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

v3.11.0

Browse Source
This commit is contained in:
Jay D Dee
2020-01-02 23:54:08 -05:00
parent 241bc26767
commit 3572cb53c4
118 changed files with 7030 additions and 1575 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
algo/quark/anime-4way.c
View File

@@ -54,10 +54,10 @@ void anime_4way_hash( void *state, const void *input )
anime_4way_ctx_holder ctx;
memcpy( &ctx, &anime_4way_ctx, sizeof(anime_4way_ctx) );
bmw512_4way( &ctx.bmw, input, 80 );
bmw512_4way_update( &ctx.bmw, input, 80 );
bmw512_4way_close( &ctx.bmw, vhash );
blake512_4way( &ctx.blake, vhash, 64 );
blake512_4way_update( &ctx.blake, vhash, 64 );
blake512_4way_close( &ctx.blake, vhash );
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ), zero );
@@ -92,7 +92,7 @@ void anime_4way_hash( void *state, const void *input )
if ( mm256_anybits0( vh_mask ) )
{
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_update( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhashB );
}
@@ -111,7 +111,7 @@ void anime_4way_hash( void *state, const void *input )
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_update( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ), zero );
@@ -119,23 +119,23 @@ void anime_4way_hash( void *state, const void *input )
if ( mm256_anybits1( vh_mask ) )
{
blake512_4way_init( &ctx.blake );
blake512_4way( &ctx.blake, vhash, 64 );
blake512_4way_update( &ctx.blake, vhash, 64 );
blake512_4way_close( &ctx.blake, vhashA );
}
if ( mm256_anybits0( vh_mask ) )
{
bmw512_4way_init( &ctx.bmw );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_update( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhashB );
}
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_update( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
skein512_4way_init( &ctx.skein );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_update( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ), zero );
@@ -143,13 +143,13 @@ void anime_4way_hash( void *state, const void *input )
if ( mm256_anybits1( vh_mask ) )
{
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_update( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhashA );
}
if ( mm256_anybits0( vh_mask ) )
{
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_update( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhashB );
}

217
algo/quark/hmq1725-4way.c
View File

@@ -21,6 +21,11 @@
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/haval/haval-hash-4way.h"
#include "algo/sha/sha-hash-4way.h"
#if defined(__VAES__)
#include "algo/groestl/groestl512-hash-4way.h"
#include "algo/shavite/shavite-hash-4way.h"
#include "algo/echo/echo-hash-4way.h"
#endif
#if defined(HMQ1725_8WAY)
@@ -28,21 +33,27 @@ union _hmq1725_8way_context_overlay
{
blake512_8way_context blake;
bmw512_8way_context bmw;
hashState_groestl groestl;
skein512_8way_context skein;
jh512_8way_context jh;
keccak512_8way_context keccak;
luffa_4way_context luffa;
cube_4way_context cube;
sph_shavite512_context shavite;
simd_4way_context simd;
hashState_echo echo;
hamsi512_8way_context hamsi;
sph_fugue512_context fugue;
shabal512_8way_context shabal;
sph_whirlpool_context whirlpool;
sha512_8way_context sha512;
haval256_5_8way_context haval;
#if defined(__VAES__)
groestl512_4way_context groestl;
shavite512_4way_context shavite;
echo_4way_context echo;
#else
hashState_groestl groestl;
sph_shavite512_context shavite;
hashState_echo echo;
#endif
} __attribute__ ((aligned (64)));
typedef union _hmq1725_8way_context_overlay hmq1725_8way_context_overlay;
@@ -52,6 +63,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
uint32_t vhash [16<<3] __attribute__ ((aligned (128)));
uint32_t vhashA[16<<3] __attribute__ ((aligned (64)));
uint32_t vhashB[16<<3] __attribute__ ((aligned (64)));
uint32_t vhashC[16<<3] __attribute__ ((aligned (64)));
uint32_t hash0 [16] __attribute__ ((aligned (64)));
uint32_t hash1 [16] __attribute__ ((aligned (64)));
uint32_t hash2 [16] __attribute__ ((aligned (64)));
@@ -67,6 +79,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
__m512i* vh = (__m512i*)vhash;
__m512i* vhA = (__m512i*)vhashA;
__m512i* vhB = (__m512i*)vhashB;
__m512i* vhC = (__m512i*)vhashC;
bmw512_8way_init( &ctx.bmw );
bmw512_8way_update( &ctx.bmw, input, 80 );
@@ -106,6 +119,28 @@ extern void hmq1725_8way_hash(void *state, const void *input)
m512_zero );
// A
#if defined(__VAES__)
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
if ( likely( ( vh_mask & 0x0f ) != 0x0f ) )
{
groestl512_4way_init( &ctx.groestl, 64 );
groestl512_4way_update_close( &ctx.groestl, vhashA, vhashA, 512 );
}
if ( likely( ( vh_mask & 0xf0 ) != 0xf0 ) )
{
groestl512_4way_init( &ctx.groestl, 64 );
groestl512_4way_update_close( &ctx.groestl, vhashB, vhashB, 512 );
}
rintrlv_4x128_8x64( vhashC, vhashA, vhashB, 512 );
#else
dintrlv_8x64_512( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
vhash );
if ( hash0[0] & mask )
{
init_groestl( &ctx.groestl, 64 );
@@ -140,13 +175,13 @@ extern void hmq1725_8way_hash(void *state, const void *input)
{
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash5,
(char*)hash5, 512 );
(char*)hash5, 512 );
}
if ( hash6[0] & mask )
{
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash6,
(char*)hash6, 512 );
(char*)hash6, 512 );
}
if ( hash7[0] & mask )
{
@@ -155,9 +190,11 @@ extern void hmq1725_8way_hash(void *state, const void *input)
(char*)hash7, 512 );
}
intrlv_8x64_512( vhashA, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
intrlv_8x64_512( vhashC, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
hash7 );
#endif
// B
if ( likely( vh_mask & 0xff ) )
{
@@ -166,7 +203,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
skein512_8way_close( &ctx.skein, vhashB );
}
mm512_blend_hash_8x64( vh, vhA, vhB, vh_mask );
mm512_blend_hash_8x64( vh, vhC, vhB, vh_mask );
jh512_8way_init( &ctx.jh );
jh512_8way_update( &ctx.jh, vhash, 64 );
@@ -225,6 +262,20 @@ extern void hmq1725_8way_hash(void *state, const void *input)
}
mm512_blend_hash_8x64( vh, vhA, vhB, vh_mask );
#if defined(__VAES__)
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
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 );
rintrlv_4x128_8x64( vhash, vhashA, vhashB, 512 );
#else
dintrlv_8x64_512( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
vhash );
@@ -256,6 +307,8 @@ extern void hmq1725_8way_hash(void *state, const void *input)
intrlv_4x128_512( vhashA, hash0, hash1, hash2, hash3 );
intrlv_4x128_512( vhashB, hash4, hash5, hash6, hash7 );
#endif
simd_4way_init( &ctx.simd, 512 );
simd_4way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_4way_init( &ctx.simd, 512 );
@@ -334,6 +387,20 @@ extern void hmq1725_8way_hash(void *state, const void *input)
}
mm512_blend_hash_8x64( vh, vhA, vhB, vh_mask );
#if defined(__VAES__)
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
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 );
rintrlv_4x128_8x64( vhash, vhashA, vhashB, 512 );
#else
dintrlv_8x64_512( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
vhash );
@@ -365,17 +432,38 @@ extern void hmq1725_8way_hash(void *state, const void *input)
intrlv_8x64_512( vhash, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
hash7 );
#endif
blake512_8way_init( &ctx.blake );
blake512_8way_update( &ctx.blake, vhash, 64 );
blake512_8way_close( &ctx.blake, vhash );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], vmask ),
m512_zero );
dintrlv_8x64_512( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
vhash );
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
// A
#if defined(__VAES__)
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
if ( likely( ( vh_mask & 0x0f ) != 0x0f ) )
{
shavite512_4way_init( &ctx.shavite );
shavite512_4way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
}
if ( likely( ( vh_mask & 0xf0 ) != 0xf0 ) )
{
shavite512_4way_init( &ctx.shavite );
shavite512_4way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
}
rintrlv_4x128_8x64( vhashC, vhashA, vhashB, 512 );
#else
dintrlv_8x64_512( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
vhash );
if ( hash0[0] & mask )
{
sph_shavite512_init( &ctx.shavite );
@@ -425,19 +513,28 @@ extern void hmq1725_8way_hash(void *state, const void *input)
sph_shavite512_close( &ctx.shavite, hash7 ); //8
}
intrlv_8x64_512( vhashC, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
hash7 );
#endif
// B
if ( likely( vh_mask & 0xff ) )
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
if ( likely( vh_mask & 0x0f ) )
{
luffa_4way_init( &ctx.luffa, 512 );
luffa_4way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
}
if ( likely( vh_mask & 0xf0 ) )
{
luffa_4way_init( &ctx.luffa, 512 );
luffa_4way_update_close( &ctx.luffa, vhash, vhashB, 64 );
rintrlv_4x128_8x64( vhashB, vhashA, vhash, 512 );
}
intrlv_8x64_512( vhashA, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
hash7 );
mm512_blend_hash_8x64( vh, vhA, vhB, vh_mask );
rintrlv_4x128_8x64( vhashB, vhashA, vhash, 512 );
mm512_blend_hash_8x64( vh, vhC, vhB, vh_mask );
hamsi512_8way_init( &ctx.hamsi );
hamsi512_8way_update( &ctx.hamsi, vhash, 64 );
@@ -475,8 +572,27 @@ extern void hmq1725_8way_hash(void *state, const void *input)
hash7 );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], vmask ),
m512_zero );
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
// A
#if defined(__VAES__)
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
if ( likely( ( vh_mask & 0x0f ) != 0x0f ) )
{
echo_4way_init( &ctx.echo, 512 );
echo_4way_update_close( &ctx.echo, vhashA, vhashA, 512 );
}
if ( likely( ( vh_mask & 0xf0 ) != 0xf0 ) )
{
echo_4way_init( &ctx.echo, 512 );
echo_4way_update_close( &ctx.echo, vhashB, vhashB, 512 );
}
rintrlv_4x128_8x64( vhashC, vhashA, vhashB, 512 );
#else
if ( hash0[0] & mask ) //4
{
init_echo( &ctx.echo, 512 );
@@ -526,19 +642,29 @@ extern void hmq1725_8way_hash(void *state, const void *input)
(const BitSequence *)hash7, 512 );
}
intrlv_8x64_512( vhashC, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
hash7 );
#endif
// B
if ( likely( vh_mask & 0xff ) )
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
if ( likely( vh_mask & 0x0f ) )
{
simd_4way_init( &ctx.simd, 512 );
simd_4way_update_close( &ctx.simd, vhashA, vhashA, 512 );
}
if ( likely( vh_mask & 0xf0 ) )
{
simd_4way_init( &ctx.simd, 512 );
simd_4way_update_close( &ctx.simd, vhash, vhashB, 512 );
rintrlv_4x128_8x64( vhashB, vhashA, vhash, 512 );
}
intrlv_8x64_512( vhashA, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
hash7 );
mm512_blend_hash_8x64( vh, vhA, vhB, vh_mask );
rintrlv_4x128_8x64( vhashB, vhashA, vhash, 512 );
mm512_blend_hash_8x64( vh, vhC, vhB, vh_mask );
rintrlv_8x64_8x32( vhashA, vhash, 512 );
shabal512_8way_init( &ctx.shabal );
@@ -641,6 +767,20 @@ extern void hmq1725_8way_hash(void *state, const void *input)
}
mm512_blend_hash_8x64( vh, vhA, vhB, vh_mask );
#if defined(__VAES__)
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 );
rintrlv_4x128_8x64( vhash, vhashA, vhashB, 512 );
#else
dintrlv_8x64_512( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
vhash );
@@ -664,6 +804,8 @@ extern void hmq1725_8way_hash(void *state, const void *input)
intrlv_8x64_512( vhash, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
hash7 );
#endif
sha512_8way_init( &ctx.sha512 );
sha512_8way_update( &ctx.sha512, vhash, 64 );
sha512_8way_close( &ctx.sha512, vhash );
@@ -830,7 +972,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
__m256i* vhB = (__m256i*)vhashB;
bmw512_4way_init( &ctx.bmw );
bmw512_4way( &ctx.bmw, input, 80 );
bmw512_4way_update( &ctx.bmw, input, 80 );
bmw512_4way_close( &ctx.bmw, vhash );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
@@ -889,18 +1031,18 @@ extern void hmq1725_4way_hash(void *state, const void *input)
if ( mm256_anybits1( vh_mask ) )
{
skein512_4way_init( &ctx.skein );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_update( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhashB );
}
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_update( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_update( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
// second fork, A = blake parallel, B= bmw parallel.
@@ -911,14 +1053,14 @@ extern void hmq1725_4way_hash(void *state, const void *input)
if ( mm256_anybits0( vh_mask ) )
{
blake512_4way_init( &ctx.blake );
blake512_4way( &ctx.blake, vhash, 64 );
blake512_4way_update( &ctx.blake, vhash, 64 );
blake512_4way_close( &ctx.blake, vhashA );
}
if ( mm256_anybits1( vh_mask ) )
{
bmw512_4way_init( &ctx.bmw );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_update( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhashB );
}
@@ -962,14 +1104,14 @@ extern void hmq1725_4way_hash(void *state, const void *input)
if ( mm256_anybits0( vh_mask ) )
{
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_update( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhashA );
}
if ( mm256_anybits1( vh_mask ) )
{
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_update( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhashB );
}
@@ -990,7 +1132,6 @@ extern void hmq1725_4way_hash(void *state, const void *input)
sph_shavite512 ( &ctx.shavite, hash3, 64 );
sph_shavite512_close( &ctx.shavite, hash3 );
intrlv_2x128_512( vhashA, hash0, hash1 );
intrlv_2x128_512( vhashB, hash2, hash3 );
@@ -1042,7 +1183,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
if ( mm256_anybits1( vh_mask ) )
{
haval256_5_4way_init( &ctx.haval );
haval256_5_4way( &ctx.haval, vhash, 64 );
haval256_5_4way_update( &ctx.haval, vhash, 64 );
haval256_5_4way_close( &ctx.haval, vhash );
memset( &vhash[8<<2], 0, 32<<2 );
rintrlv_4x32_4x64( vhashB, vhash, 512 );
@@ -1068,7 +1209,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
blake512_4way_init( &ctx.blake );
blake512_4way( &ctx.blake, vhash, 64 );
blake512_4way_update( &ctx.blake, vhash, 64 );
blake512_4way_close( &ctx.blake, vhash );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
@@ -1130,7 +1271,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way_init( &ctx.hamsi );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_update( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
@@ -1214,7 +1355,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way_init( &ctx.shabal );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_update( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
@@ -1269,7 +1410,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
if ( mm256_anybits1( vh_mask ) )
{
sha512_4way_init( &ctx.sha512 );
sha512_4way( &ctx.sha512, vhash, 64 );
sha512_4way_update( &ctx.sha512, vhash, 64 );
sha512_4way_close( &ctx.sha512, vhashB );
}
@@ -1289,7 +1430,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
sha512_4way_init( &ctx.sha512 );
sha512_4way( &ctx.sha512, vhash, 64 );
sha512_4way_update( &ctx.sha512, vhash, 64 );
sha512_4way_close( &ctx.sha512, vhash );
// A = haval parallel, B = Whirlpool serial
@@ -1305,7 +1446,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
if ( mm256_anybits0( vh_mask ) )
{
haval256_5_4way_init( &ctx.haval );
haval256_5_4way( &ctx.haval, vhash, 64 );
haval256_5_4way_update( &ctx.haval, vhash, 64 );
haval256_5_4way_close( &ctx.haval, vhash );
memset( &vhash[8<<2], 0, 32<<2 );
rintrlv_4x32_4x64( vhashA, vhash, 512 );
@@ -1341,7 +1482,7 @@ extern void hmq1725_4way_hash(void *state, const void *input)
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
bmw512_4way_init( &ctx.bmw );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_update( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
memcpy(state, vhash, 32<<2 );

2
algo/quark/hmq1725-gate.c
View File

@@ -13,7 +13,7 @@ bool register_hmq1725_algo( algo_gate_t* gate )
gate->scanhash = (void*)&scanhash_hmq1725;
gate->hash = (void*)&hmq1725hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
opt_target_factor = 65536.0;
return true;
};

77
algo/quark/quark-4way.c
View File

@@ -9,16 +9,23 @@
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/groestl/aes_ni/hash-groestl.h"
#if defined(__VAES__)
#include "algo/groestl/groestl512-hash-4way.h"
#endif
#if defined (QUARK_8WAY)
typedef struct {
blake512_8way_context blake;
bmw512_8way_context bmw;
hashState_groestl groestl;
jh512_8way_context jh;
skein512_8way_context skein;
keccak512_8way_context keccak;
#if defined(__VAES__)
groestl512_4way_context groestl;
#else
hashState_groestl groestl;
#endif
} quark_8way_ctx_holder;
quark_8way_ctx_holder quark_8way_ctx __attribute__ ((aligned (128)));
@@ -27,10 +34,14 @@ void init_quark_8way_ctx()
{
blake512_8way_init( &quark_8way_ctx.blake );
bmw512_8way_init( &quark_8way_ctx.bmw );
init_groestl( &quark_8way_ctx.groestl, 64 );
skein512_8way_init( &quark_8way_ctx.skein );
jh512_8way_init( &quark_8way_ctx.jh );
keccak512_8way_init( &quark_8way_ctx.keccak );
#if defined(__VAES__)
groestl512_4way_init( &quark_8way_ctx.groestl, 64 );
#else
init_groestl( &quark_8way_ctx.groestl, 64 );
#endif
}
void quark_8way_hash( void *state, const void *input )
@@ -38,6 +49,7 @@ void quark_8way_hash( void *state, const void *input )
uint64_t vhash[8*8] __attribute__ ((aligned (128)));
uint64_t vhashA[8*8] __attribute__ ((aligned (64)));
uint64_t vhashB[8*8] __attribute__ ((aligned (64)));
uint64_t vhashC[8*8] __attribute__ ((aligned (64)));
uint64_t hash0[8] __attribute__ ((aligned (64)));
uint64_t hash1[8] __attribute__ ((aligned (64)));
uint64_t hash2[8] __attribute__ ((aligned (64)));
@@ -49,6 +61,7 @@ void quark_8way_hash( void *state, const void *input )
__m512i* vh = (__m512i*)vhash;
__m512i* vhA = (__m512i*)vhashA;
__m512i* vhB = (__m512i*)vhashB;
__m512i* vhC = (__m512i*)vhashC;
__mmask8 vh_mask;
quark_8way_ctx_holder ctx;
const uint32_t mask = 8;
@@ -66,6 +79,25 @@ void quark_8way_hash( void *state, const void *input )
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], bit3_mask ),
zero );
#if defined(__VAES__)
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
if ( ( vh_mask & 0x0f ) != 0x0f )
{
groestl512_4way_init( &ctx.groestl, 64 );
groestl512_4way_update_close( &ctx.groestl, vhashA, vhashA, 512 );
}
if ( ( vh_mask & 0xf0 ) != 0xf0 )
{
groestl512_4way_init( &ctx.groestl, 64 );
groestl512_4way_update_close( &ctx.groestl, vhashB, vhashB, 512 );
}
rintrlv_4x128_8x64( vhashC, vhashA, vhashB, 512 );
#else
dintrlv_8x64( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
vhash, 512 );
@@ -117,16 +149,31 @@ void quark_8way_hash( void *state, const void *input )
(char*)hash7, 512 );
}
intrlv_8x64( vhashA, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
intrlv_8x64( vhashC, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
hash7, 512 );
#endif
if ( vh_mask & 0xff )
{
skein512_8way_update( &ctx.skein, vhash, 64 );
skein512_8way_close( &ctx.skein, vhashB );
}
mm512_blend_hash_8x64( vh, vhA, vhB, vh_mask );
mm512_blend_hash_8x64( vh, vhC, vhB, vh_mask );
#if defined(__VAES__)
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 );
rintrlv_4x128_8x64( vhash, vhashA, vhashB, 512 );
#else
dintrlv_8x64( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
vhash, 512 );
@@ -151,6 +198,8 @@ void quark_8way_hash( void *state, const void *input )
intrlv_8x64( vhash, hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
512 );
#endif
jh512_8way_update( &ctx.jh, vhash, 64 );
jh512_8way_close( &ctx.jh, vhash );
@@ -289,10 +338,10 @@ void quark_4way_hash( void *state, const void *input )
memcpy( &ctx, &quark_4way_ctx, sizeof(quark_4way_ctx) );
blake512_4way( &ctx.blake, input, 80 );
blake512_4way_update( &ctx.blake, input, 80 );
blake512_4way_close( &ctx.blake, vhash );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_update( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ), zero );
@@ -327,7 +376,7 @@ void quark_4way_hash( void *state, const void *input )
if ( mm256_anybits1( vh_mask ) )
{
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_update( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhashB );
}
@@ -346,7 +395,7 @@ void quark_4way_hash( void *state, const void *input )
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_update( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ), zero );
@@ -354,24 +403,24 @@ void quark_4way_hash( void *state, const void *input )
if ( mm256_anybits0( vh_mask ) )
{
blake512_4way_init( &ctx.blake );
blake512_4way( &ctx.blake, vhash, 64 );
blake512_4way_update( &ctx.blake, vhash, 64 );
blake512_4way_close( &ctx.blake, vhashA );
}
if ( mm256_anybits1( vh_mask ) )
{
bmw512_4way_init( &ctx.bmw );
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_update( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhashB );
}
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_update( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
skein512_4way_init( &ctx.skein );
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_update( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
vh_mask = _mm256_cmpeq_epi64( _mm256_and_si256( vh[0], bit3_mask ), zero );
@@ -379,14 +428,14 @@ void quark_4way_hash( void *state, const void *input )
if ( mm256_anybits0( vh_mask ) )
{
keccak512_4way_init( &ctx.keccak );
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_update( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhashA );
}
if ( mm256_anybits1( vh_mask ) )
{
jh512_4way_init( &ctx.jh );
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_update( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhashB );
}

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

@@ -15,7 +15,7 @@ bool register_quark_algo( algo_gate_t* gate )
gate->scanhash = (void*)&scanhash_quark;
gate->hash = (void*)&quark_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
return true;
};