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

v3.20.3

Browse Source
This commit is contained in:
Jay D Dee
2022-10-21 23:12:18 -04:00
parent 58030e2788
commit bd84f199fe
35 changed files with 983 additions and 4938 deletions
Download Patch File Download Diff File Expand all files Collapse all files

61
algo/quark/anime-4way.c
View File

@@ -15,7 +15,8 @@
#if defined (ANIME_8WAY)
typedef struct {
union _anime_8way_context_overlay
{
blake512_8way_context blake;
bmw512_8way_context bmw;
#if defined(__VAES__)
@@ -26,23 +27,9 @@ typedef struct {
jh512_8way_context jh;
skein512_8way_context skein;
keccak512_8way_context keccak;
} anime_8way_ctx_holder;
} __attribute__ ((aligned (64)));
anime_8way_ctx_holder anime_8way_ctx __attribute__ ((aligned (64)));
void init_anime_8way_ctx()
{
blake512_8way_init( &anime_8way_ctx.blake );
bmw512_8way_init( &anime_8way_ctx.bmw );
#if defined(__VAES__)
groestl512_4way_init( &anime_8way_ctx.groestl, 64 );
#else
init_groestl( &anime_8way_ctx.groestl, 64 );
#endif
skein512_8way_init( &anime_8way_ctx.skein );
jh512_8way_init( &anime_8way_ctx.jh );
keccak512_8way_init( &anime_8way_ctx.keccak );
}
typedef union _anime_8way_context_overlay anime_8way_context_overlay;
void anime_8way_hash( void *state, const void *input )
{
@@ -65,17 +52,14 @@ void anime_8way_hash( void *state, const void *input )
__m512i* vhB = (__m512i*)vhashB;
__m512i* vhC = (__m512i*)vhashC;
const __m512i bit3_mask = m512_const1_64( 8 );
const __m512i zero = _mm512_setzero_si512();
__mmask8 vh_mask;
anime_8way_ctx_holder ctx;
memcpy( &ctx, &anime_8way_ctx, sizeof(anime_8way_ctx) );
anime_8way_context_overlay ctx __attribute__ ((aligned (64)));
bmw512_8way_full( &ctx.bmw, vhash, input, 80 );
blake512_8way_full( &ctx.blake, vhash, vhash, 64 );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], bit3_mask ),
zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], bit3_mask );
#if defined(__VAES__)
@@ -152,8 +136,7 @@ void anime_8way_hash( void *state, const void *input )
jh512_8way_update( &ctx.jh, vhash, 64 );
jh512_8way_close( &ctx.jh, vhash );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], bit3_mask ),
zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], bit3_mask );
if ( ( vh_mask & 0xff ) != 0xff )
blake512_8way_full( &ctx.blake, vhashA, vhash, 64 );
@@ -168,8 +151,7 @@ void anime_8way_hash( void *state, const void *input )
skein512_8way_full( &ctx.skein, vhash, vhash, 64 );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], bit3_mask ),
zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], bit3_mask );
if ( ( vh_mask & 0xff ) != 0xff )
{
@@ -237,14 +219,20 @@ int scanhash_anime_8way( struct work *work, uint32_t max_nonce,
#elif defined (ANIME_4WAY)
typedef struct {
union _anime_4way_context_overlay
{
blake512_4way_context blake;
bmw512_4way_context bmw;
hashState_groestl groestl;
jh512_4way_context jh;
skein512_4way_context skein;
keccak512_4way_context keccak;
} anime_4way_ctx_holder;
#if defined(__VAES__)
groestl512_2way_context groestl2;
#endif
} __attribute__ ((aligned (64)));
typedef union _anime_4way_context_overlay anime_4way_context_overlay;
void anime_4way_hash( void *state, const void *input )
{
@@ -262,7 +250,7 @@ void anime_4way_hash( void *state, const void *input )
int h_mask;
const __m256i bit3_mask = m256_const1_64( 8 );
const __m256i zero = _mm256_setzero_si256();
anime_4way_ctx_holder ctx;
anime_4way_context_overlay ctx __attribute__ ((aligned (64)));
bmw512_4way_init( &ctx.bmw );
bmw512_4way_update( &ctx.bmw, input, 80 );
@@ -293,7 +281,18 @@ void anime_4way_hash( void *state, const void *input )
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
#if defined(__VAES__)
rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
groestl512_2way_full( &ctx.groestl2, vhashA, vhashA, 64 );
groestl512_2way_full( &ctx.groestl2, vhashB, vhashB, 64 );
rintrlv_2x128_4x64( vhash, vhashA, vhashB, 512 );
#else
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
groestl512_full( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
groestl512_full( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
@@ -302,6 +301,8 @@ void anime_4way_hash( void *state, const void *input )
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
#endif
jh512_4way_init( &ctx.jh );
jh512_4way_update( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );

69
algo/quark/hmq1725-4way.c
View File

@@ -13,6 +13,7 @@
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/shavite/shavite-hash-2way.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/hamsi/hamsi-hash-4way.h"
@@ -98,8 +99,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
intrlv_8x64_512( vhash, hash0, hash1, hash2, hash3,
hash4, hash5, hash6, hash7 );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], vmask ),
m512_zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], vmask );
// A
#if defined(__VAES__)
@@ -154,8 +154,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
keccak512_8way_update( &ctx.keccak, vhash, 64 );
keccak512_8way_close( &ctx.keccak, vhash );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], vmask ),
m512_zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], vmask );
// A
if ( ( vh_mask & 0xff ) != 0xff )
@@ -174,8 +173,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
cube_4way_full( &ctx.cube, vhashB, 512, vhashB, 64 );
rintrlv_4x128_8x64( vhash, vhashA, vhashB, 512 );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], vmask ),
m512_zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], vmask );
if ( likely( ( vh_mask & 0xff ) != 0xff ) )
{
@@ -223,8 +221,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
simd512_4way_full( &ctx.simd, vhashB, vhashB, 64 );
rintrlv_4x128_8x64( vhash, vhashA, vhashB, 512 );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], vmask ),
m512_zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], vmask );
dintrlv_8x64_512( hash0, hash1, hash2, hash3,
hash4, hash5, hash6, hash7, vhash );
// 4x32 for haval
@@ -302,8 +299,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
blake512_8way_full( &ctx.blake, vhash, vhash, 64 );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], vmask ),
m512_zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], vmask );
// A
#if defined(__VAES__)
@@ -374,8 +370,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
intrlv_8x64_512( vhash, hash0, hash1, hash2, hash3,
hash4, hash5, hash6, hash7 );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], vmask ),
m512_zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], vmask );
// A
#if defined(__VAES__)
@@ -455,8 +450,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
intrlv_8x64_512( vhash, hash0, hash1, hash2, hash3,
hash4, hash5, hash6, hash7 );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], vmask ),
m512_zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], vmask );
if ( hash0[0] & mask )
fugue512_full( &ctx.fugue, hash0, hash0, 64 );
@@ -520,8 +514,7 @@ extern void hmq1725_8way_hash(void *state, const void *input)
sha512_8way_update( &ctx.sha512, vhash, 64 );
sha512_8way_close( &ctx.sha512, vhash );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], vmask ),
m512_zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], vmask );
dintrlv_8x64_512( hash0, hash1, hash2, hash3,
hash4, hash5, hash6, hash7, vhash );
@@ -625,6 +618,7 @@ union _hmq1725_4way_context_overlay
cube_2way_context cube2;
sph_shavite512_context shavite;
hashState_sd sd;
shavite512_2way_context shavite2;
simd_2way_context simd;
hashState_echo echo;
hamsi512_4way_context hamsi;
@@ -633,6 +627,10 @@ union _hmq1725_4way_context_overlay
sph_whirlpool_context whirlpool;
sha512_4way_context sha512;
haval256_5_4way_context haval;
#if defined(__VAES__)
groestl512_2way_context groestl2;
echo_2way_context echo2;
#endif
} __attribute__ ((aligned (64)));
typedef union _hmq1725_4way_context_overlay hmq1725_4way_context_overlay;
@@ -750,15 +748,10 @@ extern void hmq1725_4way_hash(void *state, const void *input)
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
shavite512_full( &ctx.shavite, hash0, hash0, 64 );
shavite512_full( &ctx.shavite, hash1, hash1, 64 );
shavite512_full( &ctx.shavite, hash2, hash2, 64 );
shavite512_full( &ctx.shavite, hash3, hash3, 64 );
intrlv_2x128_512( vhashA, hash0, hash1 );
intrlv_2x128_512( vhashB, hash2, hash3 );
shavite512_2way_full( &ctx.shavite2, vhashA, vhashA, 64 );
shavite512_2way_full( &ctx.shavite2, vhashB, vhashB, 64 );
simd512_2way_full( &ctx.simd, vhashA, vhashA, 64 );
simd512_2way_full( &ctx.simd, vhashB, vhashB, 64 );
@@ -795,6 +788,17 @@ extern void hmq1725_4way_hash(void *state, const void *input)
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
#if defined(__VAES__)
rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
echo_2way_full( &ctx.echo2, vhashA, 512, vhashA, 64 );
echo_2way_full( &ctx.echo2, vhashB, 512, vhashB, 64 );
rintrlv_2x128_4x64( vhash, vhashA, vhashB, 512 );
#else
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
echo_full( &ctx.echo, (BitSequence *)hash0, 512,
@@ -807,7 +811,9 @@ extern void hmq1725_4way_hash(void *state, const void *input)
(const BitSequence *)hash3, 64 );
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
#endif
blake512_4way_full( &ctx.blake, vhash, vhash, 64 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
@@ -939,6 +945,17 @@ extern void hmq1725_4way_hash(void *state, const void *input)
mm256_blend_hash_4x64( vh, vhA, vhB, vh_mask );
#if defined(__VAES__)
rintrlv_4x64_2x128( vhashA, vhashB, vhash, 512 );
groestl512_2way_full( &ctx.groestl2, vhashA, vhashA, 64 );
groestl512_2way_full( &ctx.groestl2, vhashB, vhashB, 64 );
rintrlv_2x128_4x64( vhash, vhashA, vhashB, 512 );
#else
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
groestl512_full( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -948,6 +965,8 @@ extern void hmq1725_4way_hash(void *state, const void *input)
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
#endif
sha512_4way_init( &ctx.sha512 );
sha512_4way_update( &ctx.sha512, vhash, 64 );
sha512_4way_close( &ctx.sha512, vhash );

11
algo/quark/quark-4way.c
View File

@@ -68,7 +68,6 @@ void quark_8way_hash( void *state, const void *input )
quark_8way_ctx_holder ctx;
const uint32_t mask = 8;
const __m512i bit3_mask = m512_const1_64( mask );
const __m512i zero = _mm512_setzero_si512();
memcpy( &ctx, &quark_8way_ctx, sizeof(quark_8way_ctx) );
@@ -76,9 +75,7 @@ void quark_8way_hash( void *state, const void *input )
bmw512_8way_full( &ctx.bmw, vhash, vhash, 64 );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], bit3_mask ),
zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], bit3_mask );
#if defined(__VAES__)
@@ -154,8 +151,7 @@ void quark_8way_hash( void *state, const void *input )
jh512_8way_update( &ctx.jh, vhash, 64 );
jh512_8way_close( &ctx.jh, vhash );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], bit3_mask ),
zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], bit3_mask );
if ( ( vh_mask & 0xff ) != 0xff )
blake512_8way_full( &ctx.blake, vhashA, vhash, 64 );
@@ -169,8 +165,7 @@ void quark_8way_hash( void *state, const void *input )
skein512_8way_full( &ctx.skein, vhash, vhash, 64 );
vh_mask = _mm512_cmpeq_epi64_mask( _mm512_and_si512( vh[0], bit3_mask ),
zero );
vh_mask = _mm512_testn_epi64_mask( vh[0], bit3_mask );
if ( ( vh_mask & 0xff ) != 0xff )
{