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

v3.9.5.1

Browse Source
This commit is contained in:
Jay D Dee
2019-07-02 15:10:38 -04:00
parent 0d48d573ce
commit 0d769ee0fe
53 changed files with 1755 additions and 1170 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
algo/x13/skunk-4way.c
View File

@@ -78,29 +78,23 @@ int scanhash_skunk_4way( struct work *work, uint32_t max_nonce,
{
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 *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 )
((uint32_t*)ptarget)[7] = 0x0cff;
for ( int k = 0; k < 19; k++ )
be32enc( &endiandata[k], pdata[k] );
uint64_t *edata = (uint64_t*)endiandata;
mm256_intrlv_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 );
skunk_4way_hash( hash, vdata );
pdata[19] = n;

15
algo/x13/x13-4way.c
View File

@@ -189,12 +189,11 @@ int scanhash_x13_4way( struct work *work, uint32_t max_nonce,
{
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 *noncep = vdata + 73; // 9*8 + 1
__m256i *noncev = (__m256i*)vdata + 9; // aligned
int thr_id = mythr->id; // thr_id arg is deprecated
const uint32_t Htarg = ptarget[7];
uint64_t htmax[] = { 0, 0xF, 0xFF,
@@ -202,11 +201,7 @@ int scanhash_x13_4way( struct work *work, uint32_t max_nonce,
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_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
mm256_bswap_intrlv80_4x64( vdata, pdata );
for ( int m=0; m < 6; m++ )
if ( Htarg <= htmax[m] )
@@ -214,10 +209,8 @@ int scanhash_x13_4way( struct work *work, uint32_t max_nonce,
uint32_t mask = masks[m];
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 );
x13_4way_hash( hash, vdata );
pdata[19] = n;

19
algo/x13/x13sm3-4way.c
View File

@@ -166,7 +166,7 @@ void x13sm3_4way_hash( void *state, const void *input )
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
mm128_intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
intrlv_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
// SM3 parallel 32 bit
uint32_t sm3_vhash[32*4] __attribute__ ((aligned (64)));
@@ -182,7 +182,7 @@ void x13sm3_4way_hash( void *state, const void *input )
sm3_4way( &ctx.sm3, vhash, 64 );
sm3_4way_close( &ctx.sm3, sm3_vhash );
mm128_dintrlv_4x32( hash0, hash1, hash2, hash3, sm3_vhash, 512 );
dintrlv_4x32( hash0, hash1, hash2, hash3, sm3_vhash, 512 );
// Hamsi parallel 4x32x2
mm256_intrlv_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
@@ -214,12 +214,11 @@ int scanhash_x13sm3_4way( struct work *work, uint32_t max_nonce,
{
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 *noncep = vdata + 73; // 9*8 + 1
__m256i *noncev = (__m256i*)vdata + 9; // aligned
int thr_id = mythr->id; // thr_id arg is deprecated
const uint32_t Htarg = ptarget[7];
uint64_t htmax[] = { 0, 0xF, 0xFF,
@@ -227,11 +226,7 @@ int scanhash_x13sm3_4way( struct work *work, uint32_t max_nonce,
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_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
mm256_bswap_intrlv80_4x64( vdata, pdata );
blake512_4way_init( &x13sm3_ctx_mid );
blake512_4way( &x13sm3_ctx_mid, vdata, 64 );
@@ -242,10 +237,8 @@ int scanhash_x13sm3_4way( struct work *work, uint32_t max_nonce,
uint32_t mask = masks[m];
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 );
x13sm3_4way_hash( hash, vdata );
pdata[19] = n;