mirror of
https://github.com/JayDDee/cpuminer-opt.git
synced 2025-09-17 23:44:27 +00:00
Initial upload v3.4.7
This commit is contained in:
Jay D Dee
143
algo/argon2/ar2/blake2/blake2-impl.h
Normal file
143
algo/argon2/ar2/blake2/blake2-impl.h
Normal file
@@ -0,0 +1,143 @@
|
||||
#ifndef PORTABLE_BLAKE2_IMPL_H
|
||||
#define PORTABLE_BLAKE2_IMPL_H
|
||||
|
||||
#include <stdint.h>
|
||||
#include <string.h>
|
||||
|
||||
#if defined(_MSC_VER)
|
||||
#define BLAKE2_INLINE __inline
|
||||
#elif defined(__GNUC__) || defined(__clang__)
|
||||
#define BLAKE2_INLINE __inline__
|
||||
#else
|
||||
#define BLAKE2_INLINE
|
||||
#endif
|
||||
|
||||
/* Argon2 Team - Begin Code */
|
||||
/*
|
||||
Not an exhaustive list, but should cover the majority of modern platforms
|
||||
Additionally, the code will always be correct---this is only a performance
|
||||
tweak.
|
||||
*/
|
||||
#if (defined(__BYTE_ORDER__) && \
|
||||
(__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)) || \
|
||||
defined(__LITTLE_ENDIAN__) || defined(__ARMEL__) || defined(__MIPSEL__) || \
|
||||
defined(__AARCH64EL__) || defined(__amd64__) || defined(__i386__) || \
|
||||
defined(_M_IX86) || defined(_M_X64) || defined(_M_AMD64) || \
|
||||
defined(_M_ARM)
|
||||
#define NATIVE_LITTLE_ENDIAN
|
||||
#endif
|
||||
/* Argon2 Team - End Code */
|
||||
|
||||
static BLAKE2_INLINE uint32_t load32(const void *src) {
|
||||
#if defined(NATIVE_LITTLE_ENDIAN)
|
||||
uint32_t w;
|
||||
memcpy(&w, src, sizeof w);
|
||||
return w;
|
||||
#else
|
||||
const uint8_t *p = (const uint8_t *)src;
|
||||
uint32_t w = *p++;
|
||||
w |= (uint32_t)(*p++) << 8;
|
||||
w |= (uint32_t)(*p++) << 16;
|
||||
w |= (uint32_t)(*p++) << 24;
|
||||
return w;
|
||||
#endif
|
||||
}
|
||||
|
||||
static BLAKE2_INLINE uint64_t load64(const void *src) {
|
||||
#if defined(NATIVE_LITTLE_ENDIAN)
|
||||
uint64_t w;
|
||||
memcpy(&w, src, sizeof w);
|
||||
return w;
|
||||
#else
|
||||
const uint8_t *p = (const uint8_t *)src;
|
||||
uint64_t w = *p++;
|
||||
w |= (uint64_t)(*p++) << 8;
|
||||
w |= (uint64_t)(*p++) << 16;
|
||||
w |= (uint64_t)(*p++) << 24;
|
||||
w |= (uint64_t)(*p++) << 32;
|
||||
w |= (uint64_t)(*p++) << 40;
|
||||
w |= (uint64_t)(*p++) << 48;
|
||||
w |= (uint64_t)(*p++) << 56;
|
||||
return w;
|
||||
#endif
|
||||
}
|
||||
|
||||
static BLAKE2_INLINE void store32(void *dst, uint32_t w) {
|
||||
#if defined(NATIVE_LITTLE_ENDIAN)
|
||||
memcpy(dst, &w, sizeof w);
|
||||
#else
|
||||
uint8_t *p = (uint8_t *)dst;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
#endif
|
||||
}
|
||||
|
||||
static BLAKE2_INLINE void store64(void *dst, uint64_t w) {
|
||||
#if defined(NATIVE_LITTLE_ENDIAN)
|
||||
memcpy(dst, &w, sizeof w);
|
||||
#else
|
||||
uint8_t *p = (uint8_t *)dst;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
#endif
|
||||
}
|
||||
|
||||
static BLAKE2_INLINE uint64_t load48(const void *src) {
|
||||
const uint8_t *p = (const uint8_t *)src;
|
||||
uint64_t w = *p++;
|
||||
w |= (uint64_t)(*p++) << 8;
|
||||
w |= (uint64_t)(*p++) << 16;
|
||||
w |= (uint64_t)(*p++) << 24;
|
||||
w |= (uint64_t)(*p++) << 32;
|
||||
w |= (uint64_t)(*p++) << 40;
|
||||
return w;
|
||||
}
|
||||
|
||||
static BLAKE2_INLINE void store48(void *dst, uint64_t w) {
|
||||
uint8_t *p = (uint8_t *)dst;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
w >>= 8;
|
||||
*p++ = (uint8_t)w;
|
||||
}
|
||||
|
||||
static BLAKE2_INLINE uint32_t rotr32(const uint32_t w, const unsigned c) {
|
||||
return (w >> c) | (w << (32 - c));
|
||||
}
|
||||
|
||||
static BLAKE2_INLINE uint64_t rotr64(const uint64_t w, const unsigned c) {
|
||||
return (w >> c) | (w << (64 - c));
|
||||
}
|
||||
|
||||
/* prevents compiler optimizing out memset() */
|
||||
static BLAKE2_INLINE void burn(void *v, size_t n) {
|
||||
static void *(*const volatile memset_v)(void *, int, size_t) = &memset;
|
||||
memset_v(v, 0, n);
|
||||
}
|
||||
|
||||
#endif
|
||||
76
algo/argon2/ar2/blake2/blake2.h
Normal file
76
algo/argon2/ar2/blake2/blake2.h
Normal file
@@ -0,0 +1,76 @@
|
||||
#ifndef PORTABLE_BLAKE2_H
|
||||
#define PORTABLE_BLAKE2_H
|
||||
|
||||
#include <stddef.h>
|
||||
#include <stdint.h>
|
||||
#include <limits.h>
|
||||
|
||||
#if defined(__cplusplus)
|
||||
extern "C" {
|
||||
#endif
|
||||
|
||||
enum blake2b_constant {
|
||||
BLAKE2B_BLOCKBYTES = 128,
|
||||
BLAKE2B_OUTBYTES = 64,
|
||||
BLAKE2B_KEYBYTES = 64,
|
||||
BLAKE2B_SALTBYTES = 16,
|
||||
BLAKE2B_PERSONALBYTES = 16
|
||||
};
|
||||
|
||||
#pragma pack(push, 1)
|
||||
typedef struct __blake2b_param {
|
||||
uint8_t digest_length; /* 1 */
|
||||
uint8_t key_length; /* 2 */
|
||||
uint8_t fanout; /* 3 */
|
||||
uint8_t depth; /* 4 */
|
||||
uint32_t leaf_length; /* 8 */
|
||||
uint64_t node_offset; /* 16 */
|
||||
uint8_t node_depth; /* 17 */
|
||||
uint8_t inner_length; /* 18 */
|
||||
uint8_t reserved[14]; /* 32 */
|
||||
uint8_t salt[BLAKE2B_SALTBYTES]; /* 48 */
|
||||
uint8_t personal[BLAKE2B_PERSONALBYTES]; /* 64 */
|
||||
} blake2b_param;
|
||||
#pragma pack(pop)
|
||||
|
||||
typedef struct __blake2b_state {
|
||||
uint64_t h[8];
|
||||
uint64_t t[2];
|
||||
uint64_t f[2];
|
||||
unsigned buflen;
|
||||
unsigned outlen;
|
||||
uint8_t last_node;
|
||||
uint8_t buf[BLAKE2B_BLOCKBYTES];
|
||||
} blake2b_state;
|
||||
|
||||
/* Ensure param structs have not been wrongly padded */
|
||||
/* Poor man's static_assert */
|
||||
enum {
|
||||
blake2_size_check_0 = 1 / !!(CHAR_BIT == 8),
|
||||
blake2_size_check_2 =
|
||||
1 / !!(sizeof(blake2b_param) == sizeof(uint64_t) * CHAR_BIT)
|
||||
};
|
||||
|
||||
/* Streaming API */
|
||||
int blake2b_init(blake2b_state *S, size_t outlen);
|
||||
int blake2b_init_key(blake2b_state *S, size_t outlen, const void *key,
|
||||
size_t keylen);
|
||||
int blake2b_init_param(blake2b_state *S, const blake2b_param *P);
|
||||
int blake2b_update(blake2b_state *S, const void *in, size_t inlen);
|
||||
void my_blake2b_update(blake2b_state *S, const void *in, size_t inlen);
|
||||
int blake2b_final(blake2b_state *S, void *out, size_t outlen);
|
||||
|
||||
/* Simple API */
|
||||
int blake2b(void *out, const void *in, const void *key, size_t keylen);
|
||||
|
||||
/* Argon2 Team - Begin Code */
|
||||
int blake2b_long(void *out, const void *in);
|
||||
/* Argon2 Team - End Code */
|
||||
/* Miouyouyou */
|
||||
void blake2b_too(void *out, const void *in);
|
||||
|
||||
#if defined(__cplusplus)
|
||||
}
|
||||
#endif
|
||||
|
||||
#endif
|
||||
162
algo/argon2/ar2/blake2/blamka-round-opt.h
Normal file
162
algo/argon2/ar2/blake2/blamka-round-opt.h
Normal file
@@ -0,0 +1,162 @@
|
||||
#ifndef BLAKE_ROUND_MKA_OPT_H
|
||||
#define BLAKE_ROUND_MKA_OPT_H
|
||||
|
||||
#include "blake2-impl.h"
|
||||
|
||||
#if defined(_MSC_VER)
|
||||
#include <intrin.h>
|
||||
#endif
|
||||
|
||||
#include <immintrin.h>
|
||||
#if defined(__XOP__) && (defined(__GNUC__) || defined(__clang__))
|
||||
#include <x86intrin.h>
|
||||
#endif
|
||||
|
||||
#if !defined(__XOP__)
|
||||
#if defined(__SSSE3__)
|
||||
#define r16 \
|
||||
(_mm_setr_epi8(2, 3, 4, 5, 6, 7, 0, 1, 10, 11, 12, 13, 14, 15, 8, 9))
|
||||
#define r24 \
|
||||
(_mm_setr_epi8(3, 4, 5, 6, 7, 0, 1, 2, 11, 12, 13, 14, 15, 8, 9, 10))
|
||||
#define _mm_roti_epi64(x, c) \
|
||||
(-(c) == 32) \
|
||||
? _mm_shuffle_epi32((x), _MM_SHUFFLE(2, 3, 0, 1)) \
|
||||
: (-(c) == 24) \
|
||||
? _mm_shuffle_epi8((x), r24) \
|
||||
: (-(c) == 16) \
|
||||
? _mm_shuffle_epi8((x), r16) \
|
||||
: (-(c) == 63) \
|
||||
? _mm_xor_si128(_mm_srli_epi64((x), -(c)), \
|
||||
_mm_add_epi64((x), (x))) \
|
||||
: _mm_xor_si128(_mm_srli_epi64((x), -(c)), \
|
||||
_mm_slli_epi64((x), 64 - (-(c))))
|
||||
#else /* defined(__SSE2__) */
|
||||
#define _mm_roti_epi64(r, c) \
|
||||
_mm_xor_si128(_mm_srli_epi64((r), -(c)), _mm_slli_epi64((r), 64 - (-(c))))
|
||||
#endif
|
||||
#else
|
||||
#endif
|
||||
|
||||
static BLAKE2_INLINE __m128i fBlaMka(__m128i x, __m128i y) {
|
||||
const __m128i z = _mm_mul_epu32(x, y);
|
||||
return _mm_add_epi64(_mm_add_epi64(x, y), _mm_add_epi64(z, z));
|
||||
}
|
||||
|
||||
#define G1(A0, B0, C0, D0, A1, B1, C1, D1) \
|
||||
do { \
|
||||
A0 = fBlaMka(A0, B0); \
|
||||
A1 = fBlaMka(A1, B1); \
|
||||
\
|
||||
D0 = _mm_xor_si128(D0, A0); \
|
||||
D1 = _mm_xor_si128(D1, A1); \
|
||||
\
|
||||
D0 = _mm_roti_epi64(D0, -32); \
|
||||
D1 = _mm_roti_epi64(D1, -32); \
|
||||
\
|
||||
C0 = fBlaMka(C0, D0); \
|
||||
C1 = fBlaMka(C1, D1); \
|
||||
\
|
||||
B0 = _mm_xor_si128(B0, C0); \
|
||||
B1 = _mm_xor_si128(B1, C1); \
|
||||
\
|
||||
B0 = _mm_roti_epi64(B0, -24); \
|
||||
B1 = _mm_roti_epi64(B1, -24); \
|
||||
} while ((void)0, 0)
|
||||
|
||||
#define G2(A0, B0, C0, D0, A1, B1, C1, D1) \
|
||||
do { \
|
||||
A0 = fBlaMka(A0, B0); \
|
||||
A1 = fBlaMka(A1, B1); \
|
||||
\
|
||||
D0 = _mm_xor_si128(D0, A0); \
|
||||
D1 = _mm_xor_si128(D1, A1); \
|
||||
\
|
||||
D0 = _mm_roti_epi64(D0, -16); \
|
||||
D1 = _mm_roti_epi64(D1, -16); \
|
||||
\
|
||||
C0 = fBlaMka(C0, D0); \
|
||||
C1 = fBlaMka(C1, D1); \
|
||||
\
|
||||
B0 = _mm_xor_si128(B0, C0); \
|
||||
B1 = _mm_xor_si128(B1, C1); \
|
||||
\
|
||||
B0 = _mm_roti_epi64(B0, -63); \
|
||||
B1 = _mm_roti_epi64(B1, -63); \
|
||||
} while ((void)0, 0)
|
||||
|
||||
#if defined(__SSSE3__)
|
||||
#define DIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1) \
|
||||
do { \
|
||||
__m128i t0 = _mm_alignr_epi8(B1, B0, 8); \
|
||||
__m128i t1 = _mm_alignr_epi8(B0, B1, 8); \
|
||||
B0 = t0; \
|
||||
B1 = t1; \
|
||||
\
|
||||
t0 = C0; \
|
||||
C0 = C1; \
|
||||
C1 = t0; \
|
||||
\
|
||||
t0 = _mm_alignr_epi8(D1, D0, 8); \
|
||||
t1 = _mm_alignr_epi8(D0, D1, 8); \
|
||||
D0 = t1; \
|
||||
D1 = t0; \
|
||||
} while ((void)0, 0)
|
||||
|
||||
#define UNDIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1) \
|
||||
do { \
|
||||
__m128i t0 = _mm_alignr_epi8(B0, B1, 8); \
|
||||
__m128i t1 = _mm_alignr_epi8(B1, B0, 8); \
|
||||
B0 = t0; \
|
||||
B1 = t1; \
|
||||
\
|
||||
t0 = C0; \
|
||||
C0 = C1; \
|
||||
C1 = t0; \
|
||||
\
|
||||
t0 = _mm_alignr_epi8(D0, D1, 8); \
|
||||
t1 = _mm_alignr_epi8(D1, D0, 8); \
|
||||
D0 = t1; \
|
||||
D1 = t0; \
|
||||
} while ((void)0, 0)
|
||||
#else /* SSE2 */
|
||||
#define DIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1) \
|
||||
do { \
|
||||
__m128i t0 = D0; \
|
||||
__m128i t1 = B0; \
|
||||
D0 = C0; \
|
||||
C0 = C1; \
|
||||
C1 = D0; \
|
||||
D0 = _mm_unpackhi_epi64(D1, _mm_unpacklo_epi64(t0, t0)); \
|
||||
D1 = _mm_unpackhi_epi64(t0, _mm_unpacklo_epi64(D1, D1)); \
|
||||
B0 = _mm_unpackhi_epi64(B0, _mm_unpacklo_epi64(B1, B1)); \
|
||||
B1 = _mm_unpackhi_epi64(B1, _mm_unpacklo_epi64(t1, t1)); \
|
||||
} while ((void)0, 0)
|
||||
|
||||
#define UNDIAGONALIZE(A0, B0, C0, D0, A1, B1, C1, D1) \
|
||||
do { \
|
||||
__m128i t0 = C0; \
|
||||
C0 = C1; \
|
||||
C1 = t0; \
|
||||
t0 = B0; \
|
||||
__m128i t1 = D0; \
|
||||
B0 = _mm_unpackhi_epi64(B1, _mm_unpacklo_epi64(B0, B0)); \
|
||||
B1 = _mm_unpackhi_epi64(t0, _mm_unpacklo_epi64(B1, B1)); \
|
||||
D0 = _mm_unpackhi_epi64(D0, _mm_unpacklo_epi64(D1, D1)); \
|
||||
D1 = _mm_unpackhi_epi64(D1, _mm_unpacklo_epi64(t1, t1)); \
|
||||
} while ((void)0, 0)
|
||||
#endif
|
||||
|
||||
#define BLAKE2_ROUND(A0, A1, B0, B1, C0, C1, D0, 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)
|
||||
|
||||
#endif
|
||||
39
algo/argon2/ar2/blake2/blamka-round-ref.h
Normal file
39
algo/argon2/ar2/blake2/blamka-round-ref.h
Normal file
@@ -0,0 +1,39 @@
|
||||
#ifndef BLAKE_ROUND_MKA_H
|
||||
#define BLAKE_ROUND_MKA_H
|
||||
|
||||
#include "blake2.h"
|
||||
#include "blake2-impl.h"
|
||||
|
||||
/*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);
|
||||
return x + y + 2 * xy;
|
||||
}
|
||||
|
||||
#define G(a, b, c, d) \
|
||||
do { \
|
||||
a = fBlaMka(a, b); \
|
||||
d = rotr64(d ^ a, 32); \
|
||||
c = fBlaMka(c, d); \
|
||||
b = rotr64(b ^ c, 24); \
|
||||
a = fBlaMka(a, b); \
|
||||
d = rotr64(d ^ a, 16); \
|
||||
c = fBlaMka(c, d); \
|
||||
b = rotr64(b ^ c, 63); \
|
||||
} while ((void)0, 0)
|
||||
|
||||
#define BLAKE2_ROUND_NOMSG(v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, \
|
||||
v12, v13, v14, v15) \
|
||||
do { \
|
||||
G(v0, v4, v8, v12); \
|
||||
G(v1, v5, v9, v13); \
|
||||
G(v2, v6, v10, v14); \
|
||||
G(v3, v7, v11, v15); \
|
||||
G(v0, v5, v10, v15); \
|
||||
G(v1, v6, v11, v12); \
|
||||
G(v2, v7, v8, v13); \
|
||||
G(v3, v4, v9, v14); \
|
||||
} while ((void)0, 0)
|
||||
|
||||
#endif
|
||||
Reference in New Issue
Block a user