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

v3.8.5

Browse Source
This commit is contained in:
Jay D Dee
2018-03-27 20:20:05 -04:00
parent 3363d61524
commit f449c6725f
105 changed files with 4560 additions and 1846 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
algo/blake/blake-hash-4way.c
View File

@@ -30,7 +30,7 @@
* @author Thomas Pornin <thomas.pornin@cryptolog.com>
*/
#if defined (__AVX__)
#if defined (__SSE4_2__)
#include <stddef.h>
#include <string.h>
@@ -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_rotr_32( _mm_xor_si128( d, a ), 16 ); \
d = mm_ror_32( _mm_xor_si128( d, a ), 16 ); \
c = _mm_add_epi32( c, d ); \
b = mm_rotr_32( _mm_xor_si128( b, c ), 12 ); \
b = mm_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_rotr_32( _mm_xor_si128( d, a ), 8 ); \
d = mm_ror_32( _mm_xor_si128( d, a ), 8 ); \
c = _mm_add_epi32( c, d ); \
b = mm_rotr_32( _mm_xor_si128( b, c ), 7 ); \
b = mm_ror_32( _mm_xor_si128( b, c ), 7 ); \
} while (0)
#if SPH_COMPACT_BLAKE_32
@@ -419,14 +419,14 @@ do { \
do { \
a = _mm256_add_epi32( _mm256_add_epi32( _mm256_xor_si256( \
_mm256_set1_epi32( c1 ), m0 ), b ), a ); \
d = mm256_rotr_32( _mm256_xor_si256( d, a ), 16 ); \
d = mm256_ror_32( _mm256_xor_si256( d, a ), 16 ); \
c = _mm256_add_epi32( c, d ); \
b = mm256_rotr_32( _mm256_xor_si256( b, c ), 12 ); \
b = mm256_ror_32( _mm256_xor_si256( b, c ), 12 ); \
a = _mm256_add_epi32( _mm256_add_epi32( _mm256_xor_si256( \
_mm256_set1_epi32( c0 ), m1 ), b ), a ); \
d = mm256_rotr_32( _mm256_xor_si256( d, a ), 8 ); \
d = mm256_ror_32( _mm256_xor_si256( d, a ), 8 ); \
c = _mm256_add_epi32( c, d ); \
b = mm256_rotr_32( _mm256_xor_si256( b, c ), 7 ); \
b = mm256_ror_32( _mm256_xor_si256( b, c ), 7 ); \
} while (0)
#define ROUND_S_8WAY(r) do { \
@@ -445,14 +445,14 @@ do { \
#define GB_4WAY(m0, m1, c0, c1, a, b, c, d) do { \
a = _mm256_add_epi64( _mm256_add_epi64( _mm256_xor_si256( \
_mm256_set_epi64x( c1, c1, c1, c1 ), m0 ), b ), a ); \
d = mm256_rotr_64( _mm256_xor_si256( d, a ), 32 ); \
d = mm256_ror_64( _mm256_xor_si256( d, a ), 32 ); \
c = _mm256_add_epi64( c, d ); \
b = mm256_rotr_64( _mm256_xor_si256( b, c ), 25 ); \
b = mm256_ror_64( _mm256_xor_si256( b, c ), 25 ); \
a = _mm256_add_epi64( _mm256_add_epi64( _mm256_xor_si256( \
_mm256_set_epi64x( c0, c0, c0, c0 ), m1 ), b ), a ); \
d = mm256_rotr_64( _mm256_xor_si256( d, a ), 16 ); \
d = mm256_ror_64( _mm256_xor_si256( d, a ), 16 ); \
c = _mm256_add_epi64( c, d ); \
b = mm256_rotr_64( _mm256_xor_si256( b, c ), 11 ); \
b = mm256_ror_64( _mm256_xor_si256( b, c ), 11 ); \
} while (0)
#if SPH_COMPACT_BLAKE_64

4
algo/blake/blake-hash-4way.h
View File

@@ -37,7 +37,7 @@
#ifndef __BLAKE_HASH_4WAY__
#define __BLAKE_HASH_4WAY__ 1
#ifdef __AVX__
#ifdef __SSE4_2__
#ifdef __cplusplus
extern "C"{
@@ -51,7 +51,7 @@ extern "C"{
#define SPH_SIZE_blake512 512
// With AVX only Blake-256 4 way is available.
// With SSE4.2 only Blake-256 4 way is available.
// With AVX2 Blake-256 8way & Blake-512 4 way are also available.
// Blake-256 4 way

2
algo/blake/blake2s-gate.c
View File

@@ -20,7 +20,7 @@ bool register_blake2s_algo( algo_gate_t* gate )
gate->hash = (void*)&blake2s_hash;
#endif
gate->get_max64 = (void*)&blake2s_get_max64;
gate->optimizations = AVX_OPT | AVX2_OPT;
gate->optimizations = SSE42_OPT | AVX_OPT | AVX2_OPT;
return true;
};

2
algo/blake/blake2s-gate.h
View File

@@ -4,7 +4,7 @@
#include <stdint.h>
#include "algo-gate-api.h"
#if defined(__AVX__)
#if defined(__SSE4_2__)
#define BLAKE2S_4WAY
#endif
#if defined(__AVX2__)

20
algo/blake/blake2s-hash-4way.c
View File

@@ -17,7 +17,7 @@
#include <string.h>
#include <stdio.h>
#if defined(__AVX__)
#if defined(__SSE4_2__)
static const uint32_t blake2s_IV[8] =
{
@@ -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_rotr_32( _mm_xor_si128( d, a ), 16 ); \
d = mm_ror_32( _mm_xor_si128( d, a ), 16 ); \
c = _mm_add_epi32( c, d ); \
b = mm_rotr_32( _mm_xor_si128( b, c ), 12 ); \
b = mm_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_rotr_32( _mm_xor_si128( d, a ), 8 ); \
d = mm_ror_32( _mm_xor_si128( d, a ), 8 ); \
c = _mm_add_epi32( c, d ); \
b = mm_rotr_32( _mm_xor_si128( b, c ), 7 ); \
b = mm_ror_32( _mm_xor_si128( b, c ), 7 ); \
} while(0)
#define ROUND4W(r) \
@@ -210,14 +210,14 @@ int blake2s_8way_compress( blake2s_8way_state *S, const __m256i *block )
do { \
a = _mm256_add_epi32( _mm256_add_epi32( a, b ), \
m[ blake2s_sigma[r][2*i+0] ] ); \
d = mm256_rotr_32( _mm256_xor_si256( d, a ), 16 ); \
d = mm256_ror_32( _mm256_xor_si256( d, a ), 16 ); \
c = _mm256_add_epi32( c, d ); \
b = mm256_rotr_32( _mm256_xor_si256( b, c ), 12 ); \
b = mm256_ror_32( _mm256_xor_si256( b, c ), 12 ); \
a = _mm256_add_epi32( _mm256_add_epi32( a, b ), \
m[ blake2s_sigma[r][2*i+1] ] ); \
d = mm256_rotr_32( _mm256_xor_si256( d, a ), 8 ); \
d = mm256_ror_32( _mm256_xor_si256( d, a ), 8 ); \
c = _mm256_add_epi32( c, d ); \
b = mm256_rotr_32( _mm256_xor_si256( b, c ), 7 ); \
b = mm256_ror_32( _mm256_xor_si256( b, c ), 7 ); \
} while(0)
#define ROUND8W(r) \
@@ -359,4 +359,4 @@ int blake2s( uint8_t *out, const void *in, const void *key, const uint8_t outlen
}
#endif
#endif // __AVX__
#endif // __SSE4_2__

4
algo/blake/blake2s-hash-4way.h
View File

@@ -14,7 +14,7 @@
#ifndef __BLAKE2S_HASH_4WAY_H__
#define __BLAKE2S_HASH_4WAY_H__ 1
#if defined(__AVX__)
#if defined(__SSE4_2__)
#include "avxdefs.h"
@@ -107,6 +107,6 @@ int blake2s_8way_final( blake2s_8way_state *S, void *out, uint8_t outlen );
}
#endif
#endif // __AVX__
#endif // __SSE4_2__
#endif

2
algo/blake/blakecoin-gate.c
View File

@@ -22,7 +22,7 @@ bool register_vanilla_algo( algo_gate_t* gate )
gate->scanhash = (void*)&scanhash_blakecoin;
gate->hash = (void*)&blakecoinhash;
#endif
gate->optimizations = AVX_OPT | AVX2_OPT;
gate->optimizations = SSE42_OPT | AVX_OPT | AVX2_OPT;
gate->get_max64 = (void*)&blakecoin_get_max64;
return true;
}

2
algo/blake/blakecoin-gate.h
View File

@@ -4,7 +4,7 @@
#include "algo-gate-api.h"
#include <stdint.h>
#if defined(__AVX__)
#if defined(__SSE4_2__)
#define BLAKECOIN_4WAY
#endif
#if defined(__AVX2__)

2
algo/blake/decred-gate.h
View File

@@ -18,7 +18,7 @@
// uint64_t *hashes_done );
#endif
#if defined(__AVX2__)
#if defined(__SSE4_2__)
#define DECRED_4WAY
#endif