This commit is contained in:
Jay D Dee
2018-01-16 15:11:44 -05:00
parent bee78eac76
commit a90d75b8f5
77 changed files with 3408 additions and 1214 deletions

128
algo/lyra2/lyra2h-4way.c Normal file
View File

@@ -0,0 +1,128 @@
#include "lyra2h-gate.h"
#ifdef LYRA2H_4WAY
#include <memory.h>
#include <mm_malloc.h>
#include "lyra2.h"
#include "algo/blake/sph_blake.h"
#include "algo/blake/blake-hash-4way.h"
__thread uint64_t* lyra2h_4way_matrix;
bool lyra2h_4way_thread_init()
{
return ( lyra2h_4way_matrix = _mm_malloc( LYRA2H_MATRIX_SIZE, 64 ) );
}
static __thread blake256_4way_context l2h_4way_blake_mid;
void lyra2h_4way_midstate( const void* input )
{
blake256_4way_init( &l2h_4way_blake_mid );
blake256_4way( &l2h_4way_blake_mid, input, 64 );
}
void lyra2h_4way_hash( void *state, const void *input )
{
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 vhash[8*4] __attribute__ ((aligned (64)));
blake256_4way_context ctx_blake __attribute__ ((aligned (64)));
memcpy( &ctx_blake, &l2h_4way_blake_mid, sizeof l2h_4way_blake_mid );
blake256_4way( &ctx_blake, input + (64*4), 16 );
blake256_4way_close( &ctx_blake, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
LYRA2Z( lyra2h_4way_matrix, hash0, 32, hash0, 32, hash0, 32, 16, 16, 16 );
LYRA2Z( lyra2h_4way_matrix, hash1, 32, hash1, 32, hash1, 32, 16, 16, 16 );
LYRA2Z( lyra2h_4way_matrix, hash2, 32, hash2, 32, hash2, 32, 16, 16, 16 );
LYRA2Z( lyra2h_4way_matrix, hash3, 32, hash3, 32, hash3, 32, 16, 16, 16 );
memcpy( state, hash0, 32 );
memcpy( state+32, hash1, 32 );
memcpy( state+64, hash2, 32 );
memcpy( state+96, hash3, 32 );
}
int scanhash_lyra2h_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 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 + 76; // 19*4
uint32_t *noncep1 = vdata + 77;
uint32_t *noncep2 = vdata + 78;
uint32_t *noncep3 = vdata + 79;
if ( opt_benchmark )
ptarget[7] = 0x0000ff;
for ( int i=0; i < 19; i++ )
be32enc( &edata[i], pdata[i] );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
lyra2h_4way_midstate( vdata );
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 );
be32enc( &edata[19], n );
lyra2h_4way_hash( hash, vdata );
if ( hash[7] <= Htarg && fulltest( hash, ptarget ) )
{
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 ) )
{
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 ) )
{
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 ) )
{
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

25
algo/lyra2/lyra2h-gate.c Normal file
View File

@@ -0,0 +1,25 @@
#include "lyra2h-gate.h"
#include "lyra2.h"
void lyra2h_set_target( struct work* work, double job_diff )
{
work_set_target( work, job_diff / (256.0 * opt_diff_factor) );
}
bool register_lyra2h_algo( algo_gate_t* gate )
{
#ifdef LYRA2H_4WAY
gate->miner_thread_init = (void*)&lyra2h_4way_thread_init;
gate->scanhash = (void*)&scanhash_lyra2h_4way;
gate->hash = (void*)&lyra2h_4way_hash;
#else
gate->miner_thread_init = (void*)&lyra2h_thread_init;
gate->scanhash = (void*)&scanhash_lyra2h;
gate->hash = (void*)&lyra2h_hash;
#endif
gate->optimizations = AVX_OPT | AVX2_OPT | FOUR_WAY_OPT;
gate->get_max64 = (void*)&get_max64_0xffffLL;
gate->set_target = (void*)&lyra2h_set_target;
return true;
};

32
algo/lyra2/lyra2h-gate.h Normal file
View File

@@ -0,0 +1,32 @@
#ifndef LYRA2H_GATE_H__
#define LYRA2H_GATE_H__
#include "algo-gate-api.h"
#include <stdint.h>
#if defined(HASH_4WAY)
#define LYRA2H_4WAY
#endif
#define LYRA2H_MATRIX_SIZE BLOCK_LEN_INT64 * 16 * 16 * 8
#if defined(LYRA2H_4WAY)
void lyra2h_4way_hash( void *state, const void *input );
int scanhash_lyra2h_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool lyra2h_4way_thread_init();
#endif
void lyra2h_hash( void *state, const void *input );
int scanhash_lyra2h( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool lyra2h_thread_init();
#endif

View File

@@ -1,6 +1,6 @@
#include "lyra2h-gate.h"
#include <memory.h>
#include <mm_malloc.h>
#include "algo-gate-api.h"
#include "lyra2.h"
#include "algo/blake/sph_blake.h"
@@ -8,8 +8,7 @@ __thread uint64_t* lyra2h_matrix;
bool lyra2h_thread_init()
{
const int i = 16 * 16 * 96;
lyra2h_matrix = _mm_malloc( i, 64 );
lyra2h_matrix = _mm_malloc( LYRA2H_MATRIX_SIZE, 64 );
return lyra2h_matrix;
}
@@ -74,20 +73,3 @@ int scanhash_lyra2h( int thr_id, struct work *work, uint32_t max_nonce,
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
void lyra2h_set_target( struct work* work, double job_diff )
{
work_set_target( work, job_diff / (256.0 * opt_diff_factor) );
}
bool register_lyra2h_algo( algo_gate_t* gate )
{
gate->optimizations = AVX_OPT | AVX2_OPT;
gate->miner_thread_init = (void*)&lyra2h_thread_init;
gate->scanhash = (void*)&scanhash_lyra2h;
gate->hash = (void*)&lyra2h_hash;
gate->get_max64 = (void*)&get_max64_0xffffLL;
gate->set_target = (void*)&lyra2h_set_target;
return true;
};

View File

@@ -106,6 +106,7 @@ int scanhash_lyra2re(int thr_id, struct work *work,
{
pdata[19] = nonce;
*hashes_done = pdata[19] - first_nonce;
work_set_target_ratio( work, hash );
return 1;
}
}

177
algo/lyra2/lyra2rev2-4way.c Normal file
View File

@@ -0,0 +1,177 @@
#include "lyra2rev2-gate.h"
#include <memory.h>
#ifdef __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;
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 );
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 );
sph_bmw256( &ctx.bmw, hash0, 32 );
sph_bmw256_close( &ctx.bmw, hash0 );
memcpy( &ctx.bmw, &l2v2_4way_ctx.bmw, sizeof ctx.bmw );
sph_bmw256( &ctx.bmw, hash1, 32 );
sph_bmw256_close( &ctx.bmw, hash1 );
memcpy( &ctx.bmw, &l2v2_4way_ctx.bmw, sizeof ctx.bmw );
sph_bmw256( &ctx.bmw, hash2, 32 );
sph_bmw256_close( &ctx.bmw, hash2 );
memcpy( &ctx.bmw, &l2v2_4way_ctx.bmw, sizeof ctx.bmw );
sph_bmw256( &ctx.bmw, hash3, 32 );
sph_bmw256_close( &ctx.bmw, hash3 );
memcpy( state, hash0, 32 );
memcpy( state+32, hash1, 32 );
memcpy( state+64, hash2, 32 );
memcpy( state+96, hash3, 32 );
}
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 ) )
{
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 ) )
{
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 ) )
{
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 ) )
{
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

View File

@@ -0,0 +1,38 @@
#include "lyra2rev2-gate.h"
__thread uint64_t* l2v2_wholeMatrix;
void lyra2rev2_set_target( struct work* work, double job_diff )
{
work_set_target( work, job_diff / (256.0 * opt_diff_factor) );
}
bool lyra2rev2_thread_init()
{
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * 4; // nCols
const int64_t ROW_LEN_BYTES = ROW_LEN_INT64 * 8;
int i = (int64_t)ROW_LEN_BYTES * 4; // nRows;
l2v2_wholeMatrix = _mm_malloc( i, 64 );
return l2v2_wholeMatrix;
}
bool register_lyra2rev2_algo( algo_gate_t* gate )
{
#if defined (LYRA2REV2_4WAY)
init_lyra2rev2_4way_ctx();
gate->scanhash = (void*)&scanhash_lyra2rev2_4way;
gate->hash = (void*)&lyra2rev2_4way_hash;
#else
init_lyra2rev2_ctx();
gate->scanhash = (void*)&scanhash_lyra2rev2;
gate->hash = (void*)&lyra2rev2_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX_OPT | AVX2_OPT | FOUR_WAY_OPT;
gate->miner_thread_init = (void*)&lyra2rev2_thread_init;
gate->set_target = (void*)&lyra2rev2_set_target;
return true;
};

View File

@@ -0,0 +1,35 @@
#ifndef LYRA2REV2_GATE_H__
#define LYRA2REV2_GATE_H__ 1
#include "algo-gate-api.h"
#include <stdint.h>
#include "lyra2.h"
#if defined(HASH_4WAY)
#define LYRA2REV2_4WAY
#endif
extern __thread uint64_t* l2v2_wholeMatrix;
bool register_lyra2rev2_algo( algo_gate_t* gate );
#if defined(LYRA2REV2_4WAY)
void lyra2rev2_4way_hash( void *state, const void *input );
int scanhash_lyra2rev2_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
void init_lyra2rev2_4way_ctx();
#endif
void lyra2rev2_hash( void *state, const void *input );
int scanhash_lyra2rev2( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
void init_lyra2rev2_ctx();
#endif

View File

@@ -1,20 +1,12 @@
#include "lyra2rev2-gate.h"
#include <memory.h>
#include "algo-gate-api.h"
#include "algo/blake/sph_blake.h"
#include "algo/cubehash/sph_cubehash.h"
#include "algo/keccak/sph_keccak.h"
#include "algo/skein/sph_skein.h"
#include "algo/bmw/sph_bmw.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "lyra2.h"
#include "avxdefs.h"
// This gets allocated when miner_thread starts up and is never freed.
// It's not a leak because the only way to allocate it again is to exit
// the thread and that only occurs when the entire program exits.
__thread uint64_t* l2v2_wholeMatrix;
//#include "lyra2.h"
typedef struct {
cubehashParam cube1;
@@ -106,6 +98,7 @@ int scanhash_lyra2rev2(int thr_id, struct work *work,
if( fulltest(hash, ptarget) )
{
pdata[19] = nonce;
work_set_target_ratio( work, hash );
*hashes_done = pdata[19] - first_nonce;
return 1;
}
@@ -119,30 +112,3 @@ int scanhash_lyra2rev2(int thr_id, struct work *work,
return 0;
}
void lyra2rev2_set_target( struct work* work, double job_diff )
{
work_set_target( work, job_diff / (256.0 * opt_diff_factor) );
}
bool lyra2rev2_thread_init()
{
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * 4; // nCols
const int64_t ROW_LEN_BYTES = ROW_LEN_INT64 * 8;
int i = (int64_t)ROW_LEN_BYTES * 4; // nRows;
l2v2_wholeMatrix = _mm_malloc( i, 64 );
return l2v2_wholeMatrix;
}
bool register_lyra2rev2_algo( algo_gate_t* gate )
{
init_lyra2rev2_ctx();
gate->optimizations = AVX_OPT | AVX2_OPT;
gate->miner_thread_init = (void*)&lyra2rev2_thread_init;
gate->scanhash = (void*)&scanhash_lyra2rev2;
gate->hash = (void*)&lyra2rev2_hash;
gate->set_target = (void*)&lyra2rev2_set_target;
return true;
};

View File

@@ -85,8 +85,8 @@ int scanhash_lyra2z_4way( int thr_id, struct work *work, uint32_t max_nonce,
be32enc( noncep2, n+2 );
be32enc( noncep3, n+3 );
be32enc( &edata[19], n );
lyra2z_4way_hash( hash, vdata );
pdata[19] = n;
if ( hash[7] <= Htarg && fulltest( hash, ptarget ) )
{

View File

@@ -82,41 +82,3 @@ int scanhash_lyra2z( int thr_id, struct work *work, uint32_t max_nonce,
return 0;
}
/*
//int64_t get_max64_0xffffLL() { return 0xffffLL; };
void lyra2z_set_target( struct work* work, double job_diff )
{
work_set_target( work, job_diff / (256.0 * opt_diff_factor) );
}
bool zcoin_get_work_height( struct work* work, struct stratum_ctx* sctx )
{
work->height = sctx->bloc_height;
return false;
}
bool lyra2z_thread_init()
{
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * 8; // nCols
const int64_t ROW_LEN_BYTES = ROW_LEN_INT64 * 8;
int i = (int64_t)ROW_LEN_BYTES * 8; // nRows;
lyra2z_wholeMatrix = _mm_malloc( i, 64 );
return lyra2z_wholeMatrix;
}
bool register_lyra2z_algo( algo_gate_t* gate )
{
gate->optimizations = SSE2_OPT | AES_OPT | AVX_OPT | AVX2_OPT;
gate->miner_thread_init = (void*)&lyra2z_thread_init;
gate->scanhash = (void*)&scanhash_lyra2z;
gate->hash = (void*)&lyra2z_hash;
gate->get_max64 = (void*)&get_max64_0xffffLL;
gate->set_target = (void*)&lyra2z_set_target;
// gate->prevent_dupes = (void*)&zcoin_get_work_height;
return true;
};
*/