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

v3.8.1

Browse Source
This commit is contained in:
Jay D Dee
2018-02-07 16:38:45 -05:00
parent 54b8fd7362
commit a28daca3ce
83 changed files with 5153 additions and 3924 deletions
Download Patch File Download Diff File Expand all files Collapse all files

130
algo/qubit/deep-2way.c Normal file
View File

@@ -0,0 +1,130 @@
#include "deep-gate.h"
#if defined(DEEP_2WAY)
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/echo/aes_ni/hash_api.h"
typedef struct
{
luffa_2way_context luffa;
cubehashParam cube;
sph_shavite512_context shavite;
hashState_echo echo;
} deep_2way_ctx_holder;
deep_2way_ctx_holder deep_2way_ctx;
void init_deep_2way_ctx()
{
luffa_2way_init( &deep_2way_ctx.luffa, 512 );
cubehashInit(&deep_2way_ctx.cube,512,16,32);
sph_shavite512_init(&deep_2way_ctx.shavite);
init_echo(&deep_2way_ctx.echo, 512);
};
void deep_2way_hash( void *output, const void *input )
{
uint64_t hash0[8] __attribute__ ((aligned (64)));
uint64_t hash1[8] __attribute__ ((aligned (64)));
uint64_t vhash[8*2] __attribute__ ((aligned (64)));
deep_2way_ctx_holder ctx;
memcpy( &ctx, &deep_2way_ctx, sizeof(deep_2way_ctx) );
luffa_2way_update( &ctx.luffa, input + (64<<1), 16 );
luffa_2way_close( &ctx.luffa, vhash );
mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
cubehashUpdateDigest( &ctx.cube, (byte*)hash0,
(const byte*) hash0, 64 );
memcpy( &ctx.cube, &deep_2way_ctx.cube, sizeof(cubehashParam) );
cubehashUpdateDigest( &ctx.cube, (byte*)hash1, (const byte*) hash1, 64 );
sph_shavite512( &ctx.shavite, hash0, 64 );
sph_shavite512_close( &ctx.shavite, hash0 );
memcpy( &ctx.shavite, &deep_2way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash1, 64 );
sph_shavite512_close( &ctx.shavite, hash1 );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
memcpy( &ctx.echo, &deep_2way_ctx.echo, sizeof(hashState_echo) );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
memcpy( output, hash0, 32 );
memcpy( output+32, hash1, 32 );
}
int scanhash_deep_2way( 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 + 32+3; // 4*8 + 3
uint32_t *noncep1 = vdata + 32+7;
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_2x128( (uint64_t*)vdata, edata, edata, 640 );
luffa_2way_init( &deep_2way_ctx.luffa, 512 );
luffa_2way_update( &deep_2way_ctx.luffa, vdata, 64 );
for ( int m=0; m < 6; m++ ) if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
do
{
found[0] = found[1] = false;
be32enc( noncep0, n );
be32enc( noncep1, n+1 );
deep_2way_hash( hash, vdata );
pdata[19] = n;
if ( !( hash[7] & mask ) && fulltest( hash, ptarget) )
{
found[0] = true;
num_found++;
nonces[0] = n;
work_set_target_ratio( work, hash );
}
if ( !( (hash+8)[7] & mask ) && fulltest( hash+8, ptarget) )
{
found[1] = true;
num_found++;
nonces[1] = n+1;
work_set_target_ratio( work, hash+64 );
}
n += 2;
} while ( ( num_found == 0 ) && ( n < max_nonce )
&& !work_restart[thr_id].restart );
break;
}
*hashes_done = n - first_nonce + 1;
return num_found;
}
#endif

17
algo/qubit/deep-gate.c Normal file
View File

@@ -0,0 +1,17 @@
#include "deep-gate.h"
bool register_deep_algo( algo_gate_t* gate )
{
#if defined (DEEP_2WAY)
init_deep_2way_ctx();
gate->scanhash = (void*)&scanhash_deep_2way;
gate->hash = (void*)&deep_2way_hash;
#else
init_deep_ctx();
gate->scanhash = (void*)&scanhash_deep;
gate->hash = (void*)&deep_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX_OPT | AVX2_OPT;
return true;
};

32
algo/qubit/deep-gate.h Normal file
View File

@@ -0,0 +1,32 @@
#ifndef DEEP_GATE_H__
#define DEEP_GATE_H__ 1
#include "algo-gate-api.h"
#include <stdint.h>
#if defined(__AVX2__) && defined(__AES__)
#define DEEP_2WAY
#endif
bool register_deep_algo( algo_gate_t* gate );
#if defined(DEEP_2WAY)
void deep_2way_hash( void *state, const void *input );
int scanhash_deep_2way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
void init_deep_2way_ctx();
#endif
void deep_hash( void *state, const void *input );
int scanhash_deep( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
void init_deep_ctx();
#endif

13
algo/qubit/deep.c
View File

@@ -1,9 +1,9 @@
#include "algo-gate-api.h"
#include "deep-gate.h"
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "algo/luffa/sse2/luffa_for_sse2.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#ifndef NO_AES_NI
#include "algo/echo/aes_ni/hash_api.h"
@@ -139,12 +139,3 @@ int scanhash_deep( int thr_id, struct work *work, uint32_t max_nonce,
return 0;
}
bool register_deep_algo( algo_gate_t* gate )
{
gate->optimizations = SSE2_OPT | AES_OPT | AVX_OPT | AVX2_OPT;
init_deep_ctx();
gate->scanhash = (void*)&scanhash_deep;
gate->hash = (void*)&deep_hash;
return true;
};

138
algo/qubit/qubit-2way.c Normal file
View File

@@ -0,0 +1,138 @@
#include "qubit-gate.h"
#if defined(QUBIT_2WAY)
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/echo/aes_ni/hash_api.h"
typedef struct
{
luffa_2way_context luffa;
cubehashParam cube;
sph_shavite512_context shavite;
simd_2way_context simd;
hashState_echo echo;
} qubit_2way_ctx_holder;
qubit_2way_ctx_holder qubit_2way_ctx;
void init_qubit_2way_ctx()
{
luffa_2way_init( &qubit_2way_ctx.luffa, 512 );
cubehashInit(&qubit_2way_ctx.cube,512,16,32);
sph_shavite512_init(&qubit_2way_ctx.shavite);
simd_2way_init( &qubit_2way_ctx.simd, 512 );
init_echo(&qubit_2way_ctx.echo, 512);
};
void qubit_2way_hash( void *output, const void *input )
{
uint64_t hash0[8] __attribute__ ((aligned (64)));
uint64_t hash1[8] __attribute__ ((aligned (64)));
uint64_t vhash[8*2] __attribute__ ((aligned (64)));
qubit_2way_ctx_holder ctx;
memcpy( &ctx, &qubit_2way_ctx, sizeof(qubit_2way_ctx) );
luffa_2way_update( &ctx.luffa, input + (64<<1), 16 );
luffa_2way_close( &ctx.luffa, vhash );
mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
cubehashUpdateDigest( &ctx.cube, (byte*)hash0,
(const byte*) hash0, 64 );
memcpy( &ctx.cube, &qubit_2way_ctx.cube, sizeof(cubehashParam) );
cubehashUpdateDigest( &ctx.cube, (byte*)hash1, (const byte*) hash1, 64 );
sph_shavite512( &ctx.shavite, hash0, 64 );
sph_shavite512_close( &ctx.shavite, hash0 );
memcpy( &ctx.shavite, &qubit_2way_ctx.shavite,
sizeof(sph_shavite512_context) );
sph_shavite512( &ctx.shavite, hash1, 64 );
sph_shavite512_close( &ctx.shavite, hash1 );
mm256_interleave_2x128( vhash, hash0, hash1, 512 );
simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
memcpy( &ctx.echo, &qubit_2way_ctx.echo, sizeof(hashState_echo) );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
memcpy( output, hash0, 32 );
memcpy( output+32, hash1, 32 );
}
int scanhash_qubit_2way( 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 + 32+3; // 4*8 + 3
uint32_t *noncep1 = vdata + 32+7;
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_2x128( (uint64_t*)vdata, edata, edata, 640 );
luffa_2way_init( &qubit_2way_ctx.luffa, 512 );
luffa_2way_update( &qubit_2way_ctx.luffa, vdata, 64 );
for ( int m=0; m < 6; m++ ) if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];
do
{
found[0] = found[1] = false;
be32enc( noncep0, n );
be32enc( noncep1, n+1 );
qubit_2way_hash( hash, vdata );
pdata[19] = n;
if ( !( hash[7] & mask ) && fulltest( hash, ptarget) )
{
found[0] = true;
num_found++;
nonces[0] = n;
work_set_target_ratio( work, hash );
}
if ( !( (hash+8)[7] & mask ) && fulltest( hash+8, ptarget) )
{
found[1] = true;
num_found++;
nonces[1] = n+1;
work_set_target_ratio( work, hash+8 );
}
n += 2;
} while ( ( num_found == 0 ) && ( n < max_nonce )
&& !work_restart[thr_id].restart );
break;
}
*hashes_done = n - first_nonce + 1;
return num_found;
}
#endif

17
algo/qubit/qubit-gate.c Normal file
View File

@@ -0,0 +1,17 @@
#include "qubit-gate.h"
bool register_qubit_algo( algo_gate_t* gate )
{
#if defined (QUBIT_2WAY)
init_qubit_2way_ctx();
gate->scanhash = (void*)&scanhash_qubit_2way;
gate->hash = (void*)&qubit_2way_hash;
#else
init_qubit_ctx();
gate->scanhash = (void*)&scanhash_qubit;
gate->hash = (void*)&qubit_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | AVX_OPT | AVX2_OPT;
return true;
};

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

@@ -0,0 +1,32 @@
#ifndef QUBIT_GATE_H__
#define QUBIT_GATE_H__ 1
#include "algo-gate-api.h"
#include <stdint.h>
#if defined(__AVX2__) && defined(__AES__)
#define QUBIT_2WAY
#endif
bool register_qubit_algo( algo_gate_t* gate );
#if defined(QUBIT_2WAY)
void qubit_2way_hash( void *state, const void *input );
int scanhash_qubit_2way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
void init_qubit_2way_ctx();
#endif
void qubit_hash( void *state, const void *input );
int scanhash_qubit( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
void init_qubit_ctx();
#endif

19
algo/qubit/qubit.c
View File

@@ -1,11 +1,11 @@
#include "algo-gate-api.h"
#include "qubit-gate.h"
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "algo/luffa/sse2/luffa_for_sse2.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/simd/sse2/nist.h"
#include "algo/simd/nist.h"
#include "algo/shavite/sph_shavite.h"
#ifndef NO_AES_NI
#include "algo/echo/aes_ni/hash_api.h"
@@ -48,7 +48,7 @@ void qubit_luffa_midstate( const void* input )
update_luffa( &qubit_luffa_mid, input, 64 );
}
void qubithash(void *output, const void *input)
void qubit_hash(void *output, const void *input)
{
unsigned char hash[128] __attribute((aligned(64)));
#define hashB hash+64
@@ -115,7 +115,7 @@ int scanhash_qubit(int thr_id, struct work *work,
{
pdata[19] = ++n;
be32enc(&endiandata[19], n);
qubithash(hash64, endiandata);
qubit_hash(hash64, endiandata);
#ifndef DEBUG_ALGO
if (!(hash64[7] & mask))
{
@@ -151,12 +151,3 @@ int scanhash_qubit(int thr_id, struct work *work,
return 0;
}
bool register_qubit_algo( algo_gate_t* gate )
{
gate->optimizations = SSE2_OPT | AES_OPT | AVX_OPT | AVX2_OPT;
init_qubit_ctx();
gate->scanhash = (void*)&scanhash_qubit;
gate->hash = (void*)&qubithash;
return true;
};