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

v3.9.4

Browse Source
This commit is contained in:
Jay D Dee
2019-06-18 13:15:45 -04:00
parent 71d6b97ee8
commit d6e8d7a46e
75 changed files with 1790 additions and 781 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
algo/x14/axiom.c
View File

@@ -44,12 +44,13 @@ void axiomhash(void *output, const void *input)
}
int scanhash_axiom(int thr_id, struct work *work,
uint32_t max_nonce, uint64_t *hashes_done)
uint32_t max_nonce, uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t _ALIGN(64) hash64[8];
uint32_t _ALIGN(64) endiandata[20];
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];

72
algo/x14/polytimos-4way.c
View File

@@ -13,26 +13,16 @@
#include "algo/gost/sph_gost.h"
#include "algo/echo/aes_ni/hash_api.h"
typedef struct {
union _poly_4way_context_overlay
{
skein512_4way_context skein;
shabal512_4way_context shabal;
hashState_echo echo;
luffa_2way_context luffa;
sph_fugue512_context fugue;
sph_gost512_context gost;
} poly_4way_ctx_holder;
poly_4way_ctx_holder poly_4way_ctx;
void init_polytimos_4way_ctx()
{
skein512_4way_init( &poly_4way_ctx.skein );
shabal512_4way_init( &poly_4way_ctx.shabal );
init_echo( &poly_4way_ctx.echo, 512 );
luffa_2way_init( &poly_4way_ctx.luffa, 512 );
sph_fugue512_init( &poly_4way_ctx.fugue );
sph_gost512_init( &poly_4way_ctx.gost );
}
};
typedef union _poly_4way_context_overlay poly_4way_context_overlay;
void polytimos_4way_hash( void *output, const void *input )
{
@@ -41,51 +31,57 @@ void polytimos_4way_hash( void *output, 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)));
poly_4way_ctx_holder ctx __attribute__ ((aligned (64)));
memcpy( &ctx, &poly_4way_ctx, sizeof(poly_4way_ctx) );
poly_4way_context_overlay ctx;
skein512_4way_init( &ctx.skein );
skein512_4way( &ctx.skein, input, 80 );
skein512_4way_close( &ctx.skein, vhash );
// Need to convert from 64 bit interleaved to 32 bit interleaved.
uint32_t vhash32[16*4];
mm256_rintrlv_4x64_4x32( vhash32, vhash, 512 );
shabal512_4way_init( &ctx.shabal );
shabal512_4way( &ctx.shabal, vhash32, 64 );
shabal512_4way_close( &ctx.shabal, vhash32 );
mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash32, 512 );
init_echo( &ctx.echo, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *)hash0, 512 );
memcpy( &ctx.echo, &poly_4way_ctx.echo, sizeof(hashState_echo) );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash1,
(const BitSequence *) hash1, 512 );
memcpy( &ctx.echo, &poly_4way_ctx.echo, sizeof(hashState_echo) );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash2,
(const BitSequence *) hash2, 512 );
memcpy( &ctx.echo, &poly_4way_ctx.echo, sizeof(hashState_echo) );
init_echo( &ctx.echo, 512 );
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash0, 64 );
sph_fugue512_close( &ctx.fugue, hash0 );
memcpy( &ctx.fugue, &poly_4way_ctx.fugue, sizeof(sph_fugue512_context) );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash1, 64 );
sph_fugue512_close( &ctx.fugue, hash1 );
memcpy( &ctx.fugue, &poly_4way_ctx.fugue, sizeof(sph_fugue512_context) );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash2, 64 );
sph_fugue512_close( &ctx.fugue, hash2 );
memcpy( &ctx.fugue, &poly_4way_ctx.fugue, sizeof(sph_fugue512_context) );
sph_fugue512_init( &ctx.fugue );
sph_fugue512( &ctx.fugue, hash3, 64 );
sph_fugue512_close( &ctx.fugue, hash3 );
sph_gost512_init( &ctx.gost );
sph_gost512( &ctx.gost, hash0, 64 );
sph_gost512_close( &ctx.gost, hash0 );
sph_gost512_init( &ctx.gost );
@@ -104,51 +100,43 @@ void polytimos_4way_hash( void *output, const void *input )
memcpy( output+96, hash3, 32 );
}
int scanhash_polytimos_4way( int thr_id, struct work *work, uint32_t max_nonce, uint64_t *hashes_done )
int scanhash_polytimos_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
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;
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
uint32_t *nonces = work->nonces;
int num_found = 0;
uint32_t *noncep = vdata + 73; // 9*8 + 1
__m256i *noncev = (__m256i*)vdata + 9; // aligned
const uint32_t Htarg = ptarget[7];
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
volatile uint8_t *restart = &(work_restart[thr_id].restart);
if ( opt_benchmark )
ptarget[7] = 0x0cff;
for ( int i=0; i < 19; i++ )
be32enc( &endiandata[i], pdata[i] );
uint64_t *edata = (uint64_t*)endiandata;
mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
mm256_bswap_intrlv80_4x64( vdata, pdata );
do {
be32enc( noncep, n );
be32enc( noncep+2, n+1 );
be32enc( noncep+4, n+2 );
be32enc( noncep+6, n+3 );
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
polytimos_4way_hash(hash, vdata);
pdata[19] = n;
for ( int i = 0; i < 4; i++ )
if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget ) )
for ( int i = 0; i < 4; i++ ) if ( (hash+(i<<3))[7] <= Htarg )
if( fulltest( hash+(i<<3), ptarget ) && !opt_benchmark )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
submit_solution( work, hash+(i<<3), mythr, i );
}
n += 4;
} while ( ( num_found == 0 ) && ( n < max_nonce ) && !(*restart));
} while ( ( n < max_nonce-4 ) && !(*restart));
*hashes_done = n - first_nonce + 1;
return num_found;
return 0;
}
#endif

1
algo/x14/polytimos-gate.c
View File

@@ -4,7 +4,6 @@ bool register_polytimos_algo( algo_gate_t* gate )
{
gate->optimizations = SSE2_OPT | AES_OPT | AVX2_OPT;
#ifdef POLYTIMOS_4WAY
init_polytimos_4way_ctx();
gate->scanhash = (void*)&scanhash_polytimos_4way;
gate->hash = (void*)&polytimos_4way_hash;
#else

9
algo/x14/polytimos-gate.h
View File

@@ -13,19 +13,14 @@ bool register_polytimos_algo( algo_gate_t* gate );
#if defined(POLYTIMOS_4WAY)
void polytimos_4way_hash( void *state, const void *input );
int scanhash_polytimos_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
void init_polytimos_4way_ctx();
uint64_t *hashes_done, struct thr_info *mythr );
#endif
void polytimos_hash( void *state, const void *input );
int scanhash_polytimos( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
uint64_t *hashes_done, struct thr_info *mythr );
void init_polytimos_ctx();
#endif

5
algo/x14/polytimos.c
View File

@@ -76,13 +76,14 @@ void polytimos_hash(void *output, const void *input)
memcpy(output, hashA, 32);
}
int scanhash_polytimos(int thr_id, struct work *work, uint32_t max_nonce, uint64_t *hashes_done)
int scanhash_polytimos( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t _ALIGN(128) hash[8];
uint32_t _ALIGN(128) endiandata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t nonce = first_nonce;

11
algo/x14/veltor-4way.c
View File

@@ -40,7 +40,7 @@ void veltor_4way_hash( void *output, const void *input )
skein512_4way( &ctx.skein, input, 80 );
skein512_4way_close( &ctx.skein, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
sph_shavite512( &ctx.shavite, hash0, 64 );
sph_shavite512_close( &ctx.shavite, hash0 );
@@ -54,10 +54,10 @@ void veltor_4way_hash( void *output, const void *input )
sph_shavite512( &ctx.shavite, hash3, 64 );
sph_shavite512_close( &ctx.shavite, hash3 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
sph_gost512( &ctx.gost, hash0, 64 );
sph_gost512_close( &ctx.gost, hash0 );
@@ -78,7 +78,7 @@ void veltor_4way_hash( void *output, const void *input )
}
int scanhash_veltor_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
@@ -91,6 +91,7 @@ int scanhash_veltor_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint32_t *nonces = work->nonces;
int num_found = 0;
uint32_t *noncep = vdata + 73; // 9*8 + 1
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
volatile uint8_t *restart = &(work_restart[thr_id].restart);
if ( opt_benchmark )
@@ -101,7 +102,7 @@ int scanhash_veltor_4way( int thr_id, struct work *work, uint32_t max_nonce,
}
uint64_t *edata = (uint64_t*)endiandata;
mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
do
{
be32enc( noncep, n );

4
algo/x14/veltor-gate.h
View File

@@ -15,7 +15,7 @@ bool register_veltor_algo( algo_gate_t* gate );
void veltor_4way_hash( void *state, const void *input );
int scanhash_veltor_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
uint64_t *hashes_done, struct thr_info *mythr );
void init_veltor_4way_ctx();
@@ -24,7 +24,7 @@ void init_veltor_4way_ctx();
void veltor_hash( void *state, const void *input );
int scanhash_veltor( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
uint64_t *hashes_done, struct thr_info *mythr );
void init_veltor_ctx();

4
algo/x14/veltor.c
View File

@@ -61,12 +61,14 @@ void veltor_hash(void *output, const void *input)
memcpy(output, hashB, 32);
}
int scanhash_veltor(int thr_id, struct work *work, uint32_t max_nonce, uint64_t *hashes_done)
int scanhash_veltor( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t _ALIGN(128) hash[8];
uint32_t _ALIGN(128) endiandata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];

33
algo/x14/x14-4way.c
View File

@@ -78,7 +78,7 @@ void x14_4way_hash( void *state, const void *input )
bmw512_4way_close( &ctx.bmw, vhash );
// Serial
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// 3 Groestl
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 512 );
@@ -90,7 +90,7 @@ void x14_4way_hash( void *state, const void *input )
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 512 );
// Parallel 4way
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
// 4 Skein
skein512_4way( &ctx.skein, vhash, 64 );
@@ -105,16 +105,16 @@ void x14_4way_hash( void *state, const void *input )
keccak512_4way_close( &ctx.keccak, vhash );
// Serial
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// 7 Luffa
mm256_interleave_2x128( vhash, hash0, hash1, 512 );
mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
mm256_interleave_2x128( vhash, hash2, hash3, 512 );
mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
// 8 Cubehash
cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
@@ -142,13 +142,13 @@ void x14_4way_hash( void *state, const void *input )
sph_shavite512_close( &ctx.shavite, hash3 );
// 10 Simd
mm256_interleave_2x128( vhash, hash0, hash1, 512 );
mm256_intrlv_2x128( vhash, hash0, hash1, 512 );
simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
mm256_interleave_2x128( vhash, hash2, hash3, 512 );
mm256_dintrlv_2x128( hash0, hash1, vhash, 512 );
mm256_intrlv_2x128( vhash, hash2, hash3, 512 );
simd_2way_init( &ctx.simd, 512 );
simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
mm256_dintrlv_2x128( hash2, hash3, vhash, 512 );
// 11 Echo
update_final_echo( &ctx.echo, (BitSequence *)hash0,
@@ -164,10 +164,10 @@ void x14_4way_hash( void *state, const void *input )
(const BitSequence *) hash3, 512 );
// 12 Hamsi parallel 4way 32 bit
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// 13 Fugue serial
sph_fugue512( &ctx.fugue, hash0, 64 );
@@ -183,14 +183,14 @@ void x14_4way_hash( void *state, const void *input )
sph_fugue512_close( &ctx.fugue, hash3 );
// 14 Shabal, parallel 32 bit
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, state );
}
int scanhash_x14_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[4*16] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
@@ -203,6 +203,7 @@ int scanhash_x14_4way( int thr_id, struct work *work, uint32_t max_nonce,
int num_found = 0;
uint32_t *noncep = vdata + 73; // 9*8 + 1
const uint32_t Htarg = ptarget[7];
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
uint64_t htmax[] = { 0, 0xF, 0xFF,
0xFFF, 0xFFFF, 0x10000000 };
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
@@ -212,7 +213,7 @@ int scanhash_x14_4way( int thr_id, struct work *work, uint32_t max_nonce,
swab32_array( endiandata, pdata, 20 );
uint64_t *edata = (uint64_t*)endiandata;
mm256_interleave_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
for ( int m=0; m < 6; m++ )
if ( Htarg <= htmax[m] )

8
algo/x14/x14-gate.h
View File

@@ -13,19 +13,15 @@ bool register_x14_algo( algo_gate_t* gate );
#if defined(X14_4WAY)
void x14_4way_hash( void *state, const void *input );
int scanhash_x14_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
uint64_t *hashes_done, struct thr_info *mythr );
void init_x14_4way_ctx();
#endif
void x14hash( void *state, const void *input );
int scanhash_x14( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
uint64_t *hashes_done, struct thr_info *mythr );
void init_x14_ctx();
#endif

5
algo/x14/x14.c
View File

@@ -180,8 +180,8 @@ void x14hash(void *output, const void *input)
memcpy(output, hash, 32);
}
int scanhash_x14(int thr_id, struct work *work,
uint32_t max_nonce, uint64_t *hashes_done)
int scanhash_x14( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t endiandata[20] __attribute__((aligned(64)));
uint32_t hash64[8] __attribute__((aligned(64)));
@@ -190,6 +190,7 @@ int scanhash_x14(int thr_id, struct work *work,
uint32_t n = pdata[19] - 1;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
/* int */ thr_id = mythr->id; // thr_id arg is deprecated
uint64_t htmax[] = {
0,