GOSTSec
/
sgminer
mirror of https://github.com/GOSTSec/sgminer
1
0
Fork 0
Code Issues Releases Wiki Activity
Browse Source

Fixed marucoin mod 

Improve hamsi and fugue hashing (base on girino x13 version)
use SPH_KECCAK_UNROLL   1
and SPH_LUFFA_PARALLEL 1
to improve hashrate even further
djm34
aznboy84 11 years ago committed by ystarnaud
parent
12a2accd7e
commit
85fa1637e3
1 changed files with 851 additions and 1084 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 585
      kernel/marucoin-mod.cl

585
kernel/marucoin-mod.cl
Unescape View File

@ -34,9 +34,9 @@ @@ -34,9 +34,9 @@
#define X13MOD_CL
#if __ENDIAN_LITTLE__
#define SPH_LITTLE_ENDIAN 1
#define SPH_LITTLE_ENDIAN 1
#else
#define SPH_BIG_ENDIAN 1
#define SPH_BIG_ENDIAN 1
#endif
#define SPH_UPTR sph_u64
@ -44,24 +44,24 @@ @@ -44,24 +44,24 @@
typedef unsigned int sph_u32;
typedef int sph_s32;
#ifndef __OPENCL_VERSION__
typedef unsigned long long sph_u64;
typedef long long sph_s64;
typedef unsigned long long sph_u64;
typedef long long sph_s64;
#else
typedef unsigned long sph_u64;
typedef long sph_s64;
typedef unsigned long sph_u64;
typedef long sph_s64;
#endif
#define SPH_64 1
#define SPH_64_TRUE 1
#define SPH_C32(x) ((sph_u32)(x ## U))
#define SPH_T32(x) ((as_uint(x)) & SPH_C32(0xFFFFFFFF))
#define SPH_ROTL32(x, n) SPH_T32(((x) << (n)) | ((x) >> (32 - (n))))
#define SPH_T32(x) (as_uint(x))
#define SPH_ROTL32(x, n) rotate(as_uint(x), as_uint(n))
#define SPH_ROTR32(x, n) SPH_ROTL32(x, (32 - (n)))
#define SPH_C64(x) ((sph_u64)(x ## UL))
#define SPH_T64(x) ((as_ulong(x)) & SPH_C64(0xFFFFFFFFFFFFFFFF))
#define SPH_ROTL64(x, n) SPH_T64(((x) << (n)) | ((x) >> (64 - (n))))
#define SPH_T64(x) (as_ulong(x))
#define SPH_ROTL64(x, n) rotate(as_ulong(x), (n) & 0xFFFFFFFFFFFFFFFFUL)
#define SPH_ROTR64(x, n) SPH_ROTL64(x, (64 - (n)))
#define SPH_ECHO_64 1
@ -70,14 +70,12 @@ typedef int sph_s32; @@ -70,14 +70,12 @@ typedef int sph_s32;
#define SPH_SIMD_NOCOPY 0
#define SPH_KECCAK_NOCOPY 0
#define SPH_COMPACT_BLAKE_64 0
#define SPH_LUFFA_PARALLEL 0
#define SPH_LUFFA_PARALLEL 1
#define SPH_SMALL_FOOTPRINT_GROESTL 0
#define SPH_GROESTL_BIG_ENDIAN 0
#define SPH_CUBEHASH_UNROLL 0
#define SPH_KECCAK_UNROLL 0
#if !defined SPH_HAMSI_EXPAND_BIG
#define SPH_HAMSI_EXPAND_BIG 4
#endif
#define SPH_KECCAK_UNROLL 1
#define SPH_HAMSI_EXPAND_BIG 1
#include "blake.cl"
#include "bmw.cl"
@ -104,14 +102,6 @@ typedef int sph_s32; @@ -104,14 +102,6 @@ typedef int sph_s32;
#define DEC64BE(x) SWAP8(*(const __global sph_u64 *) (x));
#endif
#define SHL(x, n) ((x) << (n))
#define SHR(x, n) ((x) >> (n))
#define CONST_EXP2 q[i+0] + SPH_ROTL64(q[i+1], 5) + q[i+2] + SPH_ROTL64(q[i+3], 11) + \
q[i+4] + SPH_ROTL64(q[i+5], 27) + q[i+6] + SPH_ROTL64(q[i+7], 32) + \
q[i+8] + SPH_ROTL64(q[i+9], 37) + q[i+10] + SPH_ROTL64(q[i+11], 43) + \
q[i+12] + SPH_ROTL64(q[i+13], 53) + (SHR(q[i+14],1) ^ q[i+14]) + (SHR(q[i+15],2) ^ q[i+15])
typedef union {
unsigned char h1[64];
uint h4[16];
@ -133,8 +123,9 @@ __kernel void search(__global unsigned char* block, __global hash_t* hashes) @@ -133,8 +123,9 @@ __kernel void search(__global unsigned char* block, __global hash_t* hashes)
sph_u64 T0 = SPH_C64(0xFFFFFFFFFFFFFC00) + (80 << 3), T1 = 0xFFFFFFFFFFFFFFFF;;
if ((T0 = SPH_T64(T0 + 1024)) < 1024)
{
T1 = SPH_T64(T1 + 1);
}
sph_u64 M0, M1, M2, M3, M4, M5, M6, M7;
sph_u64 M8, M9, MA, MB, MC, MD, ME, MF;
sph_u64 V0, V1, V2, V3, V4, V5, V6, V7;
@ -169,6 +160,7 @@ __kernel void search(__global unsigned char* block, __global hash_t* hashes) @@ -169,6 +160,7 @@ __kernel void search(__global unsigned char* block, __global hash_t* hashes)
hash->h8[6] = H6;
hash->h8[7] = H7;
barrier(CLK_GLOBAL_MEM_FENCE);
}
@ -177,268 +169,58 @@ __kernel void search1(__global hash_t* hashes) @@ -177,268 +169,58 @@ __kernel void search1(__global hash_t* hashes)
{
uint gid = get_global_id(0);
__global hash_t *hash = &(hashes[gid-get_global_offset(0)]);
// bmw
sph_u64 BMW_H[16];
#pragma unroll 16
for(unsigned u = 0; u < 16; u++)
BMW_H[u] = BMW_IV512[u];
sph_u64 mv[16],q[32];
sph_u64 tmp;
mv[0] = SWAP8(hash->h8[0]);
mv[1] = SWAP8(hash->h8[1]);
mv[2] = SWAP8(hash->h8[2]);
mv[3] = SWAP8(hash->h8[3]);
mv[4] = SWAP8(hash->h8[4]);
mv[5] = SWAP8(hash->h8[5]);
mv[6] = SWAP8(hash->h8[6]);
mv[7] = SWAP8(hash->h8[7]);
mv[8] = 0x80;
mv[9] = 0;
sph_u64 BMW_h1[16], BMW_h2[16];
sph_u64 mv[16];
mv[ 0] = SWAP8(hash->h8[0]);
mv[ 1] = SWAP8(hash->h8[1]);
mv[ 2] = SWAP8(hash->h8[2]);
mv[ 3] = SWAP8(hash->h8[3]);
mv[ 4] = SWAP8(hash->h8[4]);
mv[ 5] = SWAP8(hash->h8[5]);
mv[ 6] = SWAP8(hash->h8[6]);
mv[ 7] = SWAP8(hash->h8[7]);
mv[ 8] = 0x80;
mv[ 9] = 0;
mv[10] = 0;
mv[11] = 0;
mv[12] = 0;
mv[13] = 0;
mv[14] = 0;
mv[15] = SPH_C64(512);
tmp = (mv[5] ^ BMW_H[5]) - (mv[7] ^ BMW_H[7]) + (mv[10] ^ BMW_H[10]) + (mv[13] ^ BMW_H[13]) + (mv[14] ^ BMW_H[14]);
q[0] = (SHR(tmp, 1) ^ SHL(tmp, 3) ^ SPH_ROTL64(tmp, 4) ^ SPH_ROTL64(tmp, 37)) + BMW_H[1];
tmp = (mv[6] ^ BMW_H[6]) - (mv[8] ^ BMW_H[8]) + (mv[11] ^ BMW_H[11]) + (mv[14] ^ BMW_H[14]) - (mv[15] ^ BMW_H[15]);
q[1] = (SHR(tmp, 1) ^ SHL(tmp, 2) ^ SPH_ROTL64(tmp, 13) ^ SPH_ROTL64(tmp, 43)) + BMW_H[2];
tmp = (mv[0] ^ BMW_H[0]) + (mv[7] ^ BMW_H[7]) + (mv[9] ^ BMW_H[9]) - (mv[12] ^ BMW_H[12]) + (mv[15] ^ BMW_H[15]);
q[2] = (SHR(tmp, 2) ^ SHL(tmp, 1) ^ SPH_ROTL64(tmp, 19) ^ SPH_ROTL64(tmp, 53)) + BMW_H[3];
tmp = (mv[0] ^ BMW_H[0]) - (mv[1] ^ BMW_H[1]) + (mv[8] ^ BMW_H[8]) - (mv[10] ^ BMW_H[10]) + (mv[13] ^ BMW_H[13]);
q[3] = (SHR(tmp, 2) ^ SHL(tmp, 2) ^ SPH_ROTL64(tmp, 28) ^ SPH_ROTL64(tmp, 59)) + BMW_H[4];
tmp = (mv[1] ^ BMW_H[1]) + (mv[2] ^ BMW_H[2]) + (mv[9] ^ BMW_H[9]) - (mv[11] ^ BMW_H[11]) - (mv[14] ^ BMW_H[14]);
q[4] = (SHR(tmp, 1) ^ tmp) + BMW_H[5];
tmp = (mv[3] ^ BMW_H[3]) - (mv[2] ^ BMW_H[2]) + (mv[10] ^ BMW_H[10]) - (mv[12] ^ BMW_H[12]) + (mv[15] ^ BMW_H[15]);
q[5] = (SHR(tmp, 1) ^ SHL(tmp, 3) ^ SPH_ROTL64(tmp, 4) ^ SPH_ROTL64(tmp, 37)) + BMW_H[6];
tmp = (mv[4] ^ BMW_H[4]) - (mv[0] ^ BMW_H[0]) - (mv[3] ^ BMW_H[3]) - (mv[11] ^ BMW_H[11]) + (mv[13] ^ BMW_H[13]);
q[6] = (SHR(tmp, 1) ^ SHL(tmp, 2) ^ SPH_ROTL64(tmp, 13) ^ SPH_ROTL64(tmp, 43)) + BMW_H[7];
tmp = (mv[1] ^ BMW_H[1]) - (mv[4] ^ BMW_H[4]) - (mv[5] ^ BMW_H[5]) - (mv[12] ^ BMW_H[12]) - (mv[14] ^ BMW_H[14]);
q[7] = (SHR(tmp, 2) ^ SHL(tmp, 1) ^ SPH_ROTL64(tmp, 19) ^ SPH_ROTL64(tmp, 53)) + BMW_H[8];
tmp = (mv[2] ^ BMW_H[2]) - (mv[5] ^ BMW_H[5]) - (mv[6] ^ BMW_H[6]) + (mv[13] ^ BMW_H[13]) - (mv[15] ^ BMW_H[15]);
q[8] = (SHR(tmp, 2) ^ SHL(tmp, 2) ^ SPH_ROTL64(tmp, 28) ^ SPH_ROTL64(tmp, 59)) + BMW_H[9];
tmp = (mv[0] ^ BMW_H[0]) - (mv[3] ^ BMW_H[3]) + (mv[6] ^ BMW_H[6]) - (mv[7] ^ BMW_H[7]) + (mv[14] ^ BMW_H[14]);
q[9] = (SHR(tmp, 1) ^ tmp) + BMW_H[10];
tmp = (mv[8] ^ BMW_H[8]) - (mv[1] ^ BMW_H[1]) - (mv[4] ^ BMW_H[4]) - (mv[7] ^ BMW_H[7]) + (mv[15] ^ BMW_H[15]);
q[10] = (SHR(tmp, 1) ^ SHL(tmp, 3) ^ SPH_ROTL64(tmp, 4) ^ SPH_ROTL64(tmp, 37)) + BMW_H[11];
tmp = (mv[8] ^ BMW_H[8]) - (mv[0] ^ BMW_H[0]) - (mv[2] ^ BMW_H[2]) - (mv[5] ^ BMW_H[5]) + (mv[9] ^ BMW_H[9]);
q[11] = (SHR(tmp, 1) ^ SHL(tmp, 2) ^ SPH_ROTL64(tmp, 13) ^ SPH_ROTL64(tmp, 43)) + BMW_H[12];
tmp = (mv[1] ^ BMW_H[1]) + (mv[3] ^ BMW_H[3]) - (mv[6] ^ BMW_H[6]) - (mv[9] ^ BMW_H[9]) + (mv[10] ^ BMW_H[10]);
q[12] = (SHR(tmp, 2) ^ SHL(tmp, 1) ^ SPH_ROTL64(tmp, 19) ^ SPH_ROTL64(tmp, 53)) + BMW_H[13];
tmp = (mv[2] ^ BMW_H[2]) + (mv[4] ^ BMW_H[4]) + (mv[7] ^ BMW_H[7]) + (mv[10] ^ BMW_H[10]) + (mv[11] ^ BMW_H[11]);
q[13] = (SHR(tmp, 2) ^ SHL(tmp, 2) ^ SPH_ROTL64(tmp, 28) ^ SPH_ROTL64(tmp, 59)) + BMW_H[14];
tmp = (mv[3] ^ BMW_H[3]) - (mv[5] ^ BMW_H[5]) + (mv[8] ^ BMW_H[8]) - (mv[11] ^ BMW_H[11]) - (mv[12] ^ BMW_H[12]);
q[14] = (SHR(tmp, 1) ^ tmp) + BMW_H[15];
tmp = (mv[12] ^ BMW_H[12]) - (mv[4] ^ BMW_H[4]) - (mv[6] ^ BMW_H[6]) - (mv[9] ^ BMW_H[9]) + (mv[13] ^ BMW_H[13]);
q[15] = (SHR(tmp, 1) ^ SHL(tmp, 3) ^ SPH_ROTL64(tmp, 4) ^ SPH_ROTL64(tmp, 37)) + BMW_H[0];
#pragma unroll 2
for(int i=0;i<2;i++)
{
q[i+16] =
(SHR(q[i], 1) ^ SHL(q[i], 2) ^ SPH_ROTL64(q[i], 13) ^ SPH_ROTL64(q[i], 43)) +
(SHR(q[i+1], 2) ^ SHL(q[i+1], 1) ^ SPH_ROTL64(q[i+1], 19) ^ SPH_ROTL64(q[i+1], 53)) +
(SHR(q[i+2], 2) ^ SHL(q[i+2], 2) ^ SPH_ROTL64(q[i+2], 28) ^ SPH_ROTL64(q[i+2], 59)) +
(SHR(q[i+3], 1) ^ SHL(q[i+3], 3) ^ SPH_ROTL64(q[i+3], 4) ^ SPH_ROTL64(q[i+3], 37)) +
(SHR(q[i+4], 1) ^ SHL(q[i+4], 2) ^ SPH_ROTL64(q[i+4], 13) ^ SPH_ROTL64(q[i+4], 43)) +
(SHR(q[i+5], 2) ^ SHL(q[i+5], 1) ^ SPH_ROTL64(q[i+5], 19) ^ SPH_ROTL64(q[i+5], 53)) +
(SHR(q[i+6], 2) ^ SHL(q[i+6], 2) ^ SPH_ROTL64(q[i+6], 28) ^ SPH_ROTL64(q[i+6], 59)) +
(SHR(q[i+7], 1) ^ SHL(q[i+7], 3) ^ SPH_ROTL64(q[i+7], 4) ^ SPH_ROTL64(q[i+7], 37)) +
(SHR(q[i+8], 1) ^ SHL(q[i+8], 2) ^ SPH_ROTL64(q[i+8], 13) ^ SPH_ROTL64(q[i+8], 43)) +
(SHR(q[i+9], 2) ^ SHL(q[i+9], 1) ^ SPH_ROTL64(q[i+9], 19) ^ SPH_ROTL64(q[i+9], 53)) +
(SHR(q[i+10], 2) ^ SHL(q[i+10], 2) ^ SPH_ROTL64(q[i+10], 28) ^ SPH_ROTL64(q[i+10], 59)) +
(SHR(q[i+11], 1) ^ SHL(q[i+11], 3) ^ SPH_ROTL64(q[i+11], 4) ^ SPH_ROTL64(q[i+11], 37)) +
(SHR(q[i+12], 1) ^ SHL(q[i+12], 2) ^ SPH_ROTL64(q[i+12], 13) ^ SPH_ROTL64(q[i+12], 43)) +
(SHR(q[i+13], 2) ^ SHL(q[i+13], 1) ^ SPH_ROTL64(q[i+13], 19) ^ SPH_ROTL64(q[i+13], 53)) +
(SHR(q[i+14], 2) ^ SHL(q[i+14], 2) ^ SPH_ROTL64(q[i+14], 28) ^ SPH_ROTL64(q[i+14], 59)) +
(SHR(q[i+15], 1) ^ SHL(q[i+15], 3) ^ SPH_ROTL64(q[i+15], 4) ^ SPH_ROTL64(q[i+15], 37)) +
(( ((i+16)*(0x0555555555555555ull)) + SPH_ROTL64(mv[i], i+1) +
SPH_ROTL64(mv[i+3], i+4) - SPH_ROTL64(mv[i+10], i+11) ) ^ BMW_H[i+7]);
}
#pragma unroll 4
for(int i=2;i<6;i++)
{
q[i+16] = CONST_EXP2 +
(( ((i+16)*(0x0555555555555555ull)) + SPH_ROTL64(mv[i], i+1) +
SPH_ROTL64(mv[i+3], i+4) - SPH_ROTL64(mv[i+10], i+11) ) ^ BMW_H[i+7]);
}
mv[15] = 0x200;
#define M(x) (mv[x])
#define H(x) (BMW_H[x])
#define dH(x) (BMW_h2[x])
#pragma unroll 3
for(int i=6;i<9;i++)
{
q[i+16] = CONST_EXP2 +
(( ((i+16)*(0x0555555555555555ull)) + SPH_ROTL64(mv[i], i+1) +
SPH_ROTL64(mv[i+3], i+4) - SPH_ROTL64(mv[i-6], (i-6)+1) ) ^ BMW_H[i+7]);
}
FOLDb;
#pragma unroll 4
for(int i=9;i<13;i++)
{
q[i+16] = CONST_EXP2 +
(( ((i+16)*(0x0555555555555555ull)) + SPH_ROTL64(mv[i], i+1) +
SPH_ROTL64(mv[i+3], i+4) - SPH_ROTL64(mv[i-6], (i-6)+1) ) ^ BMW_H[i-9]);
}
#undef M
#undef H
#undef dH
#pragma unroll 3
for(int i=13;i<16;i++)
{
q[i+16] = CONST_EXP2 +
(( ((i+16)*(0x0555555555555555ull)) + SPH_ROTL64(mv[i], i+1) +
SPH_ROTL64(mv[i-13], (i-13)+1) - SPH_ROTL64(mv[i-6], (i-6)+1) ) ^ BMW_H[i-9]);
}
#define M(x) (BMW_h2[x])
#define H(x) (final_b[x])
#define dH(x) (BMW_h1[x])
sph_u64 XL64 = q[16]^q[17]^q[18]^q[19]^q[20]^q[21]^q[22]^q[23];
sph_u64 XH64 = XL64^q[24]^q[25]^q[26]^q[27]^q[28]^q[29]^q[30]^q[31];
BMW_H[0] = (SHL(XH64, 5) ^ SHR(q[16],5) ^ mv[0]) + ( XL64 ^ q[24] ^ q[0]);
BMW_H[1] = (SHR(XH64, 7) ^ SHL(q[17],8) ^ mv[1]) + ( XL64 ^ q[25] ^ q[1]);
BMW_H[2] = (SHR(XH64, 5) ^ SHL(q[18],5) ^ mv[2]) + ( XL64 ^ q[26] ^ q[2]);
BMW_H[3] = (SHR(XH64, 1) ^ SHL(q[19],5) ^ mv[3]) + ( XL64 ^ q[27] ^ q[3]);
BMW_H[4] = (SHR(XH64, 3) ^ q[20] ^ mv[4]) + ( XL64 ^ q[28] ^ q[4]);
BMW_H[5] = (SHL(XH64, 6) ^ SHR(q[21],6) ^ mv[5]) + ( XL64 ^ q[29] ^ q[5]);
BMW_H[6] = (SHR(XH64, 4) ^ SHL(q[22],6) ^ mv[6]) + ( XL64 ^ q[30] ^ q[6]);
BMW_H[7] = (SHR(XH64,11) ^ SHL(q[23],2) ^ mv[7]) + ( XL64 ^ q[31] ^ q[7]);
BMW_H[8] = SPH_ROTL64(BMW_H[4], 9) + ( XH64 ^ q[24] ^ mv[8]) + (SHL(XL64,8) ^ q[23] ^ q[8]);
BMW_H[9] = SPH_ROTL64(BMW_H[5],10) + ( XH64 ^ q[25] ^ mv[9]) + (SHR(XL64,6) ^ q[16] ^ q[9]);
BMW_H[10] = SPH_ROTL64(BMW_H[6],11) + ( XH64 ^ q[26] ^ mv[10]) + (SHL(XL64,6) ^ q[17] ^ q[10]);
BMW_H[11] = SPH_ROTL64(BMW_H[7],12) + ( XH64 ^ q[27] ^ mv[11]) + (SHL(XL64,4) ^ q[18] ^ q[11]);
BMW_H[12] = SPH_ROTL64(BMW_H[0],13) + ( XH64 ^ q[28] ^ mv[12]) + (SHR(XL64,3) ^ q[19] ^ q[12]);
BMW_H[13] = SPH_ROTL64(BMW_H[1],14) + ( XH64 ^ q[29] ^ mv[13]) + (SHR(XL64,4) ^ q[20] ^ q[13]);
BMW_H[14] = SPH_ROTL64(BMW_H[2],15) + ( XH64 ^ q[30] ^ mv[14]) + (SHR(XL64,7) ^ q[21] ^ q[14]);
BMW_H[15] = SPH_ROTL64(BMW_H[3],16) + ( XH64 ^ q[31] ^ mv[15]) + (SHR(XL64,2) ^ q[22] ^ q[15]);
#pragma unroll 16
for(int i=0;i<16;i++)
{
mv[i] = BMW_H[i];
BMW_H[i] = 0xaaaaaaaaaaaaaaa0ull + (sph_u64)i;
}
FOLDb;
tmp = (mv[5] ^ BMW_H[5]) - (mv[7] ^ BMW_H[7]) + (mv[10] ^ BMW_H[10]) + (mv[13] ^ BMW_H[13]) + (mv[14] ^ BMW_H[14]);
q[0] = (SHR(tmp, 1) ^ SHL(tmp, 3) ^ SPH_ROTL64(tmp, 4) ^ SPH_ROTL64(tmp, 37)) + BMW_H[1];
tmp = (mv[6] ^ BMW_H[6]) - (mv[8] ^ BMW_H[8]) + (mv[11] ^ BMW_H[11]) + (mv[14] ^ BMW_H[14]) - (mv[15] ^ BMW_H[15]);
q[1] = (SHR(tmp, 1) ^ SHL(tmp, 2) ^ SPH_ROTL64(tmp, 13) ^ SPH_ROTL64(tmp, 43)) + BMW_H[2];
tmp = (mv[0] ^ BMW_H[0]) + (mv[7] ^ BMW_H[7]) + (mv[9] ^ BMW_H[9]) - (mv[12] ^ BMW_H[12]) + (mv[15] ^ BMW_H[15]);
q[2] = (SHR(tmp, 2) ^ SHL(tmp, 1) ^ SPH_ROTL64(tmp, 19) ^ SPH_ROTL64(tmp, 53)) + BMW_H[3];
tmp = (mv[0] ^ BMW_H[0]) - (mv[1] ^ BMW_H[1]) + (mv[8] ^ BMW_H[8]) - (mv[10] ^ BMW_H[10]) + (mv[13] ^ BMW_H[13]);
q[3] = (SHR(tmp, 2) ^ SHL(tmp, 2) ^ SPH_ROTL64(tmp, 28) ^ SPH_ROTL64(tmp, 59)) + BMW_H[4];
tmp = (mv[1] ^ BMW_H[1]) + (mv[2] ^ BMW_H[2]) + (mv[9] ^ BMW_H[9]) - (mv[11] ^ BMW_H[11]) - (mv[14] ^ BMW_H[14]);
q[4] = (SHR(tmp, 1) ^ tmp) + BMW_H[5];
tmp = (mv[3] ^ BMW_H[3]) - (mv[2] ^ BMW_H[2]) + (mv[10] ^ BMW_H[10]) - (mv[12] ^ BMW_H[12]) + (mv[15] ^ BMW_H[15]);
q[5] = (SHR(tmp, 1) ^ SHL(tmp, 3) ^ SPH_ROTL64(tmp, 4) ^ SPH_ROTL64(tmp, 37)) + BMW_H[6];
tmp = (mv[4] ^ BMW_H[4]) - (mv[0] ^ BMW_H[0]) - (mv[3] ^ BMW_H[3]) - (mv[11] ^ BMW_H[11]) + (mv[13] ^ BMW_H[13]);
q[6] = (SHR(tmp, 1) ^ SHL(tmp, 2) ^ SPH_ROTL64(tmp, 13) ^ SPH_ROTL64(tmp, 43)) + BMW_H[7];
tmp = (mv[1] ^ BMW_H[1]) - (mv[4] ^ BMW_H[4]) - (mv[5] ^ BMW_H[5]) - (mv[12] ^ BMW_H[12]) - (mv[14] ^ BMW_H[14]);
q[7] = (SHR(tmp, 2) ^ SHL(tmp, 1) ^ SPH_ROTL64(tmp, 19) ^ SPH_ROTL64(tmp, 53)) + BMW_H[8];
tmp = (mv[2] ^ BMW_H[2]) - (mv[5] ^ BMW_H[5]) - (mv[6] ^ BMW_H[6]) + (mv[13] ^ BMW_H[13]) - (mv[15] ^ BMW_H[15]);
q[8] = (SHR(tmp, 2) ^ SHL(tmp, 2) ^ SPH_ROTL64(tmp, 28) ^ SPH_ROTL64(tmp, 59)) + BMW_H[9];
tmp = (mv[0] ^ BMW_H[0]) - (mv[3] ^ BMW_H[3]) + (mv[6] ^ BMW_H[6]) - (mv[7] ^ BMW_H[7]) + (mv[14] ^ BMW_H[14]);
q[9] = (SHR(tmp, 1) ^ tmp) + BMW_H[10];
tmp = (mv[8] ^ BMW_H[8]) - (mv[1] ^ BMW_H[1]) - (mv[4] ^ BMW_H[4]) - (mv[7] ^ BMW_H[7]) + (mv[15] ^ BMW_H[15]);
q[10] = (SHR(tmp, 1) ^ SHL(tmp, 3) ^ SPH_ROTL64(tmp, 4) ^ SPH_ROTL64(tmp, 37)) + BMW_H[11];
tmp = (mv[8] ^ BMW_H[8]) - (mv[0] ^ BMW_H[0]) - (mv[2] ^ BMW_H[2]) - (mv[5] ^ BMW_H[5]) + (mv[9] ^ BMW_H[9]);
q[11] = (SHR(tmp, 1) ^ SHL(tmp, 2) ^ SPH_ROTL64(tmp, 13) ^ SPH_ROTL64(tmp, 43)) + BMW_H[12];
tmp = (mv[1] ^ BMW_H[1]) + (mv[3] ^ BMW_H[3]) - (mv[6] ^ BMW_H[6]) - (mv[9] ^ BMW_H[9]) + (mv[10] ^ BMW_H[10]);
q[12] = (SHR(tmp, 2) ^ SHL(tmp, 1) ^ SPH_ROTL64(tmp, 19) ^ SPH_ROTL64(tmp, 53)) + BMW_H[13];
tmp = (mv[2] ^ BMW_H[2]) + (mv[4] ^ BMW_H[4]) + (mv[7] ^ BMW_H[7]) + (mv[10] ^ BMW_H[10]) + (mv[11] ^ BMW_H[11]);
q[13] = (SHR(tmp, 2) ^ SHL(tmp, 2) ^ SPH_ROTL64(tmp, 28) ^ SPH_ROTL64(tmp, 59)) + BMW_H[14];
tmp = (mv[3] ^ BMW_H[3]) - (mv[5] ^ BMW_H[5]) + (mv[8] ^ BMW_H[8]) - (mv[11] ^ BMW_H[11]) - (mv[12] ^ BMW_H[12]);
q[14] = (SHR(tmp, 1) ^ tmp) + BMW_H[15];
tmp = (mv[12] ^ BMW_H[12]) - (mv[4] ^ BMW_H[4]) - (mv[6] ^ BMW_H[6]) - (mv[9] ^ BMW_H[9]) + (mv[13] ^ BMW_H[13]);
q[15] = (SHR(tmp, 1) ^ SHL(tmp, 3) ^ SPH_ROTL64(tmp, 4) ^ SPH_ROTL64(tmp, 37)) + BMW_H[0];
#pragma unroll 2
for(int i=0;i<2;i++)
{
q[i+16] =
(SHR(q[i], 1) ^ SHL(q[i], 2) ^ SPH_ROTL64(q[i], 13) ^ SPH_ROTL64(q[i], 43)) +
(SHR(q[i+1], 2) ^ SHL(q[i+1], 1) ^ SPH_ROTL64(q[i+1], 19) ^ SPH_ROTL64(q[i+1], 53)) +
(SHR(q[i+2], 2) ^ SHL(q[i+2], 2) ^ SPH_ROTL64(q[i+2], 28) ^ SPH_ROTL64(q[i+2], 59)) +
(SHR(q[i+3], 1) ^ SHL(q[i+3], 3) ^ SPH_ROTL64(q[i+3], 4) ^ SPH_ROTL64(q[i+3], 37)) +
(SHR(q[i+4], 1) ^ SHL(q[i+4], 2) ^ SPH_ROTL64(q[i+4], 13) ^ SPH_ROTL64(q[i+4], 43)) +
(SHR(q[i+5], 2) ^ SHL(q[i+5], 1) ^ SPH_ROTL64(q[i+5], 19) ^ SPH_ROTL64(q[i+5], 53)) +
(SHR(q[i+6], 2) ^ SHL(q[i+6], 2) ^ SPH_ROTL64(q[i+6], 28) ^ SPH_ROTL64(q[i+6], 59)) +
(SHR(q[i+7], 1) ^ SHL(q[i+7], 3) ^ SPH_ROTL64(q[i+7], 4) ^ SPH_ROTL64(q[i+7], 37)) +
(SHR(q[i+8], 1) ^ SHL(q[i+8], 2) ^ SPH_ROTL64(q[i+8], 13) ^ SPH_ROTL64(q[i+8], 43)) +
(SHR(q[i+9], 2) ^ SHL(q[i+9], 1) ^ SPH_ROTL64(q[i+9], 19) ^ SPH_ROTL64(q[i+9], 53)) +
(SHR(q[i+10], 2) ^ SHL(q[i+10], 2) ^ SPH_ROTL64(q[i+10], 28) ^ SPH_ROTL64(q[i+10], 59)) +
(SHR(q[i+11], 1) ^ SHL(q[i+11], 3) ^ SPH_ROTL64(q[i+11], 4) ^ SPH_ROTL64(q[i+11], 37)) +
(SHR(q[i+12], 1) ^ SHL(q[i+12], 2) ^ SPH_ROTL64(q[i+12], 13) ^ SPH_ROTL64(q[i+12], 43)) +
(SHR(q[i+13], 2) ^ SHL(q[i+13], 1) ^ SPH_ROTL64(q[i+13], 19) ^ SPH_ROTL64(q[i+13], 53)) +
(SHR(q[i+14], 2) ^ SHL(q[i+14], 2) ^ SPH_ROTL64(q[i+14], 28) ^ SPH_ROTL64(q[i+14], 59)) +
(SHR(q[i+15], 1) ^ SHL(q[i+15], 3) ^ SPH_ROTL64(q[i+15], 4) ^ SPH_ROTL64(q[i+15], 37)) +
(( ((i+16)*(0x0555555555555555ull)) + SPH_ROTL64(mv[i], i+1) +
SPH_ROTL64(mv[i+3], i+4) - SPH_ROTL64(mv[i+10], i+11) ) ^ BMW_H[i+7]);
}
#undef M
#undef H
#undef dH
#pragma unroll 4
for(int i=2;i<6;i++)
{
q[i+16] = CONST_EXP2 +
(( ((i+16)*(0x0555555555555555ull)) + SPH_ROTL64(mv[i], i+1) +
SPH_ROTL64(mv[i+3], i+4) - SPH_ROTL64(mv[i+10], i+11) ) ^ BMW_H[i+7]);
}
#pragma unroll 3
for(int i=6;i<9;i++)
{
q[i+16] = CONST_EXP2 +
(( ((i+16)*(0x0555555555555555ull)) + SPH_ROTL64(mv[i], i+1) +
SPH_ROTL64(mv[i+3], i+4) - SPH_ROTL64(mv[i-6], (i-6)+1) ) ^ BMW_H[i+7]);
}
#pragma unroll 4
for(int i=9;i<13;i++)
{
q[i+16] = CONST_EXP2 +
(( ((i+16)*(0x0555555555555555ull)) + SPH_ROTL64(mv[i], i+1) +
SPH_ROTL64(mv[i+3], i+4) - SPH_ROTL64(mv[i-6], (i-6)+1) ) ^ BMW_H[i-9]);
}
#pragma unroll 3
for(int i=13;i<16;i++)
{
q[i+16] = CONST_EXP2 +
(( ((i+16)*(0x0555555555555555ull)) + SPH_ROTL64(mv[i], i+1) +
SPH_ROTL64(mv[i-13], (i-13)+1) - SPH_ROTL64(mv[i-6], (i-6)+1) ) ^ BMW_H[i-9]);
}
XL64 = q[16]^q[17]^q[18]^q[19]^q[20]^q[21]^q[22]^q[23];
XH64 = XL64^q[24]^q[25]^q[26]^q[27]^q[28]^q[29]^q[30]^q[31];
BMW_H[0] = (SHL(XH64, 5) ^ SHR(q[16],5) ^ mv[0]) + ( XL64 ^ q[24] ^ q[0]);
BMW_H[1] = (SHR(XH64, 7) ^ SHL(q[17],8) ^ mv[1]) + ( XL64 ^ q[25] ^ q[1]);
BMW_H[2] = (SHR(XH64, 5) ^ SHL(q[18],5) ^ mv[2]) + ( XL64 ^ q[26] ^ q[2]);
BMW_H[3] = (SHR(XH64, 1) ^ SHL(q[19],5) ^ mv[3]) + ( XL64 ^ q[27] ^ q[3]);
BMW_H[4] = (SHR(XH64, 3) ^ q[20] ^ mv[4]) + ( XL64 ^ q[28] ^ q[4]);
BMW_H[5] = (SHL(XH64, 6) ^ SHR(q[21],6) ^ mv[5]) + ( XL64 ^ q[29] ^ q[5]);
BMW_H[6] = (SHR(XH64, 4) ^ SHL(q[22],6) ^ mv[6]) + ( XL64 ^ q[30] ^ q[6]);
BMW_H[7] = (SHR(XH64,11) ^ SHL(q[23],2) ^ mv[7]) + ( XL64 ^ q[31] ^ q[7]);
BMW_H[8] = SPH_ROTL64(BMW_H[4], 9) + ( XH64 ^ q[24] ^ mv[8]) + (SHL(XL64,8) ^ q[23] ^ q[8]);
BMW_H[9] = SPH_ROTL64(BMW_H[5],10) + ( XH64 ^ q[25] ^ mv[9]) + (SHR(XL64,6) ^ q[16] ^ q[9]);
BMW_H[10] = SPH_ROTL64(BMW_H[6],11) + ( XH64 ^ q[26] ^ mv[10]) + (SHL(XL64,6) ^ q[17] ^ q[10]);
BMW_H[11] = SPH_ROTL64(BMW_H[7],12) + ( XH64 ^ q[27] ^ mv[11]) + (SHL(XL64,4) ^ q[18] ^ q[11]);
BMW_H[12] = SPH_ROTL64(BMW_H[0],13) + ( XH64 ^ q[28] ^ mv[12]) + (SHR(XL64,3) ^ q[19] ^ q[12]);
BMW_H[13] = SPH_ROTL64(BMW_H[1],14) + ( XH64 ^ q[29] ^ mv[13]) + (SHR(XL64,4) ^ q[20] ^ q[13]);
BMW_H[14] = SPH_ROTL64(BMW_H[2],15) + ( XH64 ^ q[30] ^ mv[14]) + (SHR(XL64,7) ^ q[21] ^ q[14]);
BMW_H[15] = SPH_ROTL64(BMW_H[3],16) + ( XH64 ^ q[31] ^ mv[15]) + (SHR(XL64,2) ^ q[22] ^ q[15]);
hash->h8[0] = SWAP8(BMW_H[8]);
hash->h8[1] = SWAP8(BMW_H[9]);
hash->h8[2] = SWAP8(BMW_H[10]);
hash->h8[3] = SWAP8(BMW_H[11]);
hash->h8[4] = SWAP8(BMW_H[12]);
hash->h8[5] = SWAP8(BMW_H[13]);
hash->h8[6] = SWAP8(BMW_H[14]);
hash->h8[7] = SWAP8(BMW_H[15]);
hash->h8[0] = SWAP8(BMW_h1[8]);
hash->h8[1] = SWAP8(BMW_h1[9]);
hash->h8[2] = SWAP8(BMW_h1[10]);
hash->h8[3] = SWAP8(BMW_h1[11]);
hash->h8[4] = SWAP8(BMW_h1[12]);
hash->h8[5] = SWAP8(BMW_h1[13]);
hash->h8[6] = SWAP8(BMW_h1[14]);
hash->h8[7] = SWAP8(BMW_h1[15]);
barrier(CLK_GLOBAL_MEM_FENCE);
}
@ -449,95 +231,65 @@ __kernel void search2(__global hash_t* hashes) @@ -449,95 +231,65 @@ __kernel void search2(__global hash_t* hashes)
uint gid = get_global_id(0);
__global hash_t *hash = &(hashes[gid-get_global_offset(0)]);
#if !SPH_SMALL_FOOTPRINT_GROESTL
__local sph_u64 T0_C[256], T1_C[256], T2_C[256], T3_C[256];
__local sph_u64 T4_C[256], T5_C[256], T6_C[256], T7_C[256];
#else
__local sph_u64 T0_C[256], T4_C[256];
#endif
__local sph_u64 T0_L[256], T1_L[256], T2_L[256], T3_L[256], T4_L[256], T5_L[256], T6_L[256], T7_L[256];
int init = get_local_id(0);
int step = get_local_size(0);
for (int i = init; i < 256; i += step)
{
T0_C[i] = T0[i];
T4_C[i] = T4[i];
#if !SPH_SMALL_FOOTPRINT_GROESTL
T1_C[i] = T1[i];
T2_C[i] = T2[i];
T3_C[i] = T3[i];
T5_C[i] = T5[i];
T6_C[i] = T6[i];
T7_C[i] = T7[i];
#endif
T0_L[i] = T0[i];
T1_L[i] = T1[i];
T2_L[i] = T2[i];
T3_L[i] = T3[i];
T4_L[i] = T4[i];
T5_L[i] = T5[i];
T6_L[i] = T6[i];
T7_L[i] = T7[i];
}
barrier(CLK_LOCAL_MEM_FENCE);
barrier(CLK_LOCAL_MEM_FENCE); // groestl
#define T0 T0_C
#define T1 T1_C
#define T2 T2_C
#define T3 T3_C
#define T4 T4_C
#define T5 T5_C
#define T6 T6_C
#define T7 T7_C
#define T0 T0_L
#define T1 T1_L
#define T2 T2_L
#define T3 T3_L
#define T4 T4_L
#define T5 T5_L
#define T6 T6_L
#define T7 T7_L
// groestl
sph_u64 H[16];
for (unsigned int u = 0; u < 15; u ++)
H[u] = 0;
#if USE_LE
H[15] = ((sph_u64)(512 & 0xFF) << 56) | ((sph_u64)(512 & 0xFF00) << 40);
#else
H[15] = (sph_u64)512;
#endif
sph_u64 H[16] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0x0002000000000000};
sph_u64 g[16], m[16];
m[0] = DEC64E(hash->h8[0]);
m[1] = DEC64E(hash->h8[1]);
m[2] = DEC64E(hash->h8[2]);
m[3] = DEC64E(hash->h8[3]);
m[4] = DEC64E(hash->h8[4]);
m[5] = DEC64E(hash->h8[5]);
m[6] = DEC64E(hash->h8[6]);
m[7] = DEC64E(hash->h8[7]);
for (unsigned int u = 0; u < 16; u ++)
g[u] = m[u] ^ H[u];
m[8] = 0x80; g[8] = m[8] ^ H[8];
m[9] = 0; g[9] = m[9] ^ H[9];
m[10] = 0; g[10] = m[10] ^ H[10];
m[11] = 0; g[11] = m[11] ^ H[11];
m[12] = 0; g[12] = m[12] ^ H[12];
m[13] = 0; g[13] = m[13] ^ H[13];
m[14] = 0; g[14] = m[14] ^ H[14];
m[15] = 0x100000000000000; g[15] = m[15] ^ H[15];
g[0] = m[0] = DEC64E(hash->h8[0]);
g[1] = m[1] = DEC64E(hash->h8[1]);
g[2] = m[2] = DEC64E(hash->h8[2]);
g[3] = m[3] = DEC64E(hash->h8[3]);
g[4] = m[4] = DEC64E(hash->h8[4]);
g[5] = m[5] = DEC64E(hash->h8[5]);
g[6] = m[6] = DEC64E(hash->h8[6]);
g[7] = m[7] = DEC64E(hash->h8[7]);
g[8] = m[8] = 0x80;
g[9] = m[9] = 0;
g[10] = m[10] = 0;
g[11] = m[11] = 0;
g[12] = m[12] = 0;
g[13] = m[13] = 0;
g[14] = m[14] = 0;
g[15] = 0x102000000000000;
m[15] = 0x100000000000000;
PERM_BIG_P(g);
PERM_BIG_Q(m);
for (unsigned int u = 0; u < 16; u ++)
H[u] ^= g[u] ^ m[u];
sph_u64 xH[16];
for (unsigned int u = 0; u < 16; u ++)
xH[u] = H[u];
xH[u] = H[u] ^= g[u] ^ m[u];
PERM_BIG_P(xH);
for (unsigned int u = 0; u < 16; u ++)
H[u] ^= xH[u];
for (unsigned int u = 0; u < 8; u ++)
hash->h8[u] = DEC64E(H[u + 8]);
for (unsigned int u = 8; u < 16; u ++)
hash->h8[u-8] = DEC64E(H[u] ^ xH[u]);
barrier(CLK_GLOBAL_MEM_FENCE);
}
@ -591,8 +343,7 @@ __kernel void search4(__global hash_t* hashes) @@ -591,8 +343,7 @@ __kernel void search4(__global hash_t* hashes)
for(int i = 0; i < 2; i++)
{
if (i == 0)
{
if (i == 0) {
h0h ^= DEC64E(hash->h8[0]);
h0l ^= DEC64E(hash->h8[1]);
h1h ^= DEC64E(hash->h8[2]);
@ -601,9 +352,7 @@ __kernel void search4(__global hash_t* hashes) @@ -601,9 +352,7 @@ __kernel void search4(__global hash_t* hashes)
h2l ^= DEC64E(hash->h8[5]);
h3h ^= DEC64E(hash->h8[6]);
h3l ^= DEC64E(hash->h8[7]);
}
else if(i == 1)
{
} else if(i == 1) {
h4h ^= DEC64E(hash->h8[0]);
h4l ^= DEC64E(hash->h8[1]);
h5h ^= DEC64E(hash->h8[2]);
@ -633,6 +382,7 @@ __kernel void search4(__global hash_t* hashes) @@ -633,6 +382,7 @@ __kernel void search4(__global hash_t* hashes)
barrier(CLK_GLOBAL_MEM_FENCE);
}
__attribute__((reqd_work_group_size(WORKSIZE, 1, 1)))
__kernel void search5(__global hash_t* hashes)
{
@ -710,8 +460,7 @@ __kernel void search6(__global hash_t* hashes) @@ -710,8 +460,7 @@ __kernel void search6(__global hash_t* hashes)
MI5;
LUFFA_P5;
if(i == 0)
{
if(i == 0) {
M0 = hash->h4[9];
M1 = hash->h4[8];
M2 = hash->h4[11];
@ -720,16 +469,12 @@ __kernel void search6(__global hash_t* hashes) @@ -720,16 +469,12 @@ __kernel void search6(__global hash_t* hashes)
M5 = hash->h4[12];
M6 = hash->h4[15];
M7 = hash->h4[14];
}
else if(i == 1)
{
} else if(i == 1) {
M0 = 0x80000000;
M1 = M2 = M3 = M4 = M5 = M6 = M7 = 0;
}
else if(i == 2)
} else if(i == 2) {
M0 = M1 = M2 = M3 = M4 = M5 = M6 = M7 = 0;
else if(i == 3)
{
} else if(i == 3) {
hash->h4[1] = V00 ^ V10 ^ V20 ^ V30 ^ V40;
hash->h4[0] = V01 ^ V11 ^ V21 ^ V31 ^ V41;
hash->h4[3] = V02 ^ V12 ^ V22 ^ V32 ^ V42;
@ -779,12 +524,10 @@ __kernel void search7(__global hash_t* hashes) @@ -779,12 +524,10 @@ __kernel void search7(__global hash_t* hashes)
x6 ^= SWAP4(hash->h4[7]);
x7 ^= SWAP4(hash->h4[6]);
for (int i = 0; i < 13; i ++)
{
for (int i = 0; i < 13; i ++) {
SIXTEEN_ROUNDS;
if (i == 0)
{
if (i == 0) {
x0 ^= SWAP4(hash->h4[9]);
x1 ^= SWAP4(hash->h4[8]);
x2 ^= SWAP4(hash->h4[11]);
@ -793,12 +536,12 @@ __kernel void search7(__global hash_t* hashes) @@ -793,12 +536,12 @@ __kernel void search7(__global hash_t* hashes)
x5 ^= SWAP4(hash->h4[12]);
x6 ^= SWAP4(hash->h4[15]);
x7 ^= SWAP4(hash->h4[14]);
}
else if(i == 1)
} else if(i == 1) {
x0 ^= 0x80;
else if (i == 2)
} else if (i == 2) {
xv ^= SPH_C32(1);
}
}
hash->h4[0] = x0;
hash->h4[1] = x1;
@ -908,10 +651,8 @@ __kernel void search9(__global hash_t* hashes) @@ -908,10 +651,8 @@ __kernel void search9(__global hash_t* hashes)
// simd
s32 q[256];
unsigned char x[128];
for(unsigned int i = 0; i < 64; i++)
x[i] = hash->h1[i];
for(unsigned int i = 64; i < 128; i++)
x[i] = 0;
@ -1025,9 +766,9 @@ __attribute__((reqd_work_group_size(WORKSIZE, 1, 1))) @@ -1025,9 +766,9 @@ __attribute__((reqd_work_group_size(WORKSIZE, 1, 1)))
__kernel void search10(__global hash_t* hashes)
{
uint gid = get_global_id(0);
uint offset = get_global_offset(0);
hash_t hash;
__global hash_t *hashp = &(hashes[gid-offset]);
__global hash_t *hash = &(hashes[gid-get_global_offset(0)]);
__local sph_u32 AES0[256], AES1[256], AES2[256], AES3[256];
@ -1044,8 +785,9 @@ __kernel void search10(__global hash_t* hashes) @@ -1044,8 +785,9 @@ __kernel void search10(__global hash_t* hashes)
barrier(CLK_LOCAL_MEM_FENCE);
for (int i = 0; i < 8; i++)
hash.h8[i] = hashes[gid-offset].h8[i];
// echo
sph_u64 W00, W01, W10, W11, W20, W21, W30, W31, W40, W41, W50, W51, W60, W61, W70, W71, W80, W81, W90, W91, WA0, WA1, WB0, WB1, WC0, WC1, WD0, WD1, WE0, WE1, WF0, WF1;
@ -1074,14 +816,14 @@ __kernel void search10(__global hash_t* hashes) @@ -1074,14 +816,14 @@ __kernel void search10(__global hash_t* hashes)
W61 = Vb61;
W70 = Vb70;
W71 = Vb71;
W80 = hash.h8[0];
W81 = hash.h8[1];
W90 = hash.h8[2];
W91 = hash.h8[3];
WA0 = hash.h8[4];
WA1 = hash.h8[5];
WB0 = hash.h8[6];
WB1 = hash.h8[7];
W80 = hash->h8[0];
W81 = hash->h8[1];
W90 = hash->h8[2];
W91 = hash->h8[3];
WA0 = hash->h8[4];
WA1 = hash->h8[5];
WB0 = hash->h8[6];
WB1 = hash->h8[7];
WC0 = 0x80;
WC1 = 0;
WD0 = 0;
@ -1091,18 +833,27 @@ __kernel void search10(__global hash_t* hashes) @@ -1091,18 +833,27 @@ __kernel void search10(__global hash_t* hashes)
WF0 = 0x200;
WF1 = 0;
for (unsigned u = 0; u < 10; u ++)
for (unsigned u = 0; u < 10; u ++) {
BIG_ROUND;
}
hashp->h8[0] = hash.h8[0] ^ Vb00 ^ W00 ^ W80;
hashp->h8[1] = hash.h8[1] ^ Vb01 ^ W01 ^ W81;
hashp->h8[2] = hash.h8[2] ^ Vb10 ^ W10 ^ W90;
hashp->h8[3] = hash.h8[3] ^ Vb11 ^ W11 ^ W91;
hashp->h8[4] = hash.h8[4] ^ Vb20 ^ W20 ^ WA0;
hashp->h8[5] = hash.h8[5] ^ Vb21 ^ W21 ^ WA1;
hashp->h8[6] = hash.h8[6] ^ Vb30 ^ W30 ^ WB0;
hashp->h8[7] = hash.h8[7] ^ Vb31 ^ W31 ^ WB1;
hash->h8[0] = Vb00 ^ hash->h8[0] ^ W00 ^ W80;
hash->h8[1] = Vb01 ^ hash->h8[1] ^ W01 ^ W81;
hash->h8[2] = Vb10 ^ hash->h8[2] ^ W10 ^ W90;
hash->h8[3] = Vb11 ^ hash->h8[3] ^ W11 ^ W91;
hash->h8[4] = Vb20 ^ hash->h8[4] ^ W20 ^ WA0;
hash->h8[5] = Vb21 ^ hash->h8[5] ^ W21 ^ WA1;
hash->h8[6] = Vb30 ^ hash->h8[6] ^ W30 ^ WB0;
hash->h8[7] = Vb31 ^ hash->h8[7] ^ W31 ^ WB1;
barrier(CLK_GLOBAL_MEM_FENCE);
}
@ -1110,9 +861,19 @@ __attribute__((reqd_work_group_size(WORKSIZE, 1, 1))) @@ -1110,9 +861,19 @@ __attribute__((reqd_work_group_size(WORKSIZE, 1, 1)))
__kernel void search11(__global hash_t* hashes)
{
uint gid = get_global_id(0);
uint offset = get_global_offset(0);
__global hash_t *hash = &(hashes[gid-offset]);
__global hash_t *hash = &(hashes[gid-get_global_offset(0)]);
__local sph_u32 T512_L[1024];
__constant const sph_u32 *T512_C = &T512[0][0];
int init = get_local_id(0);
int step = get_local_size(0);
for (int i = init; i < 1024; i += step)
{
T512_L[i] = T512_C[i];
}
barrier(CLK_LOCAL_MEM_FENCE);
{
sph_u32 c0 = HAMSI_IV512[0], c1 = HAMSI_IV512[1], c2 = HAMSI_IV512[2], c3 = HAMSI_IV512[3];
sph_u32 c4 = HAMSI_IV512[4], c5 = HAMSI_IV512[5], c6 = HAMSI_IV512[6], c7 = HAMSI_IV512[7];
sph_u32 c8 = HAMSI_IV512[8], c9 = HAMSI_IV512[9], cA = HAMSI_IV512[10], cB = HAMSI_IV512[11];
@ -1121,29 +882,26 @@ __kernel void search11(__global hash_t* hashes) @@ -1121,29 +882,26 @@ __kernel void search11(__global hash_t* hashes)
sph_u32 m8, m9, mA, mB, mC, mD, mE, mF;
sph_u32 h[16] = { c0, c1, c2, c3, c4, c5, c6, c7, c8, c9, cA, cB, cC, cD, cE, cF };
#define buf(u) hash->h1[i + u]
for(int i = 0; i < 64; i += 8)
{
INPUT_BIG;
#define buf(u) hash->h1[i + u]
for(int i = 0; i < 64; i += 8) {
INPUT_BIG_LOCAL;
P_BIG;
T_BIG;
}
#undef buf
#define buf(u) (u == 0 ? 0x80 : 0)
INPUT_BIG;
#undef buf
#define buf(u) (u == 0 ? 0x80 : 0)
INPUT_BIG_LOCAL;
P_BIG;
T_BIG;
#undef buf
#define buf(u) (u == 6 ? 2 : 0)
INPUT_BIG;
#undef buf
#define buf(u) (u == 6 ? 2 : 0)
INPUT_BIG_LOCAL;
PF_BIG;
T_BIG;
for(unsigned u = 0; u < 16; u ++)
for (unsigned u = 0; u < 16; u ++)
hash->h4[u] = h[u];
}
barrier(CLK_GLOBAL_MEM_FENCE);
}
@ -1151,9 +909,20 @@ __attribute__((reqd_work_group_size(WORKSIZE, 1, 1))) @@ -1151,9 +909,20 @@ __attribute__((reqd_work_group_size(WORKSIZE, 1, 1)))
__kernel void search12(__global hash_t* hashes, __global uint* output, const ulong target)
{
uint gid = get_global_id(0);
uint offset = get_global_offset(0);
__global hash_t *hash = &(hashes[gid-offset]);
__global hash_t *hash = &(hashes[gid-get_global_offset(0)]);
// mixtab
__local sph_u32 mixtab0[256], mixtab1[256], mixtab2[256], mixtab3[256];
int init = get_local_id(0);
int step = get_local_size(0);
for (int i = init; i < 256; i += step)
{
mixtab0[i] = mixtab0_c[i];
mixtab1[i] = mixtab1_c[i];
mixtab2[i] = mixtab2_c[i];
mixtab3[i] = mixtab3_c[i];
}
barrier(CLK_LOCAL_MEM_FENCE);
// fugue
sph_u32 S00, S01, S02, S03, S04, S05, S06, S07, S08, S09;
sph_u32 S10, S11, S12, S13, S14, S15, S16, S17, S18, S19;
@ -1178,15 +947,12 @@ __kernel void search12(__global hash_t* hashes, __global uint* output, const ulo @@ -1178,15 +947,12 @@ __kernel void search12(__global hash_t* hashes, __global uint* output, const ulo
// apply round shift if necessary
int i;
for (i = 0; i < 32; i ++)
{
for (i = 0; i < 32; i ++) {
ROR3;
CMIX36(S00, S01, S02, S04, S05, S06, S18, S19, S20);
SMIX(S00, S01, S02, S03);
}
for (i = 0; i < 13; i ++)
{
for (i = 0; i < 13; i ++) {
S04 ^= S00;
S09 ^= S00;
S18 ^= S00;
@ -1236,9 +1002,10 @@ __kernel void search12(__global hash_t* hashes, __global uint* output, const ulo @@ -1236,9 +1002,10 @@ __kernel void search12(__global hash_t* hashes, __global uint* output, const ulo
bool result = (hash->h8[3] <= target);
if (result)
output[atomic_inc(output+0xFF)] = SWAP4(gid);
output[output[0xFF]++] = SWAP4(gid);
barrier(CLK_GLOBAL_MEM_FENCE);
}
#endif // X13MOD_CL

Write Preview
Loading…
Cancel
Save