From 7e595a36ea69027c8a28023399540a761e7686c3 Mon Sep 17 00:00:00 2001 From: Tanguy Pruvot Date: Tue, 2 Sep 2014 21:13:37 +0200 Subject: [PATCH] blake: cleanup, remove d_hash buf, not in a chain host: only bencode if gpu hash was found --- Makefile.am | 3 ++ blake32.cu | 121 +++++++++++++++------------------------------------- 2 files changed, 38 insertions(+), 86 deletions(-) diff --git a/Makefile.am b/Makefile.am index c73d9d2..520dff0 100644 --- a/Makefile.am +++ b/Makefile.am @@ -60,6 +60,9 @@ nvcc_FLAGS += $(JANSSON_INCLUDES) .cu.o: $(NVCC) $(nvcc_FLAGS) @CFLAGS@ --maxrregcount=128 -o $@ -c $< +blake32.o: blake32.cu + $(NVCC) $(nvcc_FLAGS) @CFLAGS@ --maxrregcount=64 -o $@ -c $< + # Luffa and Echo are faster with 80 registers than 128 x11/cuda_x11_luffa512.o: x11/cuda_x11_luffa512.cu $(NVCC) $(nvcc_FLAGS) @CFLAGS@ --maxrregcount=80 -o $@ -c $< diff --git a/blake32.cu b/blake32.cu index 814be2d..9755f93 100644 --- a/blake32.cu +++ b/blake32.cu @@ -35,11 +35,8 @@ extern int device_map[8]; extern cudaError_t MyStreamSynchronize(cudaStream_t stream, int situation, int thr_id); -// shared for 8 threads of addresses (cudaMalloc) -uint32_t* d_hash[8]; - __constant__ -static uint32_t pTarget[8]; +static uint32_t c_Target[8]; __constant__ static uint32_t __align__(32) c_PaddedMessage80[32]; // padded message (80 bytes + padding) @@ -181,7 +178,7 @@ extern __device__ __device_builtin__ void __nvvm_memset(uint8_t *, unsigned char #endif __global__ -void blake256_gpu_hash_80(int threads, uint32_t startNounce, uint32_t* outputHash) +void blake256_gpu_hash_80(int threads, uint32_t startNounce, uint32_t *resNounce) { uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x); if (thread < (uint32_t) threads) @@ -198,9 +195,12 @@ void blake256_gpu_hash_80(int threads, uint32_t startNounce, uint32_t* outputHas // ------ Close: Bytes 64 to 80 ------ -#if 0 /* __CUDA_ARCH__ >= 200 */ - __nvvm_memset((uint8_t*)(&msg[4]), 0, sizeof(msg)-16, 8); -#else + msg[0] = c_PaddedMessage80[16]; + msg[1] = c_PaddedMessage80[17]; + msg[2] = c_PaddedMessage80[18]; + msg[3] = nounce; /* our tested value */ + msg[4] = 0x80000000UL; //cuda_swab32(0x80U); + msg[5] = 0; // uchar[17 to 55] msg[6] = 0; msg[7] = 0; @@ -210,144 +210,93 @@ void blake256_gpu_hash_80(int threads, uint32_t startNounce, uint32_t* outputHas msg[11] = 0; msg[12] = 0; - msg[14] = 0; -#endif - msg[0] = c_PaddedMessage80[16]; - msg[1] = c_PaddedMessage80[17]; - msg[2] = c_PaddedMessage80[18]; - msg[3] = nounce; /* our tested value */ - msg[4] = 0x80000000UL; //cuda_swab32(0x80U); - msg[13] = 1; - msg[15] = 0x280; // 60-63 - - blake256_compress(h, msg, c_sigma, c_u256, 0x280); // or 0x80 - - uint32_t *outHash = &outputHash[thread<<3]; - //#pragma unroll 8 - for (int i=0; i < 8; i++) { - outHash[i] = cuda_swab32(h[i]); - } - } -} - -__host__ -void blake256_cpu_hash_80(int thr_id, int threads, uint32_t startNounce, uint32_t *d_outputHash, int order) -{ - const int threadsperblock = TPB; - - dim3 grid((threads + threadsperblock-1)/threadsperblock); - dim3 block(threadsperblock); - - size_t shared_size = 0; + msg[14] = 0; + msg[15] = 0x280; - blake256_gpu_hash_80<<>>(threads, startNounce, d_outputHash); + blake256_compress(h, msg, c_sigma, c_u256, 0x280); - MyStreamSynchronize(NULL, order, thr_id); -} - -__global__ -void gpu_check_hash_64(int threads, uint32_t startNounce, uint32_t *g_hash, uint32_t *resNounce) -{ - uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x); - if (thread < threads) - { - uint32_t* pHash = &g_hash[thread<<3]; for (int i = 7; i >= 0; i--) { - uint32_t hash = pHash[i]; - if (hash > pTarget[i]) { + uint32_t hash = cuda_swab32(h[i]); + if (hash > c_Target[i]) { return; } - if (hash < pTarget[i]) { + if (hash < c_Target[i]) { break; } } - uint32_t nounce = startNounce + thread; + /* keep the smallest nounce, hmm... */ if(resNounce[0] > nounce) resNounce[0] = nounce; } } __host__ -uint32_t cpu_check_hash_64(int thr_id, int threads, uint32_t startNounce, uint32_t *d_inputHash, int order) +uint32_t blake256_cpu_hash_80(int thr_id, int threads, uint32_t startNounce) { const int threadsperblock = TPB; - uint32_t result = 0xffffffff; - - cudaMemset(d_resNounce[thr_id], 0xff, sizeof(uint32_t)); dim3 grid((threads + threadsperblock-1)/threadsperblock); dim3 block(threadsperblock); - size_t shared_size = 0; - gpu_check_hash_64 <<>>(threads, startNounce, d_inputHash, d_resNounce[thr_id]); - MyStreamSynchronize(NULL, order, thr_id); + uint32_t result = 0xffffffffU; + cudaMemset(d_resNounce[thr_id], 0xff, sizeof(uint32_t)); + + blake256_gpu_hash_80<<>>(threads, startNounce, d_resNounce[thr_id]); + MyStreamSynchronize(NULL, 1, thr_id); if (cudaSuccess == cudaMemcpy(h_resNounce[thr_id], d_resNounce[thr_id], sizeof(uint32_t), cudaMemcpyDeviceToHost)) { cudaThreadSynchronize(); result = *h_resNounce[thr_id]; } - return result; } -__host__ -void blake256_cpu_init(int thr_id) -{ - CUDA_SAFE_CALL(cudaMemcpyToSymbol(c_sigma, host_sigma, sizeof(host_sigma), 0, cudaMemcpyHostToDevice)); - CUDA_SAFE_CALL(cudaMallocHost(&h_resNounce[thr_id], sizeof(uint32_t))); - CUDA_SAFE_CALL(cudaMalloc(&d_resNounce[thr_id], sizeof(uint32_t))); -} - __host__ void blake256_cpu_setBlock_80(uint32_t *pdata, const void *ptarget) { uint32_t PaddedMessage[32]; memcpy(PaddedMessage, pdata, 80); memset(&PaddedMessage[20], 0, 48); - - CUDA_SAFE_CALL(cudaMemcpyToSymbol(pTarget, ptarget, 32, 0, cudaMemcpyHostToDevice)); CUDA_SAFE_CALL(cudaMemcpyToSymbol(c_PaddedMessage80, PaddedMessage, sizeof(PaddedMessage), 0, cudaMemcpyHostToDevice)); + CUDA_SAFE_CALL(cudaMemcpyToSymbol(c_sigma, host_sigma, sizeof(host_sigma), 0, cudaMemcpyHostToDevice)); + CUDA_SAFE_CALL(cudaMemcpyToSymbol(c_Target, ptarget, 32, 0, cudaMemcpyHostToDevice)); } extern "C" int scanhash_blake32(int thr_id, uint32_t *pdata, const uint32_t *ptarget, uint32_t max_nonce, unsigned long *hashes_done) { const uint32_t first_nonce = pdata[19]; - const int throughput = TPB * 2048; - static bool init[8] = {0,0,0,0,0,0,0,0}; - uint32_t endiandata[20]; - uint32_t Htarg = ptarget[7]; + const int throughput = TPB * 2048; /* 2048 threads is the max on a 750Ti */ + static bool init[8] = { 0, 0, 0, 0, 0, 0, 0, 0 }; int rc = 0; if (opt_benchmark) - ((uint32_t*)ptarget)[7] = Htarg = 0x00000f; + ((uint32_t*)ptarget)[7] = 0x00000f; if (!init[thr_id]) { CUDA_SAFE_CALL(cudaSetDevice(device_map[thr_id])); - CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], 32 * throughput)); /* 32 bytes x 256K Threads (to be removed soon) */ - blake256_cpu_init(thr_id); + CUDA_SAFE_CALL(cudaMallocHost(&h_resNounce[thr_id], sizeof(uint32_t))); + CUDA_SAFE_CALL(cudaMalloc(&d_resNounce[thr_id], sizeof(uint32_t))); init[thr_id] = true; } blake256_cpu_setBlock_80(pdata, (void*)ptarget); - for (int k=0; k < 20; k++) - be32enc(&endiandata[k], pdata[k]); - do { - int order = 0; - uint32_t foundNonce; - // GPU HASH - blake256_cpu_hash_80(thr_id, throughput, pdata[19], d_hash[thr_id], order++); - - foundNonce = cpu_check_hash_64(thr_id, throughput, pdata[19], d_hash[thr_id], order++); + uint32_t foundNonce = blake256_cpu_hash_80(thr_id, throughput, pdata[19]); if (foundNonce != 0xffffffff) { + uint32_t endiandata[20]; uint32_t vhashcpu[8]; + uint32_t Htarg = ptarget[7]; + + for (int k=0; k < 20; k++) + be32enc(&endiandata[k], pdata[k]); + be32enc(&endiandata[19], foundNonce); blake32hash(vhashcpu, endiandata);