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

v3.19.3

Browse Source
This commit is contained in:
Jay D Dee
2022-01-07 12:07:38 -05:00
parent 0e3945ddb5
commit 17ccbc328f
17 changed files with 657 additions and 606 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -37,6 +37,13 @@
#if defined(__AVX512F__)
static inline __m512i blamka( __m512i x, __m512i y )
{
__m512i xy = _mm512_mul_epu32( x, y );
return _mm512_add_epi64( _mm512_add_epi64( x, y ),
_mm512_add_epi64( xy, xy ) );
}
static void fill_block( __m512i *state, const block *ref_block,
block *next_block, int with_xor )
{

29
algo/argon2/argon2d/blake2/blamka-round-opt.h
View File

@@ -328,9 +328,7 @@ static BLAKE2_INLINE __m128i fBlaMka(__m128i x, __m128i y) {
#include <immintrin.h>
#define ROR64(x, n) _mm512_ror_epi64((x), (n))
static __m512i muladd(__m512i x, __m512i y)
static inline __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));
@@ -344,8 +342,8 @@ static __m512i muladd(__m512i x, __m512i y)
D0 = _mm512_xor_si512(D0, A0); \
D1 = _mm512_xor_si512(D1, A1); \
\
D0 = ROR64(D0, 32); \
D1 = ROR64(D1, 32); \
D0 = _mm512_ror_epi64(D0, 32); \
D1 = _mm512_ror_epi64(D1, 32); \
\
C0 = muladd(C0, D0); \
C1 = muladd(C1, D1); \
@@ -353,8 +351,8 @@ static __m512i muladd(__m512i x, __m512i y)
B0 = _mm512_xor_si512(B0, C0); \
B1 = _mm512_xor_si512(B1, C1); \
\
B0 = ROR64(B0, 24); \
B1 = ROR64(B1, 24); \
B0 = _mm512_ror_epi64(B0, 24); \
B1 = _mm512_ror_epi64(B1, 24); \
} while ((void)0, 0)
#define G2(A0, B0, C0, D0, A1, B1, C1, D1) \
@@ -365,8 +363,8 @@ static __m512i muladd(__m512i x, __m512i y)
D0 = _mm512_xor_si512(D0, A0); \
D1 = _mm512_xor_si512(D1, A1); \
\
D0 = ROR64(D0, 16); \
D1 = ROR64(D1, 16); \
D0 = _mm512_ror_epi64(D0, 16); \
D1 = _mm512_ror_epi64(D1, 16); \
\
C0 = muladd(C0, D0); \
C1 = muladd(C1, D1); \
@@ -374,8 +372,8 @@ static __m512i muladd(__m512i x, __m512i y)
B0 = _mm512_xor_si512(B0, C0); \
B1 = _mm512_xor_si512(B1, C1); \
\
B0 = ROR64(B0, 63); \
B1 = ROR64(B1, 63); \
B0 = _mm512_ror_epi64(B0, 63); \
B1 = _mm512_ror_epi64(B1, 63); \
} while ((void)0, 0)
#define DIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1) \
@@ -417,11 +415,10 @@ static __m512i muladd(__m512i x, __m512i y)
#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; \
__m512i t; \
t = _mm512_shuffle_i64x2(A0, A1, _MM_SHUFFLE(1, 0, 1, 0)); \
A1 = _mm512_shuffle_i64x2(A0, A1, _MM_SHUFFLE(3, 2, 3, 2)); \
A0 = t; \
} while((void)0, 0)
#define SWAP_QUARTERS(A0, A1) \