mirror of
https://github.com/JayDDee/cpuminer-opt.git
synced 2025-09-17 23:44:27 +00:00
186 lines
6.7 KiB
C
186 lines
6.7 KiB
C
#include "lyra2rev2-gate.h"
|
|
#include <memory.h>
|
|
|
|
#if defined (__AVX2__)
|
|
|
|
#include "algo/blake/blake-hash-4way.h"
|
|
#include "algo/keccak/keccak-hash-4way.h"
|
|
#include "algo/skein/skein-hash-4way.h"
|
|
#include "algo/bmw/bmw-hash-4way.h"
|
|
|
|
#include "algo/cubehash/sph_cubehash.h"
|
|
//#include "algo/bmw/sph_bmw.h"
|
|
#include "algo/cubehash/sse2/cubehash_sse2.h"
|
|
|
|
typedef struct {
|
|
blake256_4way_context blake;
|
|
keccak256_4way_context keccak;
|
|
cubehashParam cube;
|
|
skein256_4way_context skein;
|
|
bmw256_4way_context bmw;
|
|
// sph_bmw256_context bmw;
|
|
} lyra2v2_4way_ctx_holder;
|
|
|
|
static lyra2v2_4way_ctx_holder l2v2_4way_ctx;
|
|
|
|
void init_lyra2rev2_4way_ctx()
|
|
{
|
|
// blake256_4way_init( &l2v2_4way_ctx.blake );
|
|
keccak256_4way_init( &l2v2_4way_ctx.keccak );
|
|
cubehashInit( &l2v2_4way_ctx.cube, 256, 16, 32 );
|
|
skein256_4way_init( &l2v2_4way_ctx.skein );
|
|
bmw256_4way_init( &l2v2_4way_ctx.bmw );
|
|
// sph_bmw256_init( &l2v2_4way_ctx.bmw );
|
|
}
|
|
|
|
void lyra2rev2_4way_hash( void *state, const void *input )
|
|
{
|
|
uint32_t hash0[8] __attribute__ ((aligned (64)));
|
|
uint32_t hash1[8] __attribute__ ((aligned (32)));
|
|
uint32_t hash2[8] __attribute__ ((aligned (32)));
|
|
uint32_t hash3[8] __attribute__ ((aligned (32)));
|
|
uint32_t vhash[8*4] __attribute__ ((aligned (64)));
|
|
uint64_t vhash64[4*4] __attribute__ ((aligned (64)));
|
|
lyra2v2_4way_ctx_holder ctx __attribute__ ((aligned (64)));
|
|
memcpy( &ctx, &l2v2_4way_ctx, sizeof(l2v2_4way_ctx) );
|
|
|
|
blake256_4way( &ctx.blake, input + (64<<2), 16 );
|
|
// blake256_4way( &ctx.blake, input, 80 );
|
|
blake256_4way_close( &ctx.blake, vhash );
|
|
|
|
mm256_reinterleave_4x64( vhash64, vhash, 256 );
|
|
keccak256_4way( &ctx.keccak, vhash64, 32 );
|
|
keccak256_4way_close( &ctx.keccak, vhash64 );
|
|
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
|
|
|
|
cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*) hash0, 32 );
|
|
memcpy( &ctx.cube, &l2v2_4way_ctx.cube, sizeof ctx.cube );
|
|
cubehashUpdateDigest( &ctx.cube, (byte*) hash1, (const byte*) hash1, 32 );
|
|
memcpy( &ctx.cube, &l2v2_4way_ctx.cube, sizeof ctx.cube );
|
|
cubehashUpdateDigest( &ctx.cube, (byte*) hash2, (const byte*) hash2, 32 );
|
|
memcpy( &ctx.cube, &l2v2_4way_ctx.cube, sizeof ctx.cube );
|
|
cubehashUpdateDigest( &ctx.cube, (byte*) hash3, (const byte*) hash3, 32 );
|
|
|
|
LYRA2REV2( l2v2_wholeMatrix, hash0, 32, hash0, 32, hash0, 32, 1, 4, 4 );
|
|
LYRA2REV2( l2v2_wholeMatrix, hash1, 32, hash1, 32, hash1, 32, 1, 4, 4 );
|
|
LYRA2REV2( l2v2_wholeMatrix, hash2, 32, hash2, 32, hash2, 32, 1, 4, 4 );
|
|
LYRA2REV2( l2v2_wholeMatrix, hash3, 32, hash3, 32, hash3, 32, 1, 4, 4 );
|
|
|
|
mm256_interleave_4x64( vhash64, hash0, hash1, hash2, hash3, 256 );
|
|
skein256_4way( &ctx.skein, vhash64, 32 );
|
|
skein256_4way_close( &ctx.skein, vhash64 );
|
|
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash64, 256 );
|
|
|
|
memcpy( &ctx.cube, &l2v2_4way_ctx.cube, sizeof ctx.cube );
|
|
cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*) hash0, 32 );
|
|
memcpy( &ctx.cube, &l2v2_4way_ctx.cube, sizeof ctx.cube );
|
|
cubehashUpdateDigest( &ctx.cube, (byte*) hash1, (const byte*) hash1, 32 );
|
|
memcpy( &ctx.cube, &l2v2_4way_ctx.cube, sizeof ctx.cube );
|
|
cubehashUpdateDigest( &ctx.cube, (byte*) hash2, (const byte*) hash2, 32 );
|
|
memcpy( &ctx.cube, &l2v2_4way_ctx.cube, sizeof ctx.cube );
|
|
cubehashUpdateDigest( &ctx.cube, (byte*) hash3, (const byte*) hash3, 32 );
|
|
|
|
|
|
// BMW256 4way has a lane corruption problem, only lanes 0 & 2 produce
|
|
// good hash. As a result this ugly workaround of running bmw256-4way
|
|
// twice with data shuffled to get all 4 lanes of good hash.
|
|
// The hash is then shuffled back into the appropriate lanes for output.
|
|
// Not as fast but still faster than using sph serially.
|
|
|
|
// shift lane 1 data to lane 2.
|
|
mm_interleave_4x32( vhash, hash0, hash0, hash1, hash1, 256 );
|
|
bmw256_4way( &ctx.bmw, vhash, 32 );
|
|
bmw256_4way_close( &ctx.bmw, vhash );
|
|
uint32_t trash[8] __attribute__ ((aligned (32)));
|
|
// extract lane 0 as usual and lane2 containing lane 1 hash
|
|
mm_deinterleave_4x32( state, trash, state+32, trash, vhash, 256 );
|
|
// shift lane2 data to lane 0 and lane 3 data to lane 2
|
|
mm_interleave_4x32( vhash, hash2, hash2, hash3, hash3, 256 );
|
|
bmw256_4way_init( &ctx.bmw );
|
|
bmw256_4way( &ctx.bmw, vhash, 32 );
|
|
bmw256_4way_close( &ctx.bmw, vhash );
|
|
// extract lane 2 hash from lane 0 and lane 3 hash from lane 2.
|
|
mm_deinterleave_4x32( state+64, trash, state+96, trash, vhash, 256 );
|
|
}
|
|
|
|
int scanhash_lyra2rev2_4way( int thr_id, struct work *work, uint32_t max_nonce,
|
|
uint64_t *hashes_done )
|
|
{
|
|
uint32_t hash[8*4] __attribute__ ((aligned (64)));
|
|
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
|
|
uint32_t _ALIGN(64) edata[20];
|
|
uint32_t *pdata = work->data;
|
|
uint32_t *ptarget = work->target;
|
|
const uint32_t first_nonce = pdata[19];
|
|
uint32_t n = first_nonce;
|
|
const uint32_t Htarg = ptarget[7];
|
|
uint32_t *nonces = work->nonces;
|
|
bool *found = work->nfound;
|
|
int num_found = 0;
|
|
uint32_t *noncep0 = vdata + 76; // 19*4
|
|
uint32_t *noncep1 = vdata + 77;
|
|
uint32_t *noncep2 = vdata + 78;
|
|
uint32_t *noncep3 = vdata + 79;
|
|
|
|
if ( opt_benchmark )
|
|
( (uint32_t*)ptarget )[7] = 0x0000ff;
|
|
|
|
swab32_array( edata, pdata, 20 );
|
|
|
|
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
|
|
|
|
blake256_4way_init( &l2v2_4way_ctx.blake );
|
|
blake256_4way( &l2v2_4way_ctx.blake, vdata, 64 );
|
|
|
|
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 );
|
|
|
|
lyra2rev2_4way_hash( hash, vdata );
|
|
pdata[19] = n;
|
|
|
|
if ( hash[7] <= Htarg && fulltest( hash, ptarget ) )
|
|
{
|
|
//printf("found0\n");
|
|
found[0] = true;
|
|
num_found++;
|
|
nonces[0] = pdata[19] = n;
|
|
work_set_target_ratio( work, hash );
|
|
}
|
|
if ( (hash+8)[7] <= Htarg && fulltest( hash+8, ptarget ) )
|
|
{
|
|
//printf("found1\n");
|
|
found[1] = true;
|
|
num_found++;
|
|
nonces[1] = n+1;
|
|
work_set_target_ratio( work, hash+8 );
|
|
}
|
|
if ( (hash+16)[7] <= Htarg && fulltest( hash+16, ptarget ) )
|
|
{
|
|
//printf("found2\n");
|
|
found[2] = true;
|
|
num_found++;
|
|
nonces[2] = n+2;
|
|
work_set_target_ratio( work, hash+16 );
|
|
}
|
|
if ( (hash+24)[7] <= Htarg && fulltest( hash+24, ptarget ) )
|
|
{
|
|
//printf("found3\n");
|
|
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-4)
|
|
&& !work_restart[thr_id].restart);
|
|
|
|
*hashes_done = n - first_nonce + 1;
|
|
return num_found;
|
|
}
|
|
|
|
#endif
|