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

v3.7.4

Browse Source
This commit is contained in:
Jay D Dee
2021-09-29 17:31:16 -04:00
parent 9b905fccc8
commit 2cd1507c2e
80 changed files with 8145 additions and 2097 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
algo/scrypt/neoscrypt.c
View File

@@ -69,8 +69,12 @@ typedef unsigned int uint;
#define SCRYPT_HASH_BLOCK_SIZE 64U
#define SCRYPT_HASH_DIGEST_SIZE 32U
#define ROTL32(a,b) (((a) << (b)) | ((a) >> (32 - b)))
#define ROTR32(a,b) (((a) >> (b)) | ((a) << (32 - b)))
//#define ROTL32(a,b) (((a) << (b)) | ((a) >> (32 - b)))
//#define ROTR32(a,b) (((a) >> (b)) | ((a) << (32 - b)))
#define ROTL32(a,b) rol32(a,b)
#define ROTR32(a,b) ror32(a,b)
#define U8TO32_BE(p) \
(((uint32_t)((p)[0]) << 24) | ((uint32_t)((p)[1]) << 16) | \

3981
algo/scrypt/scrypt-core-4way.c Normal file
View File

File diff suppressed because it is too large Load Diff

70
algo/scrypt/scrypt-core-4way.h Normal file
View File

@@ -0,0 +1,70 @@
#ifndef SCRYPT_CORE_4WAY_H__
#define SCRYPT_CORE_4WAY_H__
#include "simd-utils.h"
#include <stdlib.h>
#include <stdint.h>
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
void scrypt_core_16way( __m512i *X, __m512i *V, const uint32_t N );
// Serial SIMD over 4 way parallel
void scrypt_core_simd128_4way( __m128i *X, __m128i *V, const uint32_t N );
// 4 way parallel over serial SIMD
void scrypt_core_4way_simd128( __m512i *X, __m512i *V, const uint32_t N );
#endif
#if defined(__AVX2__)
void scrypt_core_8way( __m256i *X, __m256i *V, uint32_t N );
// 2 way parallel over SIMD128
void scrypt_core_2way_simd128( __m256i *X, __m256i *V, const uint32_t N );
// Double buffered 2 way parallel over SIMD128
void scrypt_core_2way_simd128_2buf( __m256i *X, __m256i *V, const uint32_t N );
// Triplee buffered 2 way parallel over SIMD128
void scrypt_core_2way_simd128_3buf( __m256i *X, __m256i *V, const uint32_t N );
// Serial SIMD128 over 2 way parallel
void scrypt_core_simd128_2way( uint64_t *X, uint64_t *V, const uint32_t N );
// Double buffered simd over parallel
void scrypt_core_simd128_2way_2buf( uint64_t *X, uint64_t *V, const uint32_t N );
// Triple buffered 2 way
void scrypt_core_simd128_2way_3buf( uint64_t *X, uint64_t *V, const uint32_t N );
// Quadruple buffered
void scrypt_core_simd128_2way_4buf( uint64_t *X, uint64_t *V, const uint32_t N );
#endif
#if defined(__SSE2__)
// Parallel 4 way, 4x memory
void scrypt_core_4way( __m128i *X, __m128i *V, const uint32_t N );
// Linear SIMD 1 way, 1x memory, lowest
void scrypt_core_simd128( uint32_t *X, uint32_t *V, const uint32_t N );
// Double buffered, 2x memory
void scrypt_core_simd128_2buf( uint32_t *X, uint32_t *V, const uint32_t N );
// Triple buffered
void scrypt_core_simd128_3buf( uint32_t *X, uint32_t *V, const uint32_t N );
// Quadruple buffered, 4x memory
void scrypt_core_simd128_4buf( uint32_t *X, uint32_t *V, const uint32_t N );
#endif
// For reference only
void scrypt_core_1way( uint32_t *X, uint32_t *V, const uint32_t N );
#endif

206
algo/scrypt/scrypt-core-ref.c Normal file
View File

@@ -0,0 +1,206 @@
#include "scrypt-core-ref.h"
#define ROTL(a, b) (((a) << (b)) | ((a) >> (32 - (b))))
static void xor_salsa8(uint32_t * const B, const uint32_t * const C)
{
uint32_t x0 = (B[ 0] ^= C[ 0]),
x1 = (B[ 1] ^= C[ 1]),
x2 = (B[ 2] ^= C[ 2]),
x3 = (B[ 3] ^= C[ 3]);
uint32_t x4 = (B[ 4] ^= C[ 4]),
x5 = (B[ 5] ^= C[ 5]),
x6 = (B[ 6] ^= C[ 6]),
x7 = (B[ 7] ^= C[ 7]);
uint32_t x8 = (B[ 8] ^= C[ 8]),
x9 = (B[ 9] ^= C[ 9]),
xa = (B[10] ^= C[10]),
xb = (B[11] ^= C[11]);
uint32_t xc = (B[12] ^= C[12]),
xd = (B[13] ^= C[13]),
xe = (B[14] ^= C[14]),
xf = (B[15] ^= C[15]);
/* Operate on columns. */
x4 ^= ROTL(x0 + xc, 7);
x9 ^= ROTL(x5 + x1, 7);
xe ^= ROTL(xa + x6, 7);
x3 ^= ROTL(xf + xb, 7);
x8 ^= ROTL(x4 + x0, 9);
xd ^= ROTL(x9 + x5, 9);
x2 ^= ROTL(xe + xa, 9);
x7 ^= ROTL(x3 + xf, 9);
xc ^= ROTL(x8 + x4, 13);
x1 ^= ROTL(xd + x9, 13);
x6 ^= ROTL(x2 + xe, 13);
xb ^= ROTL(x7 + x3, 13);
x0 ^= ROTL(xc + x8, 18);
x5 ^= ROTL(x1 + xd, 18);
xa ^= ROTL(x6 + x2, 18);
xf ^= ROTL(xb + x7, 18);
/* Operate on rows. */
x1 ^= ROTL(x0 + x3, 7);
x6 ^= ROTL(x5 + x4, 7);
xb ^= ROTL(xa + x9, 7);
xc ^= ROTL(xf + xe, 7);
x2 ^= ROTL(x1 + x0, 9);
x7 ^= ROTL(x6 + x5, 9);
x8 ^= ROTL(xb + xa, 9);
xd ^= ROTL(xc + xf, 9);
x3 ^= ROTL(x2 + x1, 13);
x4 ^= ROTL(x7 + x6, 13);
x9 ^= ROTL(x8 + xb, 13);
xe ^= ROTL(xd + xc, 13);
x0 ^= ROTL(x3 + x2, 18);
x5 ^= ROTL(x4 + x7, 18);
xa ^= ROTL(x9 + x8, 18);
xf ^= ROTL(xe + xd, 18);
/* Operate on columns. */
x4 ^= ROTL(x0 + xc, 7);
x9 ^= ROTL(x5 + x1, 7);
xe ^= ROTL(xa + x6, 7);
x3 ^= ROTL(xf + xb, 7);
x8 ^= ROTL(x4 + x0, 9);
xd ^= ROTL(x9 + x5, 9);
x2 ^= ROTL(xe + xa, 9);
x7 ^= ROTL(x3 + xf, 9);
xc ^= ROTL(x8 + x4, 13);
x1 ^= ROTL(xd + x9, 13);
x6 ^= ROTL(x2 + xe, 13);
xb ^= ROTL(x7 + x3, 13);
x0 ^= ROTL(xc + x8, 18);
x5 ^= ROTL(x1 + xd, 18);
xa ^= ROTL(x6 + x2, 18);
xf ^= ROTL(xb + x7, 18);
/* Operate on rows. */
x1 ^= ROTL(x0 + x3, 7);
x6 ^= ROTL(x5 + x4, 7);
xb ^= ROTL(xa + x9, 7);
xc ^= ROTL(xf + xe, 7);
x2 ^= ROTL(x1 + x0, 9);
x7 ^= ROTL(x6 + x5, 9);
x8 ^= ROTL(xb + xa, 9);
xd ^= ROTL(xc + xf, 9);
x3 ^= ROTL(x2 + x1, 13);
x4 ^= ROTL(x7 + x6, 13);
x9 ^= ROTL(x8 + xb, 13);
xe ^= ROTL(xd + xc, 13);
x0 ^= ROTL(x3 + x2, 18);
x5 ^= ROTL(x4 + x7, 18);
xa ^= ROTL(x9 + x8, 18);
xf ^= ROTL(xe + xd, 18);
/* Operate on columns. */
x4 ^= ROTL(x0 + xc, 7);
x9 ^= ROTL(x5 + x1, 7);
xe ^= ROTL(xa + x6, 7);
x3 ^= ROTL(xf + xb, 7);
x8 ^= ROTL(x4 + x0, 9);
xd ^= ROTL(x9 + x5, 9);
x2 ^= ROTL(xe + xa, 9);
x7 ^= ROTL(x3 + xf, 9);
xc ^= ROTL(x8 + x4, 13);
x1 ^= ROTL(xd + x9, 13);
x6 ^= ROTL(x2 + xe, 13);
xb ^= ROTL(x7 + x3, 13);
x0 ^= ROTL(xc + x8, 18);
x5 ^= ROTL(x1 + xd, 18);
xa ^= ROTL(x6 + x2, 18);
xf ^= ROTL(xb + x7, 18);
/* Operate on rows. */
x1 ^= ROTL(x0 + x3, 7);
x6 ^= ROTL(x5 + x4, 7);
xb ^= ROTL(xa + x9, 7);
xc ^= ROTL(xf + xe, 7);
x2 ^= ROTL(x1 + x0, 9);
x7 ^= ROTL(x6 + x5, 9);
x8 ^= ROTL(xb + xa, 9);
xd ^= ROTL(xc + xf, 9);
x3 ^= ROTL(x2 + x1, 13);
x4 ^= ROTL(x7 + x6, 13);
x9 ^= ROTL(x8 + xb, 13);
xe ^= ROTL(xd + xc, 13);
x0 ^= ROTL(x3 + x2, 18);
x5 ^= ROTL(x4 + x7, 18);
xa ^= ROTL(x9 + x8, 18);
xf ^= ROTL(xe + xd, 18);
/* Operate on columns. */
x4 ^= ROTL(x0 + xc, 7);
x9 ^= ROTL(x5 + x1, 7);
xe ^= ROTL(xa + x6, 7);
x3 ^= ROTL(xf + xb, 7);
x8 ^= ROTL(x4 + x0, 9);
xd ^= ROTL(x9 + x5, 9);
x2 ^= ROTL(xe + xa, 9);
x7 ^= ROTL(x3 + xf, 9);
xc ^= ROTL(x8 + x4, 13);
x1 ^= ROTL(xd + x9, 13);
x6 ^= ROTL(x2 + xe, 13);
xb ^= ROTL(x7 + x3, 13);
x0 ^= ROTL(xc + x8, 18);
x5 ^= ROTL(x1 + xd, 18);
xa ^= ROTL(x6 + x2, 18);
xf ^= ROTL(xb + x7, 18);
/* Operate on rows. */
x1 ^= ROTL(x0 + x3, 7);
x6 ^= ROTL(x5 + x4, 7);
xb ^= ROTL(xa + x9, 7);
xc ^= ROTL(xf + xe, 7);
x2 ^= ROTL(x1 + x0, 9);
x7 ^= ROTL(x6 + x5, 9);
x8 ^= ROTL(xb + xa, 9);
xd ^= ROTL(xc + xf, 9);
x3 ^= ROTL(x2 + x1, 13);
x4 ^= ROTL(x7 + x6, 13);
x9 ^= ROTL(x8 + xb, 13);
xe ^= ROTL(xd + xc, 13);
x0 ^= ROTL(x3 + x2, 18);
x5 ^= ROTL(x4 + x7, 18);
xa ^= ROTL(x9 + x8, 18);
xf ^= ROTL(xe + xd, 18);
B[ 0] += x0;
B[ 1] += x1;
B[ 2] += x2;
B[ 3] += x3;
B[ 4] += x4;
B[ 5] += x5;
B[ 6] += x6;
B[ 7] += x7;
B[ 8] += x8;
B[ 9] += x9;
B[10] += xa;
B[11] += xb;
B[12] += xc;
B[13] += xd;
B[14] += xe;
B[15] += xf;
}
/**
* @param X input/ouput
* @param V scratch buffer
* @param N factor (def. 1024)
*/
void scrypt_core_ref(uint32_t *X, uint32_t *V, uint32_t N)
{
for (uint32_t i = 0; i < N; i++) {
memcpy(&V[i * 32], X, 128);
xor_salsa8(&X[0], &X[16]);
xor_salsa8(&X[16], &X[0]);
}
for (uint32_t i = 0; i < N; i++) {
uint32_t j = 32 * (X[16] & (N - 1));
for (uint8_t k = 0; k < 32; k++)
X[k] ^= V[j + k];
xor_salsa8(&X[0], &X[16]);
xor_salsa8(&X[16], &X[0]);
}
}

1476
algo/scrypt/scrypt.c
View File

File diff suppressed because it is too large Load Diff