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

v3.9.5.4

Browse Source
This commit is contained in:
Jay D Dee
2019-07-15 17:00:26 -04:00
parent e625ed5420
commit e2d5762ef2
63 changed files with 1973 additions and 2980 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
algo/skein/skein-4way.c
View File

@@ -12,7 +12,7 @@
void skeinhash_4way( void *state, const void *input )
{
uint64_t vhash64[8*4] __attribute__ ((aligned (64)));
uint64_t vhash64[16*4] __attribute__ ((aligned (64)));
skein512_4way_context ctx_skein;
#if defined(__SHA__)
uint32_t hash0[16] __attribute__ ((aligned (64)));
@@ -30,7 +30,7 @@ void skeinhash_4way( void *state, const void *input )
skein512_4way_close( &ctx_skein, vhash64 );
#if defined(__SHA__)
mm256_dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 512 );
dintrlv_4x64( hash0, hash1, hash2, hash3, vhash64, 512 );
SHA256_Init( &ctx_sha256 );
SHA256_Update( &ctx_sha256, (unsigned char*)hash0, 64 );
@@ -50,7 +50,7 @@ void skeinhash_4way( void *state, const void *input )
intrlv_4x32( state, hash0, hash1, hash2, hash3, 256 );
#else
mm256_rintrlv_4x64_4x32( vhash32, vhash64, 512 );
rintrlv_4x64_4x32( vhash32, vhash64, 512 );
sha256_4way_init( &ctx_sha256 );
sha256_4way( &ctx_sha256, vhash32, 64 );
@@ -62,8 +62,7 @@ int scanhash_skein_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t hash[8*4] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t hash[16*4] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t *pdata = work->data;
@@ -74,9 +73,7 @@ int scanhash_skein_4way( struct work *work, uint32_t max_nonce,
__m256i *noncev = (__m256i*)vdata + 9; // aligned
int thr_id = mythr->id; // thr_id arg is deprecated
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
mm256_bswap32_intrlv80_4x64( vdata, pdata );
do
{
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(

13
algo/skein/skein-hash-4way.c
View File

@@ -279,10 +279,7 @@ do { \
_mm256_xor_si256( k2, k3 ) ), \
_mm256_xor_si256( _mm256_xor_si256( k4, k5 ), \
_mm256_xor_si256( k6, k7 ) ) ), \
_mm256_set_epi64x( SPH_C64(0x1BD11BDAA9FC1A22), \
SPH_C64(0x1BD11BDAA9FC1A22), \
SPH_C64(0x1BD11BDAA9FC1A22), \
SPH_C64(0x1BD11BDAA9FC1A22) ) ); \
m256_const1_64( 0x1BD11BDAA9FC1A22) ); \
t2 = t0 ^ t1; \
} while (0)
@@ -294,13 +291,11 @@ do { \
w3 = _mm256_add_epi64( w3, SKBI(k,s,3) ); \
w4 = _mm256_add_epi64( w4, SKBI(k,s,4) ); \
w5 = _mm256_add_epi64( w5, _mm256_add_epi64( SKBI(k,s,5), \
_mm256_set_epi64x( SKBT(t,s,0), SKBT(t,s,0), \
SKBT(t,s,0), SKBT(t,s,0) ) ) ); \
m256_const1_64( SKBT(t,s,0) ) ) ); \
w6 = _mm256_add_epi64( w6, _mm256_add_epi64( SKBI(k,s,6), \
_mm256_set_epi64x( SKBT(t,s,1), SKBT(t,s,1), \
SKBT(t,s,1), SKBT(t,s,1) ) ) ); \
m256_const1_64( SKBT(t,s,1) ) ) ); \
w7 = _mm256_add_epi64( w7, _mm256_add_epi64( SKBI(k,s,7), \
_mm256_set_epi64x( s, s, s, s ) ) ); \
m256_const1_64( s ) ) ); \
} while (0)

19
algo/skein/skein2-4way.c
View File

@@ -8,7 +8,7 @@
void skein2hash_4way( void *output, const void *input )
{
skein512_4way_context ctx;
uint64_t hash[8*4] __attribute__ ((aligned (64)));
uint64_t hash[16*4] __attribute__ ((aligned (64)));
skein512_4way_init( &ctx );
skein512_4way( &ctx, input, 80 );
@@ -22,8 +22,7 @@ void skein2hash_4way( void *output, const void *input )
int scanhash_skein2_4way( struct work *work, uint32_t max_nonce,
uint64_t *hashes_done, struct thr_info *mythr )
{
uint32_t hash[8*4] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t hash[16*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[25]);
@@ -34,20 +33,10 @@ int scanhash_skein2_4way( struct work *work, uint32_t max_nonce,
uint32_t n = first_nonce;
__m256i *noncev = (__m256i*)vdata + 9; // aligned
int thr_id = mythr->id; // thr_id arg is deprecated
// uint32_t *noncep = vdata + 73; // 9*8 + 1
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
mm256_bswap32_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 );
@@ -56,7 +45,7 @@ int scanhash_skein2_4way( struct work *work, uint32_t max_nonce,
for ( int lane = 0; lane < 4; lane++ )
if ( hash7[ lane<<1 ] <= Htarg )
{
mm256_extr_lane_4x64( lane_hash, hash, lane, 256 );
extr_lane_4x64( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) && !opt_benchmark )
{
pdata[19] = n + lane;