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

v3.9.6.2

Browse Source
This commit is contained in:
Jay D Dee
2019-07-30 10:16:43 -04:00
parent a51f59086b
commit 9d49e0be7a
66 changed files with 1949 additions and 1470 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
algo/argon2/argon2d/argon2d/core.c
View File

@@ -28,6 +28,7 @@
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <mm_malloc.h>
#include "core.h"
#include "argon2d_thread.h"
@@ -99,7 +100,8 @@ int allocate_memory(const argon2_context *context, uint8_t **memory,
if (context->allocate_cbk) {
(context->allocate_cbk)(memory, memory_size);
} else {
*memory = malloc(memory_size);
*memory = _mm_malloc( memory_size, 64 );
// *memory = malloc(memory_size);
}
if (*memory == NULL) {
@@ -116,7 +118,8 @@ void free_memory(const argon2_context *context, uint8_t *memory,
if (context->free_cbk) {
(context->free_cbk)(memory, memory_size);
} else {
free(memory);
// free(memory);
_mm_free( memory );
}
}

8
algo/argon2/argon2d/argon2d/opt.c
View File

@@ -96,14 +96,14 @@ static void fill_block(__m256i *state, const block *ref_block,
if (with_xor) {
for (i = 0; i < ARGON2_HWORDS_IN_BLOCK; i++) {
state[i] = _mm256_xor_si256(
state[i], _mm256_loadu_si256((const __m256i *)ref_block->v + i));
state[i], _mm256_load_si256((const __m256i *)ref_block->v + i));
block_XY[i] = _mm256_xor_si256(
state[i], _mm256_loadu_si256((const __m256i *)next_block->v + i));
state[i], _mm256_load_si256((const __m256i *)next_block->v + i));
}
} else {
for (i = 0; i < ARGON2_HWORDS_IN_BLOCK; i++) {
block_XY[i] = state[i] = _mm256_xor_si256(
state[i], _mm256_loadu_si256((const __m256i *)ref_block->v + i));
state[i], _mm256_load_si256((const __m256i *)ref_block->v + i));
}
}
@@ -139,7 +139,7 @@ static void fill_block(__m256i *state, const block *ref_block,
for (i = 0; i < ARGON2_HWORDS_IN_BLOCK; i++) {
state[i] = _mm256_xor_si256(state[i], block_XY[i]);
_mm256_storeu_si256((__m256i *)next_block->v + i, state[i]);
_mm256_store_si256((__m256i *)next_block->v + i, state[i]);
}
}

105
algo/argon2/argon2d/blake2/blamka-round-opt.h
View File

@@ -29,6 +29,8 @@
#include <x86intrin.h>
#endif
#include "simd-utils.h"
#if !defined(__AVX512F__)
#if !defined(__AVX2__)
#if !defined(__XOP__)
@@ -182,64 +184,63 @@ static BLAKE2_INLINE __m128i fBlaMka(__m128i x, __m128i y) {
#include <immintrin.h>
#define rotr32(x) _mm256_shuffle_epi32(x, _MM_SHUFFLE(2, 3, 0, 1))
#define rotr24(x) _mm256_shuffle_epi8(x, _mm256_setr_epi8(3, 4, 5, 6, 7, 0, 1, 2, 11, 12, 13, 14, 15, 8, 9, 10, 3, 4, 5, 6, 7, 0, 1, 2, 11, 12, 13, 14, 15, 8, 9, 10))
#define rotr16(x) _mm256_shuffle_epi8(x, _mm256_setr_epi8(2, 3, 4, 5, 6, 7, 0, 1, 10, 11, 12, 13, 14, 15, 8, 9, 2, 3, 4, 5, 6, 7, 0, 1, 10, 11, 12, 13, 14, 15, 8, 9))
#define rotr63(x) _mm256_xor_si256(_mm256_srli_epi64((x), 63), _mm256_add_epi64((x), (x)))
#define rotr32 mm256_swap32_64
#define rotr24 mm256_ror3x8_64
#define rotr16 mm256_ror1x16_64
#define rotr63( x ) mm256_rol_64( x, 1 )
//#define rotr32(x) _mm256_shuffle_epi32(x, _MM_SHUFFLE(2, 3, 0, 1))
//#define rotr24(x) _mm256_shuffle_epi8(x, _mm256_setr_epi8(3, 4, 5, 6, 7, 0, 1, 2, 11, 12, 13, 14, 15, 8, 9, 10, 3, 4, 5, 6, 7, 0, 1, 2, 11, 12, 13, 14, 15, 8, 9, 10))
//#define rotr16(x) _mm256_shuffle_epi8(x, _mm256_setr_epi8(2, 3, 4, 5, 6, 7, 0, 1, 10, 11, 12, 13, 14, 15, 8, 9, 2, 3, 4, 5, 6, 7, 0, 1, 10, 11, 12, 13, 14, 15, 8, 9))
//#define rotr63(x) _mm256_xor_si256(_mm256_srli_epi64((x), 63), _mm256_add_epi64((x), (x)))
#define G1_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
do { \
__m256i ml = _mm256_mul_epu32(A0, B0); \
ml = _mm256_add_epi64(ml, ml); \
A0 = _mm256_add_epi64(A0, _mm256_add_epi64(B0, ml)); \
__m256i ml0, ml1; \
ml0 = _mm256_mul_epu32(A0, B0); \
ml1 = _mm256_mul_epu32(A1, B1); \
ml0 = _mm256_add_epi64(ml0, ml0); \
ml1 = _mm256_add_epi64(ml1, ml1); \
A0 = _mm256_add_epi64(A0, _mm256_add_epi64(B0, ml0)); \
A1 = _mm256_add_epi64(A1, _mm256_add_epi64(B1, ml1)); \
D0 = _mm256_xor_si256(D0, A0); \
D0 = rotr32(D0); \
\
ml = _mm256_mul_epu32(C0, D0); \
ml = _mm256_add_epi64(ml, ml); \
C0 = _mm256_add_epi64(C0, _mm256_add_epi64(D0, ml)); \
\
B0 = _mm256_xor_si256(B0, C0); \
B0 = rotr24(B0); \
\
ml = _mm256_mul_epu32(A1, B1); \
ml = _mm256_add_epi64(ml, ml); \
A1 = _mm256_add_epi64(A1, _mm256_add_epi64(B1, ml)); \
D1 = _mm256_xor_si256(D1, A1); \
D0 = rotr32(D0); \
D1 = rotr32(D1); \
\
ml = _mm256_mul_epu32(C1, D1); \
ml = _mm256_add_epi64(ml, ml); \
C1 = _mm256_add_epi64(C1, _mm256_add_epi64(D1, ml)); \
\
ml0 = _mm256_mul_epu32(C0, D0); \
ml1 = _mm256_mul_epu32(C1, D1); \
ml0 = _mm256_add_epi64(ml0, ml0); \
ml1 = _mm256_add_epi64(ml1, ml1); \
C0 = _mm256_add_epi64(C0, _mm256_add_epi64(D0, ml0)); \
C1 = _mm256_add_epi64(C1, _mm256_add_epi64(D1, ml1)); \
B0 = _mm256_xor_si256(B0, C0); \
B1 = _mm256_xor_si256(B1, C1); \
B0 = rotr24(B0); \
B1 = rotr24(B1); \
} while((void)0, 0);
#define G2_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
do { \
__m256i ml = _mm256_mul_epu32(A0, B0); \
ml = _mm256_add_epi64(ml, ml); \
A0 = _mm256_add_epi64(A0, _mm256_add_epi64(B0, ml)); \
__m256i ml0, ml1; \
ml0 = _mm256_mul_epu32(A0, B0); \
ml1 = _mm256_mul_epu32(A1, B1); \
ml0 = _mm256_add_epi64(ml0, ml0); \
ml1 = _mm256_add_epi64(ml1, ml1); \
A0 = _mm256_add_epi64(A0, _mm256_add_epi64(B0, ml0)); \
A1 = _mm256_add_epi64(A1, _mm256_add_epi64(B1, ml1)); \
D0 = _mm256_xor_si256(D0, A0); \
D0 = rotr16(D0); \
\
ml = _mm256_mul_epu32(C0, D0); \
ml = _mm256_add_epi64(ml, ml); \
C0 = _mm256_add_epi64(C0, _mm256_add_epi64(D0, ml)); \
B0 = _mm256_xor_si256(B0, C0); \
B0 = rotr63(B0); \
\
ml = _mm256_mul_epu32(A1, B1); \
ml = _mm256_add_epi64(ml, ml); \
A1 = _mm256_add_epi64(A1, _mm256_add_epi64(B1, ml)); \
D1 = _mm256_xor_si256(D1, A1); \
D0 = rotr16(D0); \
D1 = rotr16(D1); \
\
ml = _mm256_mul_epu32(C1, D1); \
ml = _mm256_add_epi64(ml, ml); \
C1 = _mm256_add_epi64(C1, _mm256_add_epi64(D1, ml)); \
ml0 = _mm256_mul_epu32(C0, D0); \
ml1 = _mm256_mul_epu32(C1, D1); \
ml0 = _mm256_add_epi64(ml0, ml0); \
ml1 = _mm256_add_epi64(ml1, ml1); \
C0 = _mm256_add_epi64(C0, _mm256_add_epi64(D0, ml0)); \
C1 = _mm256_add_epi64(C1, _mm256_add_epi64(D1, ml1)); \
B0 = _mm256_xor_si256(B0, C0); \
B1 = _mm256_xor_si256(B1, C1); \
B0 = rotr63(B0); \
B1 = rotr63(B1); \
} while((void)0, 0);
@@ -259,16 +260,14 @@ static BLAKE2_INLINE __m128i fBlaMka(__m128i x, __m128i y) {
__m256i tmp1 = _mm256_blend_epi32(B0, B1, 0xCC); \
__m256i tmp2 = _mm256_blend_epi32(B0, B1, 0x33); \
B1 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
B0 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
\
tmp1 = C0; \
B0 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
C0 = C1; \
C1 = tmp1; \
\
tmp1 = _mm256_blend_epi32(D0, D1, 0xCC); \
tmp2 = _mm256_blend_epi32(D0, D1, 0x33); \
D0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
C1 = tmp1; \
tmp1 = _mm256_blend_epi32(D0, D1, 0xCC); \
D1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
D0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
} while(0);
#define UNDIAGONALIZE_1(A0, B0, C0, D0, A1, B1, C1, D1) \
@@ -287,16 +286,14 @@ static BLAKE2_INLINE __m128i fBlaMka(__m128i x, __m128i y) {
__m256i tmp1 = _mm256_blend_epi32(B0, B1, 0xCC); \
__m256i tmp2 = _mm256_blend_epi32(B0, B1, 0x33); \
B0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
B1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
\
tmp1 = C0; \
B1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
C0 = C1; \
C1 = tmp1; \
\
tmp1 = _mm256_blend_epi32(D0, D1, 0x33); \
tmp2 = _mm256_blend_epi32(D0, D1, 0xCC); \
D0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
C1 = tmp1; \
tmp1 = _mm256_blend_epi32(D0, D1, 0x33); \
D1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
D0 = _mm256_permute4x64_epi64(tmp1, _MM_SHUFFLE(2,3,0,1)); \
} while((void)0, 0);
#define BLAKE2_ROUND_1(A0, A1, B0, B1, C0, C1, D0, D1) \