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

v3.8.6.1

Browse Source
This commit is contained in:
Jay D Dee
2018-04-06 11:42:01 -04:00
parent dd5e552357
commit c7efa50aad
22 changed files with 773 additions and 393 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
algo/argon2/argon2d/blake2/blake2-impl.h
View File

@@ -4,7 +4,7 @@
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*
* You may use this work under the terms of a Creative Commons CC0 1.0
* You may use this work under the terms of a Creative Commons CC0 1.0
* License/Waiver or the Apache Public License 2.0, at your option. The terms of
* these licenses can be found at:
*
@@ -153,4 +153,4 @@ static BLAKE2_INLINE uint64_t rotr64(const uint64_t w, const unsigned c) {
void clear_internal_memory(void *v, size_t n);
#endif
#endif

6
algo/argon2/argon2d/blake2/blake2.h
View File

@@ -4,7 +4,7 @@
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*
* You may use this work under the terms of a Creative Commons CC0 1.0
* You may use this work under the terms of a Creative Commons CC0 1.0
* License/Waiver or the Apache Public License 2.0, at your option. The terms of
* these licenses can be found at:
*
@@ -78,7 +78,7 @@ int blake2b_final(blake2b_state *S, void *out, size_t outlen);
/* Simple API */
int blake2b(void *out, size_t outlen, const void *in, size_t inlen,
const void *key, size_t keylen);
const void *key, size_t keylen);
/* Argon2 Team - Begin Code */
int blake2b_long(void *out, size_t outlen, const void *in, size_t inlen);
@@ -88,4 +88,4 @@ int blake2b_long(void *out, size_t outlen, const void *in, size_t inlen);
}
#endif
#endif
#endif

4
algo/argon2/argon2d/blake2/blake2b.c
View File

@@ -4,7 +4,7 @@
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*
* You may use this work under the terms of a Creative Commons CC0 1.0
* You may use this work under the terms of a Creative Commons CC0 1.0
* License/Waiver or the Apache Public License 2.0, at your option. The terms of
* these licenses can be found at:
*
@@ -387,4 +387,4 @@ fail:
return ret;
#undef TRY
}
/* Argon2 Team - End Code */
/* Argon2 Team - End Code */

295
algo/argon2/argon2d/blake2/blamka-round-opt.h
View File

@@ -4,7 +4,7 @@
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*
* You may use this work under the terms of a Creative Commons CC0 1.0
* You may use this work under the terms of a Creative Commons CC0 1.0
* License/Waiver or the Apache Public License 2.0, at your option. The terms of
* these licenses can be found at:
*
@@ -29,6 +29,8 @@
#include <x86intrin.h>
#endif
#if !defined(__AVX512F__)
#if !defined(__AVX2__)
#if !defined(__XOP__)
#if defined(__SSSE3__)
#define r16 \
@@ -176,5 +178,294 @@ static BLAKE2_INLINE __m128i fBlaMka(__m128i x, __m128i y) {
\
UNDIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1); \
} while ((void)0, 0)
#else /* __AVX2__ */
#endif
#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 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)); \
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); \
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)); \
\
B1 = _mm256_xor_si256(B1, C1); \
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)); \
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); \
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)); \
B1 = _mm256_xor_si256(B1, C1); \
B1 = rotr63(B1); \
} while((void)0, 0);
#define DIAGONALIZE_1(A0, B0, C0, D0, A1, B1, C1, D1) \
do { \
B0 = _mm256_permute4x64_epi64(B0, _MM_SHUFFLE(0, 3, 2, 1)); \
C0 = _mm256_permute4x64_epi64(C0, _MM_SHUFFLE(1, 0, 3, 2)); \
D0 = _mm256_permute4x64_epi64(D0, _MM_SHUFFLE(2, 1, 0, 3)); \
\
B1 = _mm256_permute4x64_epi64(B1, _MM_SHUFFLE(0, 3, 2, 1)); \
C1 = _mm256_permute4x64_epi64(C1, _MM_SHUFFLE(1, 0, 3, 2)); \
D1 = _mm256_permute4x64_epi64(D1, _MM_SHUFFLE(2, 1, 0, 3)); \
} while((void)0, 0);
#define DIAGONALIZE_2(A0, A1, B0, B1, C0, C1, D0, D1) \
do { \
__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; \
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)); \
D1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
} while(0);
#define UNDIAGONALIZE_1(A0, B0, C0, D0, A1, B1, C1, D1) \
do { \
B0 = _mm256_permute4x64_epi64(B0, _MM_SHUFFLE(2, 1, 0, 3)); \
C0 = _mm256_permute4x64_epi64(C0, _MM_SHUFFLE(1, 0, 3, 2)); \
D0 = _mm256_permute4x64_epi64(D0, _MM_SHUFFLE(0, 3, 2, 1)); \
\
B1 = _mm256_permute4x64_epi64(B1, _MM_SHUFFLE(2, 1, 0, 3)); \
C1 = _mm256_permute4x64_epi64(C1, _MM_SHUFFLE(1, 0, 3, 2)); \
D1 = _mm256_permute4x64_epi64(D1, _MM_SHUFFLE(0, 3, 2, 1)); \
} while((void)0, 0);
#define UNDIAGONALIZE_2(A0, A1, B0, B1, C0, C1, D0, D1) \
do { \
__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; \
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)); \
D1 = _mm256_permute4x64_epi64(tmp2, _MM_SHUFFLE(2,3,0,1)); \
} while((void)0, 0);
#define BLAKE2_ROUND_1(A0, A1, B0, B1, C0, C1, D0, D1) \
do{ \
G1_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
G2_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
\
DIAGONALIZE_1(A0, B0, C0, D0, A1, B1, C1, D1) \
\
G1_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
G2_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
\
UNDIAGONALIZE_1(A0, B0, C0, D0, A1, B1, C1, D1) \
} while((void)0, 0);
#define BLAKE2_ROUND_2(A0, A1, B0, B1, C0, C1, D0, D1) \
do{ \
G1_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
G2_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
\
DIAGONALIZE_2(A0, A1, B0, B1, C0, C1, D0, D1) \
\
G1_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
G2_AVX2(A0, A1, B0, B1, C0, C1, D0, D1) \
\
UNDIAGONALIZE_2(A0, A1, B0, B1, C0, C1, D0, D1) \
} while((void)0, 0);
#endif /* __AVX2__ */
#else /* __AVX512F__ */
#include <immintrin.h>
#define ror64(x, n) _mm512_ror_epi64((x), (n))
static __m512i muladd(__m512i x, __m512i y)
{
__m512i z = _mm512_mul_epu32(x, y);
return _mm512_add_epi64(_mm512_add_epi64(x, y), _mm512_add_epi64(z, z));
}
#define G1(A0, B0, C0, D0, A1, B1, C1, D1) \
do { \
A0 = muladd(A0, B0); \
A1 = muladd(A1, B1); \
\
D0 = _mm512_xor_si512(D0, A0); \
D1 = _mm512_xor_si512(D1, A1); \
\
D0 = ror64(D0, 32); \
D1 = ror64(D1, 32); \
\
C0 = muladd(C0, D0); \
C1 = muladd(C1, D1); \
\
B0 = _mm512_xor_si512(B0, C0); \
B1 = _mm512_xor_si512(B1, C1); \
\
B0 = ror64(B0, 24); \
B1 = ror64(B1, 24); \
} while ((void)0, 0)
#define G2(A0, B0, C0, D0, A1, B1, C1, D1) \
do { \
A0 = muladd(A0, B0); \
A1 = muladd(A1, B1); \
\
D0 = _mm512_xor_si512(D0, A0); \
D1 = _mm512_xor_si512(D1, A1); \
\
D0 = ror64(D0, 16); \
D1 = ror64(D1, 16); \
\
C0 = muladd(C0, D0); \
C1 = muladd(C1, D1); \
\
B0 = _mm512_xor_si512(B0, C0); \
B1 = _mm512_xor_si512(B1, C1); \
\
B0 = ror64(B0, 63); \
B1 = ror64(B1, 63); \
} while ((void)0, 0)
#define DIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1) \
do { \
B0 = _mm512_permutex_epi64(B0, _MM_SHUFFLE(0, 3, 2, 1)); \
B1 = _mm512_permutex_epi64(B1, _MM_SHUFFLE(0, 3, 2, 1)); \
\
C0 = _mm512_permutex_epi64(C0, _MM_SHUFFLE(1, 0, 3, 2)); \
C1 = _mm512_permutex_epi64(C1, _MM_SHUFFLE(1, 0, 3, 2)); \
\
D0 = _mm512_permutex_epi64(D0, _MM_SHUFFLE(2, 1, 0, 3)); \
D1 = _mm512_permutex_epi64(D1, _MM_SHUFFLE(2, 1, 0, 3)); \
} while ((void)0, 0)
#define UNDIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1) \
do { \
B0 = _mm512_permutex_epi64(B0, _MM_SHUFFLE(2, 1, 0, 3)); \
B1 = _mm512_permutex_epi64(B1, _MM_SHUFFLE(2, 1, 0, 3)); \
\
C0 = _mm512_permutex_epi64(C0, _MM_SHUFFLE(1, 0, 3, 2)); \
C1 = _mm512_permutex_epi64(C1, _MM_SHUFFLE(1, 0, 3, 2)); \
\
D0 = _mm512_permutex_epi64(D0, _MM_SHUFFLE(0, 3, 2, 1)); \
D1 = _mm512_permutex_epi64(D1, _MM_SHUFFLE(0, 3, 2, 1)); \
} while ((void)0, 0)
#define BLAKE2_ROUND(A0, B0, C0, D0, A1, B1, C1, D1) \
do { \
G1(A0, B0, C0, D0, A1, B1, C1, D1); \
G2(A0, B0, C0, D0, A1, B1, C1, D1); \
\
DIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1); \
\
G1(A0, B0, C0, D0, A1, B1, C1, D1); \
G2(A0, B0, C0, D0, A1, B1, C1, D1); \
\
UNDIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1); \
} while ((void)0, 0)
#define SWAP_HALVES(A0, A1) \
do { \
__m512i t0, t1; \
t0 = _mm512_shuffle_i64x2(A0, A1, _MM_SHUFFLE(1, 0, 1, 0)); \
t1 = _mm512_shuffle_i64x2(A0, A1, _MM_SHUFFLE(3, 2, 3, 2)); \
A0 = t0; \
A1 = t1; \
} while((void)0, 0)
#define SWAP_QUARTERS(A0, A1) \
do { \
SWAP_HALVES(A0, A1); \
A0 = _mm512_permutexvar_epi64(_mm512_setr_epi64(0, 1, 4, 5, 2, 3, 6, 7), A0); \
A1 = _mm512_permutexvar_epi64(_mm512_setr_epi64(0, 1, 4, 5, 2, 3, 6, 7), A1); \
} while((void)0, 0)
#define UNSWAP_QUARTERS(A0, A1) \
do { \
A0 = _mm512_permutexvar_epi64(_mm512_setr_epi64(0, 1, 4, 5, 2, 3, 6, 7), A0); \
A1 = _mm512_permutexvar_epi64(_mm512_setr_epi64(0, 1, 4, 5, 2, 3, 6, 7), A1); \
SWAP_HALVES(A0, A1); \
} while((void)0, 0)
#define BLAKE2_ROUND_1(A0, C0, B0, D0, A1, C1, B1, D1) \
do { \
SWAP_HALVES(A0, B0); \
SWAP_HALVES(C0, D0); \
SWAP_HALVES(A1, B1); \
SWAP_HALVES(C1, D1); \
BLAKE2_ROUND(A0, B0, C0, D0, A1, B1, C1, D1); \
SWAP_HALVES(A0, B0); \
SWAP_HALVES(C0, D0); \
SWAP_HALVES(A1, B1); \
SWAP_HALVES(C1, D1); \
} while ((void)0, 0)
#define BLAKE2_ROUND_2(A0, A1, B0, B1, C0, C1, D0, D1) \
do { \
SWAP_QUARTERS(A0, A1); \
SWAP_QUARTERS(B0, B1); \
SWAP_QUARTERS(C0, C1); \
SWAP_QUARTERS(D0, D1); \
BLAKE2_ROUND(A0, B0, C0, D0, A1, B1, C1, D1); \
UNSWAP_QUARTERS(A0, A1); \
UNSWAP_QUARTERS(B0, B1); \
UNSWAP_QUARTERS(C0, C1); \
UNSWAP_QUARTERS(D0, D1); \
} while ((void)0, 0)
#endif /* __AVX512F__ */
#endif /* BLAKE_ROUND_MKA_OPT_H */

6
algo/argon2/argon2d/blake2/blamka-round-ref.h
View File

@@ -4,7 +4,7 @@
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*
* You may use this work under the terms of a Creative Commons CC0 1.0
* You may use this work under the terms of a Creative Commons CC0 1.0
* License/Waiver or the Apache Public License 2.0, at your option. The terms of
* these licenses can be found at:
*
@@ -21,7 +21,7 @@
#include "blake2.h"
#include "blake2-impl.h"
/*designed by the Lyra PHC team */
/* designed by the Lyra PHC team */
static BLAKE2_INLINE uint64_t fBlaMka(uint64_t x, uint64_t y) {
const uint64_t m = UINT64_C(0xFFFFFFFF);
const uint64_t xy = (x & m) * (y & m);
@@ -53,4 +53,4 @@ static BLAKE2_INLINE uint64_t fBlaMka(uint64_t x, uint64_t y) {
G(v3, v4, v9, v14); \
} while ((void)0, 0)
#endif
#endif