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

v3.9.5.2

Browse Source
This commit is contained in:
Jay D Dee
2019-07-04 12:12:11 -04:00
parent 0d769ee0fe
commit 9abc19a30a
29 changed files with 220 additions and 155 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
algo/jh/jha-4way.c
View File

@@ -88,6 +88,7 @@ int scanhash_jha_4way( struct work *work, uint32_t max_nonce,
{
uint32_t hash[8*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[25]);
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *pdata = work->data;
@@ -115,12 +116,11 @@ int scanhash_jha_4way( struct work *work, uint32_t max_nonce,
0
};
// for ( int i=0; i < 19; i++ )
// be32enc( &endiandata[i], pdata[i] );
for ( int i=0; i < 19; i++ )
be32enc( &edata[i], pdata[i] );
// uint64_t *edata = (uint64_t*)endiandata;
// mm256_intrlv_4x64( (uint64_t*)vdata, edata, edata, edata, edata, 640 );
mm256_bswap_intrlv80_4x64( vdata, pdata );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
for ( int m = 0; m < 6; m++ )
{

5
algo/keccak/keccak-4way.c
View File

@@ -21,6 +21,7 @@ int scanhash_keccak_4way( struct work *work, uint32_t max_nonce,
{
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t hash[8*4] __attribute__ ((aligned (32)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *hash7 = &(hash[25]); // 3*8+1
uint32_t *pdata = work->data;
@@ -31,7 +32,9 @@ int scanhash_keccak_4way( struct work *work, uint32_t max_nonce,
// const uint32_t Htarg = ptarget[7];
int thr_id = mythr->id; // thr_id arg is deprecated
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
do {
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );

6
algo/lyra2/lyra2rev3-4way.c
View File

@@ -91,6 +91,7 @@ int scanhash_lyra2rev3_8way( struct work *work, uint32_t max_nonce,
{
uint32_t hash[8*8] __attribute__ ((aligned (64)));
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[7<<3]);
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *pdata = work->data;
@@ -104,7 +105,10 @@ int scanhash_lyra2rev3_8way( struct work *work, uint32_t max_nonce,
if ( opt_benchmark )
( (uint32_t*)ptarget )[7] = 0x0000ff;
mm256_bswap_intrlv80_8x32( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_8x32( vdata, pdata );
do
{
*noncev = mm256_bswap_32( _mm256_set_epi32( n+7, n+6, n+5, n+4,

6
algo/lyra2/lyra2z-4way.c
View File

@@ -146,6 +146,7 @@ int scanhash_lyra2z_8way( struct work *work, uint32_t max_nonce,
{
uint32_t hash[8*8] __attribute__ ((aligned (64)));
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t Htarg = ptarget[7];
@@ -157,7 +158,10 @@ int scanhash_lyra2z_8way( struct work *work, uint32_t max_nonce,
if ( opt_benchmark )
ptarget[7] = 0x0000ff;
mm256_bswap_intrlv80_8x32( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_8x32( vdata, pdata );
lyra2z_8way_midstate( vdata );
do {

5
algo/quark/anime-4way.c
View File

@@ -165,6 +165,7 @@ int scanhash_anime_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 edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
@@ -189,7 +190,9 @@ int scanhash_anime_4way( struct work *work, uint32_t max_nonce,
0
};
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
for (int m=0; m < 6; m++)
if (Htarg <= htmax[m])

5
algo/quark/quark-4way.c
View File

@@ -170,6 +170,7 @@ int scanhash_quark_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 edata[20] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[25]);
uint32_t *pdata = work->data;
@@ -179,7 +180,9 @@ int scanhash_quark_4way( struct work *work, uint32_t max_nonce,
__m256i *noncev = (__m256i*)vdata + 9; // aligned
int thr_id = mythr->id; // thr_id arg is deprecated
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
do
{
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(

11
algo/sha/sha256q-4way.c
View File

@@ -36,6 +36,7 @@ int scanhash_sha256q_8way( struct work *work, uint32_t max_nonce,
{
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t hash[8*8] __attribute__ ((aligned (32)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
@@ -59,7 +60,10 @@ int scanhash_sha256q_8way( struct work *work, uint32_t max_nonce,
0 };
// Need big endian data
mm256_bswap_intrlv80_8x32( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_8x32( vdata, pdata );
sha256_8way_init( &sha256_ctx8 );
sha256_8way( &sha256_ctx8, vdata, 64 );
@@ -129,6 +133,7 @@ int scanhash_sha256q_4way( struct work *work, uint32_t max_nonce,
{
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t hash[8*4] __attribute__ ((aligned (32)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *pdata = work->data;
@@ -152,7 +157,9 @@ int scanhash_sha256q_4way( struct work *work, uint32_t max_nonce,
0xFFFF0000,
0 };
mm128_bswap_intrlv80_4x32( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm128_intrlv_4x32( vdata, edata, edata, edata, edata, 640 );
// mm128_bswap_intrlv80_4x32( vdata, pdata );
sha256_4way_init( &sha256_ctx4 );
sha256_4way( &sha256_ctx4, vdata, 64 );

13
algo/sha/sha256t-4way.c
View File

@@ -163,6 +163,7 @@ int scanhash_sha256t_8way( struct work *work, uint32_t max_nonce,
{
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t hash[8*8] __attribute__ ((aligned (32)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *hash7 = &(hash[7<<3]);
uint32_t *pdata = work->data;
@@ -186,8 +187,12 @@ int scanhash_sha256t_8way( struct work *work, uint32_t max_nonce,
0xFFFF0000,
0 };
swab32_array( edata, pdata, 20 );
mm256_intrlv_8x32( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
// Need big endian data
mm256_bswap_intrlv80_8x32( vdata, pdata );
// mm256_bswap_intrlv80_8x32( vdata, pdata );
sha256_8way_init( &sha256_ctx8 );
sha256_8way( &sha256_ctx8, vdata, 64 );
@@ -248,6 +253,7 @@ int scanhash_sha256t_4way( struct work *work, uint32_t max_nonce,
{
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t hash[8*4] __attribute__ ((aligned (32)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t *pdata = work->data;
@@ -271,7 +277,10 @@ int scanhash_sha256t_4way( struct work *work, uint32_t max_nonce,
0xFFFF0000,
0 };
mm128_bswap_intrlv80_4x32( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm128_intrlv_4x32( vdata, edata, edata, edata, edata, 640 );
// mm128_bswap_intrlv80_4x32( vdata, pdata );
sha256_4way_init( &sha256_ctx4 );
sha256_4way( &sha256_ctx4, vdata, 64 );

5
algo/skein/skein-4way.c
View File

@@ -63,6 +63,7 @@ int scanhash_skein_4way( struct work *work, uint32_t max_nonce,
{
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 lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t *pdata = work->data;
@@ -73,7 +74,9 @@ 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
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
do
{
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(

17
algo/skein/skein2-4way.c
View File

@@ -32,25 +32,24 @@ int scanhash_skein2_4way( struct work *work, uint32_t max_nonce,
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
// __m256i *noncev = (__m256i*)vdata + 9; // aligned
__m256i *noncev = (__m256i*)vdata + 9; // aligned
int thr_id = mythr->id; // thr_id arg is deprecated
uint32_t *noncep = vdata + 73; // 9*8 + 1
// 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 );
do
{
be32enc( noncep, n );
be32enc( noncep+2, n+1 );
be32enc( noncep+4, n+2 );
be32enc( noncep+6, n+3 );
// 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 );
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );
skein2hash_4way( hash, vdata );

6
algo/x11/c11-4way.c
View File

@@ -165,7 +165,7 @@ int scanhash_c11_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 edata[20] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
@@ -178,7 +178,9 @@ int scanhash_c11_4way( struct work *work, uint32_t max_nonce,
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
0xFFFFF000, 0xFFFF0000, 0 };
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
for (int m=0; m < 6; m++)
if (Htarg <= htmax[m])

5
algo/x11/tribus-4way.c
View File

@@ -64,6 +64,7 @@ int scanhash_tribus_4way( struct work *work, uint32_t max_nonce,
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
@@ -86,7 +87,9 @@ int scanhash_tribus_4way( struct work *work, uint32_t max_nonce,
0xFFFF0000,
0 };
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
// precalc midstate
// doing it one way then then interleaving would be faster but too

6
algo/x11/x11-4way.c
View File

@@ -164,7 +164,7 @@ int scanhash_x11_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 edata[20] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
@@ -177,7 +177,9 @@ int scanhash_x11_4way( struct work *work, uint32_t max_nonce,
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
0xFFFFF000, 0xFFFF0000, 0 };
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
for (int m=0; m < 6; m++)
if (Htarg <= htmax[m])

5
algo/x11/x11gost-4way.c
View File

@@ -171,6 +171,7 @@ int scanhash_x11gost_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 edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
@@ -183,7 +184,9 @@ int scanhash_x11gost_4way( struct work *work, uint32_t max_nonce,
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
0xFFFFF000, 0xFFFF0000, 0 };
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
for (int m=0; m < 6; m++)
if (Htarg <= htmax[m])

5
algo/x13/phi1612-4way.c
View File

@@ -114,6 +114,7 @@ int scanhash_phi1612_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 edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
@@ -124,7 +125,9 @@ int scanhash_phi1612_4way( struct work *work, uint32_t max_nonce,
if ( opt_benchmark )
( (uint32_t*)ptarget )[7] = 0x0cff;
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
do {
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(

5
algo/x13/skunk-4way.c
View File

@@ -78,6 +78,7 @@ 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 edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
@@ -90,7 +91,9 @@ int scanhash_skunk_4way( struct work *work, uint32_t max_nonce,
if ( opt_benchmark )
((uint32_t*)ptarget)[7] = 0x0cff;
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
do
{
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(

5
algo/x13/x13-4way.c
View File

@@ -189,6 +189,7 @@ 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 edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
@@ -201,7 +202,9 @@ int scanhash_x13_4way( struct work *work, uint32_t max_nonce,
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
0xFFFFF000, 0xFFFF0000, 0 };
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
for ( int m=0; m < 6; m++ )
if ( Htarg <= htmax[m] )

5
algo/x13/x13sm3-4way.c
View File

@@ -214,6 +214,7 @@ 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 edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
@@ -226,7 +227,9 @@ int scanhash_x13sm3_4way( struct work *work, uint32_t max_nonce,
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
0xFFFFF000, 0xFFFF0000, 0 };
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( 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 );

5
algo/x14/polytimos-4way.c
View File

@@ -105,6 +105,7 @@ int scanhash_polytimos_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 edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
@@ -117,7 +118,9 @@ int scanhash_polytimos_4way( struct work *work, uint32_t max_nonce,
if ( opt_benchmark )
ptarget[7] = 0x0cff;
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
do {
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(
_mm256_set_epi32( n+3, 0, n+2, 0, n+1, 0, n, 0 ) ), *noncev );

5
algo/x14/veltor-4way.c
View File

@@ -82,6 +82,7 @@ int scanhash_veltor_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 edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t Htarg = ptarget[7];
@@ -94,7 +95,9 @@ int scanhash_veltor_4way( struct work *work, uint32_t max_nonce,
if ( opt_benchmark )
ptarget[7] = 0x0cff;
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
do
{

5
algo/x14/x14-4way.c
View File

@@ -193,6 +193,7 @@ int scanhash_x14_4way( struct work *work, uint32_t max_nonce,
{
uint32_t hash[4*16] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
@@ -205,7 +206,9 @@ int scanhash_x14_4way( struct work *work, uint32_t max_nonce,
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
0xFFFFF000, 0xFFFF0000, 0 };
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
for ( int m=0; m < 6; m++ )
if ( Htarg <= htmax[m] )

6
algo/x15/x15-4way.c
View File

@@ -217,6 +217,7 @@ int scanhash_x15_4way( 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 edata[20] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
@@ -230,7 +231,10 @@ int scanhash_x15_4way( struct work *work, uint32_t max_nonce,
uint32_t masks[] = { 0xFFFFFFFF, 0xFFFFFFF0, 0xFFFFFF00,
0xFFFFF000, 0xFFFF0000, 0 };
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
for ( int m=0; m < 6; m++ )
if ( Htarg <= htmax[m] )

5
algo/x17/sonoa-4way.c
View File

@@ -808,6 +808,7 @@ int scanhash_sonoa_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 edata[20] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t *pdata = work->data;
@@ -823,7 +824,9 @@ int scanhash_sonoa_4way( struct work *work, uint32_t max_nonce,
0xFFFFF000, 0xFFFF0000, 0 };
// Need big endian data
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
for ( int m=0; m < 6; m++ ) if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[m];

5
algo/x17/x17-4way.c
View File

@@ -207,6 +207,7 @@ int scanhash_x17_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 edata[20] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t *pdata = work->data;
@@ -222,7 +223,9 @@ int scanhash_x17_4way( struct work *work, uint32_t max_nonce,
0xFFFFF000, 0xFFFF0000, 0 };
// Need big endian data
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
for ( int m = 0; m < 6; m++ ) if ( Htarg <= htmax[m] )
{
uint32_t mask = masks[ m ];

5
algo/x17/xevan-4way.c
View File

@@ -334,6 +334,7 @@ int scanhash_xevan_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 edata[20] __attribute__ ((aligned (64)));
uint32_t lane_hash[8] __attribute__ ((aligned (32)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t *pdata = work->data;
@@ -348,7 +349,9 @@ int scanhash_xevan_4way( struct work *work, uint32_t max_nonce,
if ( opt_benchmark )
ptarget[7] = 0x0cff;
mm256_bswap_intrlv80_4x64( vdata, pdata );
swab32_array( edata, pdata, 20 );
mm256_intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
// mm256_bswap_intrlv80_4x64( vdata, pdata );
do {
*noncev = mm256_intrlv_blend_32( mm256_bswap_32(
_mm256_set_epi32( n+3, 0,n+2, 0,n+1, 0, n, 0 ) ), *noncev );