GOSTSec
/
ccminer-gostd-lite
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
Browse Source

improve jh512 with vectors (nist5,quark,sib,x11+,zr5) 

the main improvement is to reduce asm calls to read global mem

but, a few more regs are used (68 mini vs 64 on SM 5.2)
so reduce the forced launch bounds to allow 80 or 128 regs per thread

Note: cuda 6.5 seems not able to store with v4.u32... (7.5 is fine)
        st.global.v4.u32        [%rd2], {%r3783, %r3824, %r3823, %r3822};
        st.global.v2.u32        [%rd2+16], {%r3821, %r3820};
        st.global.u32   [%rd2+24], %r3819;
        st.global.u32   [%rd2+28], %r3818;
        st.global.u32   [%rd2+44], %r3814;
        st.global.u32   [%rd2+40], %r3815;
        ...

todo, check alexis variant.. but wanted to keep this code before in git...
master
Tanguy Pruvot 8 years ago
parent
6440a9bf41
commit
b9da6c67f5
2 changed files with 36 additions and 42 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Unified View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 3
      Makefile.am
  2. 75
      quark/cuda_jh512.cu

3
Makefile.am
Unescape View File

@ -139,6 +139,9 @@ x17/cuda_x17_sha512.o: x17/cuda_x17_sha512.cu
quark/cuda_quark_blake512.o: quark/cuda_quark_blake512.cu quark/cuda_quark_blake512.o: quark/cuda_quark_blake512.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $< $(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
quark/cuda_jh512.o: quark/cuda_jh512.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
quark/cuda_quark_keccak512.o: quark/cuda_quark_keccak512.cu quark/cuda_quark_keccak512.o: quark/cuda_quark_keccak512.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=88 -o $@ -c $< $(NVCC) $(nvcc_FLAGS) --maxrregcount=88 -o $@ -c $<

75
quark/cuda_jh512.cu
Unescape View File

@ -1,9 +1,9 @@
#include "cuda_helper.h" #include <cuda_helper.h>
// #include <stdio.h> // printf // #include <stdio.h> // printf
// #include <unistd.h> // sleep // #include <unistd.h> // sleep
/* 1344 bytes */ /* 1344 bytes, align 16 is there to allow ld.const.v4 (made auto. by the compiler) */
__constant__ static __align__(16) uint32_t c_E8_bslice32[42][8] = { __constant__ static __align__(16) uint32_t c_E8_bslice32[42][8] = {
// Round 0 (Function0) // Round 0 (Function0)
{ 0xa2ded572, 0x90d6ab81, 0x67f815df, 0xf6875a4d, 0x0a15847b, 0xc54f9f4e, 0x571523b7, 0x402bd1c3 }, { 0xa2ded572, 0x90d6ab81, 0x67f815df, 0xf6875a4d, 0x0a15847b, 0xc54f9f4e, 0x571523b7, 0x402bd1c3 },
@ -58,11 +58,11 @@ __constant__ static __align__(16) uint32_t c_E8_bslice32[42][8] = {
/*swapping bits 32i||32i+1||......||32i+15 with bits 32i+16||32i+17||......||32i+31 of 32-bit x*/ /*swapping bits 32i||32i+1||......||32i+15 with bits 32i+16||32i+17||......||32i+31 of 32-bit x*/
//#define SWAP16(x) (x) = ((((x) & 0x0000ffffUL) << 16) | (((x) & 0xffff0000UL) >> 16)); //#define SWAP16(x) (x) = ((((x) & 0x0000ffffUL) << 16) | (((x) & 0xffff0000UL) >> 16));
#define SWAP16(x) (x) = __byte_perm(x, x, 0x1032); #define SWAP16(x) (x) = __byte_perm(x, 0, 0x1032);
/*swapping bits 16i||16i+1||......||16i+7 with bits 16i+8||16i+9||......||16i+15 of 32-bit x*/ /*swapping bits 16i||16i+1||......||16i+7 with bits 16i+8||16i+9||......||16i+15 of 32-bit x*/
//#define SWAP8(x) (x) = ((((x) & 0x00ff00ffUL) << 8) | (((x) & 0xff00ff00UL) >> 8)); //#define SWAP8(x) (x) = ((((x) & 0x00ff00ffUL) << 8) | (((x) & 0xff00ff00UL) >> 8));
#define SWAP8(x) (x) = __byte_perm(x, x, 0x2301); #define SWAP8(x) (x) = __byte_perm(x, 0, 0x2301);
/* /*
__device__ __forceinline__ __device__ __forceinline__
@ -90,10 +90,9 @@ static void SWAP4x4(uint32_t *x) {
#pragma nounroll #pragma nounroll
// y is used as tmp register too // y is used as tmp register too
for (uint32_t y=0; y<4; y++, ++x) { for (uint32_t y=0; y<4; y++, ++x) {
asm("and.b32 %1, %0, 0xF0F0F0F0;" asm("and.b32 %1, %0, 0xF0F0F0F0;\n\t"
"xor.b32 %0, %0, %1;" "xor.b32 %0, %0, %1; shr.b32 %1, %1, 4;\n\t"
"shr.b32 %1, %1, 4;" "vshl.u32.u32.u32.clamp.add %0, %0, 4, %1;"
"vshl.u32.u32.u32.clamp.add %0, %0, 4, %1;\n\t"
: "+r"(*x) : "r"(y)); : "+r"(*x) : "r"(y));
} }
} }
@ -103,10 +102,9 @@ static void SWAP2x4(uint32_t *x) {
#pragma nounroll #pragma nounroll
// y is used as tmp register too // y is used as tmp register too
for (uint32_t y=0; y<4; y++, ++x) { for (uint32_t y=0; y<4; y++, ++x) {
asm("and.b32 %1, %0, 0xCCCCCCCC;" asm("and.b32 %1, %0, 0xCCCCCCCC;\n\t"
"xor.b32 %0, %0, %1;" "xor.b32 %0, %0, %1; shr.b32 %1, %1, 2; \n\t"
"shr.b32 %1, %1, 2;" "vshl.u32.u32.u32.clamp.add %0, %0, 2, %1;"
"vshl.u32.u32.u32.clamp.add %0, %0, 2, %1;\n\t"
: "+r"(*x) : "r"(y)); : "+r"(*x) : "r"(y));
} }
} }
@ -116,10 +114,9 @@ static void SWAP1x4(uint32_t *x) {
#pragma nounroll #pragma nounroll
// y is used as tmp register too // y is used as tmp register too
for (uint32_t y=0; y<4; y++, ++x) { for (uint32_t y=0; y<4; y++, ++x) {
asm("and.b32 %1, %0, 0xAAAAAAAA;" asm("and.b32 %1, %0, 0xAAAAAAAA;\n\t"
"xor.b32 %0, %0, %1;" "xor.b32 %0, %0, %1; shr.b32 %1, %1, 1; \n\t"
"shr.b32 %1, %1, 1;" "vshl.u32.u32.u32.clamp.add %0, %0, 1, %1;"
"vshl.u32.u32.u32.clamp.add %0, %0, 1, %1;\n\t"
: "+r"(*x) : "r"(y)); : "+r"(*x) : "r"(y));
} }
} }
@ -272,15 +269,23 @@ static void E8(uint32_t x[8][4])
} }
} }
__global__ __launch_bounds__(256, 4) __global__
void quark_jh512_gpu_hash_64(uint32_t threads, uint32_t startNounce, uint32_t* g_hash, const uint32_t *const __restrict__ g_nonceVector) //__launch_bounds__(256,2)
void quark_jh512_gpu_hash_64(const uint32_t threads, const uint32_t startNounce, uint32_t* g_hash, uint32_t * g_nonceVector)
{ {
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x); const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads) if (thread < threads)
{ {
const uint32_t nounce = (g_nonceVector != NULL) ? g_nonceVector[thread] : (startNounce + thread); const uint32_t nounce = (g_nonceVector != NULL) ? g_nonceVector[thread] : (startNounce + thread);
uint32_t hashPosition = nounce - startNounce; const uint32_t hashPosition = nounce - startNounce;
uint32_t *Hash = &g_hash[hashPosition * 16U]; uint32_t *Hash = &g_hash[(size_t)16 * hashPosition];
uint32_t h[16];
AS_UINT4(&h[ 0]) = AS_UINT4(&Hash[ 0]);
AS_UINT4(&h[ 4]) = AS_UINT4(&Hash[ 4]);
AS_UINT4(&h[ 8]) = AS_UINT4(&Hash[ 8]);
AS_UINT4(&h[12]) = AS_UINT4(&Hash[12]);
uint32_t x[8][4] = { /* init */ uint32_t x[8][4] = { /* init */
{ 0x964bd16f, 0x17aa003e, 0x052e6a63, 0x43d5157a }, { 0x964bd16f, 0x17aa003e, 0x052e6a63, 0x43d5157a },
{ 0x8d5e228a, 0x0bef970c, 0x591234e9, 0x61c3b3f2 }, { 0x8d5e228a, 0x0bef970c, 0x591234e9, 0x61c3b3f2 },
@ -294,40 +299,26 @@ void quark_jh512_gpu_hash_64(uint32_t threads, uint32_t startNounce, uint32_t* g
#pragma unroll #pragma unroll
for (int i = 0; i < 16; i++) for (int i = 0; i < 16; i++)
x[i/4][i & 3] ^= Hash[i]; x[i/4][i & 3] ^= h[i];
E8(x); E8(x);
#pragma unroll #pragma unroll
for (uint8_t i = 0; i < 16; i++) for (int i = 0; i < 16; i++)
x[(i+16)/4][(i+16) & 3] ^= Hash[i]; x[(i+16)/4][(i+16) & 3] ^= h[i];
x[0][0] ^= 0x80U; x[0][0] ^= 0x80U;
x[3][3] ^= 0x00020000U; x[3][3] ^= 0x00020000U;
E8(x); E8(x);
x[4][0] ^= 0x80U; x[4][0] ^= 0x80U;
x[7][3] ^= 0x00020000U; x[7][3] ^= 0x00020000U;
Hash[0] = x[4][0]; AS_UINT4(&Hash[ 0]) = AS_UINT4(&x[4][0]);
Hash[1] = x[4][1]; AS_UINT4(&Hash[ 4]) = AS_UINT4(&x[5][0]);
Hash[2] = x[4][2]; AS_UINT4(&Hash[ 8]) = AS_UINT4(&x[6][0]);
Hash[3] = x[4][3]; AS_UINT4(&Hash[12]) = AS_UINT4(&x[7][0]);
Hash[4] = x[5][0];
Hash[5] = x[5][1];
Hash[6] = x[5][2];
Hash[7] = x[5][3];
Hash[8] = x[6][0];
Hash[9] = x[6][1];
Hash[10] = x[6][2];
Hash[11] = x[6][3];
Hash[12] = x[7][0];
Hash[13] = x[7][1];
Hash[14] = x[7][2];
Hash[15] = x[7][3];
} }
} }

Write Preview
Loading…
Cancel
Save