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

v3.9.5.4

Browse Source
This commit is contained in:
Jay D Dee
2019-07-15 17:00:26 -04:00
parent e625ed5420
commit e2d5762ef2
63 changed files with 1973 additions and 2980 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
algo/quark/anime-4way.c
View File

@@ -62,7 +62,7 @@ void anime_4way_hash( void *state, const void *input )
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ), zero );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
if ( hash0[0] & mask )
{
@@ -88,7 +88,7 @@ void anime_4way_hash( void *state, const void *input )
(char*)hash3, 512 );
}
mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
if ( mm256_anybits0( vh_mask ) )
{
@@ -98,7 +98,7 @@ void anime_4way_hash( void *state, const void *input )
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -109,7 +109,7 @@ void anime_4way_hash( void *state, const void *input )
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
@@ -155,7 +155,7 @@ void anime_4way_hash( void *state, const void *input )
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
mm256_dintrlv_4x64( state, state+32, state+64, state+96, vhash, 256 );
dintrlv_4x64( state, state+32, state+64, state+96, vhash, 256 );
}
int scanhash_anime_4way( struct work *work, uint32_t max_nonce,
@@ -163,7 +163,6 @@ int scanhash_anime_4way( struct work *work, uint32_t max_nonce,
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __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];
@@ -188,9 +187,7 @@ int scanhash_anime_4way( struct work *work, uint32_t max_nonce,
0
};
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
mm256_bswap32_intrlv80_4x64( vdata, pdata );
for (int m=0; m < 6; m++)
if (Htarg <= htmax[m])

44
algo/quark/hmq1725-4way.c
View File

@@ -67,7 +67,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
bmw512_4way( &ctx.bmw, input, 80 );
bmw512_4way_close( &ctx.bmw, vhash );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash0, 64 );
@@ -84,7 +84,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
// first fork, A is groestl serial, B is skein parallel.
mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
m256_zero );
@@ -116,7 +116,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
(char*)hash3, 512 );
// }
mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
// B
@@ -158,7 +158,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash0,
@@ -186,7 +186,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
cubehashUpdateDigest( &ctx.cube, (BitSequence *)hash3,
(const BitSequence *)hash3, 64 );
mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
// A= keccak parallel, B= jh parallel
@@ -209,7 +209,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
sph_shavite512_init( &ctx.shavite );
sph_shavite512 ( &ctx.shavite, hash0, 64 );
@@ -240,7 +240,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
// A is whirlpool serial, B is haval parallel.
mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
m256_zero );
@@ -271,7 +271,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
sph_whirlpool_close( &ctx.whirlpool, hash3 );
// }
mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
// B
@@ -285,7 +285,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -300,13 +300,13 @@ extern void hmq1725_4way_hash(void *state, const void *input)
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *)hash3, 512 );
mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
blake512_4way_init( &ctx.blake );
blake512_4way( &ctx.blake, vhash, 64 );
blake512_4way_close( &ctx.blake, vhash );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// shavite & luffa, both serial, select individually.
@@ -362,13 +362,13 @@ extern void hmq1725_4way_hash(void *state, const void *input)
(const BitSequence *)hash3, 64 );
}
mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way_init( &ctx.hamsi );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -438,13 +438,13 @@ extern void hmq1725_4way_hash(void *state, const void *input)
(const BitSequence *)hash3, 512 );
}
mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way_init( &ctx.shabal );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, hash0, 64 );
@@ -461,7 +461,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
// A = fugue serial, B = sha512 prarallel
mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
m256_zero );
@@ -491,7 +491,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
sph_fugue512_close( &ctx.fugue, hash3 );
// }
mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
// if ( mm256_any_clr_256( vh_mask ) )
// {
@@ -502,7 +502,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -513,7 +513,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
sha512_4way_init( &ctx.sha512 );
sha512_4way( &ctx.sha512, vhash, 64 );
@@ -524,7 +524,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], vmask ),
m256_zero );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// if ( mm256_any_set_256( vh_mask ) ) //4
// {
@@ -559,7 +559,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
sph_whirlpool_close( &ctx.whirlpool, hash3 );
// }
mm256_intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhashB, hash0, hash1, hash2, hash3, 512 );
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
@@ -589,7 +589,7 @@ int scanhash_hmq1725_4way( struct work *work, uint32_t max_nonce,
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
0xFFFFF000, 0xFFFF0000, 0 };
mm256_bswap_intrlv80_4x64( vdata, pdata );
mm256_bswap32_intrlv80_4x64( vdata, pdata );
for ( int m = 0; m < 6; m++ ) if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[ m ];

15
algo/quark/quark-4way.c
View File

@@ -63,7 +63,7 @@ void quark_4way_hash( void *state, const void *input )
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ), zero );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
if ( hash0[0] & mask )
{
@@ -89,7 +89,7 @@ void quark_4way_hash( void *state, const void *input )
(char*)hash3, 512 );
}
mm256_intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhashA, hash0, hash1, hash2, hash3, 512 );
if ( mm256_anybits0( vh_mask ) )
{
@@ -99,7 +99,7 @@ void quark_4way_hash( void *state, const void *input )
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -110,7 +110,7 @@ void quark_4way_hash( void *state, const void *input )
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
@@ -168,7 +168,6 @@ int scanhash_quark_4way( struct work *work, uint32_t max_nonce,
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[25]);
uint32_t *pdata = work->data;
@@ -178,9 +177,7 @@ int scanhash_quark_4way( struct work *work, uint32_t max_nonce,
__m256i *noncev = (__m256i*)vdata + 9; // aligned
int thr_id = mythr->id; // thr_id arg is deprecated
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
mm256_bswap32_intrlv80_4x64( vdata, pdata );
do
{
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(
@@ -192,7 +189,7 @@ int scanhash_quark_4way( struct work *work, uint32_t max_nonce,
for ( int i = 0; i < 4; i++ )
if ( ( hash7[ i<<1 ] & 0xFFFFFF00 ) == 0 )
{
mm256_extr_lane_4x64( lane_hash, hash, i, 256 );
extr_lane_4x64( lane_hash, hash, i, 256 );
if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
{
pdata[19] = n+i;