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

30
algo/x13/phi1612-4way.c
View File

@@ -10,6 +10,9 @@
#include "algo/fugue/sph_fugue.h"
#include "algo/gost/sph_gost.h"
#include "algo/echo/aes_ni/hash_api.h"
#if defined(__VAES__)
#include "algo/echo/echo-hash-4way.h"
#endif
#if defined(PHI1612_8WAY)
@@ -19,7 +22,11 @@ typedef struct {
cube_4way_context cube;
sph_fugue512_context fugue;
sph_gost512_context gost;
#if defined(__VAES__)
echo_4way_context echo;
#else
hashState_echo echo;
#endif
} phi1612_8way_ctx_holder;
phi1612_8way_ctx_holder phi1612_8way_ctx __attribute__ ((aligned (64)));
@@ -31,7 +38,11 @@ void init_phi1612_8way_ctx()
cube_4way_init( &phi1612_8way_ctx.cube, 512, 16, 32 );
sph_fugue512_init( &phi1612_8way_ctx.fugue );
sph_gost512_init( &phi1612_8way_ctx.gost );
#if defined(__VAES__)
echo_4way_init( &phi1612_8way_ctx.echo, 512 );
#else
init_echo( &phi1612_8way_ctx.echo, 512 );
#endif
};
void phi1612_8way_hash( void *state, const void *input )
@@ -118,6 +129,19 @@ void phi1612_8way_hash( void *state, const void *input )
sph_gost512_close( &ctx.gost, hash7 );
// Echo
#if defined(__VAES__)
intrlv_4x128_512( vhash, hash0, hash1, hash2, hash3 );
echo_4way_update_close( &ctx.echo, vhash, vhash, 512 );
dintrlv_4x128_512( hash0, hash1, hash2, hash3, vhash );
intrlv_4x128_512( vhash, hash4, hash5, hash6, hash7 );
echo_4way_init( &ctx.echo, 512 );
echo_4way_update_close( &ctx.echo, vhash, vhash, 512 );
dintrlv_4x128_512( hash4, hash5, hash6, hash7, vhash );
#else
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
init_echo( &ctx.echo, 512 );
@@ -142,6 +166,8 @@ void phi1612_8way_hash( void *state, const void *input )
update_final_echo( &ctx.echo, (BitSequence *)hash7,
(const BitSequence *) hash7, 512 );
#endif
memcpy( state, hash0, 32 );
memcpy( state+ 32, hash1, 32 );
memcpy( state+ 64, hash2, 32 );
@@ -225,11 +251,11 @@ void phi1612_4way_hash( void *state, const void *input )
memcpy( &ctx, &phi1612_4way_ctx, sizeof(phi1612_4way_ctx) );
// Skein parallel 4way
skein512_4way( &ctx.skein, input, 80 );
skein512_4way_update( &ctx.skein, input, 80 );
skein512_4way_close( &ctx.skein, vhash );
// JH
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_update( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
// Serial to the end

2
algo/x13/phi1612-gate.c
View File

@@ -15,7 +15,7 @@ bool register_phi1612_algo( algo_gate_t* gate )
gate->scanhash = (void*)&scanhash_phi1612;
gate->hash = (void*)&phi1612_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;
};

2
algo/x13/skunk-4way.c
View File

@@ -168,7 +168,7 @@ void skunk_4way_hash( void *output, const void *input )
skunk_4way_ctx_holder ctx __attribute__ ((aligned (64)));
memcpy( &ctx, &skunk_4way_ctx, sizeof(skunk_4way_ctx) );
skein512_4way( &ctx.skein, input, 80 );
skein512_4way_update( &ctx.skein, input, 80 );
skein512_4way_close( &ctx.skein, vhash );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );

114
algo/x13/x13-4way.c
View File

@@ -17,23 +17,34 @@
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/hamsi/hamsi-hash-4way.h"
#include "algo/fugue/sph_fugue.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(X13_8WAY)
typedef struct {
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;
#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
} x13_8way_ctx_holder;
x13_8way_ctx_holder x13_8way_ctx;
@@ -42,24 +53,30 @@ void init_x13_8way_ctx()
{
blake512_8way_init( &x13_8way_ctx.blake );
bmw512_8way_init( &x13_8way_ctx.bmw );
init_groestl( &x13_8way_ctx.groestl, 64 );
skein512_8way_init( &x13_8way_ctx.skein );
jh512_8way_init( &x13_8way_ctx.jh );
keccak512_8way_init( &x13_8way_ctx.keccak );
luffa_4way_init( &x13_8way_ctx.luffa, 512 );
cube_4way_init( &x13_8way_ctx.cube, 512, 16, 32 );
sph_shavite512_init( &x13_8way_ctx.shavite );
simd_4way_init( &x13_8way_ctx.simd, 512 );
init_echo( &x13_8way_ctx.echo, 512 );
hamsi512_8way_init( &x13_8way_ctx.hamsi );
sph_fugue512_init( &x13_8way_ctx.fugue );
#if defined(__VAES__)
groestl512_4way_init( &x13_8way_ctx.groestl, 64 );
shavite512_4way_init( &x13_8way_ctx.shavite );
echo_4way_init( &x13_8way_ctx.echo, 512 );
#else
init_groestl( &x13_8way_ctx.groestl, 64 );
sph_shavite512_init( &x13_8way_ctx.shavite );
init_echo( &x13_8way_ctx.echo, 512 );
#endif
}
void x13_8way_hash( void *state, const void *input )
{
uint64_t vhash[8*8] __attribute__ ((aligned (128)));
uint64_t vhash0[4*8] __attribute__ ((aligned (64)));
uint64_t vhash1[4*8] __attribute__ ((aligned (64)));
uint64_t vhashA[4*8] __attribute__ ((aligned (64)));
uint64_t vhashB[4*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)));
@@ -76,6 +93,19 @@ void x13_8way_hash( void *state, const void *input )
bmw512_8way_update( &ctx.bmw, vhash, 64 );
bmw512_8way_close( &ctx.bmw, vhash );
#if defined(__VAES__)
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
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 );
@@ -104,6 +134,9 @@ void x13_8way_hash( void *state, const void *input )
intrlv_8x64_512( vhash, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
hash7 );
#endif
skein512_8way_update( &ctx.skein, vhash, 64 );
skein512_8way_close( &ctx.skein, vhash );
@@ -112,21 +145,27 @@ void x13_8way_hash( void *state, const void *input )
keccak512_8way_update( &ctx.keccak, vhash, 64 );
keccak512_8way_close( &ctx.keccak, vhash );
dintrlv_8x64_512( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
vhash );
rintrlv_8x64_4x128( vhash0, vhash1, vhash, 512 );
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
luffa_4way_update_close( &ctx.luffa, vhash0, vhash0, 64 );
luffa_4way_update_close( &ctx.luffa, vhashA, vhashA, 64 );
luffa_4way_init( &ctx.luffa, 512 );
luffa_4way_update_close( &ctx.luffa, vhash1, vhash1, 64 );
luffa_4way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
cube_4way_update_close( &ctx.cube, vhash0, vhash0, 64 );
cube_4way_update_close( &ctx.cube, vhashA, vhashA, 64 );
cube_4way_init( &ctx.cube, 512, 16, 32 );
cube_4way_update_close( &ctx.cube, vhash1, vhash1, 64 );
cube_4way_update_close( &ctx.cube, vhashB, vhashB, 64 );
dintrlv_4x128_512( hash0, hash1, hash2, hash3, vhash0 );
dintrlv_4x128_512( hash4, hash5, hash6, hash7, vhash1 );
#if defined(__VAES__)
shavite512_4way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_4way_init( &ctx.shavite );
shavite512_4way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
#else
dintrlv_4x128_512( hash0, hash1, hash2, hash3, vhashA );
dintrlv_4x128_512( hash4, hash5, hash6, hash7, vhashB );
sph_shavite512( &ctx.shavite, hash0, 64 );
sph_shavite512_close( &ctx.shavite, hash0 );
@@ -159,13 +198,27 @@ void x13_8way_hash( void *state, const void *input )
sph_shavite512( &ctx.shavite, hash7, 64 );
sph_shavite512_close( &ctx.shavite, hash7 );
intrlv_4x128_512( vhash, hash0, hash1, hash2, hash3 );
simd_4way_update_close( &ctx.simd, vhash, vhash, 512 );
dintrlv_4x128_512( hash0, hash1, hash2, hash3, vhash );
intrlv_4x128_512( vhash, hash4, hash5, hash6, hash7 );
intrlv_4x128_512( vhashA, hash0, hash1, hash2, hash3 );
intrlv_4x128_512( vhashB, hash4, hash5, hash6, hash7 );
#endif
simd_4way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_4way_init( &ctx.simd, 512 );
simd_4way_update_close( &ctx.simd, vhash, vhash, 512 );
dintrlv_4x128_512( hash4, hash5, hash6, hash7, vhash );
simd_4way_update_close( &ctx.simd, vhashB, vhashB, 512 );
#if defined(__VAES__)
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_4x128_512( hash0, hash1, hash2, hash3, vhashA );
dintrlv_4x128_512( hash4, hash5, hash6, hash7, vhashB );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
@@ -193,6 +246,9 @@ void x13_8way_hash( void *state, const void *input )
intrlv_8x64_512( vhash, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
hash7 );
#endif
hamsi512_8way_update( &ctx.hamsi, vhash, 64 );
hamsi512_8way_close( &ctx.hamsi, vhash );
dintrlv_8x64_512( hash0, hash1, hash2, hash3, hash4, hash5, hash6, hash7,
@@ -321,11 +377,11 @@ void x13_4way_hash( void *state, const void *input )
memcpy( &ctx, &x13_4way_ctx, sizeof(x13_4way_ctx) );
// 1 Blake
blake512_4way( &ctx.blake, input, 80 );
blake512_4way_update( &ctx.blake, input, 80 );
blake512_4way_close( &ctx.blake, vhash );
// 2 Bmw
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_update( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
// Serial
@@ -344,15 +400,15 @@ void x13_4way_hash( void *state, const void *input )
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
// 4 Skein
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_update( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
// 5 JH
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_update( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
// 6 Keccak
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_update( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
// Serial
@@ -416,7 +472,7 @@ void x13_4way_hash( void *state, const void *input )
// 12 Hamsi parallel 4way 32 bit
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
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 );

2
algo/x13/x13-gate.c
View File

@@ -15,7 +15,7 @@ bool register_x13_algo( algo_gate_t* gate )
gate->scanhash = (void*)&scanhash_x13;
gate->hash = (void*)&x13hash;
#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;
};

389
algo/x13/x13bcd-4way.c
View File

@@ -1,7 +1,4 @@
#include "x13sm3-gate.h"
#if defined(X13SM3_4WAY)
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
@@ -13,12 +10,328 @@
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/cubehash/cube-hash-2way.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/sm3/sm3-hash-4way.h"
#include "algo/hamsi/hamsi-hash-4way.h"
#include "algo/fugue/sph_fugue.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(X13BCD_8WAY)
typedef struct {
blake512_8way_context blake;
bmw512_8way_context bmw;
skein512_8way_context skein;
jh512_8way_context jh;
keccak512_8way_context keccak;
cube_4way_context cube;
simd_4way_context simd;
sm3_8way_ctx_t sm3;
hamsi512_8way_context hamsi;
sph_fugue512_context fugue;
#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
} x13bcd_8way_ctx_holder;
x13bcd_8way_ctx_holder x13bcd_8way_ctx __attribute__ ((aligned (64)));
static __thread blake512_8way_context x13bcd_8way_ctx_mid;
void init_x13bcd_8way_ctx()
{
blake512_8way_init( &x13bcd_8way_ctx.blake );
bmw512_8way_init( &x13bcd_8way_ctx.bmw );
skein512_8way_init( &x13bcd_8way_ctx.skein );
jh512_8way_init( &x13bcd_8way_ctx.jh );
keccak512_8way_init( &x13bcd_8way_ctx.keccak );
cube_4way_init( &x13bcd_8way_ctx.cube, 512, 16, 32 );
simd_4way_init( &x13bcd_8way_ctx.simd, 512 );
sm3_8way_init( &x13bcd_8way_ctx.sm3 );
hamsi512_8way_init( &x13bcd_8way_ctx.hamsi );
sph_fugue512_init( &x13bcd_8way_ctx.fugue );
#if defined(__VAES__)
groestl512_4way_init( &x13bcd_8way_ctx.groestl, 64 );
shavite512_4way_init( &x13bcd_8way_ctx.shavite );
echo_4way_init( &x13bcd_8way_ctx.echo, 512 );
#else
init_groestl( &x13bcd_8way_ctx.groestl, 64 );
sph_shavite512_init( &x13bcd_8way_ctx.shavite );
init_echo( &x13bcd_8way_ctx.echo, 512 );
#endif
};
void x13bcd_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 hash0[8] __attribute__ ((aligned (64)));
uint64_t hash1[8] __attribute__ ((aligned (64)));
uint64_t hash2[8] __attribute__ ((aligned (64)));
uint64_t hash3[8] __attribute__ ((aligned (64)));
uint64_t hash4[8] __attribute__ ((aligned (64)));
uint64_t hash5[8] __attribute__ ((aligned (64)));
uint64_t hash6[8] __attribute__ ((aligned (64)));
uint64_t hash7[8] __attribute__ ((aligned (64)));
x13bcd_8way_ctx_holder ctx;
memcpy( &ctx, &x13bcd_8way_ctx, sizeof(x13bcd_8way_ctx) );
// Blake
memcpy( &ctx.blake, &x13bcd_8way_ctx_mid, sizeof(x13bcd_8way_ctx_mid) );
blake512_8way_update( &ctx.blake, input + (64<<3), 16 );
blake512_8way_close( &ctx.blake, vhash );
// Bmw
bmw512_8way_update( &ctx.bmw, vhash, 64 );
bmw512_8way_close( &ctx.bmw, vhash );
#if defined(__VAES__)
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 512 );
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 );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash4, (char*)hash4, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash5, (char*)hash5, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash6, (char*)hash6, 512 );
reinit_groestl( &ctx.groestl );
update_and_final_groestl( &ctx.groestl, (char*)hash7, (char*)hash7, 512 );
intrlv_8x64_512( vhash, hash0, hash1, hash2, hash3,
hash4, hash5, hash6, hash7 );
#endif
skein512_8way_update( &ctx.skein, vhash, 64 );
skein512_8way_close( &ctx.skein, vhash );
// JH
jh512_8way_update( &ctx.jh, vhash, 64 );
jh512_8way_close( &ctx.jh, vhash );
// Keccak
keccak512_8way_update( &ctx.keccak, vhash, 64 );
keccak512_8way_close( &ctx.keccak, vhash );
// SM3 parallel 32 bit
rintrlv_8x64_8x32( vhashA, vhash, 512 );
memset( vhash, 0, sizeof vhash );
sm3_8way_update( &ctx.sm3, vhashA, 64 );
sm3_8way_close( &ctx.sm3, vhash );
rintrlv_8x32_4x128( vhashA, vhashB, vhash, 512 );
// Cube
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 );
#if defined(__VAES__)
shavite512_4way_update_close( &ctx.shavite, vhashA, vhashA, 64 );
shavite512_4way_init( &ctx.shavite );
shavite512_4way_update_close( &ctx.shavite, vhashB, vhashB, 64 );
#else
dintrlv_4x128_512( hash0, hash1, hash2, hash3, vhashA );
dintrlv_4x128_512( hash4, hash5, hash6, hash7, vhashB );
sph_shavite512( &ctx.shavite, hash0, 64 );
sph_shavite512_close( &ctx.shavite, hash0 );
memcpy( &ctx.shavite, &x13bcd_8way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash1, 64 );
sph_shavite512_close( &ctx.shavite, hash1 );
memcpy( &ctx.shavite, &x13bcd_8way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash2, 64 );
sph_shavite512_close( &ctx.shavite, hash2 );
memcpy( &ctx.shavite, &x13bcd_8way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash3, 64 );
sph_shavite512_close( &ctx.shavite, hash3 );
memcpy( &ctx.shavite, &x13bcd_8way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash4, 64 );
sph_shavite512_close( &ctx.shavite, hash4 );
memcpy( &ctx.shavite, &x13bcd_8way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash5, 64 );
sph_shavite512_close( &ctx.shavite, hash5 );
memcpy( &ctx.shavite, &x13bcd_8way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash6, 64 );
sph_shavite512_close( &ctx.shavite, hash6 );
memcpy( &ctx.shavite, &x13bcd_8way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash7, 64 );
sph_shavite512_close( &ctx.shavite, hash7 );
intrlv_4x128_512( vhashA, hash0, hash1, hash2, hash3 );
intrlv_4x128_512( vhashB, hash4, hash5, hash6, hash7 );
#endif
simd_4way_update_close( &ctx.simd, vhashA, vhashA, 512 );
simd_4way_init( &ctx.simd, 512 );
simd_4way_update_close( &ctx.simd, vhashB, vhashB, 512 );
#if defined(__VAES__)
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_4x128_512( hash0, hash1, hash2, hash3, vhashA );
dintrlv_4x128_512( hash4, hash5, hash6, hash7, vhashB );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
memcpy( &ctx.echo, &x13bcd_8way_ctx.echo, sizeof(hashState_echo) );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
memcpy( &ctx.echo, &x13bcd_8way_ctx.echo, sizeof(hashState_echo) );
update_final_echo( &ctx.echo, (BitSequence *)hash2,
(const BitSequence *) hash2, 512 );
memcpy( &ctx.echo, &x13bcd_8way_ctx.echo, sizeof(hashState_echo) );
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
memcpy( &ctx.echo, &x13bcd_8way_ctx.echo, sizeof(hashState_echo) );
update_final_echo( &ctx.echo, (BitSequence *)hash4,
(const BitSequence *) hash4, 512 );
memcpy( &ctx.echo, &x13bcd_8way_ctx.echo, sizeof(hashState_echo) );
update_final_echo( &ctx.echo, (BitSequence *)hash5,
(const BitSequence *) hash5, 512 );
memcpy( &ctx.echo, &x13bcd_8way_ctx.echo, sizeof(hashState_echo) );
update_final_echo( &ctx.echo, (BitSequence *)hash6,
(const BitSequence *) hash6, 512 );
memcpy( &ctx.echo, &x13bcd_8way_ctx.echo, sizeof(hashState_echo) );
update_final_echo( &ctx.echo, (BitSequence *)hash7,
(const BitSequence *) hash7, 512 );
intrlv_8x64_512( vhash, hash0, hash1, hash2, hash3,
hash4, hash5, hash6, hash7 );
#endif
hamsi512_8way_update( &ctx.hamsi, vhash, 64 );
hamsi512_8way_close( &ctx.hamsi, vhash );
dintrlv_8x64_512( hash0, hash1, hash2, hash3,
hash4, hash5, hash6, hash7, vhash );
// Fugue serial
sph_fugue512( &ctx.fugue, hash0, 64 );
sph_fugue512_close( &ctx.fugue, state );
memcpy( &ctx.fugue, &x13bcd_8way_ctx.fugue,
sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash1, 64 );
sph_fugue512_close( &ctx.fugue, state+32 );
memcpy( &ctx.fugue, &x13bcd_8way_ctx.fugue,
sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash2, 64 );
sph_fugue512_close( &ctx.fugue, state+64 );
memcpy( &ctx.fugue, &x13bcd_8way_ctx.fugue,
sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash3, 64 );
sph_fugue512_close( &ctx.fugue, state+96 );
memcpy( &ctx.fugue, &x13bcd_8way_ctx.fugue,
sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash4, 64 );
sph_fugue512_close( &ctx.fugue, state+128 );
memcpy( &ctx.fugue, &x13bcd_8way_ctx.fugue,
sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash5, 64 );
sph_fugue512_close( &ctx.fugue, state+160 );
memcpy( &ctx.fugue, &x13bcd_8way_ctx.fugue,
sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash6, 64 );
sph_fugue512_close( &ctx.fugue, state+192 );
memcpy( &ctx.fugue, &x13bcd_8way_ctx.fugue,
sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash7, 64 );
sph_fugue512_close( &ctx.fugue, state+224 );
}
int scanhash_x13bcd_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[24*8] __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 - 8;
__m512i *noncev = (__m512i*)vdata + 9; // aligned
int thr_id = mythr->id; // thr_id arg is deprecated
const uint32_t Htarg = ptarget[7];
mm512_bswap32_intrlv80_8x64( vdata, pdata );
blake512_8way_init( &x13bcd_8way_ctx_mid );
blake512_8way_update( &x13bcd_8way_ctx_mid, vdata, 64 );
do
{
*noncev = mm512_intrlv_blend_32( mm512_bswap_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 );
x13bcd_8way_hash( hash, vdata );
pdata[19] = n;
for ( int i = 0; i < 8; i++ )
if ( (hash+(i<<3))[7] <= Htarg )
if ( fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
{
pdata[19] = n+i;
submit_lane_solution( work, hash+(i<<3), mythr, i );
}
n += 8;
} while ( ( n < last_nonce ) && !work_restart[thr_id].restart );
*hashes_done = n - first_nonce;
return 0;
}
#elif defined(X13BCD_4WAY)
typedef struct {
blake512_4way_context blake;
@@ -68,11 +381,11 @@ void x13bcd_4way_hash( void *state, const void *input )
// Blake
memcpy( &ctx.blake, &x13bcd_ctx_mid, sizeof(x13bcd_ctx_mid) );
blake512_4way( &ctx.blake, input + (64<<2), 16 );
blake512_4way_update( &ctx.blake, input + (64<<2), 16 );
blake512_4way_close( &ctx.blake, vhash );
// Bmw
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_update( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
// Serial
@@ -91,15 +404,15 @@ void x13bcd_4way_hash( void *state, const void *input )
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
// Skein
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_update( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
// JH
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_update( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
// Keccak
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_update( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
@@ -118,7 +431,7 @@ void x13bcd_4way_hash( void *state, const void *input )
uint32_t sm3_hash3[32] __attribute__ ((aligned (32)));
memset( sm3_hash3, 0, sizeof sm3_hash3 );
sm3_4way( &ctx.sm3, vhash, 64 );
sm3_4way_update( &ctx.sm3, vhash, 64 );
sm3_4way_close( &ctx.sm3, sm3_vhash );
dintrlv_4x32( hash0, hash1, hash2, hash3, sm3_vhash, 512 );
@@ -171,20 +484,23 @@ void x13bcd_4way_hash( void *state, const void *input )
// Hamsi parallel 4x32x2
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
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 );
// Fugue serial
sph_fugue512( &ctx.fugue, hash0, 64 );
sph_fugue512_close( &ctx.fugue, hash0 );
memcpy( &ctx.fugue, &x13bcd_4way_ctx.fugue, sizeof(sph_fugue512_context) );
memcpy( &ctx.fugue, &x13bcd_4way_ctx.fugue,
sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash1, 64 );
sph_fugue512_close( &ctx.fugue, hash1 );
memcpy( &ctx.fugue, &x13bcd_4way_ctx.fugue, sizeof(sph_fugue512_context) );
memcpy( &ctx.fugue, &x13bcd_4way_ctx.fugue,
sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash2, 64 );
sph_fugue512_close( &ctx.fugue, hash2 );
memcpy( &ctx.fugue, &x13bcd_4way_ctx.fugue, sizeof(sph_fugue512_context) );
memcpy( &ctx.fugue, &x13bcd_4way_ctx.fugue,
sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash3, 64 );
sph_fugue512_close( &ctx.fugue, hash3 );
@@ -203,44 +519,33 @@ int scanhash_x13bcd_4way( struct work *work, uint32_t max_nonce,
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
int thr_id = mythr->id;
const uint32_t Htarg = ptarget[7];
uint64_t htmax[] = { 0, 0xF, 0xFF,
0xFFF, 0xFFFF, 0x10000000 };
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
0xFFFFF000, 0xFFFF0000, 0 };
mm256_bswap32_intrlv80_4x64( vdata, pdata );
blake512_4way_init( &x13bcd_ctx_mid );
blake512_4way( &x13bcd_ctx_mid, vdata, 64 );
do
{
*noncev = mm256_intrlv_blend_32( mm256_bswap_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] )
{
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 );
x13bcd_4way_hash( hash, vdata );
pdata[19] = n;
x13bcd_4way_hash( hash, vdata );
pdata[19] = n;
for ( int i = 0; i < 4; i++ )
if ( ( ( (hash+(i<<3))[7] & mask ) == 0 ) )
if ( 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;
}
*hashes_done = n - first_nonce + 1;
for ( int i = 0; i < 4; i++ )
if ( (hash+(i<<3))[7] <= Htarg )
if ( fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
{
pdata[19] = n+i;
submit_lane_solution( work, hash+(i<<3), mythr, i );
}
n += 4;
} while ( ( n < last_nonce ) && !work_restart[thr_id].restart );
*hashes_done = n - first_nonce;
return 0;
}

16
algo/x13/x13sm3-4way.c
View File

@@ -71,13 +71,11 @@ void x13sm3_4way_hash( void *state, const void *input )
// Blake
memcpy( &ctx.blake, &x13sm3_ctx_mid, sizeof(x13sm3_ctx_mid) );
blake512_4way( &ctx.blake, input + (64<<2), 16 );
// blake512_4way( &ctx.blake, input, 80 );
blake512_4way_update( &ctx.blake, input + (64<<2), 16 );
blake512_4way_close( &ctx.blake, vhash );
// Bmw
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_update( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
// Serial
@@ -96,15 +94,15 @@ void x13sm3_4way_hash( void *state, const void *input )
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
// Skein
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_update( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
// JH
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_update( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
// Keccak
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_update( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
// Serial to the end
@@ -180,13 +178,13 @@ void x13sm3_4way_hash( void *state, const void *input )
uint32_t sm3_hash3[32] __attribute__ ((aligned (32)));
memset( sm3_hash3, 0, sizeof sm3_hash3 );
sm3_4way( &ctx.sm3, vhash, 64 );
sm3_4way_update( &ctx.sm3, vhash, 64 );
sm3_4way_close( &ctx.sm3, sm3_vhash );
dintrlv_4x32( hash0, hash1, hash2, hash3, sm3_vhash, 512 );
// Hamsi parallel 4x32x2
intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
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 );

8
algo/x13/x13sm3-gate.c
View File

@@ -17,7 +17,11 @@ bool register_x13sm3_algo( algo_gate_t* gate )
bool register_x13bcd_algo( algo_gate_t* gate )
{
#if defined (X13SM3_4WAY)
#if defined (X13BCD_8WAY)
init_x13bcd_8way_ctx();
gate->scanhash = (void*)&scanhash_x13bcd_8way;
gate->hash = (void*)&x13bcd_8way_hash;
#elif defined (X13BCD_4WAY)
init_x13bcd_4way_ctx();
gate->scanhash = (void*)&scanhash_x13bcd_4way;
gate->hash = (void*)&x13bcd_4way_hash;
@@ -26,7 +30,7 @@ bool register_x13bcd_algo( algo_gate_t* gate )
gate->scanhash = (void*)&scanhash_x13bcd;
gate->hash = (void*)&x13bcd_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT;
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT | AVX512_OPT | VAES_OPT;
return true;
};

38
algo/x13/x13sm3-gate.h
View File

@@ -5,13 +5,11 @@
#include <stdint.h>
#if defined(__AVX2__) && defined(__AES__)
#define X13SM3_4WAY
#define X13SM3_4WAY 1
#endif
bool register_x13sm3_algo( algo_gate_t* gate );
bool register_x13bcd_algo( algo_gate_t* gate );
#if defined(X13SM3_4WAY)
void x13sm3_4way_hash( void *state, const void *input );
@@ -19,18 +17,39 @@ int scanhash_x13sm3_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
void init_x13sm3_4way_ctx();
void x13bcd_4way_hash( void *state, const void *input );
int scanhash_x13bcd_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
void init_x13bcd_4way_ctx();
#endif
#else
void x13sm3_hash( void *state, const void *input );
int scanhash_x13sm3( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
void init_x13sm3_ctx();
#endif
#if defined(__AVX512F__) && defined(__AVX512VL__) && defined(__AVX512DQ__) && defined(__AVX512BW__)
#define X13BCD_8WAY 1
#elif defined(__AVX2__) && defined(__AES__)
#define X13BCD_4WAY 1
#endif
bool register_x13bcd_algo( algo_gate_t* gate );
#if defined(X13BCD_8WAY)
void x13bcd_8way_hash( void *state, const void *input );
int scanhash_x13bcd_8way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
void init_x13bcd_8way_ctx();
#elif defined(X13BCD_4WAY)
void x13bcd_4way_hash( void *state, const void *input );
int scanhash_x13bcd_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
void init_x13bcd_4way_ctx();
#else
void x13bcd_hash( void *state, const void *input );
int scanhash_x13bcd( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr );
@@ -38,3 +57,4 @@ void init_x13bcd_ctx();
#endif
#endif