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

v3.8.0

Browse Source
This commit is contained in:
Jay D Dee
2018-01-23 21:02:16 -05:00
parent a90d75b8f5
commit ad2275f74a
121 changed files with 4662 additions and 467 deletions
Download Patch File Download Diff File Expand all files Collapse all files

396
algo/x17/x16r-4way.c Normal file
View File

@@ -0,0 +1,396 @@
/**
* x16r algo implementation
*
* Implementation by tpruvot@github Jan 2018
* Optimized by JayDDee@github Jan 2018
*/
#include "x16r-gate.h"
#if defined (X16R_4WAY)
#include <stdio.h>
#include <stdlib.h>
#include <string.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/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/shavite/sph_shavite.h"
#include "algo/luffa/sse2/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/simd/sse2/nist.h"
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/hamsi/hamsi-hash-4way.h"
#include "algo/fugue/sph_fugue.h"
#include "algo/shabal/shabal-hash-4way.h"
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/sha/sha2-hash-4way.h"
static __thread uint32_t s_ntime = UINT32_MAX;
static __thread char hashOrder[X16R_HASH_FUNC_COUNT + 1] = { 0 };
typedef struct {
blake512_4way_context blake;
bmw512_4way_context bmw;
hashState_echo echo;
hashState_groestl groestl;
skein512_4way_context skein;
jh512_4way_context jh;
keccak512_4way_context keccak;
hashState_luffa luffa;
cubehashParam cube;
sph_shavite512_context shavite;
hashState_sd simd;
hamsi512_4way_context hamsi;
sph_fugue512_context fugue;
shabal512_4way_context shabal;
sph_whirlpool_context whirlpool;
sha512_4way_context sha512;
} x16r_4way_ctx_holder;
x16r_4way_ctx_holder x16r_4way_ctx __attribute__ ((aligned (64)));
// Cube needs one full init so fast reinits can be done in the hash loop.
void init_x16r_4way_ctx()
{
cubehashInit( &x16r_4way_ctx.cube, 512, 16, 32 );
};
void x16r_4way_hash( void* output, const void* input )
{
uint32_t hash0[16] __attribute__ ((aligned (64)));
uint32_t hash1[16] __attribute__ ((aligned (64)));
uint32_t hash2[16] __attribute__ ((aligned (64)));
uint32_t hash3[16] __attribute__ ((aligned (64)));
uint32_t vhash[16*4] __attribute__ ((aligned (64)));
uint32_t inp0[24] __attribute__ ((aligned (64)));
uint32_t inp1[24] __attribute__ ((aligned (64)));
uint32_t inp2[24] __attribute__ ((aligned (64)));
uint32_t inp3[24] __attribute__ ((aligned (64)));
x16r_4way_ctx_holder ctx;
void *in0 = (void*) inp0;
void *in1 = (void*) inp1;
void *in2 = (void*) inp2;
void *in3 = (void*) inp3;
int size = 80;
mm256_deinterleave_4x64( inp0, inp1, inp2, inp3, input, 640 );
if ( s_ntime == UINT32_MAX )
{
const uint8_t* tmp = (uint8_t*) inp0;
x16r_getAlgoString( &tmp[4], hashOrder );
}
// Input data is both 64 bit interleaved (input)
// and deinterleaved in inp0-3.
// If First function uses 64 bit data it is not required to interleave inp
// first. It may use the inerleaved data dmost convenient, ie 4way 64 bit.
// All other functions assume data is deinterleaved in hash0-3
// All functions must exit with data deinterleaved in hash0-3.
// Alias in0-3 points to either inp0-3 or hash0-3 according to
// its hashOrder position. Size is also set accordingly.
for ( int i = 0; i < 16; i++ )
{
const char elem = hashOrder[i];
const uint8_t algo = elem >= 'A' ? elem - 'A' + 10 : elem - '0';
switch ( algo )
{
case BLAKE:
blake512_4way_init( &ctx.blake );
if ( i == 0 )
blake512_4way( &ctx.blake, input, size );
else
{
mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
blake512_4way( &ctx.blake, vhash, size );
}
blake512_4way_close( &ctx.blake, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
break;
case BMW:
bmw512_4way_init( &ctx.bmw );
if ( i == 0 )
bmw512_4way( &ctx.bmw, input, size );
else
{
mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
bmw512_4way( &ctx.bmw, vhash, size );
}
bmw512_4way_close( &ctx.bmw, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
break;
case GROESTL:
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash0,
(const char*)in0, size<<3 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash1,
(const char*)in1, size<<3 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash2,
(const char*)in2, size<<3 );
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash3,
(const char*)in3, size<<3 );
break;
case SKEIN:
skein512_4way_init( &ctx.skein );
if ( i == 0 )
skein512_4way( &ctx.skein, input, size );
else
{
mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
skein512_4way( &ctx.skein, vhash, size );
}
skein512_4way_close( &ctx.skein, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
break;
case JH:
jh512_4way_init( &ctx.jh );
if ( i == 0 )
jh512_4way( &ctx.jh, input, size );
else
{
mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
jh512_4way( &ctx.jh, vhash, size );
}
jh512_4way_close( &ctx.jh, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
break;
case KECCAK:
keccak512_4way_init( &ctx.keccak );
if ( i == 0 )
keccak512_4way( &ctx.keccak, input, size );
else
{
mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
keccak512_4way( &ctx.keccak, vhash, size );
}
keccak512_4way_close( &ctx.keccak, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
break;
case LUFFA:
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash0,
(const BitSequence*)in0, size );
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash1,
(const BitSequence*)in1, size );
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash2,
(const BitSequence*)in2, size );
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash3,
(const BitSequence*)in3, size );
break;
case CUBEHASH:
cubehashReinit( &ctx.cube );
cubehashUpdateDigest( &ctx.cube, (byte*) hash0,
(const byte*)in0, size );
cubehashReinit( &ctx.cube );
cubehashUpdateDigest( &ctx.cube, (byte*) hash1,
(const byte*)in1, size );
cubehashReinit( &ctx.cube );
cubehashUpdateDigest( &ctx.cube, (byte*) hash2,
(const byte*)in2, size );
cubehashReinit( &ctx.cube );
cubehashUpdateDigest( &ctx.cube, (byte*) hash3,
(const byte*)in3, size );
break;
case SHAVITE:
sph_shavite512_init( &ctx.shavite );
sph_shavite512( &ctx.shavite, in0, size );
sph_shavite512_close( &ctx.shavite, hash0 );
sph_shavite512_init( &ctx.shavite );
sph_shavite512( &ctx.shavite, in1, size );
sph_shavite512_close( &ctx.shavite, hash1 );
sph_shavite512_init( &ctx.shavite );
sph_shavite512( &ctx.shavite, in2, size );
sph_shavite512_close( &ctx.shavite, hash2 );
sph_shavite512_init( &ctx.shavite );
sph_shavite512( &ctx.shavite, in3, size );
sph_shavite512_close( &ctx.shavite, hash3 );
break;
case SIMD:
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash0,
(const BitSequence*)in0, size<<3 );
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash1,
(const BitSequence*)in1, size<<3 );
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash2,
(const BitSequence*)in2, size<<3 );
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash3,
(const BitSequence*)in3, size<<3 );
break;
case ECHO:
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash0,
(const BitSequence*)in0, size<<3 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash1,
(const BitSequence*)in1, size<<3 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash2,
(const BitSequence*)in2, size<<3 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash3,
(const BitSequence*)in3, size<<3 );
break;
case HAMSI:
mm_interleave_4x32( vhash, in0, in1, in2, in3, size<<3 );
hamsi512_4way_init( &ctx.hamsi );
hamsi512_4way( &ctx.hamsi, vhash, size );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
break;
case FUGUE:
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, in0, size );
sph_fugue512_close( &ctx.fugue, hash0 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, in1, size );
sph_fugue512_close( &ctx.fugue, hash1 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, in2, size );
sph_fugue512_close( &ctx.fugue, hash2 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, in3, size );
sph_fugue512_close( &ctx.fugue, hash3 );
break;
case SHABAL:
mm_interleave_4x32( vhash, in0, in1, in2, in3, size<<3 );
shabal512_4way_init( &ctx.shabal );
shabal512_4way( &ctx.shabal, vhash, size );
shabal512_4way_close( &ctx.shabal, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
break;
case WHIRLPOOL:
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, in0, size );
sph_whirlpool_close( &ctx.whirlpool, hash0 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, in1, size );
sph_whirlpool_close( &ctx.whirlpool, hash1 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, in2, size );
sph_whirlpool_close( &ctx.whirlpool, hash2 );
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, in3, size );
sph_whirlpool_close( &ctx.whirlpool, hash3 );
break;
case SHA_512:
mm256_interleave_4x64( vhash, in0, in1, in2, in3, size<<3 );
sha512_4way_init( &ctx.sha512 );
sha512_4way( &ctx.sha512, vhash, size );
sha512_4way_close( &ctx.sha512, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
break;
}
in0 = (void*) hash0;
in1 = (void*) hash1;
in2 = (void*) hash2;
in3 = (void*) hash3;
size = 64;
}
memcpy( output, hash0, 32 );
memcpy( output+32, hash1, 32 );
memcpy( output+64, hash2, 32 );
memcpy( output+96, hash3, 32 );
}
int scanhash_x16r_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
{
uint32_t hash[4*16] __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;
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
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;
volatile uint8_t *restart = &(work_restart[thr_id].restart);
for ( int k=0; k < 19; k++ )
be32enc( &endiandata[k], pdata[k] );
if ( s_ntime != pdata[17] )
{
uint32_t ntime = swab32(pdata[17]);
x16r_getAlgoString( (const char*) (&endiandata[1]), hashOrder );
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_DEBUG, "hash order %s (%08x)", hashOrder, ntime );
}
if ( opt_benchmark )
ptarget[7] = 0x0cff;
uint64_t *edata = (uint64_t*)endiandata;
mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
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 );
x16r_4way_hash( hash, vdata );
pdata[19] = n;
if ( hash[7] <= Htarg && fulltest( hash, ptarget ) )
{
found[0] = true;
num_found++;
nonces[0] = n;
work_set_target_ratio( work, hash );
}
if ( (hash+8)[7] <= Htarg && fulltest( hash, ptarget ) )
{
found[1] = true;
num_found++;
nonces[1] = n+1;
work_set_target_ratio( work, hash+8 );
}
if ( (hash+16)[7] <= Htarg && fulltest( hash, ptarget ) )
{
found[2] = true;
num_found++;
nonces[2] = n+2;
work_set_target_ratio( work, hash+16 );
}
if ( (hash+24)[7] <= Htarg && fulltest( hash, 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 ) && !(*restart) );
*hashes_done = n - first_nonce + 1;
return num_found;
}
#endif

35
algo/x17/x16r-gate.c Normal file
View File

@@ -0,0 +1,35 @@
#include "x16r-gate.h"
void x16r_getAlgoString( const uint8_t* prevblock, char *output )
{
char *sptr = output;
for ( int j = 0; j < X16R_HASH_FUNC_COUNT; j++ )
{
uint8_t b = (15 - j) >> 1; // 16 first ascii hex chars (lsb in uint256)
uint8_t algoDigit = (j & 1) ? prevblock[b] & 0xF : prevblock[b] >> 4;
if (algoDigit >= 10)
sprintf(sptr, "%c", 'A' + (algoDigit - 10));
else
sprintf(sptr, "%u", (uint32_t) algoDigit);
sptr++;
}
*sptr = '\0';
}
bool register_x16r_algo( algo_gate_t* gate )
{
#if defined (X16R_4WAY)
init_x16r_4way_ctx();
gate->scanhash = (void*)&scanhash_x16r_4way;
gate->hash = (void*)&x16r_4way_hash;
#else
init_x16r_ctx();
gate->scanhash = (void*)&scanhash_x16r;
gate->hash = (void*)&x16r_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX_OPT | AVX2_OPT;
gate->set_target = (void*)&alt_set_target;
return true;
};

54
algo/x17/x16r-gate.h Normal file
View File

@@ -0,0 +1,54 @@
#ifndef X16R_GATE_H__
#define X16R_GATE_H__ 1
#include "algo-gate-api.h"
#include "avxdefs.h"
#include <stdint.h>
#if defined(__AVX2__) && defined(__AES__)
#define X16R_4WAY
#endif
enum x16r_Algo {
BLAKE = 0,
BMW,
GROESTL,
JH,
KECCAK,
SKEIN,
LUFFA,
CUBEHASH,
SHAVITE,
SIMD,
ECHO,
HAMSI,
FUGUE,
SHABAL,
WHIRLPOOL,
SHA_512,
X16R_HASH_FUNC_COUNT
};
bool register_x16r_algo( algo_gate_t* gate );
void x16r_getAlgoString( const uint8_t* prevblock, char *output );
#if defined(X16R_4WAY)
void x16r_4way_hash( void *state, const void *input );
int scanhash_x16r_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
void init_x16r_4way_ctx();
#endif
void x16r_hash( void *state, const void *input );
int scanhash_x16r( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
void init_x16r_ctx();
#endif

252
algo/x17/x16r.c Normal file
View File

@@ -0,0 +1,252 @@
/**
* x16r algo implementation
*
* Implementation by tpruvot@github Jan 2018
* Optimized by JayDDee@github Jan 2018
*/
#include "x16r-gate.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "algo/blake/sph_blake.h"
#include "algo/bmw/sph_bmw.h"
#include "algo/groestl/sph_groestl.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/sse2/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/simd/sse2/nist.h"
#include "algo/echo/sph_echo.h"
#include "algo/hamsi/sph_hamsi.h"
#include "algo/fugue/sph_fugue.h"
#include "algo/shabal/sph_shabal.h"
#include "algo/whirlpool/sph_whirlpool.h"
#include <openssl/sha.h>
#ifndef NO_AES_NI
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/groestl/aes_ni/hash-groestl.h"
#endif
static __thread uint32_t s_ntime = UINT32_MAX;
static __thread char hashOrder[X16R_HASH_FUNC_COUNT + 1] = { 0 };
typedef struct {
#ifdef NO_AES_NI
sph_groestl512_context groestl;
sph_echo512_context echo;
#else
hashState_echo echo;
hashState_groestl groestl;
#endif
sph_blake512_context blake;
sph_bmw512_context bmw;
sph_skein512_context skein;
sph_jh512_context jh;
sph_keccak512_context keccak;
hashState_luffa luffa;
cubehashParam cube;
sph_shavite512_context shavite;
hashState_sd simd;
sph_hamsi512_context hamsi;
sph_fugue512_context fugue;
sph_shabal512_context shabal;
sph_whirlpool_context whirlpool;
SHA512_CTX sha512;
} x16r_ctx_holder;
x16r_ctx_holder x16r_ctx __attribute__ ((aligned (64)));
void init_x16r_ctx()
{
//#ifdef NO_AES_NI
// sph_groestl512_init(&x16r_ctx.groestl );
// sph_echo512_init(&x16r_ctx.echo);
//#else
// init_echo( &x16r_ctx.echo, 512 );
// init_groestl( &x16r_ctx.groestl, 64 );
//#endif
// sph_blake512_init( &x16r_ctx.blake );
// sph_bmw512_init( &x16r_ctx.bmw );
// sph_skein512_init( &x16r_ctx.bmw );
// sph_jh512_init( &x16r_ctx.jh );
// sph_keccak512_init( &x16r_ctx.keccak );
// init_luffa( &x16r_ctx.luffa, 512 );
cubehashInit( &x16r_ctx.cube, 512, 16, 32 );
// sph_shavite512_init( &x16r_ctx.shavite );
// init_sd( &x16r_ctx.simd, 512 );
// sph_hamsi512_init( &x16r_ctx.hamsi );
// sph_fugue512_init( &x16r_ctx.fugue );
// sph_shabal512_init( &x16r_ctx.shabal );
// sph_whirlpool_init( &x16r_ctx.whirlpool );
// SHA512_Init( &x16r_ctx.sha512 );
};
void x16r_hash( void* output, const void* input )
{
uint32_t _ALIGN(128) hash[16];
x16r_ctx_holder ctx;
void *in = (void*) input;
int size = 80;
if ( s_ntime == UINT32_MAX )
{
const uint8_t* in8 = (uint8_t*) input;
x16r_getAlgoString( &in8[4], hashOrder );
}
for ( int i = 0; i < 16; i++ )
{
const char elem = hashOrder[i];
const uint8_t algo = elem >= 'A' ? elem - 'A' + 10 : elem - '0';
switch ( algo )
{
case BLAKE:
sph_blake512_init( &ctx.blake );
sph_blake512( &ctx.blake, in, size );
sph_blake512_close( &ctx.blake, hash );
break;
case BMW:
sph_bmw512_init( &ctx.bmw );
sph_bmw512(&ctx.bmw, in, size);
sph_bmw512_close(&ctx.bmw, hash);
break;
case GROESTL:
#ifdef NO_AES_NI
sph_groestl512_init( &ctx.groestl );
sph_groestl512( &ctx.groestl, in, size<<3 );
sph_groestl512_close(&ctx.groestl, hash);
#else
init_groestl( &ctx.groestl, 64 );
update_and_final_groestl( &ctx.groestl, (char*)hash,
(const char*)in, size<<3 );
#endif
break;
case SKEIN:
sph_skein512_init( &ctx.skein );
sph_skein512( &ctx.skein, in, size );
sph_skein512_close( &ctx.skein, hash );
break;
case JH:
sph_jh512_init( &ctx.jh );
sph_jh512(&ctx.jh, in, size );
sph_jh512_close(&ctx.jh, hash );
break;
case KECCAK:
sph_keccak512_init( &ctx.keccak );
sph_keccak512( &ctx.keccak, in, size );
sph_keccak512_close( &ctx.keccak, hash );
break;
case LUFFA:
init_luffa( &ctx.luffa, 512 );
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash,
(const BitSequence*)in, size );
break;
case CUBEHASH:
memcpy( &ctx.cube, &x16r_ctx.cube, sizeof(cubehashParam) );
cubehashUpdateDigest( &ctx.cube, (byte*) hash,
(const byte*)in, size );
break;
case SHAVITE:
sph_shavite512_init( &ctx.shavite );
sph_shavite512( &ctx.shavite, in, size );
sph_shavite512_close( &ctx.shavite, hash );
break;
case SIMD:
init_sd( &ctx.simd, 512 );
update_final_sd( &ctx.simd, (BitSequence *)hash,
(const BitSequence*)in, size<<3 );
break;
case ECHO:
#ifdef NO_AES_NI
sph_echo512_init( &ctx.echo );
sph_echo512( &ctx.echo, in, size );
sph_echo512_close( &ctx.echo, hash );
#else
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash,
(const BitSequence*)in, size<<3 );
#endif
break;
case HAMSI:
sph_hamsi512_init( &ctx.hamsi );
sph_hamsi512( &ctx.hamsi, in, size );
sph_hamsi512_close( &ctx.hamsi, hash );
break;
case FUGUE:
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, in, size );
sph_fugue512_close( &ctx.fugue, hash );
break;
case SHABAL:
sph_shabal512_init( &ctx.shabal );
sph_shabal512( &ctx.shabal, in, size );
sph_shabal512_close( &ctx.shabal, hash );
break;
case WHIRLPOOL:
sph_whirlpool_init( &ctx.whirlpool );
sph_whirlpool( &ctx.whirlpool, in, size );
sph_whirlpool_close( &ctx.whirlpool, hash );
break;
case SHA_512:
SHA512_Init( &ctx.sha512 );
SHA512_Update( &ctx.sha512, in, size );
SHA512_Final( (unsigned char*) hash, &ctx.sha512 );
break;
}
in = (void*) hash;
size = 64;
}
memcpy(output, hash, 32);
}
int scanhash_x16r( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
{
uint32_t _ALIGN(128) hash32[8];
uint32_t _ALIGN(128) endiandata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t nonce = first_nonce;
volatile uint8_t *restart = &(work_restart[thr_id].restart);
for ( int k=0; k < 19; k++ )
be32enc( &endiandata[k], pdata[k] );
if ( s_ntime != pdata[17] )
{
uint32_t ntime = swab32(pdata[17]);
x16r_getAlgoString( (const char*) (&endiandata[1]), hashOrder );
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_DEBUG, "hash order %s (%08x)", hashOrder, ntime );
}
if ( opt_benchmark )
ptarget[7] = 0x0cff;
do
{
be32enc( &endiandata[19], nonce );
x16r_hash( hash32, endiandata );
if ( hash32[7] <= Htarg && fulltest( hash32, ptarget ) )
{
work_set_target_ratio( work, hash32 );
pdata[19] = nonce;
*hashes_done = pdata[19] - first_nonce;
return 1;
}
nonce++;
} while ( nonce < max_nonce && !(*restart) );
pdata[19] = nonce;
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}

81
algo/x17/x17-4way.c
View File

@@ -1,6 +1,6 @@
#include "x17-gate.h"
#if defined(__AVX2__) && defined(__AES__)
#if defined(X17_4WAY)
#include <stdlib.h>
#include <stdint.h>
@@ -17,12 +17,12 @@
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/sse2/nist.h"
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/hamsi/sph_hamsi.h"
#include "algo/hamsi/hamsi-hash-4way.h"
#include "algo/fugue/sph_fugue.h"
#include "algo/shabal/shabal-hash-4way.h"
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/haval/sph-haval.h"
#include <openssl/sha.h>
#include "algo/haval/haval-hash-4way.h"
#include "algo/sha/sha2-hash-4way.h"
typedef struct {
blake512_4way_context blake;
@@ -36,12 +36,12 @@ typedef struct {
sph_shavite512_context shavite;
hashState_sd simd;
hashState_echo echo;
sph_hamsi512_context hamsi;
hamsi512_4way_context hamsi;
sph_fugue512_context fugue;
shabal512_4way_context shabal;
sph_whirlpool_context whirlpool;
SHA512_CTX sha512;
sph_haval256_5_context haval;
sha512_4way_context sha512;
haval256_5_4way_context haval;
} x17_4way_ctx_holder;
x17_4way_ctx_holder x17_4way_ctx __attribute__ ((aligned (64)));
@@ -59,11 +59,11 @@ void init_x17_4way_ctx()
sph_shavite512_init( &x17_4way_ctx.shavite );
init_sd( &x17_4way_ctx.simd, 512 );
init_echo( &x17_4way_ctx.echo, 512 );
sph_hamsi512_init( &x17_4way_ctx.hamsi );
hamsi512_4way_init( &x17_4way_ctx.hamsi );
sph_fugue512_init( &x17_4way_ctx.fugue );
shabal512_4way_init( &x17_4way_ctx.shabal );
SHA512_Init( &x17_4way_ctx.sha512 );
sph_haval256_5_init( &x17_4way_ctx.haval );
sha512_4way_init( &x17_4way_ctx.sha512 );
haval256_5_4way_init( &x17_4way_ctx.haval );
};
void x17_4way_hash( void *state, const void *input )
@@ -73,6 +73,7 @@ void x17_4way_hash( void *state, const void *input )
uint64_t hash2[8] __attribute__ ((aligned (64)));
uint64_t hash3[8] __attribute__ ((aligned (64)));
uint64_t vhash[8*4] __attribute__ ((aligned (64)));
uint64_t vhash32[8*4] __attribute__ ((aligned (64)));
x17_4way_ctx_holder ctx;
memcpy( &ctx, &x17_4way_ctx, sizeof(x17_4way_ctx) );
@@ -111,10 +112,9 @@ void x17_4way_hash( void *state, const void *input )
keccak512_4way( &ctx.keccak, vhash, 64 );
keccak512_4way_close( &ctx.keccak, vhash );
// Serial to the end
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// 7 Luffa
// 7 Luffa serial
update_and_final_luffa( &ctx.luffa, (BitSequence*)hash0,
(const BitSequence*)hash0, 64 );
memcpy( &ctx.luffa, &x17_4way_ctx.luffa, sizeof(hashState_luffa) );
@@ -178,18 +178,11 @@ void x17_4way_hash( void *state, const void *input )
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
// 12 Hamsi
sph_hamsi512( &ctx.hamsi, hash0, 64 );
sph_hamsi512_close( &ctx.hamsi, hash0 );
memcpy( &ctx.hamsi, &x17_4way_ctx.hamsi, sizeof(sph_hamsi512_context) );
sph_hamsi512( &ctx.hamsi, hash1, 64 );
sph_hamsi512_close( &ctx.hamsi, hash1 );
memcpy( &ctx.hamsi, &x17_4way_ctx.hamsi, sizeof(sph_hamsi512_context) );
sph_hamsi512( &ctx.hamsi, hash2, 64 );
sph_hamsi512_close( &ctx.hamsi, hash2 );
memcpy( &ctx.hamsi, &x17_4way_ctx.hamsi, sizeof(sph_hamsi512_context) );
sph_hamsi512( &ctx.hamsi, hash3, 64 );
sph_hamsi512_close( &ctx.hamsi, hash3 );
// 12 Hamsi parallel 4way 32 bit
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
// 13 Fugue
sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -226,39 +219,17 @@ void x17_4way_hash( void *state, const void *input )
sph_whirlpool( &ctx.whirlpool, hash3, 64 );
sph_whirlpool_close( &ctx.whirlpool, hash3 );
// 16 SHA512
SHA512_Update( &ctx.sha512, hash0, 64 );
SHA512_Final( (unsigned char*)hash0, &ctx.sha512 );
memcpy( &ctx.sha512, &x17_4way_ctx.sha512, sizeof(SHA512_CTX) );
SHA512_Update( &ctx.sha512, hash1, 64 );
SHA512_Final( (unsigned char*)hash1, &ctx.sha512 );
memcpy( &ctx.sha512, &x17_4way_ctx.sha512, sizeof(SHA512_CTX) );
SHA512_Update( &ctx.sha512, hash2, 64 );
SHA512_Final( (unsigned char*)hash2, &ctx.sha512 );
memcpy( &ctx.sha512, &x17_4way_ctx.sha512, sizeof(SHA512_CTX) );
SHA512_Update( &ctx.sha512, hash3, 64 );
SHA512_Final( (unsigned char*)hash3, &ctx.sha512 );
// 16 SHA512 parallel 64 bit
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
sha512_4way( &ctx.sha512, vhash, 64 );
sha512_4way_close( &ctx.sha512, vhash );
// 17 Haval
sph_haval256_5( &ctx.haval, (const void*)hash0, 64 );
sph_haval256_5_close( &ctx.haval, hash0 );
memcpy( &ctx.haval, &x17_4way_ctx.haval,
sizeof(sph_haval256_5_context) );
sph_haval256_5( &ctx.haval, (const void*)hash1, 64 );
sph_haval256_5_close( &ctx.haval, hash1 );
memcpy( &ctx.haval, &x17_4way_ctx.haval,
sizeof(sph_haval256_5_context) );
sph_haval256_5( &ctx.haval, (const void*)hash2, 64 );
sph_haval256_5_close( &ctx.haval, hash2 );
memcpy( &ctx.haval, &x17_4way_ctx.haval,
sizeof(sph_haval256_5_context) );
sph_haval256_5( &ctx.haval, (const void*)hash3, 64 );
sph_haval256_5_close( &ctx.haval, hash3 );
// 17 Haval parallel 32 bit
mm256_reinterleave_4x32( vhash32, vhash, 512 );
haval256_5_4way( &ctx.haval, vhash32, 64 );
haval256_5_4way_close( &ctx.haval, vhash );
memcpy( state, hash0, 32 );
memcpy( state+32, hash1, 32 );
memcpy( state+64, hash2, 32 );
memcpy( state+96, hash3, 32 );
mm_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
}
int scanhash_x17_4way( int thr_id, struct work *work, uint32_t max_nonce,

2
algo/x17/x17-gate.c
View File

@@ -11,7 +11,7 @@ bool register_x17_algo( algo_gate_t* gate )
gate->scanhash = (void*)&scanhash_x17;
gate->hash = (void*)&x17_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX_OPT | AVX2_OPT | FOUR_WAY_OPT;
gate->optimizations = SSE2_OPT | AES_OPT | AVX_OPT | AVX2_OPT;
return true;
};

2
algo/x17/x17-gate.h
View File

@@ -4,7 +4,7 @@
#include "algo-gate-api.h"
#include <stdint.h>
#if defined(HASH_4WAY) && defined(__AES__)
#if defined(__AVX2__) && defined(__AES__)
#define X17_4WAY
#endif

124
algo/x17/xevan-4way.c
View File

@@ -1,6 +1,6 @@
#include "xevan-gate.h"
#if defined(__AVX2__) && defined(__AES__)
#if defined(XEVAN_4WAY)
#include <stdlib.h>
#include <stdint.h>
@@ -17,13 +17,12 @@
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/simd/sse2/nist.h"
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/hamsi/sph_hamsi.h"
#include "algo/hamsi/hamsi-hash-4way.h"
#include "algo/fugue/sph_fugue.h"
#include "algo/shabal/shabal-hash-4way.h"
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/sha/sph_sha2.h"
#include "algo/haval/sph-haval.h"
#include <openssl/sha.h>
#include "algo/sha/sha2-hash-4way.h"
#include "algo/haval/haval-hash-4way.h"
typedef struct {
blake512_4way_context blake;
@@ -37,12 +36,12 @@ typedef struct {
sph_shavite512_context shavite;
hashState_sd simd;
hashState_echo echo;
sph_hamsi512_context hamsi;
hamsi512_4way_context hamsi;
sph_fugue512_context fugue;
shabal512_4way_context shabal;
sph_whirlpool_context whirlpool;
SHA512_CTX sha512;
sph_haval256_5_context haval;
sha512_4way_context sha512;
haval256_5_4way_context haval;
} xevan_4way_ctx_holder;
xevan_4way_ctx_holder xevan_4way_ctx __attribute__ ((aligned (64)));
@@ -62,12 +61,12 @@ void init_xevan_4way_ctx()
sph_shavite512_init( &xevan_4way_ctx.shavite );
init_sd( &xevan_4way_ctx.simd, 512 );
init_echo( &xevan_4way_ctx.echo, 512 );
sph_hamsi512_init( &xevan_4way_ctx.hamsi );
hamsi512_4way_init( &xevan_4way_ctx.hamsi );
sph_fugue512_init( &xevan_4way_ctx.fugue );
shabal512_4way_init( &xevan_4way_ctx.shabal );
sph_whirlpool_init( &xevan_4way_ctx.whirlpool );
SHA512_Init( &xevan_4way_ctx.sha512 );
sph_haval256_5_init( &xevan_4way_ctx.haval );
sha512_4way_init( &xevan_4way_ctx.sha512 );
haval256_5_4way_init( &xevan_4way_ctx.haval );
};
void xevan_4way_blake512_midstate( const void* input )
@@ -84,6 +83,7 @@ void xevan_4way_hash( void *output, const void *input )
uint64_t hash2[16] __attribute__ ((aligned (64)));
uint64_t hash3[16] __attribute__ ((aligned (64)));
uint64_t vhash[16<<2] __attribute__ ((aligned (64)));
uint64_t vhash32[16<<2] __attribute__ ((aligned (64)));
const int dataLen = 128;
const int midlen = 64; // bytes
const int tail = 80 - midlen; // 16
@@ -193,17 +193,11 @@ void xevan_4way_hash( void *output, const void *input )
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, dataLen<<3 );
sph_hamsi512( &ctx.hamsi, hash0, dataLen );
sph_hamsi512_close( &ctx.hamsi, hash0 );
memcpy( &ctx.hamsi, &xevan_4way_ctx.hamsi, sizeof(sph_hamsi512_context) );
sph_hamsi512( &ctx.hamsi, hash1, dataLen );
sph_hamsi512_close( &ctx.hamsi, hash1 );
memcpy( &ctx.hamsi, &xevan_4way_ctx.hamsi, sizeof(sph_hamsi512_context) );
sph_hamsi512( &ctx.hamsi, hash2, dataLen );
sph_hamsi512_close( &ctx.hamsi, hash2 );
memcpy( &ctx.hamsi, &xevan_4way_ctx.hamsi, sizeof(sph_hamsi512_context) );
sph_hamsi512( &ctx.hamsi, hash3, dataLen );
sph_hamsi512_close( &ctx.hamsi, hash3 );
// Parallel 32 bit
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
hamsi512_4way( &ctx.hamsi, vhash, dataLen );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
sph_fugue512( &ctx.fugue, hash0, dataLen );
sph_fugue512_close( &ctx.fugue, hash0 );
@@ -217,7 +211,7 @@ void xevan_4way_hash( void *output, const void *input )
sph_fugue512( &ctx.fugue, hash3, dataLen );
sph_fugue512_close( &ctx.fugue, hash3 );
// Parallel 4way
// Parallel 4way 32 bit
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
shabal512_4way( &ctx.shabal, vhash, dataLen );
shabal512_4way_close( &ctx.shabal, vhash );
@@ -239,32 +233,14 @@ void xevan_4way_hash( void *output, const void *input )
sph_whirlpool( &ctx.whirlpool, hash3, dataLen );
sph_whirlpool_close( &ctx.whirlpool, hash3 );
SHA512_Update( &ctx.sha512, hash0, dataLen );
SHA512_Final( (unsigned char*)hash0, &ctx.sha512 );
memcpy( &ctx.sha512, &xevan_4way_ctx.sha512, sizeof(SHA512_CTX) );
SHA512_Update( &ctx.sha512, hash1, dataLen );
SHA512_Final( (unsigned char*)hash1, &ctx.sha512 );
memcpy( &ctx.sha512, &xevan_4way_ctx.sha512, sizeof(SHA512_CTX) );
SHA512_Update( &ctx.sha512, hash2, dataLen );
SHA512_Final( (unsigned char*)hash2, &ctx.sha512 );
memcpy( &ctx.sha512, &xevan_4way_ctx.sha512, sizeof(SHA512_CTX) );
SHA512_Update( &ctx.sha512, hash3, dataLen );
SHA512_Final( (unsigned char*)hash3, &ctx.sha512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
sha512_4way( &ctx.sha512, vhash, dataLen );
sha512_4way_close( &ctx.sha512, vhash );
sph_haval256_5( &ctx.haval, (const void*)hash0, dataLen );
sph_haval256_5_close( &ctx.haval, hash0 );
memcpy( &ctx.haval, &xevan_4way_ctx.haval,
sizeof(sph_haval256_5_context) );
sph_haval256_5( &ctx.haval, (const void*)hash1, dataLen );
sph_haval256_5_close( &ctx.haval, hash1 );
memcpy( &ctx.haval, &xevan_4way_ctx.haval,
sizeof(sph_haval256_5_context) );
sph_haval256_5( &ctx.haval, (const void*)hash2, dataLen );
sph_haval256_5_close( &ctx.haval, hash2 );
memcpy( &ctx.haval, &xevan_4way_ctx.haval,
sizeof(sph_haval256_5_context) );
sph_haval256_5( &ctx.haval, (const void*)hash3, dataLen );
sph_haval256_5_close( &ctx.haval, hash3 );
mm256_reinterleave_4x32( vhash32, vhash, dataLen<<3 );
haval256_5_4way( &ctx.haval, vhash32, dataLen );
haval256_5_4way_close( &ctx.haval, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
memset( &vhash[ 4<<2 ], 0, (dataLen-32) << 2 );
@@ -366,17 +342,10 @@ void xevan_4way_hash( void *output, const void *input )
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, dataLen<<3 );
sph_hamsi512( &ctx.hamsi, hash0, dataLen );
sph_hamsi512_close( &ctx.hamsi, hash0 );
memcpy( &ctx.hamsi, &xevan_4way_ctx.hamsi, sizeof(sph_hamsi512_context) );
sph_hamsi512( &ctx.hamsi, hash1, dataLen );
sph_hamsi512_close( &ctx.hamsi, hash1 );
memcpy( &ctx.hamsi, &xevan_4way_ctx.hamsi, sizeof(sph_hamsi512_context) );
sph_hamsi512( &ctx.hamsi, hash2, dataLen );
sph_hamsi512_close( &ctx.hamsi, hash2 );
memcpy( &ctx.hamsi, &xevan_4way_ctx.hamsi, sizeof(sph_hamsi512_context) );
sph_hamsi512( &ctx.hamsi, hash3, dataLen );
sph_hamsi512_close( &ctx.hamsi, hash3 );
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
hamsi512_4way( &ctx.hamsi, vhash, dataLen );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
sph_fugue512( &ctx.fugue, hash0, dataLen );
sph_fugue512_close( &ctx.fugue, hash0 );
@@ -410,37 +379,16 @@ void xevan_4way_hash( void *output, const void *input )
sph_whirlpool( &ctx.whirlpool, hash3, dataLen );
sph_whirlpool_close( &ctx.whirlpool, hash3 );
SHA512_Update( &ctx.sha512, hash0, dataLen );
SHA512_Final( (unsigned char*)hash0, &ctx.sha512 );
memcpy( &ctx.sha512, &xevan_4way_ctx.sha512, sizeof(SHA512_CTX) );
SHA512_Update( &ctx.sha512, hash1, dataLen );
SHA512_Final( (unsigned char*)hash1, &ctx.sha512 );
memcpy( &ctx.sha512, &xevan_4way_ctx.sha512, sizeof(SHA512_CTX) );
SHA512_Update( &ctx.sha512, hash2, dataLen );
SHA512_Final( (unsigned char*)hash2, &ctx.sha512 );
memcpy( &ctx.sha512, &xevan_4way_ctx.sha512, sizeof(SHA512_CTX) );
SHA512_Update( &ctx.sha512, hash3, dataLen );
SHA512_Final( (unsigned char*)hash3, &ctx.sha512 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
sha512_4way( &ctx.sha512, vhash, dataLen );
sha512_4way_close( &ctx.sha512, vhash );
sph_haval256_5( &ctx.haval, (const void*)hash0, dataLen );
sph_haval256_5_close( &ctx.haval, hash0 );
memcpy( &ctx.haval, &xevan_4way_ctx.haval,
sizeof(sph_haval256_5_context) );
sph_haval256_5( &ctx.haval, (const void*)hash1, dataLen );
sph_haval256_5_close( &ctx.haval, hash1 );
memcpy( &ctx.haval, &xevan_4way_ctx.haval,
sizeof(sph_haval256_5_context) );
sph_haval256_5( &ctx.haval, (const void*)hash2, dataLen );
sph_haval256_5_close( &ctx.haval, hash2 );
memcpy( &ctx.haval, &xevan_4way_ctx.haval,
sizeof(sph_haval256_5_context) );
sph_haval256_5( &ctx.haval, (const void*)hash3, dataLen );
sph_haval256_5_close( &ctx.haval, hash3 );
mm256_reinterleave_4x32( vhash32, vhash, dataLen<<3 );
haval256_5_4way( &ctx.haval, vhash32, dataLen );
haval256_5_4way_close( &ctx.haval, vhash32 );
memcpy( output, hash0, 32 );
memcpy( output+32, hash1, 32 );
memcpy( output+64, hash2, 32 );
memcpy( output+96, hash3, 32 );
mm_deinterleave_4x32( output, output+32, output+64, output+96,
vhash32, 256 );
}
int scanhash_xevan_4way( int thr_id, struct work *work, uint32_t max_nonce,

2
algo/x17/xevan-gate.c
View File

@@ -16,7 +16,7 @@ bool register_xevan_algo( algo_gate_t* gate )
gate->scanhash = (void*)&scanhash_xevan;
gate->hash = (void*)&xevan_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX_OPT | AVX2_OPT | FOUR_WAY_OPT;
gate->optimizations = SSE2_OPT | AES_OPT | AVX_OPT | AVX2_OPT;
gate->set_target = (void*)&xevan_set_target;
gate->get_max64 = (void*)&get_max64_0xffffLL;
return true;

2
algo/x17/xevan-gate.h
View File

@@ -4,7 +4,7 @@
#include "algo-gate-api.h"
#include <stdint.h>
#if defined(HASH_4WAY) && defined(__AES__)
#if defined(__AVX2__) && defined(__AES__)
#define XEVAN_4WAY
#endif