Browse Source

Improve Lyra2RE2 Performance

Improved Lyra2Re2 Performance by 1 %
pull/2/head
Myrinia 7 years ago committed by Tanguy Pruvot
parent
commit
18d29914ec
  1. 349
      Algo256/cuda_blake256.cu
  2. 472
      Algo256/cuda_cubehash256.cu

349
Algo256/cuda_blake256.cu

@ -11,6 +11,23 @@ extern "C" { @@ -11,6 +11,23 @@ extern "C" {
#include <memory.h>
#ifdef __INTELLISENSE__
/* just for vstudio code colors */
__device__ uint32_t __byte_perm(uint32_t a, uint32_t b, uint32_t c);
#endif
#define UINT2(x,y) make_uint2(x,y)
__device__ __inline__ uint2 ROR8(const uint2 a)
{
uint2 result;
result.x = __byte_perm(a.y, a.x, 0x0765);
result.y = __byte_perm(a.x, a.y, 0x0765);
return result;
}
static __device__ uint64_t cuda_swab32ll(uint64_t x) {
return MAKE_ULONGLONG(cuda_swab32(_LODWORD(x)), cuda_swab32(_HIDWORD(x)));
}
@ -58,9 +75,25 @@ static const uint32_t c_u256[16] = { @@ -58,9 +75,25 @@ static const uint32_t c_u256[16] = {
0x3F84D5B5, 0xB5470917
};
__constant__ uint2 keccak_round_constants35[24] = {
{ 0x00000001ul, 0x00000000 }, { 0x00008082ul, 0x00000000 },
{ 0x0000808aul, 0x80000000 }, { 0x80008000ul, 0x80000000 },
{ 0x0000808bul, 0x00000000 }, { 0x80000001ul, 0x00000000 },
{ 0x80008081ul, 0x80000000 }, { 0x00008009ul, 0x80000000 },
{ 0x0000008aul, 0x00000000 }, { 0x00000088ul, 0x00000000 },
{ 0x80008009ul, 0x00000000 }, { 0x8000000aul, 0x00000000 },
{ 0x8000808bul, 0x00000000 }, { 0x0000008bul, 0x80000000 },
{ 0x00008089ul, 0x80000000 }, { 0x00008003ul, 0x80000000 },
{ 0x00008002ul, 0x80000000 }, { 0x00000080ul, 0x80000000 },
{ 0x0000800aul, 0x00000000 }, { 0x8000000aul, 0x80000000 },
{ 0x80008081ul, 0x80000000 }, { 0x00008080ul, 0x80000000 },
{ 0x80000001ul, 0x00000000 }, { 0x80008008ul, 0x80000000 }
};
#define GS2(a,b,c,d,x) { \
const uint32_t idx1 = sigma[r][x]; \
const uint32_t idx2 = sigma[r][x+1]; \
const uint32_t idx2 = sigma[r][(x)+1]; \
v[a] += (m[idx1] ^ u256[idx2]) + v[b]; \
v[d] = SPH_ROTL32(v[d] ^ v[a], 16); \
v[c] += v[d]; \
@ -76,7 +109,7 @@ static const uint32_t c_u256[16] = { @@ -76,7 +109,7 @@ static const uint32_t c_u256[16] = {
//#define ROTR32(x, n) (((x) >> (n)) | ((x) << (32 - (n))))
#define hostGS(a,b,c,d,x) { \
const uint32_t idx1 = c_sigma[r][x]; \
const uint32_t idx2 = c_sigma[r][x+1]; \
const uint32_t idx2 = c_sigma[r][(x)+1]; \
v[a] += (m[idx1] ^ c_u256[idx2]) + v[b]; \
v[d] = ROTR32(v[d] ^ v[a], 16); \
v[c] += v[d]; \
@ -88,6 +121,17 @@ static const uint32_t c_u256[16] = { @@ -88,6 +121,17 @@ static const uint32_t c_u256[16] = {
v[b] = ROTR32(v[b] ^ v[c], 7); \
}
#define GSPREC(a,b,c,d,x,y) { \
v[a] += (m[x] ^ u256[y]) + v[b]; \
v[d] = __byte_perm(v[d] ^ v[a],0, 0x1032); \
v[c] += v[d]; \
v[b] = SPH_ROTR32(v[b] ^ v[c], 12); \
v[a] += (m[y] ^ u256[x]) + v[b]; \
v[d] = __byte_perm(v[d] ^ v[a],0, 0x0321); \
v[c] += v[d]; \
v[b] = SPH_ROTR32(v[b] ^ v[c], 7); \
}
/* Second part (64-80) msg never change, store it */
__device__ __constant__ static const uint32_t c_Padding[16] = {
0, 0, 0, 0,
@ -189,6 +233,284 @@ static void blake256_compress2nd(uint32_t *h, const uint32_t *block, const uint3 @@ -189,6 +233,284 @@ static void blake256_compress2nd(uint32_t *h, const uint32_t *block, const uint3
}
}
static void __forceinline__ __device__ keccak_block(uint2 *s)
{
uint2 bc[5], tmpxor[5], u, v;
// uint2 s[25];
#pragma unroll 1
for (int i = 0; i < 24; i++)
{
#pragma unroll
for (uint32_t x = 0; x < 5; x++)
tmpxor[x] = s[x] ^ s[x + 5] ^ s[x + 10] ^ s[x + 15] ^ s[x + 20];
bc[0] = tmpxor[0] ^ ROL2(tmpxor[2], 1);
bc[1] = tmpxor[1] ^ ROL2(tmpxor[3], 1);
bc[2] = tmpxor[2] ^ ROL2(tmpxor[4], 1);
bc[3] = tmpxor[3] ^ ROL2(tmpxor[0], 1);
bc[4] = tmpxor[4] ^ ROL2(tmpxor[1], 1);
u = s[1] ^ bc[0];
s[0] ^= bc[4];
s[1] = ROL2(s[6] ^ bc[0], 44);
s[6] = ROL2(s[9] ^ bc[3], 20);
s[9] = ROL2(s[22] ^ bc[1], 61);
s[22] = ROL2(s[14] ^ bc[3], 39);
s[14] = ROL2(s[20] ^ bc[4], 18);
s[20] = ROL2(s[2] ^ bc[1], 62);
s[2] = ROL2(s[12] ^ bc[1], 43);
s[12] = ROL2(s[13] ^ bc[2], 25);
s[13] = ROL8(s[19] ^ bc[3]);
s[19] = ROR8(s[23] ^ bc[2]);
s[23] = ROL2(s[15] ^ bc[4], 41);
s[15] = ROL2(s[4] ^ bc[3], 27);
s[4] = ROL2(s[24] ^ bc[3], 14);
s[24] = ROL2(s[21] ^ bc[0], 2);
s[21] = ROL2(s[8] ^ bc[2], 55);
s[8] = ROL2(s[16] ^ bc[0], 45);
s[16] = ROL2(s[5] ^ bc[4], 36);
s[5] = ROL2(s[3] ^ bc[2], 28);
s[3] = ROL2(s[18] ^ bc[2], 21);
s[18] = ROL2(s[17] ^ bc[1], 15);
s[17] = ROL2(s[11] ^ bc[0], 10);
s[11] = ROL2(s[7] ^ bc[1], 6);
s[7] = ROL2(s[10] ^ bc[4], 3);
s[10] = ROL2(u, 1);
u = s[0]; v = s[1]; s[0] ^= (~v) & s[2]; s[1] ^= (~s[2]) & s[3]; s[2] ^= (~s[3]) & s[4]; s[3] ^= (~s[4]) & u; s[4] ^= (~u) & v;
u = s[5]; v = s[6]; s[5] ^= (~v) & s[7]; s[6] ^= (~s[7]) & s[8]; s[7] ^= (~s[8]) & s[9]; s[8] ^= (~s[9]) & u; s[9] ^= (~u) & v;
u = s[10]; v = s[11]; s[10] ^= (~v) & s[12]; s[11] ^= (~s[12]) & s[13]; s[12] ^= (~s[13]) & s[14]; s[13] ^= (~s[14]) & u; s[14] ^= (~u) & v;
u = s[15]; v = s[16]; s[15] ^= (~v) & s[17]; s[16] ^= (~s[17]) & s[18]; s[17] ^= (~s[18]) & s[19]; s[18] ^= (~s[19]) & u; s[19] ^= (~u) & v;
u = s[20]; v = s[21]; s[20] ^= (~v) & s[22]; s[21] ^= (~s[22]) & s[23]; s[22] ^= (~s[23]) & s[24]; s[23] ^= (~s[24]) & u; s[24] ^= (~u) & v;
s[0] ^= keccak_round_constants35[i];
}
}
//__launch_bounds__(256)
__global__
void blakeKeccak256_gpu_hash_80(const uint32_t threads, const uint32_t startNonce, uint32_t * Hash)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
const uint32_t nonce = startNonce + thread;
uint32_t h[8];
// uint32_t input[4];
const uint32_t T0 = 640;
#pragma unroll 8
for (int i = 0; i<8; i++) { h[i] = cpu_h[i]; }
uint32_t v[16];
const uint32_t c_Padding[12] = {
0x80000000, 0, 0, 0,
0, 0, 0, 0,
0, 1, 0, 640
};
const uint32_t u256[16] =
{
0x243F6A88, 0x85A308D3,
0x13198A2E, 0x03707344,
0xA4093822, 0x299F31D0,
0x082EFA98, 0xEC4E6C89,
0x452821E6, 0x38D01377,
0xBE5466CF, 0x34E90C6C,
0xC0AC29B7, 0xC97C50DD,
0x3F84D5B5, 0xB5470917
};
uint32_t m[16] =
{
c_data[0], c_data[1], c_data[2], nonce,
c_Padding[0], c_Padding[1], c_Padding[2], c_Padding[3],
c_Padding[4], c_Padding[5], c_Padding[6], c_Padding[7],
c_Padding[8], c_Padding[9], c_Padding[10], c_Padding[11]
};
#pragma unroll 8
for (int i = 0; i < 8; i++)
v[i] = h[i];
v[8] = u256[0];
v[9] = u256[1];
v[10] = u256[2];
v[11] = u256[3];
v[12] = u256[4] ^ T0;
v[13] = u256[5] ^ T0;
v[14] = u256[6];
v[15] = u256[7];
// { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
GSPREC(0, 4, 0x8, 0xC, 0, 1);
GSPREC(1, 5, 0x9, 0xD, 2, 3);
GSPREC(2, 6, 0xA, 0xE, 4, 5);
GSPREC(3, 7, 0xB, 0xF, 6, 7);
GSPREC(0, 5, 0xA, 0xF, 8, 9);
GSPREC(1, 6, 0xB, 0xC, 10, 11);
GSPREC(2, 7, 0x8, 0xD, 12, 13);
GSPREC(3, 4, 0x9, 0xE, 14, 15);
// { 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 },
GSPREC(0, 4, 0x8, 0xC, 14, 10);
GSPREC(1, 5, 0x9, 0xD, 4, 8);
GSPREC(2, 6, 0xA, 0xE, 9, 15);
GSPREC(3, 7, 0xB, 0xF, 13, 6);
GSPREC(0, 5, 0xA, 0xF, 1, 12);
GSPREC(1, 6, 0xB, 0xC, 0, 2);
GSPREC(2, 7, 0x8, 0xD, 11, 7);
GSPREC(3, 4, 0x9, 0xE, 5, 3);
// { 11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4 },
GSPREC(0, 4, 0x8, 0xC, 11, 8);
GSPREC(1, 5, 0x9, 0xD, 12, 0);
GSPREC(2, 6, 0xA, 0xE, 5, 2);
GSPREC(3, 7, 0xB, 0xF, 15, 13);
GSPREC(0, 5, 0xA, 0xF, 10, 14);
GSPREC(1, 6, 0xB, 0xC, 3, 6);
GSPREC(2, 7, 0x8, 0xD, 7, 1);
GSPREC(3, 4, 0x9, 0xE, 9, 4);
// { 7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8 },
GSPREC(0, 4, 0x8, 0xC, 7, 9);
GSPREC(1, 5, 0x9, 0xD, 3, 1);
GSPREC(2, 6, 0xA, 0xE, 13, 12);
GSPREC(3, 7, 0xB, 0xF, 11, 14);
GSPREC(0, 5, 0xA, 0xF, 2, 6);
GSPREC(1, 6, 0xB, 0xC, 5, 10);
GSPREC(2, 7, 0x8, 0xD, 4, 0);
GSPREC(3, 4, 0x9, 0xE, 15, 8);
// { 9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13 },
GSPREC(0, 4, 0x8, 0xC, 9, 0);
GSPREC(1, 5, 0x9, 0xD, 5, 7);
GSPREC(2, 6, 0xA, 0xE, 2, 4);
GSPREC(3, 7, 0xB, 0xF, 10, 15);
GSPREC(0, 5, 0xA, 0xF, 14, 1);
GSPREC(1, 6, 0xB, 0xC, 11, 12);
GSPREC(2, 7, 0x8, 0xD, 6, 8);
GSPREC(3, 4, 0x9, 0xE, 3, 13);
// { 2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9 },
GSPREC(0, 4, 0x8, 0xC, 2, 12);
GSPREC(1, 5, 0x9, 0xD, 6, 10);
GSPREC(2, 6, 0xA, 0xE, 0, 11);
GSPREC(3, 7, 0xB, 0xF, 8, 3);
GSPREC(0, 5, 0xA, 0xF, 4, 13);
GSPREC(1, 6, 0xB, 0xC, 7, 5);
GSPREC(2, 7, 0x8, 0xD, 15, 14);
GSPREC(3, 4, 0x9, 0xE, 1, 9);
// { 12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11 },
GSPREC(0, 4, 0x8, 0xC, 12, 5);
GSPREC(1, 5, 0x9, 0xD, 1, 15);
GSPREC(2, 6, 0xA, 0xE, 14, 13);
GSPREC(3, 7, 0xB, 0xF, 4, 10);
GSPREC(0, 5, 0xA, 0xF, 0, 7);
GSPREC(1, 6, 0xB, 0xC, 6, 3);
GSPREC(2, 7, 0x8, 0xD, 9, 2);
GSPREC(3, 4, 0x9, 0xE, 8, 11);
// { 13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10 },
GSPREC(0, 4, 0x8, 0xC, 13, 11);
GSPREC(1, 5, 0x9, 0xD, 7, 14);
GSPREC(2, 6, 0xA, 0xE, 12, 1);
GSPREC(3, 7, 0xB, 0xF, 3, 9);
GSPREC(0, 5, 0xA, 0xF, 5, 0);
GSPREC(1, 6, 0xB, 0xC, 15, 4);
GSPREC(2, 7, 0x8, 0xD, 8, 6);
GSPREC(3, 4, 0x9, 0xE, 2, 10);
// { 6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5 },
GSPREC(0, 4, 0x8, 0xC, 6, 15);
GSPREC(1, 5, 0x9, 0xD, 14, 9);
GSPREC(2, 6, 0xA, 0xE, 11, 3);
GSPREC(3, 7, 0xB, 0xF, 0, 8);
GSPREC(0, 5, 0xA, 0xF, 12, 2);
GSPREC(1, 6, 0xB, 0xC, 13, 7);
GSPREC(2, 7, 0x8, 0xD, 1, 4);
GSPREC(3, 4, 0x9, 0xE, 10, 5);
// { 10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13, 0 },
GSPREC(0, 4, 0x8, 0xC, 10, 2);
GSPREC(1, 5, 0x9, 0xD, 8, 4);
GSPREC(2, 6, 0xA, 0xE, 7, 6);
GSPREC(3, 7, 0xB, 0xF, 1, 5);
GSPREC(0, 5, 0xA, 0xF, 15, 11);
GSPREC(1, 6, 0xB, 0xC, 9, 14);
GSPREC(2, 7, 0x8, 0xD, 3, 12);
GSPREC(3, 4, 0x9, 0xE, 13, 0);
// { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
GSPREC(0, 4, 0x8, 0xC, 0, 1);
GSPREC(1, 5, 0x9, 0xD, 2, 3);
GSPREC(2, 6, 0xA, 0xE, 4, 5);
GSPREC(3, 7, 0xB, 0xF, 6, 7);
GSPREC(0, 5, 0xA, 0xF, 8, 9);
GSPREC(1, 6, 0xB, 0xC, 10, 11);
GSPREC(2, 7, 0x8, 0xD, 12, 13);
GSPREC(3, 4, 0x9, 0xE, 14, 15);
// { 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 },
GSPREC(0, 4, 0x8, 0xC, 14, 10);
GSPREC(1, 5, 0x9, 0xD, 4, 8);
GSPREC(2, 6, 0xA, 0xE, 9, 15);
GSPREC(3, 7, 0xB, 0xF, 13, 6);
GSPREC(0, 5, 0xA, 0xF, 1, 12);
GSPREC(1, 6, 0xB, 0xC, 0, 2);
GSPREC(2, 7, 0x8, 0xD, 11, 7);
GSPREC(3, 4, 0x9, 0xE, 5, 3);
// { 11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4 },
GSPREC(0, 4, 0x8, 0xC, 11, 8);
GSPREC(1, 5, 0x9, 0xD, 12, 0);
GSPREC(2, 6, 0xA, 0xE, 5, 2);
GSPREC(3, 7, 0xB, 0xF, 15, 13);
GSPREC(0, 5, 0xA, 0xF, 10, 14);
GSPREC(1, 6, 0xB, 0xC, 3, 6);
GSPREC(2, 7, 0x8, 0xD, 7, 1);
GSPREC(3, 4, 0x9, 0xE, 9, 4);
// { 7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8 },
GSPREC(0, 4, 0x8, 0xC, 7, 9);
GSPREC(1, 5, 0x9, 0xD, 3, 1);
GSPREC(2, 6, 0xA, 0xE, 13, 12);
GSPREC(3, 7, 0xB, 0xF, 11, 14);
GSPREC(0, 5, 0xA, 0xF, 2, 6);
GSPREC(1, 6, 0xB, 0xC, 5, 10);
GSPREC(2, 7, 0x8, 0xD, 4, 0);
GSPREC(3, 4, 0x9, 0xE, 15, 8);
h[0] = cuda_swab32(h[0] ^ v[0] ^ v[8]);
h[1] = cuda_swab32(h[1] ^ v[1] ^ v[9]);
h[2] = cuda_swab32(h[2] ^ v[2] ^ v[10]);
h[3] = cuda_swab32(h[3] ^ v[3] ^ v[11]);
h[4] = cuda_swab32(h[4] ^ v[4] ^ v[12]);
h[5] = cuda_swab32(h[5] ^ v[5] ^ v[13]);
h[6] = cuda_swab32(h[6] ^ v[6] ^ v[14]);
h[7] = cuda_swab32(h[7] ^ v[7] ^ v[15]);
uint2 keccak_gpu_state[25] = { 0 };
keccak_gpu_state[0].x = h[0];
keccak_gpu_state[0].y = h[1];
keccak_gpu_state[1].x = h[2];
keccak_gpu_state[1].y = h[3];
keccak_gpu_state[2].x = h[4];
keccak_gpu_state[2].y = h[5];
keccak_gpu_state[3].x = h[6];
keccak_gpu_state[3].y = h[7];
keccak_gpu_state[4] = UINT2(1, 0);
keccak_gpu_state[16] = UINT2(0, 0x80000000);
keccak_block(keccak_gpu_state);
uint64_t *outputHash = (uint64_t *)Hash;
#pragma unroll 4
for (int i = 0; i<4; i++)
outputHash[i*threads + thread] = devectorize(keccak_gpu_state[i]);
}
}
__global__ __launch_bounds__(256, 3)
void blake256_gpu_hash_80(const uint32_t threads, const uint32_t startNonce, uint64_t * Hash)
{
@ -242,7 +564,28 @@ void blake256_cpu_setBlock_80(uint32_t *pdata) @@ -242,7 +564,28 @@ void blake256_cpu_setBlock_80(uint32_t *pdata)
__host__
void blake256_cpu_init(int thr_id, uint32_t threads)
{
cuda_get_arch(thr_id);
cudaMemcpyToSymbol(u256, c_u256, sizeof(c_u256), 0, cudaMemcpyHostToDevice);
cudaMemcpyToSymbol(sigma, c_sigma, sizeof(c_sigma), 0, cudaMemcpyHostToDevice);
}
__host__
void blakeKeccak256_cpu_hash_80(const int thr_id, const uint32_t threads, const uint32_t startNonce, uint64_t *Hash, int order)
{
const uint32_t threadsperblock = 256;
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
blakeKeccak256_gpu_hash_80 << <grid, block >> > (threads, startNonce, (uint32_t *)Hash);
}
__host__
void blakeKeccak256_cpu_hash_80(const int thr_id, const uint32_t threads, const uint32_t startNonce, uint64_t *Hash, int order, cudaStream_t stream)
{
const uint32_t threadsperblock = 256;
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
blakeKeccak256_gpu_hash_80 << <grid, block, 0, stream >> > (threads, startNonce, (uint32_t *)Hash);
}

472
Algo256/cuda_cubehash256.cu

@ -3,179 +3,247 @@ @@ -3,179 +3,247 @@
#define CUBEHASH_ROUNDS 16 /* this is r for CubeHashr/b */
#define CUBEHASH_BLOCKBYTES 32 /* this is b for CubeHashr/b */
#ifdef __INTELLISENSE__
/* just for vstudio code colors */
#define __CUDA_ARCH__ 520
#endif
#if __CUDA_ARCH__ < 350
#define LROT(x,bits) ((x << bits) | (x >> (32 - bits)))
#else
#define LROT(x, bits) __funnelshift_l(x, x, bits)
#endif
#if __CUDA_ARCH__ < 500
#define TPB 576
#else
#define TPB 1024
#endif
#define TPB35 576
#define TPB50 1024
#define ROTATEUPWARDS7(a) LROT(a,7)
#define ROTATEUPWARDS11(a) LROT(a,11)
//#define SWAP(a,b) { uint32_t u = a; a = b; b = u; }
#define SWAP(a,b) { a ^= b; b ^= a; a ^= b; }
__device__ __forceinline__ void rrounds(uint32_t x[2][2][2][2][2])
{
int r;
int j;
int k;
int l;
int m;
#pragma unroll 2
for (r = 0; r < CUBEHASH_ROUNDS; ++r) {
/* "add x_0jklm into x_1jklmn modulo 2^32" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
x[1][j][k][l][m] += x[0][j][k][l][m];
uint32_t x0[2][2][2][2];
uint32_t x1[2][2][2][2];
for (r = 0; r < CUBEHASH_ROUNDS; r += 2) {
/* "rotate x_0jklm upwards by 7 bits" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
x[0][j][k][l][m] = ROTATEUPWARDS7(x[0][j][k][l][m]);
/* "swap x_00klm with x_01klm" */
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
SWAP(x[0][0][k][l][m], x[0][1][k][l][m])
/* "xor x_1jklm into x_0jklm" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
x[0][j][k][l][m] ^= x[1][j][k][l][m];
/* "swap x_1jk0m with x_1jk1m" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (m = 0; m < 2; ++m)
SWAP(x[1][j][k][0][m], x[1][j][k][1][m])
x0[0][0][0][0] = ROTATEUPWARDS7(x[0][0][0][0][0]);
x0[0][0][0][1] = ROTATEUPWARDS7(x[0][0][0][0][1]);
x0[0][0][1][0] = ROTATEUPWARDS7(x[0][0][0][1][0]);
x0[0][0][1][1] = ROTATEUPWARDS7(x[0][0][0][1][1]);
x0[0][1][0][0] = ROTATEUPWARDS7(x[0][0][1][0][0]);
x0[0][1][0][1] = ROTATEUPWARDS7(x[0][0][1][0][1]);
x0[0][1][1][0] = ROTATEUPWARDS7(x[0][0][1][1][0]);
x0[0][1][1][1] = ROTATEUPWARDS7(x[0][0][1][1][1]);
x0[1][0][0][0] = ROTATEUPWARDS7(x[0][1][0][0][0]);
x0[1][0][0][1] = ROTATEUPWARDS7(x[0][1][0][0][1]);
x0[1][0][1][0] = ROTATEUPWARDS7(x[0][1][0][1][0]);
x0[1][0][1][1] = ROTATEUPWARDS7(x[0][1][0][1][1]);
x0[1][1][0][0] = ROTATEUPWARDS7(x[0][1][1][0][0]);
x0[1][1][0][1] = ROTATEUPWARDS7(x[0][1][1][0][1]);
x0[1][1][1][0] = ROTATEUPWARDS7(x[0][1][1][1][0]);
x0[1][1][1][1] = ROTATEUPWARDS7(x[0][1][1][1][1]);
/* "add x_0jklm into x_1jklm modulo 2^32" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
x[1][j][k][l][m] += x[0][j][k][l][m];
x1[0][0][0][0] = x[1][0][0][0][0] + x[0][0][0][0][0];
x1[0][0][0][1] = x[1][0][0][0][1] + x[0][0][0][0][1];
x1[0][0][1][0] = x[1][0][0][1][0] + x[0][0][0][1][0];
x1[0][0][1][1] = x[1][0][0][1][1] + x[0][0][0][1][1];
x1[0][1][0][0] = x[1][0][1][0][0] + x[0][0][1][0][0];
x1[0][1][0][1] = x[1][0][1][0][1] + x[0][0][1][0][1];
x1[0][1][1][0] = x[1][0][1][1][0] + x[0][0][1][1][0];
x1[0][1][1][1] = x[1][0][1][1][1] + x[0][0][1][1][1];
x1[1][0][0][0] = x[1][1][0][0][0] + x[0][1][0][0][0];
x1[1][0][0][1] = x[1][1][0][0][1] + x[0][1][0][0][1];
x1[1][0][1][0] = x[1][1][0][1][0] + x[0][1][0][1][0];
x1[1][0][1][1] = x[1][1][0][1][1] + x[0][1][0][1][1];
x1[1][1][0][0] = x[1][1][1][0][0] + x[0][1][1][0][0];
x1[1][1][0][1] = x[1][1][1][0][1] + x[0][1][1][0][1];
x1[1][1][1][0] = x[1][1][1][1][0] + x[0][1][1][1][0];
x1[1][1][1][1] = x[1][1][1][1][1] + x[0][1][1][1][1];
/* "xor x_1~jklm into x_0jklm" */
x[0][0][0][0][0] = x0[0][0][0][0] ^ x1[1][0][0][0];
x[0][0][0][0][1] = x0[0][0][0][1] ^ x1[1][0][0][1];
x[0][0][0][1][0] = x0[0][0][1][0] ^ x1[1][0][1][0];
x[0][0][0][1][1] = x0[0][0][1][1] ^ x1[1][0][1][1];
x[0][0][1][0][0] = x0[0][1][0][0] ^ x1[1][1][0][0];
x[0][0][1][0][1] = x0[0][1][0][1] ^ x1[1][1][0][1];
x[0][0][1][1][0] = x0[0][1][1][0] ^ x1[1][1][1][0];
x[0][0][1][1][1] = x0[0][1][1][1] ^ x1[1][1][1][1];
x[0][1][0][0][0] = x0[1][0][0][0] ^ x1[0][0][0][0];
x[0][1][0][0][1] = x0[1][0][0][1] ^ x1[0][0][0][1];
x[0][1][0][1][0] = x0[1][0][1][0] ^ x1[0][0][1][0];
x[0][1][0][1][1] = x0[1][0][1][1] ^ x1[0][0][1][1];
x[0][1][1][0][0] = x0[1][1][0][0] ^ x1[0][1][0][0];
x[0][1][1][0][1] = x0[1][1][0][1] ^ x1[0][1][0][1];
x[0][1][1][1][0] = x0[1][1][1][0] ^ x1[0][1][1][0];
x[0][1][1][1][1] = x0[1][1][1][1] ^ x1[0][1][1][1];
/* "rotate x_0jklm upwards by 11 bits" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
x[0][j][k][l][m] = ROTATEUPWARDS11(x[0][j][k][l][m]);
/* "swap x_0j0lm with x_0j1lm" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
SWAP(x[0][j][0][l][m], x[0][j][1][l][m])
/* "xor x_1jklm into x_0jklm" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
x[0][j][k][l][m] ^= x[1][j][k][l][m];
/* "swap x_1jkl0 with x_1jkl1" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
SWAP(x[1][j][k][l][0], x[1][j][k][l][1])
}
}
__device__ __forceinline__ void block_tox(const uint32_t *in, uint32_t x[2][2][2][2][2])
{
x[0][0][0][0][0] ^= in[0];
x[0][0][0][0][1] ^= in[1];
x[0][0][0][1][0] ^= in[2];
x[0][0][0][1][1] ^= in[3];
x[0][0][1][0][0] ^= in[4];
x[0][0][1][0][1] ^= in[5];
x[0][0][1][1][0] ^= in[6];
x[0][0][1][1][1] ^= in[7];
}
__device__ __forceinline__ void hash_fromx(uint32_t *out, uint32_t x[2][2][2][2][2])
{
out[0] = x[0][0][0][0][0];
out[1] = x[0][0][0][0][1];
out[2] = x[0][0][0][1][0];
out[3] = x[0][0][0][1][1];
out[4] = x[0][0][1][0][0];
out[5] = x[0][0][1][0][1];
out[6] = x[0][0][1][1][0];
out[7] = x[0][0][1][1][1];
x0[0][0][0][0] = ROTATEUPWARDS11(x[0][0][0][0][0]);
x0[0][0][0][1] = ROTATEUPWARDS11(x[0][0][0][0][1]);
x0[0][0][1][0] = ROTATEUPWARDS11(x[0][0][0][1][0]);
x0[0][0][1][1] = ROTATEUPWARDS11(x[0][0][0][1][1]);
x0[0][1][0][0] = ROTATEUPWARDS11(x[0][0][1][0][0]);
x0[0][1][0][1] = ROTATEUPWARDS11(x[0][0][1][0][1]);
x0[0][1][1][0] = ROTATEUPWARDS11(x[0][0][1][1][0]);
x0[0][1][1][1] = ROTATEUPWARDS11(x[0][0][1][1][1]);
x0[1][0][0][0] = ROTATEUPWARDS11(x[0][1][0][0][0]);
x0[1][0][0][1] = ROTATEUPWARDS11(x[0][1][0][0][1]);
x0[1][0][1][0] = ROTATEUPWARDS11(x[0][1][0][1][0]);
x0[1][0][1][1] = ROTATEUPWARDS11(x[0][1][0][1][1]);
x0[1][1][0][0] = ROTATEUPWARDS11(x[0][1][1][0][0]);
x0[1][1][0][1] = ROTATEUPWARDS11(x[0][1][1][0][1]);
x0[1][1][1][0] = ROTATEUPWARDS11(x[0][1][1][1][0]);
x0[1][1][1][1] = ROTATEUPWARDS11(x[0][1][1][1][1]);
/* "add x_0jklm into x_1~jk~lm modulo 2^32" */
x[1][1][0][1][0] = x1[1][0][1][0] + x[0][0][0][0][0];
x[1][1][0][1][1] = x1[1][0][1][1] + x[0][0][0][0][1];
x[1][1][0][0][0] = x1[1][0][0][0] + x[0][0][0][1][0];
x[1][1][0][0][1] = x1[1][0][0][1] + x[0][0][0][1][1];
x[1][1][1][1][0] = x1[1][1][1][0] + x[0][0][1][0][0];
x[1][1][1][1][1] = x1[1][1][1][1] + x[0][0][1][0][1];
x[1][1][1][0][0] = x1[1][1][0][0] + x[0][0][1][1][0];
x[1][1][1][0][1] = x1[1][1][0][1] + x[0][0][1][1][1];
x[1][0][0][1][0] = x1[0][0][1][0] + x[0][1][0][0][0];
x[1][0][0][1][1] = x1[0][0][1][1] + x[0][1][0][0][1];
x[1][0][0][0][0] = x1[0][0][0][0] + x[0][1][0][1][0];
x[1][0][0][0][1] = x1[0][0][0][1] + x[0][1][0][1][1];
x[1][0][1][1][0] = x1[0][1][1][0] + x[0][1][1][0][0];
x[1][0][1][1][1] = x1[0][1][1][1] + x[0][1][1][0][1];
x[1][0][1][0][0] = x1[0][1][0][0] + x[0][1][1][1][0];
x[1][0][1][0][1] = x1[0][1][0][1] + x[0][1][1][1][1];
/* "xor x_1~j~k~lm into x_0jklm" */
x[0][0][0][0][0] = x0[0][0][0][0] ^ x[1][1][1][1][0];
x[0][0][0][0][1] = x0[0][0][0][1] ^ x[1][1][1][1][1];
x[0][0][0][1][0] = x0[0][0][1][0] ^ x[1][1][1][0][0];
x[0][0][0][1][1] = x0[0][0][1][1] ^ x[1][1][1][0][1];
x[0][0][1][0][0] = x0[0][1][0][0] ^ x[1][1][0][1][0];
x[0][0][1][0][1] = x0[0][1][0][1] ^ x[1][1][0][1][1];
x[0][0][1][1][0] = x0[0][1][1][0] ^ x[1][1][0][0][0];
x[0][0][1][1][1] = x0[0][1][1][1] ^ x[1][1][0][0][1];
x[0][1][0][0][0] = x0[1][0][0][0] ^ x[1][0][1][1][0];
x[0][1][0][0][1] = x0[1][0][0][1] ^ x[1][0][1][1][1];
x[0][1][0][1][0] = x0[1][0][1][0] ^ x[1][0][1][0][0];
x[0][1][0][1][1] = x0[1][0][1][1] ^ x[1][0][1][0][1];
x[0][1][1][0][0] = x0[1][1][0][0] ^ x[1][0][0][1][0];
x[0][1][1][0][1] = x0[1][1][0][1] ^ x[1][0][0][1][1];
x[0][1][1][1][0] = x0[1][1][1][0] ^ x[1][0][0][0][0];
x[0][1][1][1][1] = x0[1][1][1][1] ^ x[1][0][0][0][1];
}
/* "rotate x_0jklm upwards by 7 bits" */
x0[0][0][0][0] = ROTATEUPWARDS7(x[0][0][0][0][0]);
x0[0][0][0][1] = ROTATEUPWARDS7(x[0][0][0][0][1]);
x0[0][0][1][0] = ROTATEUPWARDS7(x[0][0][0][1][0]);
x0[0][0][1][1] = ROTATEUPWARDS7(x[0][0][0][1][1]);
x0[0][1][0][0] = ROTATEUPWARDS7(x[0][0][1][0][0]);
x0[0][1][0][1] = ROTATEUPWARDS7(x[0][0][1][0][1]);
x0[0][1][1][0] = ROTATEUPWARDS7(x[0][0][1][1][0]);
x0[0][1][1][1] = ROTATEUPWARDS7(x[0][0][1][1][1]);
x0[1][0][0][0] = ROTATEUPWARDS7(x[0][1][0][0][0]);
x0[1][0][0][1] = ROTATEUPWARDS7(x[0][1][0][0][1]);
x0[1][0][1][0] = ROTATEUPWARDS7(x[0][1][0][1][0]);
x0[1][0][1][1] = ROTATEUPWARDS7(x[0][1][0][1][1]);
x0[1][1][0][0] = ROTATEUPWARDS7(x[0][1][1][0][0]);
x0[1][1][0][1] = ROTATEUPWARDS7(x[0][1][1][0][1]);
x0[1][1][1][0] = ROTATEUPWARDS7(x[0][1][1][1][0]);
x0[1][1][1][1] = ROTATEUPWARDS7(x[0][1][1][1][1]);
/* "add x_0jklm into x_1~j~k~l~m modulo 2^32" */
x1[1][1][1][1] = x[1][1][1][1][1] + x[0][0][0][0][0];
x1[1][1][1][0] = x[1][1][1][1][0] + x[0][0][0][0][1];
x1[1][1][0][1] = x[1][1][1][0][1] + x[0][0][0][1][0];
x1[1][1][0][0] = x[1][1][1][0][0] + x[0][0][0][1][1];
x1[1][0][1][1] = x[1][1][0][1][1] + x[0][0][1][0][0];
x1[1][0][1][0] = x[1][1][0][1][0] + x[0][0][1][0][1];
x1[1][0][0][1] = x[1][1][0][0][1] + x[0][0][1][1][0];
x1[1][0][0][0] = x[1][1][0][0][0] + x[0][0][1][1][1];
x1[0][1][1][1] = x[1][0][1][1][1] + x[0][1][0][0][0];
x1[0][1][1][0] = x[1][0][1][1][0] + x[0][1][0][0][1];
x1[0][1][0][1] = x[1][0][1][0][1] + x[0][1][0][1][0];
x1[0][1][0][0] = x[1][0][1][0][0] + x[0][1][0][1][1];
x1[0][0][1][1] = x[1][0][0][1][1] + x[0][1][1][0][0];
x1[0][0][1][0] = x[1][0][0][1][0] + x[0][1][1][0][1];
x1[0][0][0][1] = x[1][0][0][0][1] + x[0][1][1][1][0];
x1[0][0][0][0] = x[1][0][0][0][0] + x[0][1][1][1][1];
/* "xor x_1j~k~l~m into x_0jklm" */
x[0][0][0][0][0] = x0[0][0][0][0] ^ x1[0][1][1][1];
x[0][0][0][0][1] = x0[0][0][0][1] ^ x1[0][1][1][0];
x[0][0][0][1][0] = x0[0][0][1][0] ^ x1[0][1][0][1];
x[0][0][0][1][1] = x0[0][0][1][1] ^ x1[0][1][0][0];
x[0][0][1][0][0] = x0[0][1][0][0] ^ x1[0][0][1][1];
x[0][0][1][0][1] = x0[0][1][0][1] ^ x1[0][0][1][0];
x[0][0][1][1][0] = x0[0][1][1][0] ^ x1[0][0][0][1];
x[0][0][1][1][1] = x0[0][1][1][1] ^ x1[0][0][0][0];
x[0][1][0][0][0] = x0[1][0][0][0] ^ x1[1][1][1][1];
x[0][1][0][0][1] = x0[1][0][0][1] ^ x1[1][1][1][0];
x[0][1][0][1][0] = x0[1][0][1][0] ^ x1[1][1][0][1];
x[0][1][0][1][1] = x0[1][0][1][1] ^ x1[1][1][0][0];
x[0][1][1][0][0] = x0[1][1][0][0] ^ x1[1][0][1][1];
x[0][1][1][0][1] = x0[1][1][0][1] ^ x1[1][0][1][0];
x[0][1][1][1][0] = x0[1][1][1][0] ^ x1[1][0][0][1];
x[0][1][1][1][1] = x0[1][1][1][1] ^ x1[1][0][0][0];
__device__ __forceinline__
void Update32(uint32_t x[2][2][2][2][2], const uint32_t *data)
{
/* "xor the block into the first b bytes of the state" */
/* "and then transform the state invertibly through r identical rounds" */
block_tox(data, x);
rrounds(x);
/* "rotate x_0jklm upwards by 11 bits" */
x0[0][0][0][0] = ROTATEUPWARDS11(x[0][0][0][0][0]);
x0[0][0][0][1] = ROTATEUPWARDS11(x[0][0][0][0][1]);
x0[0][0][1][0] = ROTATEUPWARDS11(x[0][0][0][1][0]);
x0[0][0][1][1] = ROTATEUPWARDS11(x[0][0][0][1][1]);
x0[0][1][0][0] = ROTATEUPWARDS11(x[0][0][1][0][0]);
x0[0][1][0][1] = ROTATEUPWARDS11(x[0][0][1][0][1]);
x0[0][1][1][0] = ROTATEUPWARDS11(x[0][0][1][1][0]);
x0[0][1][1][1] = ROTATEUPWARDS11(x[0][0][1][1][1]);
x0[1][0][0][0] = ROTATEUPWARDS11(x[0][1][0][0][0]);
x0[1][0][0][1] = ROTATEUPWARDS11(x[0][1][0][0][1]);
x0[1][0][1][0] = ROTATEUPWARDS11(x[0][1][0][1][0]);
x0[1][0][1][1] = ROTATEUPWARDS11(x[0][1][0][1][1]);
x0[1][1][0][0] = ROTATEUPWARDS11(x[0][1][1][0][0]);
x0[1][1][0][1] = ROTATEUPWARDS11(x[0][1][1][0][1]);
x0[1][1][1][0] = ROTATEUPWARDS11(x[0][1][1][1][0]);
x0[1][1][1][1] = ROTATEUPWARDS11(x[0][1][1][1][1]);
/* "add x_0jklm into x_1j~kl~m modulo 2^32" */
x[1][0][1][0][1] = x1[0][1][0][1] + x[0][0][0][0][0];
x[1][0][1][0][0] = x1[0][1][0][0] + x[0][0][0][0][1];
x[1][0][1][1][1] = x1[0][1][1][1] + x[0][0][0][1][0];
x[1][0][1][1][0] = x1[0][1][1][0] + x[0][0][0][1][1];
x[1][0][0][0][1] = x1[0][0][0][1] + x[0][0][1][0][0];
x[1][0][0][0][0] = x1[0][0][0][0] + x[0][0][1][0][1];
x[1][0][0][1][1] = x1[0][0][1][1] + x[0][0][1][1][0];
x[1][0][0][1][0] = x1[0][0][1][0] + x[0][0][1][1][1];
x[1][1][1][0][1] = x1[1][1][0][1] + x[0][1][0][0][0];
x[1][1][1][0][0] = x1[1][1][0][0] + x[0][1][0][0][1];
x[1][1][1][1][1] = x1[1][1][1][1] + x[0][1][0][1][0];
x[1][1][1][1][0] = x1[1][1][1][0] + x[0][1][0][1][1];
x[1][1][0][0][1] = x1[1][0][0][1] + x[0][1][1][0][0];
x[1][1][0][0][0] = x1[1][0][0][0] + x[0][1][1][0][1];
x[1][1][0][1][1] = x1[1][0][1][1] + x[0][1][1][1][0];
x[1][1][0][1][0] = x1[1][0][1][0] + x[0][1][1][1][1];
/* "xor x_1jkl~m into x_0jklm" */
x[0][0][0][0][0] = x0[0][0][0][0] ^ x[1][0][0][0][1];
x[0][0][0][0][1] = x0[0][0][0][1] ^ x[1][0][0][0][0];
x[0][0][0][1][0] = x0[0][0][1][0] ^ x[1][0][0][1][1];
x[0][0][0][1][1] = x0[0][0][1][1] ^ x[1][0][0][1][0];
x[0][0][1][0][0] = x0[0][1][0][0] ^ x[1][0][1][0][1];
x[0][0][1][0][1] = x0[0][1][0][1] ^ x[1][0][1][0][0];
x[0][0][1][1][0] = x0[0][1][1][0] ^ x[1][0][1][1][1];
x[0][0][1][1][1] = x0[0][1][1][1] ^ x[1][0][1][1][0];
x[0][1][0][0][0] = x0[1][0][0][0] ^ x[1][1][0][0][1];
x[0][1][0][0][1] = x0[1][0][0][1] ^ x[1][1][0][0][0];
x[0][1][0][1][0] = x0[1][0][1][0] ^ x[1][1][0][1][1];
x[0][1][0][1][1] = x0[1][0][1][1] ^ x[1][1][0][1][0];
x[0][1][1][0][0] = x0[1][1][0][0] ^ x[1][1][1][0][1];
x[0][1][1][0][1] = x0[1][1][0][1] ^ x[1][1][1][0][0];
x[0][1][1][1][0] = x0[1][1][1][0] ^ x[1][1][1][1][1];
x[0][1][1][1][1] = x0[1][1][1][1] ^ x[1][1][1][1][0];
}
__device__ __forceinline__
void Update32_const(uint32_t x[2][2][2][2][2])
{
x[0][0][0][0][0] ^= 0x80;
rrounds(x);
}
__device__ __forceinline__
@ -185,27 +253,44 @@ void Final(uint32_t x[2][2][2][2][2], uint32_t *hashval) @@ -185,27 +253,44 @@ void Final(uint32_t x[2][2][2][2][2], uint32_t *hashval)
x[1][1][1][1][1] ^= 1U;
/* "the state is then transformed invertibly through 10r identical rounds" */
#pragma unroll 2
for (int i = 0; i < 10; ++i) rrounds(x);
/* "output the first h/8 bytes of the state" */
hash_fromx(hashval, x);
hashval[0] = x[0][0][0][0][0];
hashval[1] = x[0][0][0][0][1];
hashval[2] = x[0][0][0][1][0];
hashval[3] = x[0][0][0][1][1];
hashval[4] = x[0][0][1][0][0];
hashval[5] = x[0][0][1][0][1];
hashval[6] = x[0][0][1][1][0];
hashval[7] = x[0][0][1][1][1];
}
#if __CUDA_ARCH__ >= 500
__global__ __launch_bounds__(TPB, 1)
__global__ __launch_bounds__(TPB50, 1)
#else
__global__ __launch_bounds__(TPB35, 1)
#endif
void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *g_hash)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
#if __CUDA_ARCH__ >= 500
uint2 Hash[4];
Hash[0] = __ldg(&g_hash[thread]);
Hash[1] = __ldg(&g_hash[thread + 1 * threads]);
Hash[2] = __ldg(&g_hash[thread + 2 * threads]);
Hash[3] = __ldg(&g_hash[thread + 3 * threads]);
#else
uint32_t Hash[8];
LOHI(Hash[0], Hash[1], __ldg(&((uint64_t*)g_hash)[thread]));
LOHI(Hash[2], Hash[3], __ldg(&((uint64_t*)g_hash)[thread + 1 * threads]));
LOHI(Hash[4], Hash[5], __ldg(&((uint64_t*)g_hash)[thread + 2 * threads]));
LOHI(Hash[6], Hash[7], __ldg(&((uint64_t*)g_hash)[thread + 3 * threads]));
#endif
uint32_t x[2][2][2][2][2] =
{
@ -219,6 +304,7 @@ void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *g_ha @@ -219,6 +304,7 @@ void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *g_ha
0x15815AEB, 0x4AB6AAD6, 0x9CDAF8AF, 0xD6032C0A
};
#if __CUDA_ARCH__ >= 500
x[0][0][0][0][0] ^= Hash[0].x;
x[0][0][0][0][1] ^= Hash[0].y;
x[0][0][0][1][0] ^= Hash[1].x;
@ -227,48 +313,7 @@ void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *g_ha @@ -227,48 +313,7 @@ void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *g_ha
x[0][0][1][0][1] ^= Hash[2].y;
x[0][0][1][1][0] ^= Hash[3].x;
x[0][0][1][1][1] ^= Hash[3].y;
rrounds(x);
x[0][0][0][0][0] ^= 0x80U;
rrounds(x);
Final(x, (uint32_t*) Hash);
g_hash[thread] = Hash[0];
g_hash[1 * threads + thread] = Hash[1];
g_hash[2 * threads + thread] = Hash[2];
g_hash[3 * threads + thread] = Hash[3];
}
}
#else
__global__ __launch_bounds__(TPB, 1)
void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *d_hash)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
uint32_t Hash[8];
uint64_t* g_hash = (uint64_t*) d_hash;
LOHI(Hash[0], Hash[1], __ldg(&g_hash[thread]));
LOHI(Hash[2], Hash[3], __ldg(&g_hash[thread + 1 * threads]));
LOHI(Hash[4], Hash[5], __ldg(&g_hash[thread + 2 * threads]));
LOHI(Hash[6], Hash[7], __ldg(&g_hash[thread + 3 * threads]));
uint32_t x[2][2][2][2][2] =
{
0xEA2BD4B4, 0xCCD6F29F, 0x63117E71, 0x35481EAE,
0x22512D5B, 0xE5D94E63, 0x7E624131, 0xF4CC12BE,
0xC2D0B696, 0x42AF2070, 0xD0720C35, 0x3361DA8C,
0x28CCECA4, 0x8EF8AD83, 0x4680AC00, 0x40E5FBAB,
0xD89041C3, 0x6107FBD5, 0x6C859D41, 0xF0B26679,
0x09392549, 0x5FA25603, 0x65C892FD, 0x93CB6285,
0x2AF2B5AE, 0x9E4B4E60, 0x774ABFDD, 0x85254725,
0x15815AEB, 0x4AB6AAD6, 0x9CDAF8AF, 0xD6032C0A
};
x[0][0][0][0][0] ^= Hash[0];
x[0][0][0][0][1] ^= Hash[1];
x[0][0][0][1][0] ^= Hash[2];
@ -277,29 +322,48 @@ void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *d_ha @@ -277,29 +322,48 @@ void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *d_ha
x[0][0][1][0][1] ^= Hash[5];
x[0][0][1][1][0] ^= Hash[6];
x[0][0][1][1][1] ^= Hash[7];
#endif
rrounds(x);
x[0][0][0][0][0] ^= 0x80U;
rrounds(x);
#if __CUDA_ARCH__ >= 500
Final(x, (uint32_t*)Hash);
g_hash[thread] = Hash[0];
g_hash[1 * threads + thread] = Hash[1];
g_hash[2 * threads + thread] = Hash[2];
g_hash[3 * threads + thread] = Hash[3];
#else
Final(x, Hash);
g_hash[thread] = ((uint64_t*)Hash)[0];
g_hash[1 * threads + thread] = ((uint64_t*)Hash)[1];
g_hash[2 * threads + thread] = ((uint64_t*)Hash)[2];
g_hash[3 * threads + thread] = ((uint64_t*)Hash)[3];
((uint64_t*)g_hash)[thread] = ((uint64_t*)Hash)[0];
((uint64_t*)g_hash)[1 * threads + thread] = ((uint64_t*)Hash)[1];
((uint64_t*)g_hash)[2 * threads + thread] = ((uint64_t*)Hash)[2];
((uint64_t*)g_hash)[3 * threads + thread] = ((uint64_t*)Hash)[3];
#endif
}
}
#endif
__host__
void cubehash256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint64_t *d_hash, int order)
{
uint32_t tpb = TPB;
uint32_t tpb = TPB35;
if (cuda_arch[thr_id] >= 500) tpb = TPB50;
dim3 grid((threads + tpb - 1) / tpb);
dim3 block(tpb);
cubehash256_gpu_hash_32 << <grid, block >> > (threads, startNounce, (uint2*)d_hash);
}
__host__
void cubehash256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint64_t *d_hash, int order, cudaStream_t stream)
{
uint32_t tpb = TPB35;
if (cuda_arch[thr_id] >= 500) tpb = TPB50;
dim3 grid((threads + tpb - 1) / tpb);
dim3 block(tpb);
cubehash256_gpu_hash_32 << <grid, block, 0, stream >> > (threads, startNounce, (uint2*)d_hash);
}
Loading…
Cancel
Save