_bswap_32( *(buf+14) ); \
MF = mm128_bswap_32( *(buf+15) ); \
ROUND_S_4WAY(0); \
ROUND_S_4WAY(1); \
ROUND_S_4WAY(2); \
@@ -1073,8 +1073,8 @@ blake32_4way_close( blake_4way_small_context *sc, unsigned ub, unsigned n,
if (out_size_w32 == 8)
buf[52>>2] = _mm_or_si128( buf[52>>2],
_mm_set1_epi32( 0x01000000UL ) );
*(buf+(56>>2)) = mm_bswap_32( _mm_set1_epi32( th ) );
*(buf+(60>>2)) = mm_bswap_32( _mm_set1_epi32( tl ) );
*(buf+(56>>2)) = mm128_bswap_32( _mm_set1_epi32( th ) );
*(buf+(60>>2)) = mm128_bswap_32( _mm_set1_epi32( tl ) );
blake32_4way( sc, buf + (ptr>>2), 64 - ptr );
}
else
@@ -1086,13 +1086,13 @@ blake32_4way_close( blake_4way_small_context *sc, unsigned ub, unsigned n,
memset_zero_128( buf, 56>>2 );
if (out_size_w32 == 8)
buf[52>>2] = _mm_set1_epi32( 0x01000000UL );
*(buf+(56>>2)) = mm_bswap_32( _mm_set1_epi32( th ) );
*(buf+(60>>2)) = mm_bswap_32( _mm_set1_epi32( tl ) );
*(buf+(56>>2)) = mm128_bswap_32( _mm_set1_epi32( th ) );
*(buf+(60>>2)) = mm128_bswap_32( _mm_set1_epi32( tl ) );
blake32_4way( sc, buf, 64 );
}
out = (__m128i*)dst;
for ( k = 0; k < out_size_w32; k++ )
out[k] = mm_bswap_32( sc->H[k] );
out[k] = mm128_bswap_32( sc->H[k] );
}
#if defined (__AVX2__)

5
algo/blake/blake2s-4way.c
View File

@@ -85,7 +85,8 @@ void blake2s_4way_hash( void *output, const void *input )
blake2s_4way_update( &ctx, input + (64<<2), 16 );
blake2s_4way_final( &ctx, vhash, BLAKE2S_OUTBYTES );
mm_deinterleave_4x32( output, output+32, output+64, output+96, vhash, 256 );
mm128_deinterleave_4x32( output, output+32, output+64, output+96,
vhash, 256 );
}
int scanhash_blake2s_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -104,7 +105,7 @@ int scanhash_blake2s_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint32_t *noncep = vdata + 76; // 19*4
swab32_array( edata, pdata, 20 );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
blake2s_4way_init( &blake2s_4w_ctx, BLAKE2S_OUTBYTES );
blake2s_4way_update( &blake2s_4w_ctx, vdata, 64 );

8
algo/blake/blake2s-hash-4way.c
View File

@@ -92,13 +92,13 @@ int blake2s_4way_compress( blake2s_4way_state *S, const __m128i* block )
#define G4W(r,i,a,b,c,d) \
do { \
a = _mm_add_epi32( _mm_add_epi32( a, b ), m[ blake2s_sigma[r][2*i+0] ] ); \
d = mm_ror_32( _mm_xor_si128( d, a ), 16 ); \
d = mm128_ror_32( _mm_xor_si128( d, a ), 16 ); \
c = _mm_add_epi32( c, d ); \
b = mm_ror_32( _mm_xor_si128( b, c ), 12 ); \
b = mm128_ror_32( _mm_xor_si128( b, c ), 12 ); \
a = _mm_add_epi32( _mm_add_epi32( a, b ), m[ blake2s_sigma[r][2*i+1] ] ); \
d = mm_ror_32( _mm_xor_si128( d, a ), 8 ); \
d = mm128_ror_32( _mm_xor_si128( d, a ), 8 ); \
c = _mm_add_epi32( c, d ); \
b = mm_ror_32( _mm_xor_si128( b, c ), 7 ); \
b = mm128_ror_32( _mm_xor_si128( b, c ), 7 ); \
} while(0)
#define ROUND4W(r) \

4
algo/blake/blakecoin-4way.c
View File

@@ -17,7 +17,7 @@ void blakecoin_4way_hash(void *state, const void *input)
blake256r8_4way( &ctx, input + (64<<2), 16 );
blake256r8_4way_close( &ctx, vhash );
mm_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
mm128_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
}
int scanhash_blakecoin_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -37,7 +37,7 @@ int scanhash_blakecoin_4way( int thr_id, struct work *work, uint32_t max_nonce,
HTarget = 0x7f;
swab32_array( edata, pdata, 20 );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
blake256r8_4way_init( &blakecoin_4w_ctx );
blake256r8_4way( &blakecoin_4w_ctx, vdata, 64 );

4
algo/blake/decred-4way.c
View File

@@ -23,7 +23,7 @@ void decred_hash_4way( void *state, const void *input )
memcpy( &ctx, &blake_mid, sizeof(blake_mid) );
blake256_4way( &ctx, tail, tail_len );
blake256_4way_close( &ctx, vhash );
mm_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
mm128_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
}
int scanhash_decred_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -44,7 +44,7 @@ int scanhash_decred_4way( int thr_id, struct work *work, uint32_t max_nonce,
memcpy( edata, pdata, 180 );
// use the old way until new way updated for size.
mm_interleave_4x32x( vdata, edata, edata, edata, edata, 180*8 );
mm128_interleave_4x32x( vdata, edata, edata, edata, edata, 180*8 );
blake256_4way_init( &blake_mid );
blake256_4way( &blake_mid, vdata, DECRED_MIDSTATE_LEN );

5
algo/blake/decred-gate.c
View File

@@ -140,6 +140,7 @@ bool decred_ready_to_mine( struct work* work, struct stratum_ctx* stratum,
return true;
}
int decred_get_work_data_size() { return DECRED_DATA_SIZE; }
bool register_decred_algo( algo_gate_t* gate )
{
@@ -154,7 +155,7 @@ bool register_decred_algo( algo_gate_t* gate )
gate->optimizations = AVX2_OPT;
gate->get_nonceptr = (void*)&decred_get_nonceptr;
gate->get_max64 = (void*)&get_max64_0x3fffffLL;
gate->display_extra_data = (void*)&decred_decode_extradata;
gate->decode_extra_data = (void*)&decred_decode_extradata;
gate->build_stratum_request = (void*)&decred_be_build_stratum_request;
gate->work_decode = (void*)&std_be_work_decode;
gate->submit_getwork_result = (void*)&std_be_submit_getwork_result;
@@ -163,7 +164,7 @@ bool register_decred_algo( algo_gate_t* gate )
gate->nbits_index = DECRED_NBITS_INDEX;
gate->ntime_index = DECRED_NTIME_INDEX;
gate->nonce_index = DECRED_NONCE_INDEX;
gate->work_data_size = DECRED_DATA_SIZE;
gate->get_work_data_size = (void*)&decred_get_work_data_size;
gate->work_cmp_size = DECRED_WORK_COMPARE_SIZE;
allow_mininginfo = false;
have_gbt = false;

2
algo/blake/decred.c
View File

@@ -268,7 +268,7 @@ bool register_decred_algo( algo_gate_t* gate )
gate->hash = (void*)&decred_hash;
gate->get_nonceptr = (void*)&decred_get_nonceptr;
gate->get_max64 = (void*)&get_max64_0x3fffffLL;
gate->display_extra_data = (void*)&decred_decode_extradata;
gate->decode_extra_data = (void*)&decred_decode_extradata;
gate->build_stratum_request = (void*)&decred_be_build_stratum_request;
gate->work_decode = (void*)&std_be_work_decode;
gate->submit_getwork_result = (void*)&std_be_submit_getwork_result;