mirror of
https://github.com/JayDDee/cpuminer-opt.git
synced 2025-09-17 23:44:27 +00:00
v3.12.2
This commit is contained in:
@@ -7,27 +7,8 @@
|
||||
#include "algo/cubehash/cubehash_sse2.h"
|
||||
#include "algo/cubehash/cube-hash-2way.h"
|
||||
|
||||
#if defined (LYRA2REV2_8WAY)
|
||||
|
||||
typedef struct {
|
||||
blake256_8way_context blake;
|
||||
keccak256_8way_context keccak;
|
||||
cube_4way_context cube;
|
||||
skein256_8way_context skein;
|
||||
bmw256_8way_context bmw;
|
||||
} lyra2v2_8way_ctx_holder __attribute__ ((aligned (64)));
|
||||
|
||||
static lyra2v2_8way_ctx_holder l2v2_8way_ctx;
|
||||
|
||||
bool init_lyra2rev2_8way_ctx()
|
||||
{
|
||||
keccak256_8way_init( &l2v2_8way_ctx.keccak );
|
||||
cube_4way_init( &l2v2_8way_ctx.cube, 256, 16, 32 );
|
||||
skein256_8way_init( &l2v2_8way_ctx.skein );
|
||||
bmw256_8way_init( &l2v2_8way_ctx.bmw );
|
||||
return true;
|
||||
}
|
||||
|
||||
#if 0
|
||||
void lyra2rev2_8way_hash( void *state, const void *input )
|
||||
{
|
||||
uint32_t vhash[8*8] __attribute__ ((aligned (128)));
|
||||
@@ -52,14 +33,24 @@ void lyra2rev2_8way_hash( void *state, const void *input )
|
||||
keccak256_8way_update( &ctx.keccak, vhashA, 32 );
|
||||
keccak256_8way_close( &ctx.keccak, vhash );
|
||||
|
||||
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 256 );
|
||||
dintrlv_8x64( hash0, hash1, hash2, hash3,
|
||||
hash4, hash5, hash6, hash7, vhash, 256 );
|
||||
|
||||
cube_4way_update_close( &ctx.cube, vhashA, vhashA, 32 );
|
||||
cube_4way_init( &ctx.cube, 256, 16, 32 );
|
||||
cube_4way_update_close( &ctx.cube, vhashB, vhashB, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash0, 256, (const byte*) hash0, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash1, 256, (const byte*) hash1, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash2, 256, (const byte*) hash2, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash3, 256, (const byte*) hash3, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash4, 256, (const byte*) hash4, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash5, 256, (const byte*) hash5, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash6, 256, (const byte*) hash6, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash7, 256, (const byte*) hash7, 32 );
|
||||
|
||||
dintrlv_4x128( hash0, hash1, hash2, hash3, vhashA, 256 );
|
||||
dintrlv_4x128( hash4, hash5, hash6, hash7, vhashB, 256 );
|
||||
// cube_4way_update_close( &ctx.cube, vhashA, vhashA, 32 );
|
||||
// cube_4way_init( &ctx.cube, 256, 16, 32 );
|
||||
// cube_4way_update_close( &ctx.cube, vhashB, vhashB, 32 );
|
||||
//
|
||||
// dintrlv_4x128( hash0, hash1, hash2, hash3, vhashA, 256 );
|
||||
// dintrlv_4x128( hash4, hash5, hash6, hash7, vhashB, 256 );
|
||||
|
||||
intrlv_2x256( vhash, hash0, hash1, 256 );
|
||||
LYRA2REV2_2WAY( l2v2_wholeMatrix, vhash, 32, vhash, 32, 1, 4, 4 );
|
||||
@@ -80,15 +71,123 @@ void lyra2rev2_8way_hash( void *state, const void *input )
|
||||
skein256_8way_update( &ctx.skein, vhash, 32 );
|
||||
skein256_8way_close( &ctx.skein, vhash );
|
||||
|
||||
rintrlv_8x64_4x128( vhashA, vhashB, vhash, 256 );
|
||||
dintrlv_8x64( hash0, hash1, hash2, hash3,
|
||||
hash4, hash5, hash6, hash7, vhash, 256 );
|
||||
|
||||
cube_4way_init( &ctx.cube, 256, 16, 32 );
|
||||
cube_4way_update_close( &ctx.cube, vhashA, vhashA, 32 );
|
||||
cube_4way_init( &ctx.cube, 256, 16, 32 );
|
||||
cube_4way_update_close( &ctx.cube, vhashB, vhashB, 32 );
|
||||
|
||||
dintrlv_4x128( hash0, hash1, hash2, hash3, vhashA, 256 );
|
||||
dintrlv_4x128( hash4, hash5, hash6, hash7, vhashB, 256 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash0, 256, (const byte*) hash0, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash1, 256, (const byte*) hash1, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash2, 256, (const byte*) hash2, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash3, 256, (const byte*) hash3, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash4, 256, (const byte*) hash4, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash5, 256, (const byte*) hash5, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash6, 256, (const byte*) hash6, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash7, 256, (const byte*) hash7, 32 );
|
||||
|
||||
// cube_4way_init( &ctx.cube, 256, 16, 32 );
|
||||
// cube_4way_update_close( &ctx.cube, vhashA, vhashA, 32 );
|
||||
// cube_4way_init( &ctx.cube, 256, 16, 32 );
|
||||
// cube_4way_update_close( &ctx.cube, vhashB, vhashB, 32 );
|
||||
//
|
||||
// dintrlv_4x128( hash0, hash1, hash2, hash3, vhashA, 256 );
|
||||
// dintrlv_4x128( hash4, hash5, hash6, hash7, vhashB, 256 );
|
||||
|
||||
intrlv_8x32( vhash, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
|
||||
hash7, 256 );
|
||||
|
||||
bmw256_8way_update( &ctx.bmw, vhash, 32 );
|
||||
bmw256_8way_close( &ctx.bmw, state );
|
||||
}
|
||||
#endif
|
||||
|
||||
|
||||
|
||||
|
||||
#if defined (LYRA2REV2_8WAY)
|
||||
|
||||
typedef struct {
|
||||
blake256_8way_context blake;
|
||||
keccak256_8way_context keccak;
|
||||
cubehashParam cube;
|
||||
skein256_8way_context skein;
|
||||
bmw256_8way_context bmw;
|
||||
} lyra2v2_8way_ctx_holder __attribute__ ((aligned (64)));
|
||||
|
||||
static lyra2v2_8way_ctx_holder l2v2_8way_ctx;
|
||||
|
||||
bool init_lyra2rev2_8way_ctx()
|
||||
{
|
||||
keccak256_8way_init( &l2v2_8way_ctx.keccak );
|
||||
cubehashInit( &l2v2_8way_ctx.cube, 256, 16, 32 );
|
||||
skein256_8way_init( &l2v2_8way_ctx.skein );
|
||||
bmw256_8way_init( &l2v2_8way_ctx.bmw );
|
||||
return true;
|
||||
}
|
||||
|
||||
void lyra2rev2_8way_hash( void *state, const void *input )
|
||||
{
|
||||
uint32_t vhash[8*8] __attribute__ ((aligned (128)));
|
||||
uint32_t vhashA[8*8] __attribute__ ((aligned (64)));
|
||||
uint32_t hash0[8] __attribute__ ((aligned (64)));
|
||||
uint32_t hash1[8] __attribute__ ((aligned (64)));
|
||||
uint32_t hash2[8] __attribute__ ((aligned (64)));
|
||||
uint32_t hash3[8] __attribute__ ((aligned (64)));
|
||||
uint32_t hash4[8] __attribute__ ((aligned (64)));
|
||||
uint32_t hash5[8] __attribute__ ((aligned (64)));
|
||||
uint32_t hash6[8] __attribute__ ((aligned (64)));
|
||||
uint32_t hash7[8] __attribute__ ((aligned (64)));
|
||||
lyra2v2_8way_ctx_holder ctx __attribute__ ((aligned (64)));
|
||||
memcpy( &ctx, &l2v2_8way_ctx, sizeof(l2v2_8way_ctx) );
|
||||
|
||||
blake256_8way_update( &ctx.blake, input + (64<<3), 16 );
|
||||
blake256_8way_close( &ctx.blake, vhash );
|
||||
|
||||
rintrlv_8x32_8x64( vhashA, vhash, 256 );
|
||||
|
||||
keccak256_8way_update( &ctx.keccak, vhashA, 32 );
|
||||
keccak256_8way_close( &ctx.keccak, vhash );
|
||||
|
||||
dintrlv_8x64( hash0, hash1, hash2, hash3,
|
||||
hash4, hash5, hash6, hash7, vhash, 256 );
|
||||
|
||||
cubehash_full( &ctx.cube, (byte*) hash0, 256, (const byte*) hash0, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash1, 256, (const byte*) hash1, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash2, 256, (const byte*) hash2, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash3, 256, (const byte*) hash3, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash4, 256, (const byte*) hash4, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash5, 256, (const byte*) hash5, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash6, 256, (const byte*) hash6, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash7, 256, (const byte*) hash7, 32 );
|
||||
|
||||
intrlv_2x256( vhash, hash0, hash1, 256 );
|
||||
LYRA2REV2_2WAY( l2v2_wholeMatrix, vhash, 32, vhash, 32, 1, 4, 4 );
|
||||
dintrlv_2x256( hash0, hash1, vhash, 256 );
|
||||
intrlv_2x256( vhash, hash2, hash3, 256 );
|
||||
LYRA2REV2_2WAY( l2v2_wholeMatrix, vhash, 32, vhash, 32, 1, 4, 4 );
|
||||
dintrlv_2x256( hash2, hash3, vhash, 256 );
|
||||
intrlv_2x256( vhash, hash4, hash5, 256 );
|
||||
LYRA2REV2_2WAY( l2v2_wholeMatrix, vhash, 32, vhash, 32, 1, 4, 4 );
|
||||
dintrlv_2x256( hash4, hash5, vhash, 256 );
|
||||
intrlv_2x256( vhash, hash6, hash7, 256 );
|
||||
LYRA2REV2_2WAY( l2v2_wholeMatrix, vhash, 32, vhash, 32, 1, 4, 4 );
|
||||
dintrlv_2x256( hash6, hash7, vhash, 256 );
|
||||
|
||||
intrlv_8x64( vhash, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
|
||||
hash7, 256 );
|
||||
|
||||
skein256_8way_update( &ctx.skein, vhash, 32 );
|
||||
skein256_8way_close( &ctx.skein, vhash );
|
||||
|
||||
dintrlv_8x64( hash0, hash1, hash2, hash3,
|
||||
hash4, hash5, hash6, hash7, vhash, 256 );
|
||||
|
||||
cubehash_full( &ctx.cube, (byte*) hash0, 256, (const byte*) hash0, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash1, 256, (const byte*) hash1, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash2, 256, (const byte*) hash2, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash3, 256, (const byte*) hash3, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash4, 256, (const byte*) hash4, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash5, 256, (const byte*) hash5, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash6, 256, (const byte*) hash6, 32 );
|
||||
cubehash_full( &ctx.cube, (byte*) hash7, 256, (const byte*) hash7, 32 );
|
||||
|
||||
intrlv_8x32( vhash, hash0, hash1, hash2, hash3, hash4, hash5, hash6,
|
||||
hash7, 256 );
|
||||
@@ -97,49 +196,49 @@ void lyra2rev2_8way_hash( void *state, const void *input )
|
||||
bmw256_8way_close( &ctx.bmw, state );
|
||||
}
|
||||
|
||||
int scanhash_lyra2rev2_8way( struct work *work, uint32_t max_nonce,
|
||||
int scanhash_lyra2rev2_8way( struct work *work, const uint32_t max_nonce,
|
||||
uint64_t *hashes_done, struct thr_info *mythr )
|
||||
{
|
||||
uint32_t hash[8*8] __attribute__ ((aligned (128)));
|
||||
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
|
||||
uint32_t *hash7 = &(hash[7<<3]);
|
||||
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
|
||||
uint32_t *hashd7 = &hash[7*8];
|
||||
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t last_nonce = max_nonce - 8;
|
||||
uint32_t n = first_nonce;
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
__m256i *noncev = (__m256i*)vdata + 19; // aligned
|
||||
int thr_id = mythr->id;
|
||||
const uint32_t targ32 = ptarget[7];
|
||||
__m256i *noncev = (__m256i*)vdata + 19;
|
||||
const int thr_id = mythr->id;
|
||||
const bool bench = opt_benchmark;
|
||||
|
||||
if ( opt_benchmark )
|
||||
( (uint32_t*)ptarget )[7] = 0x0000ff;
|
||||
if ( bench ) ptarget[7] = 0x0000ff;
|
||||
|
||||
mm256_bswap32_intrlv80_8x32( vdata, pdata );
|
||||
|
||||
*noncev = _mm256_set_epi32( n+7, n+6, n+5, n+4, n+3, n+2, n+1, n );
|
||||
blake256_8way_init( &l2v2_8way_ctx.blake );
|
||||
blake256_8way_update( &l2v2_8way_ctx.blake, vdata, 64 );
|
||||
|
||||
do
|
||||
{
|
||||
*noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,
|
||||
n+3, n+2, n+1, n ) );
|
||||
|
||||
lyra2rev2_8way_hash( hash, vdata );
|
||||
pdata[19] = n;
|
||||
|
||||
for ( int lane = 0; lane < 8; lane++ ) if ( hash7[lane] <= Htarg )
|
||||
for ( int lane = 0; lane < 8; lane++ )
|
||||
if ( unlikely( hashd7[lane] <= targ32 ) )
|
||||
{
|
||||
extr_lane_8x32( lane_hash, hash, lane, 256 );
|
||||
if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
|
||||
if ( likely( valid_hash( lane_hash, ptarget ) && !bench ) )
|
||||
{
|
||||
pdata[19] = n + lane;
|
||||
submit_lane_solution( work, lane_hash, mythr, lane );
|
||||
pdata[19] = bswap_32( n + lane );
|
||||
submit_lane_solution( work, lane_hash, mythr, lane );
|
||||
}
|
||||
}
|
||||
*noncev = _mm256_add_epi32( *noncev, m256_const1_32( 8 ) );
|
||||
n += 8;
|
||||
} while ( (n < last_nonce) && !work_restart[thr_id].restart);
|
||||
} while ( likely( (n < last_nonce) && !work_restart[thr_id].restart ) );
|
||||
pdata[19] = n;
|
||||
*hashes_done = n - first_nonce;
|
||||
return 0;
|
||||
}
|
||||
@@ -226,15 +325,16 @@ int scanhash_lyra2rev2_4way( struct work *work, uint32_t max_nonce,
|
||||
{
|
||||
uint32_t hash[8*4] __attribute__ ((aligned (64)));
|
||||
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
|
||||
uint32_t *hash7 = &(hash[7<<2]);
|
||||
uint32_t *hashd7 = &(hash[7<<2]);
|
||||
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
|
||||
uint32_t *pdata = work->data;
|
||||
uint32_t *ptarget = work->target;
|
||||
const uint32_t first_nonce = pdata[19];
|
||||
const uint32_t last_nonce = max_nonce - 4;
|
||||
uint32_t n = first_nonce;
|
||||
const uint32_t Htarg = ptarget[7];
|
||||
__m128i *noncev = (__m128i*)vdata + 19; // aligned
|
||||
int thr_id = mythr->id; // thr_id arg is deprecated
|
||||
const uint32_t targ32 = ptarget[7];
|
||||
__m128i *noncev = (__m128i*)vdata + 19;
|
||||
int thr_id = mythr->id;
|
||||
|
||||
if ( opt_benchmark )
|
||||
( (uint32_t*)ptarget )[7] = 0x0000ff;
|
||||
@@ -249,20 +349,20 @@ int scanhash_lyra2rev2_4way( struct work *work, uint32_t max_nonce,
|
||||
*noncev = mm128_bswap_32( _mm_set_epi32( n+3, n+2, n+1, n ) );
|
||||
|
||||
lyra2rev2_4way_hash( hash, vdata );
|
||||
pdata[19] = n;
|
||||
|
||||
for ( int lane = 0; lane < 4; lane++ ) if ( hash7[lane] <= Htarg )
|
||||
for ( int lane = 0; lane < 4; lane++ ) if ( hashd7[lane] <= targ32 )
|
||||
{
|
||||
extr_lane_4x32( lane_hash, hash, lane, 256 );
|
||||
if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
|
||||
if ( valid_hash( lane_hash, ptarget ) && !opt_benchmark )
|
||||
{
|
||||
pdata[19] = n + lane;
|
||||
submit_lane_solution( work, lane_hash, mythr, lane );
|
||||
}
|
||||
}
|
||||
n += 4;
|
||||
} while ( (n < max_nonce-4) && !work_restart[thr_id].restart);
|
||||
*hashes_done = n - first_nonce + 1;
|
||||
} while ( (n < last_nonce) && !work_restart[thr_id].restart);
|
||||
pdata[19] = n;
|
||||
*hashes_done = n - first_nonce;
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
Reference in New Issue
Block a user