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

v25.3

Browse Source
This commit is contained in:
Jay D Dee
2025-01-16 12:31:53 -05:00
parent 1ed18bf22e
commit dd99580a4c
14 changed files with 1961 additions and 2031 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
Makefile.am
View File

@@ -301,7 +301,7 @@ if USE_ASM
disable_flags =
cpuminer_SOURCES += asm/neoscrypt_asm.S
else
disable_flags += -DNOASM
disable_flags = -DNOASM
endif
cpuminer_LDFLAGS = @LDFLAGS@

6
README.md
View File

@@ -55,9 +55,9 @@ Supported Algorithms
allium Garlicoin
anime Animecoin
argon2 Argon2 coin (AR2)
argon2d250 argon2d-crds, Credits (CRDS)
argon2d500 argon2d-dyn, Dynamic (DYN)
argon2d4096 argon2d-uis, Unitus, (UIS)
argon2d250
argon2d500
argon2d4096
blake Blake-256
blake2b Blake2-512
blake2s Blake2-256

7
RELEASE_NOTES
View File

@@ -75,6 +75,13 @@ If not what makes it happen or not happen?
Change Log
----------
v25.3
#442, #443: Fixed a regression in Makefile.am.
Updated dockerfile.
Removed algo features log display.
Some code cleanup.
v25.2
ARM: Fixed regression from v25.1 that could cause build fail.

6
algo-gate-api.c
View File

@@ -295,8 +295,8 @@ bool register_algo_gate( int algo, algo_gate_t *gate )
{
case ALGO_ALLIUM: rc = register_allium_algo ( gate ); break;
case ALGO_ANIME: rc = register_anime_algo ( gate ); break;
case ALGO_ARGON2D250: rc = register_argon2d_crds_algo ( gate ); break;
case ALGO_ARGON2D500: rc = register_argon2d_dyn_algo ( gate ); break;
case ALGO_ARGON2D250: rc = register_argon2d250_algo ( gate ); break;
case ALGO_ARGON2D500: rc = register_argon2d500_algo ( gate ); break;
case ALGO_ARGON2D4096: rc = register_argon2d4096_algo ( gate ); break;
case ALGO_AXIOM: rc = register_axiom_algo ( gate ); break;
case ALGO_BLAKE: rc = register_blake_algo ( gate ); break;
@@ -416,8 +416,6 @@ void exec_hash_function( int algo, void *output, const void *pdata )
const char* const algo_alias_map[][2] =
{
// alias proper
{ "argon2d-dyn", "argon2d500" },
{ "argon2d-uis", "argon2d4096" },
{ "bcd", "x13bcd" },
{ "bitcore", "timetravel10" },
{ "bitzeny", "yescryptr8" },

30
algo/argon2d/argon2d-gate.c
View File

@@ -6,9 +6,7 @@ static const size_t INPUT_BYTES = 80; // Lenth of a block header in bytes. Inpu
static const size_t OUTPUT_BYTES = 32; // Length of output needed for a 256-bit hash
static const unsigned int DEFAULT_ARGON2_FLAG = 2; //Same as ARGON2_DEFAULT_FLAGS
// Credits
void argon2d_crds_hash( void *output, const void *input )
void argon2d250_hash( void *output, const void *input )
{
argon2_context context;
context.out = (uint8_t *)output;
@@ -34,7 +32,7 @@ void argon2d_crds_hash( void *output, const void *input )
argon2_ctx( &context, Argon2_d );
}
int scanhash_argon2d_crds( struct work *work, uint32_t max_nonce,
int scanhash_argon2d250( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t _ALIGN(64) edata[20];
@@ -50,7 +48,7 @@ int scanhash_argon2d_crds( struct work *work, uint32_t max_nonce,
do {
be32enc(&edata[19], nonce);
argon2d_crds_hash( hash, edata );
argon2d250_hash( hash, edata );
if ( hash[7] <= Htarg && fulltest( hash, ptarget ) && !opt_benchmark )
{
pdata[19] = nonce;
@@ -64,18 +62,16 @@ int scanhash_argon2d_crds( struct work *work, uint32_t max_nonce,
return 0;
}
bool register_argon2d_crds_algo( algo_gate_t* gate )
bool register_argon2d250_algo( algo_gate_t* gate )
{
gate->scanhash = (void*)&scanhash_argon2d_crds;
gate->hash = (void*)&argon2d_crds_hash;
gate->scanhash = (void*)&scanhash_argon2d250;
gate->hash = (void*)&argon2d250_hash;
gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT | NEON_OPT;
opt_target_factor = 65536.0;
return true;
}
// Dynamic
void argon2d_dyn_hash( void *output, const void *input )
void argon2d500_hash( void *output, const void *input )
{
argon2_context context;
context.out = (uint8_t *)output;
@@ -101,7 +97,7 @@ void argon2d_dyn_hash( void *output, const void *input )
argon2_ctx( &context, Argon2_d );
}
int scanhash_argon2d_dyn( struct work *work, uint32_t max_nonce,
int scanhash_argon2d500( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t _ALIGN(64) edata[20];
@@ -118,7 +114,7 @@ int scanhash_argon2d_dyn( struct work *work, uint32_t max_nonce,
do
{
edata[19] = nonce;
argon2d_dyn_hash( hash, edata );
argon2d500_hash( hash, edata );
if ( unlikely( valid_hash( (uint64_t*)hash, (uint64_t*)ptarget )
&& !bench ) )
{
@@ -133,17 +129,15 @@ int scanhash_argon2d_dyn( struct work *work, uint32_t max_nonce,
return 0;
}
bool register_argon2d_dyn_algo( algo_gate_t* gate )
bool register_argon2d500_algo( algo_gate_t* gate )
{
gate->scanhash = (void*)&scanhash_argon2d_dyn;
gate->hash = (void*)&argon2d_dyn_hash;
gate->scanhash = (void*)&scanhash_argon2d500;
gate->hash = (void*)&argon2d500_hash;
gate->optimizations = SSE2_OPT | AVX2_OPT | AVX512_OPT | NEON_OPT;
opt_target_factor = 65536.0;
return true;
}
// Unitus
int scanhash_argon2d4096( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{

12
algo/argon2d/argon2d-gate.h
View File

@@ -5,19 +5,19 @@
#include <stdint.h>
// Credits: version = 0x10, m_cost = 250.
bool register_argon2d_crds_algo( algo_gate_t* gate );
bool register_argon2d250_algo( algo_gate_t* gate );
void argon2d_crds_hash( void *state, const void *input );
void argon2d250_hash( void *state, const void *input );
int scanhash_argon2d_crds( struct work *work, uint32_t max_nonce,
int scanhash_argon2d250( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
// Dynamic: version = 0x10, m_cost = 500.
bool register_argon2d_dyn_algo( algo_gate_t* gate );
bool register_argon2d500_algo( algo_gate_t* gate );
void argon2d_dyn_hash( void *state, const void *input );
void argon2d500_hash( void *state, const void *input );
int scanhash_argon2d_dyn( struct work *work, uint32_t max_nonce,
int scanhash_argon2d500( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );

104
algo/scrypt/scrypt.c
View File

@@ -274,9 +274,6 @@ static inline void PBKDF2_SHA256_128_32_SHA_2BUF( uint32_t *tstate0,
#endif // SHA
static const uint32_t keypad_4way[ 4*12 ] __attribute((aligned(32))) =
{
0x80000000, 0x80000000, 0x80000000, 0x80000000,
@@ -447,7 +444,7 @@ static inline void PBKDF2_SHA256_128_32_4way( uint32_t *tstate,
output[i] = bswap_32( ostate[i] );
}
#ifdef HAVE_SHA256_8WAY
#if defined(__AVX2__)
/*
static const uint32_t _ALIGN(32) finalblk_8way[8 * 16] = {
@@ -590,7 +587,7 @@ static inline void PBKDF2_SHA256_128_32_8way( uint32_t *tstate,
output[i] = bswap_32(ostate[i]);
}
#endif /* HAVE_SHA256_8WAY */
#endif //AVX2
#if defined(SIMD512)
@@ -724,25 +721,10 @@ static inline void PBKDF2_SHA256_128_32_16way( uint32_t *tstate,
#endif // AVX512
#define SCRYPT_MAX_WAYS 12
#define HAVE_SCRYPT_3WAY 1
void scrypt_core(uint32_t *X, uint32_t *V, int N);
void scrypt_core_3way(uint32_t *X, uint32_t *V, int N);
#if defined(__AVX2__)
#undef SCRYPT_MAX_WAYS
#define SCRYPT_MAX_WAYS 24
#define HAVE_SCRYPT_6WAY 1
void scrypt_core_6way(uint32_t *X, uint32_t *V, int N);
#endif
#ifndef SCRYPT_MAX_WAYS
#define SCRYPT_MAX_WAYS 1
#endif
#include "scrypt-core-4way.h"
/*
#if ( SCRYPT_THROUGHPUT == 1 )
static bool scrypt_N_1_1_256( const uint32_t *input, uint32_t *output,
uint32_t *midstate, int N, int thr_id )
{
@@ -752,15 +734,12 @@ static bool scrypt_N_1_1_256( const uint32_t *input, uint32_t *output,
memcpy(tstate, midstate, 32);
HMAC_SHA256_80_init(input, tstate, ostate);
PBKDF2_SHA256_80_128(tstate, ostate, input, X);
scrypt_core_simd128( X, scratchbuf, N ); // woring
// scrypt_core_1way( X, V, N ); // working
// scrypt_core(X, V, N);
scrypt_core_1way( X, scratchbuf, N );
PBKDF2_SHA256_128_32(tstate, ostate, X, output);
return true;
}
*/
#endif
#if ( SCRYPT_THROUGHPUT == 8 )
@@ -1201,20 +1180,6 @@ static int scrypt_N_1_1_256_16way( const uint32_t *input, uint32_t *output,
if ( work_restart[thrid].restart ) return 0;
scrypt_core_simd128_2buf( X+448, V, N );
********************/
/*
scrypt_core_3way( X, V, N );
if ( work_restart[thrid].restart ) return 0;
scrypt_core_3way( X+ 96, V, N );
if ( work_restart[thrid].restart ) return 0;
scrypt_core_simd128_2buf( X+192, V, N );
if ( work_restart[thrid].restart ) return 0;
scrypt_core_3way( X+256, V, N );
if ( work_restart[thrid].restart ) return 0;
scrypt_core_3way( X+352, V, N );
if ( work_restart[thrid].restart ) return 0;
scrypt_core_simd128_2buf( X+448, V, N );
*/
if ( work_restart[thrid].restart ) return 0;
@@ -1321,8 +1286,7 @@ static int scrypt_N_1_1_256_4way_sha( const uint32_t *input, uint32_t *output,
return 1;
}
#else
// SSE2
#elif defined(__SSE2__) || defined(__ARM_NEON)
static int scrypt_N_1_1_256_4way( const uint32_t *input, uint32_t *output,
uint32_t *midstate, int N, int thrid )
@@ -1481,7 +1445,7 @@ bool scrypt_miner_thread_init( int thr_id )
bool register_scrypt_algo( algo_gate_t* gate )
{
#if defined(__SHA__) || defined(__ARM_FEATURE_SHA2)
gate->optimizations = SSE2_OPT | SHA256_OPT | NEON_OPT;
gate->optimizations = SSE2_OPT | SSE42_OPT | AVX_OPT | SHA256_OPT | NEON_OPT;
#else
gate->optimizations = SSE2_OPT | SSE42_OPT | AVX_OPT | AVX2_OPT | AVX512_OPT | NEON_OPT;
#endif
@@ -1491,31 +1455,31 @@ bool register_scrypt_algo( algo_gate_t* gate )
opt_param_n = opt_param_n ? opt_param_n : 1024;
applog( LOG_INFO,"Scrypt paramaters: N= %d, R= 1", opt_param_n );
// scrypt_throughput defined at compile time and used to replace
// MAX_WAYS to reduce memory usage.
#if defined(SIMD512)
// scrypt_throughput = 16;
if ( opt_param_n > 0x4000 )
scratchbuf_size = opt_param_n * 3 * 128; // 3 buf
else
scratchbuf_size = opt_param_n * 4 * 128; // 4 way
#elif defined(__SHA__) || defined(__ARM_FEATURE_SHA2)
// scrypt_throughput = 2;
scratchbuf_size = opt_param_n * 2 * 128; // 2 buf
#elif defined(__AVX2__)
// scrypt_throughput = 8;
if ( opt_param_n > 0x4000 )
scratchbuf_size = opt_param_n * 3 * 128; // 3 buf
else
scratchbuf_size = opt_param_n * 2 * 128; // 2 way
#else
// scrypt_throughput = 4;
if ( opt_param_n > 0x4000 )
scratchbuf_size = opt_param_n * 2 * 128; // 2 buf
else
scratchbuf_size = opt_param_n * 4 * 128; // 4 way
#endif
switch ( SCRYPT_THROUGHPUT )
{
case 16: // AVX512
if ( opt_param_n > 0x4000 )
scratchbuf_size = opt_param_n * 3 * 128; // 3 buf
else
scratchbuf_size = opt_param_n * 4 * 128; // 4 way
break;
case 2: // SHA256
scratchbuf_size = opt_param_n * 2 * 128; // 2 buf
break;
case 8: // AVX2
if ( opt_param_n > 0x4000 )
scratchbuf_size = opt_param_n * 3 * 128; // 3 buf
else
scratchbuf_size = opt_param_n * 2 * 128; // 2 way
break;
case 4: // SSE2, NEON
if ( opt_param_n > 0x4000 )
scratchbuf_size = opt_param_n * 2 * 128; // 2 buf
else
scratchbuf_size = opt_param_n * 4 * 128; // 4 way
default:
scratchbuf_size = opt_param_n; // 1 way
}
char t_units[4] = {0};
char d_units[4] = {0};

20
build-avx2.sh
View File

@@ -1,27 +1,9 @@
#!/bin/bash
#if [ "$OS" = "Windows_NT" ]; then
# ./mingw64.sh
# exit 0
#fi
#!/bin/sh
# Linux build
make distclean || echo clean
rm -f config.status
./autogen.sh || echo done
# Ubuntu 10.04 (gcc 4.4)
# extracflags="-O3 -march=native -Wall -D_REENTRANT -funroll-loops -fvariable-expansion-in-unroller -fmerge-all-constants -fbranch-target-load-optimize2 -fsched2-use-superblocks -falign-loops=16 -falign-functions=16 -falign-jumps=16 -falign-labels=16"
# Debian 7.7 / Ubuntu 14.04 (gcc 4.7+)
#extracflags="$extracflags -Ofast -flto -fuse-linker-plugin -ftree-loop-if-convert-stores"
#CFLAGS="-O3 -march=native -Wall" ./configure --with-curl --with-crypto=$HOME/usr
CFLAGS="-O3 -march=haswell -maes -Wall" ./configure --with-curl
#CFLAGS="-O3 -march=native -Wall" CXXFLAGS="$CFLAGS -std=gnu++11" ./configure --with-curl
make -j $(nproc)
strip -s cpuminer

1881
configure vendored
View File

File diff suppressed because it is too large Load Diff

2
configure.ac
View File

@@ -1,4 +1,4 @@
AC_INIT([cpuminer-opt], [25.2])
AC_INIT([cpuminer-opt], [25.3])
AC_PREREQ([2.59c])
AC_CANONICAL_SYSTEM

1877
configure~
View File

File diff suppressed because it is too large Load Diff

8
cpu-miner.c
View File

@@ -2878,6 +2878,7 @@ static bool cpu_capability( bool display_only )
bool sw_has_vaes = false;
bool sw_has_sha256 = false; // x86_64 or AArch64
bool sw_has_sha512 = false; // x86_64 or AArch64
/*
set_t algo_features = algo_gate.optimizations;
bool algo_has_sse2 = set_incl( SSE2_OPT, algo_features );
bool algo_has_sse42 = set_incl( SSE42_OPT, algo_features );
@@ -2900,7 +2901,7 @@ static bool cpu_capability( bool display_only )
bool use_sha512;
bool use_neon;
bool use_none;
*/
#if defined(__x86_64__)
sw_has_x86_64 = true;
#elif defined(__aarch64__)
@@ -3057,6 +3058,9 @@ static bool cpu_capability( bool display_only )
if ( sw_has_sha512 ) printf( " SHA512" );
else if ( sw_has_sha256 ) printf( " SHA256" );
printf("\n");
/*
if ( !display_only )
{
printf("\nAlgo features:");
@@ -3076,6 +3080,7 @@ static bool cpu_capability( bool display_only )
}
printf("\n");
if ( display_only ) return true;
// Determine mining options
@@ -3108,6 +3113,7 @@ static bool cpu_capability( bool display_only )
else if ( use_sha256 ) printf( " SHA256" );
printf( "\n" );
}
*/
return true;
}

25
miner.h
View File

@@ -289,26 +289,6 @@ static inline void le16enc(void *pp, uint16_t x)
json_t* json_load_url(char* cfg_url, json_error_t *err);
//void sha256_init(uint32_t *state);
//void sha256_transform(uint32_t *state, const uint32_t *block, int swap);
//void sha256d(unsigned char *hash, const unsigned char *data, int len);
#ifdef USE_ASM
#if defined(__ARM_NEON__) || defined(__i386__) || defined(__x86_64__)
#define HAVE_SHA256_4WAY 1
int sha256_use_4way();
void sha256_init_4way(uint32_t *state);
void sha256_transform_4way(uint32_t *state, const uint32_t *block, int swap);
#endif
//#if defined(__x86_64__) && defined(USE_AVX2)
#if defined(__x86_64__) && defined(__AVX2__)
#define HAVE_SHA256_8WAY 1
int sha256_use_8way();
void sha256_init_8way(uint32_t *state);
void sha256_transform_8way(uint32_t *state, const uint32_t *block, int swap);
#endif
#endif
struct work;
void work_free(struct work *w);
@@ -851,10 +831,9 @@ Options:\n\
-a, --algo=ALGO specify the algorithm to use\n\
allium Garlicoin (GRLC)\n\
anime Animecoin (ANI)\n\
argon2 Argon2 Coin (AR2)\n\
argon2d250\n\
argon2d500 argon2d-dyn, Dynamic (DYN)\n\
argon2d4096 argon2d-uis, Unitus (UIS)\n\
argon2d500\n\
argon2d4096\n\
axiom Shabal-256 MemoHash\n\
blake blake256r14 (SFR)\n\
blake2b Blake2b 256\n\

12
winbuild-cross.sh
View File

@@ -120,12 +120,12 @@ CFLAGS="-msse2 $DEFAULT_CFLAGS_OLD" ./configure $CONFIGURE_ARGS
make -j 8
strip -s cpuminer.exe
mv cpuminer.exe release/cpuminer-sse2.exe
make clean || echo clean
#make clean || echo clean
# Native with CPU groups ennabled
make clean || echo clean
rm -f config.status
CFLAGS="-march=native $DEFAULT_CFLAGS_OLD" ./configure $CONFIGURE_ARGS
make -j 8
strip -s cpuminer.exe
#make clean || echo clean
#rm -f config.status
#CFLAGS="-march=native $DEFAULT_CFLAGS_OLD" ./configure $CONFIGURE_ARGS
#make -j 8
#strip -s cpuminer.exe