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

v3.12.2

Browse Source
This commit is contained in:
Jay D Dee
2020-02-09 13:30:40 -05:00
parent dc2f8d81d3
commit 3da2b958cf
39 changed files with 1496 additions and 1518 deletions
Download Patch File Download Diff File Expand all files Collapse all files

297
algo/x22/x22i-4way.c
View File

@@ -87,64 +87,40 @@ void x22i_8way_hash( void *output, const void *input )
unsigned char hashA7[64] __attribute__((aligned(32))) = {0};
x22i_8way_ctx_overlay ctx;
blake512_8way_init( &ctx.blake );
blake512_8way_update( &ctx.blake, input, 80 );
blake512_8way_close( &ctx.blake, vhash );
blake512_8way_full( &ctx.blake, vhash, input, 80 );
bmw512_8way_init( &ctx.bmw );
bmw512_8way_update( &ctx.bmw, vhash, 64 );
bmw512_8way_close( &ctx.bmw, vhash );
bmw512_8way_full( &ctx.bmw, vhash, vhash, 64 );
#if defined(__VAES__)
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
groestl512_4way_init( &ctx.groestl, 64 );
groestl512_4way_update_close( &ctx.groestl, vhashA, vhashA, 512 );
groestl512_4way_init( &ctx.groestl, 64 );
groestl512_4way_update_close( &ctx.groestl, vhashB, vhashB, 512 );
groestl512_4way_full( &ctx.groestl, vhashA, vhashA, 64 );
groestl512_4way_full( &ctx.groestl, vhashB, vhashB, 64 );
rintrlv_4x128_8x64( vhash, vhashA, vhashB, 512 );
rintrlv_4x128_8x64( vhash, vhashA, vhashB, 512 );
#else
dintrlv_8x64_512( hash0, hash1, hash2, hash3,
hash4, hash5, hash6, hash7, vhash );
dintrlv_8x64_512( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
vhash );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash0,
(const char*)hash0, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash1,
(const char*)hash1, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash2,
(const char*)hash2, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3,
(const char*)hash3, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash4,
(const char*)hash4, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash5,
(const char*)hash5, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash6,
(const char*)hash6, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash7,
(const char*)hash7, 512 );
groestl512_full( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
groestl512_full( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
groestl512_full( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
groestl512_full( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
groestl512_full( &ctx.groestl, (char*)hash4, (char*)hash4, 512 );
groestl512_full( &ctx.groestl, (char*)hash5, (char*)hash5, 512 );
groestl512_full( &ctx.groestl, (char*)hash6, (char*)hash6, 512 );
groestl512_full( &ctx.groestl, (char*)hash7, (char*)hash7, 512 );
intrlv_8x64_512( vhash, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
hash7 );
intrlv_8x64_512( vhash, hash0, hash1, hash2, hash3,
hash4, hash5, hash6, hash7 );
#endif
skein512_8way_init( &ctx.skein );
skein512_8way_update( &ctx.skein, vhash, 64 );
skein512_8way_close( &ctx.skein, vhash );
skein512_8way_full( &ctx.skein, vhash, vhash, 64 );
jh512_8way_init( &ctx.jh );
jh512_8way_update( &ctx.jh, vhash, 64 );
jh512_8way_close( &ctx.jh, vhash );
@@ -155,22 +131,16 @@ void x22i_8way_hash( void *output, const void *input )
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
luffa_4way_init( &ctx.luffa, 512 );
luffa_4way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
luffa_4way_init( &ctx.luffa, 512 );
luffa_4way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
luffa512_4way_full( &ctx.luffa, vhashA, vhashA, 64 );
luffa512_4way_full( &ctx.luffa, vhashB, vhashB, 64 );
cube_4way_init( &ctx.cube, 512, 16, 32 );
cube_4way_update_close( &ctx.cube, vhashA, vhashA, 64 );
cube_4way_init( &ctx.cube, 512, 16, 32 );
cube_4way_update_close( &ctx.cube, vhashB, vhashB, 64 );
cube_4way_full( &ctx.cube, vhashA, 512, vhashA, 64 );
cube_4way_full( &ctx.cube, vhashB, 512, vhashB, 64 );
#if defined(__VAES__)
shavite512_4way_init( &ctx.shavite );
shavite512_4way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_4way_init( &ctx.shavite );
shavite512_4way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
shavite512_4way_full( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_4way_full( &ctx.shavite, vhashB, vhashB, 64 );
#else
@@ -207,17 +177,13 @@ void x22i_8way_hash( void *output, const void *input )
#endif
simd_4way_init( &ctx.simd, 512 );
simd_4way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_4way_init( &ctx.simd, 512 );
simd_4way_update_close( &ctx.simd, vhashB, vhashB, 512 );
simd512_4way_full( &ctx.simd, vhashA, vhashA, 64 );
simd512_4way_full( &ctx.simd, vhashB, vhashB, 64 );
#if defined(__VAES__)
echo_4way_init( &ctx.echo, 512 );
echo_4way_update_close( &ctx.echo, vhashA, vhashA, 512 );
echo_4way_init( &ctx.echo, 512 );
echo_4way_update_close( &ctx.echo, vhashB, vhashB, 512 );
echo_4way_full( &ctx.echo, vhashA, 512, vhashA, 64 );
echo_4way_full( &ctx.echo, vhashB, 512, vhashB, 64 );
rintrlv_4x128_8x64( vhash, vhashA, vhashB, 512 );
@@ -226,30 +192,22 @@ void x22i_8way_hash( void *output, const void *input )
dintrlv_4x128_512( hash0, hash1, hash2, hash3, vhashA );
dintrlv_4x128_512( hash4, hash5, hash6, hash7, vhashB );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash0,
(const BitSequence*)hash0, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash1,
(const BitSequence*)hash1, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash2,
(const BitSequence*)hash2, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash3,
(const BitSequence*)hash3, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash4,
(const BitSequence*)hash4, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash5,
(const BitSequence*)hash5, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash6,
(const BitSequence*)hash6, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash7,
(const BitSequence*)hash7, 512 );
echo_full( &ctx.echo, (BitSequence *)hash0, 512,
(const BitSequence *)hash0, 64 );
echo_full( &ctx.echo, (BitSequence *)hash1, 512,
(const BitSequence *)hash1, 64 );
echo_full( &ctx.echo, (BitSequence *)hash2, 512,
(const BitSequence *)hash2, 64 );
echo_full( &ctx.echo, (BitSequence *)hash3, 512,
(const BitSequence *)hash3, 64 );
echo_full( &ctx.echo, (BitSequence *)hash4, 512,
(const BitSequence *)hash4, 64 );
echo_full( &ctx.echo, (BitSequence *)hash5, 512,
(const BitSequence *)hash5, 64 );
echo_full( &ctx.echo, (BitSequence *)hash6, 512,
(const BitSequence *)hash6, 64 );
echo_full( &ctx.echo, (BitSequence *)hash7, 512,
(const BitSequence *)hash7, 64 );
intrlv_8x64_512( vhash, hash0, hash1, hash2, hash3,
hash4, hash5, hash6, hash7 );
@@ -443,6 +401,55 @@ void x22i_8way_hash( void *output, const void *input )
sha256_8way_close( &ctx.sha256, output );
}
int scanhash_x22i_8way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[8*8] __attribute__ ((aligned (128)));
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t *hashd7 = &(hash[7*8]);
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const uint32_t last_nonce = max_nonce - 8;
__m512i *noncev = (__m512i*)vdata + 9;
uint32_t n = first_nonce;
const int thr_id = mythr->id;
const uint32_t targ32 = ptarget[7];
const bool bench = opt_benchmark;
if ( bench ) ptarget[7] = 0x08ff;
InitializeSWIFFTX();
mm512_bswap32_intrlv80_8x64( vdata, pdata );
*noncev = mm512_intrlv_blend_32(
_mm512_set_epi32( n+7, 0, n+6, 0, n+5, 0, n+4, 0,
n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
do
{
x22i_8way_hash( hash, vdata );
for ( int lane = 0; lane < 8; lane++ )
if ( unlikely( ( hashd7[ lane ] <= targ32 ) && !bench ) )
{
extr_lane_8x32( lane_hash, hash, lane, 256 );
if ( likely( valid_hash( lane_hash, ptarget ) ) )
{
pdata[19] = bswap_32( n + lane );
submit_lane_solution( work, lane_hash, mythr, lane );
}
}
*noncev = _mm512_add_epi32( *noncev,
m512_const1_64( 0x0000000800000000 ) );
n += 8;
} while ( likely( ( n < last_nonce ) && !work_restart[thr_id].restart ) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}
/*
int scanhash_x22i_8way( struct work* work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
@@ -488,6 +495,7 @@ int scanhash_x22i_8way( struct work* work, uint32_t max_nonce,
*hashes_done = n - first_nonce;
return 0;
}
*/
#elif defined(X22I_4WAY)
@@ -531,33 +539,21 @@ void x22i_4way_hash( void *output, const void *input )
unsigned char hashA3[64] __attribute__((aligned(32))) = {0};
x22i_ctx_overlay ctx;
blake512_4way_init( &ctx.blake );
blake512_4way_update( &ctx.blake, input, 80 );
blake512_4way_close( &ctx.blake, vhash );
blake512_4way_full( &ctx.blake, vhash, input, 80 );
bmw512_4way_init( &ctx.bmw );
bmw512_4way_update( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
dintrlv_4x64_512( hash0, hash1, hash2, hash3, vhash );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash0,
(const char*)hash0, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash1,
(const char*)hash1, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash2,
(const char*)hash2, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3,
(const char*)hash3, 512 );
groestl512_full( &ctx.groestl, (char*)hash0, (const char*)hash0, 512 );
groestl512_full( &ctx.groestl, (char*)hash1, (const char*)hash1, 512 );
groestl512_full( &ctx.groestl, (char*)hash2, (const char*)hash2, 512 );
groestl512_full( &ctx.groestl, (char*)hash3, (const char*)hash3, 512 );
intrlv_4x64_512( vhash, hash0, hash1, hash2, hash3 );
skein512_4way_init( &ctx.skein );
skein512_4way_update( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
skein512_4way_full( &ctx.skein, vhash, vhash, 64 );
jh512_4way_init( &ctx.jh );
jh512_4way_update( &ctx.jh, vhash, 64 );
@@ -569,41 +565,29 @@ void x22i_4way_hash( void *output, const void *input )
rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
luffa512_2way_full( &ctx.luffa, vhashA, vhashA, 64 );
luffa512_2way_full( &ctx.luffa, vhashB, vhashB, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashA, vhashA, 64 );
cube_2way_init( &ctx.cube, 512, 16, 32 );
cube_2way_update_close( &ctx.cube, vhashB, vhashB, 64 );
cube_2way_full( &ctx.cube, vhashA, 512, vhashA, 64 );
cube_2way_full( &ctx.cube, vhashB, 512, vhashB, 64 );
shavite512_2way_full( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_2way_full( &ctx.shavite, vhashB, vhashB, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_2way_init( &ctx.shavite );
shavite512_2way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhashB, vhashB, 512 );
simd512_2way_full( &ctx.simd, vhashA, vhashA, 64 );
simd512_2way_full( &ctx.simd, vhashB, vhashB, 64 );
dintrlv_2x128_512( hash0, hash1, vhashA );
dintrlv_2x128_512( hash2, hash3, vhashB );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash0,
(const BitSequence*)hash0, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash1,
(const BitSequence*)hash1, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash2,
(const BitSequence*)hash2, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash3,
(const BitSequence*)hash3, 512 );
echo_full( &ctx.echo, (BitSequence *)hash0, 512,
(const BitSequence *)hash0, 64 );
echo_full( &ctx.echo, (BitSequence *)hash1, 512,
(const BitSequence *)hash1, 64 );
echo_full( &ctx.echo, (BitSequence *)hash2, 512,
(const BitSequence *)hash2, 64 );
echo_full( &ctx.echo, (BitSequence *)hash3, 512,
(const BitSequence *)hash3, 64 );
intrlv_4x64_512( vhash, hash0, hash1, hash2, hash3 );
@@ -722,44 +706,47 @@ void x22i_4way_hash( void *output, const void *input )
int scanhash_x22i_4way( struct work* work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[4*16] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t hash[8*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t *hashd7 = &(hash[ 7*4 ]);
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
__m256i *noncev = (__m256i*)vdata + 9; // aligned
const uint32_t last_nonce = max_nonce - 4;
__m256i *noncev = (__m256i*)vdata + 9;
uint32_t n = first_nonce;
const int thr_id = mythr->id;
const uint32_t Htarg = ptarget[7];
const uint32_t targ32 = ptarget[7];
const bool bench = opt_benchmark;
if ( bench ) ptarget[7] = 0x08ff;
if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x08ff;
InitializeSWIFFTX();
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 );
do
{
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
x22i_4way_hash( hash, vdata );
for ( int lane = 0; lane < 4; lane++ )
if unlikely( ( hash7[ lane ] <= Htarg ) )
if ( unlikely( hashd7[ lane ] <= targ32 && !bench ) )
{
extr_lane_4x32( lane_hash, hash, lane, 256 );
if ( likely( fulltest( lane_hash, ptarget ) && !opt_benchmark ) )
if ( valid_hash( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
pdata[19] = bswap_32( n + lane );
submit_lane_solution( work, lane_hash, mythr, lane );
}
}
*noncev = _mm256_add_epi32( *noncev,
m256_const1_64( 0x0000000400000000 ) );
n += 4;
} while ( likely( ( n < max_nonce - 4 ) && !work_restart[thr_id].restart ) );
*hashes_done = n - first_nonce + 1;
} while ( likely( ( n <= last_nonce ) && !work_restart[thr_id].restart ) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}

52
algo/x22/x22i.c
View File

@@ -167,40 +167,38 @@ void x22i_hash( void *output, const void *input )
memcpy(output, hash, 32);
}
int scanhash_x22i( struct work* work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
int scanhash_x22i( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t endiandata[20] __attribute__((aligned(64)));
uint32_t hash[8] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
uint32_t n = first_nonce;
uint32_t edata[20] __attribute__((aligned(64)));
uint32_t hash64[8] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
const uint32_t first_nonce = n;
const int thr_id = mythr->id;
const bool bench = opt_benchmark;
if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x08ff;
for (int k=0; k < 20; k++)
be32enc(&endiandata[k], pdata[k]);
if ( bench ) ptarget[7] = 0x08ff;
mm128_bswap32_80( edata, pdata );
InitializeSWIFFTX();
do
{
pdata[19] = ++n;
be32enc( &endiandata[19], n );
x22i_hash( hash, endiandata );
if ( hash[7] < Htarg )
if ( fulltest( hash, ptarget ) && !opt_benchmark )
submit_solution( work, hash, mythr );
} while ( n < max_nonce && !work_restart[thr_id].restart );
*hashes_done = pdata[19] - first_nonce;
return 0;
edata[19] = n;
x22i_hash( hash64, edata );
if ( unlikely( valid_hash( hash64, ptarget ) && !bench ) )
{
pdata[19] = bswap_32( n );
submit_solution( work, hash64, mythr );
}
n++;
} while ( n < max_nonce && !work_restart[thr_id].restart );
*hashes_done = n - first_nonce;
pdata[19] = n;
return 0;
}
#endif

190
algo/x22/x25x-4way.c
View File

@@ -530,6 +530,55 @@ void x25x_8way_hash( void *output, const void *input )
blake2s_8way_full_blocks( &ctx.blake2s, output, vhashX, 64*24 );
}
int scanhash_x25x_8way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[8*8] __attribute__ ((aligned (128)));
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t *hashd7 = &(hash[7*8]);
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const uint32_t last_nonce = max_nonce - 8;
__m512i *noncev = (__m512i*)vdata + 9;
uint32_t n = first_nonce;
const int thr_id = mythr->id;
const uint32_t targ32 = ptarget[7];
const bool bench = opt_benchmark;
if ( bench ) ptarget[7] = 0x08ff;
InitializeSWIFFTX();
mm512_bswap32_intrlv80_8x64( vdata, pdata );
*noncev = mm512_intrlv_blend_32(
_mm512_set_epi32( n+7, 0, n+6, 0, n+5, 0, n+4, 0,
n+3, 0, n+2, 0, n+1, 0, n, 0 ), *noncev );
do
{
x25x_8way_hash( hash, vdata );
for ( int lane = 0; lane < 8; lane++ )
if ( unlikely( ( hashd7[ lane ] <= targ32 ) && !bench ) )
{
extr_lane_8x32( lane_hash, hash, lane, 256 );
if ( likely( valid_hash( lane_hash, ptarget ) ) )
{
pdata[19] = bswap_32( n + lane );
submit_lane_solution( work, lane_hash, mythr, lane );
}
}
*noncev = _mm512_add_epi32( *noncev,
m512_const1_64( 0x0000000800000000 ) );
n += 8;
} while ( likely( ( n < last_nonce ) && !work_restart[thr_id].restart ) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}
/*
int scanhash_x25x_8way( struct work* work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
@@ -574,6 +623,7 @@ int scanhash_x25x_8way( struct work* work, uint32_t max_nonce,
*hashes_done = n - first_nonce;
return 0;
}
*/
#elif defined(X25X_4WAY)
@@ -614,9 +664,7 @@ void x25x_4way_hash( void *output, const void *input )
unsigned char vhashX[24][64*4] __attribute__ ((aligned (64)));
x25x_4way_ctx_overlay ctx __attribute__ ((aligned (64)));
blake512_4way_init( &ctx.blake );
blake512_4way_update( &ctx.blake, input, 80 );
blake512_4way_close( &ctx.blake, vhash );
blake512_4way_full( &ctx.blake, vhash, input, 80 );
dintrlv_4x64_512( hash0[0], hash1[0], hash2[0], hash3[0], vhash );
bmw512_4way_init( &ctx.bmw );
@@ -624,24 +672,13 @@ void x25x_4way_hash( void *output, const void *input )
bmw512_4way_close( &ctx.bmw, vhash );
dintrlv_4x64_512( hash0[1], hash1[1], hash2[1], hash3[1], vhash );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash0[2],
(const char*)hash0[1], 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash1[2],
(const char*)hash1[1], 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash2[2],
(const char*)hash2[1], 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3[2],
(const char*)hash3[1], 512 );
groestl512_full( &ctx.groestl, (char*)hash0[2], (const char*)hash0[1], 512 );
groestl512_full( &ctx.groestl, (char*)hash1[2], (const char*)hash1[1], 512 );
groestl512_full( &ctx.groestl, (char*)hash2[2], (const char*)hash2[1], 512 );
groestl512_full( &ctx.groestl, (char*)hash3[2], (const char*)hash3[1], 512 );
intrlv_4x64_512( vhash, hash0[2], hash1[2], hash2[2], hash3[2] );
skein512_4way_init( &ctx.skein );
skein512_4way_update( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
skein512_4way_full( &ctx.skein, vhash, vhash, 64 );
dintrlv_4x64_512( hash0[3], hash1[3], hash2[3], hash3[3], vhash );
jh512_4way_init( &ctx.jh );
@@ -654,32 +691,20 @@ void x25x_4way_hash( void *output, const void *input )
keccak512_4way_close( &ctx.keccak, vhash );
dintrlv_4x64_512( hash0[5], hash1[5], hash2[5], hash3[5], vhash );
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash0[6],
(const BitSequence*)hash0[5], 64 );
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash1[6],
(const BitSequence*)hash1[5], 64 );
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash2[6],
(const BitSequence*)hash2[5], 64 );
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash3[6],
(const BitSequence*)hash3[5], 64 );
cubehashInit( &ctx.cube, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash0[7],
(const byte*)hash0[6], 64 );
cubehashInit( &ctx.cube, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash1[7],
(const byte*)hash1[6], 64 );
cubehashInit( &ctx.cube, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash2[7],
(const byte*)hash2[6], 64 );
cubehashInit( &ctx.cube, 512, 16, 32 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash3[7],
(const byte*)hash3[6], 64 );
luffa_full( &ctx.luffa, (BitSequence*)hash0[6], 512,
(const BitSequence*)hash0[5], 64 );
luffa_full( &ctx.luffa, (BitSequence*)hash1[6], 512,
(const BitSequence*)hash1[5], 64 );
luffa_full( &ctx.luffa, (BitSequence*)hash2[6], 512,
(const BitSequence*)hash2[5], 64 );
luffa_full( &ctx.luffa, (BitSequence*)hash3[6], 512,
(const BitSequence*)hash3[5], 64 );
cubehash_full( &ctx.cube, (byte*)hash0[7], 512, (const byte*)hash0[6], 64 );
cubehash_full( &ctx.cube, (byte*)hash1[7], 512, (const byte*)hash1[6], 64 );
cubehash_full( &ctx.cube, (byte*)hash2[7], 512, (const byte*)hash2[6], 64 );
cubehash_full( &ctx.cube, (byte*)hash3[7], 512, (const byte*)hash3[6], 64 );
sph_shavite512_init(&ctx.shavite);
sph_shavite512(&ctx.shavite, (const void*) hash0[7], 64);
sph_shavite512_close(&ctx.shavite, hash0[8]);
@@ -693,31 +718,23 @@ void x25x_4way_hash( void *output, const void *input )
sph_shavite512(&ctx.shavite, (const void*) hash3[7], 64);
sph_shavite512_close(&ctx.shavite, hash3[8]);
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence*)hash0[9],
(const BitSequence*)hash0[8], 512 );
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence*)hash1[9],
(const BitSequence*)hash1[8], 512 );
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence*)hash2[9],
(const BitSequence*)hash2[8], 512 );
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence*)hash3[9],
(const BitSequence*)hash3[8], 512 );
simd_full( &ctx.simd, (BitSequence*)hash0[9],
(const BitSequence*)hash0[8], 512 );
simd_full( &ctx.simd, (BitSequence*)hash1[9],
(const BitSequence*)hash1[8], 512 );
simd_full( &ctx.simd, (BitSequence*)hash2[9],
(const BitSequence*)hash2[8], 512 );
simd_full( &ctx.simd, (BitSequence*)hash3[9],
(const BitSequence*)hash3[8], 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash0[10],
(const BitSequence*)hash0[9], 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash1[10],
(const BitSequence*)hash1[9], 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash2[10],
(const BitSequence*)hash2[9], 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence*)hash3[10],
(const BitSequence*)hash3[9], 512 );
echo_full( &ctx.echo, (BitSequence *)hash0[10], 512,
(const BitSequence *)hash0[ 9], 64 );
echo_full( &ctx.echo, (BitSequence *)hash1[10], 512,
(const BitSequence *)hash1[ 9], 64 );
echo_full( &ctx.echo, (BitSequence *)hash2[10], 512,
(const BitSequence *)hash2[ 9], 64 );
echo_full( &ctx.echo, (BitSequence *)hash3[10], 512,
(const BitSequence *)hash3[ 9], 64 );
intrlv_4x64_512( vhash, hash0[10], hash1[10], hash2[10], hash3[10] );
@@ -870,43 +887,46 @@ void x25x_4way_hash( void *output, const void *input )
int scanhash_x25x_4way( struct work* work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[16*4] __attribute__ ((aligned (128)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t hash[8*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t *hashd7 = &(hash[ 7*4 ]);
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
__m256i *noncev = (__m256i*)vdata + 9; // aligned
uint32_t n = first_nonce;
const uint32_t last_nonce = max_nonce - 4;
__m256i *noncev = (__m256i*)vdata + 9;
uint32_t n = first_nonce;
const int thr_id = mythr->id;
const uint32_t Htarg = ptarget[7];
const uint32_t targ32 = ptarget[7];
const bool bench = opt_benchmark;
if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x08ff;
if ( bench ) ptarget[7] = 0x08ff;
InitializeSWIFFTX();
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 );
do
{
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
x25x_4way_hash( hash, vdata );
for ( int lane = 0; lane < 4; lane++ ) if ( hash7[lane] <= Htarg )
for ( int lane = 0; lane < 4; lane++ )
if ( unlikely( hashd7[ lane ] <= targ32 && !bench ) )
{
extr_lane_4x32( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
if ( valid_hash( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
submit_lane_solution( work, lane_hash, mythr, lane );
pdata[19] = bswap_32( n + lane );
submit_lane_solution( work, lane_hash, mythr, lane );
}
}
*noncev = _mm256_add_epi32( *noncev,
m256_const1_64( 0x0000000400000000 ) );
n += 4;
} while ( likely( ( n < last_nonce ) && !work_restart[thr_id].restart ) );
} while ( likely( ( n <= last_nonce ) && !work_restart[thr_id].restart ) );
pdata[19] = n;
*hashes_done = n - first_nonce;
return 0;
}

46
algo/x22/x25x.c
View File

@@ -201,42 +201,38 @@ void x25x_hash( void *output, const void *input )
memcpy(output, &hash[24], 32);
}
int scanhash_x25x( struct work* work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
int scanhash_x25x( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr)
{
uint32_t edata[20] __attribute__((aligned(64)));
uint32_t hash[8] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
uint32_t n = first_nonce;
uint32_t hash64[8] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
const uint32_t first_nonce = n;
const int thr_id = mythr->id;
const bool bench = opt_benchmark;
if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x08ff;
if ( bench ) ptarget[7] = 0x08ff;
mm128_bswap32_80( edata, pdata );
for (int k=0; k < 20; k++)
be32enc(&edata[k], pdata[k]);
InitializeSWIFFTX();
do
{
pdata[19] = ++n;
be32enc( &edata[19], n );
x25x_hash( hash, edata );
if ( hash[7] < Htarg )
if ( fulltest( hash, ptarget ) && !opt_benchmark )
submit_solution( work, hash, mythr );
} while ( n < max_nonce && !work_restart[thr_id].restart );
*hashes_done = pdata[19] - first_nonce;
return 0;
edata[19] = n;
x25x_hash( hash64, edata );
if ( unlikely( valid_hash( hash64, ptarget ) && !bench ) )
{
pdata[19] = bswap_32( n );
submit_solution( work, hash64, mythr );
}
n++;
} while ( n < max_nonce && !work_restart[thr_id].restart );
*hashes_done = n - first_nonce;
pdata[19] = n;
return 0;
}
#endif