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

v23.5

Browse Source
This commit is contained in:
Jay D Dee
2023-10-25 20:36:20 -04:00
parent 31c4dedf59
commit 160608cce5
180 changed files with 10318 additions and 13097 deletions
Download Patch File Download Diff File Expand all files Collapse all files

55
algo/simd/nist.c
View File

@@ -7,7 +7,7 @@
/* #define NO_PRECOMPUTED_IV */
#if defined(__SSE2__) // || defined(__ARM_NEON)
/*
* Increase the counter.
@@ -16,7 +16,7 @@ void IncreaseCounter(hashState_sd *state, DataLength databitlen) {
#ifdef HAS_64
state->count += databitlen;
#else
u32 old_count = state->count_low;
uint32_t old_count = state->count_low;
state->count_low += databitlen;
if (state->count_low < old_count)
state->count_high++;
@@ -28,14 +28,9 @@ void IncreaseCounter(hashState_sd *state, DataLength databitlen) {
* Initialize the hashState_sd with a given IV.
* If the IV is NULL, initialize with zeros.
*/
HashReturn InitIV(hashState_sd *state, int hashbitlen, const u32 *IV) {
int InitIV(hashState_sd *state, int hashbitlen, const u32 *IV) {
int n;
if (!SupportedLength(hashbitlen))
return BAD_HASHBITLEN;
n = 8;
int n = 8;
state->hashbitlen = hashbitlen;
state->n_feistels = n;
@@ -71,15 +66,15 @@ HashReturn InitIV(hashState_sd *state, int hashbitlen, const u32 *IV) {
// free(state->buffer);
// free(state->A);
return SUCCESS;
return 0;
}
/*
* Initialize the hashState_sd.
*/
HashReturn init_sd(hashState_sd *state, int hashbitlen) {
HashReturn r;
int init_sd(hashState_sd *state, int hashbitlen) {
int r;
char *init;
#ifndef NO_PRECOMPUTED_IV
@@ -91,7 +86,7 @@ HashReturn init_sd(hashState_sd *state, int hashbitlen) {
// r=InitIV(state, hashbitlen, IV_384);
// else
if (hashbitlen == 512)
r=InitIV(state, hashbitlen, IV_512);
r = InitIV(state, hashbitlen, IV_512);
else
#endif
{
@@ -99,7 +94,7 @@ HashReturn init_sd(hashState_sd *state, int hashbitlen) {
* Nonstandart length: IV is not precomputed.
*/
r=InitIV(state, hashbitlen, NULL);
if (r != SUCCESS)
if (r != 0)
return r;
init = malloc(state->blocksize);
@@ -115,7 +110,7 @@ HashReturn init_sd(hashState_sd *state, int hashbitlen) {
return r;
}
HashReturn update_sd( hashState_sd *state, const BitSequence *data,
int update_sd( hashState_sd *state, const BitSequence *data,
DataLength databitlen )
{
unsigned current;
@@ -135,7 +130,7 @@ HashReturn update_sd( hashState_sd *state, const BitSequence *data,
{
// The number of hashed bits is not a multiple of 8.
// Very painfull to implement and not required by the NIST API.
return FAIL;
return 1;
}
while ( databitlen > 0 )
@@ -156,7 +151,7 @@ HashReturn update_sd( hashState_sd *state, const BitSequence *data,
{
memcpy( state->buffer+current/8, data, (databitlen+7)/8 );
IncreaseCounter( state, databitlen );
return SUCCESS;
return 0;
}
else
{
@@ -169,16 +164,16 @@ HashReturn update_sd( hashState_sd *state, const BitSequence *data,
}
}
}
return SUCCESS;
return 0;
}
HashReturn final_sd( hashState_sd *state, BitSequence *hashval )
int final_sd( hashState_sd *state, BitSequence *hashval )
{
#ifdef HAS_64
u64 l;
uint64_t l;
int current = state->count & (state->blocksize - 1);
#else
u32 l;
uint32_t l;
int current = state->count_low & (state->blocksize - 1);
#endif
unsigned int i;
@@ -248,10 +243,10 @@ HashReturn final_sd( hashState_sd *state, BitSequence *hashval )
BitSequence mask = 0xff << ( 8 - (state->hashbitlen % 8) );
hashval[state->hashbitlen/8 + 1] = bs[state->hashbitlen/8 + 1] & mask;
}
return SUCCESS;
return 0;
}
HashReturn update_final_sd( hashState_sd *state, BitSequence *hashval,
int update_final_sd( hashState_sd *state, BitSequence *hashval,
const BitSequence *data, DataLength databitlen )
{
int current, i;
@@ -259,7 +254,7 @@ HashReturn update_final_sd( hashState_sd *state, BitSequence *hashval,
static int align = -1;
BitSequence out[64];
int isshort = 1;
u64 l;
uint64_t l;
if (align == -1)
align = RequiredAlignment();
@@ -274,7 +269,7 @@ HashReturn update_final_sd( hashState_sd *state, BitSequence *hashval,
{
// The number of hashed bits is not a multiple of 8.
// Very painfull to implement and not required by the NIST API.
return FAIL;
return 1;
}
while ( databitlen > 0 )
@@ -357,7 +352,7 @@ HashReturn update_final_sd( hashState_sd *state, BitSequence *hashval,
BitSequence mask = 0xff << ( 8 - (state->hashbitlen % 8) );
hashval[state->hashbitlen/8 + 1] = out[state->hashbitlen/8 + 1] & mask;
}
return SUCCESS;
return 0;
}
int simd_full( hashState_sd *state, BitSequence *hashval,
@@ -372,7 +367,7 @@ int simd_full( hashState_sd *state, BitSequence *hashval,
static int align = -1;
BitSequence out[64];
int isshort = 1;
u64 l;
uint64_t l;
if (align == -1)
align = RequiredAlignment();
@@ -387,7 +382,7 @@ int simd_full( hashState_sd *state, BitSequence *hashval,
{
// The number of hashed bits is not a multiple of 8.
// Very painfull to implement and not required by the NIST API.
return FAIL;
return 1;
}
while ( databitlen > 0 )
@@ -470,6 +465,8 @@ int simd_full( hashState_sd *state, BitSequence *hashval,
BitSequence mask = 0xff << ( 8 - (state->hashbitlen % 8) );
hashval[state->hashbitlen/8 + 1] = out[state->hashbitlen/8 + 1] & mask;
}
return SUCCESS;
return 0;
}
#endif

24
algo/simd/nist.h
View File

@@ -20,16 +20,16 @@ typedef struct {
unsigned int n_feistels;
#ifdef HAS_64
u64 count;
uint64_t count;
#else
u32 count_low;
u32 count_high;
uint32_t count_low;
uint32_t count_high;
#endif
DATA_ALIGN(u32 A[32]);
u32 *B;
u32 *C;
u32 *D;
DATA_ALIGN(uint32_t A[32]);
uint32_t *B;
uint32_t *C;
uint32_t *D;
DATA_ALIGN(unsigned char buffer[128]);
} hashState_sd;
@@ -38,13 +38,13 @@ typedef struct {
* NIST API
*/
HashReturn init_sd(hashState_sd *state, int hashbitlen);
int init_sd(hashState_sd *state, int hashbitlen);
HashReturn update_sd(hashState_sd *state, const BitSequence *data, DataLength databitlen);
int update_sd(hashState_sd *state, const BitSequence *data, DataLength databitlen);
HashReturn final_sd(hashState_sd *state, BitSequence *hashval);
int final_sd(hashState_sd *state, BitSequence *hashval);
HashReturn update_final_sd( hashState_sd *state, BitSequence *hashval,
int update_final_sd( hashState_sd *state, BitSequence *hashval,
const BitSequence *data, DataLength databitlen );
int simd_full( hashState_sd *state, BitSequence *hashval,
@@ -54,7 +54,7 @@ int simd_full( hashState_sd *state, BitSequence *hashval,
* Internal API
*/
int SupportedLength(int hashbitlen);
//int SupportedLength(int hashbitlen);
int RequiredAlignment(void);
void SIMD_Compress(hashState_sd * state, const unsigned char *M, int final);

3
algo/simd/vector.c
View File

@@ -5,16 +5,19 @@
#include "vector.h"
//#if defined(__SSE2__) || defined(__ARM_NEON)
#if defined(__SSE2__)
#define PRINT_SOME 0
/*
int SupportedLength(int hashbitlen) {
if (hashbitlen <= 0 || hashbitlen > 512)
return 0;
else
return 1;
}
*/
int RequiredAlignment(void) {
return 16;

14
algo/simd/vector.h
View File

@@ -2,6 +2,7 @@
#define __VECTOR_H__
#include "compat.h"
#include "simd-utils.h"
/*******************************
* Using GCC vector extensions *
@@ -133,13 +134,13 @@ union u32 {
#define vec_or(x,y) v128_or( x, y )
#define vec_xor(x,y) v128_xor( x, y )
#define v16_and vec_and
#define v16_or vec_or
#define v16_xor vec_xor
#define v16_and v128_and
#define v16_or v128_or
#define v16_xor v128_xor
#define v32_and vec_and
#define v32_or vec_or
#define v32_xor vec_xor
#define v32_and v128_and
#define v32_or v128_or
#define v32_xor v128_xor
#define vec_andn( x,y ) v128_andnot( x, y )
#define v16_andn vec_andn
@@ -158,7 +159,6 @@ union u32 {
#define v16_interleavel v128_unpacklo16
#define v16_interleaveh v128_unpackhi16
// the builtins compile for arm, so ???
#define v16_mergel(a,b) V1632(__builtin_ia32_punpcklwd128(a,b))
#define v16_mergeh(a,b) V1632(__builtin_ia32_punpckhwd128(a,b))