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

m7: forgot to indent some cuda files 

and remove unused variables...
Tanguy Pruvot 10 years ago
parent
871add21f1
commit
fcd381cda2
6 changed files with 679 additions and 725 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. 174
      m7/cuda_m7_sha256.cu
  2. 83
      m7/cuda_mul2.cu
  3. 74
      m7/cuda_ripemd160.cu
  4. 85
      m7/cuda_tiger192.cu
  5. 6
      m7/m7.cu
  6. 40
      m7/m7_keccak512.cu

174
m7/cuda_m7_sha256.cu
Unescape View File

@ -1,10 +1,10 @@
#include <stdio.h> #include <stdio.h>
#include <memory.h> #include <memory.h>
#include "cuda_helper.h"
#include "sph/sph_types.h" #include "sph/sph_types.h"
#include "cuda_helper.h"
extern cudaError_t MyStreamSynchronize(cudaStream_t stream, int situation, int thr_id); extern cudaError_t MyStreamSynchronize(cudaStream_t stream, int situation, int thr_id);
//#define SPH_C64(x) ((uint64_t)(x ## ULL)) //#define SPH_C64(x) ((uint64_t)(x ## ULL))
@ -67,28 +67,35 @@ static const uint32_t cpu_K[64] = {
}; };
static __device__ __forceinline__ uint32_t bsg2_0(uint32_t x) __device__ __forceinline__
static uint32_t bsg2_0(uint32_t x)
{ {
uint32_t r1 = SPH_ROTR32(x,2); uint32_t r1 = SPH_ROTR32(x,2);
uint32_t r2 = SPH_ROTR32(x,13); uint32_t r2 = SPH_ROTR32(x,13);
uint32_t r3 = SPH_ROTR32(x,22); uint32_t r3 = SPH_ROTR32(x,22);
return xor3b(r1,r2,r3); return xor3b(r1,r2,r3);
} }
static __device__ __forceinline__ uint32_t bsg2_1(uint32_t x)
__device__ __forceinline__
static uint32_t bsg2_1(uint32_t x)
{ {
uint32_t r1 = SPH_ROTR32(x,6); uint32_t r1 = SPH_ROTR32(x,6);
uint32_t r2 = SPH_ROTR32(x,11); uint32_t r2 = SPH_ROTR32(x,11);
uint32_t r3 = SPH_ROTR32(x,25); uint32_t r3 = SPH_ROTR32(x,25);
return xor3b(r1,r2,r3); return xor3b(r1,r2,r3);
} }
static __device__ __forceinline__ uint32_t ssg2_0(uint32_t x)
__device__ __forceinline__
static uint32_t ssg2_0(uint32_t x)
{ {
uint64_t r1 = SPH_ROTR32(x,7); uint64_t r1 = SPH_ROTR32(x,7);
uint64_t r2 = SPH_ROTR32(x,18); uint64_t r2 = SPH_ROTR32(x,18);
uint64_t r3 = shr_t32(x,3); uint64_t r3 = shr_t32(x,3);
return xor3b(r1,r2,r3); return xor3b(r1,r2,r3);
} }
static __device__ __forceinline__ uint32_t ssg2_1(uint32_t x)
__device__ __forceinline__
static uint32_t ssg2_1(uint32_t x)
{ {
uint64_t r1 = SPH_ROTR32(x,17); uint64_t r1 = SPH_ROTR32(x,17);
uint64_t r2 = SPH_ROTR32(x,19); uint64_t r2 = SPH_ROTR32(x,19);
@ -96,7 +103,8 @@ static __device__ __forceinline__ uint32_t ssg2_1(uint32_t x)
return xor3b(r1,r2,r3); return xor3b(r1,r2,r3);
} }
static __device__ __forceinline__ void sha2_step1(uint32_t a,uint32_t b,uint32_t c,uint32_t &d,uint32_t e,uint32_t f,uint32_t g,uint32_t &h, __device__ __forceinline__
static void sha2_step1(uint32_t a,uint32_t b,uint32_t c,uint32_t &d,uint32_t e,uint32_t f,uint32_t g,uint32_t &h,
uint32_t in,const uint32_t Kshared) uint32_t in,const uint32_t Kshared)
{ {
uint32_t t1,t2; uint32_t t1,t2;
@ -111,12 +119,10 @@ d = d + t1;
h = t1 + t2; h = t1 + t2;
} }
static __forceinline__ void sha2_step1_host(uint32_t a,uint32_t b,uint32_t c,uint32_t &d,uint32_t e,uint32_t f,uint32_t g,uint32_t &h, __host__ __forceinline__
static void sha2_step1_host(uint32_t a,uint32_t b,uint32_t c,uint32_t &d,uint32_t e,uint32_t f,uint32_t g,uint32_t &h,
uint32_t in,const uint32_t Kshared) uint32_t in,const uint32_t Kshared)
{ {
uint32_t t1,t2; uint32_t t1,t2;
uint32_t vxandx = (((f) ^ (g)) & (e)) ^ (g); // xandx(e, f, g); uint32_t vxandx = (((f) ^ (g)) & (e)) ^ (g); // xandx(e, f, g);
uint32_t bsg21 =ROTR(e, 6) ^ ROTR(e, 11) ^ ROTR(e, 25); // bsg2_1(e); uint32_t bsg21 =ROTR(e, 6) ^ ROTR(e, 11) ^ ROTR(e, 25); // bsg2_1(e);
@ -129,7 +135,8 @@ d = d + t1;
h = t1 + t2; h = t1 + t2;
} }
static __device__ __forceinline__ void sha2_step2(uint32_t a,uint32_t b,uint32_t c,uint32_t &d,uint32_t e,uint32_t f,uint32_t g,uint32_t &h, __device__ __forceinline__
static void sha2_step2(uint32_t a,uint32_t b,uint32_t c,uint32_t &d,uint32_t e,uint32_t f,uint32_t g,uint32_t &h,
uint32_t* in,uint32_t pc,const uint32_t Kshared) uint32_t* in,uint32_t pc,const uint32_t Kshared)
{ {
uint32_t t1,t2; uint32_t t1,t2;
@ -156,10 +163,10 @@ t1 = h + bsg21 + vxandx + Kshared + in[pc];
t2 = bsg20 + andorv; t2 = bsg20 + andorv;
d = d + t1; d = d + t1;
h = t1 + t2; h = t1 + t2;
} }
static __forceinline__ void sha2_step2_host(uint32_t a,uint32_t b,uint32_t c,uint32_t &d,uint32_t e,uint32_t f,uint32_t g,uint32_t &h, __host__ __forceinline__
static void sha2_step2_host(uint32_t a,uint32_t b,uint32_t c,uint32_t &d,uint32_t e,uint32_t f,uint32_t g,uint32_t &h,
uint32_t* in,uint32_t pc,const uint32_t Kshared) uint32_t* in,uint32_t pc,const uint32_t Kshared)
{ {
uint32_t t1,t2; uint32_t t1,t2;
@ -172,7 +179,6 @@ uint32_t inx1 = in[pcidx1];
uint32_t inx2 = in[pcidx2]; uint32_t inx2 = in[pcidx2];
uint32_t inx3 = in[pcidx3]; uint32_t inx3 = in[pcidx3];
uint32_t ssg21 = ROTR(inx1, 17) ^ ROTR(inx1, 19) ^ SPH_T32((inx1) >> 10); //ssg2_1(inx1); uint32_t ssg21 = ROTR(inx1, 17) ^ ROTR(inx1, 19) ^ SPH_T32((inx1) >> 10); //ssg2_1(inx1);
uint32_t ssg20 = ROTR(inx3, 7) ^ ROTR(inx3, 18) ^ SPH_T32((inx3) >> 3); //ssg2_0(inx3); uint32_t ssg20 = ROTR(inx3, 7) ^ ROTR(inx3, 18) ^ SPH_T32((inx3) >> 3); //ssg2_0(inx3);
uint32_t vxandx = (((f) ^ (g)) & (e)) ^ (g); // xandx(e, f, g); uint32_t vxandx = (((f) ^ (g)) & (e)) ^ (g); // xandx(e, f, g);
@ -186,14 +192,12 @@ t1 = h + bsg21 + vxandx + Kshared + in[pc];
t2 = bsg20 + andorv; t2 = bsg20 + andorv;
d = d + t1; d = d + t1;
h = t1 + t2; h = t1 + t2;
} }
static __device__ __forceinline__ void sha2_round_body(uint32_t* in, uint32_t* r,const uint32_t* Kshared) __device__ __forceinline__
static void sha2_round_body(uint32_t* in, uint32_t* r,const uint32_t* Kshared)
{ {
uint32_t a=r[0]; uint32_t a=r[0];
uint32_t b=r[1]; uint32_t b=r[1];
uint32_t c=r[2]; uint32_t c=r[2];
@ -221,8 +225,8 @@ static __device__ __forceinline__ void sha2_round_body(uint32_t* in, uint32_t* r
sha2_step1(b,c,d,e,f,g,h,a,in[15],Kshared[15]); sha2_step1(b,c,d,e,f,g,h,a,in[15],Kshared[15]);
#pragma unroll 3 #pragma unroll 3
for (int i=0;i<3;i++) { for (int i=0;i<3;i++)
{
sha2_step2(a,b,c,d,e,f,g,h,in,0,Kshared[16+16*i]); sha2_step2(a,b,c,d,e,f,g,h,in,0,Kshared[16+16*i]);
sha2_step2(h,a,b,c,d,e,f,g,in,1,Kshared[17+16*i]); sha2_step2(h,a,b,c,d,e,f,g,in,1,Kshared[17+16*i]);
sha2_step2(g,h,a,b,c,d,e,f,in,2,Kshared[18+16*i]); sha2_step2(g,h,a,b,c,d,e,f,in,2,Kshared[18+16*i]);
@ -239,11 +243,8 @@ static __device__ __forceinline__ void sha2_round_body(uint32_t* in, uint32_t* r
sha2_step2(d,e,f,g,h,a,b,c,in,13,Kshared[29+16*i]); sha2_step2(d,e,f,g,h,a,b,c,in,13,Kshared[29+16*i]);
sha2_step2(c,d,e,f,g,h,a,b,in,14,Kshared[30+16*i]); sha2_step2(c,d,e,f,g,h,a,b,in,14,Kshared[30+16*i]);
sha2_step2(b,c,d,e,f,g,h,a,in,15,Kshared[31+16*i]); sha2_step2(b,c,d,e,f,g,h,a,in,15,Kshared[31+16*i]);
} }
r[0] = r[0] + a; r[0] = r[0] + a;
r[1] = r[1] + b; r[1] = r[1] + b;
r[2] = r[2] + c; r[2] = r[2] + c;
@ -254,10 +255,9 @@ static __device__ __forceinline__ void sha2_round_body(uint32_t* in, uint32_t* r
r[7] = r[7] + h; r[7] = r[7] + h;
} }
static __forceinline__ void sha2_round_body_host(uint32_t* in, uint32_t* r,const uint32_t* Kshared) __forceinline__
static void sha2_round_body_host(uint32_t* in, uint32_t* r,const uint32_t* Kshared)
{ {
uint32_t a=r[0]; uint32_t a=r[0];
uint32_t b=r[1]; uint32_t b=r[1];
uint32_t c=r[2]; uint32_t c=r[2];
@ -285,8 +285,8 @@ static __forceinline__ void sha2_round_body_host(uint32_t* in, uint32_t* r,const
sha2_step1_host(b,c,d,e,f,g,h,a,in[15],Kshared[15]); sha2_step1_host(b,c,d,e,f,g,h,a,in[15],Kshared[15]);
#pragma unroll 3 #pragma unroll 3
for (int i=0;i<3;i++) { for (int i=0;i<3;i++)
{
sha2_step2_host(a,b,c,d,e,f,g,h,in,0,Kshared[16+16*i]); sha2_step2_host(a,b,c,d,e,f,g,h,in,0,Kshared[16+16*i]);
sha2_step2_host(h,a,b,c,d,e,f,g,in,1,Kshared[17+16*i]); sha2_step2_host(h,a,b,c,d,e,f,g,in,1,Kshared[17+16*i]);
sha2_step2_host(g,h,a,b,c,d,e,f,in,2,Kshared[18+16*i]); sha2_step2_host(g,h,a,b,c,d,e,f,in,2,Kshared[18+16*i]);
@ -303,7 +303,6 @@ static __forceinline__ void sha2_round_body_host(uint32_t* in, uint32_t* r,const
sha2_step2_host(d,e,f,g,h,a,b,c,in,13,Kshared[29+16*i]); sha2_step2_host(d,e,f,g,h,a,b,c,in,13,Kshared[29+16*i]);
sha2_step2_host(c,d,e,f,g,h,a,b,in,14,Kshared[30+16*i]); sha2_step2_host(c,d,e,f,g,h,a,b,in,14,Kshared[30+16*i]);
sha2_step2_host(b,c,d,e,f,g,h,a,in,15,Kshared[31+16*i]); sha2_step2_host(b,c,d,e,f,g,h,a,in,15,Kshared[31+16*i]);
} }
r[0] = r[0] + a; r[0] = r[0] + a;
@ -316,27 +315,12 @@ static __forceinline__ void sha2_round_body_host(uint32_t* in, uint32_t* r,const
r[7] = r[7] + h; r[7] = r[7] + h;
} }
__global__
__global__ void m7_sha256_gpu_hash_120(int threads, uint32_t startNounce, uint64_t *outputHash) void m7_sha256_gpu_hash_120(int threads, uint32_t startNounce, uint64_t *outputHash)
{ {
/*
__shared__ uint32_t Kshared[64];
if (threadIdx.x < 64) {
Kshared[threadIdx.x]=K[threadIdx.x];
}
__syncthreads();
*/
union {
uint8_t h1[64];
uint32_t h4[16];
uint64_t h8[8];
} hash;
//uint32_t buf[8];
int thread = (blockDim.x * blockIdx.x + threadIdx.x); int thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads) if (thread < threads)
{ {
uint32_t nounce = startNounce + thread ; // original implementation uint32_t nounce = startNounce + thread ; // original implementation
uint32_t buf[8]; uint32_t buf[8];
@ -344,56 +328,36 @@ uint64_t h8[8];
uint32_t in3[16]={0}; uint32_t in3[16]={0};
#pragma unroll 13 #pragma unroll 13
for (int i=0;i<13;i++) {in2[i]= cuda_swab32(c_PaddedMessage80[i+16]);} for (int i=0; i<13; i++)
in2[i]= cuda_swab32(c_PaddedMessage80[i+16]);
in2[13]=cuda_swab32(nounce); in2[13]=cuda_swab32(nounce);
in2[14]=cuda_swab32(c_PaddedMessage80[30]); in2[14]=cuda_swab32(c_PaddedMessage80[30]);
in3[15]=0x3d0; in3[15]=0x3d0;
#pragma unroll 8 #pragma unroll 8
for (int i=0;i<8;i++) {buf[i]= pbuf[i];} for (int i=0; i<8; i++)
buf[i] = pbuf[i];
sha2_round_body(in2,buf,K); sha2_round_body(in2,buf,K);
sha2_round_body(in3,buf,K); sha2_round_body(in3,buf,K);
//#pragma unroll 8
//for (int i=0;i<8;i++) {hash.h4[i]=cuda_swab32(buf[i]);}
#pragma unroll 4 #pragma unroll 4
for (int i=0;i<4;i++) {outputHash[i*threads+thread]=cuda_swab32ll(((uint64_t*)buf)[i]);} for (int i=0; i<4; i++) {
outputHash[i*threads+thread] = cuda_swab32ll(((uint64_t*)buf)[i]);
}
////////////////////////////////////////////////////////////////////////////////////////////////// } // thread
} // threads
} }
__global__
__global__ void m7_sha256_gpu_hash_300(int threads, uint32_t startNounce, uint64_t *g_hash1, uint64_t *g_nonceVector, uint32_t *resNounce) void m7_sha256_gpu_hash_300(int threads, uint32_t startNounce, uint64_t *g_hash1, uint64_t *g_nonceVector, uint32_t *resNounce)
{ {
/*
__shared__ uint32_t Kshared[64];
if (threadIdx.x < 64) {
Kshared[threadIdx.x]=K[threadIdx.x];
}
__syncthreads();
*/
int thread = (blockDim.x * blockIdx.x + threadIdx.x); int thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads) if (thread < threads)
{ {
union {
uint8_t h1[304];
uint32_t h4[76];
uint64_t h8[38];
} hash;
uint32_t in[16],buf[8]; uint32_t in[16],buf[8];
#pragma unroll 8 #pragma unroll 8
for (int i=0;i<8;i++) {((uint64_t*)in)[i]= cuda_swab32ll(g_hash1[threads*i+thread]);} for (int i=0;i<8;i++) {((uint64_t*)in)[i]= cuda_swab32ll(g_hash1[threads*i+thread]);}
#pragma unroll 8 #pragma unroll 8
@ -415,20 +379,22 @@ uint64_t h8[38];
#pragma unroll 5 #pragma unroll 5
for (int i=0;i<5;i++) {((uint64_t*)in)[i]= cuda_swab32ll(g_hash1[threads*(i+32)+thread]);} for (int i=0;i<5;i++) {((uint64_t*)in)[i]= cuda_swab32ll(g_hash1[threads*(i+32)+thread]);}
((uint64_t*)in)[5] = g_hash1[threads*(5+32)+thread]; ((uint64_t*)in)[5] = g_hash1[threads*(5+32)+thread];
in[11]=0; in[11]=0;
in[12]=0; in[12]=0;
in[13]=0; in[13]=0;
in[14]=0; in[14]=0;
in[15]=0x968; in[15]=0x968;
int it=0; int it=0;
do { do {
in[15]-=8; in[15]-=8;
it++; it++;
} while (((uint8_t*)in)[44-it]==0); } while (((uint8_t*)in)[44-it]==0);
((uint8_t*)in)[44-it+1]=0x80; ((uint8_t*)in)[44-it+1]=0x80;
((uint64_t*)in)[5]= cuda_swab32ll(((uint64_t*)in)[5]); ((uint64_t*)in)[5]= cuda_swab32ll(((uint64_t*)in)[5]);
@ -438,73 +404,62 @@ uint64_t h8[38];
uint32_t nounce = startNounce +thread; uint32_t nounce = startNounce +thread;
bool rc = false; bool rc = false;
#pragma unroll 4 #pragma unroll 4
for (int i = 0; i < 4; i++) for (int i = 0; i < 4; i++)
{ {
if (cuda_swab32ll(((uint64_t*)buf)[i]) != ((uint64_t*)pTarget)[i]) { if (cuda_swab32ll(((uint64_t*)buf)[i]) != ((uint64_t*)pTarget)[i]) {
if (cuda_swab32ll(((uint64_t*)buf)[i]) < ((uint64_t*)pTarget)[i]) {rc = true;} else {rc = false;} if (cuda_swab32ll(((uint64_t*)buf)[i]) < ((uint64_t*)pTarget)[i])
rc = true;
else
rc = false;
//if cuda_swab32(((uint64_t*)buf)[3]) < ((uint64_t*)pTarget)[3]) {rc = true;} //if cuda_swab32(((uint64_t*)buf)[3]) < ((uint64_t*)pTarget)[3]) {rc = true;}
} }
} }
if (rc && resNounce[0] > nounce)
if(rc == true)
{
if(resNounce[0] > nounce)
resNounce[0] = nounce; resNounce[0] = nounce;
} // thread
}
////
} // threads
} }
__host__
void m7_sha256_cpu_init(int thr_id, int threads)
__host__ void m7_sha256_cpu_init(int thr_id, int threads)
{ {
// Kopiere die Hash-Tabellen in den GPU-Speicher
cudaMemcpyToSymbol( H256,cpu_H256,sizeof(cpu_H256),0, cudaMemcpyHostToDevice ); cudaMemcpyToSymbol( H256,cpu_H256,sizeof(cpu_H256),0, cudaMemcpyHostToDevice );
cudaMemcpyToSymbol( K,cpu_K,sizeof(cpu_K),0, cudaMemcpyHostToDevice ); cudaMemcpyToSymbol( K,cpu_K,sizeof(cpu_K),0, cudaMemcpyHostToDevice );
cudaMalloc(&d_MNonce[thr_id], sizeof(uint32_t)); cudaMalloc(&d_MNonce[thr_id], sizeof(uint32_t));
cudaMallocHost(&d_mnounce[thr_id], 1*sizeof(uint32_t)); cudaMallocHost(&d_mnounce[thr_id], 1*sizeof(uint32_t));
} }
__host__
__host__ uint32_t m7_sha256_cpu_hash_300(int thr_id, int threads, uint32_t startNounce, uint64_t *d_nonceVector,uint64_t *d_hash, int order) uint32_t m7_sha256_cpu_hash_300(int thr_id, int threads, uint32_t startNounce, uint64_t *d_nonceVector,uint64_t *d_hash, int order)
{ {
const int threadsperblock = 384;
uint32_t result = 0xffffffff; uint32_t result = 0xffffffff;
cudaMemset(d_MNonce[thr_id], 0xff, sizeof(uint32_t));
const int threadsperblock = 384; // Alignment mit mixtob Grösse. NICHT ÄNDERN
cudaMemset(d_MNonce[thr_id], 0xff, sizeof(uint32_t));
dim3 grid((threads + threadsperblock-1)/threadsperblock); dim3 grid((threads + threadsperblock-1)/threadsperblock);
dim3 block(threadsperblock); dim3 block(threadsperblock);
size_t shared_size = 0; size_t shared_size = 0;
m7_sha256_gpu_hash_300<<<grid, block, shared_size>>>(threads, startNounce, d_hash, d_nonceVector, d_MNonce[thr_id]); m7_sha256_gpu_hash_300<<<grid, block, shared_size>>>(threads, startNounce, d_hash, d_nonceVector, d_MNonce[thr_id]);
cudaMemcpy(d_mnounce[thr_id], d_MNonce[thr_id], sizeof(uint32_t), cudaMemcpyDeviceToHost); cudaMemcpy(d_mnounce[thr_id], d_MNonce[thr_id], sizeof(uint32_t), cudaMemcpyDeviceToHost);
MyStreamSynchronize(NULL, order, thr_id); MyStreamSynchronize(NULL, order, thr_id);
result = *d_mnounce[thr_id]; result = *d_mnounce[thr_id];
return result; return result;
} }
__host__
__host__ void m7_sha256_cpu_hash_120(int thr_id, int threads, uint32_t startNounce, uint64_t *d_outputHash, int order) void m7_sha256_cpu_hash_120(int thr_id, int threads, uint32_t startNounce, uint64_t *d_outputHash, int order)
{ {
const int threadsperblock = 512;
const int threadsperblock = 512; // Alignment mit mixtob Grösse. NICHT ÄNDERN
// berechne wie viele Thread Blocks wir brauchen
dim3 grid((threads + threadsperblock-1)/threadsperblock); dim3 grid((threads + threadsperblock-1)/threadsperblock);
dim3 block(threadsperblock); dim3 block(threadsperblock);
// dim3 grid(1);
// dim3 block(1);
size_t shared_size = 0; size_t shared_size = 0;
m7_sha256_gpu_hash_120<<<grid, block, shared_size>>>(threads, startNounce, d_outputHash); m7_sha256_gpu_hash_120<<<grid, block, shared_size>>>(threads, startNounce, d_outputHash);
@ -512,7 +467,8 @@ __host__ void m7_sha256_cpu_hash_120(int thr_id, int threads, uint32_t startNoun
MyStreamSynchronize(NULL, order, thr_id); MyStreamSynchronize(NULL, order, thr_id);
} }
__host__ void m7_sha256_setBlock_120(void *pdata,const void *ptarget) //not useful __host__
void m7_sha256_setBlock_120(void *pdata,const void *ptarget) //not useful
{ {
unsigned char PaddedMessage[128]; unsigned char PaddedMessage[128];
uint8_t ending =0x80; uint8_t ending =0x80;

83
m7/cuda_mul2.cu
Unescape View File

@ -85,7 +85,8 @@ ulonglong2 umul64wide (unsigned long long int a,
} }
__device__ __forceinline__ void umul_ppmmT4(t4_t *h, t4_t *l, t4_t m, t4_t n) __device__ __forceinline__
void umul_ppmmT4(t4_t *h, t4_t *l, t4_t m, t4_t n)
{ {
asm ("{\n\t" asm ("{\n\t"
".reg .u32 o0, o1, o2, o3, o4; \n\t" ".reg .u32 o0, o1, o2, o3, o4; \n\t"
@ -187,7 +188,8 @@ __device__ __forceinline__ void umul_ppmmT4(t4_t *h, t4_t *l, t4_t m, t4_t n){
#endif #endif
__device__ __forceinline__ t4_t T4(uint32_t thread, uint32_t threads, uint32_t idx, uint64_t *g){ __device__ __forceinline__
t4_t T4(uint32_t thread, uint32_t threads, uint32_t idx, uint64_t *g){
t4_t ret; t4_t ret;
ret.high = g[(idx*2 + 1)*threads + thread]; ret.high = g[(idx*2 + 1)*threads + thread];
ret.low = g[(idx*2)*threads + thread]; ret.low = g[(idx*2)*threads + thread];
@ -199,7 +201,8 @@ __device__ __forceinline__ t4_t T4(uint32_t thread, uint32_t threads, uint32_t i
return ret; return ret;
} }
__device__ __forceinline__ void T4_store(uint32_t thread, uint32_t threads, uint32_t idx, uint64_t *g, t4_t val){ __device__ __forceinline__
void T4_store(uint32_t thread, uint32_t threads, uint32_t idx, uint64_t *g, t4_t val){
g[(idx*2 + 1)*threads + thread]=val.high; g[(idx*2 + 1)*threads + thread]=val.high;
g[(idx*2)*threads + thread]=val.low; g[(idx*2)*threads + thread]=val.low;
@ -209,12 +212,14 @@ __device__ __forceinline__ void T4_store(uint32_t thread, uint32_t threads, uint
} }
__device__ __forceinline__ void T4_set(t4_t *d, uint64_t v){ __device__ __forceinline__
void T4_set(t4_t *d, uint64_t v){
d->high = 0; d->high = 0;
d->low = v; d->low = v;
} }
__device__ __forceinline__ t4_t T4_add(t4_t a, t4_t b){ __device__ __forceinline__
t4_t T4_add(t4_t a, t4_t b){
t4_t ret; t4_t ret;
uint32_t c=0; uint32_t c=0;
ret.low = a.low + b.low; ret.low = a.low + b.low;
@ -224,7 +229,8 @@ __device__ __forceinline__ t4_t T4_add(t4_t a, t4_t b){
return ret; return ret;
} }
__device__ __forceinline__ t4_t T4_add(uint64_t a, t4_t b){ __device__ __forceinline__
t4_t T4_add(uint64_t a, t4_t b){
t4_t ret; t4_t ret;
uint32_t c=0; uint32_t c=0;
ret.low = a + b.low; ret.low = a + b.low;
@ -234,8 +240,8 @@ __device__ __forceinline__ t4_t T4_add(uint64_t a, t4_t b){
return ret; return ret;
} }
__device__ __forceinline__
__device__ __forceinline__ uint32_t T4_lt(t4_t a, t4_t b){ uint32_t T4_lt(t4_t a, t4_t b){
if(a.high < b.high) if(a.high < b.high)
return 1; return 1;
if(a.high == b.high && a.low < b.low) if(a.high == b.high && a.low < b.low)
@ -243,7 +249,8 @@ __device__ __forceinline__ uint32_t T4_lt(t4_t a, t4_t b){
return 0; return 0;
} }
__device__ __forceinline__ uint32_t T4_gt(t4_t a, uint64_t b){ __device__ __forceinline__
uint32_t T4_gt(t4_t a, uint64_t b){
if(a.high) if(a.high)
return 1; return 1;
if(a.low > b) if(a.low > b)
@ -252,7 +259,8 @@ __device__ __forceinline__ uint32_t T4_gt(t4_t a, uint64_t b){
} }
__device__ void mulScalarT4(uint32_t thread, uint32_t threads, uint32_t len, uint64_t* g_p, uint64_t* g_v, t4_t sml, uint32_t *size){ __device__
void mulScalarT4(uint32_t thread, uint32_t threads, uint32_t len, uint64_t* g_p, uint64_t* g_v, t4_t sml, uint32_t *size){
t4_t ul, cl, hpl, lpl; t4_t ul, cl, hpl, lpl;
uint32_t i; uint32_t i;
T4_set(&cl,0); T4_set(&cl,0);
@ -289,13 +297,14 @@ __device__ void mulScalar(uint32_t thread, uint32_t threads, uint32_t len, uint6
*size = len + (cl != 0); *size = len + (cl != 0);
} }
uint64_t __device__ addmul_1g (uint32_t thread, uint32_t threads, uint64_t *sum, uint32_t sofst, uint64_t *x, uint64_t xsz, uint64_t a){ uint64_t __device__ addmul_1g (uint32_t thread, uint32_t threads, uint64_t *sum, uint32_t sofst, uint64_t *x, uint64_t xsz, uint64_t a)
{
uint64_t carry=0; uint64_t carry=0;
uint32_t i; uint32_t i;
uint64_t ul,lpl,hpl,rl; uint64_t ul,lpl,hpl,rl;
for(i=0; i < xsz; i++){ for(i=0; i < xsz; i++)
{
ul = x[i*threads + thread]; ul = x[i*threads + thread];
umul_ppmm (hpl, lpl, ul, a); umul_ppmm (hpl, lpl, ul, a);
@ -311,13 +320,15 @@ uint64_t __device__ addmul_1g (uint32_t thread, uint32_t threads, uint64_t *sum,
return carry; return carry;
} }
t4_t __device__ addmul_1gT4 (uint32_t thread, uint32_t threads, uint64_t *sum, uint32_t sofst, uint64_t *x, uint64_t xsz, t4_t a){ __device__
t4_t addmul_1gT4 (uint32_t thread, uint32_t threads, uint64_t *sum, uint32_t sofst, uint64_t *x, uint64_t xsz, t4_t a)
{
t4_t carry; t4_t carry;
uint32_t i; uint32_t i;
t4_t ul,lpl,hpl,rl; t4_t ul,lpl,hpl,rl;
T4_set(&carry,0); T4_set(&carry,0);
for(i=0; i < xsz; i++){ for(i=0; i < xsz; i++)
{
ul = T4(thread,threads,i,x); ul = T4(thread,threads,i,x);
umul_ppmmT4 (&hpl, &lpl, ul, a); umul_ppmmT4 (&hpl, &lpl, ul, a);
@ -333,9 +344,8 @@ t4_t __device__ addmul_1gT4 (uint32_t thread, uint32_t threads, uint64_t *sum, u
return carry; return carry;
} }
__global__
void gpu_mul(int threads, uint32_t ulegs, uint32_t vlegs, uint64_t *g_u, uint64_t *g_v, uint64_t *g_p)
__global__ void gpu_mul(int threads, uint32_t ulegs, uint32_t vlegs, uint64_t *g_u, uint64_t *g_v, uint64_t *g_p)
{ {
int thread = (blockDim.x * blockIdx.x + threadIdx.x); int thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads) if (thread < threads)
@ -354,7 +364,6 @@ __global__ void gpu_mul(int threads, uint32_t ulegs, uint32_t vlegs, uint64_t *g
mulScalar(thread,threads,ulegs,g_p,g_u,g_v[thread],&psize); mulScalar(thread,threads,ulegs,g_p,g_u,g_v[thread],&psize);
#if 1 #if 1
while (vofst < vlegs) { while (vofst < vlegs) {
//clear high word //TODO: right //clear high word //TODO: right
// printf("Size: %d\n", rp->size[tid]); // printf("Size: %d\n", rp->size[tid]);
@ -368,19 +377,17 @@ __global__ void gpu_mul(int threads, uint32_t ulegs, uint32_t vlegs, uint64_t *g
// if(D_REF(rp->d,up->size[tid] + vp->size[tid] - 1,tid) != (uint64_t)0) // if(D_REF(rp->d,up->size[tid] + vp->size[tid] - 1,tid) != (uint64_t)0)
// rp->size[tid]++; // rp->size[tid]++;
#endif #endif
} }
} }
__global__ void gpu_mulT4(int threads, uint32_t ulegs, uint32_t vlegs, uint64_t *g_u, uint64_t *g_v, uint64_t *g_p) __global__
void gpu_mulT4(int threads, uint32_t ulegs, uint32_t vlegs, uint64_t *g_u, uint64_t *g_v, uint64_t *g_p)
{ {
int thread = (blockDim.x * blockIdx.x + threadIdx.x); int thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads) if (thread < threads)
{ {
if(ulegs < vlegs) { // everything written the other way around... are you kidding me ?!
if(ulegs < vlegs){ ///everything written the other way around... are you kidding me ?!
uint64_t t1=ulegs; uint64_t t1=ulegs;
ulegs = vlegs; ulegs = vlegs;
vlegs = t1; vlegs = t1;
@ -396,8 +403,6 @@ __global__ void gpu_mulT4(int threads, uint32_t ulegs, uint32_t vlegs, uint64_t
// cuPrintf("U: %d V: %d\n", ulegs, vlegs); // cuPrintf("U: %d V: %d\n", ulegs, vlegs);
} }
uint32_t vofst=1,rofst=1,psize=0; uint32_t vofst=1,rofst=1,psize=0;
mulScalarT4(thread,threads,ulegs,g_p,g_u,T4(thread,threads,0,g_v),&psize); mulScalarT4(thread,threads,ulegs,g_p,g_u,T4(thread,threads,0,g_v),&psize);
@ -405,21 +410,17 @@ __global__ void gpu_mulT4(int threads, uint32_t ulegs, uint32_t vlegs, uint64_t
t4_t zero; t4_t zero;
T4_set(&zero,0); T4_set(&zero,0);
// while (vofst < vlegs) { // while (vofst < vlegs) {
#pragma unroll #pragma unroll
for (vofst=1;vofst<vlegs;vofst++) { for (vofst=1;vofst<vlegs;vofst++)
{
T4_store(thread,threads,psize,g_p,zero); T4_store(thread,threads,psize,g_p,zero);
T4_store(thread,threads,ulegs+rofst,g_p, addmul_1gT4(thread, threads, g_p, rofst, g_u, ulegs, T4(thread,threads,vofst,g_v))); T4_store(thread,threads,ulegs+rofst,g_p, addmul_1gT4(thread, threads, g_p, rofst, g_u, ulegs, T4(thread,threads,vofst,g_v)));
// vofst++; // vofst++;
rofst++; rofst++;
psize++; psize++;
} }
#endif #endif
} }
} }
@ -434,10 +435,9 @@ inline void gpuAssert(cudaError_t code, char *file, int line, bool abort=true)
} }
} }
__host__
__host__ void cpu_mul(int thr_id, int threads, uint32_t alegs, uint32_t blegs, uint64_t *g_a, uint64_t *g_b, uint64_t *g_p,int order) void cpu_mul(int thr_id, int threads, uint32_t alegs, uint32_t blegs, uint64_t *g_a, uint64_t *g_b, uint64_t *g_p,int order)
{ {
const int threadsperblock = 512; // Alignment mit mixtab Gr\F6sse. NICHT \C4NDERN const int threadsperblock = 512; // Alignment mit mixtab Gr\F6sse. NICHT \C4NDERN
// berechne wie viele Thread Blocks wir brauchen // berechne wie viele Thread Blocks wir brauchen
@ -445,13 +445,13 @@ __host__ void cpu_mul(int thr_id, int threads, uint32_t alegs, uint32_t blegs, u
dim3 block(threadsperblock); dim3 block(threadsperblock);
size_t shared_size = 0; size_t shared_size = 0;
gpu_mul<<<grid, block, shared_size>>>(threads, alegs, blegs, g_a, g_b, g_p) ;
gpu_mul<<<grid, block, shared_size>>>(threads, alegs, blegs, g_a, g_b, g_p) ;
} }
__host__ void cpu_mulT4(int thr_id, int threads, uint32_t alegs, uint32_t blegs, uint64_t *g_a, uint64_t *g_b, uint64_t *g_p, int order) __host__
void cpu_mulT4(int thr_id, int threads, uint32_t alegs, uint32_t blegs, uint64_t *g_a, uint64_t *g_b, uint64_t *g_p, int order)
{ {
const int threadsperblock = 256; // better occupancy (for both 780 and 750 ti's) const int threadsperblock = 256; // better occupancy (for both 780 and 750 ti's)
// berechne wie viele Thread Blocks wir brauchen // berechne wie viele Thread Blocks wir brauchen
@ -459,10 +459,13 @@ __host__ void cpu_mulT4(int thr_id, int threads, uint32_t alegs, uint32_t blegs,
dim3 block(threadsperblock); dim3 block(threadsperblock);
size_t shared_size = 0; size_t shared_size = 0;
//gpu_mulT4<<<grid, block, shared_size>>>(threads, alegs, blegs, g_a, g_b, g_p) ; //gpu_mulT4<<<grid, block, shared_size>>>(threads, alegs, blegs, g_a, g_b, g_p) ;
gpu_mulT4<<<grid, block, shared_size>>>(threads, blegs, alegs, g_b, g_a, g_p) ; gpu_mulT4<<<grid, block, shared_size>>>(threads, blegs, alegs, g_b, g_a, g_p) ;
} }
__host__ void mul_init(){ __host__
void mul_init()
{
} }

74
m7/cuda_ripemd160.cu
Unescape View File

@ -282,14 +282,12 @@ static const uint32_t IV[5] = {
(h)[0] = tmp; \ (h)[0] = tmp; \
} }
__global__
__global__ void m7_ripemd160_gpu_hash_120(int threads, uint32_t startNounce, uint64_t *outputHash) void m7_ripemd160_gpu_hash_120(int threads, uint32_t startNounce, uint64_t *outputHash)
{ {
int thread = (blockDim.x * blockIdx.x + threadIdx.x); int thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads) if (thread < threads)
{ {
uint32_t nounce = startNounce + thread; uint32_t nounce = startNounce + thread;
union { union {
uint8_t h1[64]; uint8_t h1[64];
@ -308,51 +306,52 @@ uint64_t h8[8];
#define F3(x, y, z) xornot64(x,y,z) #define F3(x, y, z) xornot64(x,y,z)
#define F4(x, y, z) xandx(z,x,y) #define F4(x, y, z) xandx(z,x,y)
#define F5(x, y, z) xornt64(x,y,z) #define F5(x, y, z) xornt64(x,y,z)
uint32_t in2[16],in3[16]; uint32_t in2[16],in3[16];
uint32_t in[16],buf[5]; uint32_t in[16],buf[5];
// #pragma unroll 16
// for (int i=0;i<16;i++) {in[i]= c_PaddedMessage80[i];}
#pragma unroll 16 #pragma unroll 16
for (int i=0;i<16;i++) {if ((i+16)<29) {in2[i]= c_PaddedMessage80[i+16];} for (int i=0;i<16;i++) {
else if ((i+16)==29) {in2[i]= nounce;} if ((i+16) < 29)
else if ((i+16)==30) {in2[i]= c_PaddedMessage80[i+16];} in2[i] = c_PaddedMessage80[i+16];
else {in2[i]= 0;}} else if ((i+16)==29)
in2[i] = nounce;
else if ((i+16)==30)
in2[i] = c_PaddedMessage80[i+16];
else
in2[i] = 0;
}
#pragma unroll 16 #pragma unroll 16
for (int i=0;i<16;i++) {in3[i]=0;} for (int i=0;i<16;i++)
in3[i]=0;
in3[14]=0x3d0; in3[14]=0x3d0;
// #pragma unroll 5
// for (int i=0;i<5;i++) {buf[i]=gpu_IV[i];}
#pragma unroll 5 #pragma unroll 5
for (int i=0;i<5;i++) {buf[i]=bufo[i];} for (int i=0;i<5;i++)
// RIPEMD160_ROUND_BODY(in, buf); //no need to calculate it several time (need to moved) buf[i]=bufo[i];
RIPEMD160_ROUND_BODY(in2, buf); RIPEMD160_ROUND_BODY(in2, buf);
RIPEMD160_ROUND_BODY(in3, buf); RIPEMD160_ROUND_BODY(in3, buf);
hash.h4[5]=0; hash.h4[5]=0;
#pragma unroll 5 #pragma unroll 5
for (int i=0; i<5; i++) for (int i=0; i<5; i++)
{hash.h4[i]=buf[i]; hash.h4[i]=buf[i];
}
//uint64_t *outHash = (uint64_t *)outputHash + 8 * thread;
//#pragma unroll 3
//for (int i=0;i<3;i++) {outHash[i]=hash.h8[i];}
#pragma unroll 3 #pragma unroll 3
for (int i=0;i<3;i++) {outputHash[i*threads+thread]=hash.h8[i];} for (int i=0;i<3;i++) {
//#pragma unroll 8 outputHash[i*threads+thread] = hash.h8[i];
//for (int i=0;i<8;i++) { if (i<3) {outputHash[i*threads+thread]=hash.h8[i];} else {outputHash[i*threads+thread]=0;}} }
} }
} }
void ripemd160_cpu_init(int thr_id, int threads) void ripemd160_cpu_init(int thr_id, int threads)
{ {
cudaMemcpyToSymbol(gpu_IV,IV,sizeof(IV),0, cudaMemcpyHostToDevice); cudaMemcpyToSymbol(gpu_IV,IV,sizeof(IV),0, cudaMemcpyHostToDevice);
} }
__host__ void ripemd160_setBlock_120(void *pdata) __host__
void ripemd160_setBlock_120(void *pdata)
{ {
unsigned char PaddedMessage[128]; unsigned char PaddedMessage[128];
uint8_t ending =0x80; uint8_t ending =0x80;
@ -371,29 +370,28 @@ __host__ void ripemd160_setBlock_120(void *pdata)
#define F3(x, y, z) (((x) | ~(y)) ^ (z)) #define F3(x, y, z) (((x) | ~(y)) ^ (z))
#define F4(x, y, z) ((((x) ^ (y)) & (z)) ^ (y)) #define F4(x, y, z) ((((x) ^ (y)) & (z)) ^ (y))
#define F5(x, y, z) ((x) ^ ((y) | ~(z))) #define F5(x, y, z) ((x) ^ ((y) | ~(z)))
uint32_t* alt_data =(uint32_t*)pdata; uint32_t* alt_data =(uint32_t*)pdata;
uint32_t in[16],buf[5]; uint32_t in[16],buf[5];
for (int i=0;i<16;i++)
in[i]= alt_data[i];
for (int i=0;i<16;i++) {in[i]= alt_data[i];} for (int i=0;i<5;i++)
buf[i]=IV[i];
for (int i=0;i<5;i++) {buf[i]=IV[i];}
RIPEMD160_ROUND_BODY(in, buf); //no need to calculate it several time (need to moved) RIPEMD160_ROUND_BODY(in, buf); //no need to calculate it several time (need to moved)
cudaMemcpyToSymbol(bufo, buf, 5*sizeof(uint32_t), 0, cudaMemcpyHostToDevice); cudaMemcpyToSymbol(bufo, buf, 5*sizeof(uint32_t), 0, cudaMemcpyHostToDevice);
} }
__host__ void m7_ripemd160_cpu_hash_120(int thr_id, int threads, uint32_t startNounce, uint64_t *d_outputHash, int order) __host__
void m7_ripemd160_cpu_hash_120(int thr_id, int threads, uint32_t startNounce, uint64_t *d_outputHash, int order)
{ {
const int threadsperblock = 256;
const int threadsperblock = 256; // Alignment mit mixtab Grösse. NICHT ÄNDERN
dim3 grid((threads + threadsperblock-1)/threadsperblock); dim3 grid((threads + threadsperblock-1)/threadsperblock);
dim3 block(threadsperblock); dim3 block(threadsperblock);
//dim3 grid(1);
//dim3 block(1);
size_t shared_size =0; size_t shared_size =0;
m7_ripemd160_gpu_hash_120<<<grid, block, shared_size>>>(threads, startNounce, d_outputHash); m7_ripemd160_gpu_hash_120<<<grid, block, shared_size>>>(threads, startNounce, d_outputHash);

85
m7/cuda_tiger192.cu
Unescape View File

@ -50,11 +50,13 @@ extern cudaError_t MyStreamSynchronize(cudaStream_t stream, int situation, int t
__constant__ uint64_t c_PaddedMessage80[16]; // padded message (80 bytes + padding) __constant__ uint64_t c_PaddedMessage80[16]; // padded message (80 bytes + padding)
__constant__ uint64_t bufo[3]; __constant__ uint64_t bufo[3];
static __constant__ uint64_t gpu_III[3]; static __constant__ uint64_t gpu_III[3];
static __constant__ uint64_t T1[256]; static __constant__ uint64_t T1[256];
static __constant__ uint64_t T2[256]; static __constant__ uint64_t T2[256];
static __constant__ uint64_t T3[256]; static __constant__ uint64_t T3[256];
static __constant__ uint64_t T4[256]; static __constant__ uint64_t T4[256];
static const uint64_t III[3] = { static const uint64_t III[3] = {
SPH_C64(0x0123456789ABCDEF),SPH_C64(0xFEDCBA9876543210),SPH_C64(0xF096A5B4C3B2E187) SPH_C64(0x0123456789ABCDEF),SPH_C64(0xFEDCBA9876543210),SPH_C64(0xF096A5B4C3B2E187)
}; };
@ -649,29 +651,24 @@ static const uint64_t cpu_T4[256] = {
(r)[2] = SPH_T64(C + (r)[2]); \ (r)[2] = SPH_T64(C + (r)[2]); \
} }
__global__
__global__ void m7_tiger192_gpu_hash_120(int threads, uint32_t startNounce, uint64_t *outputHash) void m7_tiger192_gpu_hash_120(int threads, uint32_t startNounce, uint64_t *outputHash)
{ {
__shared__ uint64_t sharedMem[1024]; __shared__ uint64_t sharedMem[1024];
if(threadIdx.x < 256)
{ if(threadIdx.x < 256) {
sharedMem[threadIdx.x] = T1[threadIdx.x]; sharedMem[threadIdx.x] = T1[threadIdx.x];
sharedMem[threadIdx.x+256] = T2[threadIdx.x]; sharedMem[threadIdx.x+256] = T2[threadIdx.x];
sharedMem[threadIdx.x+512] = T3[threadIdx.x]; sharedMem[threadIdx.x+512] = T3[threadIdx.x];
sharedMem[threadIdx.x+768] = T4[threadIdx.x]; sharedMem[threadIdx.x+768] = T4[threadIdx.x];
} }
__syncthreads(); __syncthreads();
int thread = (blockDim.x * blockIdx.x + threadIdx.x); int thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads) if (thread < threads)
{ {
uint32_t nounce = startNounce + thread; uint32_t nounce = startNounce + thread;
union {
uint8_t h1[64];
uint32_t h4[16];
uint64_t h8[8];
} hash;
/* /*
#undef MUL5 #undef MUL5
#undef MUL7 #undef MUL7
@ -691,8 +688,6 @@ uint64_t h8[8];
ROUND(b, c, a, X7, mul); \ ROUND(b, c, a, X7, mul); \
} }
#define ROUND(a, b, c, x, mul) { \ #define ROUND(a, b, c, x, mul) { \
c ^= x; \ c ^= x; \
a = SPH_T64(a - (sharedMem[c & 0xFF] ^ sharedMem[((c >> 16) & 0xFF)+256] \ a = SPH_T64(a - (sharedMem[c & 0xFF] ^ sharedMem[((c >> 16) & 0xFF)+256] \
@ -702,64 +697,72 @@ uint64_t h8[8];
b = mul(b); \ b = mul(b); \
} }
uint64_t in2[8];
uint64_t in[8],buf[3];
uint64_t in2[8],in3[8];
#pragma unroll 8 #pragma unroll 8
for (int i=0;i<8;i++) {in2[i]= c_PaddedMessage80[i+8];} for (int i=0; i<8; i++)
in2[i] = c_PaddedMessage80[i+8];
uint32_t* Mess = (uint32_t*)in2; uint32_t* Mess = (uint32_t*)in2;
Mess[13] = nounce; Mess[13] = nounce;
uint64_t in3[8];
#pragma unroll 8 #pragma unroll 8
for (int i=0;i<8;i++) {in3[i]=0;} for (int i=0; i<8; i++)
in3[i]=0;
in3[7]=0x3d0; in3[7]=0x3d0;
#pragma unroll 3
for (int i=0;i<3;i++) {buf[i]=bufo[i];} uint64_t buf[3];
#pragma unroll 3
for (int i=0; i<3; i++)
buf[i]=bufo[i];
TIGER_ROUND_BODY(in2, buf); TIGER_ROUND_BODY(in2, buf);
TIGER_ROUND_BODY(in3, buf); TIGER_ROUND_BODY(in3, buf);
#pragma unroll 8 #pragma unroll 8
for (int i=0;i<8;i++) { if (i<3) {outputHash[i*threads+thread]=buf[i];} else {outputHash[i*threads+thread]=0;}} for (int i=0;i<8;i++) {
} //// threads if (i<3) {
outputHash[i*threads+thread] = buf[i];
} else {
outputHash[i*threads+thread] = 0;
}
}
} // thread
} }
__host__
void tiger192_cpu_init(int thr_id, int threads) void tiger192_cpu_init(int thr_id, int threads)
{ {
cudaMemcpyToSymbol(gpu_III,III,sizeof(III),0, cudaMemcpyHostToDevice); cudaMemcpyToSymbol(gpu_III,III,sizeof(III),0, cudaMemcpyHostToDevice);
cudaMemcpyToSymbol(T1,cpu_T1,sizeof(cpu_T1),0, cudaMemcpyHostToDevice); cudaMemcpyToSymbol(T1,cpu_T1,sizeof(cpu_T1),0, cudaMemcpyHostToDevice);
cudaMemcpyToSymbol(T2,cpu_T2,sizeof(cpu_T2),0, cudaMemcpyHostToDevice); cudaMemcpyToSymbol(T2,cpu_T2,sizeof(cpu_T2),0, cudaMemcpyHostToDevice);
cudaMemcpyToSymbol(T3,cpu_T3,sizeof(cpu_T3),0, cudaMemcpyHostToDevice); cudaMemcpyToSymbol(T3,cpu_T3,sizeof(cpu_T3),0, cudaMemcpyHostToDevice);
cudaMemcpyToSymbol(T4,cpu_T4,sizeof(cpu_T4),0, cudaMemcpyHostToDevice); cudaMemcpyToSymbol(T4,cpu_T4,sizeof(cpu_T4),0, cudaMemcpyHostToDevice);
} }
__host__ void m7_tiger192_cpu_hash_120(int thr_id, int threads, uint32_t startNounce, uint64_t *d_outputHash, int order) __host__
void m7_tiger192_cpu_hash_120(int thr_id, int threads, uint32_t startNounce, uint64_t *d_outputHash, int order)
{ {
const int threadsperblock = 640; // 256
const int threadsperblock = 640; // Alignment mit mixtab Grösse. NICHT ÄNDERN
// const int threadsperblock = 256;
dim3 grid((threads + threadsperblock-1)/threadsperblock); dim3 grid((threads + threadsperblock-1)/threadsperblock);
dim3 block(threadsperblock); dim3 block(threadsperblock);
//dim3 grid(1);
//dim3 block(1);
size_t shared_size = 0; size_t shared_size = 0;
m7_tiger192_gpu_hash_120<<<grid, block, shared_size>>>(threads, startNounce, d_outputHash); m7_tiger192_gpu_hash_120<<<grid, block, shared_size>>>(threads, startNounce, d_outputHash);
MyStreamSynchronize(NULL, order, thr_id); MyStreamSynchronize(NULL, order, thr_id);
} }
__host__
__host__ void tiger192_setBlock_120(void *pdata) void tiger192_setBlock_120(void *pdata)
{ {
unsigned char PaddedMessage[128]; unsigned char PaddedMessage[128];
uint8_t ending =0x01; uint8_t ending =0x01;
memcpy(PaddedMessage, pdata, 122); memcpy(PaddedMessage, pdata, 122);
memset(PaddedMessage+122,ending,1); memset(PaddedMessage+122,ending,1);
memset(PaddedMessage+123, 0, 5); //useless memset(PaddedMessage+123, 0, 5); //useless
@ -782,14 +785,16 @@ __host__ void tiger192_setBlock_120(void *pdata)
b = mul(b); \ b = mul(b); \
} }
uint64_t* alt_data = (uint64_t*) pdata; uint64_t* alt_data = (uint64_t*) pdata;
uint64_t in[8],buf[3]; uint64_t in[8],buf[3];
for (int i=0;i<8;i++) {in[i]= alt_data[i];}
for (int i=0;i<3;i++) {buf[i]=III[i];}
TIGER_ROUND_BODY(in, buf) for (int i=0;i<8;i++)
cudaMemcpyToSymbol( bufo, buf, 3*sizeof(uint64_t), 0, cudaMemcpyHostToDevice); in[i] = alt_data[i];
for (int i=0;i<3;i++)
buf[i] = III[i];
TIGER_ROUND_BODY(in, buf)
cudaMemcpyToSymbol(bufo, buf, 3*sizeof(uint64_t), 0, cudaMemcpyHostToDevice);
} }

6
m7/m7.cu
Unescape View File

@ -22,7 +22,7 @@ extern "C"
extern int device_map[8]; extern int device_map[8];
extern bool opt_benchmark; extern bool opt_benchmark;
static uint64_t *d_hash[8]; //static uint64_t *d_hash[8];
static uint64_t *FinalHash[8]; static uint64_t *FinalHash[8];
static uint64_t *KeccakH[8]; static uint64_t *KeccakH[8];
static uint64_t *WhirlpoolH[8]; static uint64_t *WhirlpoolH[8];
@ -112,11 +112,9 @@ extern "C" void m7_hash(void *state, const void *input,uint32_t TheNonce, int de
{ {
// sha256(sha256*sha512*keccak512*ripemd160*haval*tiger1*whirlpool) // sha256(sha256*sha512*keccak512*ripemd160*haval*tiger1*whirlpool)
char data_str[245], hash_str[65], target_str[65];
uint8_t *bdata = 0; uint8_t *bdata = 0;
mpz_t bns[7]; mpz_t bns[7];
mpz_t product; mpz_t product;
int rc = 0;
for(int i=0; i < 7; i++) { for(int i=0; i < 7; i++) {
mpz_init(bns[i]); mpz_init(bns[i]);
@ -292,7 +290,6 @@ extern "C" int scanhash_m7(int thr_id, uint32_t *pdata,
tiger192_setBlock_120((void*)pdata); tiger192_setBlock_120((void*)pdata);
cuda_check_cpu_setTarget(ptarget); cuda_check_cpu_setTarget(ptarget);
uint32_t TheNonce = pdata[29];
do { do {
int order = 0; int order = 0;
@ -328,7 +325,6 @@ extern "C" int scanhash_m7(int thr_id, uint32_t *pdata,
MyStreamSynchronize(0,order++,thr_id); MyStreamSynchronize(0,order++,thr_id);
m7_bigmul_unroll2_cpu(0, throughput, RipemdH[thr_id], d_prod0[thr_id], d_prod1[thr_id],order); m7_bigmul_unroll2_cpu(0, throughput, RipemdH[thr_id], d_prod0[thr_id], d_prod1[thr_id],order);
MyStreamSynchronize(0,order++,thr_id); MyStreamSynchronize(0,order++,thr_id);
foundNonce = m7_sha256_cpu_hash_300(thr_id, throughput, pdata[29], NULL, d_prod1[thr_id], order); foundNonce = m7_sha256_cpu_hash_300(thr_id, throughput, pdata[29], NULL, d_prod1[thr_id], order);

40
m7/m7_keccak512.cu
Unescape View File

@ -5,6 +5,8 @@
extern cudaError_t MyStreamSynchronize(cudaStream_t stream, int situation, int thr_id); extern cudaError_t MyStreamSynchronize(cudaStream_t stream, int situation, int thr_id);
__constant__ uint64_t c_PaddedMessage80[16]; // padded message (80 bytes + padding)
static __constant__ uint64_t stateo[25]; static __constant__ uint64_t stateo[25];
static __constant__ uint64_t RC[24]; static __constant__ uint64_t RC[24];
static const uint64_t cpu_RC[24] = { static const uint64_t cpu_RC[24] = {
@ -22,7 +24,9 @@ static const uint64_t cpu_RC[24] = {
0x0000000080000001ull, 0x8000000080008008ull 0x0000000080000001ull, 0x8000000080008008ull
}; };
static __device__ __forceinline__ void keccak_block(uint64_t *s, const uint64_t *keccak_round_constants) { __device__ __forceinline__
static void keccak_block(uint64_t *s, const uint64_t *keccak_round_constants)
{
size_t i; size_t i;
uint64_t t[5], u[5], v, w; uint64_t t[5], u[5], v, w;
@ -136,8 +140,9 @@ static __device__ __forceinline__ void keccak_block(uint64_t *s, const uint64_t
} }
} }
__host__ __forceinline__
static __forceinline__ void keccak_block_host(uint64_t *s, const uint64_t *keccak_round_constants) { static void keccak_block_host(uint64_t *s, const uint64_t *keccak_round_constants)
{
size_t i; size_t i;
uint64_t t[5], u[5], v, w; uint64_t t[5], u[5], v, w;
@ -204,19 +209,12 @@ static __forceinline__ void keccak_block_host(uint64_t *s, const uint64_t *kecca
} }
} }
__global__ /* __launch_bounds__(256, 2) */
void m7_keccak512_gpu_hash_120(int threads, uint32_t startNounce, uint64_t *outputHash)
__constant__ uint64_t c_PaddedMessage80[16]; // padded message (80 bytes + padding)
__global__ void m7_keccak512_gpu_hash_120(int threads, uint32_t startNounce, uint64_t *outputHash)
{ {
int thread = (blockDim.x * blockIdx.x + threadIdx.x); int thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads) if (thread < threads)
{ {
uint32_t nounce = startNounce + thread; uint32_t nounce = startNounce + thread;
uint64_t state[25]; uint64_t state[25];
@ -237,38 +235,36 @@ __global__ void m7_keccak512_gpu_hash_120(int threads, uint32_t startNounce, uin
keccak_block(state,RC); keccak_block(state,RC);
#pragma unroll 8 #pragma unroll 8
for (int i=0;i<8;i++) {outputHash[i*threads+thread]=state[i];} for (int i=0;i<8;i++) {
outputHash[i*threads+thread] = state[i];
}
} //thread } //thread
} }
void m7_keccak512_cpu_init(int thr_id, int threads) void m7_keccak512_cpu_init(int thr_id, int threads)
{ {
cudaMemcpyToSymbol( RC,cpu_RC,sizeof(cpu_RC),0,cudaMemcpyHostToDevice); cudaMemcpyToSymbol( RC,cpu_RC,sizeof(cpu_RC),0,cudaMemcpyHostToDevice);
} }
__host__ void m7_keccak512_setBlock_120(void *pdata) __host__ void m7_keccak512_setBlock_120(void *pdata)
{ {
unsigned char PaddedMessage[128]; unsigned char PaddedMessage[128];
uint8_t ending =0x01; uint8_t ending =0x01;
memcpy(PaddedMessage, pdata, 122); memcpy(PaddedMessage, pdata, 122);
memset(PaddedMessage+122,ending,1); memset(PaddedMessage+122,ending,1);
memset(PaddedMessage+123, 0, 5); memset(PaddedMessage+123, 0, 5);
cudaMemcpyToSymbol( c_PaddedMessage80, PaddedMessage, 16*sizeof(uint64_t), 0, cudaMemcpyHostToDevice); cudaMemcpyToSymbol( c_PaddedMessage80, PaddedMessage, 16*sizeof(uint64_t), 0, cudaMemcpyHostToDevice);
uint64_t* alt_data = (uint64_t*) pdata; uint64_t* alt_data = (uint64_t*) pdata;
uint64_t state[25]; uint64_t state[25];
for(int i=0;i<25;i++) {state[i]=0;} for(int i=0;i<9;i++)
state[i] = alt_data[i];
for(int i=10;i<25;i++)
for (int i=0;i<9;i++) {state[i] ^= alt_data[i];} state[i] = 0;
keccak_block_host(state,cpu_RC); keccak_block_host(state,cpu_RC);
cudaMemcpyToSymbol(stateo, state, 25*sizeof(uint64_t), 0, cudaMemcpyHostToDevice); cudaMemcpyToSymbol(stateo, state, 25*sizeof(uint64_t), 0, cudaMemcpyHostToDevice);
} }

Write Preview
Loading…
Cancel
Save