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

v3.12.0.1

Browse Source
This commit is contained in:
Jay D Dee
2020-02-06 22:50:20 -05:00
parent 13523a12f9
commit fc97ef174a
7 changed files with 82 additions and 102 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
RELEASE_NOTES
View File

@@ -65,6 +65,17 @@ If not what makes it happen or not happen?
Change Log
----------
v3.12.0.1
Fixed hodl rejects, git issue #237.
Fixed debug code added in v3.12.0 to work with AVX2 to be enabled only
after low difficulty share have been seen to avoid unnecessarily excessive
log outout.
Added more digits of precision to diff in log output to help diagnose
low difficulty shares.
v3.12.0
Faster phi2 AVX2 +62%, AVX512 +150% on Intel CPUs. AMD Ryzen AVX2 is

2
algo/hodl/hodl-gate.c
View File

@@ -144,7 +144,7 @@ int hodl_scanhash( struct work* work, uint32_t max_nonce,
#if defined(__AES__)
GenRandomGarbage( (CacheEntry*)hodl_scratchbuf, work->data, mythr->id );
pthread_barrier_wait( &hodl_barrier );
return scanhash_hodl_wolf( work, max_nonce, hashes_done, thr_info );
return scanhash_hodl_wolf( work, max_nonce, hashes_done, mythr );
#endif
return false;
}

62
algo/quark/anime-4way.c
View File

@@ -162,59 +162,35 @@ int scanhash_anime_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
const uint32_t first_nonce = pdata[19];
const uint32_t last_nonce = max_nonce - 4;
__m256i *noncev = (__m256i*)vdata + 9; // aligned
int thr_id = mythr->id; // thr_id arg is deprecated
const uint32_t Htarg = ptarget[7];
uint64_t htmax[] = {
0,
0xF,
0xFF,
0xFFF,
0xFFFF,
0x10000000
};
uint32_t masks[] = {
0xFFFFFFFF,
0xFFFFFFF0,
0xFFFFFF00,
0xFFFFF000,
0xFFFF0000,
0
};
const int thr_id = mythr->id;
mm256_bswap32_intrlv80_4x64( vdata, pdata );
*noncev = mm256_intrlv_blend_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
for (int m=0; m < 6; m++)
if (Htarg <= htmax[m])
do
{
anime_4way_hash( hash, vdata );
for ( int i = 0; i < 4; i++ )
if ( valid_hash( hash+(i<<3), ptarget ) && !opt_benchmark )
{
uint32_t mask = masks[m];
do
{
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
anime_4way_hash( hash, vdata );
pdata[19] = n;
for ( int i = 0; i < 4; i++ )
if ( ( ( (hash+(i<<3))[7] & mask ) == 0 )
&& fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
{
pdata[19] = n+i;
submit_lane_solution( work, hash+(i<<3), mythr, i );
}
n += 4;
} while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
break;
pdata[19] = bswap_32( n+i );
submit_solution( work, hash+(i<<3), mythr );
}
*hashes_done = n - first_nonce + 1;
*noncev = _mm256_add_epi32( *noncev,
m256_const1_64( 0x0000000400000000 ) );
n += 4;
} while ( ( n < last_nonce ) && !work_restart[thr_id].restart );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}

55
algo/quark/anime.c
View File

@@ -126,49 +126,28 @@ int scanhash_anime( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t hash[8] __attribute__ ((aligned (64)));
uint32_t endiandata[20] __attribute__((aligned(64)));
uint32_t edata[20] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
const uint32_t first_nonce = pdata[19];
int thr_id = mythr->id; // thr_id arg is deprecated
const uint32_t Htarg = ptarget[7];
uint64_t htmax[] = {
0,
0xF,
0xFF,
0xFFF,
0xFFFF,
0x10000000
};
uint32_t masks[] = {
0xFFFFFFFF,
0xFFFFFFF0,
0xFFFFFF00,
0xFFFFF000,
0xFFFF0000,
0
};
const int thr_id = mythr->id;
const int bench = opt_benchmark;
swab32_array( edata, pdata, 20 );
swab32_array( endiandata, pdata, 20 );
for (int m=0; m < 6; m++)
if (Htarg <= htmax[m])
{
uint32_t mask = masks[m];
do
{
be32enc( &endiandata[19], n );
anime_hash( hash, endiandata );
pdata[19] = n;
if ( ( hash[7] & mask ) == 0 && fulltest( hash, ptarget ) )
submit_solution( work, hash, mythr );
n++;
} while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
break;
}
*hashes_done = n - first_nonce + 1;
do
{
edata[19] = n;
anime_hash( hash, edata );
if ( valid_hash( hash, ptarget ) && !bench )
{
be32enc( &pdata[19], n );
submit_solution( work, hash, mythr );
}
n++;
} while ( ( n < max_nonce ) && !work_restart[thr_id].restart );
*hashes_done = n - first_nonce;
pdata[19] = n;
return 0;
}

20
configure vendored
View File

@@ -1,6 +1,6 @@
#! /bin/sh
# Guess values for system-dependent variables and create Makefiles.
# Generated by GNU Autoconf 2.69 for cpuminer-opt 3.12.0.
# Generated by GNU Autoconf 2.69 for cpuminer-opt 3.12.0.1.
#
#
# Copyright (C) 1992-1996, 1998-2012 Free Software Foundation, Inc.
@@ -577,8 +577,8 @@ MAKEFLAGS=
# Identity of this package.
PACKAGE_NAME='cpuminer-opt'
PACKAGE_TARNAME='cpuminer-opt'
PACKAGE_VERSION='3.12.0'
PACKAGE_STRING='cpuminer-opt 3.12.0'
PACKAGE_VERSION='3.12.0.1'
PACKAGE_STRING='cpuminer-opt 3.12.0.1'
PACKAGE_BUGREPORT=''
PACKAGE_URL=''
@@ -1332,7 +1332,7 @@ if test "$ac_init_help" = "long"; then
# Omit some internal or obsolete options to make the list less imposing.
# This message is too long to be a string in the A/UX 3.1 sh.
cat <<_ACEOF
\`configure' configures cpuminer-opt 3.12.0 to adapt to many kinds of systems.
\`configure' configures cpuminer-opt 3.12.0.1 to adapt to many kinds of systems.
Usage: $0 [OPTION]... [VAR=VALUE]...
@@ -1404,7 +1404,7 @@ fi
if test -n "$ac_init_help"; then
case $ac_init_help in
short | recursive ) echo "Configuration of cpuminer-opt 3.12.0:";;
short | recursive ) echo "Configuration of cpuminer-opt 3.12.0.1:";;
esac
cat <<\_ACEOF
@@ -1509,7 +1509,7 @@ fi
test -n "$ac_init_help" && exit $ac_status
if $ac_init_version; then
cat <<\_ACEOF
cpuminer-opt configure 3.12.0
cpuminer-opt configure 3.12.0.1
generated by GNU Autoconf 2.69
Copyright (C) 2012 Free Software Foundation, Inc.
@@ -2012,7 +2012,7 @@ cat >config.log <<_ACEOF
This file contains any messages produced by compilers while
running configure, to aid debugging if configure makes a mistake.
It was created by cpuminer-opt $as_me 3.12.0, which was
It was created by cpuminer-opt $as_me 3.12.0.1, which was
generated by GNU Autoconf 2.69. Invocation command line was
$ $0 $@
@@ -2993,7 +2993,7 @@ fi
# Define the identity of the package.
PACKAGE='cpuminer-opt'
VERSION='3.12.0'
VERSION='3.12.0.1'
cat >>confdefs.h <<_ACEOF
@@ -6690,7 +6690,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
# report actual input values of CONFIG_FILES etc. instead of their
# values after options handling.
ac_log="
This file was extended by cpuminer-opt $as_me 3.12.0, which was
This file was extended by cpuminer-opt $as_me 3.12.0.1, which was
generated by GNU Autoconf 2.69. Invocation command line was
CONFIG_FILES = $CONFIG_FILES
@@ -6756,7 +6756,7 @@ _ACEOF
cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
ac_cs_config="`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`"
ac_cs_version="\\
cpuminer-opt config.status 3.12.0
cpuminer-opt config.status 3.12.0.1
configured by $0, generated by GNU Autoconf 2.69,
with options \\"\$ac_cs_config\\"

2
configure.ac
View File

@@ -1,4 +1,4 @@
AC_INIT([cpuminer-opt], [3.12.0])
AC_INIT([cpuminer-opt], [3.12.0.1])
AC_PREREQ([2.59c])
AC_CANONICAL_SYSTEM

32
cpu-miner.c
View File

@@ -998,7 +998,7 @@ void report_summary_log( bool force )
solved_block_count );
}
bool lowdiff_debug = true;
bool lowdiff_debug = false;
static int share_result( int result, struct work *null_work,
const char *reason )
@@ -1068,7 +1068,7 @@ static int share_result( int result, struct work *null_work,
else
{
rejected_share_count++;
if ( strstr( reason, "Low diff " ) ) lowdiff_debug = true;
lowdiff_debug = true;
}
}
@@ -1145,7 +1145,7 @@ static int share_result( int result, struct work *null_work,
bres, share_time, latency );
if ( have_stratum && !opt_quiet )
applog2( LOG_NOTICE, "Diff %.3g (%.3g%), %sBlock %d, %sJob %s" CL_WHT,
applog2( LOG_NOTICE, "Diff %.5g (%.3g%), %sBlock %d, %sJob %s" CL_WHT,
my_stats.share_diff, share_ratio, bcol, stratum.block_height,
scol, my_stats.job_id );
@@ -1163,13 +1163,15 @@ static int share_result( int result, struct work *null_work,
for ( int i = 0; i < 8; i++ )
be32enc( str2 + i, str1[7 - i] );
bin2hex( str3, (unsigned char*)str2, 12 );
applog2( LOG_INFO, "Share diff: %g, Hash: %s...", my_stats.share_diff, str3 );
applog2( LOG_INFO, "Share diff: %.5g, Hash: %s...",
my_stats.share_diff, str3 );
diff_to_target( str1, my_stats.target_diff );
for ( int i = 0; i < 8; i++ )
be32enc( str2 + i, str1[7 - i] );
bin2hex( str3, (unsigned char*)str2, 12 );
applog2( LOG_INFO, "Target diff: %g, Targ: %s...", str3 );
applog2( LOG_INFO, "Target diff: %.5g, Targ: %s...",
my_stats.target_diff, str3 );
}
if ( unlikely( opt_reset_on_stale && stale ) )
@@ -1710,9 +1712,9 @@ if ( lowdiff_debug )
{
uint32_t* h = (uint32_t*)hash;
uint32_t* t = (uint32_t*)work->target;
applog(LOG_INFO,"Hash[7:0}: %08x %08x %08x %08x %08x %08x %08x %08x",
applog(LOG_INFO,"Hash[7:0]: %08x %08x %08x %08x %08x %08x %08x %08x",
h[7],h[6],h[5],h[4],h[3],h[2],h[1],h[0]);
applog(LOG_INFO,"Targ[7:0}: %08x %08x %08x %08x %08x %08x %08x %08x",
applog(LOG_INFO,"Targ[7:0]: %08x %08x %08x %08x %08x %08x %08x %08x",
t[7],t[6],t[5],t[4],t[3],t[2],t[1],t[0]);
}
return true;
@@ -1733,6 +1735,18 @@ bool submit_lane_solution( struct work *work, const void *hash,
if ( !opt_quiet )
applog( LOG_NOTICE, "%d submitted by thread %d, lane %d, job %s",
submitted_share_count, thr->id, lane, work->job_id );
if ( lowdiff_debug )
{
uint32_t* h = (uint32_t*)hash;
uint32_t* t = (uint32_t*)work->target;
applog(LOG_INFO,"Hash[7:0]: %08x %08x %08x %08x %08x %08x %08x %08x",
h[7],h[6],h[5],h[4],h[3],h[2],h[1],h[0]);
applog(LOG_INFO,"Targ[7:0]: %08x %08x %08x %08x %08x %08x %08x %08x",
t[7],t[6],t[5],t[4],t[3],t[2],t[1],t[0]);
}
return true;
}
else
@@ -2117,7 +2131,6 @@ static void *miner_thread( void *userdata )
int lo_freq, hi_freq;
linux_cpu_hilo_freq( &lo_freq, &hi_freq );
memcpy( &cpu_temp_time, &tv_end, sizeof(cpu_temp_time) );
if ( temp > hi_temp ) hi_temp = temp;
if ( use_colors && ( temp >= 70 ) )
{
if ( temp >= 80 )
@@ -2129,6 +2142,7 @@ static void *miner_thread( void *userdata )
sprintf( tempstr, "%d C", temp );
applog( LOG_INFO,"CPU temp: curr %s (max %d), Freq: %.3f/%.3f GHz",
tempstr, hi_temp, (float)lo_freq / 1e6, (float)hi_freq/ 1e6 );
if ( temp > hi_temp ) hi_temp = temp;
}
}
#endif
@@ -2483,7 +2497,7 @@ void std_stratum_gen_work( struct stratum_ctx *sctx, struct work *g_work )
if ( !opt_quiet )
{
applog2( LOG_INFO, "%s: %s", algo_names[opt_algo], short_url );
applog2( LOG_INFO, "Diff: Net %.3g, Stratum %.3g, Target %.3g",
applog2( LOG_INFO, "Diff: Net %.5g, Stratum %.5g, Target %.5g",
net_diff, stratum_diff, last_targetdiff );
if ( likely( hr > 0. ) )