This commit is contained in:
Jay D Dee
2018-02-15 14:48:50 -05:00
parent e4265a6f11
commit d60a268972
57 changed files with 3469 additions and 2135 deletions

273
algo/x12/x12-4way.c Normal file
View File

@@ -0,0 +1,273 @@
#include "x12-gate.h"
#if defined(X12_4WAY)
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "algo/blake/blake-hash-4way.h"
#include "algo/bmw/bmw-hash-4way.h"
#include "algo/groestl/aes_ni/hash-groestl.h"
#include "algo/skein/skein-hash-4way.h"
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/hamsi/hamsi-hash-4way.h"
//#include "algo/fugue/sph_fugue.h"
typedef struct {
blake512_4way_context blake;
bmw512_4way_context bmw;
hashState_groestl groestl;
skein512_4way_context skein;
jh512_4way_context jh;
keccak512_4way_context keccak;
luffa_2way_context luffa;
cubehashParam cube;
sph_shavite512_context shavite;
simd_2way_context simd;
hashState_echo echo;
hamsi512_4way_context hamsi;
// sph_fugue512_context fugue;
} x12_4way_ctx_holder;
x12_4way_ctx_holder x12_4way_ctx __attribute__ ((aligned (64)));
void init_x12_4way_ctx()
{
blake512_4way_init( &x12_4way_ctx.blake );
bmw512_4way_init( &x12_4way_ctx.bmw );
init_groestl( &x12_4way_ctx.groestl, 64 );
skein512_4way_init( &x12_4way_ctx.skein );
jh512_4way_init( &x12_4way_ctx.jh );
keccak512_4way_init( &x12_4way_ctx.keccak );
luffa_2way_init( &x12_4way_ctx.luffa, 512 );
cubehashInit( &x12_4way_ctx.cube, 512, 16, 32 );
sph_shavite512_init( &x12_4way_ctx.shavite );
simd_2way_init( &x12_4way_ctx.simd, 512 );
init_echo( &x12_4way_ctx.echo, 512 );
hamsi512_4way_init( &x12_4way_ctx.hamsi );
// sph_fugue512_init( &x12_4way_ctx.fugue );
};
void x12_4way_hash( void *state, const void *input )
{
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 vhash[8*4] __attribute__ ((aligned (64)));
x12_4way_ctx_holder ctx;
memcpy( &ctx, &x12_4way_ctx, sizeof(x12_4way_ctx) );
// 1 Blake
blake512_4way( &ctx.blake, input, 80 );
blake512_4way_close( &ctx.blake, vhash );
// 2 Bmw
bmw512_4way( &ctx.bmw, vhash, 64 );
bmw512_4way_close( &ctx.bmw, vhash );
// Serial
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// 3 Groestl
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
memcpy( &ctx.groestl, &x12_4way_ctx.groestl, sizeof(hashState_groestl) );
update_and_final_groestl( &ctx.groestl, (char*)hash1, (char*)hash1, 512 );
memcpy( &ctx.groestl, &x12_4way_ctx.groestl, sizeof(hashState_groestl) );
update_and_final_groestl( &ctx.groestl, (char*)hash2, (char*)hash2, 512 );
memcpy( &ctx.groestl, &x12_4way_ctx.groestl, sizeof(hashState_groestl) );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
// Parallel 4way 64 bit
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
// 4 Skein
skein512_4way( &ctx.skein, vhash, 64 );
skein512_4way_close( &ctx.skein, vhash );
// 5 JH
jh512_4way( &ctx.jh, vhash, 64 );
jh512_4way_close( &ctx.jh, vhash );
// 6 Keccak
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
// Serial
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// 7 Luffa
mm256_interleave_2x128( vhash, hash0, hash1, 512 );
luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
mm256_interleave_2x128( vhash, hash2, hash3, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
// 8 Cubehash
cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
cubehashReinit( &ctx.cube );
cubehashUpdateDigest( &ctx.cube, (byte*)hash1, (const byte*) hash1, 64 );
cubehashReinit( &ctx.cube );
cubehashUpdateDigest( &ctx.cube, (byte*)hash2, (const byte*) hash2, 64 );
cubehashReinit( &ctx.cube );
cubehashUpdateDigest( &ctx.cube, (byte*)hash3, (const byte*) hash3, 64 );
// 9 Shavite
sph_shavite512( &ctx.shavite, hash0, 64 );
sph_shavite512_close( &ctx.shavite, hash0 );
memcpy( &ctx.shavite, &x12_4way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash1, 64 );
sph_shavite512_close( &ctx.shavite, hash1 );
memcpy( &ctx.shavite, &x12_4way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash2, 64 );
sph_shavite512_close( &ctx.shavite, hash2 );
memcpy( &ctx.shavite, &x12_4way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash3, 64 );
sph_shavite512_close( &ctx.shavite, hash3 );
// 10 Simd
mm256_interleave_2x128( vhash, hash0, hash1, 512 );
simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
mm256_interleave_2x128( vhash, hash2, hash3, 512 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
// 11 Echo
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
memcpy( &ctx.echo, &x12_4way_ctx.echo, sizeof(hashState_echo) );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
memcpy( &ctx.echo, &x12_4way_ctx.echo, sizeof(hashState_echo) );
update_final_echo( &ctx.echo, (BitSequence *)hash2,
(const BitSequence *) hash2, 512 );
memcpy( &ctx.echo, &x12_4way_ctx.echo, sizeof(hashState_echo) );
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
// 12 Hamsi parallel 4way 32 bit
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm256_deinterleave_4x64( state, state+32, state+64, state+96, vhash, 256 );
/*
// 13 Fugue serial
sph_fugue512( &ctx.fugue, hash0, 64 );
sph_fugue512_close( &ctx.fugue, hash0 );
memcpy( &ctx.fugue, &x13_4way_ctx.fugue, sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash1, 64 );
sph_fugue512_close( &ctx.fugue, hash1 );
memcpy( &ctx.fugue, &x13_4way_ctx.fugue, sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash2, 64 );
sph_fugue512_close( &ctx.fugue, hash2 );
memcpy( &ctx.fugue, &x13_4way_ctx.fugue, sizeof(sph_fugue512_context) );
sph_fugue512( &ctx.fugue, hash3, 64 );
sph_fugue512_close( &ctx.fugue, hash3 );
memcpy( state, hash0, 32 );
memcpy( state+32, hash1, 32 );
memcpy( state+64, hash2, 32 );
memcpy( state+96, hash3, 32 );
*/
}
int scanhash_x12_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t endiandata[20] __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];
uint32_t *nonces = work->nonces;
bool *found = work->nfound;
int num_found = 0;
uint32_t *noncep0 = vdata + 73; // 9*8 + 1
uint32_t *noncep1 = vdata + 75;
uint32_t *noncep2 = vdata + 77;
uint32_t *noncep3 = vdata + 79;
const uint32_t Htarg = ptarget[7];
uint64_t htmax[] = { 0, 0xF, 0xFF,
0xFFF, 0xFFFF, 0x10000000 };
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
0xFFFFF000, 0xFFFF0000, 0 };
// big endian encode 0..18 uint32_t, 64 bits at a time
swab32_array( endiandata, pdata, 20 );
uint64_t *edata = (uint64_t*)endiandata;
mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
for ( int m=0; m < 6; m++ )
if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
do
{
found[0] = found[1] = found[2] = found[3] = false;
be32enc( noncep0, n );
be32enc( noncep1, n+1 );
be32enc( noncep2, n+2 );
be32enc( noncep3, n+3 );
x12_4way_hash( hash, vdata );
pdata[19] = n;
if ( ( hash[7] & mask ) == 0 && fulltest( hash, ptarget ) )
{
found[0] = true;
num_found++;
nonces[0] = n;
work_set_target_ratio( work, hash );
}
if ( ( (hash+8)[7] & mask ) == 0 && fulltest( hash+8, ptarget ) )
{
found[1] = true;
num_found++;
nonces[1] = n+1;
work_set_target_ratio( work, hash+8 );
}
if ( ( (hash+16)[7] & mask ) == 0 && fulltest( hash+16, ptarget ) )
{
found[2] = true;
num_found++;
nonces[2] = n+2;
work_set_target_ratio( work, hash+16 );
}
if ( ( (hash+24)[7] & mask ) == 0 && fulltest( hash+24, ptarget ) )
{
found[3] = true;
num_found++;
nonces[3] = n+3;
work_set_target_ratio( work, hash+24 );
}
n += 4;
} while ( ( num_found == 0 ) && ( n < max_nonce )
&& !work_restart[thr_id].restart );
break;
}
*hashes_done = n - first_nonce + 1;
return num_found;
}
#endif

18
algo/x12/x12-gate.c Normal file
View File

@@ -0,0 +1,18 @@
#include "x12-gate.h"
bool register_x12_algo( algo_gate_t* gate )
{
#if defined (X12_4WAY)
init_x12_4way_ctx();
gate->scanhash = (void*)&scanhash_x12_4way;
gate->hash = (void*)&x12_4way_hash;
#else
init_x12_ctx();
gate->scanhash = (void*)&scanhash_x12;
gate->hash = (void*)&x12hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT;
gate->get_max64 = (void*)&get_max64_0x3ffff;
return true;
};

32
algo/x12/x12-gate.h Normal file
View File

@@ -0,0 +1,32 @@
#ifndef X12_GATE_H__
#define X12_GATE_H__ 1
#include "algo-gate-api.h"
#include <stdint.h>
#if defined(__AVX2__) && defined(__AES__)
#define X12_4WAY
#endif
bool register_x12_algo( algo_gate_t* gate );
#if defined(X12_4WAY)
void x12_4way_hash( void *state, const void *input );
int scanhash_x12_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
void init_x12_4way_ctx();
#endif
void x12hash( void *state, const void *input );
int scanhash_x12( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
void init_x12_ctx();
#endif

252
algo/x12/x12.c Normal file
View File

@@ -0,0 +1,252 @@
#include "x12-gate.h"
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "algo/groestl/sph_groestl.h"
#include "algo/blake/sph_blake.h"
#include "algo/bmw/sph_bmw.h"
#include "algo/jh/sph_jh.h"
#include "algo/keccak/sph_keccak.h"
#include "algo/skein/sph_skein.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/luffa/sph_luffa.h"
#include "algo/cubehash/sph_cubehash.h"
#include "algo/simd/sph_simd.h"
#include "algo/echo/sph_echo.h"
#include "algo/hamsi/sph_hamsi.h"
//#include "algo/fugue/sph_fugue.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/blake/sse2/blake.c"
#include "algo/bmw/sse2/bmw.c"
#include "algo/keccak/sse2/keccak.c"
#include "algo/skein/sse2/skein.c"
#include "algo/jh/sse2/jh_sse2_opt64.h"
#ifndef NO_AES_NI
#include "algo/groestl/aes_ni/hash-groestl.h"
#include "algo/echo/aes_ni/hash_api.h"
#endif
typedef struct {
#ifdef NO_AES_NI
sph_groestl512_context groestl;
sph_echo512_context echo;
#else
hashState_groestl groestl;
hashState_echo echo;
#endif
hashState_luffa luffa;
cubehashParam cubehash;
sph_shavite512_context shavite;
hashState_sd simd;
sph_hamsi512_context hamsi;
// sph_fugue512_context fugue;
} x12_ctx_holder;
x12_ctx_holder x12_ctx;
void init_x12_ctx()
{
#ifdef NO_AES_NI
sph_groestl512_init(&x12_ctx.groestl);
sph_echo512_init(&x12_ctx.echo);
#else
init_echo( &x12_ctx.echo, 512 );
init_groestl (&x12_ctx.groestl, 64 );
#endif
init_luffa( &x12_ctx.luffa, 512 );
cubehashInit( &x12_ctx.cubehash, 512, 16, 32 );
sph_shavite512_init( &x12_ctx.shavite );
init_sd( &x12_ctx.simd, 512 );
sph_hamsi512_init( &x12_ctx.hamsi );
// sph_fugue512_init( &x13_ctx.fugue );
};
void x12hash(void *output, const void *input)
{
unsigned char hash[128] __attribute__ ((aligned (32)));
#define hashB hash+64
x12_ctx_holder ctx;
memcpy( &ctx, &x12_ctx, sizeof(x12_ctx) );
// X11 algos
unsigned char hashbuf[128];
size_t hashptr;
sph_u64 hashctA;
sph_u64 hashctB;
//---blake1---
DECL_BLK;
BLK_I;
BLK_W;
BLK_C;
//---bmw2---
DECL_BMW;
BMW_I;
BMW_U;
#define M(x) sph_dec64le_aligned(data + 8 * (x))
#define H(x) (h[x])
#define dH(x) (dh[x])
BMW_C;
#undef M
#undef H
#undef dH
//---groetl----
#ifdef NO_AES_NI
sph_groestl512 (&ctx.groestl, hash, 64);
sph_groestl512_close(&ctx.groestl, hash);
#else
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)hash, 512 );
#endif
//---skein4---
DECL_SKN;
SKN_I;
SKN_U;
SKN_C;
//---jh5------
DECL_JH;
JH_H;
//---keccak6---
DECL_KEC;
KEC_I;
KEC_U;
KEC_C;
//--- luffa7
update_and_final_luffa( &ctx.luffa, (BitSequence*)hashB,
(const BitSequence*)hash, 64 );
// 8 Cube
cubehashUpdateDigest( &ctx.cubehash, (byte*) hash,
(const byte*)hashB, 64 );
// 9 Shavite
sph_shavite512( &ctx.shavite, hash, 64);
sph_shavite512_close( &ctx.shavite, hashB);
// 10 Simd
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence *)hashB, 512 );
//11---echo---
#ifdef NO_AES_NI
sph_echo512(&ctx.echo, hash, 64);
sph_echo512_close(&ctx.echo, hashB);
#else
update_final_echo ( &ctx.echo, (BitSequence *)hashB,
(const BitSequence *)hash, 512 );
#endif
// 12 Hamsi
sph_hamsi512(&ctx.hamsi, hashB, 64);
sph_hamsi512_close(&ctx.hamsi, hash);
/*
// 13 Fugue
sph_fugue512(&ctx.fugue, hash, 64);
sph_fugue512_close(&ctx.fugue, hashB);
*/
asm volatile ("emms");
memcpy(output, hashB, 32);
}
int scanhash_x12(int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done)
{
uint32_t endiandata[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] - 1;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
uint64_t htmax[] = {
0,
0xF,
0xFF,
0xFFF,
0xFFFF,
0x10000000
};
uint32_t masks[] = {
0xFFFFFFFF,
0xFFFFFFF0,
0xFFFFFF00,
0xFFFFF000,
0xFFFF0000,
0
};
// we need bigendian data...
swab32_array( endiandata, pdata, 20 );
#ifdef DEBUG_ALGO
printf("[%d] Htarg=%X\n", thr_id, Htarg);
#endif
for (int m=0; m < 6; m++) {
if (Htarg <= htmax[m]) {
uint32_t mask = masks[m];
do {
pdata[19] = ++n;
be32enc(&endiandata[19], n);
x12hash(hash64, endiandata);
#ifndef DEBUG_ALGO
if (!(hash64[7] & mask))
{
if ( fulltest(hash64, ptarget) )
{
*hashes_done = n - first_nonce + 1;
return true;
}
// else
// {
// applog(LOG_INFO, "Result does not validate on CPU!");
// }
}
#else
if (!(n % 0x1000) && !thr_id) printf(".");
if (!(hash64[7] & mask)) {
printf("[%d]",thr_id);
if (fulltest(hash64, ptarget)) {
work_set_target_ratio( work, hash );
*hashes_done = n - first_nonce + 1;
return true;
}
}
#endif
} while (n < max_nonce && !work_restart[thr_id].restart);
// see blake.c if else to understand the loop on htmax => mask
break;
}
}
*hashes_done = n - first_nonce + 1;
pdata[19] = n;
return 0;
}