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

v24.1

Browse Source
This commit is contained in:
Jay D Dee
2024-04-16 21:31:35 -04:00
parent 9d3a46c355
commit 4f930574cc
33 changed files with 757 additions and 609 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
algo/m7m/m7m.c
View File

@@ -21,7 +21,7 @@
#define EPS1 DBL_EPSILON
#define EPS2 3.0e-11
inline double exp_n( double xt )
static inline double exp_n( double xt )
{
if ( xt < -700.0 )
return 0;
@@ -33,7 +33,7 @@ inline double exp_n( double xt )
return exp( xt );
}
inline double exp_n2( double x1, double x2 )
static inline double exp_n2( double x1, double x2 )
{
double p1 = -700., p2 = -37., p3 = -0.8e-8, p4 = 0.8e-8,
p5 = 37., p6 = 700.;

75
algo/m7m/magimath.cpp
View File

@@ -1,75 +0,0 @@
// Copyright (c) 2014 The Magi developers
// Distributed under the MIT/X11 software license, see the accompanying
// file COPYING or http://www.opensource.org/licenses/mit-license.php.
#include <iostream>
#include <cfloat>
#include <limits>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
#include "magimath.h"
#define EPS1 (std::numeric_limits<double>::epsilon())
#define EPS2 3.0e-11
static void gauleg(double x1, double x2, double x[], double w[], const int n)
{
int m,j,i;
double z1, z, xm, xl, pp, p3, p2, p1;
m=(n+1)/2;
xm=0.5*(x2+x1);
xl=0.5*(x2-x1);
for (i=1;i<=m;i++) {
z=cos(3.141592654*(i-0.25)/(n+0.5));
do {
p1=1.0;
p2=0.0;
for (j=1;j<=n;j++) {
p3=p2;
p2=p1;
p1=((2.0*j-1.0)*z*p2-(j-1.0)*p3)/j;
}
pp=n*(z*p1-p2)/(z*z-1.0);
z1=z;
z=z1-p1/pp;
} while (fabs(z-z1) > EPS2);
x[i]=xm-xl*z;
x[n+1-i]=xm+xl*z;
w[i]=2.0*xl/((1.0-z*z)*pp*pp);
w[n+1-i]=w[i];
}
}
static double GaussianQuad_N(double func(const double), const double a2, const double b2, const int NptGQ)
{
double s=0.0;
#ifdef _MSC_VER
#define SW_DIVS 23
double x[SW_DIVS+1], w[SW_DIVS+1];
#else
double x[NptGQ+1], w[NptGQ+1];
#endif
gauleg(a2, b2, x, w, NptGQ);
for (int j=1; j<=NptGQ; j++) {
s += w[j]*func(x[j]);
}
return s;
}
static double swit_(double wvnmb)
{
return pow( (5.55243*(exp_n(-0.3*wvnmb/15.762) - exp_n(-0.6*wvnmb/15.762)))*wvnmb, 0.5)
/ 1034.66 * pow(sin(wvnmb/65.), 2.);
}
uint32_t sw_(int nnounce, int divs)
{
double wmax = ((sqrt((double)(nnounce))*(1.+EPS1))/450+100);
return ((uint32_t)(GaussianQuad_N(swit_, 0., wmax, divs)*(1.+EPS1)*1.e6));
}

54
algo/m7m/magimath.h
View File

@@ -1,54 +0,0 @@
// Copyright (c) 2014 The Magi developers
// Distributed under the MIT/X11 software license, see the accompanying
// file COPYING or http://www.opensource.org/licenses/mit-license.php.
#ifndef MAGI_MATH_H
#define MAGI_MATH_H
#include <math.h>
#ifdef __cplusplus
extern "C" {
#endif
uint32_t sw_(int nnounce, int divs);
#ifdef __cplusplus
}
#endif
inline double exp_n(double xt)
{
double p1 = -700.0, p3 = -0.8e-8, p4 = 0.8e-8, p6 = 700.0;
if(xt < p1)
return 0;
else if(xt > p6)
return 1e200;
else if(xt > p3 && xt < p4)
return (1.0 + xt);
else
return exp(xt);
}
// 1 / (1 + exp(x1-x2))
inline double exp_n2(double x1, double x2)
{
double p1 = -700., p2 = -37., p3 = -0.8e-8, p4 = 0.8e-8, p5 = 37., p6 = 700.;
double xt = x1 - x2;
if (xt < p1+1.e-200)
return 1.;
else if (xt > p1 && xt < p2 + 1.e-200)
return ( 1. - exp(xt) );
else if (xt > p2 && xt < p3 + 1.e-200)
return ( 1. / (1. + exp(xt)) );
else if (xt > p3 && xt < p4)
return ( 1. / (2. + xt) );
else if (xt > p4 - 1.e-200 && xt < p5)
return ( exp(-xt) / (1. + exp(-xt)) );
else if (xt > p5 - 1.e-200 && xt < p6)
return ( exp(-xt) );
else //if (xt > p6 - 1.e-200)
return 0.;
}
#endif

2
algo/sha/sha1-hash.c
View File

@@ -205,7 +205,7 @@ void sha1_x86_sha_transform_be( uint32_t *state_out, const void *input,
#endif
#if defined(__aarch64__) && defined(__ARM_FEATURE_SHA2)
#if defined(__ARM_NEON) && defined(__ARM_FEATURE_SHA2)
#define sha1_neon_rounds( state_out, data, state_in ) \
{ \

14
algo/sha/sha256-hash.c
View File

@@ -1,6 +1,6 @@
#include "sha256-hash.h"
#if ( defined(__x86_64__) && defined(__SHA__) ) || defined(__ARM_NEON) && defined(__ARM_FEATURE_SHA2)
#if ( defined(__x86_64__) && defined(__SHA__) ) || ( defined(__ARM_NEON) && defined(__ARM_FEATURE_SHA2) )
static const uint32_t SHA256_IV[8] =
{
@@ -189,7 +189,7 @@ static const uint32_t SHA256_IV[8] =
_mm_store_si128( (__m128i*) &state_out[4], STATE1 ); \
}
void sha256_opt_transform_le( uint32_t *state_out, const void *input,
void sha256_x86_sha_transform_le( uint32_t *state_out, const void *input,
const uint32_t *state_in )
{
#define load_msg( m, i ) casti_v128( m, i )
@@ -197,7 +197,7 @@ void sha256_opt_transform_le( uint32_t *state_out, const void *input,
#undef load_msg
}
void sha256_opt_transform_be( uint32_t *state_out, const void *input,
void sha256_x86_sha_transform_be( uint32_t *state_out, const void *input,
const uint32_t *state_in )
{
#define load_msg( m, i ) v128_bswap32( casti_v128( m, i ) )
@@ -517,7 +517,7 @@ void sha256_opt_transform_be( uint32_t *state_out, const void *input,
_mm_store_si128( (__m128i*) &out_Y[4], STATE1_Y ); \
}
void sha256_ni2x_transform_le( uint32_t *out_X, uint32_t*out_Y,
void sha256_x86_x2sha_transform_le( uint32_t *out_X, uint32_t*out_Y,
const void *msg_X, const void *msg_Y,
const uint32_t *in_X, const uint32_t *in_Y )
{
@@ -526,7 +526,7 @@ void sha256_ni2x_transform_le( uint32_t *out_X, uint32_t*out_Y,
#undef load_msg
}
void sha256_ni2x_transform_be( uint32_t *out_X, uint32_t*out_Y,
void sha256_x86_x2sha_transform_be( uint32_t *out_X, uint32_t*out_Y,
const void *msg_X, const void *msg_Y,
const uint32_t *in_X, const uint32_t *in_Y )
{
@@ -541,7 +541,7 @@ void sha256_ni2x_transform_be( uint32_t *out_X, uint32_t*out_Y,
// The goal is to avoid any redundant processing in final. Prehash is almost
// 4 rounds total, only missing the final addition of the nonce.
// Nonce must be set to zero for prehash.
void sha256_ni_prehash_3rounds( uint32_t *ostate, const void *msg,
void sha256_x86_sha_prehash_3rounds( uint32_t *ostate, const void *msg,
uint32_t *sstate, const uint32_t *istate )
{
__m128i STATE0, STATE1, MSG, TMP;
@@ -569,7 +569,7 @@ void sha256_ni_prehash_3rounds( uint32_t *ostate, const void *msg,
casti_m128i( ostate, 1 ) = STATE1;
}
void sha256_ni2x_final_rounds( uint32_t *out_X, uint32_t *out_Y,
void sha256_x86_x2sha_final_rounds( uint32_t *out_X, uint32_t *out_Y,
const void *msg_X, const void *msg_Y,
const uint32_t *state_mid_X, const uint32_t *state_mid_Y,
const uint32_t *state_save_X, const uint32_t *state_save_Y )

55
algo/sha/sha256-hash.h
View File

@@ -5,27 +5,21 @@
#include "simd-utils.h"
#include "cpuminer-config.h"
// generic interface
static const uint32_t SHA256_IV[8];
#if defined(__x86_64__) && defined(__SHA__)
typedef struct
{
unsigned char buf[64]; /* first field, for alignment */
unsigned char buf[64];
uint32_t state[8];
uint64_t count;
} sha256_context __attribute__((aligned(64)));
static const uint32_t SHA256_IV[8];
void sha256_full( void *hash, const void *data, size_t len );
void sha256_update( sha256_context *ctx, const void *data, size_t len );
void sha256_final( sha256_context *ctx, void *hash );
void sha256_ctx_init( sha256_context *ctx );
void sha256_transform_le( uint32_t *state_out, const uint32_t *data,
const uint32_t *state_in );
void sha256_transform_be( uint32_t *state_out, const uint32_t *data,
const uint32_t *state_in );
#if defined(__x86_64__) && defined(__SHA__)
void sha256_x86_sha_transform_le( uint32_t *state_out, const void *input,
const uint32_t *state_in );
@@ -50,14 +44,6 @@ void sha256_x86_x2sha_final_rounds( uint32_t *state_out_X, uint32_t *state_out_Y
const uint32_t *state_mid_X, const uint32_t *state_mid_Y,
const uint32_t *state_save_X, const uint32_t *state_save_Y );
// Temporary during name transition
#define sha256_opt_transform_le sha256_x86_sha_transform_le
#define sha256_opt_transform_be sha256_x86_sha_transform_be
#define sha256_ni2x_transform_le sha256_x86_x2sha_transform_le
#define sha256_ni2x_transform_be sha256_x86_x2sha_transform_be
#define sha256_ni_prehash_3rounds sha256_x86_sha_prehash_3rounds
#define sha256_ni2x_final_rounds sha256_x86_x2sha_final_rounds
// generic API
#define sha256_transform_le sha256_x86_sha_transform_le
#define sha256_transform_be sha256_x86_sha_transform_be
@@ -68,6 +54,20 @@ void sha256_x86_x2sha_final_rounds( uint32_t *state_out_X, uint32_t *state_out_Y
#elif defined(__ARM_NEON) && defined(__ARM_FEATURE_SHA2)
// SHA-256 AArch64 with NEON & SHA2
typedef struct
{
unsigned char buf[64];
uint32_t state[8];
uint64_t count;
} sha256_context __attribute__((aligned(64)));
void sha256_full( void *hash, const void *data, size_t len );
void sha256_update( sha256_context *ctx, const void *data, size_t len );
void sha256_final( sha256_context *ctx, void *hash );
void sha256_ctx_init( sha256_context *ctx );
void sha256_neon_sha_transform_be( uint32_t *state_out, const void *input,
const uint32_t *state_in );
void sha256_neon_sha_transform_le( uint32_t *state_out, const void *input,
@@ -89,14 +89,6 @@ void sha256_neon_x2sha_final_rounds( uint32_t *state_out_X,
const uint32_t *state_mid_X, const uint32_t *state_mid_Y,
const uint32_t *state_save_X, const uint32_t *state_save_Y );
// Temporary during name transition
#define sha256_transform_le sha256_neon_sha_transform_le
#define sha256_transform_be sha256_neon_sha_transform_be
#define sha256_2x_transform_le sha256_neon_x2sha_transform_le
#define sha256_2x_transform_be sha256_neon_x2sha_transform_be
#define sha256_prehash_3rounds sha256_neon_sha_prehash_3rounds
#define sha256_2x_final_rounds sha256_neon_x2sha_final_rounds
// generic API
#define sha256_transform_le sha256_neon_sha_transform_le
#define sha256_transform_be sha256_neon_sha_transform_be
@@ -106,9 +98,11 @@ void sha256_neon_x2sha_final_rounds( uint32_t *state_out_X,
#define sha256_2x_final_rounds sha256_neon_x2sha_final_rounds
#else
// without HW acceleration...
#include "sph_sha2.h"
#define sha256_context sph_sha256_context
#define sha256_full sph_sha256_full
#define sha256_ctx_init sph_sha256_init
#define sha256_update sph_sha256
@@ -117,12 +111,11 @@ void sha256_neon_x2sha_final_rounds( uint32_t *state_out_X,
#define sha256_transform_be sph_sha256_transform_be
#define sha256_prehash_3rounds sph_sha256_prehash_3rounds
#endif
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
// SHA-256 16 way
// SHA-256 16 way x86_64
typedef struct
{
@@ -147,7 +140,7 @@ void sha256_16x32_final_rounds( __m512i *state_out, const __m512i *data,
int sha256_16x32_transform_le_short( __m512i *state_out, const __m512i *data,
const __m512i *state_in, const uint32_t *target );
#define sha256_16way_context sha256_16x32_context
#define sha256_16way_context sha256_16x32_context
#define sha256_16way_init sha256_16x32_init
#define sha256_16way_update sha256_16x32_update
#define sha256_16way_close sha256_16x32_close
@@ -162,7 +155,7 @@ int sha256_16x32_transform_le_short( __m512i *state_out, const __m512i *data,
#if defined (__AVX2__)
// SHA-256 8 way
// SHA-256 8 way x86_64
typedef struct
{
@@ -201,7 +194,7 @@ int sha256_8x32_transform_le_short( __m256i *state_out, const __m256i *data,
#endif // AVX2
// SHA-256 4 way
// SHA-256 4 way x86_64 with SSE2 or AArch64 with NEON
typedef struct
{

4
algo/x11/timetravel-gate.c
View File

@@ -16,14 +16,14 @@ bool register_timetravel_algo( algo_gate_t* gate )
return true;
};
inline void tt_swap( int *a, int *b )
static inline void tt_swap( int *a, int *b )
{
int c = *a;
*a = *b;
*b = c;
}
inline void reverse( int *pbegin, int *pend )
static inline void reverse( int *pbegin, int *pend )
{
while ( (pbegin != pend) && (pbegin != --pend) )
{

4
algo/x11/timetravel10-gate.c
View File

@@ -16,14 +16,14 @@ bool register_timetravel10_algo( algo_gate_t* gate )
return true;
};
inline void tt10_swap( int *a, int *b )
static inline void tt10_swap( int *a, int *b )
{
int c = *a;
*a = *b;
*b = c;
}
inline void reverse( int *pbegin, int *pend )
static inline void reverse( int *pbegin, int *pend )
{
while ( (pbegin != pend) && (pbegin != --pend) )
{

2
algo/x16/x16rv2-4way.c
View File

@@ -730,7 +730,7 @@ typedef union _x16rv2_4way_context_overlay x16rv2_4way_context_overlay;
static __thread x16rv2_4way_context_overlay x16rv2_ctx;
// Pad the 24 bytes tiger hash to 64 bytes
inline void padtiger512( uint32_t* hash )
static inline void padtiger512( uint32_t* hash )
{
for ( int i = 6; i < 16; i++ ) hash[i] = 0;
}

2
algo/x16/x16rv2.c
View File

@@ -33,7 +33,7 @@ union _x16rv2_context_overlay
typedef union _x16rv2_context_overlay x16rv2_context_overlay;
// Pad the 24 bytes tiger hash to 64 bytes
inline void padtiger512(uint32_t* hash) {
static inline void padtiger512(uint32_t* hash) {
for (int i = (24/4); i < (64/4); i++) hash[i] = 0;
}