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

v3.19.9

Browse Source
This commit is contained in:
Jay D Dee
2022-07-10 11:04:00 -04:00
parent 26b8927632
commit f552f2b1e8
27 changed files with 883 additions and 396 deletions
Download Patch File Download Diff File Expand all files Collapse all files

102
algo/x16/x16r-4way.c
View File

@@ -16,7 +16,8 @@
#if defined (X16R_8WAY)
// Perform midstate prehash of hash functions with block size <= 72 bytes.
// Perform midstate prehash of hash functions with block size <= 72 bytes,
// 76 bytes for hash functions that operate on 32 bit data.
void x16r_8way_prehash( void *vdata, void *pdata )
{
@@ -44,18 +45,36 @@ void x16r_8way_prehash( void *vdata, void *pdata )
skein512_8way_update( &x16r_ctx.skein, vdata, 64 );
break;
case LUFFA:
{
hashState_luffa ctx_luffa;
mm128_bswap32_80( edata, pdata );
intrlv_4x128( vdata2, edata, edata, edata, edata, 640 );
luffa_4way_init( &x16r_ctx.luffa, 512 );
luffa_4way_update( &x16r_ctx.luffa, vdata2, 64 );
rintrlv_4x128_8x64( vdata, vdata2, vdata2, 640 );
intrlv_8x64( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
init_luffa( &ctx_luffa, 512 );
update_luffa( &ctx_luffa, (const BitSequence*)edata, 64 );
intrlv_4x128( x16r_ctx.luffa.buffer, ctx_luffa.buffer,
ctx_luffa.buffer, ctx_luffa.buffer, ctx_luffa.buffer, 512 );
intrlv_4x128( x16r_ctx.luffa.chainv, ctx_luffa.chainv,
ctx_luffa.chainv, ctx_luffa.chainv, ctx_luffa.chainv, 1280 );
x16r_ctx.luffa.hashbitlen = ctx_luffa.hashbitlen;
x16r_ctx.luffa.rembytes = ctx_luffa.rembytes;
}
break;
case CUBEHASH:
{
cubehashParam ctx_cube;
mm128_bswap32_80( edata, pdata );
intrlv_4x128( vdata2, edata, edata, edata, edata, 640 );
cube_4way_init( &x16r_ctx.cube, 512, 16, 32 );
cube_4way_update( &x16r_ctx.cube, vdata2, 64 );
rintrlv_4x128_8x64( vdata, vdata2, vdata2, 640 );
intrlv_8x64( vdata, edata, edata, edata, edata,
edata, edata, edata, edata, 640 );
cubehashInit( &ctx_cube, 512, 16, 32 );
cubehashUpdate( &ctx_cube, (const byte*)edata, 64 );
x16r_ctx.cube.hashlen = ctx_cube.hashlen;
x16r_ctx.cube.rounds = ctx_cube.rounds;
x16r_ctx.cube.blocksize = ctx_cube.blocksize;
x16r_ctx.cube.pos = ctx_cube.pos;
intrlv_4x128( x16r_ctx.cube.h, ctx_cube.x, ctx_cube.x, ctx_cube.x,
ctx_cube.x, 1024 );
}
break;
case HAMSI:
mm512_bswap32_intrlv80_8x64( vdata, pdata );
@@ -94,14 +113,14 @@ void x16r_8way_prehash( void *vdata, void *pdata )
int x16r_8way_hash_generic( void* output, const void* input, int thrid )
{
uint32_t vhash[20*8] __attribute__ ((aligned (128)));
uint32_t hash0[20] __attribute__ ((aligned (64)));
uint32_t hash1[20] __attribute__ ((aligned (64)));
uint32_t hash2[20] __attribute__ ((aligned (64)));
uint32_t hash3[20] __attribute__ ((aligned (64)));
uint32_t hash4[20] __attribute__ ((aligned (64)));
uint32_t hash5[20] __attribute__ ((aligned (64)));
uint32_t hash6[20] __attribute__ ((aligned (64)));
uint32_t hash7[20] __attribute__ ((aligned (64)));
uint32_t hash0[20] __attribute__ ((aligned (16)));
uint32_t hash1[20] __attribute__ ((aligned (16)));
uint32_t hash2[20] __attribute__ ((aligned (16)));
uint32_t hash3[20] __attribute__ ((aligned (16)));
uint32_t hash4[20] __attribute__ ((aligned (16)));
uint32_t hash5[20] __attribute__ ((aligned (16)));
uint32_t hash6[20] __attribute__ ((aligned (16)));
uint32_t hash7[20] __attribute__ ((aligned (16)));
x16r_8way_context_overlay ctx;
memcpy( &ctx, &x16r_ctx, sizeof(ctx) );
void *in0 = (void*) hash0;
@@ -476,7 +495,7 @@ int scanhash_x16r_8way( struct work *work, uint32_t max_nonce,
{
uint32_t hash[16*8] __attribute__ ((aligned (128)));
uint32_t vdata[20*8] __attribute__ ((aligned (64)));
uint32_t bedata1[2] __attribute__((aligned(64)));
uint32_t bedata1[2];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
@@ -500,7 +519,7 @@ int scanhash_x16r_8way( struct work *work, uint32_t max_nonce,
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_INFO, "hash order %s (%08x)", x16r_hash_order, ntime );
applog( LOG_INFO, "Hash order %s Ntime %08x", x16r_hash_order, ntime );
}
x16r_8way_prehash( vdata, pdata );
@@ -552,18 +571,33 @@ void x16r_4way_prehash( void *vdata, void *pdata )
skein512_4way_prehash64( &x16r_ctx.skein, vdata );
break;
case LUFFA:
{
hashState_luffa ctx_luffa;
mm128_bswap32_80( edata, pdata );
intrlv_2x128( vdata2, edata, edata, 640 );
luffa_2way_init( &x16r_ctx.luffa, 512 );
luffa_2way_update( &x16r_ctx.luffa, vdata2, 64 );
rintrlv_2x128_4x64( vdata, vdata2, vdata2, 640 );
break;
intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
init_luffa( &ctx_luffa, 512 );
update_luffa( &ctx_luffa, (const BitSequence*)edata, 64 );
intrlv_2x128( x16r_ctx.luffa.buffer, ctx_luffa.buffer,
ctx_luffa.buffer, 512 );
intrlv_2x128( x16r_ctx.luffa.chainv, ctx_luffa.chainv,
ctx_luffa.chainv, 1280 );
x16r_ctx.luffa.hashbitlen = ctx_luffa.hashbitlen;
x16r_ctx.luffa.rembytes = ctx_luffa.rembytes;
}
break;
case CUBEHASH:
{
cubehashParam ctx_cube;
mm128_bswap32_80( edata, pdata );
intrlv_2x128( vdata2, edata, edata, 640 );
cube_2way_init( &x16r_ctx.cube, 512, 16, 32 );
cube_2way_update( &x16r_ctx.cube, vdata2, 64 );
rintrlv_2x128_4x64( vdata, vdata2, vdata2, 640 );
intrlv_4x64( vdata, edata, edata, edata, edata, 640 );
cubehashInit( &ctx_cube, 512, 16, 32 );
cubehashUpdate( &ctx_cube, (const byte*)edata, 64 );
x16r_ctx.cube.hashlen = ctx_cube.hashlen;
x16r_ctx.cube.rounds = ctx_cube.rounds;
x16r_ctx.cube.blocksize = ctx_cube.blocksize;
x16r_ctx.cube.pos = ctx_cube.pos;
intrlv_2x128( x16r_ctx.cube.h, ctx_cube.x, ctx_cube.x, 1024 );
}
break;
case HAMSI:
mm256_bswap32_intrlv80_4x64( vdata, pdata );
@@ -596,10 +630,10 @@ void x16r_4way_prehash( void *vdata, void *pdata )
int x16r_4way_hash_generic( void* output, const void* input, int thrid )
{
uint32_t vhash[20*4] __attribute__ ((aligned (128)));
uint32_t hash0[20] __attribute__ ((aligned (64)));
uint32_t hash1[20] __attribute__ ((aligned (64)));
uint32_t hash2[20] __attribute__ ((aligned (64)));
uint32_t hash3[20] __attribute__ ((aligned (64)));
uint32_t hash0[20] __attribute__ ((aligned (32)));
uint32_t hash1[20] __attribute__ ((aligned (32)));
uint32_t hash2[20] __attribute__ ((aligned (32)));
uint32_t hash3[20] __attribute__ ((aligned (32)));
x16r_4way_context_overlay ctx;
memcpy( &ctx, &x16r_ctx, sizeof(ctx) );
void *in0 = (void*) hash0;
@@ -890,7 +924,7 @@ int scanhash_x16r_4way( struct work *work, uint32_t max_nonce,
{
uint32_t hash[16*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t bedata1[2] __attribute__((aligned(64)));
uint32_t bedata1[2];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
@@ -913,7 +947,7 @@ int scanhash_x16r_4way( struct work *work, uint32_t max_nonce,
x16_r_s_getAlgoString( (const uint8_t*)bedata1, x16r_hash_order );
s_ntime = ntime;
if ( opt_debug && !thr_id )
applog( LOG_INFO, "hash order %s (%08x)", x16r_hash_order, ntime );
applog( LOG_INFO, "Hash order %s Ntime %08x", x16r_hash_order, ntime );
}
x16r_4way_prehash( vdata, pdata );

8
algo/x16/x16rt-4way.c
View File

@@ -30,8 +30,8 @@ int scanhash_x16rt_8way( struct work *work, uint32_t max_nonce,
x16rt_getTimeHash( masked_ntime, &timeHash );
x16rt_getAlgoString( &timeHash[0], x16r_hash_order );
s_ntime = masked_ntime;
if ( opt_debug && !thr_id )
applog( LOG_INFO, "hash order: %s time: (%08x) time hash: (%08x)",
if ( !thr_id )
applog( LOG_INFO, "Hash order %s, Nime %08x, time hash %08x",
x16r_hash_order, bswap_32( pdata[17] ), timeHash );
}
@@ -84,8 +84,8 @@ int scanhash_x16rt_4way( struct work *work, uint32_t max_nonce,
x16rt_getTimeHash( masked_ntime, &timeHash );
x16rt_getAlgoString( &timeHash[0], x16r_hash_order );
s_ntime = masked_ntime;
if ( opt_debug && !thr_id )
applog( LOG_INFO, "hash order: %s time: (%08x) time hash: (%08x)",
if ( !thr_id )
applog( LOG_INFO, "Hash order %s, Nime %08x, time hash %08x",
x16r_hash_order, bswap_32( pdata[17] ), timeHash );
}

29
algo/x16/x16rv2-4way.c
View File

@@ -45,14 +45,14 @@ static __thread x16rv2_8way_context_overlay x16rv2_ctx;
int x16rv2_8way_hash( void* output, const void* input, int thrid )
{
uint32_t vhash[24*8] __attribute__ ((aligned (128)));
uint32_t hash0[24] __attribute__ ((aligned (64)));
uint32_t hash1[24] __attribute__ ((aligned (64)));
uint32_t hash2[24] __attribute__ ((aligned (64)));
uint32_t hash3[24] __attribute__ ((aligned (64)));
uint32_t hash4[24] __attribute__ ((aligned (64)));
uint32_t hash5[24] __attribute__ ((aligned (64)));
uint32_t hash6[24] __attribute__ ((aligned (64)));
uint32_t hash7[24] __attribute__ ((aligned (64)));
uint32_t hash0[24] __attribute__ ((aligned (32)));
uint32_t hash1[24] __attribute__ ((aligned (32)));
uint32_t hash2[24] __attribute__ ((aligned (32)));
uint32_t hash3[24] __attribute__ ((aligned (32)));
uint32_t hash4[24] __attribute__ ((aligned (32)));
uint32_t hash5[24] __attribute__ ((aligned (32)));
uint32_t hash6[24] __attribute__ ((aligned (32)));
uint32_t hash7[24] __attribute__ ((aligned (32)));
x16rv2_8way_context_overlay ctx;
memcpy( &ctx, &x16rv2_ctx, sizeof(ctx) );
void *in0 = (void*) hash0;
@@ -706,11 +706,11 @@ inline void padtiger512( uint32_t* hash )
int x16rv2_4way_hash( void* output, const void* input, int thrid )
{
uint32_t hash0[20] __attribute__ ((aligned (64)));
uint32_t hash1[20] __attribute__ ((aligned (64)));
uint32_t hash2[20] __attribute__ ((aligned (64)));
uint32_t hash3[20] __attribute__ ((aligned (64)));
uint32_t vhash[20*4] __attribute__ ((aligned (64)));
uint32_t hash0[20] __attribute__ ((aligned (32)));
uint32_t hash1[20] __attribute__ ((aligned (32)));
uint32_t hash2[20] __attribute__ ((aligned (32)));
uint32_t hash3[20] __attribute__ ((aligned (32)));
x16rv2_4way_context_overlay ctx;
memcpy( &ctx, &x16rv2_ctx, sizeof(ctx) );
void *in0 = (void*) hash0;
@@ -1054,8 +1054,8 @@ int scanhash_x16rv2_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 vdata32[20*4] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t bedata1[2] __attribute__((aligned(64)));
uint32_t edata[20];
uint32_t bedata1[2];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
@@ -1068,7 +1068,6 @@ int scanhash_x16rv2_4way( struct work *work, uint32_t max_nonce,
if ( bench ) ptarget[7] = 0x0fff;
bedata1[0] = bswap_32( pdata[1] );
bedata1[1] = bswap_32( pdata[2] );