serial_no
int64
1
24.2k
cuda_source
stringlengths
11
9.01M
19,001
#include "includes.h" __global__ void merge(unsigned char * img_all, unsigned char * img, float * selection, int n, int stride) { int x = blockIdx.x * TILE_DIM + threadIdx.x; int y = blockIdx.y * TILE_DIM + threadIdx.y; int width = gridDim.x * TILE_DIM; int idx = 0; float sum = 0; float weight = 0; for (int j = 0; j < TILE_DIM; j+= BLOCK_ROWS) { sum = 0; for (idx = 0; idx < n; idx ++) sum += selection[idx * stride + ((y+j)*width + x)]; for (idx = 0; idx < n; idx ++) selection[idx * stride + ((y+j)*width + x)] /= (sum + 1e-5); img[3*((y+j)*width + x)] = 0; img[3*((y+j)*width + x)+1] = 0; img[3*((y+j)*width + x)+2] = 0; for (idx = 0; idx < n; idx ++) { //weight = selection[idx * stride + ((y+j)*width + x)]; weight = 0.25; //weight = 0.5; img[3*((y+j)*width + x)] += (unsigned char) (img_all[idx * stride * 3 + 3*((y+j)*width + x)] * weight); img[3*((y+j)*width + x)+1] += (unsigned char) (img_all[idx * stride * 3 + 3*((y+j)*width + x) + 1] * weight); img[3*((y+j)*width + x)+2] += (unsigned char)(img_all[idx * stride * 3 + 3*((y+j)*width + x) + 2] * weight); } } }
19,002
#include <float.h> #include <cstdlib> #include "../device/device.cu" // __global__ void // reduce0(float* g_idata,float* g_odata, unsigned int n) { // extern __shared__ float temp[]; // int thid = threadIdx.x; // temp[thid] = g_idata[thid]; // __syncthreads(); // for(int offset = 1;offset < n; offset *= 2) { // if(thid >= offset) // temp[thid] += temp[thid - offset]; // __syncthreads(); // } // g_odata[thid] = temp[thid]; // } // cluster assignment using randomization __global__ void init_cluster_assignment(int k, int size, int * cluster_size, int * cluster_assignment){ for (int i=0; i<k; i++) cluster_size[i] = 0; int group = 0; int index = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x; int random = index % k ; for (int i=index; i<size; i+=stride){ group = (int) random; cluster_assignment[i] = group; cluster_size[group] += 1; } } __global__ void update_clusters(int k, float ** cluster, const int * cluster_assignment, int data_size, \ int dimensions, float ** feature_vector,const int * cluster_size, int * response){ response[0] = 0; int index = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x; float ** temp; temp = new float* [k]; for (int i=0; i<k; i++){ temp[i] = new float[dimensions]; for (int j=0; j<dimensions; j++){ temp[i][j] = (float) 0; } } for (int i=0; i<data_size; i++){ for (int j=0; j<dimensions; j++){ temp[cluster_assignment[i]][j] += feature_vector[i][j]; } } for (int i=index; i<k; i+=stride){ if (cluster_size[i] == 0){ continue; } for (int j=0; j<dimensions; j++){ if (cluster[i][j] != temp[i][j]/cluster_size[i]){ response[0] = 1; } cluster[i][j] = temp[i][j]/cluster_size[i]; } } } __global__ void update_cluster_assignment(int k, int * cluster_assignment, float ** cluster, int size, int dimension, float ** features){ int index = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x; for (int i=index; i<size; i+=stride){ cluster_assignment[i] = find_nearest_center(k, features[i], dimension, cluster); } }
19,003
#include "includes.h" __global__ void kernel_fill(float4* d_dx1, float val, int numel) { size_t col = threadIdx.x + blockIdx.x * blockDim.x; if (col >= numel) { return; } d_dx1[col].x = val; d_dx1[col].y = val; d_dx1[col].z = val; d_dx1[col].w = val; }
19,004
#include "includes.h" #define UMUL(a, b) ( (a) * (b) ) #define UMAD(a, b, c) ( UMUL((a), (b)) + (c) ) typedef unsigned int uint; typedef unsigned short ushort; typedef unsigned char uchar; #define SHARED_MEMORY_SIZE 49152 #define MERGE_THREADBLOCK_SIZE 128 static uint *d_PartialHistograms; /* * Function that maps value to bin in range 0 inclusive to binCOunt exclusive */ __global__ void mergePartialHistogramsKernel(uint *d_Histogram, uint *d_PartialHistograms, uint histogramCount, uint binCount) { for (uint bin = blockIdx.x; bin < binCount; bin += gridDim.x) { uint sum = 0; for (uint histogramIndex = threadIdx.x; histogramIndex < histogramCount; histogramIndex += MERGE_THREADBLOCK_SIZE) { sum += d_PartialHistograms[bin + histogramIndex * binCount]; } __shared__ uint data[MERGE_THREADBLOCK_SIZE]; data[threadIdx.x] = sum; for (uint stride = MERGE_THREADBLOCK_SIZE / 2; stride > 0; stride >>= 1) { __syncthreads(); if (threadIdx.x < stride) { data[threadIdx.x] += data[threadIdx.x + stride]; } } if (threadIdx.x == 0) { d_Histogram[bin] = data[0]; } } }
19,005
#include "CommonDataStructure.cuh" #pragma region GraphStructure_Part GraphStructure::GraphStructure() { vertexNum = 0; adjList.clear(); } GraphStructure::GraphStructure(int num) { vertexNum = 0; adjList.clear(); ReserveSpace(num); } GraphStructure::~GraphStructure() { } void GraphStructure::ReserveSpace(int num) { adjList.reserve(num); } void GraphStructure::SetVertexNum(int vNum) { vertexNum = vNum; adjList.resize(vertexNum); } void GraphStructure::AddEdge(int s, int e) { adjList[s].insert(e); } void GraphStructure::DeleteEdge(int s, int e) { adjList[s].erase(adjList[s].find(e)); } void GraphStructure::DeleteAllEdge(int s) { adjList[s].clear(); } #pragma endregion
19,006
/* check-thread-index.cu */ #include <stdio.h> #include <stdlib.h> #include <math.h> #include <cuda_runtime.h> #define CHECK_CUDA_CALL(call) \ { \ const cudaError_t error = call; \ \ if (error != cudaSuccess) { \ fprintf(stderr, "Error (%s:%d), code: %d, reason: %s\n", \ __FILE__, __LINE__, \ error, cudaGetErrorString(error)); \ exit(EXIT_FAILURE); \ } \ } \ void printMatrix(int* matC, int row, int col) { int i; int j; int* pC = matC; printf("Matrix (%d, %d)\n", row, col); for (i = 0; i < row; ++i) { for (j = 0; j < col; ++j) printf("%2d ", pC[j]); printf("\n"); pC += col; } printf("\n"); return; } __global__ void printThreadIndex(int* matA, int row, int col) { int x = threadIdx.x + blockIdx.x * blockDim.x; int y = threadIdx.y + blockIdx.y * blockDim.y; unsigned int i = y * col + x; printf("threadIdx: (%d, %d, %d), blockIdx: (%d, %d, %d), " "coordinate: (%d, %d), array index: %d, " "matrix value: %d\n", threadIdx.x, threadIdx.y, threadIdx.z, blockIdx.x, blockIdx.y, blockIdx.z, x, y, i, matA[i]); } int main(int argc, char** argv) { int i; int dev; cudaDeviceProp deviceProp; int matRow; int matCol; int numOfElements; int numOfBytes; int* hostMatA; int* devMatA; /* Setup device */ dev = 0; CHECK_CUDA_CALL(cudaGetDeviceProperties(&deviceProp, dev)); printf("Using device %d: %s\n", dev, deviceProp.name); CHECK_CUDA_CALL(cudaSetDevice(dev)); /* Set matrix size */ matCol = 8; matRow = 6; numOfElements = matRow * matCol; numOfBytes = numOfElements * sizeof(int); /* Allocate host memory */ hostMatA = (int*)calloc(numOfElements, sizeof(int)); for (i = 0; i < numOfElements; ++i) hostMatA[i] = i; printMatrix(hostMatA, matRow, matCol); /* Allocate device memory */ CHECK_CUDA_CALL(cudaMalloc((void**)&devMatA, numOfBytes)); /* Set execution configuration */ dim3 block(4, 2); dim3 grid((matCol + block.x - 1) / block.x, (matRow + block.y - 1) / block.y); /* Transfer matrix data from host */ CHECK_CUDA_CALL(cudaMemcpy(devMatA, hostMatA, numOfBytes, cudaMemcpyHostToDevice)); /* Call kernel from host */ printThreadIndex<<<grid, block>>>(devMatA, matRow, matCol); CHECK_CUDA_CALL(cudaDeviceSynchronize()); /* Free device and host memory */ CHECK_CUDA_CALL(cudaFree(devMatA)); free(hostMatA); /* Reset device */ CHECK_CUDA_CALL(cudaDeviceReset()); return EXIT_SUCCESS; }
19,007
#include <stdint.h> #include <unistd.h> #include <stdio.h> #include <assert.h> #include <sys/time.h> #include <time.h> #include <stdlib.h> #include <sys/mman.h> static void HandleError( cudaError_t err, const char *file, int line ) { if (err != cudaSuccess) { printf( "%s in %s at line %d\n", cudaGetErrorString( err ), file, line ); exit( EXIT_FAILURE ); } } #define HANDLE_ERROR( err ) (HandleError( err, __FILE__, __LINE__ )) #define HANDLE_NULL( a ) {if (a == NULL) { \ printf( "Host memory failed in %s at line %d\n", \ __FILE__, __LINE__ ); \ exit( EXIT_FAILURE );}} inline double diff_s(struct timeval start, struct timeval end) { return ((double) (end.tv_usec - start.tv_usec) / 1000000 + (double) (end.tv_sec - start.tv_sec)); } __global__ void readKernel(uint64_t *memory, uint64_t *memoryToRead) { int tid = threadIdx.x + blockIdx.x*blockDim.x; //memory[tid]=memoryToRead[tid]; //__shared__ uint64_t temp; uint64_t temp = memoryToRead[tid]; if(!temp) __syncthreads(); } __global__ void writeKernel(uint64_t *memory) { int tid = threadIdx.x + blockIdx.x*blockDim.x; memory[tid]=5; } __global__ void nullKernel(int *memory) { } __global__ void initCudaMallocd(uint64_t *memory, int N) { int tid =threadIdx.x; if(tid==0){ for(int k=0;k< N ;k++) memory[k]=5; } } void verify(uint64_t* memory, int N) { int error = 0; for(int i =0; i<N; i++){ if(memory[i]!=5){ error = 1; break; } } if(error) printf("error in verification\n"); else printf("verified SUCCESS\n"); } __global__ void verifyCudaMallocd(uint64_t* memory, int N) { int tid=threadIdx.x; if(tid==0) { int error = 0; for(int i =0; i<N; i++){ if(memory[i]!=5){ error = 1; break; } } if(error) printf("error in verification\n"); else printf("verified SUCCESS\n"); } } int main( int argc, char *argv[] ) { uint64_t *hostAllocd, *cudaMallocd, *cpuMallocd; int ITERATIONS = 100000; int numBytes = 1; struct timeval tv1, tv2; int opt; int read=0; //read benchmark? or write? int benchmarkType = 0; int locked = 0; //mlock data? int dryRun = 0; //dry run to measure noise TLB misses/...etc while ((opt = getopt(argc, argv, "m:b:i:r:ld")) != -1) { switch (opt) { case 'm': numBytes = atoi(optarg); //assert(numBytes%16 == 0 && numBytes<=1024); break; case 'b': benchmarkType = atoi(optarg); break; case 'i': ITERATIONS = atoi(optarg); break; case 'r': read = atoi(optarg); break; case 'l': locked = 1; break; case 'd': dryRun = 1; break; default: /* '?' */ break; } } int num_of_blocks=1; int num_of_threads_per_block=numBytes; if(numBytes>1024){ num_of_blocks = 16; num_of_threads_per_block = numBytes/16; } if(benchmarkType == 0 || benchmarkType == 1) HANDLE_ERROR(cudaFree(0)); switch (benchmarkType) { case 0: {//read/Write to hostAlloc'd data HANDLE_ERROR( cudaHostAlloc( &hostAllocd, sizeof(uint64_t)*numBytes, 0 ) ); for(int k=0;k< numBytes ;k++){ hostAllocd[k]=1; } if(read) { uint64_t *memoryToRead; HANDLE_ERROR( cudaHostAlloc( &memoryToRead, sizeof(uint64_t)*numBytes, 0 ) ); for(int k=0;k< numBytes ;k++) memoryToRead[k]=5; gettimeofday(&tv1, NULL); for(int i = 0; i < ITERATIONS; i++) { readKernel<<<num_of_blocks,num_of_threads_per_block>>>(hostAllocd,memoryToRead); HANDLE_ERROR( cudaDeviceSynchronize()); } gettimeofday(&tv2, NULL); cudaFreeHost(memoryToRead); //verify(hostAllocd,numBytes); } else { gettimeofday(&tv1, NULL); for(int i = 0; i < ITERATIONS; i++) { writeKernel<<<num_of_blocks,num_of_threads_per_block>>>(hostAllocd); HANDLE_ERROR( cudaDeviceSynchronize()); } gettimeofday(&tv2, NULL); verify(hostAllocd,numBytes); } HANDLE_ERROR( cudaGetLastError()); double elapsedTimeSeconds = diff_s(tv1,tv2); printf("HostAlloc [%s] Latency including kernel launch overhead = %f us\n",(read==1)?"read":"write",elapsedTimeSeconds*1e6/(float)ITERATIONS); cudaFreeHost(hostAllocd); break; } case 1: {//read/Write to cudaMalloc'd data cpuMallocd = (uint64_t *)malloc(sizeof(uint64_t)*numBytes); assert(cpuMallocd); for(int k=0;k< numBytes ;k++){ cpuMallocd[k]=1; } HANDLE_ERROR( cudaMalloc( &cudaMallocd, sizeof(uint64_t)*numBytes) ); HANDLE_ERROR( cudaMemcpy( cudaMallocd,cpuMallocd, sizeof(uint64_t)*numBytes,cudaMemcpyDefault) ); if(read) { uint64_t *memoryToRead; HANDLE_ERROR( cudaMalloc( &memoryToRead, sizeof(uint64_t)*numBytes ) ); initCudaMallocd<<<1,1>>>(memoryToRead,numBytes); HANDLE_ERROR( cudaDeviceSynchronize()); gettimeofday(&tv1, NULL); for(int i = 0; i < ITERATIONS; i++) { readKernel<<<num_of_blocks,num_of_threads_per_block>>>(cudaMallocd,memoryToRead); HANDLE_ERROR( cudaDeviceSynchronize()); } gettimeofday(&tv2, NULL); cudaFree(memoryToRead); //verifyCudaMallocd<<<1,1>>>(cudaMallocd,numBytes); //HANDLE_ERROR( cudaDeviceSynchronize()); } else { gettimeofday(&tv1, NULL); for(int i = 0; i < ITERATIONS; i++) { writeKernel<<<num_of_blocks,num_of_threads_per_block>>>(cudaMallocd); HANDLE_ERROR( cudaDeviceSynchronize()); } gettimeofday(&tv2, NULL); //verifyCudaMallocd<<<1,1>>>(cudaMallocd,numBytes); //HANDLE_ERROR( cudaDeviceSynchronize()); } HANDLE_ERROR( cudaGetLastError()); double elapsedTimeSeconds = diff_s(tv1,tv2); printf("CudaMalloc [%s] Latency including kernel launch overhead = %f us\n",(read==1)?"read":"write",elapsedTimeSeconds*1e6/(float)ITERATIONS); free(cpuMallocd); cudaFree(cudaMallocd); break; } case 2: { gettimeofday(&tv1, NULL); for(int i = 0; i < ITERATIONS; i++) { nullKernel<<<num_of_blocks,num_of_threads_per_block>>>(0); HANDLE_ERROR( cudaDeviceSynchronize()); } gettimeofday(&tv2, NULL); HANDLE_ERROR( cudaGetLastError()); double elapsedTimeSeconds = diff_s(tv1,tv2); printf("null kernel launch overhead = %f us\n",elapsedTimeSeconds*1e6/(float)ITERATIONS); break; } case 3: {//read/Write to cpu mallocd data uint64_t *memory_to_access = (uint64_t *)malloc(sizeof(uint64_t)*numBytes ); if(!dryRun) { if(locked) mlock(memory_to_access,sizeof(uint64_t)*numBytes); assert(memory_to_access); if(read) { for(int k=0;k< numBytes ;k++) memory_to_access[k]=5; uint64_t fake; if(numBytes<8) { gettimeofday(&tv1, NULL); for(int i=0; i<ITERATIONS; i++) { for (int j = 0; j < (numBytes); j++) { fake += memory_to_access[j]; } } gettimeofday(&tv2, NULL); } else { gettimeofday(&tv1, NULL); for(int i=0; i<ITERATIONS; i++) { for (int j = 0; j < (numBytes); j += 8) { fake += memory_to_access[j]; fake += memory_to_access[j + 1]; fake += memory_to_access[j + 2]; fake += memory_to_access[j + 3]; fake += memory_to_access[j + 4]; fake += memory_to_access[j + 5]; fake += memory_to_access[j + 6]; fake += memory_to_access[j + 7]; } } gettimeofday(&tv2, NULL); } } else { uint64_t fake=5; if(numBytes<8) { gettimeofday(&tv1, NULL); for(int i=0; i<ITERATIONS; i++) { for (int j = 0; j < (numBytes); j++) { memory_to_access[j] = fake; } } gettimeofday(&tv2, NULL); } else { gettimeofday(&tv1, NULL); for(int i=0; i<ITERATIONS; i++) { for (int j = 0; j < (numBytes); j += 8) { memory_to_access[j] = fake; memory_to_access[j + 1] = fake; memory_to_access[j + 2] = fake; memory_to_access[j + 3] = fake; memory_to_access[j + 4] = fake; memory_to_access[j + 5] = fake; memory_to_access[j + 6] = fake; memory_to_access[j + 7] = fake; } } gettimeofday(&tv2, NULL); } } double elapsedTimeSeconds = diff_s(tv1,tv2); printf("cpu malloc [%s] Latency = %f us\n",(read==1)?"read":"write",elapsedTimeSeconds*1e6/(float)ITERATIONS); } // printf("Press enter to continue...\n"); // getchar(); verify(memory_to_access,numBytes); free(memory_to_access); break; } case 4: {//read/Write to cpu but hostsllocd data uint64_t *memory_to_access; HANDLE_ERROR(cudaSetDevice(0)); HANDLE_ERROR(cudaSetDeviceFlags(cudaDeviceMapHost)); HANDLE_ERROR(cudaFree(0)); HANDLE_ERROR(cudaHostAlloc(&memory_to_access,sizeof(uint64_t)*numBytes,cudaHostAllocMapped)); if(!dryRun) { if(read) { for(int k=0;k< numBytes ;k++) memory_to_access[k]=5; uint64_t fake; if(numBytes<8) { gettimeofday(&tv1, NULL); for(int i=0; i<ITERATIONS; i++) { for (int j = 0; j < (numBytes); j++) { fake += memory_to_access[j]; } } gettimeofday(&tv2, NULL); } else { gettimeofday(&tv1, NULL); for(int i=0; i<ITERATIONS; i++) { for (int j = 0; j < (numBytes); j += 8) { fake += memory_to_access[j]; fake += memory_to_access[j + 1]; fake += memory_to_access[j + 2]; fake += memory_to_access[j + 3]; fake += memory_to_access[j + 4]; fake += memory_to_access[j + 5]; fake += memory_to_access[j + 6]; fake += memory_to_access[j + 7]; } } gettimeofday(&tv2, NULL); } } else { uint64_t fake=5; if(numBytes<8) { gettimeofday(&tv1, NULL); for(int i=0; i<ITERATIONS; i++) { for (int j = 0; j < (numBytes); j++) { memory_to_access[j] = fake; } } gettimeofday(&tv2, NULL); } else { gettimeofday(&tv1, NULL); for(int i=0; i<ITERATIONS; i++) { for (int j = 0; j < (numBytes); j += 8) { memory_to_access[j] = fake; memory_to_access[j + 1] = fake; memory_to_access[j + 2] = fake; memory_to_access[j + 3] = fake; memory_to_access[j + 4] = fake; memory_to_access[j + 5] = fake; memory_to_access[j + 6] = fake; memory_to_access[j + 7] = fake; } } gettimeofday(&tv2, NULL); } } double elapsedTimeSeconds = diff_s(tv1,tv2); printf("cpu hostAlloc [%s] Latency = %f us\n",(read==1)?"read":"write",elapsedTimeSeconds*1e6/(float)ITERATIONS); } // printf("Press enter to continue...\n"); // getchar(); verify(memory_to_access,numBytes); cudaFreeHost(memory_to_access); break; } case 5: {//read/Write to cpu but mallocManaged data uint64_t *memory_to_access; HANDLE_ERROR(cudaMallocManaged(&memory_to_access,sizeof(uint64_t)*numBytes)); if(!dryRun) { if(read) { for(int k=0;k< numBytes ;k++) memory_to_access[k]=5; uint64_t fake; if(numBytes<8) { gettimeofday(&tv1, NULL); for(int i=0; i<ITERATIONS; i++) { for (int j = 0; j < (numBytes); j++) { fake += memory_to_access[j]; } } gettimeofday(&tv2, NULL); } else { gettimeofday(&tv1, NULL); for(int i=0; i<ITERATIONS; i++) { for (int j = 0; j < (numBytes); j += 8) { fake += memory_to_access[j]; fake += memory_to_access[j + 1]; fake += memory_to_access[j + 2]; fake += memory_to_access[j + 3]; fake += memory_to_access[j + 4]; fake += memory_to_access[j + 5]; fake += memory_to_access[j + 6]; fake += memory_to_access[j + 7]; } } gettimeofday(&tv2, NULL); } } else { uint64_t fake=5; if(numBytes<8) { gettimeofday(&tv1, NULL); for(int i=0; i<ITERATIONS; i++) { for (int j = 0; j < (numBytes); j++) { memory_to_access[j] = fake; } } gettimeofday(&tv2, NULL); } else { gettimeofday(&tv1, NULL); for(int i=0; i<ITERATIONS; i++) { for (int j = 0; j < (numBytes); j += 8) { memory_to_access[j] = fake; memory_to_access[j + 1] = fake; memory_to_access[j + 2] = fake; memory_to_access[j + 3] = fake; memory_to_access[j + 4] = fake; memory_to_access[j + 5] = fake; memory_to_access[j + 6] = fake; memory_to_access[j + 7] = fake; } } gettimeofday(&tv2, NULL); } } double elapsedTimeSeconds = diff_s(tv1,tv2); printf("cpu mallocManaged [%s] Latency = %f us\n",(read==1)?"read":"write",elapsedTimeSeconds*1e6/(float)ITERATIONS); } // printf("Press enter to continue...\n"); // getchar(); verify(memory_to_access,numBytes); cudaFree(memory_to_access); break; } } if(benchmarkType == 0 || benchmarkType == 1) cudaDeviceReset(); return 0; }
19,008
#include "includes.h" __global__ void erosionRows3DKernel ( unsigned short *d_dst, unsigned short *d_src, int w, int h, int d, int kernel_radius ) { __shared__ unsigned short smem[ER_ROWS_BLOCKDIM_Z][ER_ROWS_BLOCKDIM_Y][(ER_ROWS_RESULT_STEPS + 2 * ER_ROWS_HALO_STEPS) * ER_ROWS_BLOCKDIM_X]; unsigned short *smem_thread = smem[threadIdx.z][threadIdx.y]; //Offset to the left halo edge const int baseX = (blockIdx.x * ER_ROWS_RESULT_STEPS - ER_ROWS_HALO_STEPS) * ER_ROWS_BLOCKDIM_X + threadIdx.x; const int baseY = blockIdx.y * ER_ROWS_BLOCKDIM_Y + threadIdx.y; const int baseZ = blockIdx.z * ER_ROWS_BLOCKDIM_Z + threadIdx.z; d_src += (baseZ * h + baseY) * w + baseX; d_dst += (baseZ * h + baseY) * w + baseX; //Load main data #pragma unroll for (int i = ER_ROWS_HALO_STEPS; i < ER_ROWS_HALO_STEPS + ER_ROWS_RESULT_STEPS; i++) { smem_thread[threadIdx.x + i * ER_ROWS_BLOCKDIM_X] = d_src[i * ER_ROWS_BLOCKDIM_X]; } //Load left halo #pragma unroll for (int i = 0; i < ER_ROWS_HALO_STEPS; i++) { smem_thread[threadIdx.x + i * ER_ROWS_BLOCKDIM_X] = (baseX + i * ER_ROWS_BLOCKDIM_X >= 0) ? d_src[i * ER_ROWS_BLOCKDIM_X] : 0; } //Load right halo #pragma unroll for (int i = ER_ROWS_HALO_STEPS + ER_ROWS_RESULT_STEPS; i < ER_ROWS_HALO_STEPS + ER_ROWS_RESULT_STEPS + ER_ROWS_HALO_STEPS; i++) { smem_thread[threadIdx.x + i * ER_ROWS_BLOCKDIM_X] = (baseX + i * ER_ROWS_BLOCKDIM_X < w) ? d_src[i * ER_ROWS_BLOCKDIM_X] : 0; } //Compute and store results __syncthreads(); #pragma unroll for (int i = ER_ROWS_HALO_STEPS; i < ER_ROWS_HALO_STEPS + ER_ROWS_RESULT_STEPS; i++) { unsigned short *smem_kern = &smem_thread[threadIdx.x + i * ER_ROWS_BLOCKDIM_X - kernel_radius]; unsigned short val = smem_kern[0]; //#pragma unroll for (int j = 1; j <= 2*kernel_radius; j++) { val = min(val, smem_kern[j]); } d_dst[i * ER_ROWS_BLOCKDIM_X] = val; } }
19,009
#include <curand.h> #include <curand_kernel.h> #define DIM 1600 #define PI 3.14159265 __global__ void erode(unsigned char *R_input, unsigned char *G_input, unsigned char *B_input, size_t i_size, unsigned char *r_dataC, unsigned char *g_dataC, unsigned char *b_dataC, unsigned long col, unsigned long row, unsigned int dim, int m) { int x = threadIdx.x + (blockIdx.x * blockDim.x); int y = threadIdx.y + (blockIdx.y * blockDim.y); int offset = x + y * i_size; int offset2, ximg, yimg; int c1 = 255,c2 = 255,c3 = 255; int end = dim/2, ini = -end; for (int i = ini; i <= end; i++) { ximg = x + i; for (int j = ini; j <= end; j++) { yimg = y + j; offset2 = ximg + yimg * i_size; if (ximg < col && yimg < row) if (ximg > 0 && yimg > 0) if(R_input[offset2]+G_input[offset2]+B_input[offset2]<c1+c2+c3) c1 = R_input[offset2]; c2 = G_input[offset2]; c3 = B_input[offset2]; } } r_dataC[offset] = c1; g_dataC[offset] = c2; b_dataC[offset] = c3; } __global__ void dilate(unsigned char *R_input, unsigned char *G_input, unsigned char *B_input, size_t i_size, unsigned char *r_dataC, unsigned char *g_dataC, unsigned char *b_dataC, unsigned long col, unsigned long row, unsigned int dim, int m) { int x = threadIdx.x + (blockIdx.x * blockDim.x); int y = threadIdx.y + (blockIdx.y * blockDim.y); int offset = x + y * i_size; int offset2, ximg, yimg; int c1 = 0,c2 = 0,c3 = 0; int end = dim/2, ini = -end; for (int i = ini; i <= end; i++) { ximg = x + i; for (int j = ini; j <= end; j++) { yimg = y + j; offset2 = ximg + yimg * i_size; if (ximg < col && yimg < row) if (ximg > 0 && yimg > 0) if(R_input[offset2]+G_input[offset2]+B_input[offset2]>c1+c2+c3) c1 = R_input[offset2]; c2 = G_input[offset2]; c3 = B_input[offset2]; } } r_dataC[offset] = c1; g_dataC[offset] = c2; b_dataC[offset] = c3; } __global__ void median_filter(unsigned char *R_input, unsigned char *G_input, unsigned char *B_input, size_t i_size, unsigned char *r_dataC, unsigned char *g_dataC, unsigned char *b_dataC, unsigned long col, unsigned long row, unsigned int dim) { int x = threadIdx.x + (blockIdx.x * blockDim.x); int y = threadIdx.y + (blockIdx.y * blockDim.y); int offset = x + y * i_size; int offset2, ximg, yimg; unsigned char temp_r = 0, temp_g = 0, temp_b = 0, temp; int end = dim/2, ini = -end, k = 0, n = 0, i, j; int hr[9]; int hg[9]; int hb[9]; for (i = ini; i <= end; i++) { ximg = x + i; for (j = ini; j <= end; j++) { yimg = y + j; offset2 = ximg + yimg * i_size; if (ximg < col && yimg < row) if (ximg > 0 && yimg > 0) { hr[n] = R_input[offset2]; hg[n] = G_input[offset2]; hb[n] = B_input[offset2]; n++;} k++; } } for (i = 0; i < n; i++) for (j= i + 1; j < n; j++) if (hr[j] < hr[i]) { temp = hr[j]; hr[j] = hr[i]; hr[i] = temp;} for (i = 0; i < n; i++) for (j= i + 1; j < n; j++) if (hg[j] < hg[i]) { temp = hg[j]; hg[j] = hg[i]; hg[i] = temp;} for (i = 0; i < n; i++) for (j= i + 1; j < n; j++) if (hb[j] < hb[i]) { temp = hb[j]; hb[j] = hb[i]; hb[i] = temp;} if(n%2 == 1){ temp_r = hr[(n/2)]; temp_g = hg[(n/2)]; temp_b = hb[(n/2)]; }else{ temp_r = hr[(n/2)] + hr[(n/2) - 1]; temp_g = hg[(n/2)] + hg[(n/2) - 1]; temp_b = hb[(n/2)] + hb[(n/2) - 1];} r_dataC[offset] = temp_r; g_dataC[offset] = temp_g; b_dataC[offset] = temp_b; } __global__ void Operador_Convolucion(unsigned char *R_input, unsigned char *G_input, unsigned char *B_input, size_t i_size, unsigned char *r_dataC, unsigned char *g_dataC, unsigned char *b_dataC, unsigned long col, unsigned long row, float *mask, unsigned int dim) { int x = threadIdx.x + (blockIdx.x * blockDim.x); int y = threadIdx.y + (blockIdx.y * blockDim.y); int offset = x + y * i_size; int offset2, ximg, yimg; unsigned char temp_r = 0, temp_g = 0, temp_b = 0; int end = dim/2, ini = -end, k = 0; for (int i = ini; i <= end; i++) { ximg = x + i; for (int j = ini; j <= end; j++) { yimg = y + j; offset2 = ximg + yimg * i_size; if (ximg < col && yimg < row) if (ximg > 0 && yimg > 0) { temp_r += R_input[offset2]*mask[k]; temp_g += G_input[offset2]*mask[k]; temp_b += B_input[offset2]*mask[k];} k++; } } r_dataC[offset] = temp_r; g_dataC[offset] = temp_g; b_dataC[offset] = temp_b; }
19,010
#include <stdio.h> #include <math.h> #include <cuda.h> // void Radix(int* array, int array_size, int max_digit); /* Thread function */ __host__ void rng(int* arr, int n); /* Seed function */ __host__ int max_el(int * vec, int n); __host__ int num_digit(int el); __device__ int to_digit(int el, int divider); __host__ int to_digit_host(int el, int divider); __host__ void print_array(int * array, int n); __global__ void count_to_bucket(int * data, int * bucket, int length, int digit); __host__ void countSort(int * data, int * bucket, int length, int digit); __host__ void empty_bucket(int * bucket, int size); __host__ void print_array_file(int * array, int array_len); int main(int argc,char *argv[]) { if(argc != 2) { perror("Please specify data length"); exit(1); } printf("flag 1\n"); int data_size = strtol(argv[1], NULL, 10); int numThread = 1000; float numBlocksFloat = (float) data_size / numThread; int numBlocks = ceil(numBlocksFloat); int *global_array; int *global_bucket; int max_digit; int base= 10; printf("data size : %d\n%.f\n", data_size,numBlocksFloat); printf("flag 2 thread %d block %d \n", numThread, numBlocks); // aloocating array to be accessible by both cpu and gpu cudaMallocManaged(&global_array, data_size*sizeof(int)+1); // cudaMalloc(&local_array,data_size*sizeof(int)+1); rng(global_array, data_size); // cudaMemcpy(d_a, a, sizeof(float) * N, cudaMemcpyHostToDevice); printf("flag 3\n"); printf("flag 4\n"); // global_array = (*int )malloc(data_size * sizeof(int)); // initialization data print_array(global_array, data_size); max_digit = num_digit(max_el(global_array, data_size)); printf("max digit %d\n", max_digit); int bucket_el = base*max_digit; cudaMallocManaged(&global_bucket, bucket_el*sizeof(int)+1); empty_bucket(global_bucket,bucket_el); for(int i = 1; i<= max_digit; i++){ count_to_bucket<<<numBlocks,numThread>>>(global_array,global_bucket,data_size,i); } // Wait for GPU to finish before accessing on host cudaDeviceSynchronize(); for(int i = 0; i<max_digit; i++){ countSort(global_array, global_bucket, data_size, i); } print_array(global_bucket,bucket_el); print_array(global_array, data_size); cudaFree(global_array); //cudaFree(global_bucket); return 0; } __global__ void count_to_bucket(int * data, int * bucket, int length, int digit){ int block = blockIdx.x; int thread = threadIdx.x; int i = block*1000+thread; // printf("block %d thread %d\n", digit, thread); //for(int i = (digit-1)*10 + thread;i <=(digit-1)*10+thread && i < length; i++){ if(block*1000+thread < length){ int num_bucket = to_digit(data[i], digit) + 10*(digit-1); printf("%d [%d] %d\n", data[i], digit, num_bucket); bucket[num_bucket] ++; } //} }; __host__ // void countSort(int * data, int * bucket, int length, int digit){ // int *local_sort = (int*) malloc (length * sizeof(int)); // int index = 0; // // sort // // printf("local sort "); // for(int block =0; block < digit; block++){ // for(int d = 0; d < 10; d++){ // for(int j = 0; j < length; j++){ // if(to_digit_host(data[j], block) == d){ // local_sort[index] = data[j]; // index ++; // bucket[block*10+d] --; // } // if(bucket[block*10+d] == 0) { // // printf("\n"); // break; // } // } // } // } // // printf("index ends in %d \n", index); // // copy // for(int i=0; i < length; i++){ // data[i] = local_sort[i]; // } // free(local_sort); // //empty_bucket(bucket, 10); // } void countSort(int * data, int * bucket, int length, int digit){ int * local_sort = (int*) malloc (length * sizeof(int)); int index = 0; // sort // printf("local sort "); for(int i =0; i < 10; i++){ for(int j = 0; j < length; j++){ if(to_digit_host(data[j], digit+1) == i){ local_sort[index] = data[j]; index ++; bucket[digit*10+i] --; } if(bucket[digit*10+i] == 0) { // printf("\n"); break; } } } // printf("index ends in %d \n", index); // copy for(int i=0; i < length; i++){ data[i] = local_sort[i]; } free(local_sort); empty_bucket(bucket, 10); } __host__ void empty_bucket(int * bucket, int size){ for(int i = 0; i < size; i++){ bucket[i] = 0; } } __host__ void rng(int* arr, int n) { int seed = 13516123; srand(seed); for(long i = 0; i < n; i++) { arr[i] = (int)rand(); } } __host__ int max_el(int * vec, int n){ int max = vec[0]; for(int i = 0; i < n; i++){ if(vec[i] > max) max = vec[i]; } return max; }; __device__ int to_digit(int el, int divider){ for(int i = 1; i< divider; i++){ el /= 10; } return el % 10; }; __host__ int to_digit_host(int el, int divider){ for(int i = 1; i< divider; i++){ el /= 10; } return el % 10; }; __host__ void print_array(int * array, int array_len){ int n = array_len; for(int i = 0; i < n; i++){ printf("%d ", array[i]); } printf("\n"); } __host__ void print_array_file(int * array, int array_len){ int n = array_len; FILE * fp; FILE * fo; int i; /* open the file for writing*/ fp = fopen ("../test/result.txt","w"); fo = fopen ("../output/output.txt","w"); /* write 10 lines of text into the file stream*/ for(i = 0; i < n;i++){ fprintf (fp, "%d ", array[i]); fprintf (fo, "%d ", array[i]); } fprintf (fp, "\n "); fprintf (fo, "\n "); /* close the file*/ fclose (fp); fclose (fo); } __host__ int num_digit(int el){ int digit = 1; while(el > 9){ el /= 10; digit++; } return digit; };
19,011
#include "cuda_runtime.h" #include "device_launch_parameters.h" #define _USE_MATH_DEFINES #include <cmath> #include <iostream> #include <string> #include <stdio.h> static const int DIM = 128; static const int NODES = DIM * DIM; static const double L = 1.0 * NODES; static const double TIME_OVERALL = 30.0; // seconds static const double DIRICHLET = 0.0; static const double NEUMAN = 5; __global__ void Kernel(double *v, double *vPrev, double hx, double ht, unsigned long timeSteps) { unsigned int idx_X = threadIdx.x + blockIdx.x * blockDim.x; unsigned int idx_Y = threadIdx.y + blockIdx.y * blockDim.y; unsigned int idx = idx_X + idx_Y * DIM; for (int iStep = 0; iStep < timeSteps; ++iStep) { if(idx > 0 && idx < (NODES - 1)) v[idx] = (((vPrev[idx + 1] - 2 * vPrev[idx] + vPrev[idx - 1]) * ht) / (hx * hx)) + vPrev[idx]; __syncthreads(); if (idx == 0) { v[idx] = DIRICHLET; } else if (idx == (NODES - 1)) { v[idx] = hx * NEUMAN + v[idx - 1]; } vPrev[idx] = v[idx]; __syncthreads(); } } int main() { size_t mem_size = sizeof(double) * NODES; cudaError cudaStatus; double hx = L / (NODES - 1); double ht = 1e-1;//(hx * hx) / 2; unsigned long timeSteps = TIME_OVERALL / ht; std::cout << "hx = " << hx << std::endl; std::cout << "ht = " << ht << std::endl; std::cout << "steps = " << timeSteps << std::endl; if (hx < 1e-9 || ht < 1e-14) { std::cout << "too small values of hx and (or) ht: " << std::endl; return 1; } double *devNodes; double *prevDevNodes; cudaMalloc((void **) &devNodes, mem_size); cudaMalloc((void **) &prevDevNodes, mem_size); cudaMemset(prevDevNodes, 0, mem_size); cudaMemset(devNodes, 0, mem_size); dim3 N_Grid(DIM / 32, DIM / 32, 1); dim3 N_Block(32, 32, 1); cudaStatus = cudaSetDevice(0); if (cudaStatus != cudaSuccess) { std::cout << "failed to set cuda device 0" << std::endl; cudaFree(devNodes); return 1; } Kernel <<< N_Grid, N_Block >>> (devNodes, prevDevNodes, hx, ht, timeSteps); cudaStatus = cudaGetLastError(); if(cudaStatus != cudaSuccess) { std::cout << "last error: " << cudaGetErrorString(cudaStatus) << std::endl; cudaFree(devNodes); return 1; } double *hostNodes = (double *) malloc(mem_size); cudaMemcpy(hostNodes, devNodes, mem_size, cudaMemcpyDeviceToHost); cudaFree(devNodes); cudaFree(prevDevNodes); for(size_t i = 0; i < NODES; i++) { std::cout << i << ": " << hostNodes[i] << std::endl; } free(hostNodes); return 0; }
19,012
#include <stdio.h> #include <stdlib.h> #include "cuda_runtime.h" // Defines #define GridWidth 60 #define BlockWidth 128 // Variables for host and device vectors. __global__ void AddVectors(float* A, float* B, float *C, int N) { int idx = blockIdx.x*blockDim.x + threadIdx.x; if (idx<=N) C[idx] = A[idx] + B[idx]; } // Host code performs setup and calls the kernel. int main(int argc, char** argv) { float* h_A; float* h_B; float* h_C; float* d_A; float* d_B; float* d_C; int N= GridWidth * BlockWidth; printf("N=%d\n",N); size_t size = N * sizeof(float); dim3 dimGrid(GridWidth); dim3 dimBlock(BlockWidth); h_A = (float*)malloc(size); h_B = (float*)malloc(size); h_C = (float*)malloc(size); cudaMalloc((void**)&d_A, size); cudaMalloc((void**)&d_B, size); cudaMalloc((void**)&d_C, size); // Initialize host vectors h_A and h_B for(int i=0; i<N; ++i) { h_A[i] = (float)i; h_B[i] = (float)(N-i); } // Copy host vectors h_A and h_B to device vectores d_A and d_B cudaMemcpy(d_A, h_A, size, cudaMemcpyHostToDevice); cudaMemcpy(d_B, h_B, size, cudaMemcpyHostToDevice); AddVectors<<<dimGrid, dimBlock>>>(d_A, d_B, d_C,N); cudaThreadSynchronize(); // Copy result from device memory to host memory cudaMemcpy(h_C, d_C, size, cudaMemcpyDeviceToHost); for(int i=0; i<N; ++i) { printf("%f\n",h_C[i]); } // Free device vectors cudaFree(d_A); cudaFree(d_B); cudaFree(d_C); // Free host memory free(h_A); free(h_B); free(h_C); }
19,013
#include "includes.h" //Bibliotecas Basicas //Biblioteca Thrust //Biblioteca cuRAND //PARAMETROS GLOBAIS const int QUANT_PAIS_AVALIA = 4; int POP_TAM = 200; int N_CIDADES = 20; int BLOCKSIZE = 1024; int TOTALTHREADS = 2048; int N_GERA = 100; const int MUT = 10; const int MAX = 19; const int MIN = 0; const int ELITE = 2; /* * Busca por erros nos processos da gpu */ __global__ void escolhePais(unsigned int n, unsigned int np, int *paisAle, double *fitness, int *pool) { int index = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x; for (int i=index; i<n; i+=stride) { double best = 10000.0; int best_index = -1; int idx; for (int j=0; j<QUANT_PAIS_AVALIA; j++) { idx = paisAle[i*QUANT_PAIS_AVALIA+j]; if (fitness[idx] < best) { best = fitness[idx]; best_index = idx; } } pool[i] = best_index; } }
19,014
#include "includes.h" __global__ void sobelEdgeDetectionSharedMemUnrollCoalsed(int *input, int *output, int width, int height, int thresh) { __shared__ int shMem[4 * _TILESIZE_2 * _TILESIZE_2 ]; int num = _UNROLL_; int size = num * _TILESIZE_2; int i = blockIdx.x * (num * _TILESIZE_) + threadIdx.x; int j = blockIdx.y * (num * _TILESIZE_) + threadIdx.y; int xind = threadIdx.x; int yind = threadIdx.y; for(int x = 0; x < num; x++) { for(int y = 0; y < num; y++) { int xOffset = x * (_TILESIZE_), yOffset = y * (_TILESIZE_); shMem[ size * (yind + yOffset) + (xind + xOffset)] = input[(j + yOffset) * width + (i + xOffset)]; } } __syncthreads(); if (i < width - _TILESIZE_ && j < height - _TILESIZE_ && xind > 0 && yind > 0 && xind < (_TILESIZE_2 - 1) && yind < (_TILESIZE_2 - 1)) { for(int x = 0; x < num; x++) { for(int y = 0; y < num; y++) { int xOffset = x * _TILESIZE_, yOffset = y * _TILESIZE_; int sum1 = shMem[(xind + 1 + xOffset) + size * (yind - 1 + yOffset)] - shMem[(xind - 1 + xOffset) + size * (yind - 1 + yOffset)] + 2 * shMem[(xind + 1 + xOffset) + size * (yind + yOffset)] - 2 * shMem[(xind - 1 + xOffset) + size * (yind + yOffset)] + shMem[(xind + 1 + xOffset) + size * (yind + 1 + yOffset)] - shMem[(xind - 1 + xOffset) + size * (yind + 1 + yOffset)]; int sum2 = shMem[(xind - 1 + xOffset) + size * (yind - 1 + yOffset)] + 2 * shMem[(xind + xOffset) + size * (yind - 1 + yOffset)] + shMem[(xind + 1 + xOffset) + size * (yind - 1 + yOffset)] - shMem[(xind - 1 + xOffset) + size * (yind + 1 + yOffset)] - 2 * shMem[(xind + xOffset) + size * (yind + 1 + yOffset)] - shMem[(xind + 1 + xOffset) + size * (yind + 1 + yOffset)]; int magnitude = sum1 * sum1 + sum2 * sum2; int index = (j + yOffset) * width + (i + xOffset); if(magnitude > thresh) output[index] = 255; else output[index] = 0; } } } }
19,015
#include "includes.h" __global__ void box_iou_cuda_kernel(float * box_iou, float4 * box1, float4 * box2, long M, long N, int idxJump) { int idx = blockIdx.x*blockDim.x + threadIdx.x; size_t b1_idx, b2_idx, b1_row_offset, b2_row_offset; float xmin1, xmin2, xmax1, xmax2, ymin1, ymin2, ymax1, ymax2, x_tl, y_tl, x_br, y_br, w, h, inter, area1, area2, iou; for (long i=idx; i<M*N; i+=idxJump){ b1_idx = i/N; b2_idx = i%N; b1_row_offset = b1_idx; b2_row_offset = b2_idx; xmin1 = box1[b1_row_offset].x; ymin1 = box1[b1_row_offset].y; xmax1 = box1[b1_row_offset].z; ymax1 = box1[b1_row_offset].w; xmin2 = box2[b2_row_offset].x; ymin2 = box2[b2_row_offset].y; xmax2 = box2[b2_row_offset].z; ymax2 = box2[b2_row_offset].w; x_tl = fmaxf(xmin1,xmin2); y_tl = fmaxf(ymin1,ymin2); x_br = fminf(xmax1,xmax2); y_br = fminf(ymax1,ymax2); w = (x_br - x_tl + 1)<0 ? 0.0f:(x_br - x_tl + 1); h = (y_br - y_tl + 1)<0 ? 0.0f:(y_br - y_tl + 1); inter = w*h; area1 = (xmax1-xmin1+1)*(ymax1-ymin1+1); area2 = (xmax2-xmin2+1)*(ymax2-ymin2+1); iou = inter/(area1+area2-inter); box_iou[b1_idx*N+b2_idx]=iou; } }
19,016
float h_A[]= { 0.7213110389323074, 0.7036072892760992, 0.7972208480899733, 0.5153771118746243, 0.9216180039321551, 0.9615308379788774, 0.6739112073326918, 0.7155256159701121, 0.79258098175816, 0.9020993892924711, 0.8466942281531666, 0.5537511206889679, 0.5438491782768006, 0.7116611841848426, 0.5345733881172183, 0.6305435920228089, 0.9277412059481013, 0.9601175756536102, 0.5446681834227135, 0.8851113168080971, 0.9213980885405118, 0.9915170026771223, 0.9587707645596766, 0.6219713016116879, 0.59276932445235, 0.8176418731352921, 0.6798959501796369, 0.7878206193846803, 0.9719590782979264, 0.9033969542114628, 0.5257703038302184, 0.5060736164359405, 0.7617378755679971, 0.6813167728709468, 0.9829966035233257, 0.9436321253947957, 0.7141845778293617, 0.8573416967736969, 0.5422510511098686, 0.8854271419940865, 0.7079571984354764, 0.5809701005539115, 0.7670813919644976, 0.5026467498089977, 0.9606221012390379, 0.881001105638563, 0.5452291376880389, 0.6132540058623579, 0.8398635939579402, 0.8950965387747699, 0.577723664133599, 0.65654378252201, 0.551610115160845, 0.7471229692397765, 0.6064047653161813, 0.5348684993169526, 0.5714834053077453, 0.5699335991556576, 0.9581613026614635, 0.5800884251550227, 0.5621315217726607, 0.7260545065405787, 0.8263375959182511, 0.7004871816808094, 0.6628612849215572, 0.7630559501530698, 0.6090032592882435, 0.5297518102839368, 0.9716773789062979, 0.9792908103361173, 0.813729030576051, 0.9358222319928466, 0.9552555022955985, 0.9659368142992655, 0.5503401170334505, 0.7983438229770574, 0.8612723782138829, 0.8954838168746426, 0.6707221843700835, 0.87132828504265, 0.6813242207001085, 0.8731189965641872, 0.8963920044873096, 0.6589936053193723, 0.5142121673426723, 0.831984826281815, 0.7639888826373817, 0.6420662927592042, 0.8954349479086924, 0.6547561830427948, 0.7015274218684611, 0.9130054177465503, 0.8337834495754337, 0.9077154271987949, 0.8910566519116231, 0.9933333958880843, 0.7208793338115973, 0.8732275221059247, 0.6499131589293037, 0.6225016536582735, 0.506024346065687, 0.6334473626646867, 0.8642540267703913, 0.6723920546622816, 0.6454028787559742, 0.7549523427254714, 0.805806790664657, 0.6303127937986285, 0.8937464695521591, 0.9274795925381079, 0.7702126603748829, 0.8055277344962463, 0.8157804696830382, 0.6219542206282047, 0.5431210508452308, 0.5110240954422681, 0.5513970040916557, 0.9426855261472222, 0.8748081215557051, 0.741094255177086, 0.8497124066557136, 0.6273164134428054, 0.7748003989421978, 0.6611962772062256, 0.7401352720701797, 0.7687019107934498, 0.7767744755636825, 0.5913898182077495, 0.5385837135035092, 0.6082693609582921, 0.8626094369370461, 0.9699138062719916, 0.6910816521277436, 0.8865828246797212, 0.7391787580562894, 0.8063511498264284, 0.7425674516704125, 0.6912159215757671, 0.6498378086468368, 0.5128719865057241, 0.7829125973602015, 0.7452102281974154, 0.9838934615177628, 0.7448248499577833, 0.7821284001877189, 0.8336625562667601, 0.7947946889587203, 0.6821792645687259, 0.6749657511955991, 0.82056385260541, 0.594456091081921, 0.5025934920429402, 0.5358054634484388, 0.7299644543146196, 0.800198263470014, 0.5200795780935461, 0.9127085069646359, 0.520189261488305, 0.5721914853245895, 0.8353448213929731, 0.6954853620284434, 0.5287393914234155, 0.7749133682929376, 0.9626332237047208, 0.504531933674538, 0.7416881835220519, 0.8029161141930703, 0.9512371692847112, 0.678590959823639, 0.5942002011283005, 0.7231946939255616, 0.8891894086624462, 0.8467433825580544, 0.8250284374644645, 0.5006569190556542, 0.7168051210436979, 0.7906952071921669, 0.8270348886306684, 0.6098752346678269, 0.5222347961416989, 0.7929018318998295, 0.5101055813477642, 0.7169833755909871, 0.5702461928129504, 0.9023390198515662, 0.8818647192968105, 0.963439971720073, 0.5654149968549657, 0.9693649864156556, 0.8935600740999206, 0.7186220005851536, 0.7447207297667482, 0.9237355949222104, 0.9169525724065697, 0.5802019547819947, 0.9405852601832476, 0.8876001178289146, 0.7693074221942977, 0.5587374478203172, 0.94048087746272, 0.7081619194854067, 0.5577780946399669, 0.639746138541101, 0.5677279896423741, 0.7013171406572657, 0.8531962429999775, 0.9534461688042684, 0.9544454327368554, 0.7419798377038764, 0.7781598798873539, 0.9880995476366918, 0.9085277704260113, 0.8808235015558032, 0.645561561502733, 0.6756942754185749, 0.5067823378774516, 0.506378470755337, 0.810777692389024, 0.5123445978187132, 0.8708238156107807, 0.8503474958572756, 0.954442138732992, 0.9080537030698932, 0.9018808628797211, 0.9449707700689803, 0.8191690079487154, 0.6827385944940092, 0.8929245413314859, 0.5292200559271782, 0.8624408829473031, 0.6845293685637726, 0.6978517435484533, 0.7839079461033043, 0.8375322422631433, 0.9537265678949418, 0.8208246713448575, 0.9297464038826257, 0.8533886395580749, 0.6595065038741466, 0.5367401755113856, 0.5711884859357816, 0.9956268735292508, 0.5355328946765383, 0.5455838142054785, 0.9854809795755022, 0.7816238163921971, 0.5441177260680754, 0.7809034256329053, 0.8833044171043674, 0.7976267671978272, 0.6949480150237191, 0.97620470200642, 0.8229129809550366, 0.869410101611787, 0.7857192541226444, 0.9049082880373116, 0.8217825261671052, 0.5978171261180575, 0.7516740248586333, 0.835374844356755, 0.6647492038017275, 0.654372380384227, 0.533928025397404, 0.8754977563944351, 0.5272353988484589, 0.9324332907020879, 0.8092159617417161, 0.8081259134157913, 0.7270657386338355, 0.9038540598848672, 0.9711575053233915, 0.6771643891208445, 0.9579552939218283, 0.853761852888534, 0.8960824555273585, 0.5269678068509709, 0.6943165773545628, 0.5796276656005226, 0.706014162848424, 0.8372419249150096, 0.6996409865937712, 0.5716824987141625, 0.6438334075025813, 0.5260148903229024, 0.9331886427333145, 0.9323333298241967, 0.7260844671738941, 0.9270432682797352, 0.9337665703653302, 0.6315745755524134, 0.7385872282881337, 0.7804016374726153, 0.8425343267477277, 0.8992882216574574, 0.8154295497769282, 0.9454044330995546, 0.9504649658919259, 0.7552753505697389, 0.703333042647019, 0.851159545729975, 0.6219148319282171, 0.916502088917321, 0.8613541833378122, 0.9124095174944935, 0.9286629029328552, 0.612550255934398, 0.8679719411162063, 0.9693607203380965, 0.8453412431169058, 0.8725376854813824, 0.7912174040299396, 0.5888947879582849, 0.9848924865749478, 0.8174611882014551, 0.7400586120333806, 0.9892661252731871, 0.9813085520067423, 0.9573712978603379, 0.649588198345867, 0.6250235203353991, 0.5955909287888107, 0.620373053569861, 0.526029835569561, 0.7144405404754234, 0.780285293622486, 0.8666553691054684, 0.8964231072882586, 0.8591118144087425, 0.9572999862830796, 0.9572257787849424, 0.7738885343608862, 0.6271141855599663, 0.5539161885973134, 0.674124267886359, 0.8456009864511342, 0.789085976005303, 0.9179901181123216, 0.6757078684677362, 0.554766193187559, 0.8991436828308943, 0.72861832616026, 0.8437634536967329, 0.9579330569265415, 0.9850939686630349, 0.8228998187444237, 0.8388421319341619, 0.5297061437767218, 0.6094811259115935, 0.6935463743720558, 0.8634731923807555, 0.8062166169630336, 0.7894429442727542, 0.7651759682262698, 0.9731711420190147, 0.9941330255107579, 0.7341363475012623, 0.9521718941174941, 0.9765870463957484, 0.7892140968169058, 0.7051042952693836, 0.7433897579012426, 0.7796078583896936, 0.9420146913245346, 0.5585284274921852, 0.6614237390233539, 0.8225768496189233, 0.9228783746366208, 0.5543525000804628, 0.712200484202808, 0.7458232809428115, 0.6370976984346218, 0.7614142413586191, 0.807507380810081, 0.5224827912351996, 0.6724736119732511, 0.593179868665466, 0.9306430166660243, 0.5324363741372664, 0.8889130123351461, 0.9107423542503432, 0.6550202302554903, 0.944894451539642, 0.9052118156355418, 0.5590466975750978, 0.805467627477866, 0.980986543801699, 0.5502338170531667, 0.7555823165396496, 0.5443648978107398, 0.7345272275399661, 0.5414860274201246, 0.5203901051451236, 0.5423934957350249, 0.6539933857773308, 0.9208207415813712, 0.5502128020162411, 0.6876157144549884, 0.5702853582447343, 0.5904228543623173, 0.8889623987647162, 0.8985261691881565, 0.5048154598888724, 0.6076562850468037, 0.6961908971137987, 0.6500172650564393, 0.8511948985756508, 0.66319992047087, 0.9484600768282544, 0.5089392129473143, 0.7668151252216436, 0.8848797511184859, 0.8442684141503054, 0.8293842491761947, 0.9228608712069555, 0.716848629788235, 0.9187140383200945, 0.7381151908607154, 0.9316436232594469, 0.9007970659292505, 0.8571124627981633, 0.7563461277617688, 0.513696518145864, 0.5748871483045002, 0.9520785587809371, 0.8896683038703164, 0.6998782192395402, 0.8834708853708064, 0.6918430479828297, 0.9714088838129342, 0.7358356385444389, 0.706251044986796, 0.6445022136693828, 0.6502160963722614, 0.67845490395182, 0.6989987170451886, 0.5019984250344179, 0.595813407554385, 0.7070955606691687, 0.8216025133858569, 0.6394441217980584, 0.9861550745536893, 0.5630267871638102, 0.7197774220892457, 0.5707820282987905, 0.5554174483683547, 0.5608175917649101, 0.7346401576638559, 0.6966702058717653, 0.6445646315267469, 0.8744094501599702, 0.8013922028376101, 0.8827004493426109, 0.6030222621908303, 0.8946255779552771, 0.5787758041400044, 0.7187007220827769, 0.6284769420023835, 0.7346188868724551, 0.695829407120654, 0.6386811653525657, 0.7150252903858563, 0.6293714022285715, 0.5249633990391243, 0.8647394391293555, 0.8230188310468833, 0.8976785086232437, 0.5039818379281453, 0.8211569054426457, 0.7194889824918933, 0.9289131829642242, 0.7541040312860572, 0.742051339762712, 0.5515601578515075, 0.8893119930560853, 0.7849628852472481, 0.8274409410965613, 0.9430988897880357, 0.5593617432017448, 0.9863674781085019, 0.8818683165019685, 0.5334322723076458, 0.5431232249183211, 0.956274780832098, 0.6750059580031558, 0.8033812934976199, 0.5633421647434877, 0.7588173120251935, 0.6148214442198539, 0.5306560401308527, 0.9159055555823894, 0.7829682549085655, 0.6077550001723855, 0.6768516473519923, 0.9083875292208695, 0.8238798732507031, 0.8842950293224175, 0.7509042583446859, 0.5835260581189603, 0.771432641391331, 0.8735359080421995, 0.6654501818823724, 0.8330481168914817, 0.6586063723028632, 0.5324851641008854, 0.6516233791749912, 0.8668811241368004, 0.6901352977817978, 0.8276546369593913, 0.883845084525178, 0.5816936278385099, 0.6363005501359384, 0.9231059836000148, 0.7356217224283114, 0.8089825425159602, 0.5589388507840811, 0.6306424940443877, 0.5350914779207365, 0.5810872349283129, 0.8127313666556297, 0.9468961803825973, 0.7860397530364289, 0.6428268710185303, 0.8479121114897317, 0.9319903051408973, 0.7831470045544986, 0.7486175937861625, 0.9891526397281045, 0.8517228564544181, 0.8630442177017649, 0.7009876895581848, 0.919993632123095, 0.9742906773769642, 0.5001017712541294, 0.7288059511752198, 0.5356269422575284, 0.5009525538060497, 0.506362747718705, 0.6855898077307808, 0.8186058722239021, 0.6437350504120393, 0.718239019281738, 0.5968857591904404, 0.609524616029788, 0.7346307101193603, 0.5303081011328907, 0.8945951420123923, 0.714368980223252, 0.9270788314109553, 0.7772798233338185, 0.5043792436685017, 0.5489902958837589, 0.8161466086153648, 0.9002587902927479, 0.8594978043993857, 0.7720918657186675, 0.708471889407926, 0.8553134366489652, 0.8307261279699127, 0.9196334979451786, 0.5460130796628213, 0.5322359638794936, 0.7015642375586723, 0.9454702956809771, 0.5614611510499937, 0.9886486033452997, 0.8505700200768735, 0.7705538255754345, 0.5242606557494717, 0.5487130885383176, 0.9797450718881408, 0.7620902625223458, 0.5234501381612409, 0.6015105519508768, 0.6908773949001565, 0.8052201557728094, 0.5619060781574086, 0.8365028605718405, 0.7381675988564596, 0.5518180430531955, 0.7932454856705924, 0.8872510212340574, 0.7280059763374385, 0.978804367491308, 0.9211564036702524, 0.8495645790636963, 0.9434113274257316, 0.9900948520859476, 0.5320469764868729, 0.824062921633453, 0.6392225464923971, 0.7767066000638363, 0.5870722591297224, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0, 50000.0}; int h_B[]= { 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 48, 50, 52, 54, 56, 58, 60, 62, 64, 66, 68, 70, 72, 74, 76, 78, 80, 82, 84, 86, 88, 90, 92, 94, 97, 99, 101, 103, 106, 108, 111, 113, 117, 119, 121, 123, 125, 127, 129, 131, 133, 135, 137, 139, 142, 144, 146, 148, 150, 152, 158, 160, 163, 165, 168, 170, 173, 175, 177, 179, 181, 183, 185, 187, 189, 191, 199, 201, 204, 206, 209, 211, 198, 116, 203, 198, 116, 198, 116, 203, 193, 215, 116, 156, 156, 193, 196, 196, 198, 215, 587, 589, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 768, 770, 772, 774, 776, 778, 780, 782, 784, 786, 788, 790, 792, 794, 796, 798, 800, 802, 804, 806, 808, 810, 812, 814, 816, 818, 820, 822, 824, 826, 828, 830, 832, 834, 836, 838, 840, 842, 844, 846, 848, 850, 852, 854, 856, 858, 860, 862, 864, 865, 866, 867, 868, 869, 870, 871, 872, 873, 874, 875, 876, 877, 878, 879, 880, 881, 882, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 162, 157, 162, 157, 167, 141, 105, 141, 110, 213, 208, 162, 157, 162, 157, 162, 157, 167, 141, 110, 105, 1065, 1072, 203, 213, 208, 162, 157, 141, 110, 105, 213, 208, 213, 208, 1075, 162, 157, 141, 1059, 162, 157, 172, 1065, 208, 162, 157, 141, 1059, 162, 157, 172, 1065, 213, 162, 157, 162, 157, 162, 157, 167, 141, 110, 105, 162, 157, 172, 1065, 213, 208, 162, 157, 162, 157, 167, 141, 105, 141, 110, 213, 208, 162, 157, 162, 157, 167, 141, 1036, 162, 157, 172, 1065, 203, 213, 208, 162, 157, 162, 157, 162, 157, 167, 141, 110, 105, 162, 157, 172, 110, 105, 213, 208, 213, 208, 162, 157, 141, 110, 105, 213, 208, 213, 208, 162, 157, 162, 157, 162, 157, 167, 141, 110, 105, 162, 157, 172, 1065, 213, 208, 162, 157, 162, 157, 162, 157, 167, 141, 1059, 162, 157, 162, 157, 167, 172, 1065, 203, 213, 208, 203, 213, 208, 1086, 1074, 1086, 1074, 1086, 1087, 1086, 1087, 1086, 1087, 1086, 1087, 1086, 1087, 1086, 1079, 1087, 1087, 1087, 1086, 1087, 1086, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 1280, 1281, 1282, 1283, 1284, 1285, 1286, 1287, 1288, 1289, 1290, 1291, 1292, 1293, 1294, 1295, 1296, 1297, 1298, 1299, 1300, 1301, 1303, 1304, 1305, 1306, 1307, 1308, 1309, 1310, 1311, 1312, 1313, 1314, 1316, 1317, 1318, 1319, 1320, 1321, 1322, 1323, 1324, 1325, 1326, 1327, 1328, 1329, 1330, 1331, 1332, 1333, 1334, 1335, 1336, 1337, 1338, 1339, 1340, 1341, 1342, 1343, 1344, 1345, 1346, 1347, 1348, 1349, 1350, 1351, 1352, 1353, 1354, 1355, 1356, 1357, 1358, 1359, 1360, 1361, 1362, 1363, 1364, 1365, 1366, 1367, 1368, 1369, 1370, 1371, 1372, 1373, 1374, 1375, 1376, 1377, 1378, 1379, 1380, 1381, 1382, 1383, 1384, 1385, 1386, 1387, 1388, 1389, 1390, 1391, 1392, 1393, 1394, 1395, 1396, 1397, 1398, 1399, 1400, 1401, 1402, 1403, 1404, 1405, 1406, 1407, 1408, 1409, 1410, 1411, 1412, 1413, 1414, 1415, 1416, 1417, 1418, 1419, 1420, 1421, 1422, 1423, 1424, 1425, 1426, 1427, 1428, 1429, 1430, 1431, 1432, 1433, 1434, 1435, 1436, 1437, 1438, 1439, 1440, 1441, 1442, 1443, 1444, 1445, 1087, 1086, 1087, 1086, 1446, 1447, 1448, 1449, 1450, 1451, 1452, 1453, 1454, 1455, 1456, 1457, 1458, 1459, 1460, 1461, 1462, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 1536, 1538, 1545, 1547, 1549, 1551, 1555, 1559, 1561, 1564, 1566, 1568, 1570, 1574, 1579, 1583, 1588, 1590, 1592, 1596, 1598, 1602, 1604, 1606, 1613, 1615, 1617, 1622, 1627, 1629, 1631, 1633, 1637, 1639, 1642, 1644, 1646, 1648, 1651, 1653, 1655, 1657, 1659, 1661, 1665, 1667, 1671, 1673, 1675, 1677, 1682, 1684, 1690, 1693, 1544, 1542, 1557, 1558, 1692, 1696, 1698, 1681, 1689, 1700, 1701, 1692, 1702, 1703, 1573, 1577, 1692, 1089, 1582, 1586, 1692, 1089, 1601, 1708, 1710, 1612, 1610, 1621, 1625, 1626, 1714, 1670, 1681, 1688, 1689, 1692, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 1540, 1084, 1083, 1846, 1847, 1085, 1553, 1084, 1083, 1798, 1848, 1849, 1085, 1850, 1679, 1084, 1083, 1801, 1080, 1081, 1853, 1854, 1857, 1679, 1084, 1083, 1860, 1686, 1084, 1083, 1861, 1862, 1863, 1679, 1084, 1083, 1864, 1686, 1084, 1083, 1865, 1866, 1867, 1594, 1084, 1083, 1811, 1686, 1084, 1083, 1868, 1085, 1608, 1084, 1083, 1871, 1872, 1085, 1619, 1084, 1083, 1873, 1686, 1084, 1083, 1874, 1875, 1085, 1635, 1084, 1083, 1824, 1686, 1084, 1083, 1826, 1085, 1089, 1679, 1084, 1083, 1830, 1080, 1081, 1663, 1084, 1083, 1836, 1686, 1084, 1083, 1877, 1085, 1679, 1084, 1083, 1878, 1686, 1084, 1083, 1879, 1880, 1085, 1881, 1089, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 2048, 2049, 2050, 2051, 2053, 2054, 2055, 2056, 2057, 2060, 2062, 2063, 2064, 2065, 2066, 2067, 2069, 2070, 2071, 2072, 2073, 2075, 2076, 2077, 2079, 2081, 2082, 2083, 2085, 2086, 2087, 2089, 2091, 2092, 2093, 2094, 2095, 2096, 2097, 2099, 2100, 2101, 2102, 2103, 2105, 2106, 2107, 2108, 2110, 2111, 2112, 2115, 2116, 2117, 2118, 2119, 2120, 2121, 2122, 2123, 2124, 2125, 2126, 2127, 2128, 2129, 2130, 2131, 2132, 2133, 2134, 2135, 2136, 2137, 2138, 2140, 2141, 2142, 2143, 2145, 2146, 2147, 2149, 2150, 2151, 2152, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 2304, 2309, 2314, 2320, 2321, 2322, 2325, 2328, 2329, 2332, 2335, 2336, 2340, 2344, 2349, 2352, 2356, 2360, 2366, 2372, 2376, 2380, 2383, 2386, 2388, 2308, 2389, 2313, 2319, 2318, 2389, 2343, 2348, 2355, 2365, 2364, 2371, 2370, 2389, 2379, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 2560, 2561, 2562, 2565, 2566, 2568, 2569, 2571, 2572, 2573, 2574, 2575, 2576, 2577, 2578, 2579, 2580, 2581, 2582, 2585, 2586, 2587, 2588, 2589, 2371, 2370, 2080, 2090, 2590, 2591, 2592, 2593, 2594, 2595, 2596, 2597, 2371, 2370, 2598, 2599, 2389, 2387, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 2148, 2307, 2058, 2312, 2836, 2148, 2317, 2838, 2148, 2068, 2840, 2841, 2078, 2074, 2842, 2088, 2084, 2843, 2098, 2339, 2844, 2148, 2347, 2113, 2109, 2363, 2359, 2848, 2850, 2148, 2369, 2852, 2853, 2139, 2375, 2854, 2148, 2144, 2856, 2857, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 3072, 3073, 3074, 3075, 3077, 3078, 3080, 3081, 3082, 3084, 3085, 3087, 3088, 3090, 3091, 3093, 3094, 3095, 3096, 3097, 3098, 3101, 3102, 3103, 3105, 3106, 3108, 3109, 3110, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 3328, 3330, 3332, 3334, 3337, 3339, 3341, 3343, 3345, 3347, 3349, 3352, 3354, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 3100, 3356, 2847, 3089, 3092, 2835, 3076, 3336, 2846, 3079, 3086, 3099, 3107, 3351, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 3840, 3841, 3842, 3843, 3844, 3845, 3846, 3847, 3848, 3849, 3850, 3851, 3852, 3853, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 4096, 4098, 4100, 4102, 4104, 4106, 4108, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 4352, 4354, 4356, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 4608, 4610, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 4864, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 1090, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 5376, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255}; int h_C[]= { 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31, 33, 35, 37, 39, 41, 43, 45, 47, 49, 51, 53, 55, 57, 59, 61, 63, 65, 67, 69, 71, 73, 75, 77, 79, 81, 83, 85, 87, 89, 91, 93, 95, 98, 100, 102, 104, 107, 109, 112, 114, 118, 120, 122, 124, 126, 128, 130, 132, 134, 136, 138, 140, 143, 145, 147, 149, 151, 153, 159, 161, 164, 166, 169, 171, 174, 176, 178, 180, 182, 184, 186, 188, 190, 192, 200, 202, 205, 207, 210, 212, 115, 115, 194, 115, 115, 115, 197, 195, 96, 96, 115, 154, 155, 214, 194, 195, 197, 214, 588, 590, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 769, 771, 773, 775, 777, 779, 781, 783, 785, 787, 789, 791, 793, 795, 797, 799, 801, 803, 805, 807, 809, 811, 813, 815, 817, 819, 821, 823, 825, 827, 829, 831, 833, 835, 837, 839, 841, 843, 845, 847, 849, 851, 853, 855, 857, 859, 861, 863, 238, 239, 250, 253, 254, 289, 334, 335, 343, 346, 363, 375, 376, 385, 386, 387, 388, 392, 883, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 1053, 1034, 1061, 1055, 1056, 1030, 1024, 1032, 1033, 1068, 1067, 1053, 1052, 1061, 1054, 1061, 1055, 1056, 1048, 1043, 1025, 1051, 1073, 1066, 1068, 1067, 1053, 1041, 1048, 1043, 1042, 1045, 1044, 1047, 1046, 1076, 1053, 1026, 1057, 1058, 1061, 1060, 1063, 1064, 1070, 1053, 1027, 1057, 1058, 1061, 1060, 1063, 1064, 1071, 1053, 1052, 1061, 1054, 1061, 1055, 1056, 1028, 1050, 1029, 1061, 1060, 1063, 1051, 1068, 1067, 1053, 1034, 1061, 1055, 1056, 1030, 1031, 1032, 1033, 1068, 1067, 1053, 1034, 1061, 1055, 1056, 1035, 1058, 1061, 1060, 1063, 1064, 1066, 1068, 1037, 1053, 1052, 1061, 1054, 1061, 1055, 1056, 1048, 1050, 1038, 1061, 1060, 1063, 1040, 1039, 1068, 1067, 1071, 1070, 1053, 1041, 1048, 1043, 1042, 1045, 1044, 1047, 1046, 1053, 1052, 1061, 1054, 1061, 1055, 1056, 1048, 1050, 1049, 1061, 1060, 1063, 1051, 1068, 1067, 1053, 1052, 1061, 1054, 1061, 1055, 1056, 1057, 1058, 1061, 1060, 1061, 1061, 1062, 1063, 1064, 1066, 1068, 1067, 1069, 1071, 1070, 1078, 1078, 1078, 1078, 1078, 1078, 1078, 1078, 1078, 1077, 1077, 1077, 1077, 1078, 1082, 1078, 1078, 1082, 1088, 1088, 1088, 1088, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 251, 252, 255, 256, 257, 258, 259, 260, 261, 262, 263, 264, 265, 266, 267, 268, 269, 270, 271, 272, 273, 274, 275, 276, 277, 278, 279, 280, 281, 282, 283, 284, 285, 286, 287, 288, 290, 291, 292, 293, 294, 295, 296, 297, 298, 299, 300, 301, 302, 303, 304, 305, 306, 307, 308, 309, 310, 311, 312, 313, 314, 315, 316, 317, 318, 319, 320, 321, 322, 323, 324, 325, 326, 327, 328, 329, 330, 331, 332, 333, 336, 337, 338, 339, 340, 341, 342, 344, 345, 347, 348, 349, 350, 351, 352, 353, 354, 355, 356, 357, 358, 359, 360, 361, 362, 364, 365, 366, 367, 368, 369, 370, 371, 372, 373, 374, 377, 378, 379, 380, 381, 382, 383, 384, 389, 390, 391, 398, 412, 413, 415, 416, 1315, 1315, 1315, 1315, 434, 435, 446, 447, 457, 458, 460, 461, 467, 488, 490, 491, 507, 517, 518, 521, 522, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 1537, 1539, 1546, 1548, 1550, 1552, 1556, 1560, 1562, 1565, 1567, 1569, 1571, 1575, 1580, 1584, 1589, 1591, 1593, 1597, 1599, 1603, 1605, 1607, 1614, 1616, 1618, 1623, 1628, 1630, 1632, 1634, 1638, 1640, 1643, 1645, 1647, 1649, 1652, 1654, 1656, 1658, 1660, 1662, 1666, 1668, 1672, 1674, 1676, 1678, 1683, 1685, 1691, 1694, 1543, 1541, 1687, 1302, 1315, 1697, 1699, 1650, 1315, 420, 421, 1315, 423, 424, 1572, 1576, 1078, 1578, 1581, 1585, 1078, 1587, 1600, 1709, 1711, 1611, 1609, 1620, 1624, 1078, 1715, 1669, 1680, 1687, 1088, 1088, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 1793, 1840, 1792, 396, 397, 1794, 1797, 1796, 1795, 1554, 404, 405, 1799, 407, 1841, 1840, 1800, 1563, 1802, 1803, 418, 419, 422, 1841, 1840, 1804, 428, 1843, 1843, 1805, 432, 433, 436, 1841, 1840, 1806, 440, 1843, 1843, 1807, 444, 445, 448, 1810, 1809, 1808, 1595, 1843, 1843, 1812, 456, 1813, 1815, 1840, 1814, 465, 466, 1816, 1818, 1840, 1817, 472, 1843, 1843, 1819, 476, 477, 1820, 1823, 1822, 1821, 1636, 1843, 1843, 1825, 1641, 1827, 1828, 1841, 1840, 1829, 1650, 1831, 1832, 1835, 1834, 1833, 1664, 1843, 1843, 1837, 505, 1838, 1841, 1840, 1839, 511, 1843, 1843, 1842, 515, 516, 1844, 520, 1845, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 393, 394, 395, 2052, 399, 400, 401, 402, 403, 406, 408, 409, 410, 411, 414, 417, 1855, 1858, 425, 426, 427, 429, 430, 431, 1704, 437, 438, 439, 441, 442, 443, 1706, 449, 450, 451, 452, 453, 454, 455, 459, 462, 463, 464, 2104, 468, 469, 470, 471, 473, 474, 475, 478, 479, 480, 481, 482, 483, 484, 485, 486, 487, 489, 492, 493, 494, 495, 496, 497, 498, 499, 500, 501, 502, 503, 504, 506, 508, 509, 510, 512, 513, 514, 1717, 519, 1719, 523, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 2305, 2310, 2315, 1856, 1859, 2323, 2326, 1705, 2330, 2333, 1707, 2337, 2341, 2345, 2350, 2353, 2357, 2361, 2367, 2373, 2377, 2381, 2384, 1718, 1720, 1695, 2061, 2059, 1852, 1851, 1870, 1869, 1712, 2114, 1713, 1713, 1876, 1876, 1716, 1716, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 2306, 2311, 2316, 2324, 2327, 2331, 2334, 2338, 2342, 2346, 2351, 2354, 2358, 2362, 2368, 2374, 2378, 2382, 2385, 526, 529, 530, 533, 534, 2564, 2563, 2567, 2570, 547, 548, 551, 554, 557, 558, 559, 560, 2584, 2583, 567, 568, 2584, 2583, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 2834, 2816, 2834, 2817, 2837, 2834, 2818, 2839, 2834, 2830, 537, 538, 2820, 2819, 541, 2822, 2821, 544, 2824, 2823, 2845, 2834, 2825, 2827, 2826, 2829, 2828, 2849, 2851, 2834, 2830, 563, 564, 2832, 2831, 2855, 2834, 2833, 571, 572, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 524, 525, 527, 528, 531, 532, 535, 536, 3083, 539, 540, 542, 543, 545, 546, 549, 550, 552, 553, 555, 556, 561, 562, 3104, 565, 566, 569, 570, 3111, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 3329, 3331, 3333, 3335, 3338, 3340, 3342, 3344, 3346, 3348, 3350, 3353, 3355, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 3594, 3596, 3592, 3589, 3590, 3584, 3585, 3587, 3591, 3586, 3588, 3593, 3595, 3594, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 573, 574, 575, 576, 577, 578, 579, 580, 581, 582, 583, 584, 585, 586, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 4097, 4099, 4101, 4103, 4105, 4107, 4109, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 4353, 4355, 4357, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 4609, 4358, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 4865, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 5120, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255, 591, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255}; bool h_Op[]= { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; #define THREADS_PER_BLOCK 256 #define BLOCKS_PER_GRID 1 #define SIZE_OF_IN 768 #define SIZE_OF_AC 5120 __device__ void ac(float *A, const int *B, const int *C, const bool *Op, int n_iter) { int i= blockDim.x * blockIdx.x + threadIdx.x; __shared__ float R[23*THREADS_PER_BLOCK]; const int t= THREADS_PER_BLOCK; __shared__ float final; final=0; R[i + 0*t] = A[i + 0*t]; R[i + 1*t] = A[i + 1*t]; R[i + 2*t] = A[i + 2*t]; __syncthreads(); for (int iter=0; iter< n_iter; iter++) { R[i + 3*t] = Op[i + 0*t] ? R[B[i + 0*t]] * R[C[i + 0*t]] : R[B[i + 0*t]] + R[C[i + 0*t]]; __syncthreads(); R[i + 4*t] = Op[i + 1*t] ? R[B[i + 1*t]] * R[C[i + 1*t]] : R[B[i + 1*t]] + R[C[i + 1*t]]; __syncthreads(); R[i + 5*t] = Op[i + 2*t] ? R[B[i + 2*t]] * R[C[i + 2*t]] : R[B[i + 2*t]] + R[C[i + 2*t]]; __syncthreads(); R[i + 6*t] = Op[i + 3*t] ? R[B[i + 3*t]] * R[C[i + 3*t]] : R[B[i + 3*t]] + R[C[i + 3*t]]; __syncthreads(); R[i + 7*t] = Op[i + 4*t] ? R[B[i + 4*t]] * R[C[i + 4*t]] : R[B[i + 4*t]] + R[C[i + 4*t]]; __syncthreads(); R[i + 8*t] = Op[i + 5*t] ? R[B[i + 5*t]] * R[C[i + 5*t]] : R[B[i + 5*t]] + R[C[i + 5*t]]; __syncthreads(); R[i + 9*t] = Op[i + 6*t] ? R[B[i + 6*t]] * R[C[i + 6*t]] : R[B[i + 6*t]] + R[C[i + 6*t]]; __syncthreads(); R[i + 10*t] = Op[i + 7*t] ? R[B[i + 7*t]] * R[C[i + 7*t]] : R[B[i + 7*t]] + R[C[i + 7*t]]; __syncthreads(); R[i + 11*t] = Op[i + 8*t] ? R[B[i + 8*t]] * R[C[i + 8*t]] : R[B[i + 8*t]] + R[C[i + 8*t]]; __syncthreads(); R[i + 12*t] = Op[i + 9*t] ? R[B[i + 9*t]] * R[C[i + 9*t]] : R[B[i + 9*t]] + R[C[i + 9*t]]; __syncthreads(); R[i + 13*t] = Op[i + 10*t] ? R[B[i + 10*t]] * R[C[i + 10*t]] : R[B[i + 10*t]] + R[C[i + 10*t]]; __syncthreads(); R[i + 14*t] = Op[i + 11*t] ? R[B[i + 11*t]] * R[C[i + 11*t]] : R[B[i + 11*t]] + R[C[i + 11*t]]; __syncthreads(); R[i + 15*t] = Op[i + 12*t] ? R[B[i + 12*t]] * R[C[i + 12*t]] : R[B[i + 12*t]] + R[C[i + 12*t]]; __syncthreads(); R[i + 16*t] = Op[i + 13*t] ? R[B[i + 13*t]] * R[C[i + 13*t]] : R[B[i + 13*t]] + R[C[i + 13*t]]; __syncthreads(); R[i + 17*t] = Op[i + 14*t] ? R[B[i + 14*t]] * R[C[i + 14*t]] : R[B[i + 14*t]] + R[C[i + 14*t]]; __syncthreads(); R[i + 18*t] = Op[i + 15*t] ? R[B[i + 15*t]] * R[C[i + 15*t]] : R[B[i + 15*t]] + R[C[i + 15*t]]; __syncthreads(); R[i + 19*t] = Op[i + 16*t] ? R[B[i + 16*t]] * R[C[i + 16*t]] : R[B[i + 16*t]] + R[C[i + 16*t]]; __syncthreads(); R[i + 20*t] = Op[i + 17*t] ? R[B[i + 17*t]] * R[C[i + 17*t]] : R[B[i + 17*t]] + R[C[i + 17*t]]; __syncthreads(); R[i + 21*t] = Op[i + 18*t] ? R[B[i + 18*t]] * R[C[i + 18*t]] : R[B[i + 18*t]] + R[C[i + 18*t]]; __syncthreads(); R[i + 22*t] = Op[i + 19*t] ? R[B[i + 19*t]] * R[C[i + 19*t]] : R[B[i + 19*t]] + R[C[i + 19*t]]; if (i==0) { final += R[22*t]; } __syncthreads(); } if (i==0) { A[0]= final;} }
19,017
/** * @author Eddie Davis (eddiedavis@u.boisestate.edu) * @author Jeff Pope (jeffreymithoug@u.boisestate.edu) * @file mandelbrot.cu * @brief CS530 PA4: Mandelbrot-CUDA Impementation * @date 12/4/2016 */ #include <stdlib.h> #include <stdio.h> #include <string.h> #include <math.h> #include <limits.h> #include <cuda_runtime.h> #define RMIN 0.3129928802767 #define RMAX 0.31299305009252 #define IMIN 0.0345483210604 #define IMAX 0.0345485012278 #define RADIUS_SQ 4.0 /* 2^2 */ #define DEF_WIDTH 2400 /* Image width in pixels */ #define DEF_HEIGHT 2400 /* Image height in pixels */ #define DEF_BLK_SZ 32 /* BLOCK_SIZE = GCD(WIDTH, THREADS_PER_BLOCK) = GCD(2400, 1024) */ #define MIN_BLK_SZ 1 #define MAX_COLOR UCHAR_MAX /* 255 */ #define OUT_FILE "Mandelbrot.pgm" #define DEF_ITER 1000 #define DEBUG 0 typedef int DTYPE; //typedef unsigned char DTYPE; /** * writeOutput * * Writes Mandelbrot image in PGM format. * * @param fileName Filename to write PGM data. * @param data Output array data (Mandelbrot pixels) * @param width Image width * @param height Image height */ void writeOutput(const char *fileName, DTYPE *data, int width, int height) { int i, j; /* index variables */ int max = -1; /* for pgm file output */ int size = width * height; /* PGM file format requires the largest pixel value, calculate this */ for (i = 0; i < size; ++i) { if (data[i] > max) { max = data[i]; } } /* open the file for writing. omit error checking. */ FILE * fout = fopen(fileName, "w"); /* PGM file header */ fprintf(fout, "P2\n"); fprintf(fout, "%d\t%d\n", width, height); fprintf(fout, "%d\n",max); /* throw out the data */ for (i = 0; i < height; ++i) { for (j = 0; j < width; ++j) { fprintf(fout, "%d\t", data[i * width + j]); } fprintf(fout,"\n"); } /* flush the buffer and close the file */ fflush(fout); fclose(fout); } /** * cudaAssert * * CUDA error handler. * * @param code cudaError_t error code struct. * @param file Name of file in which error occurred. * @param line Line number on which error occurred. */ #define cudaAssert(ans) { _cudaAssert((ans), __FILE__, __LINE__); } inline void _cudaAssert(cudaError_t code, const char *file, int line) { if (code != cudaSuccess) { fprintf(stderr, "cudaAssert: %s %s %d\n", cudaGetErrorString(code), file, line); exit(code); } } /** * cudaPrintDevice * * Prints revelevant information about the given CUDA device. * * @param file File pointer to write device properties. * @param prop cudaDeviceProp structure pointer. * @param dnum CUDA device number. */ void cudaPrintDevice(FILE *file, cudaDeviceProp *prop, int dnum) { fprintf(file, "Device Number: %d\n", dnum); fprintf(file, " Device name: %s\n", prop->name); fprintf(file, " Memory Clock Rate (KHz): %d\n", prop->memoryClockRate); fprintf(file, " Memory Bus Width (bits): %d\n", prop->memoryBusWidth); fprintf(file, " Peak Memory Bandwidth (GB/s): %f\n", 2.0 * prop->memoryClockRate * (prop->memoryBusWidth / 8) / 1.0e6); fprintf(file, " Compute Version: %d.%d\n", prop->major, prop->minor); fprintf(file, " Compute Mode: "); switch (prop->computeMode) { case cudaComputeModeExclusive: fprintf(file, "Exclusive"); break; case cudaComputeModeProhibited: fprintf(file, "Prohibited"); break; default: fprintf(file, "Default"); break; } fprintf(file, "\n"); fprintf(file, " SM count: %d\n", prop->multiProcessorCount); fprintf(file, " Shared mem/block: %zd\n", prop->sharedMemPerBlock); fprintf(file, " Threads per warp: %d\n", prop->warpSize); fprintf(file, " Max threads per block: %d\n", prop->maxThreadsPerBlock); fprintf(file, " Max block size: ("); for (int j = 0; j < 3; j++) { fprintf(file, "%d,", prop->maxThreadsDim[j]); } fprintf(file, ")\n Max grid size: ("); for (int j = 0; j < 3; j++) { fprintf(file, "%d,", prop->maxGridSize[j]); } fprintf(file, ")\n\n"); } /** * mand (CUDA kernel function) * * Generates the Mandelbrot set. * * @param output Output array to receive computed Mandelbrot pixels. * @param maxIter Max iterations to test for escape values. * @param width Image width. * @param height Image height. * @param realRange Range of real component. * @param imagRange Range of imaginary component. */ __global__ void mand(DTYPE* output, int maxIter, int width, int height, double realRange, double imagRange) { int col = blockDim.x * blockIdx.x + threadIdx.x; // Image col (X coord) int row = blockDim.y * blockIdx.y + threadIdx.y; // Image row (Y coord) if (col < width && row < height) { int idx = row * width + col; double cReal = RMIN + row * realRange; double cImag = IMIN + col * imagRange; double zReal = 0.0; double zImag = 0.0; double zReal2 = zReal; double zImag2 = zImag; double zCurr; double zMag; int iter = 0; for (; iter < maxIter; ++iter) { zCurr = zReal; zReal2 = zReal * zReal; zImag2 = zImag * zImag; zReal = zReal2 - zImag2 + cReal; zImag = (2.0 * zCurr * zImag) + cImag; zMag = zReal2 + zImag2; if (zMag > RADIUS_SQ) { break; } } output[idx] = (DTYPE) floor(((double) (MAX_COLOR * iter)) / (double) maxIter); } } /** * main * * Main function. * * @param argc Argument count. * @param argv Argument values. * @return */ int main(int argc, char ** argv) { int nDevices = 0; DTYPE *output = NULL; DTYPE *d_output = NULL; float time; /*timer*/ int maxIter = DEF_ITER; if (argc > 1) { maxIter = atoi(argv[1]); /* first command line argument... */ } if (maxIter < 1) { printf("usage: %s [MAX_ITERATION=%d] [WIDTH=%d] [HEIGHT=%d] [BLOCK_X=%d] [BLOCK_Y=1]\n", argv[0], DEF_ITER, DEF_WIDTH, DEF_HEIGHT, MIN_BLK_SZ); return 0; } int width = DEF_WIDTH; if (argc > 2) { width = atoi(argv[2]); } int height = DEF_HEIGHT; if (argc > 3) { height = atoi(argv[3]); } cudaAssert(cudaGetDeviceCount(&nDevices)); if (nDevices < 1) { printf("ERROR: No valid CUDA devices on this machine!\n"); return -1; } if (DEBUG) { fprintf(stderr, "nDevices = %d\n", nDevices); cudaDeviceProp prop; for (int i = 0; i < nDevices; i++) { cudaAssert(cudaGetDeviceProperties(&prop, i)); cudaPrintDevice(stderr, &prop, i); } } // Get data size... int nPixels = width * height; int nBytes = nPixels * sizeof(DTYPE); if (DEBUG) fprintf(stderr, "nPixels = %d, nBytes = %d\n", nPixels, nBytes); /* Allocate memory on host to store output values for pixels */ output = (DTYPE *) malloc(nBytes); if (output == NULL) { perror("output"); return -1; } // Set block size... int blockX = 0; if (argc > 4) { blockX = atoi(argv[4]); } if (blockX < 1) { blockX = MIN_BLK_SZ; } int blockY = 0; if (argc > 5) { blockY = atoi(argv[5]); } if (blockY < 1) { blockY = MIN_BLK_SZ; } if (blockX == MIN_BLK_SZ && blockY == MIN_BLK_SZ) { blockX = DEF_BLK_SZ; } dim3 blockSize(blockX, blockY); if (DEBUG) fprintf(stderr, "blockSize = (%d,%d,%d)\n", blockSize.x, blockSize.y, blockSize.z); // Set grid size... int gridX = width / blockSize.x; int gridY = height / blockSize.y; dim3 gridSize(gridX, gridY); if (DEBUG) fprintf(stderr, "gridSize = (%d,%d,%d)\n", gridSize.x, gridSize.y, gridSize.z); printf("Running Mandelbrot-CUDA with (w,h,mi,bx,by,gx,gy) = (%d,%d,%d,%d,%d,%d,%d)...\n", width, height, maxIter, blockSize.x, blockSize.y, gridSize.x, gridSize.y); // Create event timers... cudaEvent_t start, stop; cudaAssert(cudaEventCreate(&start)); cudaAssert(cudaEventCreate(&stop)); // Start timer... cudaAssert(cudaEventRecord(start)); // Allocate memory on device... if (DEBUG) fprintf(stderr, "cudaMalloc...\n"); cudaAssert(cudaMalloc(&d_output, nBytes)); double realRange = (RMAX - RMIN) / (double) (width - 1); double imagRange = (IMAX - IMIN) / (double) (height - 1); // Invoke the kernel... if (DEBUG) { fprintf(stderr, "kernel: mand(d_output[%d], maxIter=%d, realRange=%lf, imagRange=%lf)...\n", nPixels, maxIter, realRange, imagRange); } mand<<<gridSize, blockSize>>>(d_output, maxIter, width, height, realRange, imagRange); // cudaMemcpy is an implicit barrier so need need for sync. // Copy data back to host... if (DEBUG) fprintf(stderr, "cudaMemcpy...\n"); cudaAssert(cudaMemcpy(output, d_output, nBytes, cudaMemcpyDeviceToHost)); // Free data on device... if (DEBUG) fprintf(stderr, "cudaFree...\n"); cudaAssert(cudaFree(d_output)); // Stop timer... cudaAssert(cudaEventRecord(stop)); // Get elapsed time... if (DEBUG) fprintf(stderr, "cudaEventSynchronize...\n"); cudaAssert(cudaEventSynchronize(stop)); if (DEBUG) fprintf(stderr, "cudaEventElapsedTime...\n"); cudaAssert(cudaEventElapsedTime(&time, start, stop)); // Write the output... if (DEBUG) fprintf(stderr, "writeOutput...\n"); writeOutput(OUT_FILE, output, width, height); // Free host data... free(output); // Report timing... printf("Elapsed time: %lf sec\n", time * 1E-3); return 0; }
19,018
#include <cstdio> #include "kernel.cuh" __global__ void matmul_basic(const float* A, const float* B, float* C, const int len) { int i = blockIdx.y * blockDim.y + threadIdx.y; int j = blockIdx.x * blockDim.x + threadIdx.x; if (i<len&&j<len) { float sum=0; for (int k=0; k<len; k++) { sum += A[i*len+k]*B[k*len+j]; } C[i*len+j]=sum; } } __global__ void matmul_tiled(const float* A, const float* B, float* C, const int len, const int len_tile) { extern __shared__ float smem[]; float *sA = &smem[0]; float *sB = &smem[len_tile*len_tile]; int i = blockIdx.y * blockDim.y + threadIdx.y; int si = threadIdx.y; int j = blockIdx.x * blockDim.x + threadIdx.x; int sj = threadIdx.x; int num_tiles = (len+(len_tile-1))/len_tile; float sum = 0; for (int t=0; t<num_tiles; t++) { if (i<len && (sj+t*len_tile) < len) sA[si*len_tile+sj] = A[(i)*len+(sj+t*len_tile)]; else sA[si*len_tile+sj] = 0; if (j<len && (si+t*len_tile) < len) sB[si*len_tile+sj] = B[(si+t*len_tile)*len+(j)]; else sB[si*len_tile+sj] = 0; __syncthreads(); for (int k=0; k<len_tile; k++) { sum += sA[(si*len_tile)+k]*sB[(k*len_tile)+sj]; } __syncthreads(); } if (i<len && j<len) { C[i*len+j] = sum; } } __global__ void transpose(const float* A, float* A_T, const int len) { int i = blockIdx.y * blockDim.y + threadIdx.y; int j = blockIdx.x * blockDim.x + threadIdx.x; if (j<len && i<len) { A_T[i*len+j] = A[j*len+i]; } } __global__ void matmul_tiled_transposed(const float* A, const float* B, float* C, const int len, const int len_tile) { extern __shared__ float smem[]; float *sA = &smem[0]; float *sB = &smem[len_tile*len_tile]; int i = blockIdx.y * blockDim.y + threadIdx.y; int si = threadIdx.y; int j = blockIdx.x * blockDim.x + threadIdx.x; int sj = threadIdx.x; int num_tiles = (len+(len_tile-1))/len_tile; int itosj = ((i/len_tile)*len_tile) + sj; float sum = 0; for (int t=0; t<num_tiles; t++) { if (itosj<len && (si+t*len_tile) < len) sA[si*len_tile+sj] = A[(si+t*len_tile)*len+(itosj)]; else sA[si*len_tile+sj] = 0; if (j<len && (si+t*len_tile) < len) sB[si*len_tile+sj] = B[(si+t*len_tile)*len+(j)]; else sB[si*len_tile+sj] = 0; __syncthreads(); for (int k=0; k<len_tile; k++) { sum += sA[(k*len_tile)+si]*sB[(k*len_tile)+sj]; } __syncthreads(); } if (i<len && j<len) { C[i*len+j] = sum; } }
19,019
#include <iostream> #include <thrust/device_vector.h> #include <thrust/execution_policy.h> #include <thrust/host_vector.h> #include <thrust/scan.h> using namespace std; int main(int argc, const char *argv[]) { string N; if (argc > 1) { N = string(argv[1]); } unsigned int n = atoi(N.c_str()); thrust::host_vector<float> H(n); for (unsigned int i = 0; i < n; i++) { H[i] = 1; } thrust::device_vector<float> D = H; thrust::device_vector<float> res(n); cudaEvent_t start; cudaEvent_t stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start); thrust::exclusive_scan(D.begin(), D.end(), res.begin()); cudaEventRecord(stop); cudaEventSynchronize(stop); // Get the elapsed time in milliseconds float ms; cudaEventElapsedTime(&ms, start, stop); cout << res[n - 1] << endl; cout << ms << endl; return 0; }
19,020
#include <stdio.h> #include <cuda.h> #include <cuda_runtime.h> #include <cuda_runtime_api.h> #include <algorithm> #include <cstring> //memset #define MAX_INT_BITS 32 #define checkCudaErrors(call) \ { \ const cudaError_t error = call; \ if (error != cudaSuccess) \ { \ fprintf(stderr, "Error: %s:%d, ", __FILE__, __LINE__); \ fprintf(stderr, "code: %d, reason: %s\n", error, \ cudaGetErrorString(error)); \ exit(1); \ } \ } void printArray(unsigned int* const h_input, const size_t numElems) { for (unsigned int i = 0; i < numElems; ++i) printf("%d ", h_input[i]); printf("\n"); } __device__ void dev_printArray(unsigned int* const d_input, const size_t numElems) { for (unsigned int i = 0; i < numElems; ++i) printf("%d \t", d_input[i]); printf("\n"); } __device__ unsigned int d_count = 0; __device__ unsigned int d_curBlock = 0; __global__ void set_init_scan() { d_count = 0; d_curBlock = 0; } __global__ void exclusive_scan(const unsigned int* const d_in, unsigned int* const d_out, unsigned int* const d_prevSums, const int n) { extern __shared__ unsigned int s_blkIn[]; //blockSize __shared__ unsigned int s_prevSum; __shared__ unsigned int s_bid; __shared__ unsigned int s_endBlock; // get block idx if (0 == threadIdx.x) { s_bid = atomicAdd(&d_count, 1); s_endBlock = blockDim.x - 1; } __syncthreads(); int i = s_bid * blockDim.x + threadIdx.x; if (i == n - 1) s_endBlock = threadIdx.x; if (i >= n) return; // load data if (0 == i) s_blkIn[0] = 0; else s_blkIn[threadIdx.x] = d_in[i - 1]; __syncthreads(); /// reduction for (int stride = 1; stride < blockDim.x; stride *= 2) { int localThreadIdx = 2 * (threadIdx.x + 1) * stride - 1; if (localThreadIdx < blockDim.x) { s_blkIn[localThreadIdx] += s_blkIn[localThreadIdx - stride]; } __syncthreads(); } //post reduction for (int stride = blockDim.x / 4; stride > 0; stride /= 2) { int localThreadIdx = 2 * (threadIdx.x + 1) * stride - 1; if (localThreadIdx + stride < blockDim.x) { s_blkIn[localThreadIdx + stride] += s_blkIn[localThreadIdx]; } __syncthreads(); } if (0 == threadIdx.x) { while(atomicAdd(&d_curBlock, 0) < s_bid); s_prevSum = d_prevSums[s_bid]; d_prevSums[s_bid + 1] = s_prevSum + s_blkIn[s_endBlock]; __threadfence(); atomicAdd(&d_curBlock, 1); } __syncthreads(); // update output d_out[i] = s_blkIn[threadIdx.x] + s_prevSum; } unsigned int* host_exclusive_scan(const unsigned int* const d_in, const size_t numElems, const dim3 blockSize) { const dim3 gridSize((numElems - 1) / blockSize.x + 1); set_init_scan<<<1, 1>>>(); unsigned int* d_out; checkCudaErrors(cudaMalloc(&d_out, sizeof(unsigned int) * numElems)); unsigned int* d_prevSums; checkCudaErrors(cudaMalloc(&d_prevSums, sizeof(unsigned int) * gridSize.x)); checkCudaErrors(cudaMemset(d_prevSums, 0, sizeof(int) * gridSize.x)); int sharedSize = blockSize.x * sizeof(unsigned int) + 2 * sizeof (unsigned int) + sizeof(int); exclusive_scan<<<gridSize, blockSize, sharedSize>>>(d_in, d_out, d_prevSums, numElems); cudaDeviceSynchronize(); checkCudaErrors(cudaGetLastError()); checkCudaErrors(cudaFree(d_prevSums)); return d_out; } __global__ void histogram(const unsigned int* const d_inputVals, const size_t numElems, const size_t numBins, unsigned int* const d_hist, const unsigned int mask, const unsigned int digitOrder) { extern __shared__ unsigned int s_hist[]; // numBins for (unsigned int bin = threadIdx.x; bin < numBins; bin += blockDim.x) { s_hist[bin] = 0; } __syncthreads(); int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx < numElems) { unsigned int bin = (d_inputVals[idx] & mask) >> digitOrder; atomicAdd(&s_hist[bin], 1); } __syncthreads(); //update d_hist by s_hist for (unsigned int bin = threadIdx.x; bin < numBins; bin += blockDim.x) { int idxHist = bin * gridDim.x + blockIdx.x; d_hist[idxHist] = s_hist[bin]; // printf("bin - indexHist - val : %d - %d - %d - %d - %d\n", blockIdx.x, bin, idxHist, d_hist[idxHist]); } } unsigned int* host_histogram(const unsigned int* const d_inputVals, const size_t numElems, const size_t numBins, const unsigned int mask, const unsigned int digitOrder, const dim3 blockSize) { const dim3 gridSize((numElems - 1) / blockSize.x + 1); unsigned int *d_hist; checkCudaErrors(cudaMalloc(&d_hist, sizeof(unsigned int) * numBins * gridSize.x)); checkCudaErrors(cudaMemset(d_hist, 0, sizeof(unsigned int) * numBins * gridSize.x)); int sharedSize = numBins * sizeof(unsigned int); histogram<<<gridSize, blockSize, sharedSize>>>(d_inputVals, numElems, numBins, d_hist, mask, digitOrder); cudaDeviceSynchronize(); checkCudaErrors(cudaGetLastError()); return d_hist; } __global__ void scatter(unsigned int* const d_inputVals, unsigned int* const d_outputVals, const unsigned int* const d_histScan, const size_t numElems, const size_t numBins, const unsigned int mask, const unsigned int digitOrder) { extern __shared__ unsigned int s_blkIn[]; // blockSize unsigned int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx >= numElems) return; //load data s_blkIn[threadIdx.x] = d_inputVals[idx]; __syncthreads(); //get number of elements ( < this element and = this element but locate on prev block unsigned int bin = (s_blkIn[threadIdx.x] & mask) >> digitOrder; unsigned int numPrevElems = d_histScan[bin * gridDim.x + blockIdx.x]; //calc number of element = this element on this block unsigned int count = 0; for (int i = 0; i < threadIdx.x; ++i) { unsigned int prevBin = (s_blkIn[i] & mask) >> digitOrder; if (bin == prevBin) ++count; } // scatter to result array unsigned int rank = numPrevElems + count; d_outputVals[rank] = s_blkIn[threadIdx.x]; // printf("block - idx - rank - val: %d - %d - %d - %d\n", blockIdx.x, idx, rank, s_blkIn[threadIdx.x]); } void host_scatter(unsigned int* const d_inputVals, unsigned int* const d_outputVals, const size_t numElems, const size_t numBins, const unsigned int* const d_histScan, const unsigned int mask, const unsigned int digitOrder, const dim3 blockSize) { const dim3 gridSize((numElems - 1) / blockSize.x + 1); unsigned int sharedSize = blockSize.x * sizeof(unsigned int); scatter<<<gridSize, blockSize, sharedSize>>> (d_inputVals, d_outputVals, d_histScan, numElems,numBins, mask, digitOrder); } unsigned int getMaxNumOfBits(unsigned int* h_input, const size_t numElems) { unsigned int maxElem = *std::max_element(h_input, h_input + numElems); unsigned int mask = 1 << 31; unsigned int count = 0; while (! (mask & maxElem)) { ++count; mask >>= 1; } return MAX_INT_BITS - count; } void radix_sort(unsigned int* h_inputVals, unsigned int* d_inputVals, unsigned int* d_outputVals, const size_t numElems, const size_t numBits, const dim3 blockSize) { const dim3 gridSize((numElems - 1) / blockSize.x + 1); size_t numBins = 1 << numBits; unsigned int* pInVals = d_inputVals; unsigned int* pOutVals = d_outputVals; unsigned int maxBits = getMaxNumOfBits(h_inputVals, numElems); if (maxBits % numBits) maxBits += numBits; // loop through digits for (unsigned int i = 0; i <= maxBits; i += numBits) { unsigned int mask = (numBins - 1) << i; // printf("mask: %d\n", mask); //histogram unsigned int* d_hist = host_histogram(pInVals, numElems, numBins, mask, i, blockSize); // unsigned int* h_hist = (unsigned int*) malloc (sizeof(unsigned int) * numBins * gridSize.x); // checkCudaErrors(cudaMemcpy(h_hist, d_hist, sizeof(unsigned int) * numBins * gridSize.x, cudaMemcpyDeviceToHost)); // printArray(h_hist, numBins * gridSize.x); // free(h_hist); // exclusive scan hist unsigned int* d_histScan = host_exclusive_scan(d_hist, numBins * gridSize.x, blockSize); // unsigned int* h_histScan = (unsigned int*) malloc (sizeof(unsigned int) * numBins * gridSize.x); // checkCudaErrors(cudaMemcpy(h_histScan, d_histScan, sizeof(unsigned int) * numBins * gridSize.x, cudaMemcpyDeviceToHost)); // printArray(h_histScan, numBins * gridSize.x); // free(h_histScan); //scatter host_scatter(pInVals, pOutVals, numElems, numBins, d_histScan, mask, i, blockSize); std::swap(pInVals, pOutVals); // unsigned int* h_result = (unsigned int*) malloc (sizeof(unsigned int) * numElems); // checkCudaErrors(cudaMemcpy(h_result, d_inputVals, sizeof(unsigned int) * numElems, cudaMemcpyDeviceToHost)); // printArray(h_result, numElems); // free(h_result); // printf("\n\n\n"); checkCudaErrors(cudaFree(d_hist)); checkCudaErrors(cudaFree(d_histScan)); d_hist = NULL; d_histScan = NULL; } if (pInVals == d_outputVals) { checkCudaErrors(cudaMemcpy(d_outputVals, d_inputVals, sizeof(unsigned int) * numElems, cudaMemcpyDeviceToDevice)); } } int main() { // srand(time(NULL)); const dim3 blockSize(64); const size_t numElems = 10000000; const unsigned int numBits = 2; unsigned int* h_inputVals = (unsigned int*) malloc(sizeof(unsigned int) * numElems); for (int i = 0; i < numElems; ++i) { h_inputVals[i] = rand() % 1000000000 + 1; } // printArray(h_inputVals, numElems); unsigned int* d_inputVals; checkCudaErrors(cudaMalloc(&d_inputVals, sizeof(unsigned int) * numElems)); checkCudaErrors(cudaMemcpy(d_inputVals, h_inputVals, sizeof(unsigned int) * numElems, cudaMemcpyHostToDevice)); unsigned int* d_outputVals; checkCudaErrors(cudaMalloc(&d_outputVals, sizeof(unsigned int) * numElems)); radix_sort(h_inputVals, d_inputVals, d_outputVals, numElems, numBits, blockSize); unsigned int* h_outputVals = (unsigned int*) malloc(sizeof(unsigned int) * numElems); checkCudaErrors(cudaMemcpy(h_outputVals, d_outputVals, sizeof(unsigned int) * numElems, cudaMemcpyDeviceToHost)); // printArray(h_outputVals, numElems); bool check = true; for (int i = 1; i < numElems; ++i) { if (h_outputVals[i] < h_outputVals[i - 1]) { printf("\nfalse at index : %d\n", i); check = false; break; } } if (check) printf("\nTRUE\n"); else printf("\nFALSE\n"); free(h_inputVals); free(h_outputVals); checkCudaErrors(cudaFree(d_inputVals)); checkCudaErrors(cudaFree(d_outputVals)); return 0; }
19,021
/* #include "patchBasedObject.cuh" template <typename T> __host__ void PatchBasedObject<T>::generateSuperpixels(uint2 & pbbsize, uint2 & stride) { int spx_sz = 0; // initialize the superpixel size float noLabels = 2; // control number of superpixels using-> (int)(noLabels * sqrt( [width * height] /2 )) ; double compactness = 10; // It was used to control the superpixel shape [compactness factor], but now it is redundant as in SLICO it is defined automatically per each superpixel Superpixels<T> superpixesl; superpixesl.runStackSLIC(m_h_stack, m_h_spx_stack, spx_sz, noLabels, compactness); pbbsize = make_uint2(3 * spx_sz, 3 * spx_sz); // In order to fix the patch size for all the superpixels, we define patch size 3 times the uniform superpixel size. The extra free space is used later for dilation stride = make_uint2(0.25*spx_sz, 0.25*spx_sz); // define dilation iterations = 0.25 * superpixel_size cout << "Patch size: " << pbbsize.x << "x" << pbbsize.y << endl; cout << "Dilation iterations: " << stride.x << endl; } template<> class PatchBasedObject < float >; template<> class PatchBasedObject < double >; */
19,022
// a cuda app. we will convert this to opencl, and run it :-) #include <iostream> #include <memory> #include <cassert> using namespace std; #include <cuda_runtime.h> __global__ void setValue(char *data, int idx, char value) { if(threadIdx.x == 0) { data[idx] = value; // data[idx] = 15; } } int main(int argc, char *argv[]) { int N = 1024; char *gpuChars; cudaMalloc((void**)(&gpuChars), N * sizeof(char)); setValue<<<dim3(32, 1, 1), dim3(32, 1, 1)>>>(gpuChars, 2, (char)89); char hostChars[4]; cudaMemcpy(hostChars, gpuChars, 4 * sizeof(char), cudaMemcpyDeviceToHost); cout << "hostChars[2] " << (int)hostChars[2] << endl; assert(hostChars[2] == 89); setValue<<<dim3(32, 1, 1), dim3(32, 1, 1)>>>(gpuChars, 2, (char)23); cudaMemcpy(hostChars, gpuChars, 4 * sizeof(char), cudaMemcpyDeviceToHost); cout << "hostChars[2] " << (int)hostChars[2] << endl; assert(hostChars[2] == 23); hostChars[2] = 44; cudaMemcpy(gpuChars, hostChars, 4 * sizeof(char), cudaMemcpyHostToDevice); hostChars[2] = 55; cudaMemcpy(hostChars, gpuChars, 4 * sizeof(char), cudaMemcpyDeviceToHost); cout << "hostChars[2] " << (int)hostChars[2] << endl; assert(hostChars[2] == 44); cudaFree(gpuChars); return 0; }
19,023
#include "includes.h" __global__ void differenceImg(float *d_Octave0,float *d_Octave1,float *d_diffOctave,int pitch,int height){ int x = blockIdx.x*blockDim.x+threadIdx.x; int y = blockIdx.y*blockDim.y+threadIdx.y; int index = y * pitch + x; if(y<height) d_diffOctave[index] = (d_Octave1[index] - d_Octave0[index]); }
19,024
#include "includes.h" #define BLOCK_SIZE 16 #define BLOCKSIZE_X 16 #define BLOCKSIZE_Y 16 // STD includes // CUDA runtime // Utilities and system includes static // Print device properties __global__ void writeChannelKernel( unsigned char* image, unsigned char* channel, int imageW, int imageH, int channelToMerge, int numChannels) { int x = blockIdx.x * blockDim.x + threadIdx.x; int y = blockIdx.y * blockDim.y + threadIdx.y; int posOut = y * (imageW*numChannels) + (x*numChannels) + channelToMerge; int posIn = y * imageW + x; image[posOut] = channel[posIn]; }
19,025
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include<stdio.h> __global__ void unique_grid_id_calculation_2d_2d(int* data) { int thread_id = blockDim.x * threadIdx.y + threadIdx.x; int num_threads_in_a_block = blockDim.x * blockDim.y; int block_offset = blockIdx.x * num_threads_in_a_block; int num_threads_in_a_row = num_threads_in_a_block * gridDim.x; int row_offset = num_threads_in_a_row * blockIdx.y; int grid_id = thread_id + row_offset + block_offset; printf("blockIdx.x: %d, blockIdx.y: %d, threadIdx.x: %d, grid ID: %d, - data : %d \n", blockIdx.x, blockIdx.y, thread_id, grid_id, data[grid_id]); } /* int main() { int array_size = 16; int array_byte_size = sizeof(int) * array_size; int h_data[] = { 23, 9, 4, 53, 65, 12, 1, 33, 87, 45, 23, 12, 342, 56, 44, 99 }; int* d_data; cudaMalloc((void**)&d_data, array_byte_size); cudaMemcpy(d_data, h_data, array_byte_size, cudaMemcpyHostToDevice); dim3 block(2, 2); dim3 grid(2, 2); unique_grid_id_calculation_2d_2d<<<grid, block>>>(d_data); cudaDeviceSynchronize(); cudaDeviceReset(); return 0; } */
19,026
#include "includes.h" __global__ void softmax_trivial(float* softmaxP, float* b, int rows, int cols){ int tid = threadIdx.x; int bid = blockIdx.x; float _max = -100000000.0; float sum = 0.0; if(tid * cols + bid < rows * cols){ for(int i = 0 ; i < rows ; i++) _max = max(_max, b[i * cols + bid]); for(int i = 0 ; i < rows ; i++) softmaxP[i * cols + bid] = (b[i * cols + bid] - _max); for(int i = 0 ; i < rows ; i++) softmaxP[i * cols + bid] = __expf(softmaxP[i * cols + bid]); for(int i = 0 ; i < rows ; i++) sum += softmaxP[i * cols + bid]; for(int i = 0 ; i < rows ; i++) softmaxP[i * cols + bid] /= sum; } }
19,027
#include "includes.h" __global__ void LessThan(float * xf, bool * xb, size_t idxf, size_t idxb, size_t N) { for (int i = blockIdx.x * blockDim.x + threadIdx.x; i < N; i += blockDim.x * gridDim.x) { //printf("From less than %f %f %d \n", xf[(idxf-1)*N+i], xf[(idxf-2)*N+i], xf[(idxf-1)*N+i] < xf[(idxf-2)*N+i]); xb[idxb*N+i] = (xf[(idxf-1)*N+i] < xf[(idxf-2)*N+i]); } return; }
19,028
#include <stdio.h> #include <stdlib.h> #include <sys/time.h> #include <unistd.h> #include <string.h> #define DATA_LENGTH 100 #define CUDA_CALL(X) X; // {if(cudaError == X){printf("Error Calling %s at line %s\n", #X, __LINE__);}} float * genInput(int l); void verify(float *a, float *b, float *c, int l); __global__ void vecAdd(float *in1, float *in2, float *out, int len) { int i=0; int tid = threadIdx.x; for(i=0; i<DATA_LENGTH; i++){ // first half of first warp and second half of 2nd warp if(tid < 16 || tid > 47) out[i] = in1[i]+in2[i]; else out[i] = in1[i]+in2[i]; } for(i=0; i<DATA_LENGTH; i++){ // only even threads not compactable if(threadIdx.x%2 == 0 ) out[i] = in1[i]+in2[i]; else out[i] = in1[i]+in2[i]; } out[0] = in1[0] + in2[0]; } int main(int argc, char **argv) { int inputLength; float *hostInput1; float *hostInput2; float *hostOutput; float *deviceInput1; float *deviceInput2; float *deviceOutput; struct timeval t; gettimeofday(&t, NULL); srand(t.tv_sec); inputLength = DATA_LENGTH; hostInput1 = genInput(inputLength); hostInput2 = genInput(inputLength); hostOutput = ( float * )malloc(inputLength * sizeof(float)); //@@ Allocate GPU memory here CUDA_CALL(cudaMalloc((void**)&deviceInput1, inputLength*sizeof(float))); CUDA_CALL(cudaMalloc((void**)&deviceInput2, inputLength*sizeof(float))); CUDA_CALL(cudaMalloc((void**)&deviceOutput, inputLength*sizeof(float))); //@@ Copy memory to the GPU here CUDA_CALL(cudaMemcpy(deviceInput1, hostInput1, sizeof(float)*inputLength, cudaMemcpyHostToDevice)); CUDA_CALL(cudaMemcpy(deviceInput2, hostInput2, sizeof(float)*inputLength, cudaMemcpyHostToDevice)); //@@ Initialize the grid and block dimensions here dim3 numBlocks(1,1,1); //dim3 numThreads(ThreadsPerBlock,1,1); dim3 numThreads(64,1,1); //@@ Launch the GPU Kernel here vecAdd<<<numBlocks, numThreads>>>(deviceInput1, deviceInput2, deviceOutput, inputLength); cudaDeviceSynchronize(); //@@ Copy the GPU memory back to the CPU here CUDA_CALL(cudaMemcpy(hostOutput, deviceOutput, inputLength*sizeof(float), cudaMemcpyDeviceToHost)); //@@ Free the GPU memory here CUDA_CALL(cudaFree(deviceInput1)); CUDA_CALL(cudaFree(deviceInput2)); CUDA_CALL(cudaFree(deviceOutput)); verify(hostInput1, hostInput2, hostOutput, inputLength); free(hostInput1); free(hostInput2); free(hostOutput); return 0; } float * genInput(int l) { int i; float * arr = (float*)malloc(l*sizeof(float)); for(i=0; i<l; i++){ arr[i] = rand(); arr[i] = arr[i]/rand(); } return arr; } void verify(float *a, float *b, float *c, int l) { char buff1[50] = {0}; char buff2[50] = {0}; int i; for(i=0; i<l; i++){ float d = a[i]+b[i]; sprintf(buff1, "%1.8f", d); sprintf(buff2, "%1.8f", c[i]); if(strcmp(buff1, buff2) != 0){ printf("ERROR at index %d, Exp %1.8f Got %1.8f\n",i,d,c[i]); break; } } }
19,029
#include <stdio.h> #include <stdlib.h> #include <math.h> #include <cuda.h> // CUDA kernel. Cada thread ejecuta la operación sobrte un elemencto de c __global__ void vecAdd(double *a, double *b, double *c, int n) { // Obtención del Id global int id = blockIdx.x*blockDim.x+threadIdx.x; // Nos aseguramos de no salir de los bordes if (id < n) c[id] = a[id] + b[id]; printf("I am thread (%d, %d) in block (%d, %d). Global thread ID = %d\n",threadIdx.y, threadIdx.x, blockIdx.y, blockIdx.x); } __host__ int main( int argc, char* argv[] ) { // Tamaño de los vectores int n = 100000; // Vectores de entrada al host (CPU) double *h_a; double *h_b; // Vector de salida del host double *h_c; // Vector de entrada del device (GPU) double *d_a; double *d_b; // Vector de salida del device double *d_c; // Size, in bytes, of each vector //size_t bytes = n*sizeof(double); // Se asigna memoria para cada vector del host h_a = (double*)malloc(n*sizeof(double)); h_b = (double*)malloc(n*sizeof(double)); h_c = (double*)malloc(n*sizeof(double)); // Se asigna memoria para cada vector del device cudaMalloc(&d_a, n*sizeof(double)); cudaMalloc(&d_b, n*sizeof(double)); cudaMalloc(&d_c, n*sizeof(double)); int i; // Se inicializa los vectores del host for( i = 0; i < n; i++ ) { h_a[i] = sin(i)*sin(i); h_b[i] = cos(i)*cos(i); } // Se copia el vector del host al vector del device cudaMemcpy( d_a, h_a, n*sizeof(double), cudaMemcpyHostToDevice); cudaMemcpy( d_b, h_b, n*sizeof(double), cudaMemcpyHostToDevice); int blockSize, gridSize; // Número de threads en cada bloque blockSize = 1024; // Número de bloques en la grilla gridSize = (int)ceil((float)n/blockSize); printf("%i\n", gridSize); // Se ejecuta el kernel vecAdd<<<gridSize, blockSize>>>(d_a, d_b, d_c, n); // Se copia el vector resultante del device al host cudaMemcpy( h_c, d_c, n*sizeof(double), cudaMemcpyDeviceToHost ); // Sum up vector c and print result divided by n, this should equal 1 within error double sum = 0; for(i=0; i<n; i++) sum += h_c[i]; printf("final : %f\n", sum/n); // Se libera la memoria del device cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); // Se libera la memoria del host free(h_a); free(h_b); free(h_c); return 0; }
19,030
/* * CUDA kernel for convolution in 2D, corresponding to conv2 in Matlab * Sofie Lovdal 5.6.2018 */ __global__ void conv2(double * output, double * const input, unsigned int const numRows, unsigned int const numCols, double * const kernel, unsigned int const height_kernel, unsigned int const width_kernel) { /*global thread ID in x dimension - moving horizontally in the image*/ const int colIdx = blockIdx.x*blockDim.x + threadIdx.x; /*global thread ID in y dimension - moving vertically in the image*/ const int rowIdx = blockIdx.y*blockDim.y + threadIdx.y; int i, j, kernelIdx, imageIdx; /*Effectivize: Load local image area into shared memory space*/ /*make sure we are within image*/ if(colIdx>=numCols || rowIdx >= numRows) return; /*Linear index of pixel corresponding to current thread */ int linearIdx = rowIdx*numCols + colIdx; int kernel_radius=height_kernel/2; int imageRowIdx, imageColIdx; /*Apply convolution to linarIdx (pixel that each thread should treat) */ double sum=0.0; for (i = -kernel_radius; i <= kernel_radius; i++) { for (j = -kernel_radius; j <= kernel_radius; j++) { kernelIdx = width_kernel*(i+kernel_radius) + (j+kernel_radius); imageRowIdx = rowIdx+i; imageColIdx = colIdx+j; imageIdx = imageRowIdx*numCols + imageColIdx; /*zero padding at borders: top, bottom, left, right*/ if(imageRowIdx<0 || imageRowIdx >=numRows || imageColIdx <0 || imageColIdx >= numCols ) { sum+=0.0; } else { sum=sum+input[imageIdx]*kernel[kernelIdx]; } } } output[linearIdx] = (sum>0.0 ? (sum) : 0.0); //here: half-wave rectification?? }
19,031
/* * * Last name: Will * First name: Peter * Net ID: pcw276 * */ #include <cuda.h> #include <stdlib.h> #include <stdio.h> #include <string.h> #include <curand.h> #include <curand_kernel.h> #include <math.h> #include <thrust/host_vector.h> #include <thrust/device_vector.h> #define N_WALKERS 1000 #define MAX_THETA_SIZE 255 #define PARAMS 2 __constant__ float center_d[PARAMS]; // Function declarations __global__ void init(unsigned int seed, curandState_t *state); __global__ void init_walkers(float *walkers, int n_walkers, int n_theta, int r, curandState_t *state); __device__ float G(float a, curandState_t state); __device__ double Rosenbrock(float *point); __device__ void step_walkers(float *s1_walkers, unsigned int s1_n, float *s2_walkers, unsigned int s2_n, float a, unsigned int k_dim, curandState_t *states); __global__ void emcee_emcee(float *walkers, unsigned int n_walkers, unsigned int theta_dim, unsigned int steps, int a, curandState_t *states); /********************************************/ __global__ void init(unsigned int seed, unsigned int n_walkers, curandState_t *state) { int id = threadIdx.x + blockIdx.x *blockDim.x; if (id < n_walkers) { curand_init(seed, id, 0, &state[id]); } } __global__ void init_walkers(float *walkers, int n_walkers, int n_theta, int r, curandState_t *state) { int id = threadIdx.x + blockIdx.x * blockDim.x; if(id < n_walkers) { for(int i = 0; i < n_theta; i++) { walkers[(id+(i*n_walkers))] = center_d[i] + (curand_uniform(&state[id])-.5)*2*r; } } } __device__ float G(float a, float u) { return pow((u*(a-1)+1) / sqrtf(a), 2); } __device__ double Rosenbrock(float *point) { return ((double) exp(-((100*pow(point[1] - pow(point[0],2), 2)) + pow(1 - point[0],2)) / 20)); } __device__ void step_walkers(float *walkers, unsigned int n_walkers, unsigned int k, unsigned int offset, unsigned int theta_dim, float a, curandState_t *states) { int id = threadIdx.x + blockIdx.x * blockDim.x; curandState_t localState = states[id]; float u1 = curand_uniform(&localState); float u2 = curand_uniform(&localState); float u3 = curand_uniform(&localState); int w1_idx = id + offset; int w2_idx = k+ceil((n_walkers - k)*u1)-1 - offset; float w1[MAX_THETA_SIZE], x_prime[MAX_THETA_SIZE]; double q1, q2; float z = G(a, u2); for (int i = 0; i < theta_dim; i++) { w1[i] = walkers[w1_idx+(i*n_walkers)]; x_prime[i] = walkers[w2_idx+(i*n_walkers)] + z*( walkers[w1_idx+(i*n_walkers)] - walkers[w2_idx+(i*n_walkers)] ); } q1 = Rosenbrock(w1); q2 = Rosenbrock(x_prime); if (u3 < (powf(z,theta_dim-1)*(q2/q1))) { for (int i =0; i < theta_dim; i++) { walkers[w1_idx+(i*n_walkers)] = x_prime[i]; } } states[id] = localState; } __global__ void emcee_emcee(float *walkers, int k, int k2, unsigned int n_walkers, unsigned int theta_dim, int a, curandState_t *states) { int id = threadIdx.x + blockIdx.x * blockDim.x; if (id < k) { step_walkers(walkers, n_walkers, k, 0, a, theta_dim, states); } __syncthreads(); if (id < k2) { step_walkers(walkers, n_walkers, k2, k, a, theta_dim, states); } } int get_block_size(int n_walkers) { int factor = ceil((double)n_walkers / 4800); int blocksize = factor*32; if (blocksize > 256) { blocksize = 256; } return blocksize; } void get_mean(float *walkers, double *means, unsigned int n_walkers, unsigned int theta_dim, int step) { float *start_ind, *stop_ind; for(int i = 0; i < theta_dim; i++) { start_ind = walkers + i*n_walkers; stop_ind = walkers + (i+1)*n_walkers; thrust::device_vector<float> vec( start_ind, stop_ind ); means[i + theta_dim*step] = thrust::reduce( vec.begin(), vec.end() ) / n_walkers; } } int main(int argc, char *argv[]) { curandState_t *states; int seed = 10; int blocksize = get_block_size(N_WALKERS); int n_blocks = ceil((float) N_WALKERS / blocksize); int a = 2; int k = floor((double) N_WALKERS / 2); int k2 = N_WALKERS - k; float *walkers_h, *walkers_d; float *center_h = (float*) calloc(PARAMS, sizeof(float)); int r = 2; walkers_h = (float*) malloc(PARAMS*N_WALKERS*sizeof(float)); cudaMemcpyToSymbol(center_d, center_h, PARAMS*sizeof(float)); cudaMalloc((void**) &states, N_WALKERS*sizeof(curandState_t)); init<<<n_blocks, blocksize>>>(seed, N_WALKERS, states); cudaMalloc((void**) &walkers_d, PARAMS*N_WALKERS*sizeof(float)); init_walkers<<<n_blocks*2,blocksize>>>(walkers_d, N_WALKERS, PARAMS, r, states); printf("USING %d blocks of %d threads\n", n_blocks, blocksize); cudaMemcpy(walkers_h, walkers_d, PARAMS*N_WALKERS*sizeof(float), cudaMemcpyDeviceToHost); FILE *fp2 = fopen("out.txt", "w"); for(int i =0; i < N_WALKERS; i++) { for(int j=0; j<PARAMS; j++){ fprintf(fp2, "%f\t", walkers_h[i +j*N_WALKERS]); } fprintf(fp2, "\n"); } fclose(fp2); int STEPS = 10000; double *means; means = (double *) malloc(PARAMS*STEPS*sizeof(double)); for(int t = 0; t<STEPS; t++) { emcee_emcee<<<n_blocks,blocksize>>>(walkers_d, k, k2, N_WALKERS, PARAMS, a, states); get_mean(walkers_d, means, N_WALKERS, PARAMS, t); } FILE *fp4 = fopen("means.txt","w"); for(int j =0; j<STEPS; j++) { for(int i =0; i<PARAMS; i++) { fprintf(fp4, "%f \t", means[i +j*PARAMS]); } fprintf(fp4, "\n"); } fclose(fp4); cudaMemcpy(walkers_h, walkers_d, PARAMS*N_WALKERS*sizeof(float), cudaMemcpyDeviceToHost); FILE *fp3 = fopen("out2.txt", "w"); for(int i =0; i < N_WALKERS; i++) { for(int j=0; j<PARAMS; j++){ fprintf(fp3, "%f\t", walkers_h[i+j*N_WALKERS]); } fprintf(fp3, "\n"); } fclose(fp3); return 0; }
19,032
// Babak Poursartip // 09/14/2020 // Udemy Cuda // unique index calculation #include <cstdio> // =========================================== // 2d grid, 2d block __global__ void unique_gid_calculation_3d_3d(int *input) { int threadsPerBlock = blockDim.x * blockDim.y * blockDim.z; int threadPositionInBlock = threadIdx.x + blockDim.x * threadIdx.y + blockDim.x * blockDim.y * threadIdx.z; int blockPositionInGrid = blockIdx.x + gridDim.x * blockIdx.y + gridDim.x * gridDim.y * blockIdx.z; int tid = blockPositionInGrid * threadsPerBlock + threadPositionInBlock; printf("gridDim.x: %d, gridDim.y: %d, gridDim.z: %d, blockDim.x: %d, blockDim.y: %d, blockDim.z: %d, blockIdx.x: %d, blockIdx.y: %d, blockIdx.z: %d, threadIdx.x: %d, threadIdx.y: %d, threadIdx.z: %d, tid: %d \n", gridDim.x, gridDim.y, gridDim.z, blockDim.x, blockDim.y, blockDim.z, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y, threadIdx.z, tid); } // =========================================== // 2d grid, 2d block __global__ void unique_gid_calculation_2d_2d(int *input) { int tid = blockDim.x * threadIdx.y + threadIdx.x; int num_threads_in_a_block = blockDim.x * blockDim.y; int block_offset = blockIdx.x * num_threads_in_a_block; int num_threads_in_a_row = num_threads_in_a_block * gridDim.x; int row_offset = num_threads_in_a_row * blockIdx.y; int gid = tid + block_offset + row_offset; printf("gridDim.x: %d, gridDim.y: %d, gridDim.z: %d, blockDim.x: %d, blockDim.y: %d, blockDim.z: %d, blockIdx.x: %d, blockIdx.y: %d, blockIdx.z: %d, threadIdx.x: %d, threadIdx.y: %d, threadIdx.z: %d, gid: %d \n", gridDim.x, gridDim.y, gridDim.z, blockDim.x, blockDim.y, blockDim.z, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y, threadIdx.z, gid); } // =========================================== int main() { printf(" starts ..."); int nCol = 6; int nRow = 8; int array_size = nCol*nRow; int array_byte_size = sizeof(int) * array_size; int h_data[array_size]; for (int i = 0; i< array_size; ++i) { h_data[i] = i; } printf(" data on the host: \n"); for (int i = 0; i < array_size; ++i) { printf(" %d", h_data[i]); if((i%8)==0) printf("\n"); } printf("\n\n"); int *d_data; // array on the device cudaMalloc((void **)&d_data, array_byte_size); cudaMemcpy(d_data, h_data, array_byte_size, cudaMemcpyHostToDevice); dim3 grid(nCol, 1, 1); dim3 block(nRow, 1, 1); printf(" data on the device: \n"); //unique_gid_calculation_2d_2d<<<grid, block>>>(d_data); unique_gid_calculation_3d_3d<<<grid, block>>>(d_data); cudaDeviceSynchronize(); cudaDeviceReset(); printf(" finished."); return 0; }
19,033
#include "includes.h" __global__ void kernel(unsigned int rows, unsigned int cols , float* ddata,float* vdata ,float *results){ /* unsigned char y; int m, n ; unsigned int p = 0 ; int cases[3]; int controls[3]; int tot_cases = 1; int tot_controls= 1; int total = 1; float chisquare = 0.0f; float exp[3]; float Conexpected[3]; float Cexpected[3]; float numerator1; float numerator2; */ int i; float dp =0; int tid = threadIdx.x + blockIdx.x * blockDim.x; for(i =0; i<cols ;i++ ) { dp+= ddata[i*rows+tid]*vdata[i]; } /*cases[0]=1;cases[1]=1;cases[2]=1; controls[0]=1;controls[1]=1;controls[2]=1; i for ( m = 0 ; m < cRows ; m++ ) { y = snpdata[m * cols + tid]; if ( y == '0') { cases[0]++; } else if ( y == '1') { cases[1]++; } else if ( y == '2') { cases[2]++; } } for ( n = cRows ; n < cRows + contRows ; n++ ) { y = snpdata[n * cols + tid]; if ( y == '0' ) { controls[0]++; } else if ( y == '1') { controls[1]++; } else if ( y == '2') { controls[2]++; } } tot_cases = cases[0]+cases[1]+cases[2]; tot_controls = controls[0]+controls[1]+controls[2]; total = tot_cases + tot_controls; for( p = 0 ; p < 3; p++) { exp[p] = (float)cases[p] + controls[p]; Cexpected[p] = tot_cases * exp[p] / total; Conexpected[p] = tot_controls * exp[p] / total; numerator1 = (float)cases[p] - Cexpected[p]; numerator2 = (float)controls[p] - Conexpected[p]; chisquare += numerator1 * numerator1 / Cexpected[p] + numerator2 * numerator2 / Conexpected[p]; }*/ results[tid] = dp; }
19,034
#include <cmath> #include <cstdio> #include <iostream> #include "sobel.cuh" using namespace std; __global__ void conv_kernel_no_shmem(const float* image, const float* mask, float* output, unsigned int r, unsigned int c) { int tidx = threadIdx.x, tidy = threadIdx.y; int bidx = blockIdx.x, bidy = blockIdx.y; int bdy = blockDim.y, bdx = blockDim.x; float avg_intensity = 0; long x_idx = tidx + (long)bdx * (long)bidx; // long since can be > 2^31 -1 long y_idx = tidy + (long)bdy * (long)bidy; int idx = y_idx*c+x_idx; float temp = 0; for (int i = 0; i < 3; i++) { for (int j = 0; j < 3; j++) { if (x_idx-1+j < c && y_idx-1+i < r && x_idx-1+j >= 0 && y_idx-1+i >= 0) temp += image[(y_idx-1+i)*c + x_idx-1+j] * mask[i*3+j]; else temp += avg_intensity * mask[i*3+j]; } } if (y_idx < r && x_idx < c) output[idx] = temp; } __global__ void conv_kernel(const float* image, const float* mask, float* output, unsigned int r, unsigned int c) { int tidx = threadIdx.x, tidy = threadIdx.y; int bidx = blockIdx.x, bidy = blockIdx.y; int bdy = blockDim.y, bdx = blockDim.x; float avg_intensity = 0; // printf("%d %d %d %d\n", tidx, tidy, bidx, bidy); // if (tidx == 0 && tidy == 0 && bidx == 0 && bidy == 0) { // printf("%d %d\n", bidx, bidy); // for (int i = 0; i < 9; i++) // printf("%f ", mask[i]); // printf("\n"); // } extern __shared__ float arr[]; // Can't use "volatile" to prevent shmem data from being directly loaded onto registers float* img = &arr[0]; float* msk = &arr[(bdx + 2) * (bdy + 2)]; float* out = &arr[(bdx + 2) * (bdy + 2) + 3*3]; long x_idx = tidx + (long)bdx * (long)bidx; // long since can be > 2^31 -1 long y_idx = tidy + (long)bdy * (long)bidy; // load image elements in-place if (x_idx < c && y_idx < r) img[(tidy+1)*(bdx+2) + tidx+1] = image[y_idx * c + x_idx]; else img[(tidy+1)*(bdx+2) + tidx+1] = avg_intensity; if (tidx < 3 && tidy < 3) msk[tidy*3 + tidx] = mask[tidy*3 + tidx]; if (tidx == 0 && tidy == 0) { // leftmost top corner if (x_idx >= 1 && y_idx >= 1) img[tidy*(bdx+2) + tidx] = image[(y_idx-1) * c + x_idx-1]; else img[tidy*(bdx+2) + tidx] = avg_intensity; } else if (tidx == 0 && tidy == bdy - 1) { // leftmost bottom corner if (x_idx >= 1 && y_idx < r-1) img[(tidy+2)*(bdx+2) + tidx] = image[(y_idx+1) * c + x_idx-1]; else img[(tidy+2)*(bdx+2) + tidx] = avg_intensity; } else if (tidx == bdx - 1 && tidy == 0) { // rightmost top corner if (x_idx < c -1 && y_idx >= 1) img[tidy*(bdx+2) + tidx+2] = image[(y_idx-1) * c + x_idx+1]; else img[tidy*(bdx+2) + tidx+2] = avg_intensity; } else if (tidx == bdx - 1 && tidy == bdy -1) { // rightmost bottom corner if (x_idx < c -1 && y_idx < r-1) img[(tidy+2)*(bdx+2) + tidx+2] = image[(y_idx+1) * c + x_idx+1]; else img[(tidy+2)*(bdx+2) + tidx+2] = avg_intensity; } if (tidx == 0) { // leftmost col if (x_idx >= 1) img[(tidy+1)*(bdx+2) + tidx] = image[y_idx*c + x_idx-1]; else img[(tidy+1)*(bdx+2) + tidx] = avg_intensity; } else if (tidx == bdx - 1) { // rightmost col if (x_idx < c-1) img[(tidy+1)*(bdx+2) + tidx+2] = image[y_idx*c + x_idx+1]; else img[(tidy+1)*(bdx+2) + tidx+2] = avg_intensity; } if (tidy == 0) { // top row if (y_idx >= 1) img[tidy*(bdx+2) + tidx+1] = image[(y_idx-1)*c + x_idx]; else img[tidy*(bdx+2) + tidx+1] = avg_intensity; } else if (tidy == bdy - 1) { // bottom row if (y_idx < r-1) img[(tidy+2)*(bdx+2) + tidx+1] = image[(y_idx+1)*c + x_idx]; else img[(tidy+2)*(bdx+2) + tidx+1] = avg_intensity; } __syncthreads(); // if (tidx == 2 && tidy == 1 && bidx == 21 && bidy == 30) { // for (int i = 0; i < bdy+2; i++) { // for (int j = 0; j < bdx+2; j++ ) { // printf("%f ", img[i*(bdx+2)+j]); // } // printf("\n"); // } // for (int i = -1; i < bdy+1; i++) { // for (int j = -1; j < bdx+1; j++ ) { // printf("%f ", image[(bidy*bdy+i)*c+(bidx*bdx+j)]); // } // printf("\n"); // } // } out[tidy*bdx+tidx] = 0; for (int i = 0; i < 3; i++) { for (int j = 0; j < 3; j++) { out[tidy*bdx+tidx] += img[(tidy+i)*(bdx+2) + (tidx+j)] * msk[i*3+j]; } } __syncthreads(); if (x_idx < c && y_idx < r) output[y_idx*c+x_idx] = out[tidy*bdy+tidx]; } __global__ void conv_kernel_horiz(const float* image, const float* mask, float* output, unsigned int r, unsigned int c) { int tidx = threadIdx.x, tidy = threadIdx.y; int bidx = blockIdx.x, bidy = blockIdx.y; int bdy = blockDim.y, bdx = blockDim.x; float avg_intensity = 0; extern __shared__ float arr[]; // Can't use "volatile" to prevent shmem data from being directly loaded onto registers float* img = &arr[0]; float* msk = &arr[(bdx + 2) * bdy]; float* out = &arr[(bdx + 2) * bdy + 3]; long x_idx = tidx + (long)bdx * (long)bidx; // long since can be > 2^31 -1 long y_idx = tidy + (long)bdy * (long)bidy; // load image elements in-place if (x_idx < c && y_idx < r) img[tidy*(bdx+2) + tidx+1] = image[y_idx * c + x_idx]; else img[tidy*(bdx+2) + tidx+1] = avg_intensity; // load the mask into shmem if (tidx < 3 && tidy == 0) msk[tidx] = mask[tidx]; if (tidx == 0) { // leftmost col if (x_idx >= 1) img[tidy*(bdx+2) + tidx] = image[y_idx*c + x_idx-1]; else img[tidy*(bdx+2) + tidx] = avg_intensity; } else if (tidx == bdx - 1) { // rightmost col if (x_idx < c-1) img[tidy*(bdx+2) + tidx+2] = image[y_idx*c + x_idx+1]; else img[tidy*(bdx+2) + tidx+2] = avg_intensity; } __syncthreads(); out[tidy*bdx+tidx] = 0; for (int i = 0; i < 3; i++) { out[tidy*bdx+tidx] += img[tidy*(bdx+2) + (tidx+i)] * msk[i]; } __syncthreads(); if (x_idx < c && y_idx < r) output[y_idx*c+x_idx] = out[tidy*bdy+tidx]; } __global__ void conv_kernel_vert(const float* image, const float* mask, float* output, unsigned int r, unsigned int c) { int tidx = threadIdx.x, tidy = threadIdx.y; int bidx = blockIdx.x, bidy = blockIdx.y; int bdy = blockDim.y, bdx = blockDim.x; float avg_intensity = 0; extern __shared__ float arr[]; // Can't use "volatile" to prevent shmem data from being directly loaded onto registers float* img = &arr[0]; float* msk = &arr[(bdy + 2) * bdx]; float* out = &arr[(bdy + 2) * bdx + 3]; long x_idx = tidx + (long)bdx * (long)bidx; // long since can be > 2^31 -1 long y_idx = tidy + (long)bdy * (long)bidy; // load image elements in-place if (x_idx < c && y_idx < r) img[(tidy+1)*bdx + tidx] = image[y_idx * c + x_idx]; else img[(tidy+1)*bdx + tidx] = avg_intensity; // load the mask into shmem if (tidx == 0 && tidx < 3) msk[tidy] = mask[tidy]; if (tidy == 0) { // top row if (y_idx >= 1) img[tidy*bdx + tidx] = image[(y_idx-1)*c + x_idx]; else img[tidy*bdx + tidx] = avg_intensity; } else if (tidy == bdy - 1) { // bottom row if (y_idx < r-1) img[(tidy+2)*bdx + tidx] = image[(y_idx+1)*c + x_idx]; else img[(tidy+2)*bdx + tidx] = avg_intensity; } __syncthreads(); out[tidy*bdx+tidx] = 0; for (int i = 0; i < 3; i++) { out[tidy*bdx+tidx] += img[(tidy+i)*bdx + tidx] * msk[i]; } __syncthreads(); if (x_idx < c && y_idx < r) output[y_idx*c+x_idx] = out[tidy*bdy+tidx]; } __global__ void conv_kernel_vert_opt(const float* image, const float* mask, float* output, unsigned int r, unsigned int c) { int tidx = threadIdx.x, tidy = threadIdx.y; int bidx = blockIdx.x, bidy = blockIdx.y; int bdy = blockDim.y, bdx = blockDim.x; float avg_intensity = 0; extern __shared__ float arr[]; // Can't use "volatile" to prevent shmem data from being directly loaded onto registers float* img = &arr[0]; float* msk = &arr[(bdy + 2) * bdx]; float* out = &arr[(bdy + 2) * bdx + 3]; long x_idx = tidx + (long)bdx * (long)bidx; // long since can be > 2^31 -1 long y_idx = tidy + (long)bdy * (long)bidy; // load image elements in-place if (x_idx < c && y_idx < r) img[(tidy+1) + tidx*(bdy+2)] = image[y_idx * c + x_idx]; else img[(tidy+1) + tidx*(bdy+2)] = avg_intensity; // load the mask into shmem if (tidx == 0 && tidx < 3) msk[tidy] = mask[tidy]; if (tidy == 0) { // top row if (y_idx >= 1) img[tidy + tidx*(bdy+2)] = image[(y_idx-1)*c + x_idx]; else img[tidy + tidx*(bdy+2)] = avg_intensity; } else if (tidy == bdy - 1) { // bottom row if (y_idx < r-1) img[(tidy+2) + tidx*(bdy+2)] = image[(y_idx+1)*c + x_idx]; else img[(tidy+2) + tidx*(bdy+2)] = avg_intensity; } __syncthreads(); out[tidy+tidx*bdy] = 0; for (int i = 0; i < 3; i++) { out[tidy+tidx*bdy] += img[(tidy+i) + tidx*(bdy+2)] * msk[i]; } __syncthreads(); if (x_idx < c && y_idx < r) output[y_idx*c+x_idx] = out[tidy+tidx*bdy]; } __host__ void conv(const float* image, const float* mask, float* output, unsigned int r, unsigned int c, unsigned int bdx, unsigned int bdy) { dim3 block(bdx, bdy); dim3 grid((c + block.x - 1) / block.x, (r + block.y - 1) / block.y); // cout << bdx << " " << bdy << " " << (c + block.x - 1) / block.x << " " << (r + block.y - 1) / block.y << endl; // for (int i = 0; i < r; i++) { // for (int j = 0; j < c; j++) { // cout << image[i*c+j] << " "; // } // cout << endl; // } conv_kernel<<<grid, block, sizeof(float) * (bdx + 2) * (bdy + 2) + 3 * 3 * sizeof(float) + sizeof(float) * bdx * bdy>>>(image, mask, output, r, c); // conv_kernel_no_shmem<<<grid, block>>>(image, mask, output, r, c); cudaError_t err; // // Check for kernel launch errors // err = cudaGetLastError(); // if (err != cudaSuccess) // printf("Error: %s\n", cudaGetErrorString(err)); err = cudaDeviceSynchronize(); // cout << cudaGetErrorName(err) << endl; } __host__ void conv_opt(const float* image, const float* mask1, float* mask2, float* output, unsigned int r, unsigned int c, unsigned int bdx, unsigned int bdy) { dim3 block(bdx, bdy); dim3 grid((c + block.x - 1) / block.x, (r + block.y - 1) / block.y); cudaError_t err; float *temp; err = cudaMalloc((void **)&temp, r * c * sizeof(float)); // cout << cudaGetErrorName(err) << endl; conv_kernel_horiz<<<grid, block, sizeof(float) * bdy * (bdx+2) + 3 * sizeof(float) + sizeof(float) * bdx * bdy>>>(image, mask1, temp, r, c); err = cudaDeviceSynchronize(); // cout << cudaGetErrorName(err) << endl; conv_kernel_vert_opt<<<grid, block, sizeof(float) * (bdy+2) * bdx + 3 * sizeof(float) + sizeof(float) * bdx * bdy>>>(temp, mask2, output, r, c); err = cudaDeviceSynchronize(); // cout << cudaGetErrorName(err) << endl; err = cudaFree(temp); // cout << cudaGetErrorName(err) << endl; } __global__ void magnitude(const float* outx, const float *outy, float* out, unsigned int r, unsigned int c) { int x_idx = threadIdx.x + (long)blockDim.x * (long)blockIdx.x; int y_idx = threadIdx.y + (long)blockDim.y * (long)blockIdx.y; int idx = y_idx*c + x_idx; // Code motion if (x_idx < c && y_idx < r) out[idx] = sqrt(outx[idx]*outx[idx] + outy[idx]*outy[idx]); }
19,035
// includes, system #include <stdio.h> using namespace std; #include <float.h> #include <sys/stat.h> #include <limits> //#include "cuPrintf.cu" // includes CUDA #include <cuda_runtime.h> #include <device_launch_parameters.h> // includes, project //#include <helper_cuda.h> //#include <helper_functions.h> // helper functions for SDK examples __global__ void calc_pwtad (double *g_Pwtad,double *g_Ptszd,double *g_Pzd,double *g_Pwz,double *g_Ptrwz,double *g_Pd, int g_nDocs,int g_nZ, int g_nTs,int g_nVoc, int g_nTr) { // nTr,Nvoc //__shared__ double tmp_pwtad[100]; //y is nz x is nts const int block_width_x=128; __shared__ double tmp_pd; __shared__ double tmp_pwtad[block_width_x]; __shared__ double tmp_pzd; __shared__ double tmp_pwz; __shared__ double tmp_ptrwz; int pwtad_index; int Td_index; __shared__ double tmp_Ptszd[block_width_x]; tmp_pd=g_Pd[blockIdx.x]; /* __shared__ double tmp_pwz[block_height_y]; __shared__ double tmp_pzd[block_height_y]; __shared__ double tmp_ptrwz[block_height_y]; */ int nTs_counter=ceil((float)g_nTs/(float)block_width_x); int index_Ptszd; for (int tr=0;tr<g_nTr;tr++) { for (int w=0;w<g_nVoc;w++) { for (int ts=0;ts<nTs_counter;ts++) { tmp_pwtad[threadIdx.x]=0; __syncthreads(); for (int z=0;z<g_nZ;z++) { tmp_pwz=g_Pwz[z*g_nVoc+w]; tmp_ptrwz=g_Ptrwz[(z*g_nVoc+w)*g_nTr+tr]; tmp_Ptszd[threadIdx.x]=0; tmp_pzd=g_Pzd[blockIdx.x*g_nZ+z]; //tmp_pwz[threadIdx.y]=0; // tmp_pzd[threadIdx.y]=0; // tmp_ptrwz[threadIdx.y]=0; __syncthreads(); index_Ptszd=(blockIdx.x*g_nZ+z)*g_nTs+block_width_x*ts+threadIdx.x; /* tmp_pwz[threadIdx.y]=g_Pwz[w*g_nz+z*blockDim.y+threadIdx.y]; tmp_pzd[threadIdx.y]=g_Pzd[blockIdx.x*g_nz+z*blockDim.y+threadIdx.y]; tmp_ptrwz[threadIdx.y]=g_Ptrwz[((z*blockDim.y+threadIdx.y)*g_nVoc+w)*g_nTr+tr]; */ if ((block_width_x*ts+threadIdx.x)<g_nTs ) { tmp_Ptszd[threadIdx.x]= g_Ptszd[index_Ptszd]; } __syncthreads(); //tmp_pwtad[threadIdx.y][threadIdx.x]=tmp_pzd[threadIdx.y]*tmp_Ptszd[threadIdx.y][threadIdx.x]*tmp_pwz[threadIdx.y]*tmp_ptrwz[threadIdx.y]; //__syncthreads(); if ((block_width_x*ts+threadIdx.x)< g_nTs) { tmp_pwtad[threadIdx.x]+=tmp_pd*tmp_pzd*tmp_Ptszd[threadIdx.x]*tmp_pwz*tmp_ptrwz; } __syncthreads(); //if (threadIdx.x==0 && threadIdx.y==0) //printf("tmp_pwtad[0][0]::%0.8f \n",tmp_pwtad[0][0]); } // nz loop Td_index=tr+block_width_x*ts+threadIdx.x; //-((tr+block_width_x*ts+threadIdx.x)/(g_nTr+g_nTs-1)); if(Td_index< (g_nTs+g_nTr-1)) { pwtad_index=(blockIdx.x*(g_nTs+g_nTr-1)+Td_index)*g_nVoc+w; //g_Pwtad[(w*(g_nTr+g_nTs-1)+(tr+blockDim.x*ts+threadIdx.x))*g_nDocs+blockIdx.x]+=tmp_pwtad[0][threadIdx.x]; //g_Pwtad[(blockIdx.x*(g_nTr+g_nTs-1)+tr+blockDim.x*ts+threadIdx.x -((tr+blockDim.x*ts+threadIdx.x)/(g_nTr+g_nTs-2)))*g_nVoc+w]+=tmp_pwtad[0][threadIdx.x]; g_Pwtad[pwtad_index]+=tmp_pwtad[threadIdx.x]; } __syncthreads(); } // nts counter loop }// tr loop }// w loop } __global__ void calc_tszd(double *g_zd, double *g_tszd,double *g_Ptszd,double *g_Pzd,double *g_Pwz,double *g_Ptrwz,double *g_Pwtad,double *g_Doc,double *pd,double *pz, int g_nDocs,int g_nZ, int g_nTs,int g_nVoc, int g_nTr,float lambdaTsSparsity, bool z_prior , bool trainflag) { const int block_width_x=16; const int block_height_y=16; __shared__ double partial_tszd[block_height_y*block_width_x]; double tmp_tszd=0; double tmp_zd=0; int gdoc_index; int iterations_nvoc = ceil((float)g_nVoc/(float)block_width_x); int iterations_ntr = ceil((float)g_nTr/(float)block_height_y); __shared__ double decrease; __shared__ double shared_ptszd; __shared__ double shared_pzd; __shared__ double shared_pd; for (int ts=0;ts<g_nTs;ts++) { //tmp_sum=0; tmp_tszd=0; shared_ptszd=g_Ptszd[(blockIdx.y*g_nZ+blockIdx.x)*g_nTs+ts]; shared_pzd=g_Pzd[(blockIdx.y)*g_nZ+blockIdx.x]; shared_pd=pd[blockIdx.y]; __syncthreads(); for (unsigned int nvoc_counter=0;nvoc_counter<iterations_nvoc;nvoc_counter++) { for (unsigned int ntr_counter=0;ntr_counter<iterations_ntr;ntr_counter++) { partial_tszd[threadIdx.y*block_width_x+threadIdx.x]=0; __syncthreads(); if ((ntr_counter*blockDim.y+threadIdx.y)<g_nTr && (nvoc_counter*blockDim.x+threadIdx.x)<g_nVoc ) { gdoc_index=(blockIdx.y*(g_nTs+g_nTr-1)+ts+ntr_counter*blockDim.y+threadIdx.y-(ts+ntr_counter*blockDim.y+threadIdx.y)/(g_nTs+g_nTr-1))*g_nVoc+nvoc_counter*blockDim.x+threadIdx.x; partial_tszd[threadIdx.y*block_width_x+threadIdx.x]=(shared_pd*g_Doc[gdoc_index]*shared_pzd* shared_ptszd*g_Pwz[(blockIdx.x)*g_nVoc+nvoc_counter*blockDim.x+threadIdx.x]*g_Ptrwz[(blockIdx.x*g_nVoc+nvoc_counter*blockDim.x+threadIdx.x)*g_nTr+ntr_counter*blockDim.y+threadIdx.y])/(g_Pwtad[gdoc_index]+DBL_EPSILON); } __syncthreads(); for (unsigned int s=block_width_x*block_height_y/2;s>=1;s>>=1) {//g_tszd[s]=s; if (threadIdx.y*block_width_x+threadIdx.x<s) { partial_tszd[threadIdx.y*block_width_x+threadIdx.x]+=partial_tszd[threadIdx.y*block_width_x+threadIdx.x+s]; //shared_partial_tszd[0][][threadIdx.y]+=shared_partial_tszd[idx][threadIdx.z][threadIdx.y]; //sdata[threadIdx.x]+=sdata[threadIdx.x+s]; } __syncthreads(); } if (threadIdx.x==0 && threadIdx.y==0) { tmp_tszd+=partial_tszd[0]; } __syncthreads(); } //ntr counter } //nvoc counter if (threadIdx.x==0 && threadIdx.y==0) { if (lambdaTsSparsity!=0){ decrease = lambdaTsSparsity*pd[blockIdx.y]/(g_nZ*g_nTs); tmp_tszd-=decrease; tmp_tszd=max(DBL_EPSILON,tmp_tszd); } g_tszd[(blockIdx.y*g_nZ+blockIdx.x)*g_nTs+ts]=tmp_tszd; tmp_zd+=tmp_tszd; } __syncthreads(); } // nts loop if (threadIdx.x==0 && threadIdx.y==0) { if (z_prior && !trainflag) tmp_zd+= pd[blockIdx.y]/(g_nZ*pz[blockIdx.x]); g_zd[blockIdx.y*g_nZ+blockIdx.x]=tmp_zd; } } __global__ void calc_trwz(double *g_trwz,double *g_wz,double *g_Ptszd,double *g_Pzd,double *g_Pwz,double *g_Ptrwz,double *g_Pwtad,double *g_Doc,double *pd,double *g_trp,double *priortrwz,double *priorwz,int g_nDocs,int g_nZ, int g_nTs,int g_nVoc, int g_nTr,int totWords,float tr_wt,bool usePriorPtrwz,bool tr_prior) { const int block_width_x=64; int gdoc_index; int iterations_nTs = ceil((float)g_nTs/(float)block_width_x); __shared__ double partial_trwz[block_width_x]; int Td; __shared__ double tmp_sum; __shared__ double tmp_sum_outerloop; __shared__ double partial_zd; __shared__ double partial_pwz; __shared__ double partial_ptrwz; __shared__ double more; __shared__ double tmp_pd; __shared__ double tmp_sum_wz; __shared__ double word_sum_count[block_width_x]; double sum_count; tmp_sum_wz=0; tmp_sum=0; tmp_sum_outerloop=0; sum_count=0; partial_pwz=g_Pwz[blockIdx.y*g_nVoc+blockIdx.x]; __syncthreads(); for(int tr=0;tr<g_nTr;tr++) { partial_ptrwz=g_Ptrwz[(blockIdx.y*g_nVoc+blockIdx.x)*g_nTr+tr]; sum_count=0; __syncthreads(); for (int d=0;d<g_nDocs;d++) { partial_zd=g_Pzd[d*g_nZ+blockIdx.y]; tmp_sum=0; tmp_pd=pd[d]; __syncthreads(); for (unsigned int nts_counter=0;nts_counter<iterations_nTs;nts_counter++) { partial_trwz[threadIdx.x]=0; word_sum_count[threadIdx.x]=0; __syncthreads(); if (nts_counter*blockDim.x+threadIdx.x <g_nTs) { //gdoc_index=(blockIdx.y*(g_nTs+g_nTr-1)+ts+threadIdx.y-(ts+threadIdx.y)/(g_nTs+g_nTr-2))*g_nVoc+nvoc_counter*blockDim.x+threadIdx.x+ *g blockIdx.x; Td=(nts_counter*block_width_x+threadIdx.x)+tr; gdoc_index=(d*(g_nTs+g_nTr-1)+Td-(Td/(g_nTs+g_nTr-1)))*g_nVoc+ blockIdx.x; word_sum_count[threadIdx.x]=g_Doc[gdoc_index]; partial_trwz[threadIdx.x]=(g_Doc[gdoc_index]*tmp_pd*partial_zd*g_Ptszd[(d*g_nZ+blockIdx.y)*g_nTs+nts_counter*block_width_x+threadIdx.x]*partial_pwz*partial_ptrwz)/(g_Pwtad[gdoc_index]+DBL_MIN); // if ( blockIdx.x==0 && blockIdx.y==0 && blockIdx.z==0 && d==0 && nts_counter==0 ) // printf("tid x is %d and g_Doc[gdoc_index] is %0.8f and partial_trwz[threadIdx.x] is %0.8f \n",threadIdx.x,g_Doc[gdoc_index],partial_trwz[threadIdx.x]); } __syncthreads(); for (unsigned int s=block_width_x/2;s>=1;s>>=1) {//g_tszd[s]=s; if (threadIdx.x<s) { partial_trwz[threadIdx.x]+=partial_trwz[threadIdx.x+s]; word_sum_count[threadIdx.x]+=word_sum_count[threadIdx.x+s]; } __syncthreads(); } //__syncthreads(); if (threadIdx.x==0 ) { tmp_sum += partial_trwz[0]; sum_count+=word_sum_count[0]; } __syncthreads(); //g_tszd[(blockIdx.y*g_nZ+blockIdx.x)*g_nTs+ts]+=partial_tszd[0]; //printf("g_tszd value for index %d is %0.8f\n",(blockIdx.y*g_nZ+blockIdx.x)*g_nTs+ts,partial_tszd[0]); //__syncthreads(); } // nts counter loop if (threadIdx.x==0 ) { tmp_sum_outerloop+=tmp_sum; // if ( blockIdx.x==1 && blockIdx.y==0 && blockIdx.z==0 ) //printf(" d is %d and tmp_sum is %0.8f and tmp_sum_outerloop is %0.8f \n",d,tmp_sum,tmp_sum_outerloop); } __syncthreads(); } //nDocs for loop if (threadIdx.x==0 ) { if (usePriorPtrwz) { //double more = priorPwz->mat[w][z][0] * priorPtrwz->mat[tr][w][z] * m_totWords; tmp_sum_outerloop+=priortrwz[(blockIdx.y*g_nVoc+blockIdx.x)*g_nTr+tr]*priorwz[blockIdx.y*g_nVoc+blockIdx.x]*totWords; } else if (tr_prior){ //temp1.mat[tr][w][z] = sum_trwz + m_Options.tr_wt * (sum_count * trp->mat[tr][0][0]) / (m_nVoc * m_nAspects); //sum_wz += temp1.mat[tr][w][z]; tmp_sum_outerloop+=tr_wt*sum_count*g_trp[tr]/(g_nVoc*g_nZ);// fill it } g_trwz[(blockIdx.y*g_nVoc+blockIdx.x)*g_nTr+tr]=tmp_sum_outerloop; tmp_sum_wz+=tmp_sum_outerloop; tmp_sum_outerloop=0; tmp_sum=0; } __syncthreads(); }// ntr loop if (threadIdx.x==0 ) { g_wz[blockIdx.y*g_nVoc+blockIdx.x]=tmp_sum_wz; } __syncthreads(); }
19,036
#include "includes.h" __global__ void group_point_grad_gpu(int b, int n, int c, int m, int nsample, const float *grad_out, const int *idx, float *grad_points) { int index = threadIdx.x; idx += m*nsample*index; grad_out += m*nsample*c*index; grad_points += n*c*index; for (int j=0;j<m;++j) { for (int k=0;k<nsample;++k) { int ii = idx[j*nsample+k]; for (int l=0;l<c;++l) { grad_points[ii*c+l] += grad_out[j*nsample*c+k*c+l]; } } } }
19,037
#include "includes.h" __global__ void OPT_4(int *d_adjList, int *d_sizeAdj, int *d_lcmMatrix, int *d_LCMSize, int n_vertices) { int i = threadIdx.x + blockDim.x * blockIdx.x; if(i<n_vertices) { int indexUsed = 0, indexOffset = 0; int iStart = 0, iEnd = 0; int k = 0; if(i > 0) { k = d_sizeAdj[i-1]; indexOffset = d_LCMSize[i-1]; } iEnd = d_sizeAdj[i]; for(int j = indexOffset; j<iEnd; j++) { d_lcmMatrix[j] = 0; } __syncthreads(); for(int j = 0; j < n_vertices; j++) { if(i==j) continue; iStart = k; int jStart = 0, jEnd = 0; if(j > 0) jStart = d_sizeAdj[j-1]; jEnd = d_sizeAdj[j]; int compVec = 0; while (iStart < iEnd && jStart < jEnd) { if(d_adjList[iStart] < d_adjList[jStart]) iStart++; else if (d_adjList[jStart] < d_adjList[iStart]) jStart++; else // if arr1[i] == arr2[j] { jStart++; iStart++; compVec++; } } if (compVec > 0) { atomicAdd((int*)&d_lcmMatrix[indexUsed + indexOffset], compVec); // d_lcmMatrix[indexUsed + indexOffset] = compVec; indexUsed++; } // __syncthreads(); } // __syncthreads(); // d_LCMSize[i] = indexUsed; // __syncthreads(); } }
19,038
#include<iostream> #include <fstream> #include <string> #include <stdio.h> #include <stdlib.h> using namespace std; __global__ void kernel( float* r_gpu, float* g_gpu, float* b_gpu, int N, int n_m) { int tId = threadIdx.x + blockIdx.x * blockDim.x; int i=n_m; while(i < N && tId < n_m) { r_gpu[tId] += r_gpu[tId+i]; g_gpu[tId] += g_gpu[tId+i]; b_gpu[tId] += b_gpu[tId+i]; i+=n_m; } } __global__ void kernel2( float* r_gpu, float* g_gpu, float* b_gpu, int N, int l) { int tId = threadIdx.x + blockIdx.x * blockDim.x; if(tId < N) { r_gpu[tId] = r_gpu[tId]/l; g_gpu[tId] = g_gpu[tId]/l; b_gpu[tId] = b_gpu[tId]/l; } } void sumar(float* r1, float* g1, float* b1, float* r, float* g, float* b) { //cout << *r << "|" << 1-*r << endl; *r1 += *r; *g1 += *g; *b1 += *b; } void promedio(float* r, float* g, float* b, int l) { *r = *r/l; *g = *g/l; *b = *b/l; } int main(int argc, char const *argv[]) { FILE *pFile; int n, m, l; float *r, *g, *b; pFile = fopen ("images6.txt","r"); fscanf(pFile, "%d %d %d", &l, &m, &n); r = new float[n*m*l]; g = new float[n*m*l]; b = new float[n*m*l]; int block_size = 256; int grid_size = (int) ceil((float) n*m / block_size); float* r_gpu, *g_gpu, *b_gpu; cudaMalloc(&r_gpu, sizeof(float) * n * m * l); cudaMalloc(&g_gpu, sizeof(float) * n * m * l); cudaMalloc(&b_gpu, sizeof(float) * n * m * l); for (int j = 0; j < l; ++j){ for (int i = 0; i < n*m; ++i) { fscanf (pFile, "%f", &r[i+(j*n*m)]); } for (int i = 0; i < n*m; ++i) { fscanf (pFile, "%f", &g[i+(j*n*m)]); } for (int i = 0; i < n*m; ++i) { fscanf (pFile, "%f", &b[i+(j*n*m)]); } } fclose (pFile); cudaMemcpy(r_gpu, r, sizeof(float) * n * m * l, cudaMemcpyHostToDevice); cudaMemcpy(g_gpu, g, sizeof(float) * n * m * l, cudaMemcpyHostToDevice); cudaMemcpy(b_gpu, b, sizeof(float) * n * m * l, cudaMemcpyHostToDevice); int tamanio = n * m * l; int nm=n * m; cudaEvent_t ct1, ct2; float dt; cudaEventCreate(&ct1); cudaEventCreate(&ct2); cudaEventRecord(ct1); kernel<<<grid_size, block_size>>>(r_gpu, g_gpu, b_gpu, tamanio, nm); kernel2<<<grid_size, block_size>>>(r_gpu, g_gpu, b_gpu, nm, l); cudaEventRecord(ct2); cudaEventSynchronize(ct2); cudaEventElapsedTime(&dt, ct1, ct2); cout << "Tiempo GPU: " << dt << " [ms]" << endl; cudaMemcpy(r, r_gpu, sizeof(float) * n * m * l, cudaMemcpyDeviceToHost); cudaMemcpy(g, g_gpu, sizeof(float) * n * m * l, cudaMemcpyDeviceToHost); cudaMemcpy(b, b_gpu, sizeof(float) * n * m * l, cudaMemcpyDeviceToHost); FILE * pSalida; pSalida = fopen ("gpu_img_salida.txt","w"); fprintf(pSalida, "%d %d\n", m, n); for (int i = 0; i < n*m; ++i) { if(i == n*m - 1) { fprintf(pSalida, "%f", r[i]); } else { fprintf(pSalida, "%f ", r[i]); } } fprintf(pSalida, "\n"); for (int i = 0; i < n*m; ++i) { if(i == n*m - 1) { fprintf(pSalida, "%f", g[i]); } else { fprintf(pSalida, "%f ", g[i]); } } fprintf(pSalida, "\n"); for (int i = 0; i < n*m; ++i) { if(i == n*m - 1) { fprintf(pSalida, "%f", b[i]); } else { fprintf(pSalida, "%f ", b[i]); } } delete r; delete g; delete b; cudaFree(r_gpu); cudaFree(g_gpu); cudaFree(b_gpu); //cin.get(); return 0; }
19,039
/******************************************************************************* GPU OPTIMIZED MONTE CARLO (GOMC) 2.75 Copyright (C) 2022 GOMC Group A copy of the MIT License can be found in License.txt along with this program, also can be found at <https://opensource.org/licenses/MIT>. ********************************************************************************/ #include "CUDAMemoryManager.cuh" #ifdef GOMC_CUDA long long CUDAMemoryManager::totalAllocatedBytes = 0; std::unordered_map<void *, std::pair<unsigned int, std::string> > CUDAMemoryManager::allocatedPointers; cudaError_t CUDAMemoryManager::mallocMemory(void **address, unsigned int size, std::string var_name) { cudaError_t ret = cudaMalloc(address, size); allocatedPointers[*address] = make_pair(size, var_name); totalAllocatedBytes += size; if (size == 0) { std::cout << "Warning! You are trying to allocate " << var_name << " with a size of zero bytes!\n"; } return ret; } cudaError_t CUDAMemoryManager::freeMemory(void *address, std::string var_name) { if(allocatedPointers.find(address) != allocatedPointers.end()) { totalAllocatedBytes -= allocatedPointers[address].first; allocatedPointers.erase(address); } else if (address != nullptr) { std::cout << "Warning! You are trying to free " << var_name << " but it has already been freed\n" << "\tor was never allocated!\n"; } return cudaFree(address); } bool CUDAMemoryManager::isFreed() { bool ret = allocatedPointers.size() == 0; while(allocatedPointers.size() != 0) { auto it = allocatedPointers.begin(); std::cout << "You forgot to free memory " << it->second.second << " with " << it->second.first << " bytes allocated!\n"; std::cout << "I am going to free it for you!\n"; freeMemory(it->first, it->second.second); } return ret; } #endif
19,040
__device__ int findRoot(const int equivalenceMatrix[], int elementIndex){ while(equivalenceMatrix[elementIndex] != elementIndex) elementIndex = equivalenceMatrix[elementIndex]; return elementIndex; } __global__ void finalUpdate(const int* input, int* output, const int height, const int width){ int row = blockIdx.y*blockDim.y + threadIdx.y; int col = blockIdx.x*blockDim.x + threadIdx.x; int globalIndex = col * height + row; output[globalIndex] = findRoot(input, input[globalIndex]); }
19,041
#include <stdio.h> __global__ void helloFromGPU() { printf("Hello World from GPU! %d %d\n", threadIdx.x, blockIdx.x); } int main(int argc, char**argv) { printf("Hello World from CPU!\n"); int blocks = 1024; int threads = 1; helloFromGPU<<<blocks, threads>>>(); cudaDeviceReset(); return 0; }
19,042
// Copyright (c) 2018 John Biddiscombe // // Distributed under the Boost Software License, Version 1.0. (See accompanying // file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt) #include "cuda_runtime.h" __global__ void saxpy(int n, float a, float *x, float *y) { int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < n) y[i] = a * x[i] + y[i]; }
19,043
#include "Descriptor.cuh" Descriptor::Descriptor(int32_t size, int32_t type, bool normalized, int32_t stride, int32_t pointer) : size(size), type(type), normalized(normalized), stride(stride), pointer(pointer) { }
19,044
/* * hw04p02.cu * * Created on: Oct 04, 2015 * Author: Kazi * Usage: * It performs integer multiplication of a 16x32 matrix with a 32x1 vector * on a GPU. Does not take any arguments. Just generates predefined matrices * and reports the time taken to do the multiplication. */ #include <stdio.h> #include <stdlib.h> #include <math.h> #include <cuda.h> #include <iostream> /* * Given an array A of size (h x w) and a vector b of size (w), it takes the product * Ab and writes it to a vector c of size (h). */ __global__ void multArray(int* A, int* b, int* c, int w, int size) { int ti = threadIdx.x; int dsum = 0; //The sum for the dot product int k; //Perform the multiplication if (ti < size) { for(k=0; k<w; k++) { //Take the dot of a row of A with b dsum = dsum + A[ti*w+k]*b[k]; } c[ti] = dsum; } } /* * Entry point for the program. Currently specifies matrix and vector size. * Allocates memory on the host and device and then creates matrices on the * host. Copies them over to the device to multiply them. Copies the result * back over to the host. */ int main(int argc, char* argv[]) { //Set the size of the arrays, threads, blocks int height = 16; int width = 32; int threads = height; int blocks = 1; //Allocate memory on the host int *hA = (int*)malloc(width*height*sizeof(int)); int *hb = (int*)malloc(width*sizeof(int)); int *hc = (int*)malloc(height*sizeof(int)); //Allocate memory on the device int *dA, *db, *dc; cudaMalloc((void**) &dA, sizeof(int)*width*height); cudaMalloc((void**) &db, sizeof(int)*width); cudaMalloc((void**) &dc, sizeof(int)*height); //Generate the matrices on the host int i; int j; for(i=0; i<height; i++) { hc[i] = 0; //Let the storage array be all zeros for(j=0; j<width; j++) { hA[i*width + j] = i+j; //Set the matrix A if (i == 0) hb[j] = j; //Set the vector b } } //Start inclusive timing here cudaEvent_t startIn, stopIn; cudaEventCreate(&startIn); cudaEventCreate(&stopIn); cudaEventRecord(startIn, 0); //Copy hA,hb, hc onto dA,db, dc cudaMemcpy(dA, hA, sizeof(int)*width*height, cudaMemcpyHostToDevice); cudaMemcpy(db, hb, sizeof(int)*width, cudaMemcpyHostToDevice); cudaMemcpy(dc, hc, sizeof(int)*width, cudaMemcpyHostToDevice); //Start exclusive timing here cudaEvent_t startEx, stopEx; cudaEventCreate(&startEx); cudaEventCreate(&stopEx); cudaEventRecord(startEx, 0); //Use kernel to multiply A and b multArray <<<blocks,threads>>> (dA, db, dc, width, width*height); //Stop exclusive timing here cudaEventRecord(stopEx, 0); cudaEventSynchronize(stopEx); float exTime; cudaEventElapsedTime(&exTime, startEx, stopEx); cudaEventDestroy(startEx); cudaEventDestroy(stopEx); //Copy dc back into hc cudaMemcpy(hc, dc, sizeof(int)*height, cudaMemcpyDeviceToHost); //Stop inclusive timing here cudaEventRecord(stopIn, 0); cudaEventSynchronize(stopIn); float inTime; cudaEventElapsedTime(&inTime, startIn, stopIn); cudaEventDestroy(startIn); cudaEventDestroy(stopIn); //For testing - to see what the result vector looks like for(j=0; j<height; j++) { //printf("%d\n", hc[j]); std::cout << j << ": " << hc[j] << std::endl; } //Output timing printf("Inclusive time: %f ms. \n", inTime); printf("Exclusive time: %f ms. \n", exTime); //Get device properties cudaDeviceProp deviceProp; cudaGetDeviceProperties(&deviceProp, 0); printf("Device name: %s \n", deviceProp.name); printf("Clock rate: %d \n", deviceProp.clockRate); printf("Multiprocessors: %d \n", deviceProp.multiProcessorCount); printf("L2 Cache: %d \n", deviceProp.l2CacheSize); printf("Max threads per MP: %d \n", deviceProp.maxThreadsPerMultiProcessor); printf("Warp size: %d \n", deviceProp.warpSize); printf("Max threads per block: %d \n", deviceProp.maxThreadsPerBlock); printf("Max registers per block: %d \n", deviceProp.regsPerBlock); printf("Max blocks per MP: 32 \n"); //From table printf("Max warps per MP: 64 \n"); //From table printf("Shared memory per block (B): %d \n", deviceProp.sharedMemPerBlock); printf("Compute capability: %d.%d.\n", deviceProp.major, deviceProp.minor); //Write to file FILE *fp; fp = fopen("./problem2.out","w"); fprintf(fp, "Results of hw04p02.cu: \n"); for (i=0; i<height; i++) { fprintf(fp, "%d\n", hc[i]); } fclose(fp); //Cleanup if(dA) cudaFree(dA); if(db) cudaFree(db); if(dc) cudaFree(dc); if(hA) free(hA); if(hb) free(hb); if(hc) free(hc); return 0; }
19,045
/** * University of Pittsburgh * Department of Computer Science * CS1645: Introduction to HPC Systems * Instructor: Xiaolong Cui * This is a skeleton for implementing prefix sum using GPU, inspired * by nvidia course of similar name. */ #include <stdio.h> #include <math.h> #include <string.h> #include <stdlib.h> #include <time.h> #define N 512 /* Prototypes */ void random_floats(float* a, int n); void serial_scan(float* out, float* in, int n); /* * You should implement the parallel scan function here! */ __global__ void parallel_scan(float *g_odata, float *g_idata, int n) { extern __shared__ float temp_buf[]; int thread_name = threadIdx.x; int out_place = 0; int in_place = 1; temp_buf[out_place*n + thread_name] = (thread_name > 0) ? g_idata[thread_name - 1] : 0; /* Make sure threads are synchronized on transfer*/ __syncthreads(); for (int displacement = 1; displacement < n; displacement = displacement << 1) { out_place = 1 - out_place; in_place = 1 - out_place; if (thread_name >= displacement) { temp_buf[out_place*n + thread_name] += temp_buf[in_place*n + thread_name - displacement]; } else { temp_buf[out_place*n + thread_name] = temp_buf[in_place*n + thread_name]; } __syncthreads(); } g_odata[thread_name] = temp_buf[out_place*n + thread_name]; } /* * Fills an array a with n random floats. */ void random_floats(float* a, int n) { float d; // Comment out this line if you want consistent "random". srand(time(NULL)); for (int i = 0; i < n; ++i) { d = rand() % 8; a[i] = ((rand() % 64) / (d > 0 ? d : 1)); } } /* * Simple Serial implementation of exclusive scan. */ void serial_scan(float* out, float* in, int n) { float total_sum = 0; out[0] = 0; for (int i = 1; i < n; i++) { total_sum += in[i - 1]; out[i] = out[i - 1] + in[i - 1]; } if (total_sum != out[n - 1]) { printf("Warning: exceeding accuracy of float.\n"); } } /* * This is a simple function that confirms that the output of the scan * function matches that of a golden image (array). */ bool printError(float *gold_out, float *test_out, bool show_all) { bool firstFail = true; bool error = false; float epislon = 0.1; float diff = 0.0; for (int i = 0; i < N; ++i) { diff = abs(gold_out[i] - test_out[i]); if ((diff > epislon) && firstFail) { printf("ERROR: gold_out[%d] = %f != test_out[%d] = %f // diff = %f \n", i, gold_out[i], i, test_out[i], diff); firstFail = show_all; error = true; } } return error; } int main(void) { float *in, *out, *gold_out; // host float *d_in, *d_out, *dgold_out; // device int size = sizeof(float) * N; in = (float *)malloc(size); random_floats(in, N); out = (float *)malloc(size); gold_out = (float *)malloc(size); /* Allocate arrays on the device*/ cudaMalloc((void **)&d_in, size); cudaMalloc((void **)&d_out, size); cudaMalloc((void **)&dgold_out, size); /*Copy over the input array to the device*/ cudaMemcpy(d_in, in, size, cudaMemcpyHostToDevice); // *********** // RUN SERIAL SCAN // *********** serial_scan(gold_out, in, N); // *********** // RUN PARALLEL SCAN // *********** parallel_scan<<<1,N>>>(dgold_out, d_in, N); if (printError(gold_out, out, false)) { printf("ERROR: The parallel scan function failed to produce proper output.\n"); } else { printf("CONGRATS: The parallel scan function produced proper output.\n"); } return 0; }
19,046
#include <stdio.h> #include <stdlib.h> #include <math.h> #include <cufft.h> typedef float2 CComplex; typedef double2 ZComplex; template<typename T> static __device__ __host__ inline T operator+(const T a, const T b); template<typename T> static __device__ __host__ inline T operator*(const T a, const T b); // Complex addition template<typename T> static __device__ __host__ inline T operator+(const T a, const T b) { T c; c.x = a.x + b.x; c.y = a.y + b.y; return c; } // Complex multiplication template<typename T> static __device__ __host__ inline T operator*(const T a, const T b) { T c; c.x = a.x*b.x - a.y*b.y; c.y = a.x*b.y + a.y*b.x; return c; } /// @brief Computes: \f$ X \leftarrow B X \f$. /// @param[in,out] x On input this is the spectra of the waveform. /// On exit, this is the spectra multiplied with the /// spectra of the filter - i.e., the convolution. /// This is an array whose dimension is [nw]. /// @param[in] b The spectra of the filter coefficients. /// This is an array whose dimension is [nw]. /// @param[in] nw The number of frequencies. static __global__ void multiplySpectra(CComplex *x, const CComplex *b, const int nw) { const int numThreads = blockDim.x*gridDim.x; const int threadID = blockIdx.x*blockDim.x + threadIdx.x; for (int i=threadID; i<nw; i=i+numThreads) { x[i] = x[i]*b[i]; } } struct cufft32z_struct { cufftHandle mPlan; }; extern "C" void clear(struct cufft32z_struct *cuft); extern "C" void initialize(struct cufft32z_struct *cuft); void clear(struct cufft32z_struct *cuft) { cufftDestroy(cuft->mPlan); } void initialize(struct cufft32z_struct *cuft) { //checkCudaErrors(cufftMakePlan1d(plan_input, new_size, CUFFT_C2C, 1, worksize)); }
19,047
#include "includes.h" __global__ void suma(int a, int b, int *c){ *c = a+b; }
19,048
#include "includes.h" __global__ void cu_relu(const float* src, float* dst, int n){ int tid = threadIdx.x + blockIdx.x * blockDim.x; int stride = blockDim.x * gridDim.x; while(tid < n){ if(src[tid] > 0.0) dst[tid] = src[tid]; else dst[tid] = 0.0; tid += stride; } }
19,049
#include <cstdio> #include <cstdlib> #include <math.h> // Assertion to check for errors #define CUDA_SAFE_CALL(ans) { gpuAssert((ans), __FILE__, __LINE__); } inline void gpuAssert(cudaError_t code, char *file, int line, bool abort=true) { if (code != cudaSuccess) { fprintf(stderr,"CUDA_SAFE_CALL: %s %s %d\n", cudaGetErrorString(code), file, line); if (abort) exit(code); } } #define NUM_THREADS_PER_BLOCK 300 #define NUM_BLOCKS 20 #define PRINT_TIME 1 #define TEST_CASE 3 #define X_DIM 6 #define X_LENGTH 6000 #define START_ETA 0.1 #define ETA 1.0 #define DELTA .05 #define MAX_ITERS 40000 #define IMUL(a, b) __mul24(a, b) const char* getfield(char* line, int num); __global__ void calculate_weights(float* X, char* Y, float* W, char* misclassified,int x_length, int x_dim, double eta){ __shared__ float block_weights[NUM_THREADS_PER_BLOCK][X_DIM]; int tx = threadIdx.x; int tx_global = blockIdx.x*blockDim.x + threadIdx.x; int block_id = blockIdx.x; int i,j; if(misclassified[tx_global] == 1){ for(j= 0; j < x_dim;j++){ block_weights[tx][j] = eta*X[tx_global*x_dim+j]*Y[tx_global]; } } else{ for(j=0; j < x_dim; j++){ block_weights[tx][j] = 0; } } __syncthreads(); float sum; if(tx==1){ for(j=0;j<x_dim;j++){ sum = 0; for(i=0; i < NUM_THREADS_PER_BLOCK;i++){ sum = sum+ block_weights[i][j]; } W[block_id*x_dim+j]= sum; } } } __global__ void classify(float* X, char* Y, float* W, char* misclassified, int* not_classified, int* sum_missed, int x_dim){ float score; int tx = threadIdx.x; int tx_global = blockIdx.x*blockDim.x + threadIdx.x; int j; score = 0; sum_missed[tx_global] = 0; not_classified[tx_global] = 0; __syncthreads(); for(j=0;j < x_dim; j++){ score += X[tx_global*x_dim + j]*W[j]; } misclassified[tx_global] = score*Y[tx_global] <= 0.0 ? 1:0; if(misclassified[tx_global] == 1){ sum_missed[tx_global] = 1; not_classified[tx_global] = 1; } } int main(int argc, char **argv){ // GPU Timing variables cudaEvent_t start, stop; float elapsed_gpu; int test_case = TEST_CASE; int h_x_length = X_LENGTH; int h_x_dim = X_DIM; int line_counter = 0; int i; int total_missed =0; float verify = 0; int not_classified = 1; int iters = 0; float eta = ETA; float start_eta = START_ETA; float delta = DELTA; int sum_missed_iters[5][2]; // global variables on GPU float* g_W; float* g_X; float* g_score; char* g_Y; int* g_not_classified; char* g_misclassified; int* g_sum_missed; float* g_W_matrix; //global arrays on host float* h_W; float* h_W_matrix; float* h_X; float* h_score; char* h_misclassified; char* h_Y; int* h_not_classified; int* h_sum_missed; int missed = 0; // Select GPU CUDA_SAFE_CALL(cudaSetDevice(1)); // Allocate GPU memory size_t allocSize_X = h_x_dim * h_x_length * sizeof(float); size_t allocSize_Y = h_x_length * sizeof(char); size_t allocSize_W = h_x_dim * sizeof(float); size_t allocSize_Score = h_x_length * sizeof(float); size_t allocSize_sumMissed = sizeof(int)*h_x_length; size_t allocSize_notClassified = sizeof(int) * NUM_BLOCKS; size_t allocSize_W_mat = sizeof(float)*h_x_dim*NUM_BLOCKS; CUDA_SAFE_CALL(cudaMalloc((void **)&g_W, allocSize_W)) CUDA_SAFE_CALL(cudaMalloc((void **)&g_X, allocSize_X)); CUDA_SAFE_CALL(cudaMalloc((void **)&g_Y, allocSize_Y)); CUDA_SAFE_CALL(cudaMalloc((void **)&g_score, allocSize_Score)); CUDA_SAFE_CALL(cudaMalloc((void **)&g_misclassified, allocSize_Y)); CUDA_SAFE_CALL(cudaMalloc((void **)&g_sum_missed, allocSize_sumMissed)); CUDA_SAFE_CALL(cudaMalloc((void **)&g_not_classified, allocSize_sumMissed)); CUDA_SAFE_CALL(cudaMalloc((void **)&g_W_matrix, allocSize_W_mat)); // Allocate arrays on host memory h_X = (float *) malloc(allocSize_X); h_Y = (char *) malloc(allocSize_Y); h_W = (float *) malloc(allocSize_W); h_W_matrix = (float *) malloc(allocSize_W_mat); h_misclassified = (char *) malloc(allocSize_Y); h_score = (float *) malloc(allocSize_Score); h_sum_missed = (int *) malloc(allocSize_sumMissed); h_not_classified = (int *) malloc(allocSize_sumMissed); FILE* stream = fopen("data.csv", "r"); char line[1024]; while (fgets(line, 1024, stream) && line_counter < X_LENGTH) { char* tmp = strdup(line); int idx = line_counter*h_x_dim; h_X[idx] = 1.0; h_X[idx + 1] = strtod(getfield(tmp, 1), NULL); tmp = strdup(line); h_X[idx + 2] = strtod(getfield(tmp, 2), NULL); h_X[idx + 3] = h_X[idx + 1]*h_X[idx + 2]; // xy h_X[idx + 4] = h_X[idx + 1]*h_X[idx + 1]; // x^2 h_X[idx + 5] = h_X[idx + 2]*h_X[idx + 2]; // y^2 // NOTE strtok clobbers tmp free(tmp); line_counter++; } printf("x length by line counter: %d \n", line_counter); for(i=0; i < h_x_length; ++i){ switch(test_case) { case 1: h_Y[i] = (0.2*(h_X[i*h_x_dim + 1] - 0.5)) + (.6-h_X[i*h_x_dim + 2]) > 0 ? 1 : -1; break; case 2: h_Y[i] = (h_X[i*h_x_dim + 1]-.5)*(h_X[i*h_x_dim + 1]-.5) + (h_X[i*h_x_dim + 2]-.5)*(h_X[i*h_x_dim + 2]-.5) > 0.09 ? 1 : -1; break; case 3: h_Y[i] = 4*(h_X[i*h_x_dim + 1]-.5)*(h_X[i*h_x_dim + 1]-.5) + (.2-h_X[i*h_x_dim + 2]) > 0 ? 1 : -1; break; default: h_Y[i] = 0; } } float exec_times[5][2]; int k, j; float acc;; int index = 0; float current_eta = eta; for(k = 0; k < 5; k++){ for(i=0;i< h_x_length;i++){ h_misclassified[i] = 1; h_not_classified[i] = 0; h_sum_missed[i] = 0; } for(i=0; i < h_x_dim; i++){ h_W[i] = 0; } int h_W_size = NUM_BLOCKS* h_x_dim; for(i = 0; i < h_W_size; i++){ h_W_matrix[i] = 0; } CUDA_SAFE_CALL(cudaMemcpy(g_X, h_X, allocSize_X, cudaMemcpyHostToDevice)); CUDA_SAFE_CALL(cudaMemcpy(g_Y, h_Y, allocSize_Y, cudaMemcpyHostToDevice)); CUDA_SAFE_CALL(cudaMemcpy(g_W, h_W, allocSize_W, cudaMemcpyHostToDevice)); CUDA_SAFE_CALL(cudaMemcpy(g_misclassified, h_misclassified, allocSize_Y, cudaMemcpyHostToDevice)); CUDA_SAFE_CALL(cudaMemcpy(g_W_matrix, h_W_matrix, allocSize_W_mat, cudaMemcpyHostToDevice)); CUDA_SAFE_CALL(cudaMemcpy(g_not_classified, h_not_classified,allocSize_sumMissed, cudaMemcpyHostToDevice)); CUDA_SAFE_CALL(cudaMemcpy(g_sum_missed, h_sum_missed, allocSize_sumMissed, cudaMemcpyHostToDevice)); iters = 0; missed = 0; not_classified = 1; #ifdef PRINT_TIME cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start,0); #endif dim3 dimBlock(NUM_THREADS_PER_BLOCK, 1, 1); dim3 dimGrid(NUM_BLOCKS, 1); while(not_classified && iters <= MAX_ITERS){ // Increment iters iters++; // Set condition to zero (to avoid infinite while loop) and set it to one if there's an element that is misclassified not_classified = 0; // One block with 500 threads (one thread working on each row of data in X) calculate_weights<<<dimGrid, dimBlock>>>(g_X, g_Y,g_W_matrix,g_misclassified,h_x_length, h_x_dim, current_eta); CUDA_SAFE_CALL(cudaPeekAtLastError()); cudaDeviceSynchronize(); // Copy weight vector to host CUDA_SAFE_CALL(cudaMemcpy(h_W_matrix, g_W_matrix, allocSize_W_mat, cudaMemcpyDeviceToHost)); for(i=0;i<h_x_dim;i++){ acc =0; for(j=0;j<NUM_BLOCKS;j++){ acc += h_W_matrix[j*h_x_dim + i]; } h_W[i] += acc; } cudaDeviceSynchronize(); CUDA_SAFE_CALL(cudaMemcpy(g_W, h_W, allocSize_W, cudaMemcpyHostToDevice)); // Check classification success cudaDeviceSynchronize(); classify<<<dimGrid, dimBlock>>>(g_X, g_Y, g_W, g_misclassified, g_not_classified, g_sum_missed,h_x_dim); CUDA_SAFE_CALL(cudaPeekAtLastError()); cudaDeviceSynchronize(); // Copy arrays back to host CUDA_SAFE_CALL(cudaMemcpy(h_not_classified, g_not_classified,allocSize_sumMissed, cudaMemcpyDeviceToHost)); CUDA_SAFE_CALL(cudaMemcpy(h_sum_missed, g_sum_missed,allocSize_sumMissed, cudaMemcpyDeviceToHost)); for(i=0;i<h_x_length;i++){ not_classified += h_not_classified[i]; } } #ifdef PRINT_TIME cudaEventRecord(stop,0); cudaEventSynchronize(stop); cudaEventElapsedTime(&elapsed_gpu, start, stop); exec_times[k][0] = elapsed_gpu; printf("Exec time: %f \n", elapsed_gpu); cudaEventDestroy(start); #endif printf("\n"); for(i=0;i < h_x_length; ++i){ missed += h_sum_missed[i]; } if(missed == 0){ printf("Perfectly separated data\n"); } else{ printf("Finished MAX_ITERS (%d iters) and still %d misclassified \n", iters, missed); } sum_missed_iters[k][0] = missed; sum_missed_iters[k][1] = iters; total_missed =0; for(i = 0; i < h_x_length; i++){ verify = 0; for(j=0;j < h_x_dim; j++){ verify += h_X[i*h_x_dim + j] * h_W[j]; } verify = verify * h_Y[i]; if(verify < 0){ total_missed++; } } exec_times[k][1] = total_missed; } printf("Verification Missed: Exec time (ms) Sum Missed: Iters: \n"); int avgIters = 0; float avg_exec_time = 0; total_missed = 0; for(i=0;i<5; i++){ avgIters += sum_missed_iters[i][1]; avg_exec_time += exec_times[i][0]; total_missed += sum_missed_iters[i][0]; printf("\t%f\t\t\t%f\t\t%d\t\t\t%d\n", exec_times[i][1], exec_times[i][0], sum_missed_iters[i][0], sum_missed_iters[i][1]); } avgIters = avgIters/ 5; avg_exec_time = avg_exec_time/ 5; printf("Avg exec time: %f avg iters: %d sum missed (should be 0): %d \n", avg_exec_time, avgIters, total_missed); // Free-up device and host memory CUDA_SAFE_CALL(cudaFree(g_X)); CUDA_SAFE_CALL(cudaFree(g_Y)); CUDA_SAFE_CALL(cudaFree(g_W)); CUDA_SAFE_CALL(cudaFree(g_sum_missed)); CUDA_SAFE_CALL(cudaFree(g_not_classified)); CUDA_SAFE_CALL(cudaFree(g_score)); CUDA_SAFE_CALL(cudaFree(g_misclassified)); free(h_X); free(h_not_classified); free(h_sum_missed); free(h_Y); free(h_W); free(h_misclassified); free(h_score); return 0; } const char* getfield(char* line, int num) { const char* tok; for (tok = strtok(line, ","); tok && *tok; tok = strtok(NULL, ",\n")) { if (!--num) return tok; } return NULL; }
19,050
//source: https://github.com/lzhengchun/matrix-cuda/blob/master/matrix_cuda.cu #include <stdio.h> #include <stdlib.h> #include <sys/time.h> #include <assert.h> #define TYPE float #define TILE_DIM 32 /* Returns the current time in miliseconds. */ double getMilitime(){ struct timeval ret; gettimeofday(&ret, NULL); return ((ret.tv_sec ) * 1000000u + ret.tv_usec) / 1.e6; } void MatrixMultiplicationCPU(TYPE*A, TYPE*B, TYPE*C, int M, int N, int K){ int i,j,count; for(i=0;i<M;++i) for(j=0;j<K;++j) for(count=0;count<N;++count){ C[i*K+j] += A[i*N+count] * B[count*K+j]; } } #define err(format, ...) do { fprintf(stderr, format, ##__VA_ARGS__); exit(1); } while (0) #ifdef __CUDACC__ inline void checkCuda(cudaError_t e) { if (e != cudaSuccess) { // cudaGetErrorString() isn't always very helpful. Look up the error // number in the cudaError enum in driver_types.h in the CUDA includes // directory for a better explanation. err("CUDA Error %d: %s\n", e, cudaGetErrorString(e)); } } inline void checkLastCudaError() { checkCuda(cudaGetLastError()); } #endif void Allocate(TYPE**A, TYPE**B, TYPE**C, int M, int N, int K){ (*A) = (*B) = (*C) = NULL; (*A) = (TYPE*)malloc(sizeof(TYPE*) * M * N); (*B) = (TYPE*)malloc(sizeof(TYPE*) * N * K); (*C) = (TYPE*)malloc(sizeof(TYPE*) * M * K); assert((*A) != NULL); assert((*B) != NULL); assert((*C) != NULL); } void Fill(TYPE*A, TYPE*B, int M, int N, int K){ int i; for (i=0; i<M*N; i++) A[i]= 1.0; for (i=0; i<N*K; i++) B[i]= 2.0; } void UnAllocate(TYPE**A, TYPE**B, TYPE**C){ free((*A)); free((*B)); free((*C)); } void Print2DMatrix(TYPE*A, int M, int N) { int i; for(i = 0; i < M*N; ++i){ if((i%M)==0) printf("\n"); printf("%f ",A[i]); } printf("\n"); } //source: https://stackoverflow.com/questions/18815489/cuda-tiled-matrix-matrix-multiplication-with-shared-memory-and-matrix-size-whic __global__ void MatrixMultiplication(float* A, float* B, float* C, int ARows, int ACols, int BRows, int BCols, int CRows, int CCols) { float CValue = 0; int Row = blockIdx.y*TILE_DIM + threadIdx.y; int Col = blockIdx.x*TILE_DIM + threadIdx.x; __shared__ float As[TILE_DIM][TILE_DIM]; __shared__ float Bs[TILE_DIM][TILE_DIM]; for (int k = 0; k < (TILE_DIM + ACols - 1)/TILE_DIM; k++) { if (k*TILE_DIM + threadIdx.x < ACols && Row < ARows) As[threadIdx.y][threadIdx.x] = A[Row*ACols + k*TILE_DIM + threadIdx.x]; else As[threadIdx.y][threadIdx.x] = 0.0; if (k*TILE_DIM + threadIdx.y < BRows && Col < BCols) Bs[threadIdx.y][threadIdx.x] = B[(k*TILE_DIM + threadIdx.y)*BCols + Col]; else Bs[threadIdx.y][threadIdx.x] = 0.0; __syncthreads(); for (int n = 0; n < TILE_DIM; ++n) CValue += As[threadIdx.y][n] * Bs[n][threadIdx.x]; __syncthreads(); } if (Row < CRows && Col < CCols) C[((blockIdx.y * blockDim.y + threadIdx.y)*CCols) + (blockIdx.x * blockDim.x)+ threadIdx.x] = CValue; } int main(int argc, char* argv[]){ if(argc<4){ printf("Input Error\n"); return 1; } int M = atoi(argv[1]); int N = atoi(argv[2]); int K = atoi(argv[3]); TYPE*A,*B,*C, *A_Device, *B_Device, *C_Device; Allocate(&A,&B,&C,M,N,K); Fill(A,B,M,N,K); //Allocate in Device checkCuda(cudaMalloc(&A_Device, M*N*sizeof(TYPE))); checkCuda(cudaMalloc(&B_Device, N*K*sizeof(TYPE))); checkCuda(cudaMalloc(&C_Device, M*K*sizeof(TYPE))); //Copy to Device cudaMemcpy(A_Device, A, M*N*sizeof(TYPE), cudaMemcpyHostToDevice); cudaMemcpy(B_Device, B, N*K*sizeof(TYPE), cudaMemcpyHostToDevice); dim3 dimBlock(32, 32); dim3 dimGrid; dimGrid.x = (K + dimBlock.x - 1) / dimBlock.x; dimGrid.y = (M + dimBlock.y - 1) / dimBlock.y; printf("start timing\tm=%d,n=%d,k=%d\n",M,N,K); double start_time = getMilitime(); MatrixMultiplication<<<dimGrid,dimBlock>>>(A_Device,B_Device,C_Device,M,N,N,K,M,K); cudaDeviceSynchronize(); checkLastCudaError(); printf("elapsed time (Tiled CUDA MatMult): %f sec\n", getMilitime()-start_time); //copy to Host cudaMemcpy(C, C_Device, M*K*sizeof(TYPE), cudaMemcpyDeviceToHost); //Free in Device checkCuda(cudaFree(A_Device)); checkCuda(cudaFree(B_Device)); checkCuda(cudaFree(C_Device)); //verify results TYPE*C_CPU = (TYPE*) (TYPE*)malloc(sizeof(TYPE*) * M * K); MatrixMultiplicationCPU(A,B,C_CPU,M,N,K); int count; for(count=0;count<M * K;count++){ if(C_CPU[count]!=C[count]) {printf("Not Equal, idx: %d!",count);break;} } free(C_CPU); //free on Host UnAllocate(&A,&B,&C); return(0); }
19,051
#include<stdio.h> #include<stdlib.h> #include<string.h> #include <cuda_runtime_api.h> #define restrict __restrict__ #define PADDINGCLASS -2 #define OUTPUT_FILE "ocuda" #define INPUT_FILE "data" #define KMAX 20 #define CLASSESMAX 100 void check_error(cudaError_t err, const char *msg); void printStats(cudaEvent_t before, cudaEvent_t after, const char *msg); void readInput(FILE* file, float* coords, int* classes, int spacedim, int numels, int totalElements); void writeOutput(float* coords, int* classes, int spacedim, int numels); __global__ void knn(float* coords, float2* kOutput, int totalElements, int numels, int spacedim, int k, int* classes, int classes_num); __global__ void knnPunisher(float2* kOutput, int* d_classes, int numels, int newels, int k, int classes_num); __device__ float distance(float* coords, float* coords2, int spacedim); __device__ int insert(float2* kPoints, float2 newDist, int* size, int k, int gid); __device__ void swapfloat2(float2* d1, float2* d2); __device__ int deviceFindMode(int* kclasses, int classes_num, int k); __device__ float distanceShm(float* coords, int left, int spacedim); //Declaration of shared-memory. It's going to contains partial minimum of distances extern __shared__ int mPartial[]; int main(int argc, char *argv[]) { int newels; //number of points we want classify int k; //number of nearest points we use to classify int numels; //total element already classified int spacedim; char filePath[255]; //path + filname of input file int classes_num; //number of classes float* h_coords; //coords of existing points with a class int* h_classes; //array contains the class for each points //*** Device-variables-declaration *** float* d_coords; int2* d_determinate; int* d_classes; float2* d_kOutput; //*** end-device-declaration //***cudaEvent-declaration*** cudaEvent_t before_allocation, before_input, before_upload, before_knn, before_download; cudaEvent_t after_allocation, after_input, after_upload, after_knn, after_download; //***end-cudaEvent-declaration*** //Requisiti: numels e newels devono essere maggiori di K if (argc > 2) { strcpy(filePath, argv[1]); k = atoi(argv[2]); } else { printf("how-to-use: knn <inputfile> <k> \n"); exit(1); } //***cuda-init-event*** check_error(cudaEventCreate(&before_allocation), "create before_allocation cudaEvent"); check_error(cudaEventCreate(&before_input), "create before_input cudaEvent"); check_error(cudaEventCreate(&before_upload), "create before_upload cudaEvent"); check_error(cudaEventCreate(&before_knn), "create before_knn cudaEvent"); check_error(cudaEventCreate(&before_download), "create before_download cudaEvent"); check_error(cudaEventCreate(&after_allocation), "create after_allocation cudaEvent"); check_error(cudaEventCreate(&after_input), "create after_input cudaEvent"); check_error(cudaEventCreate(&after_upload), "create after_upload cudaEvent"); check_error(cudaEventCreate(&after_knn), "create after_knn cudaEvent"); check_error(cudaEventCreate(&after_download), "create after_download cudaEvent"); //***end-cuda-init-event*** FILE *fp; if((fp = fopen(filePath, "r")) == NULL) { printf("No such file\n"); exit(1); } fseek(fp, 0L, SEEK_END); float fileSize = ftell(fp); rewind(fp); int count = fscanf(fp, "%d,%d,%d,%d\n", &numels, &newels, &classes_num, &spacedim); int totalElements = numels + newels; //*** allocation *** cudaEventRecord(before_allocation); h_coords = (float*) malloc(sizeof(float)*totalElements*spacedim); h_classes = (int*) malloc(sizeof(int)*totalElements); //*** device-allocation *** check_error(cudaMalloc(&d_coords, totalElements*spacedim*sizeof(float)), "alloc d_coords_x"); check_error(cudaMalloc(&d_classes, totalElements*sizeof(int)), "alloc d_classes"); check_error(cudaMalloc(&d_determinate, newels*2*sizeof(int)), "alloc d_determinate"); check_error(cudaMalloc(&d_kOutput, newels*KMAX*2*sizeof(float)), "alloc d_kOutput"); //*** end-device-allocation *** cudaEventRecord(after_allocation); ///***input-from-file*** cudaEventRecord(before_input); readInput(fp, h_coords, h_classes, spacedim, numels, totalElements); cudaEventRecord(after_input); fclose(fp); ///***end-input-from-file*** //***copy-arrays-on-device*** cudaEventRecord(before_upload); check_error(cudaMemcpy(d_coords, h_coords, totalElements*spacedim*sizeof(float), cudaMemcpyHostToDevice), "copy d_coords"); check_error(cudaMemcpy(d_classes, h_classes, totalElements*sizeof(int), cudaMemcpyHostToDevice), "copy d_classes"); cudaEventRecord(after_upload); //***end-copy-arrays-on-device*** const int blockSize = 512; int numBlocks = (newels + blockSize - 1)/blockSize; cudaEventRecord(before_knn); knn<<<numBlocks, blockSize>>>(d_coords, d_kOutput, totalElements, numels, spacedim, k, d_classes, classes_num); knnPunisher<<<numBlocks, blockSize, newels*sizeof(int)*2>>>(d_kOutput, d_classes, numels, newels, k, classes_num); cudaEventRecord(after_knn); check_error(cudaMemcpy(h_classes+numels, d_classes+numels, newels*sizeof(int), cudaMemcpyDeviceToHost), "download classes"); check_error(cudaEventSynchronize(after_knn), "sync cudaEvents"); printStats(before_knn, after_knn, "knn"); writeOutput(h_coords, h_classes, spacedim, totalElements); return 0; } void check_error(cudaError_t err, const char *msg) { if (err != cudaSuccess) { fprintf(stderr, "%s : error %d (%s)\n", msg, err, cudaGetErrorString(err)); exit(err); } } float runtime; void printStats(cudaEvent_t before, cudaEvent_t after, const char *msg) { check_error(cudaEventElapsedTime(&runtime, before, after), msg); printf("%s %gms\n", msg, runtime); } __global__ void knn(float* coords, float2* kOutput, int totalElements, int numels, int spacedim, int k, int* classes, int classes_num) { int gid = numels + threadIdx.x + blockIdx.x*blockDim.x; //id del punto da determinare if (gid >= totalElements) return; float* newPointCoords = coords+spacedim*gid; float* pointCoords; float2 kPoints[KMAX]; int i = 0, size = 0, count = 0; float2 dist; for (i = 0; i < numels; i++) { pointCoords = coords+spacedim*i; dist = make_float2(distance(newPointCoords, pointCoords, spacedim), i); insert(kPoints, dist, &size, k, gid); } for (count=0; i < gid; i++) { pointCoords = coords+spacedim*i; dist = make_float2(distance(newPointCoords, pointCoords, spacedim), i); count += insert(kPoints, dist, &size, k, gid); } if (count > 0) { classes[gid] = -1; } else { int kclasses[KMAX]; for (int j = 0; j < k; j++) kclasses[j] = classes[(int)(kPoints[j].y)]; classes[gid] = deviceFindMode(kclasses, classes_num, k); } //copia kPoints in kOutput int newelId = gid-numels; for (i = 0; i < k; i++) kOutput[newelId*KMAX + i] = kPoints[i]; } __global__ void knnPunisher(float2* kOutput, int* classes, int numels, int newels, int k, int classes_num) { //Declaration of shared-memory. It's going to contains partial minimum of distances extern __shared__ int mPartial[]; int gid = threadIdx.x + blockIdx.x*blockDim.x; if (gid >= newels) return; //stiamo lancio un kernel per ogni newels //prendiamo la classe del newels e la mettiamo nella shared memory int i, id, lid, kclasses[KMAX], kPoints[KMAX], count = 1; //Se sono qui la classe per il kPoint è da determinare for (i = 0; i < k; i++) kPoints[i] = kOutput[gid*KMAX+i].y; //gid while(count != 0) { for (i = 0; i < gid; i++) mPartial[i] = classes[i+numels]; mPartial[gid] = classes[gid+numels]; if (mPartial[gid] != -1) return; __syncthreads(); //Le sue dipendenze, se già determinate stanno nella shared-memory count = 0; for (i = k-1; i >= 0; i--) { id = kPoints[i]; lid = id - numels; if (id > numels && mPartial[lid] < 0) { //segno quelli indeterminati count++; break; } } if (count == 0) { //posso determinare il punto //le sue dipendenze si trovano in shared memory for (i = 0; i < k; i++) kclasses[i] = classes[kPoints[i]]; classes[gid+numels] = deviceFindMode(kclasses, classes_num, k); } } } __device__ int deviceFindMode(int* kclasses, int classes_num, int k) { int classesCount[CLASSESMAX]; int i; int temp=0; for (i = 0; i < CLASSESMAX; i++) classesCount[i] = 0; for (i = 0; i < k; i++){ temp=kclasses[i]; classesCount[temp]+=1; } int max = 0; int maxValue = classesCount[0]; for (i = 1; i < classes_num; i++) { int value = classesCount[i]; if (value > maxValue) { max = i; maxValue = value; } else if (value != 0 && maxValue == value) { int j = 0; for (j = 0; j < k; j++) { if (kclasses[j] == i) { max = i; break; } else if (kclasses[j] == max) break; } } } return max; } //inserimento smart in kPoints __device__ int insert(float2* kPoints, float2 newDist, int* size, int k, int gid) { int inserted = 0; if (*size == 0) { //Caso base: inserimento su array vuoto kPoints[0] = newDist; *size = *size + 1; return 1; } int i = 1; float2* value = &newDist; //nuovo elemento float2* tail = &(kPoints[*size-i]); if (*size < k) { kPoints[*size] = newDist; value = &(kPoints[*size]); inserted = 1; } //partire della fine, swap se trovo elemento più grande - mi fermo se trovo elemento più piccolo while (i <= *size && (*tail).x > (*value).x) { swapfloat2(tail, value); value = tail; i++; tail = &(kPoints[*size-i]); inserted = 1; } if (inserted && *size < k) *size = *size + 1; return inserted; } __device__ void swapfloat2(float2* d1, float2* d2) { //da provate tmp = d1, d1 = d2, d2 = tmp float2 tmp; tmp.x = (*d1).x; tmp.y = (*d1).y; (*d1).x = (*d2).x; (*d1).y = (*d2).y; (*d2).x = tmp.x; (*d2).y = tmp.y; } // read input from file void readInput(FILE* file, float* coords, int* classes, int spacedim, int numels, int totalElements) { int i, j; int count; for(i=0; i<numels; i++) { for (j = 0; j < spacedim; j++) count = fscanf(file, "%f,", &(coords[i*spacedim +j])); count = fscanf(file, "%d\n", &(classes[i])); } for(; i < totalElements; i++) { for (j = 0; j < spacedim; j++) count = fscanf(file, "%f,", &(coords[i*spacedim+j])); count = fscanf(file, "-1\n"); } count++; } //Write Output on file void writeOutput(float* coords, int* classes, int spacedim, int numels) { FILE *fp; fp = fopen(OUTPUT_FILE, "w"); int i, j; for( i = 0; i < numels; i++) { for (j = 0; j < spacedim; j++) fprintf(fp, "%lf,", coords[i*spacedim+j]); fprintf(fp, "%d\n", classes[i]); } fclose(fp); } //multidimensional euclidian distance (without sqrt) __device__ float distance(float* coords, float* coords2, int spacedim) { float sum = 0; int i; for (i = 0; i < spacedim; i++) { float diff = coords[i] - coords2[i]; sum += diff*diff; } return sum; }
19,052
#include <stdio.h> #include <stdlib.h> __global__ void multAdd(float *d_in, float *d_out) { int idx = blockIdx.x * blockDim.x + threadIdx.x; int f = d_in[idx]; d_out[idx] = ((2*f) + 1); } int main(int argc, char** argv) { int ARRAY_SIZE = 128; int ARRAY_MEM = ARRAY_SIZE*sizeof(float); float* h_in = (float*)malloc(ARRAY_MEM); float* h_out = (float*)malloc(ARRAY_MEM); float *d_out, *d_in; for(int i = 0; i < ARRAY_SIZE; i++) { h_in[i] = i; } cudaMalloc(&d_in, ARRAY_MEM); cudaMalloc(&d_out, ARRAY_MEM); cudaMemcpy(d_in, h_in, ARRAY_MEM, cudaMemcpyHostToDevice); multAdd<<<1, 128>>>(d_in, d_out); cudaMemcpy(h_out, d_out, ARRAY_MEM, cudaMemcpyDeviceToHost); for(int i = 0; i < ARRAY_SIZE; i += 4) { printf("%10f %10f %10f %10f\n", h_out[i], h_out[i+1], h_out[i+2], h_out[i+3]); } cudaFree(d_in); cudaFree(d_out); free(h_in); free(h_out); return 0; }
19,053
#include "includes.h" __global__ void dirtyFixWindowsVarScaleKernel( float *xMin, float *xMax, float *yMin, float *yMax, const int size, const float h, const float w, const float minWidth) { int idx = BLOCK_SIZE * BLOCK_SIZE * blockIdx.x + threadIdx.x; if (idx < 2*size) { float paramMin, paramMax; if (idx < size) { paramMin = max(-h+1, min(h-1, xMin[idx])); paramMax = max(-h+1, min(h-1, xMax[idx])); if (paramMin + minWidth - 0.99 > paramMax) { const float mean = 0.5 * (paramMin + paramMax); paramMin = mean - 0.5 * (minWidth - 0.9); paramMax = mean + 0.5 * (minWidth - 0.9); } xMin[idx] = paramMin; xMax[idx] = paramMax; } else { idx -= size; paramMin = max(-w+1, min(w-1, yMin[idx])); paramMax = max(-w+1, min(w-1, yMax[idx])); if (paramMin + minWidth - 0.99 > paramMax) { const float mean = 0.5 * (paramMin + paramMax); paramMin = mean - 0.5 * (minWidth - 0.9); paramMax = mean + 0.5 * (minWidth - 0.9); } yMin[idx] = paramMin; yMax[idx] = paramMax; } } }
19,054
extern "C" __global__ void histgramMakerKernel_SharedMemAtomics(int *d_histgram, const unsigned char *d_text, int textLength) { __shared__ int sh_histgram[256]; for (int histPos = threadIdx.x; histPos < 256; histPos += blockDim.x) sh_histgram[histPos] = 0; __syncthreads(); int stride = gridDim.x * blockDim.x; int gid = blockDim.x * blockIdx.x + threadIdx.x; for (int pos = gid; pos < textLength; pos += stride) { int ch = d_text[pos]; atomicAdd(&sh_histgram[ch], 1); } __syncthreads(); for (int histPos = threadIdx.x; histPos < 256; histPos += blockDim.x) atomicAdd(&d_histgram[histPos], sh_histgram[histPos]); }
19,055
#include <stdio.h> #include <curand_kernel.h> #include <curand.h> // Adapted from // https://stackoverflow.com/questions/26650391/generate-random-number-within-a-function-with-curand-without-preallocation __global__ void myfunc(double *vals, size_t n) { int tId = threadIdx.x + (blockIdx.x * blockDim.x); curandState state; curand_init((unsigned long long)clock() + tId, 0, 0, &state); for (size_t i = 0; i < n; i++) { vals[i] = curand_uniform_double(&state); } } // Adapted from // https://stackoverflow.com/questions/7989039/use-of-cudamalloc-why-the-double-pointer int main() { size_t n = 20; size_t num_bytes = n * sizeof(double); double *vals_host = 0; vals_host = (double*)malloc(num_bytes); double *vals_device = 0; cudaMalloc((void**)&vals_device, num_bytes); myfunc<<<1, 1>>>(vals_device, n); cudaMemcpy(vals_host, vals_device, num_bytes, cudaMemcpyDeviceToHost); for (size_t i = 0; i < n; i++) { printf("%d = %lf\n", i, vals_host[i]); } free(vals_host); cudaFree(vals_device); return 0; }
19,056
#include <iostream> #include <assert.h> // Here you can set the device ID that was assigned to you #define MYDEVICE 0 // Simple utility function to check for CUDA runtime errors void checkCUDAError(const char *msg); // Part 2 of 4: implement the kernel __global__ void kernel( int *a, int dimx, int dimy ) { dim3 index; index.x = threadIdx.x + blockIdx.x * blockDim.x; // printf("index.x %d threadIdx.x %d blockIdx.x %d blockDim.x %d \n",index.x,threadIdx.x,blockIdx.x,blockDim.x); index.y = threadIdx.y + blockIdx.y * blockDim.y; // printf("index.y %d threadIdx.y %d blockIdx.y %d blockDim.y %d \n",index.y,threadIdx.y,blockIdx.y,blockDim.y); if (index.x < dimx && index.y < dimy ) { int i = index.x + index.y * dimx; // printf("i %d x %d y %d dimx %d \n",i,index.x,index.y,dimx); a[i] = i; } } //////////////////////////////////////////////////////////////////////////////// // Program main //////////////////////////////////////////////////////////////////////////////// int main(int argc, char* argv[]) { cudaSetDevice(MYDEVICE); // Part 1 and 4 of 4: set the dimensions of the matrix int const dimx = (argc > 4) ? std::atoi(argv[1]) : 4; int const dimy = (argc > 4) ? std::atoi(argv[2]) : 4; std::cout << "MATRIX " << dimx << "x" << dimy << std::endl; // int dimx = 4; // int dimy = 4; int num_bytes = dimx*dimy*sizeof(dimx); int *d_a=0, *h_a=0; // device and host pointers // h_a = (int*)malloc(num_bytes); cudaMallocHost( &h_a,num_bytes ); //allocate memory on the device cudaMalloc( &d_a,num_bytes ); if( NULL==h_a || NULL==d_a ) { std::cerr << "couldn't allocate memory" << std::endl; return 1; } // Part 2 of 4: define grid and block size and launch the kernel dim3 grid, block; // block.x = 2; // block.y = 2; block.x = (argc > 4) ? std::atoi(argv[3]) : 2; block.y = (argc > 4) ? std::atoi(argv[4]) : 2; std::cout << "#THREADS " << block.x << "x" << block.y << std::endl; grid.x = (dimx + block.x - 1)/block.x; grid.y = (dimy + block.y - 1)/block.y; std::cout << "#BLOCKS " << grid.x << "x" << grid.y << std::endl; kernel<<<grid, block>>>( d_a, dimx, dimy ); // block until the device has completed cudaDeviceSynchronize(); // check if kernel execution generated an error checkCUDAError("kernel execution"); // device to host copy cudaMemcpy( h_a, d_a, num_bytes, cudaMemcpyDeviceToHost ); // Check for any CUDA errors checkCUDAError("cudaMemcpy"); // verify the data returned to the host is correct for(int row=0; row<dimy; row++) { for(int col=0; col<dimx; col++) assert(h_a[row * dimx + col] == row * dimx + col); } // free host memory // free( h_a ); cudaFreeHost(h_a); // free device memory cudaFree( d_a ); // If the program makes it this far, then the results are correct and // there are no run-time errors. Good work! std::cout << "Correct!" << std::endl; return 0; } void checkCUDAError(const char *msg) { cudaError_t err = cudaGetLastError(); if( cudaSuccess != err) { std::cerr << "Cuda error: " << msg << " " << cudaGetErrorString(err) << std::endl; exit(-1); } }
19,057
#include<bits/stdc++.h> using namespace std; #define pi (2.0*acos(0.0)) #define eps 1e-6 #define ll long long #define inf (1<<29) #define vi vector<int> #define vll vector<ll> #define sc(x) scanf("%d",&x) #define scl(x) scanf("%lld",&x) #define all(v) v.begin() , v.end() #define me(a,val) memset( a , val ,sizeof(a) ) #define pb(x) push_back(x) #define pii pair<int,int> #define mp(a,b) make_pair(a,b) #define Q(x) (x) * (x) #define L(x) ((x<<1) + 1) #define R(x) ((x<<1) + 2) #define M(x,y) ((x+y)>>1) #define fi first #define se second #define MOD 1000000007 #define ios ios::sync_with_stdio(0) #define N 1024 #define TB 32 int main(){ clock_t tStart = clock(); int *a , *b , *c; size_t size = N * N * sizeof(int) ; a = (int *)malloc( size ); b = (int *)malloc( size ); c = (int *)malloc( size ); for(int i = 0; i < N; i++) for(int j = 0; j < N; j++) a[i * N + j] = i + j; for(int i = 0; i < N; i++) for(int j = 0; j < N; j++) b[i * N + j] = 1; for(int i = 0; i < N; i++) for(int j = 0; j < N; j++) c[i * N + j] = a[i * N + j] * b[i * N + j]; printf("Time taken: %.2fms\n", 1000.0 * (double)(clock() - tStart)/CLOCKS_PER_SEC); return 0; }
19,058
#include "includes.h" __device__ inline float stableLogit(float x) { if(x >= 0) { float z = expf(-x); return 1.0 / (1.0 + z); } else { float z = expf(x); return z / (1.0 + z); } } __global__ void gLSTMCellBackward(float* outCell, float* outXW, float* outSU, float* outB, const float* cell, const float* xW, const float* sU, const float* b, const float* mask, const float* adj, size_t rows, size_t cols) { for(int bid = 0; bid < rows; bid += gridDim.x) { int j = bid + blockIdx.x; if(j < rows) { float m = !mask || mask[j]; float* rowOutCell = outCell + j * cols; float* rowOutXW = outXW + j * cols * 4; float* rowOutSU = outSU + j * cols * 4; const float* rowCell = cell + j * cols; const float* xWrow = xW + j * cols * 4; const float* sUrow = sU + j * cols * 4; const float* rowAdj = adj + j * cols; for(int tid = 0; tid < cols; tid += blockDim.x) { int i = tid + threadIdx.x; if(i < cols) { float gf = stableLogit(xWrow[i] + sUrow[i] + b[i]); int k = i + cols; float gi = stableLogit(xWrow[k] + sUrow[k] + b[k]); int l = i + 2 * cols; float gc = tanhf(xWrow[l] + sUrow[l] + b[l]); float adj = rowAdj[i]; // dc/dc_{t-1} if(outCell) rowOutCell[i] += (m * gf - m + 1) * adj; // dc/d(b_f) = dc/d(xW_f) ... float dcdxf = m * rowCell[i] * gf * (1 - gf) * adj; if(outXW) rowOutXW[i] += dcdxf; if(outSU) rowOutSU[i] += dcdxf; if(outB) atomicAdd(outB + i, dcdxf); // dc/d(b_i) ... float dcdb_i = m * gc * gi * (1 - gi) * adj; if(outXW) rowOutXW[k] += dcdb_i; if(outSU) rowOutSU[k] += dcdb_i; if(outB) atomicAdd(outB + k, dcdb_i); // dc/d(b_c) ... float dcdxc = m * gi * (1 - gc * gc) * adj; if(outXW) rowOutXW[l] += dcdxc; if(outSU) rowOutSU[l] += dcdxc; if(outB) atomicAdd(outB + l, dcdxc); } } } } }
19,059
#include <iostream> #include <numeric> #include <stdlib.h> #include <stdio.h> #include<thrust/scan.h> /* Somethings so confuse me, why i can't get same correct result every time. */ /* These two kernel could be used on large array, but slow Best advice: use __syncthreads() before you want to use different index */ __global__ void hillis_steele_scan_forward(float * d_out, float * d_in, const int array_size){ int idx = blockDim.x * blockIdx.x + threadIdx.x; d_out[idx] = d_in[idx]; for(int step = 1; step < array_size; step *= 2){ if(idx + step >= array_size) return; __syncthreads(); float in1 = d_out[idx]; __syncthreads(); d_out[idx + step] += in1; } } __global__ void hillis_steele_scan_backward(float * d_out, float * d_in){ int idx = blockDim.x * blockIdx.x + threadIdx.x; d_out[idx] = d_in[idx]; for(int step = 1; step <= idx; step *= 2){ if(idx - step < 0) return; __syncthreads(); float in1 = d_out[idx - step]; __syncthreads(); d_out[idx] += in1; } } /* These two kernel could be used on small array, but fast */ __global__ void shared_hillis_steele_scan_forward(float *d_out, float *d_in, const int array_size) { extern __shared__ float sdata[]; int idx = threadIdx.x; if(idx < array_size) { sdata[idx] = d_in[idx]; } else { return; } // the code below performs iterative scan on XY for(int step = 1; step < array_size; step *= 2){ if(idx + step >= array_size) return; __syncthreads(); float in1 = sdata[idx]; __syncthreads(); sdata[idx + step] += in1; } d_out[idx] = sdata[idx]; } __global__ void shared_hillis_steele_scan_backward(float * d_out, float * d_in, const int array_size){ extern __shared__ float sdata[]; int idx = threadIdx.x; if(idx < array_size) { sdata[idx] = d_in[idx]; } else { return; } sdata[idx] = d_in[idx]; for(int step = 1; step <= idx; step *= 2){ if(idx - step < 0) return; __syncthreads(); float in1 = sdata[idx - step]; __syncthreads(); sdata[idx] += in1; } d_out[idx] = sdata[idx]; } /* This kernel will get correct result when array size is power of 2 */ __global__ void blelloch_exclusive_scan(float *g_odata, float *g_idata, int n) { extern __shared__ float temp[];// allocated on invocation int thid = threadIdx.x; int offset = 1; temp[thid] = g_idata[thid]; for (int d = n / 2; d > 0; d /= 2) // build sum in place up the tree { __syncthreads(); if (thid < d) { int ai = offset*(2*thid+1)-1; int bi = offset*(2*thid+2)-1; temp[bi] += temp[ai]; } offset <<= 1; //multiply by 2 implemented as bitwise operation } if (thid == 0) { temp[n - 1] = 0; } // clear the last element for (int d = 1; d < n; d *= 2) // traverse down tree & build scan { offset >>= 1; __syncthreads(); if (thid < d) { int ai = offset*(2*thid+1)-1; int bi = offset*(2*thid+2)-1; float t = temp[ai]; temp[ai] = temp[bi]; temp[bi] += t; } } __syncthreads(); g_odata[thid] = temp[thid]; } int main(int argc, char ** argv) { const int ARRAY_SIZE = 1025; const int ARRAY_BYTES = ARRAY_SIZE * sizeof(float); const int maxThreadPerBlock = 512; const int numBlock = ARRAY_SIZE / maxThreadPerBlock + 1; // generate the input array on the host float h_in[ARRAY_SIZE]; for (int i = 0; i < ARRAY_SIZE; i++) { h_in[i] = float(i); } float h_out[ARRAY_SIZE]; // declare GPU memory pointers float * d_in; float * d_out; // allocate GPU memory cudaMalloc((void**) &d_in, ARRAY_BYTES); cudaMalloc((void**) &d_out, ARRAY_BYTES); // transfer the array to the GPU cudaMemcpy(d_in, h_in, ARRAY_BYTES, cudaMemcpyHostToDevice); // launch the kernel //hillis_steele_scan_forward<<<numBlock, maxThreadPerBlock>>>(d_out, d_in, ARRAY_SIZE); //hillis_steele_scan_backward<<<numBlock, maxThreadPerBlock>>>(d_out, d_in); //shared_hillis_steele_scan_forward<<<numBlock, maxThreadPerBlock, maxThreadPerBlock * sizeof(float)>>>(d_out, d_in, ARRAY_SIZE); //shared_hillis_steele_scan_backward<<<numBlock, maxThreadPerBlock, maxThreadPerBlock * sizeof(float)>>>(d_out, d_in, ARRAY_SIZE); //blelloch_exclusive_scan<<<numBlock, maxThreadPerBlock, maxThreadPerBlock * sizeof(float)>>>(d_out, d_in, ARRAY_SIZE); thrust::inclusive_scan(h_in, h_in + ARRAY_SIZE, h_out); //copy back the result array to the CPU //cudaMemcpy(h_out, d_out, ARRAY_BYTES, cudaMemcpyDeviceToHost); // print out the resulting array for (int i =0; i < ARRAY_SIZE; i++) { printf("%f", h_out[i]); printf(((i % 4) != 3) ? "\t" : "\n"); } cudaFree(d_in); cudaFree(d_out); return 0; }
19,060
/* Simple Monte Carlo Pi Simulation using CUDA Primatives */ #include <curand.h> #include <iostream> #include <iomanip> __device__ int total_device_points{}; __global__ void measure_points(const float* random_x, const float* random_y) { const int i = blockIdx.x * blockDim.x + threadIdx.x; const float x = random_x[i] - 0.5F; const float y = random_y[i] - 0.5F; const int n = sqrtf(pow(x, 2) + pow(y, 2)) > 0.5F ? 0 : 1; atomicAdd(&total_device_points, n); } int main() { constexpr int width = 512; constexpr int height = 512; constexpr int count = width * height; constexpr int size = count * sizeof(float); curandGenerator_t random_generator; curandCreateGenerator(&random_generator, CURAND_RNG_PSEUDO_MTGP32); curandSetPseudoRandomGeneratorSeed(random_generator, time(nullptr)); float *random_x, *random_y; cudaMalloc(&random_x, size); cudaMalloc(&random_y, size); curandGenerateUniform(random_generator, random_x, count); curandGenerateUniform(random_generator, random_y, count); measure_points <<<width, height>>> (random_x, random_y); int total_host_points; cudaMemcpyFromSymbol(&total_host_points, total_device_points, sizeof(int)); const float estimated_pi = ((4.0F * static_cast<float>(total_host_points)) / static_cast<float>(count)); std::cout << std::setprecision(std::numeric_limits<float>::digits10 + 1) << "Using the Monte Carlo Method Pi is estimated to be: " << estimated_pi << '\n'; cudaFree(random_x); cudaFree(random_y); }
19,061
#include "conv2d-transpose-input-grad.hh" #include "graph.hh" #include "../runtime/graph.hh" #include "../runtime/node.hh" #include "../memory/alloc.hh" namespace ops { Conv2DTransposeInputGrad::Conv2DTransposeInputGrad(Op* y, Op* kernel, const int strides[], const int input_size[]) : Op("conv2d_transpose_input_grad", Shape({input_size[0], input_size[1], input_size[2], input_size[3]}), {y, kernel}) , m_strides(strides) { m_input_size[0] = input_size[0]; m_input_size[1] = input_size[1]; m_input_size[2] = input_size[2]; m_input_size[3] = input_size[3]; } void Conv2DTransposeInputGrad::compile() { auto& g = Graph::instance(); auto& cy = g.compiled(preds()[0]); auto& ckernel = g.compiled(preds()[1]); Shape out_shape({m_input_size[0], m_input_size[1], m_input_size[2], m_input_size[3]}); dbl_t* out_data = tensor_alloc(out_shape.total()); int y_size[4] = { cy.out_shape[0], cy.out_shape[1], cy.out_shape[2], cy.out_shape[3]}; int kernel_size[4] = { ckernel.out_shape[0], ckernel.out_shape[1], ckernel.out_shape[2], ckernel.out_shape[3]}; auto out_node = rt::Node::op_conv2d_transpose_input_grad(cy.out_data, ckernel.out_data, m_strides, out_data, y_size, kernel_size, m_input_size, {cy.out_node, ckernel.out_node}); g.add_compiled(this, {out_node}, {out_data}, out_node, out_shape, out_data); } }
19,062
//****************************************************************************** // // File: ModCubRoot.cu // // Version: 1.0 //****************************************************************************** // Number of threads per block. #define NT 1024 // Overall counter variable in global memory. __device__ unsigned long long int count; __device__ unsigned long long int arraySize = 3; /** * Device kernel to compute modular cube root. * * * @author Omkar Kakade */ extern "C" __global__ void computeModularCubeRoot ( unsigned long long int c, unsigned long long int N, unsigned long long int *final_M) { unsigned long long int thr, size, rank; unsigned long long int local_c; unsigned long long int local_m; unsigned long long int increment; // Determine number of threads and this thread's rank. thr = threadIdx.x; size = gridDim.x*NT; rank = blockIdx.x*NT + thr; // Initialize per-thread. local_c = 0; local_m = 0; increment = 1; unsigned long long int atom_result =0; // Compute modular cube roots. for (unsigned long long int i = rank; i < N; i += size) { unsigned long long int first_mod = (i)%N; unsigned long long int second_mod = (first_mod * i)%N; unsigned long long int third_mod = (second_mod * i)%N; local_c = third_mod; local_m = i; if (local_c == c){ // atomic counter value updation. atom_result = atomicAdd(&count,increment); if (atom_result < arraySize) { final_M[atom_result]=local_m; } } } }
19,063
// // Created by Peter Rigole on 2019-03-08. // #include "Managed.cuh" __host__ void *Managed::operator new(size_t len) { void *ptr; cudaMallocManaged(&ptr, len); return ptr; } __host__ void Managed::operator delete(void *ptr) { cudaFree(ptr); }
19,064
#include <iostream> #include <random> #include <iomanip> #include <cuda.h> #define MASK_DIM_SIZE 10 // Number of elements for one spacial dimension #define TILE_SIZE 16 __constant__ float d_mask[MASK_DIM_SIZE * MASK_DIM_SIZE]; /*------------------------------------------------------------------------------------------------- * FORWARD DECLARATION *-----------------------------------------------------------------------------------------------*/ __global__ void naive_convolution_kernel(const float *d_in, float *d_out, int M, int N, int mask_m, int mask_n); __global__ void tiled_convolution_kernel(const float *d_in, float *d_out, int M, int N, int mask_m, int mask_n); __global__ void cache_convolution_kernel(const float *d_in, float *d_out, int M, int N, int mask_m, int mask_n); void cpu_convolution(const float *A, const float *mask, float *out, int M, int N, int mask_m, int mask_n); float convolution_naive(const float *in, float *out, int M, int N, int mask_m, int mask_n); float convolution_tiled(const float *in, float *out, int M, int N, int mask_m, int mask_n); float convolution_cache(const float *in, float *out, int M, int N, int mask_m, int mask_n); float check_convolution(const float *A, const float *B, int M, int N); /*------------------------------------------------------------------------------------------------- * MAIN *-----------------------------------------------------------------------------------------------*/ int main(int argc, char **argv) { const int M = 2000, N = 2000; const int mask_m = 3, mask_n = 3; float *A = new float[M * N]; float *cpu_out = new float[M * N]; float *naive_out = new float[M * N]; float *tiled_out = new float[M * N]; float *cache_out = new float[M * N]; float *mask = new float[mask_m * mask_n]; std::random_device rd; std::mt19937 engine(rd()); std::uniform_real_distribution<float> rand0(0, 10.0); for (int i = 0; i < M * N; i++) A[i] = rand0(engine); mask[0] = 1.0; mask[1] = 2.0; mask[2] = 1.0; mask[3] = 2.0; mask[4] = 4.0; mask[5] = 2.0; mask[6] = 1.0; mask[7] = 2.0; mask[8] = 1.0; cudaMemcpyToSymbol(d_mask, mask, mask_m * mask_n * sizeof(float)); cpu_convolution(A, mask, cpu_out, M, N, mask_m, mask_n); float naive_time = convolution_naive(A, naive_out, M, N, mask_m, mask_n); float tiled_time = convolution_tiled(A, tiled_out, M, N, mask_m, mask_n); float cache_time = convolution_cache(A, cache_out, M, N, mask_m, mask_n); float naive_err = check_convolution(cpu_out, naive_out, M, N); std::cout << "Naive convolution error = " << naive_err << ". Time = " << naive_time << "ms.\n"; float tiled_err = check_convolution(cpu_out, tiled_out, M, N); std::cout << "tiled convolution error = " << tiled_err<< ". Time = " << tiled_time << "ms.\n"; float cache_err = check_convolution(cpu_out, cache_out, M, N); std::cout << "cache convolution error = " << cache_err << ". Time = " << cache_time << "ms.\n"; delete[] A; delete[] cpu_out; delete[] naive_out; delete[] tiled_out; delete[] cache_out; delete[] mask; } /*------------------------------------------------------------------------------------------------- * KERNELS *-----------------------------------------------------------------------------------------------*/ __global__ void naive_convolution_kernel(const float *d_in, float *d_out, int M, int N, int mask_m, int mask_n) { int m_idx = blockIdx.y * blockDim.y + threadIdx.y; int n_idx = blockIdx.x * blockDim.x + threadIdx.x; // Check if indices are within matrix if (m_idx < M && n_idx < N) { int start_m = m_idx - (mask_m / 2); int start_n = n_idx - (mask_n / 2); float val = 0.0; // Loop over mask for (int m = 0; m < mask_m; m++) { for (int n = 0; n < mask_n; n++) { int i = start_m + m; int j = start_n + n; if (i >= 0 && i < M && j >= 0 && j < N) val += d_in[i * N + j] * d_mask[m * mask_n + n]; } } d_out[m_idx * N + n_idx] = val; } } __global__ void tiled_convolution_kernel(const float *d_in, float *d_out, int M, int N, int mask_m, int mask_n) { int row = blockIdx.y * blockDim.y + threadIdx.y; int col = blockIdx.x * blockDim.x + threadIdx.x; int m_half = mask_m / 2; int n_half = mask_n / 2; __shared__ float tile[TILE_SIZE + MASK_DIM_SIZE - 1][TILE_SIZE + MASK_DIM_SIZE - 1]; int left_idx = (blockIdx.x - 1) * blockDim.x + threadIdx.x; int right_idx = (blockIdx.x + 1) * blockDim.x + threadIdx.x; int top_idx = (blockIdx.y - 1) * blockDim.y + threadIdx.y; int bottom_idx = (blockIdx.y + 1) * blockDim.y + threadIdx.y; // Load block matrix tile[threadIdx.y + m_half][threadIdx.x + n_half] = (row < M || col < N) ? d_in[row * N + col] : 0; // Load left and right tiles if (threadIdx.x >= blockDim.x - n_half) tile[threadIdx.y + m_half][threadIdx.x - blockDim.x + n_half] = (left_idx < 0) ? 0 : d_in[row * N + left_idx]; if (threadIdx.x < n_half) tile[threadIdx.y + m_half][threadIdx.x + blockDim.x + n_half] = (right_idx >= N) ? 0 : d_in[row * N + right_idx]; // Load top and bottom tiles, including corners if (threadIdx.y >= blockDim.y - m_half) { tile[threadIdx.y - blockDim.y + m_half][threadIdx.x + n_half] = (top_idx < 0) ? 0 : d_in[top_idx * N + col]; if (threadIdx.x >= blockDim.x - n_half) tile[threadIdx.y - blockDim.y + m_half][threadIdx.x - blockDim.x + n_half] = (top_idx < 0 || left_idx < 0) ? 0 : d_in[top_idx * N + left_idx]; else if (threadIdx.x < n_half) tile[threadIdx.y - blockDim.y + m_half][threadIdx.x + blockDim.x + n_half] = (top_idx < 0 || right_idx >= N) ? 0 : d_in[top_idx * N + right_idx]; } if (threadIdx.y < m_half) { tile[threadIdx.y + blockDim.y + m_half][threadIdx.x + n_half] = (bottom_idx >= M) ? 0 : d_in[bottom_idx * N + col]; if (threadIdx.x >= blockDim.x - n_half) tile[threadIdx.y + blockDim.y + m_half][threadIdx.x - blockDim.x + n_half] = (bottom_idx >= M || left_idx < 0) ? 0 : d_in[bottom_idx * N + left_idx]; else if (threadIdx.x < n_half) tile[threadIdx.y + blockDim.y + m_half][threadIdx.x + blockDim.x + n_half] = (bottom_idx >= M || right_idx >= N) ? 0 : d_in[bottom_idx * N + right_idx]; } __syncthreads(); if (row < M && col < N) { float sum = 0.0; // Perform convolution for (int i = 0; i < mask_m; i++) for (int j = 0; j < mask_n; j++) sum += tile[threadIdx.y + i][threadIdx.x + j] * d_mask[i * mask_n + j]; d_out[row * N + col] = sum; } } __global__ void cache_convolution_kernel(const float *d_in, float *d_out, int M, int N, int mask_m, int mask_n) { int row = blockIdx.y * blockDim.y + threadIdx.y; int col = blockIdx.x * blockDim.x + threadIdx.x; int m_half = mask_m / 2; int n_half = mask_n / 2; __shared__ float tile[TILE_SIZE][TILE_SIZE]; if (row < M && col < N) tile[threadIdx.y][threadIdx.x] = d_in[row * N + col]; __syncthreads(); int row_tile_start = blockIdx.y * blockDim.y; int row_tile_end = (blockIdx.y + 1) * blockDim.y; int col_tile_start = blockIdx.x * blockDim.x; int col_tile_end = (blockIdx.x + 1) * blockDim.x; int m_start = row - m_half; int n_start = col - n_half; float sum = 0.0; if (row < M && col < N) { // Loop over mask for (int i = 0; i < mask_m; i++) { for (int j = 0; j < mask_n; j++) { int m = m_start + i; int n = n_start + j; // Check if indicies with mask applied is within matrix if (m >= 0 && m < M && n >= 0 && n < N) { // Check if elements are in the cache if (m >= row_tile_start && m < row_tile_end && n >= col_tile_start && n < col_tile_end) sum += tile[threadIdx.y + i - m_half][threadIdx.x + j - n_half] * d_mask[i * mask_n + j]; else sum += d_in[m * N + n] * d_mask[i * mask_n + j]; } } } d_out[row * N + col] = sum; } } /*------------------------------------------------------------------------------------------------- * FUNCTIONS *-----------------------------------------------------------------------------------------------*/ void cpu_convolution(const float *A, const float *mask, float *out, int M, int N, int mask_m, int mask_n) { int half_m = mask_m / 2; int half_n = mask_n / 2; for (int i = 0; i < M; i++) { for (int j = 0; j < N; j++) { int start_m = i - half_m; int start_n = j - half_n; float val = 0.0; for (int m = 0; m < mask_m; m++) { for (int n = 0; n < mask_n; n++) { int m_idx = start_m + m; int n_idx = start_n + n; if (m_idx >= 0 && m_idx < M && n_idx >= 0 && n_idx < N) val += A[m_idx * N + n_idx] * mask[m * mask_n + n]; } } out[i * N + j] = val; } } } float convolution_naive(const float *in, float *out, int M, int N, int mask_m, int mask_n) { int size = M * N * sizeof(float); float *d_in, *d_out; cudaMalloc((void**) &d_in, size); cudaMalloc((void**) &d_out, size); cudaMemcpy(d_in, in, size, cudaMemcpyHostToDevice); int N_thd = TILE_SIZE; int M_blk = ceil(static_cast<float>(M) / static_cast<float>(N_thd)); int N_blk = ceil(static_cast<float>(N) / static_cast<float>(N_thd)); dim3 grid_dim(M_blk, N_blk, 1); dim3 blk_dim(N_thd, N_thd, 1); cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start); naive_convolution_kernel<<<grid_dim, blk_dim>>>(d_in, d_out, M, N, mask_m, mask_n); cudaEventRecord(stop); cudaMemcpy(out, d_out, size, cudaMemcpyDeviceToHost); cudaEventSynchronize(stop); float time_ms; cudaEventElapsedTime(&time_ms, start, stop); cudaFree(d_in); cudaFree(d_out); return time_ms; } float convolution_tiled(const float *in, float *out, int M, int N, int mask_m, int mask_n) { int size = M * N * sizeof(float); float *d_in, *d_out; cudaMalloc((void**) &d_in, size); cudaMalloc((void**) &d_out, size); cudaMemcpy(d_in, in, size, cudaMemcpyHostToDevice); int N_thd = TILE_SIZE; int M_blk = ceil(static_cast<float>(M) / static_cast<float>(N_thd)); int N_blk = ceil(static_cast<float>(N) / static_cast<float>(N_thd)); dim3 grid_dim(M_blk, N_blk, 1); dim3 blk_dim(N_thd, N_thd, 1); cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start); tiled_convolution_kernel<<<grid_dim, blk_dim>>>(d_in, d_out, M, N, mask_m, mask_n); cudaEventRecord(stop); cudaMemcpy(out, d_out, size, cudaMemcpyDeviceToHost); cudaEventSynchronize(stop); float time_ms; cudaEventElapsedTime(&time_ms, start, stop); cudaFree(d_in); cudaFree(d_out); return time_ms; } float convolution_cache(const float *in, float *out, int M, int N, int mask_m, int mask_n) { int size = M * N * sizeof(float); float *d_in, *d_out; cudaMalloc((void**) &d_in, size); cudaMalloc((void**) &d_out, size); cudaMemcpy(d_in, in, size, cudaMemcpyHostToDevice); int N_thd = TILE_SIZE; int M_blk = ceil(static_cast<float>(M) / static_cast<float>(N_thd)); int N_blk = ceil(static_cast<float>(N) / static_cast<float>(N_thd)); dim3 grid_dim(M_blk, N_blk, 1); dim3 blk_dim(N_thd, N_thd, 1); cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start); cache_convolution_kernel<<<grid_dim, blk_dim>>>(d_in, d_out, M, N, mask_m, mask_n); cudaEventRecord(stop); cudaMemcpy(out, d_out, size, cudaMemcpyDeviceToHost); cudaEventSynchronize(stop); float time_ms; cudaEventElapsedTime(&time_ms, start, stop); cudaFree(d_in); cudaFree(d_out); return time_ms; } float check_convolution(const float *A, const float *B, int M, int N) { float sum = 0.0; for (int i = 0; i < N; i++) sum += (A[i] - B[i]) * (A[i] - B[i]); return sqrt(sum); }
19,065
#include <array> #include <cassert> #include <chrono> #include <iostream> #include <math.h> #include <string> using namespace std; float lb = 0; float ub = 2; const int nx = 41; const int ny = 41; int nt = 500; int nit = 50; int c = 1; float dx = ub / float(nx - 1); float dy = ub / float(ny - 1); float rho = 1; float nu = 0.1; float dt = 0.001; void npprint(float *u, int dimx = ny, int dimy = nx, string msg = "OUT: ") { printf("%s\n", msg.c_str()); printf("x-------------------------------x\n"); printf("[\n"); for (int i = 0; i < dimx; i++) { printf("["); for (int k = 0; k < dimy; k++) printf("%3.4f, ", u[i * dimy + k]); printf("],\n"); } printf("]\n"); printf("x-------------------------------x\n"); } void linspace(float *x, int lb, int ub, int num) { for (int k = 0; k < num; k++) { x[k] = k * (ub - lb) / float(num - 1) + lb; } } __device__ void copy(float *lhs, float *rhs, int dimx, int dimy) { for (int k = 0; k < dimy * dimx; k++) rhs[k] = lhs[k]; } void fill(float *x, float fillnum, int dimx, int dimy) { for (int k = 0; k < dimy * dimx; k++) x[k] = fillnum; } void meshgrid(float *x, float *y, float *X, float *Y) { int dimx = ny; int dimy = ny; for (int k = 0; k < dimy; k++) for (int i = 0; i < dimx; i++) { X[k * dimx + i] = x[i]; Y[k * dimx + i] = y[k]; } } __global__ void build_up_b(float *b, float rho, float dt, float *u, float *v, float dx, float dy, int nx, int ny) { /* def build_up_b(b, rho, dt, u, v, dx, dy): return b */ int xlim = ny; int ylim = nx; int idx = blockIdx.x; int idy = threadIdx.x; assert(nx = blockDim.x); if (idx >= xlim - 2) return; if (idy >= ylim - 2) return; b[(idx + 1) * nx + idy + 1] = (rho * (1 / dt * ((u[(idx + 1) * nx + idy + 2] - u[(idx + 1) * nx + idy]) / (2 * dx) + (v[(idx + 2) * nx + idy + 1] - v[(idx)*nx + idy + 1]) / (2 * dy)) - pow(((u[(idx + 1) * nx + idy + 2] - u[(idx + 1) * nx + idy]) / (2 * dx)), 2) - 2 * ((u[(idx + 2) * nx + idy + 1] - u[(idx)*nx + idy + 1]) / (2 * dy) * (v[(idx + 1) * nx + idy + 2] - v[(idx + 1) * nx + idy]) / (2 * dx)) - pow((v[(idx + 2) * nx + idy + 1] - v[(idx)*nx + idy + 1]) / (2 * dy), 2))); } __global__ void pmargin(float *p, int nx, int ny) { int xlim = ny; int ylim = nx; int idx = blockIdx.x; int idy = threadIdx.x; p[0 * ylim + idy] = p[1 * ylim + idy]; // dp/dy = 0 at y = 0 p[(xlim - 1) * ylim + idy] = 0; // p = 0 at y = 2 __syncthreads(); p[idx * ylim + ylim - 1] = p[idx * ylim + ylim - 2]; // dp/dx = 0 at x =2 p[idx * ylim + 0] = p[idx * ylim + 1]; // dp/dx = 0 at x =0 __syncthreads(); } __global__ void pupdate(float *p, float *pn, float dx, float dy, float *b, int nx, int ny, int nit) { int xlim = ny; int ylim = nx; int idx = blockIdx.x; int idy = threadIdx.x; if (idx < xlim - 2 && idy < ylim - 2) { copy(p, pn, ny, nx); __syncthreads(); p[(idx + 1) * nx + idy + 1] = (((pn[(idx + 1) * nx + idy + 2] + pn[(idx + 1) * nx + idy]) * pow(dy, 2) + (pn[(idx + 2) * nx + idy + 1] + pn[(idx)*nx + idy + 1]) * pow(dx, 2)) / (2 * (pow(dx, 2) + pow(dy, 2))) - pow(dx, 2) * pow(dy, 2) / (2 * (pow(dx, 2) + pow(dy, 2))) * b[(idx + 1) * nx + idy + 1]); } __syncthreads(); } void pressure_poisson(float *p, float *pn, float dx, float dy, float *b, int nx, int ny, int nit) { /* def pressure_poisson(p, dx, dy, b): */ for (int q = 0; q < nit; q++) { pupdate<<<ny, nx>>>(p, pn, dx, dy, b, nx, ny, nit); cudaDeviceSynchronize(); pmargin<<<ny, nx>>>(p, nx, ny); cudaDeviceSynchronize(); } } __global__ void cupdate(int nt, float *u, float *v, float *un, float *vn, float dt, float dx, float dy, float *p, float rho, float nu) { int idx = blockIdx.x; int idy = threadIdx.x; copy(u, un, ny, nx); copy(v, vn, ny, nx); __syncthreads(); u[(idx + 1) * nx + idy + 1] = (un[(idx + 1) * nx + idy + 1] - un[(idx + 1) * nx + idy + 1] * dt / dx * (un[(idx + 1) * nx + idy + 1] - un[(idx + 1) * nx + idy]) - vn[(idx + 1) * nx + idy + 1] * dt / dy * (un[(idx + 1) * nx + idy + 1] - un[(idx)*nx + idy + 1]) - dt / (2 * rho * dx) * (p[(idx + 1) * nx + idy + 2] - p[(idx + 1) * nx + idy]) + nu * (dt / pow(dx, 2) * (un[(idx + 1) * nx + idy + 2] - 2 * un[(idx + 1) * nx + idy + 1] + un[(idx + 1) * nx + idy]) + dt / pow(dy, 2) * (un[(idx + 2) * nx + idy + 1] - 2 * un[(idx + 1) * nx + idy + 1] + un[(idx)*nx + idy + 1]))); __syncthreads(); v[(idx + 1) * nx + idy + 1] = (vn[(idx + 1) * nx + idy + 1] - un[(idx + 1) * nx + idy + 1] * dt / dx * (vn[(idx + 1) * nx + idy + 1] - vn[(idx + 1) * nx + idy]) - vn[(idx + 1) * nx + idy + 1] * dt / dy * (vn[(idx + 1) * nx + idy + 1] - vn[(idx)*nx + idy + 1]) - dt / (2 * rho * dy) * (p[(idx + 2) * nx + idy + 1] - p[(idx)*nx + idy + 1]) + nu * (dt / pow(dx, 2) * (vn[(idx + 1) * nx + idy + 2] - 2 * vn[(idx + 1) * nx + idy + 1] + vn[(idx + 1) * nx + idy]) + dt / pow(dy, 2) * (vn[(idx + 2) * nx + idy + 1] - 2 * vn[(idx + 1) * nx + idy + 1] + vn[(idx)*nx + idy + 1]))); __syncthreads(); } __global__ void cmargin(float *u, float *v, int nx, int ny) { int xlim = ny; int ylim = nx; int idx = blockIdx.x; int idy = threadIdx.x; u[0 * ylim + idy] = 0; u[(xlim - 1) * ylim + idy] = 10; v[0 * ylim + idy] = 0; v[(xlim - 1) * ylim + idy] = 0; __syncthreads(); u[idx * ylim + 0] = 0; v[idx * ylim + 0] = 0; u[idx * ylim + ylim - 1] = 0; v[idx * ylim + ylim - 1] = 0; __syncthreads(); } void cavity_flow(int nt, float *u, float *v, float *un, float *vn, float dt, float dx, float dy, float *p, float *pn, float rho, float nu) { /* def cavity_flow(nt, u, v, dt, dx, dy, p, rho, nu): */ float *b; cudaMallocManaged(&b, nx * ny * sizeof(float)); fill(b, 0, nx, ny); for (int n = 0; n < nt; n++) { build_up_b<<<nx, ny>>>(b, rho, dt, u, v, dx, dy, nx, ny); cudaDeviceSynchronize(); pressure_poisson(p, pn, dx, dy, b, nx, ny, nit); cudaDeviceSynchronize(); cupdate<<<nx, ny>>>(nt, u, v, un, vn, dt, dx, dy, p, rho, nu); cudaDeviceSynchronize(); cmargin<<<nx, ny>>>(u, v, nx, ny); cudaDeviceSynchronize(); } cudaFree(b); } int main() { float x[nx]; float y[ny]; float X[ny * nx]; float Y[ny * nx]; linspace(x, lb, ub, nx); linspace(y, lb, ub, ny); meshgrid(x, y, X, Y); float *u, *un; float *v, *vn; float *p, *pn; cudaMallocManaged(&p, nx * ny * sizeof(float)); cudaMallocManaged(&pn, nx * ny * sizeof(float)); cudaMallocManaged(&u, nx * ny * sizeof(float)); cudaMallocManaged(&v, nx * ny * sizeof(float)); cudaMallocManaged(&un, nx * ny * sizeof(float)); cudaMallocManaged(&vn, nx * ny * sizeof(float)); fill(u, 0, nx, ny); fill(v, 0, nx, ny); fill(p, 0, nx, ny); fill(pn, 0, nx, ny); fill(vn, 0, nx, ny); fill(un, 0, nx, ny); auto start = std::chrono::high_resolution_clock::now(); cavity_flow(nt, u, v, un, vn, dt, dx, dy, p, pn, rho, nu); cudaDeviceSynchronize(); auto finish = std::chrono::high_resolution_clock::now(); npprint(u, ny, nx, "U"); npprint(v, ny, nx, "V"); npprint(p, ny, nx, "P"); std::chrono::duration<double> elapsed = finish - start; printf("GPU Elapsed time: %3.3f s\n", elapsed.count()); cudaFree(p); cudaFree(pn); cudaFree(vn); cudaFree(un); cudaFree(u); cudaFree(v); }
19,066
#include <cuda.h> #include <stdio.h> #define N 32 void printMatrix (unsigned* matrix) { for (unsigned i = 0; i < N * N; i++) { printf(" %u ", matrix[i]); if (i % N == (N-1)) { printf("\n"); } } } void createMatrix(unsigned* matrix) { for (unsigned i = 0; i < N; i++) { for (unsigned j = 0; j < N; j++) { if (i == j) { matrix[i * N + j] = i + 1; } else { matrix[i * N + j] = 0; } } } } __global__ void square (unsigned* matrix, unsigned* result, unsigned matrixSize) { unsigned id = blockIdx.x * blockDim.x + threadIdx.x; for (unsigned jj = 0; jj < matrixSize; jj++) { for (unsigned kk = 0; kk < matrixSize; kk++) { result[id * matrixSize + jj] += matrix[id * matrixSize + kk] * matrix[kk * matrixSize + jj]; } } } __host__ void call_sqr (unsigned* h_in_matrix, unsigned* h_out_matrix) { // unsigned n = N; unsigned *d_in_matrix, *d_out_matrix; cudaMalloc((void **) &d_in_matrix, N * N * sizeof(unsigned)); cudaMalloc((void **) &d_out_matrix, N * N * sizeof(unsigned)); cudaMemcpy(d_in_matrix, h_in_matrix, N * N * sizeof(unsigned), cudaMemcpyHostToDevice); square<<<1, N>>>(d_in_matrix, d_out_matrix, N); cudaMemcpy(h_out_matrix, d_out_matrix, N * N * sizeof(unsigned), cudaMemcpyDeviceToHost); cudaFree(d_in_matrix); cudaFree(d_out_matrix); } int main() { unsigned* matrix; unsigned* result; matrix = (unsigned*)malloc(N * N * sizeof(unsigned)); result = (unsigned*)malloc(N * N * sizeof(unsigned)); createMatrix(matrix); call_sqr (matrix, result); printMatrix(result); free(matrix); free(result); return 0; }
19,067
#include "includes.h" #define TILE_WIDTH = 16; __global__ void matrixMultiply(float *A, float *B, float *C, int numARows, int numAColumns, int numBRows, int numBColumns, int numCRows, int numCColumns) { //@@Y-axis matrix dimension int row = blockIdx.y*blockDim.y + threadIdx.y; //@@X-axis matrix Dimension int columns = blockIdx.x*blockDim.x + threadIdx.x; //@@Initilaizing final value to add in the output matrix float pValue = 0; //@@checking for boundary condition if (row < numARows && columns < numBColumns) { //@@adding values from 0 to matrix A width and from 0 to martrix B Height for (int k = 0; k < numAColumns; k++) { //@@Summation of the rows from matrix A and columns from matrix B to pValue pValue += A[row*numAColumns + k] * B[k*numBColumns + columns]; } //@@add the final value to the output matrix C[row*numCColumns + columns] = pValue; } }
19,068
#include "includes.h" __global__ void binaryCrossEntropyCost(float* cost, float* predictions, float* target, int size) { int index = blockIdx.x * blockDim.x + threadIdx.x; if (index < size) { float partial_cost = target[index] * logf(1.0e-15+predictions[index]) + (1.0f - target[index]) * logf(1.0e-15+(1.0f - predictions[index])); atomicAdd(cost, -1.0 * partial_cost / size); } }
19,069
#include<stdio.h> #include <time.h> #define TRUE 1 #define FALSE 0 #define MIN(a,b) (a < b?a:b ) static const int N = 150; __global__ void cerca_array_device(int *array,int *valor,int *res) { int b; int id = threadIdx.x + blockIdx.x * blockDim.x; if(*res == FALSE && *valor == array[id]){ *res = TRUE; for(int i = 0; i <= 1000000; i++) //demostrar que si se hacen mas operacines en la busqueda, los threads son b = (b*70)/3; } } __host__ bool cerca_array_host(int *array, int valor) { int b; for(int i = 0 ; i < N ; ++i){ if(array[i] == valor){ return true; } for(int i = 0; i <= 1000000; i++) b = (b*70)/3; } return false; } int main() { srand(time(NULL)); int a[N],valor; for(int i=0;i<N;i++) a[i] = (int)rand()/(int)(RAND_MAX/300.0); for(int i=0;i<N;i++) printf("valor: %d \t", a[i]); printf("\nNombre a cercar: "); scanf("%d",&valor); //Execució a la CPU int res; clock_t t_host = clock(); res = cerca_array_host(a,valor); t_host = clock() - t_host; double time_taken_host = ((double)t_host)/CLOCKS_PER_SEC; printf("CPU: %f segons \n", time_taken_host); if(res == TRUE) printf("host: We found the number\n"); else printf("host: We don't found the number :(\n"); int *dev_array , *dev_value , *dev_res; cudaMalloc((void**)&dev_array, N*sizeof(int) ); cudaMalloc((void**)&dev_value, sizeof(int) ); cudaMalloc((void**)&dev_res, sizeof(int) ); res = FALSE; cudaMemcpy(dev_array, a, N*sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(dev_value, &valor, sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(dev_res, &res, sizeof(int), cudaMemcpyHostToDevice); int threads_block = MIN(512,N); while(N%threads_block != 0)--threads_block; int blocks = N / threads_block; clock_t t_device = clock(); cerca_array_device<<<blocks,threads_block>>>(dev_array,dev_value,dev_res); cudaMemcpy(&res, dev_res, sizeof(int), cudaMemcpyDeviceToHost);//Copy memory from device to host t_device = clock() - t_device; double time_taken_device = ((double)t_device)/CLOCKS_PER_SEC; printf("GPU %f segons \n", time_taken_device); cudaFree(dev_array);//free device mem cudaFree(dev_value); cudaFree(dev_res); //Printa si ha cercat el nombre if(res == TRUE) printf("device: We found the number\n"); else printf("device: We don't found the number :(\n"); return 0; }
19,070
#include<stdio.h> #include<cuda.h> __global__ void sq(float *d_out, float* d_in) { int idx = threadIdx.x; float f = d_in[idx]; d_out[idx] = f*f; } int main(int argc, char** argv) { const int ARRAY_SIZE = 64; const int ARRAY_BYTES = ARRAY_SIZE * sizeof(float); float h_in[ARRAY_SIZE]; for(int i=0; i < ARRAY_SIZE; i++) { h_in[i] = float(i); } float h_out[ARRAY_SIZE]; //declare GPU memory pointers float *d_in; float *d_out; //allocate memory for the two arrays on the device cudaMalloc((void**)&d_in,ARRAY_BYTES); cudaMalloc((void**)&d_out,ARRAY_BYTES); //transfer the array to the GPU // destination,source,size,method cudaMemcpy(d_in,h_in,ARRAY_BYTES,cudaMemcpyHostToDevice); //launch the kernel sq<<<ARRAY_SIZE,1>>>(d_out,d_in); cudaDeviceSynchronize(); cudaError_t error = cudaGetLastError(); if(error!=cudaSuccess) { fprintf(stderr,"ERROR: %s\n", cudaGetErrorString(error) ); exit(-1); } //copy the results back onto the device cudaMemcpy(h_out,d_out,ARRAY_BYTES,cudaMemcpyDeviceToHost); for(int i=0; i < ARRAY_SIZE; i++) { printf("%f:%f \t",h_in[i], h_out[i]); if (i%4==0) printf("\n"); } printf("\n"); cudaFree(d_in); cudaFree(d_out); }//end of main
19,071
#ifdef __NVCC__ // __device__ volatile int PQ[MAX_NODE]; //K in parallel template <class U> __global__ void extractMin(unsigned int* PQ, unsigned int* PQ_size, int* expandNodes,int* expandNodes_size,U* Cx,int* openList,int N,int K){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id<K && PQ_size[id]>0){ //extract min from PQ int front = id* ( (N+K-1)/K ); int node = PQ[front]; // restructure the heap PQ[front]=PQ[front+PQ_size[id]-1]; PQ_size[id]-=1; int pqIndex = 0; while(2*pqIndex+1 < PQ_size[id]){ if(2*pqIndex+2 >= PQ_size[id]){ if( Cx[PQ[front+pqIndex]] > Cx[PQ[front+2*pqIndex+1]]){ int swap = PQ[front + 2*pqIndex+1]; PQ[front + 2*pqIndex+1] = PQ[front +pqIndex]; PQ[front + pqIndex] = swap; pqIndex = 2*pqIndex+1; } else break; } else{ if( Cx[PQ[front+pqIndex]] > Cx[PQ[front+2*pqIndex+1]] && Cx[PQ[front+2*pqIndex+1]] <= Cx[PQ[front+2*pqIndex+2]] ){ int swap = PQ[front + 2*pqIndex+1]; PQ[front + 2*pqIndex+1] = PQ[front +pqIndex]; PQ[front + pqIndex] = swap; pqIndex = 2*pqIndex+1; } else if(Cx[PQ[front+pqIndex]] > Cx[PQ[front+2*pqIndex+2]] && Cx[PQ[front+2*pqIndex+2]] <= Cx[PQ[front+2*pqIndex+1]] ){ int swap = PQ[front + 2*pqIndex+2]; PQ[front + 2*pqIndex+2] = PQ[front +pqIndex]; PQ[front + pqIndex] = swap; pqIndex = 2*pqIndex+2; } else{ break; } } } //removed from openList openList[node] = -1; //added to expand next int len = atomicAdd(expandNodes_size,1); expandNodes[len]=node; } } //for K in parallel template <class T,class U> __global__ void A_star_expand(int* off,int* edge, T* W,U* Hx,int* parent,volatile U* Cx, int* expandNodes,int* expandNodes_size, int* lock ,int* flagfound,int* openList,int* nVFlag, int N,int E, int K,int dest, int flagDiff,int dE, int* diff_off,int* diff_edge,unsigned int* diff_weight ){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id< *expandNodes_size ){ int node = expandNodes[id]; //reach dest if(node == dest){ atomicOr(flagfound,1); } // expand int start = off[node]; int end = E; if(node!=N-1) end = off[node+1]; while(start < end){ int child = edge[start]; //deleted edges if(child<0){ start++; continue; } //array L initilaized with 0 //get the lock for child to update C(x) //loop till acquire the lock bool leaveLoop = false; while(leaveLoop==false){ if(atomicCAS(&lock[child],0,1)==0){ //critical section if( Cx[child] > (Cx[node] - Hx[node])+ W[start]+ Hx[child] ){ Cx[child] = (Cx[node] - Hx[node])+ W[start]+ Hx[child]; __threadfence(); parent[child] = node; if(openList[child]==-1){ nVFlag[child]=1; //add only once } } //end critical section leaveLoop = true; atomicCAS(&lock[child],1,0); } __syncthreads(); } start++; } //diff expand if(flagDiff){ start = diff_off[node]; end = dE; if(node!=N-1) end = diff_off[node+1]; while(start<end){ int child = diff_edge[start]; //deleted edges if(child<0){ start++; continue; } //array L initilaized with 0 //get the lock for child to update C(x) bool leaveLoop = false; while(!leaveLoop){ if(atomicCAS(&lock[child],0,1)==0){ //critical section if( Cx[child] > (Cx[node] - Hx[node])+ W[start]+ Hx[child] ){ Cx[child] = (Cx[node] - Hx[node])+ W[start]+ Hx[child]; __threadfence(); parent[child] = node; if(openList[child]==-1){ nVFlag[child]=1; //add only once } } //end critical section leaveLoop = true; atomicCAS(&lock[child],1,0); } __syncthreads(); } start++; } } //end diff }//end } //K in parallel -- O(N) template <class U> __global__ void keepHeapPQ(unsigned int* PQ,unsigned int* PQ_size,U* Cx,int N,int K){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id < K && PQ_size[id] > 0){ int front = id*( (N+K-1)/K ); int size = PQ_size[id]; for(int i=front;i<front+size;i++){ if(2*i+2 < front+size){ int cost = Cx[PQ[i]]; int costLeft = Cx[PQ[2*i+1]]; int costRight = Cx[PQ[2*i+2]]; if( cost > costLeft || cost > costRight ){ int index ; if(costLeft <= costRight) index = 2*i+1; else index = 2*i+2; while(index > front){ if( Cx[PQ[(index-1)/2]] > Cx[PQ[index]] ){ int swap = PQ[index]; PQ[index] = PQ[(index-1)/2]; PQ[(index-1)/2] = swap; index = (index-1)/2; } else break; } } } else if(2*i+1 < front+size){ if(Cx[PQ[i]] > Cx[PQ[2*i+1]]){ int index = 2*i+1; while(index > front){ if( Cx[PQ[(index-1)/2]] > Cx[PQ[index]] ){ int swap = PQ[index]; PQ[index] = PQ[(index-1)/2]; PQ[(index-1)/2] = swap; index = (index-1)/2; } else break; } } } } } } //N threads __global__ void setNV(int* nextFlag,int* nextV,int* nvSize,int N){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id < N){ if(nextFlag[id]==1){ int index = atomicAdd(nvSize,1); nextV[index]=id; } } } //for K in parallel template <class U> __global__ void insertPQ(unsigned int* PQ,unsigned int* PQS,int* nextV,int* nVsize,U* Cx,int K,int N,int* openList){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id < K){ int front = id*( (N+K-1)/K ); int i = id; while(i<*nVsize){ //if not already present if(openList[nextV[i]]!=-1){ i+=K; continue; } PQ[front+PQS[id]]= nextV[i]; PQS[id]+=1; //add in openList openList[nextV[i]] = id; if(PQS[id]>1){ int index = PQS[id]-1; while(index>0){ if(Cx[PQ[front+ (index-1)/2]] > Cx[PQ[front+index]]){ int swap = PQ[front+index]; PQ[front+index]=PQ[front+ (index-1)/2]; PQ[front+ (index-1)/2] = swap; index = (index-1)/2; } else break; } } i += K; } } } //for K in parallel template <class U> __global__ void checkMIN(unsigned int* PQ, unsigned int* PQ_size,int* flagEnd,U* Cx,int dest,int N,int K){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id < K && PQ_size[id] > 0 ){ int front = id* ( (N+K-1)/K ); int node = PQ[front]; //check if atleast one min, dont end the a* if( Cx[node] < Cx[dest] ){ atomicAnd(flagEnd,0); } } } template <class T, class U> __global__ void propogateDel(int* delEdgesV,int delEdge, volatile U* Cx, int* rev_offset,int* rev_edges,T* rev_weight,int N,int E, U* Hx,volatile int* parent,int* parent_old,int* addFlag, int* rev_diff_offset,int* rev_diff_edges,T* rev_diff_weight,int dE){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id<delEdge){ int node = delEdgesV[id]; //check for the parent and add to nextflag and update the cost int start = rev_offset[node]; int end = E; if(node!=N-1) end = rev_offset[node+1]; //no parent // write in parent read always from old_parent parent[node] = -1; Cx[node]=INT_MAX; addFlag[node]=1; int cost = INT_MAX; int opt_parent = -1; //if any parent can change the cost while(start< end){ int p = rev_edges[start]; //del edges if(p<0 || p==node){ start++; continue; } int weight = rev_weight[start]; int flag_cycle = false; //check parent doesn't contain node int ancestor = parent_old[p]; while(ancestor>0){ if(ancestor==node){ flag_cycle = true; break; } ancestor = parent_old[ancestor]; } //no need to lock only single parent so only one node in array so one node per thread if(!flag_cycle && Cx[p]!=INT_MAX && cost > (Cx[p]-Hx[p])+weight+Hx[node] ){ cost = (Cx[p]-Hx[p] )+weight+Hx[node]; opt_parent = p; } start++; } start = rev_diff_offset[node]; end = dE; if(node!=N-1) end = rev_diff_offset[node+1]; while(start< end){ int p = rev_diff_edges[start]; //del edges if(p<0 || p==node){ start++; continue; } int weight = rev_diff_weight[start]; int flag_cycle = false; //check parent doesn't contain node int ancestor = parent_old[p]; while(ancestor!=-1){ if(ancestor==node){ flag_cycle = true; break; } ancestor = parent_old[ancestor]; } //no need to lock only single parent so only one node in array so one node per thread if(!flag_cycle && Cx[p]!=INT_MAX && cost > (Cx[p]-Hx[p])+weight+Hx[node] ){ cost = (Cx[p]-Hx[p] )+weight+Hx[node]; opt_parent = p; } start++; } //write here if(cost!=INT_MAX){ Cx[node]=cost; parent[node]=opt_parent; } } } //add inserted edges to propogate template <class T, class U> __global__ void propogateAdd(int* diff_off, int* diff_edges,T* diff_W,U* Hx,int* addFlag, volatile U* Cx,int* lock, int* parent, int* parent_old, int N, int dE){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id < N){ int node = id; int start = diff_off[node]; int end = dE; if(node!=N-1) end = diff_off[node+1]; while(start < end ){ int child = diff_edges[start]; //deleted edges if(child<0){ start++; continue; } //array L initilaized with 0 //get the lock for child to update C(x) //loop till acquire the lock bool leaveLoop = false; while(!leaveLoop){ if(atomicCAS(&lock[child],0,1)==0){ //critical section bool flag_cycle = false; int ancestor = node; while(ancestor > 0){ if(ancestor==child){ flag_cycle = true; break; } ancestor = parent_old[ancestor]; } if(!flag_cycle && Cx[node] != INT_MAX && Cx[child] > (Cx[node] - Hx[node])+ diff_W[start]+ Hx[child] ){ Cx[child] = (Cx[node] - Hx[node])+ diff_W[start]+ Hx[child]; parent[child] = node; __threadfence(); addFlag[child]=1; } //end critical section leaveLoop = true; atomicCAS(&lock[child],1,0); } __syncthreads(); } start++; } } } template <class T,class U> __global__ void insert_propagate(int* nodes, int* size, int* off, int* edge,T* W,U* Hx, int N,int E,volatile U* Cx,int* lock, int* parent,int* addFlag, int* diff_off,int* diff_edge,T* diff_W,int dE){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id < *size){ int node = nodes[id]; int start = off[node]; int end = E; if(node!=N-1) end = off[node+1]; while(start < end ){ int child = edge[start]; //deleted edges if(child<0){ start++; continue; } bool leaveLoop = false; while(!leaveLoop){ if(atomicExch(&lock[child],1)==0){ if(Cx[node]!=INT_MAX && Cx[child] > (Cx[node] - Hx[node])+ W[start]+ Hx[child] ){ Cx[child] = (Cx[node] - Hx[node])+ W[start]+ Hx[child]; __threadfence(); parent[child] = node; addFlag[child]=1; } leaveLoop = true; atomicExch(&lock[child],0); } __syncthreads(); } start++; } start = diff_off[node]; end = dE; if(node!=N-1) end = diff_off[node+1]; while(start < end ){ int child = diff_edge[start]; //deleted edges if(child<0){ start++; continue; } bool leaveLoop = false; while(!leaveLoop){ if(atomicCAS(&lock[child],0,1)==0){ //critical section if(Cx[node]!=INT_MAX && Cx[child] > (Cx[node] - Hx[node])+ diff_W[start]+ Hx[child] ){ Cx[child] = (Cx[node] - Hx[node])+ diff_W[start]+ Hx[child]; __threadfence(); parent[child] = node; addFlag[child]=1; } //end critical section leaveLoop = true; atomicCAS(&lock[child],1,0); } __syncthreads(); } start++; } } } template <class T,class U> __global__ void delete_propagate(int* nodes, int* size, int* off, int* edge,T* W,U* Hx, int N,int E,volatile U* Cx,int* lock, int* parent,int* parent_old,int* addFlag, int* diff_off,int* diff_edge,T* diff_W,int dE, int* rev_offset,int* rev_edges,T* rev_weight, int* rev_diff_offset,int* rev_diff_edges,T* rev_diff_weight){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id < *size){ int node = nodes[id]; int start = off[node]; int end = E; if(node!=N-1) end = off[node+1]; while(start < end ){ int child = edge[start]; if(child<0){ start++; continue; } bool leaveLoop = false; while(!leaveLoop){ if(atomicExch(&lock[child],1)==0){ if(Cx[node]!=INT_MAX && Cx[child]!=INT_MAX && Cx[child] > (Cx[node] - Hx[node])+ W[start]+ Hx[child] ){ Cx[child] = (Cx[node] - Hx[node])+ W[start]+ Hx[child]; __threadfence(); parent[child] = node; addFlag[child]=1; } else if( (Cx[node]==INT_MAX && parent[child]==node ) || ( parent[child]==node && (Cx[child] < Cx[node] - Hx[node]+ W[start]+ Hx[child]) ) ){ //use back edges int rstart = rev_offset[child]; int rend = E; if(child!=N-1) rend = rev_offset[child+1]; //there is always one parent that is node. Cx[child] = INT_MAX; parent[child]=-1; while(rstart < rend){ int p = rev_edges[rstart]; if(p<0 || p == child){ rstart++; continue; } int weight = rev_weight[rstart]; bool flag_cycle = false; //check parent doesn't contain child int ancestor = parent_old[p]; while(ancestor > 0){ if(ancestor==child){ flag_cycle = true; break; } ancestor = parent_old[ancestor]; } if(!flag_cycle && Cx[p]!=INT_MAX && Cx[child] > (Cx[p]-Hx[p])+weight+Hx[child] ){ Cx[child] = (Cx[p]-Hx[p] )+weight+Hx[child]; parent[child] = p; } rstart++; } rstart = rev_diff_offset[child]; rend = dE; if(child!=N-1) rend = rev_diff_offset[child+1]; while(rstart < rend){ int p = rev_diff_edges[rstart]; if(p<0 || p==child){ rstart++; continue; } int weight = rev_diff_weight[rstart]; int flag_cycle = false; //check parent doesn't contain child int ancestor = parent_old[p]; while(ancestor!=-1){ if(ancestor==child){ flag_cycle = true; break; } ancestor = parent_old[ancestor]; } if(!flag_cycle && Cx[p]!=INT_MAX && Cx[child] > (Cx[p]-Hx[p])+weight+Hx[child] ){ Cx[child] = (Cx[p]-Hx[p] )+weight+Hx[child]; parent[child] = p; } rstart++; } addFlag[child]=1; } leaveLoop = true; atomicExch(&lock[child],0); } __syncthreads(); } start++; } start = diff_off[node]; end = dE; if(node!=N-1) end = diff_off[node+1]; while(start < end ){ int child = diff_edge[start]; //deleted edges if(child<0){ start++; continue; } //array L initilaized with 0 //get the lock for child to update C(x) //loop till acquire the lock bool leaveLoop = false; while(!leaveLoop){ if(atomicCAS(&lock[child],0,1)==0){ if(Cx[node]!=INT_MAX && Cx[child]!=INT_MAX && Cx[child] > (Cx[node] - Hx[node])+ W[start]+ Hx[child] ){ Cx[child] = (Cx[node] - Hx[node])+ W[start]+ Hx[child]; __threadfence(); parent[child] = node; addFlag[child]=1; } else if((Cx[node]==INT_MAX && parent[child]==node )|| ( parent[child]==node && (Cx[child] < Cx[node] - Hx[node]+ diff_W[start]+ Hx[child]) ) ){ //use back edges int rstart = rev_offset[child]; int rend = E; if(child!=N-1) rend = rev_offset[child+1]; //there is always one parent that is node. Cx[child] = INT_MAX; parent[child]=-1; while(rstart < rend){ int p = rev_edges[rstart]; if(p<0 || p ==child){ rstart++; continue; } int weight = rev_weight[rstart]; int flag_cycle = false; //check parent doesn't contain child int ancestor = parent_old[p]; while(ancestor!=-1){ if(ancestor==child) flag_cycle = true; ancestor = parent_old[ancestor]; } if(!flag_cycle && Cx[p]!=INT_MAX && Cx[child] > (Cx[p]-Hx[p])+weight+Hx[child] ){ Cx[child] = (Cx[p]-Hx[p] )+weight+Hx[child]; parent[child] = p; } rstart++; } rstart = rev_diff_offset[child]; rend = dE; if(child!=N-1) rend = rev_diff_offset[child+1]; while(rstart < rend){ int p = rev_diff_edges[rstart]; if(p<0 || p==child){ rstart++; continue; } int weight = rev_diff_weight[rstart]; int flag_cycle = false; //check parent doesn't contain child int ancestor = parent_old[p]; while(ancestor!=-1){ if(ancestor==child){ flag_cycle = true; break; } ancestor = parent_old[ancestor]; } if(!flag_cycle && Cx[p]!=INT_MAX && Cx[child] > (Cx[p]-Hx[p])+weight+Hx[child] ){ Cx[child] = (Cx[p]-Hx[p] )+weight+Hx[child]; parent[child] = p; } rstart++; } addFlag[child]=1; } //end critical section leaveLoop = true; atomicCAS(&lock[child],1,0); } __syncthreads(); } start++; } } } //do in 1 thread template <class U> __global__ void insertDest(unsigned int* PQ,unsigned int* PQ_size,U* Cx,int dest,int* openList){ int id = 0; int front = 0; if(openList[dest]==-1){ PQ[front+PQ_size[id]]= dest; PQ_size[id]+=1; //add in openList openList[dest] = id; if(PQ_size[id]>1){ int index = PQ_size[id]-1; while(index>0){ if(Cx[PQ[front+ (index-1)/2]] > Cx[PQ[front+index]]){ int swap = PQ[front+index]; PQ[front+index]=PQ[front+ (index-1)/2]; PQ[front+ (index-1)/2] = swap; index = (index-1)/2; } else break; } } } } template <class U> __global__ void getCx(U* Cx,int dest,U* val){ int id = blockIdx.x*blockDim.x+threadIdx.x; if(id==0){ *val = Cx[dest]; } } #endif
19,072
#include <iostream> #include <fstream> #include <algorithm> #include <cmath> #include <ctime> #include <cuda.h> #include <cuda_runtime.h> //#define WRITE_TO_FILE #define NX 4 #define NY 256 using namespace std; typedef double(*func2)(double,double); typedef double(*func3)(double,double,double); //Обработчик ошибок static void HandleError(cudaError_t err, const char *file, int line) { if (err != cudaSuccess) { printf("%s in %s at line %d\n", cudaGetErrorString(err), file, line); exit(EXIT_FAILURE); } } #define HANDLE_ERROR( error ) (HandleError( error, __FILE__, __LINE__ )) #define PSI(x,y) 0.0 #define F(x,y,t) 0.0 #define PHI(x,y) __sinf(M_PI*x)*__sinf(M_PI*y) __global__ void first_layer_kernel(double *U,double *Uprev,double tau,double a, int N1n,int N2n, double h1, double h2/*,func2 phi,func2 psi,func3 f*/) { int i=threadIdx.x+blockIdx.x*blockDim.x+1; int j=threadIdx.y+blockIdx.y*blockDim.y+1; if((i < N1n-1)&&(j<N2n-1)) U[i*N2n+j]=Uprev[i*N2n+j]+tau*PSI(i*h1,j*h2)+ tau*tau*0.5*F(i*h1,j*h2,0.0)+ a*a*tau*tau*0.5*((PHI((i+1)*h1,j*h2)-2.0*PHI(i*h1,j*h2)+PHI((i-1)*h1,j*h2))/(h1*h1)+(PHI(i*h1,(j+1)*h2)-2.0*PHI(i*h1,j*h2)+PHI(i*h1,(j-1)*h2))/(h2*h2)); } __global__ void main_kernel(double *U,double *Uprev,double *Unext,double tau,double a,double t, int N1n,int N2n, double h1, double h2/*,func2 phi,func2 psi,func3 f*/) { int i=threadIdx.x+blockIdx.x*blockDim.x+1; int j=threadIdx.y+blockIdx.y*blockDim.y+1; if((i < N1n-1)&&(j<N2n-1)) Unext[i*N2n+j]=2.0*U[i*N2n+j]-Uprev[i*N2n+j]+a*a*tau*tau*((U[(i+1)*N2n+j]-2.0*U[i*N2n+j]+U[(i-1)*N2n+j])/(h1*h1)+(U[i*N2n+(j+1)]-2.0*U[i*N2n+j]+U[i*N2n+(j-1)])/(h2*h2))+F(i*h1,j*h2,t); } double solveGPU(double a,double L1,double L2,double T,double tau,int N1,int N2,func2 phi,func2 psi,func3 f) { double *Unext,*U,*Uprev,*Uloc; double h1=L1/N1,h2=L2/N2; int N1n=N1+1,N2n=N2+1; double t=tau; float gputime=0.0; size_t size=N1n*N2n*sizeof(double); dim3 threads(NX,NY,1),blocks((N1-1)%NX==0?(N1-1)/NX:(N1-1)/NX+1,(N2-1)%NY==0?(N2-1)/NY:(N2-1)/NY+1,1); Uloc=new double[N1n*N2n]; HANDLE_ERROR( cudaMalloc(&U,size) ); HANDLE_ERROR( cudaMalloc(&Unext,size) ); HANDLE_ERROR( cudaMalloc(&Uprev,size) ); #ifdef WRITE_TO_FILE ofstream ofile("../membr2dexpl/datagpu.dat"); ofile.precision(16); #endif //Нулевой временной слой for(int i=0;i<N1n;i++) { for(int j=0;j<N2n;j++) { Uloc[i*N2n+j]=phi(i*h1,j*h2); #ifdef WRITE_TO_FILE ofile<<Uloc[i*N2n+j]<<' '; #endif } #ifdef WRITE_TO_FILE ofile<<endl; #endif } #ifdef WRITE_TO_FILE ofile<<endl; ofile<<endl; #endif HANDLE_ERROR( cudaMemcpy(Uprev,Uloc,size,cudaMemcpyHostToDevice) ); HANDLE_ERROR( cudaMemcpy(U,Uprev,size,cudaMemcpyDeviceToDevice) ); HANDLE_ERROR( cudaMemcpy(Unext,Uprev,size,cudaMemcpyDeviceToDevice) ); //Первый временной слой cudaEvent_t start,stop; HANDLE_ERROR( cudaEventCreate(&start) ); HANDLE_ERROR( cudaEventCreate(&stop) ); HANDLE_ERROR( cudaEventRecord(start) ); first_layer_kernel<<<blocks,threads>>>(U,Uprev,tau,a,N1n,N2n,h1,h2/*,phi,psi,f*/); HANDLE_ERROR( cudaGetLastError() ); HANDLE_ERROR( cudaDeviceSynchronize() ); #ifdef WRITE_TO_FILE HANDLE_ERROR( cudaMemcpy(Uloc,U,size,cudaMemcpyDeviceToHost) ); for(int i=0;i<N1n;i++) { for(int j=0;j<N2n;j++) ofile<<Uloc[i*N2n+j]<<' '; ofile<<endl; } ofile<<endl; ofile<<endl; #endif //Основной цикл while(t<T-0.5*tau) { main_kernel<<<blocks,threads>>>(U,Uprev,Unext,tau,a,t,N1n,N2n,h1,h2/*,phi,psi,f*/); HANDLE_ERROR( cudaGetLastError() ); HANDLE_ERROR( cudaDeviceSynchronize() ); #ifdef WRITE_TO_FILE HANDLE_ERROR( cudaMemcpy(Uloc,Unext,size,cudaMemcpyDeviceToHost) ); for(int i=0;i<N1n;i++) { for(int j=0;j<N2n;j++) ofile<<Uloc[i*N2n+j]<<' '; ofile<<endl; } ofile<<endl; ofile<<endl; #endif t+=tau; swap(U,Unext); swap(Uprev,Unext); } HANDLE_ERROR( cudaMemcpy(Uloc,Unext,size,cudaMemcpyDeviceToHost) ); HANDLE_ERROR( cudaEventRecord(stop) ); HANDLE_ERROR( cudaEventSynchronize(stop) ); HANDLE_ERROR( cudaEventElapsedTime(&gputime,start,stop) ); HANDLE_ERROR( cudaFree(U) ); HANDLE_ERROR( cudaFree(Unext) ); HANDLE_ERROR( cudaFree(Uprev) ); HANDLE_ERROR( cudaEventDestroy(start) ); HANDLE_ERROR( cudaEventDestroy(stop) ); delete[] Uloc; #ifdef WRITE_TO_FILE ofile.close(); #endif return (double)gputime/1000.0; } double solveCPU(double a,double L1,double L2,double T,double tau,int N1,int N2,func2 phi,func2 psi,func3 f) { double *Unext,*U,*Uprev; double h1=L1/N1,h2=L2/N2; int N1n=N1+1,N2n=N2+1; double t=tau; double cputime=0.0; U=new double[N1n*N2n]; Unext=new double[N1n*N2n]; Uprev=new double[N1n*N2n]; #ifdef WRITE_TO_FILE ofstream ofile("../membr2dexpl/datacpu.dat"); ofile.precision(16); #endif //Нулевой временной слой for(int i=0;i<N1n;i++) { for(int j=0;j<N2n;j++) { Uprev[i*N2n+j]=phi(i*h1,j*h2); #ifdef WRITE_TO_FILE ofile<<Uprev[i*N2n+j]<<' '; #endif } #ifdef WRITE_TO_FILE ofile<<endl; #endif } #ifdef WRITE_TO_FILE ofile<<endl; ofile<<endl; #endif //Первый временной слой cputime-=(double)clock(); for(int i=0;i<N1n;i++) { for(int j=0;j<N2n;j++) { if((i==0)||(j==0)||(i==N1)||(j==N2)) { U[i*N2n+j]=Uprev[i*N2n+j]; Unext[i*N2n+j]=Uprev[i*N2n+j]; } else { U[i*N2n+j]=Uprev[i*N2n+j]+tau*psi(i*h1,j*h2)+ tau*tau*0.5*f(i*h1,j*h2,0.0)+ a*a*tau*tau*0.5*((phi((i+1)*h1,j*h2)-2.0*phi(i*h1,j*h2)+phi((i-1)*h1,j*h2))/(h1*h1)+(phi(i*h1,(j+1)*h2)-2.0*phi(i*h1,j*h2)+phi(i*h1,(j-1)*h2))/(h2*h2)); } #ifdef WRITE_TO_FILE ofile<<U[i*N2n+j]<<' '; #endif } #ifdef WRITE_TO_FILE ofile<<endl; #endif } #ifdef WRITE_TO_FILE ofile<<endl; ofile<<endl; #endif //Основной цикл while(t<T-0.5*tau) { for(int i=1;i<N1n-1;i++) for(int j=1;j<N2n-1;j++) Unext[i*N2n+j]=2.0*U[i*N2n+j]-Uprev[i*N2n+j]+a*a*tau*tau*((U[(i+1)*N2n+j]-2.0*U[i*N2n+j]+U[(i-1)*N2n+j])/(h1*h1)+(U[i*N2n+(j+1)]-2.0*U[i*N2n+j]+U[i*N2n+(j-1)])/(h2*h2))+f(i*h1,j*h2,t); #ifdef WRITE_TO_FILE for(int i=0;i<N1n;i++) { for(int j=0;j<N2n;j++) ofile<<Unext[i*N2n+j]<<' '; ofile<<endl; } ofile<<endl; ofile<<endl; #endif t+=tau; swap(U,Unext); swap(Uprev,Unext); } cputime+=(double)clock(); cputime/=(double)CLOCKS_PER_SEC; delete[] U; delete[] Unext; delete[] Uprev; #ifdef WRITE_TO_FILE ofile.close(); #endif return cputime; } __host__ __device__ double zero3(double a,double b,double c) { return 0.0; } __host__ __device__ double zero2(double a,double b) { return 0.0; } __host__ __device__ double init(double x, double y) { return sin(M_PI*x)*sin(M_PI*y); } __host__ __device__ double init2(double x, double y) { double sigma=0.1; return exp(-((x-0.5)*(x-0.5)+(y-0.5)*(y-0.5))/2.0/sigma/sigma); } __host__ __device__ double init3(double x, double y) { return (x-0.5)*(x-0.5)+(y-0.5)*(y-0.5)<0.25*0.25?1.0:0.0; } int main(int argc, char *argv[]) { double cputime,gputime; cputime=solveCPU(1.0,1.0,1.0,0.1,0.0001,1000,1000,init,zero2,zero3); //cputime=solveCPU(1.0,1.0,1.0,10,0.01,25,25,init2,zero2,zero3); cout<<"CPU time: "<<cputime<<endl; gputime=solveGPU(1.0,1.0,1.0,0.1,0.0001,1000,1000,init,zero2,zero3); //gputime=solveGPU(1.0,1.0,1.0,10,0.01,25,25,init,zero2,zero3); cout<<"GPU time: "<<gputime<<endl; cout<<"Ratio: "<<cputime/gputime<<endl; return 0; }
19,073
// #include <bits/stdc++.h> #include <stdio.h> #include <stdlib.h> #include <iostream> #include <sstream> #include <string> #include <vector> #include <algorithm> #include <climits> #include <thrust/swap.h> #include <thrust/extrema.h> #include <thrust/functional.h> #include <thrust/host_vector.h> #include <thrust/device_vector.h> using namespace std; typedef long long ll; #define CUDA_ERROR(err) { \ if (err != cudaSuccess) { \ fprintf(stderr, "ERROR: CUDA failed in %s:%d: %s\n", __FILE__, __LINE__, cudaGetErrorString(err)); \ return(1); \ } \ } \ #define NUM_BLOCKS 10 #define BLOCK_SIZE 1024 __device__ void swap_step(int* nums, int* tmp, int size, int start, int stop, int step, int i) { // Using shared memory to store blocks and sort them __shared__ int sh_array[BLOCK_SIZE]; // Step for bitonic merge inside merging for (int shift = start; shift < stop; shift += step) { // New start pointer tmp = nums + shift; // Right side if (i >= BLOCK_SIZE / 2) sh_array[i] = tmp[BLOCK_SIZE * 3 / 2 - 1 - i]; else sh_array[i] = tmp[i]; __syncthreads(); // From half for (int j = BLOCK_SIZE / 2; j > 0; j /= 2) { unsigned int XOR = i ^ j; // The threads with the lowest ids sort the array if (XOR > i) { if ((i & BLOCK_SIZE) != 0) { // Step descending, swap(i, XOR) if (sh_array[i] < sh_array[XOR]) thrust::swap(sh_array[i], sh_array[XOR]); } else { // Step ascending, swap(i, XOR) if (sh_array[i] > sh_array[XOR]) thrust::swap(sh_array[i], sh_array[XOR]); } } __syncthreads(); } // Back from shared to temporary tmp[i] = sh_array[i]; } } __global__ void kernel_bitonic_merge_step(int* nums, int size, bool is_odd, bool flag) { // Temporary array for splitting into blocks int* tmp = nums; // Every thread gets exactly one value in the unsorted array unsigned int i = threadIdx.x; int id_block = blockIdx.x; int offset = gridDim.x; // For odd step if(is_odd) { swap_step(nums, tmp, size, (BLOCK_SIZE / 2) + id_block * BLOCK_SIZE, size - BLOCK_SIZE, offset * BLOCK_SIZE, i); } else { // For even step swap_step(nums, tmp, size, id_block * BLOCK_SIZE, size, offset * BLOCK_SIZE, i); } } __global__ void bitonic_sort_step(int *nums, int j, int k, int size) { // Using shared memory to store blocks and sort them __shared__ int sh_array[BLOCK_SIZE]; // Temporary array for splitting into blocks int* tmp = nums; // Every thread gets exactly one value in the unsorted array unsigned int i = threadIdx.x; int id_block = blockIdx.x; int offset = gridDim.x; // Step for bitonic sort for (int shift = id_block * BLOCK_SIZE; shift < size; shift += offset * BLOCK_SIZE) { // New start pointer tmp = nums + shift; // Store in shared memory sh_array[i] = tmp[i]; __syncthreads(); // From half for (j = k / 2; j > 0; j /= 2) { unsigned int XOR = i ^ j; // The threads with the lowest ids sort the array if (XOR > i) { if ((i & k) != 0) { // Step descending, swap(i, XOR) if (sh_array[i] < sh_array[XOR]) thrust::swap(sh_array[i], sh_array[XOR]); } else { // Step ascending, swap(i, XOR) if (sh_array[i] > sh_array[XOR]) thrust::swap(sh_array[i], sh_array[XOR]); } } __syncthreads(); } // Back from shared to temporary tmp[i] = sh_array[i]; } } int main(int argc, char *argv[]) { ios_base::sync_with_stdio(false); cin.tie(nullptr); cout.tie(nullptr); int size, upd_size; // Allocating + inputting // scanf("%d", &size); fread(&size, sizeof(int), 1, stdin); fprintf(stderr, "%d\n", size); // To the degree of 2^n (1024 max) upd_size = ceil((double)size / BLOCK_SIZE) * BLOCK_SIZE; int* data = (int*)malloc(sizeof(int) * upd_size); int* dev_data; CUDA_ERROR(cudaMalloc((void**)&dev_data, sizeof(int) * upd_size)); // for (int i = 0; i < size; ++i) { // fread(&size, sizeof(int), 1, stdin); // scanf("%d", &data[i]); // // fprintf(stderr, "%d ", data[i]); // } // fprintf(stderr, "\n"); fread(data, sizeof(int), size, stdin); for (int i = size; i < upd_size; ++i) { data[i] = INT_MAX; } // Copy to device CUDA_ERROR(cudaMemcpy(dev_data, data, sizeof(int) * upd_size, cudaMemcpyHostToDevice)); //////////////////////////////////////////////////////////////////////////////////////// // Pre sort of all blocks by bitonic sort // Main step for (int k = 2; k <= upd_size; k *= 2) { if (k > BLOCK_SIZE) break; // Merge and split step for (int j = k / 2; j > 0; j /= 2) { bitonic_sort_step<<<NUM_BLOCKS, BLOCK_SIZE>>>(dev_data, j, k, upd_size); CUDA_ERROR(cudaGetLastError()); } } //////////////////////////////////////////////////////////////////////////////////////// /* Implementation of odd-even sort Sort of buckets with bitonic merge inside | 1 3 5 7 | 2 4 6 8 | -> | 1 2 3 4 5 6 7 8| (size == 8) Including 2 steps merge + splitting */ for (int i = 0; i < 2 * (upd_size / BLOCK_SIZE); ++i) { kernel_bitonic_merge_step<<<NUM_BLOCKS, BLOCK_SIZE>>>(dev_data, upd_size, (bool)(i % 2), true); CUDA_ERROR(cudaGetLastError()); } CUDA_ERROR(cudaMemcpy(data, dev_data, sizeof(int) * upd_size, cudaMemcpyDeviceToHost)) CUDA_ERROR(cudaFree(dev_data)); for (int i = 0; i < size; ++i) { fprintf(stderr, "%d ", data[i]); } fprintf(stderr, "\n"); fwrite(data, sizeof(int), size, stdout); free(data); return 0; }
19,074
/* CUDA kernels and functions Kurt Kaminski 2016 */ #ifndef __FLUID_KERNELS__ #define __FLUID_KERNELS__ #include <cuda_runtime.h> ///////////////////////////////////////////////////////////////////////////////////////////////// ///////////////////////////////////////////////////////////////////////////////////////////////// // Get 1d index from 2d coords //__device__ int //IX(int x, int y) //{ // return x + (y * blockDim.x * gridDim.x); //} // //__device__ int //getX(int w) //{ // int x = threadIdx.x + (blockIdx.x * blockDim.x); // //if (x >= w) x = 0; if (x < 0) x = w-1; // if (x >= w) return 0; // else return x; //} // //__device__ int //getY(int h) //{ // int y = threadIdx.y + (blockIdx.y * blockDim.y); // //if (y >= h) y = 0; if (y < 0) y = h-1; // if (y >= h) return 0; // else return y; //} // //__global__ void //ClearArray(float *field, float value, int w, int h) //{ // int x = getX(w); // int y = getY(h); // int id = IX(x,y); // // field[id] = value; //} // //__global__ void //ClearArray(int *field, float value, int w, int h) //{ // int x = getX(w); // int y = getY(h); // int id = IX(x,y); // // field[id] = value; //} // //__device__ int //clamp(int i) //{ // if (i < 0) // i = 0; // if (i > 255) // i = 255; // // return i; //} // //// Functions for converting to/from a int (4 bytes, 1 byte per RGBA, which are in the range 0-255) //// to 4 floats in the range 0.0-1.0 //// Note how the data is stored in BGRA format due to how its stored on the GPU. //__device__ int //rgbaToInt(float r, float g, float b, float a) //{ // return // (clamp((int)(a * 255.0f)) << 24) | // (clamp((int)(r * 255.0f)) << 16) | // (clamp((int)(g * 255.0f)) << 8) | // (clamp((int)(b * 255.0f)) << 0); //} // //__device__ //void intToRgba(int pixel, float &r, float &g, float &b, float &a) //{ // b = float(pixel&0xff) / 255.0f; // g = float((pixel>>8)&0xff) / 255.0f; // r = float((pixel>>16)&0xff) / 255.0f; // a = float((pixel>>24)&0xff) / 255.0f; //} // // //__global__ void //sampleKernel( int* src, int inw, int inh, int *dest, int w, int h ) //{ // int tx = threadIdx.x; // int ty = threadIdx.y; // int bw = blockDim.x; // int bh = blockDim.y; // int x = blockIdx.x*bw + tx; // int y = blockIdx.y*bh + ty; // // // If the resolution isn't a multiple of the grid/thread size, or the resolutions don't match // // we need to make sure we arn't reading or writting beyond the bounds of the data // if (x >= inw || y >= inh) // { // if (x < w && y < h) // dest[y * w + x] = rgbaToInt(0.0f, 0.0f, 0.0f, 1.0); // return; // } // else if (x >= w || y >= h) // { // return; // } // else // { // int pixel = src[y * inw + x]; // float r,g,b,a; // intToRgba(pixel, r, g, b, a); // // // Simple monochrome operation // float v = r*0.3f + g*0.6f + b*0.1f; // dest[y * w + x] = rgbaToInt(v, v, v, a); // } //} ///////////////////////////////////////////////////////////////////////////////////////////////// ///////////////////////////////////////////////////////////////////////////////////////////////// __device__ int clamp(int i) { if (i < 0) i = 0; if (i > 255) i = 255; return i; } __device__ float clamp(float i, float min, float max) { if (i < min) i = min; if (i > max) i = max; return i; } __device__ float fitRange(float valueIn, float baseMin, float baseMax, float limitMin, float limitMax) { return ((limitMax - limitMin) * (valueIn - baseMin) / (baseMax - baseMin)) + limitMin; } // Get 1d index from 2d coords __device__ int IX(int x, int y) { return x + (y * blockDim.x * gridDim.x); } __device__ int getX(int w) { int x = threadIdx.x + (blockIdx.x * blockDim.x); //if (x >= w) x = 0; if (x < 0) x = w-1; if (x >= w) return 0; else return x; } __device__ int getY(int h) { int y = threadIdx.y + (blockIdx.y * blockDim.y); //if (y >= h) y = 0; if (y < 0) y = h-1; if (y >= h) return 0; else return y; } // Returns true if within the bounds of both the container edges and a user-defined boundary __device__ bool checkBounds(float *_boundary, int x, int y, int w, int h) { if (x > 1 && x < w-2 && y > 1 && y < h-2 && _boundary[4*IX(x,y)+0] < 1 ){ return true; } else { return false; } } __device__ bool checkBounds(int x, int y, int w, int h) { if (x > 1 && x < w-2 && y > 1 && y < h-2){ return true; } else { return false; } } // Functions for converting to/from a int (4 bytes, 1 byte per RGBA, which are in the range 0-255) // to 4 floats in the range 0.0-1.0 // Note how the data is stored in BGRA format due to how its stored on the GPU. __device__ int rgbaToInt(float r, float g, float b, float a) { return (clamp((int)(a * 255.0f)) << 24) | (clamp((int)(r * 255.0f)) << 16) | (clamp((int)(g * 255.0f)) << 8) | (clamp((int)(b * 255.0f)) << 0); } __device__ void intToRgba(int pixel, float &r, float &g, float &b, float &a) { b = float(pixel&0xff) / 255.0f; g = float((pixel>>8)&0xff) / 255.0f; r = float((pixel>>16)&0xff) / 255.0f; a = float((pixel>>24)&0xff) / 255.0f; } __device__ void rgbaToColor(float *dest, int id, float r, float g, float b, float a) { dest[4*id+0] = b; dest[4*id+1] = g; dest[4*id+2] = r; dest[4*id+3] = a; } // Set boundary conditions __device__ void set_bnd( int b, int x, int y, float *field, float *boundary, int w, int h) { int sz = w*h; int id = IX(x,y); bool outOfBnd = boundary[4*id+0] > 0.0 ? true : false; //if (boundary[4*id+0] > 0.0) outOfBnd = true; //if (x==0) field[id] = b==1 ? -1*field[IX(1,y)] : field[IX(1,y)]; //if (x==w-1) field[id] = b==1 ? -1*field[IX(w-2,y)] : field[IX(w-2,y)]; //if (y==0) field[id] = b==2 ? -1*field[IX(x,1)] : field[IX(x,1)]; //if (y==h-1) field[id] = b==2 ? -1*field[IX(x,h-2)] : field[IX(x,h-2)]; if (x==0 || outOfBnd) field[id] = b==1 ? -1*field[IX(1,y)] : -1 * field[IX(1,y)]; if (x==w-1 || outOfBnd) field[id] = b==1 ? -1*field[IX(w-2,y)] : -1 * field[IX(w-2,y)]; if (y==0 || outOfBnd) field[id] = b==2 ? -1*field[IX(x,1)] : -1 * field[IX(x,1)]; if (y==h-1 || outOfBnd) field[id] = b==2 ? -1*field[IX(x,h-2)] : -1 * field[IX(x,h-2)]; //if (outOfBnd){ // field[id] = -1*field[id]; // field[IX(x+1,y)] = -1*field[IX(x+1,y)]; // field[IX(x-1,y)] = -1*field[IX(x-1,y)]; // field[IX(x,y+1)] = -1*field[IX(x,y+1)]; // field[IX(x,y-1)] = -1*field[IX(x,y-1)]; //} if (id == 0) field[id] = 0.5*(field[IX(1,0)]+field[IX(0,1)]); // southwest if (id == sz-w) field[id] = 0.5*(field[IX(1,h-1)]+field[IX(0, h-2)]); // northwest if (id == w-1) field[id] = 0.5*(field[IX(w-2,0)]+field[IX(w-1,1)]); // southeast if (id == sz-1) field[id] = 0.5*(field[IX(w-2,h-1)]+field[IX(w-1,h-2)]); // northeast } __global__ void DrawSquare( float *field, float value, int w, int h ) { int x = getX(w); int y = getY(h); int id = IX(x,y); float posX = (float)x/w; float posY = (float)y/h; if ( posX < .92 && posX > .45 && posY < .51 && posY > .495 ) { field[id] = value; } } //__global__ void //DrawBnd( int *boundary, int w, int h ) { // int x = getX(w); // int y = getY(h); // int id = IX(x,y); // // float posX = (float)x/w; // float posY = (float)y/h; // if ( posX < .82 && posX > .70 && posY < .33 && posY > .21 ) { // boundary[id] = 1; // } // else boundary[id] = 0; //} __global__ void SetBoundary( int b, float *field, float *boundary, int w, int h ) { int x = getX(w); int y = getY(h); set_bnd(b, x, y, field, boundary, w, h); } __global__ void getSum( float *_data, float _sum, int w, int h ) { int x = getX(w); int y = getY(h); _sum += _data[IX(x,y)]; } __global__ void ClearArray(float *field, float value, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); field[id] = value; } __global__ void ClearArray(int *field, float value, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); field[id] = value; } __global__ void MapArray(float *field, float value, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); field[id] = float(x) * value; } // How can I template these? __global__ void AddFromUI ( float *field, float value, float dt, int x_coord, int y_coord, int w, int h ) { int x = getX(w); int y = getY(h); int id = IX(x,y); if (x>x_coord-5 && x<x_coord+5 && y>y_coord-5 && y<y_coord+5){ // if (x == x_coord && y==y_coord){ field[id] += value * dt; } else return; } __global__ void AddFromUI ( float *field, float *valueUI, int index, float dt, int w, int h ) { int x = getX(w); int y = getY(h); int id = IX(x,y); field[id] += valueUI[4*id+index] * dt; } __global__ void AddObstacleVelocity ( float *u, float *v, float *obstacle, float dt, int w, int h ) { int x = getX(w); int y = getY(h); int id = IX(x,y); u[id] += obstacle[4*id+2] * dt; //red v[id] += obstacle[4*id+1] * dt; //green } __global__ void SetFromUI ( float *A, float *B, float *valueUI, int w, int h ) { int x = getX(w); int y = getY(h); int id = IX(x,y); float v1 = valueUI[4*id+2]; //red float v2 = valueUI[4*id+1]; //green if (v1 > 0.0) A[id] = v1; if (v2 > 0.0) B[id] = v2; } __global__ void MakeSource(int *src, float *dest, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); int pixel = src[id]; float r,g,b,a; intToRgba(pixel, r, g, b, a); dest[id] = r; } // *!* This is currently only grabbing the red channel *!* __global__ void MakeSource(int *src, int *dest, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); int pixel = src[id]; float r,g,b,a; intToRgba(pixel, r, g, b, a); dest[id] = src[id]&0xff/255; } __global__ void AddSource(float *field, float *source, float dt, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); field[id] += (dt * source[id]); } __global__ void MakeColor(float *src, int *dest, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); dest[id] = rgbaToInt(src[id], src[id], src[id], 1.0); //dest[id] = rgbaToInt(1.0, src[id], src[id], 1.0); } __global__ void MakeColor(float *src0, float *src1, float *src2, float *dest, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); rgbaToColor(dest, id, src0[id], src1[id], src2[id], 1.0); } __global__ void MakeColor(float *src0, float *src1, float *src2, float *src3, float *dest, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); rgbaToColor(dest, id, src0[id], src1[id], src2[id], src3[id]); } __global__ void MakeColorLong( float *r1, float *g1, float *b1, float *a1, float *r2, float *g2, float *b2, float *a2, float *dest, int w, int h, int stride) { int idIn = getY(h) * w + (getX(w)); int idOut = getY(h) * (w*stride) + (getX(w*stride)); //Left image dest[4*idOut+0] = b1[idIn]; dest[4*idOut+1] = g1[idIn]; dest[4*idOut+2] = r1[idIn]; dest[4*idOut+3] = a1[idIn]; //Right image idOut += w; dest[4*idOut+0] = b2[idIn]; dest[4*idOut+1] = g2[idIn]; dest[4*idOut+2] = r2[idIn]; dest[4*idOut+3] = 1.0; } __device__ float bilerp(float *src, float _i, float _j, int w, int h) { int i0, j0, i1, j1; float s0, t0, s1, t1; float i; float j; // fit bounds i = (_i < 0.5f) ? 0.5f : _i; i = (_i > float(w)-2.0+0.5f) ? float(w)-2.0+0.5f : _i; j = (_j > float(h)-2.0+0.5f) ? float(h)-2.0+0.5f : _j; j = (_j < 0.5) ? 0.5 : _j; //if (_i < 0.5f) i = 0.5f; //else if (_i > float(w)-2.0+0.5f) i = float(w)-2.0+0.5f; //if (j < 0.5f) j = 0.5f; //else if (j > float(h)-2.0+0.5f) j = float(h)-2.0+0.5f; // bilinear interpolation i0 = int(i); i1 = i0+1; j0 = int(j); j1 = j0+1; s1 = (float)i-i0; s0 = (float)1-s1; t1 = (float)j-j0; t0 = (float)1-t1; return (float) s0*(t0*src[IX(i0,j0)] + t1*src[IX(i0,j1)])+ s1*(t0*src[IX(i1,j0)] + t1*src[IX(i1,j1)]); } __global__ void Advect (float *vel_u, float *vel_v, float *src_u, float *src_v, float *boundary, float *dest_u, float *dest_v, float timeStep, float diff, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); //if (x > 1 && x < w-1 && y > 1 && y < h-1){ if (checkBounds(boundary, x, y, w, h)) { float dt0 = (float)timeStep * float(w-2); float i = float(x) - dt0 * vel_u[id]; float j = float(y) - dt0 * vel_v[id]; dest_u[id] = diff * bilerp(src_u, i, j, w, h); dest_v[id] = diff * bilerp(src_v, i, j, w, h); } else { dest_u[id] = 0.0; dest_v[id] = 0.0; } } __global__ void Advect (float *vel_u, float *vel_v, float *src, float *boundary, float *dest, float timeStep, float diff, bool skipBilerp, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); if (checkBounds(boundary, x, y, w, h)) { //if (x > 1 && x < w-1 && y > 1 && y < h-1){ float dt0 = (float)timeStep * float(w-2); float i = float(x) - dt0 * vel_u[id]; float j = float(y) - dt0 * vel_v[id]; dest[id] = diff * bilerp(src, i, j, w, h); } else { dest[id] = 0.0; } } __device__ float curl(int i, int j, float *u, float *v) { float du_dy = (u[IX(i, j+1)] - u[IX(i, j-1)]) * 0.5f; float dv_dx = (v[IX(i+1, j)] - v[IX(i-1, j)]) * 0.5f; return du_dy - dv_dx; } __global__ void vorticityConfinement(float *u, float *v, float *Fvc_x, float *Fvc_y, float *_boundary, float curlAmt, float dt, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); //if (x>1 && x<w-2 && y>1 && y<h-2){ if (checkBounds(_boundary, x, y, w, h)) { // Calculate magnitude of curl(u,v) for each cell. (|w|) // curl[I(i, j)] = Math.abs(curl(i, j)); // Find derivative of the magnitude (n = del |w|) float dw_dx = ( abs(curl(x+1,y, u, v)) - abs(curl(x-1,y, u, v)) ) * 0.5f; float dw_dy = ( abs(curl(x,y+1, u, v)) - abs(curl(x,y-1, u, v)) ) * 0.5f; // Calculate vector length. (|n|) // Add small factor to prevent divide by zeros. float length = sqrt(dw_dx * dw_dx + dw_dy * dw_dy); length = length + 0.000001f; //if (length == 0.0) length -= 0.000001f; // N = ( n/|n| ) float vel = curl(x, y, u, v); // N x w // 0.5 = curl amount Fvc_y[id] = Fvc_y[id] + ((dw_dx/length) * vel * dt * curlAmt); Fvc_x[id] = Fvc_x[id] + ((dw_dy/length) * -vel * dt * curlAmt); } else { Fvc_x[id] = 0.0; Fvc_y[id] = 0.0; } } __global__ void ApplyBuoyancy( float *vel_u, float *vel_v, float *temp, float *dens, float *dest_u, float *dest_v, float ambientTemp, float buoy, float weight, float dt, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); if (checkBounds(x, y, w, h)) { dest_u[id] = vel_u[id]; dest_v[id] = vel_v[id]; float T = temp[id]; float Sigma = buoy; float Kappa = weight; if (T > ambientTemp) { float D = dens[id]; dest_u[id] += (dt * (T - ambientTemp) * Sigma - D * Kappa) * 0; dest_v[id] += (dt * (T - ambientTemp) * Sigma - D * Kappa) * .1; } else { return; } } } __global__ void ComputeDivergence( float *u, float *v, float *boundary, float *dest, int w, int h ) { int x = getX(w); int y = getY(h); int id = IX(x,y); //if (x > 2 && x < w-2 && y > 2 && y < h-2){ if (checkBounds(x, y, w, h)){ float vN, vS, vE, vW; // Find neighboring obstacles: float oN = boundary[4 * IX(x, y+1) + 0]; float oS = boundary[4 * IX(x, y-1) + 0]; float oE = boundary[4 * IX(x+1, y) + 0]; float oW = boundary[4 * IX(x-1, y) + 0]; // Find neighboring velocities, use center pressure for solid cells: vN = (oN > 0.0) ? boundary[4 * IX(x, y+1) + 1] : v[IX(x, y+1)]; vS = (oS > 0.0) ? boundary[4 * IX(x, y-1) + 1] : v[IX(x, y-1)]; vE = (oE > 0.0) ? boundary[4 * IX(x+1, y) + 2] : u[IX(x+1, y)]; vW = (oW > 0.0) ? boundary[4 * IX(x-1, y) + 2] : u[IX(x-1, y)]; //float cellSize = 1.0; //dest[id] = (0.5 / cellSize) * ( u[IX(x+1, y)] - u[IX(x-1, y)] + v[IX(x, y+1)] - v[IX(x, y-1)] ); //dest[id] = 0.5 * ( (u[IX(x+1, y)] - u[IX(x-1, y)]) + (v[IX(x, y+1)] - v[IX(x, y-1)]) ) ; dest[id] = 0.5 * ( vE - vW + vN - vS ) / float(w-2); } else { return; } } __global__ void Jacobi( float *p, float *divergence, float *boundary, float *dest, int w, int h ) //Jacobi( float *p, float *divergence, cudaTextureObject_t txBoundary, float *dest, int w, int h ) { int x = getX(w); int y = getY(h); int id = IX(x,y); if (checkBounds(x, y, w, h)){ // Find neighboring obstacles: float oN = boundary[4 * IX(x, y+1) + 0]; float oS = boundary[4 * IX(x, y-1) + 0]; float oE = boundary[4 * IX(x+1, y) + 0]; float oW = boundary[4 * IX(x-1, y) + 0]; //float oN = tex1Dfetch<float>(txBoundary, 4 * IX(x, y+1) + 0); //float oS = tex1Dfetch<float>(txBoundary, 4 * IX(x, y-1) + 0); //float oE = tex1Dfetch<float>(txBoundary, 4 * IX(x+1, y) + 0); //float oW = tex1Dfetch<float>(txBoundary, 4 * IX(x-1, y) + 0); // Find neighboring pressure, use center pressure for solid cells: //float pC = p[id]; float pN = (oN > 0.0) ? p[id] : p[IX(x, y+1)]; float pS = (oS > 0.0) ? p[id] : p[IX(x, y-1)]; float pE = (oE > 0.0) ? p[id] : p[IX(x+1, y)]; float pW = (oW > 0.0) ? p[id] : p[IX(x-1, y)]; //float cellSize = 1.0; //float Alpha = -cellSize * cellSize; float Alpha = -1.0; float bC = divergence[id]; float InverseBeta = .25; dest[id] = (pW + pE + pS + pN + Alpha * bC) * InverseBeta; } else { return; } } __global__ void SubtractGradient( float *vel_u, float *vel_v, float *p, float *boundary, float *dest_u, float *dest_v, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); if (checkBounds(x, y, w, h)){ // Find neighboring obstacles: float oN = boundary[4 * IX(x, y+1) + 0]; float oS = boundary[4 * IX(x, y-1) + 0]; float oE = boundary[4 * IX(x+1, y) + 0]; float oW = boundary[4 * IX(x-1, y) + 0]; // Use center pressure for solid cells: //float obstU = 0.0; //float obstV = 0.0; //float vMask = 1.0; // Find neighboring pressure: //float pN = p[IX(x, y+1)]; //float pS = p[IX(x, y-1)]; //float pE = p[IX(x+1, y)]; //float pW = p[IX(x-1, y)]; //float pC = p[id]; //if (oN > 0.0) {pN = p[id]; obstV = boundary[4 * IX(x, y+1) + 1]; vMask = 0.0; } //if (oS > 0.0) {pS = p[id]; obstV = boundary[4 * IX(x, y-1) + 1]; vMask = 0.0; } //if (oE > 0.0) {pE = p[id]; obstU = boundary[4 * IX(x+1, y) + 2]; vMask = 0.0; } //if (oW > 0.0) {pW = p[id]; obstU = boundary[4 * IX(x+1, y) + 2]; vMask = 0.0; } float pN = (oN > 0.0) ? p[id] : p[IX(x, y+1)]; float pS = (oS > 0.0) ? p[id] : p[IX(x, y-1)]; float pE = (oE > 0.0) ? p[id] : p[IX(x+1, y)]; float pW = (oW > 0.0) ? p[id] : p[IX(x-1, y)]; float obstV = (oN > 0.0) ? boundary[4 * IX(x, y+1) + 1] : (oS > 0.0) ? boundary[4 * IX(x, y-1) + 1] : 0.0; float obstU = (oE > 0.0) ? boundary[4 * IX(x+1, y) + 2] : (oW > 0.0) ? boundary[4 * IX(x+1, y) + 2] : 0.0; float vMask = (oN > 0.0 || oS > 0.0 || oE > 0.0 || oW > 0.0) ? 0.0 : 1.0; // Enforce the free-slip boundary condition: float old_u = vel_u[id]; float old_v = vel_v[id]; //float cellSize = 1.0; //float GradientScale = 1.125 / cellSize; float GradientScale = 0.5 * float(w-2); float grad_u = (pE - pW) * GradientScale; float grad_v = (pN - pS) * GradientScale; float new_u = old_u - grad_u; float new_v = old_v - grad_v; obstU = 0; obstV = 0; dest_u[id] = (vMask * new_u) + obstU; dest_v[id] = (vMask * new_v) + obstV; } else { dest_u[id] = 0.0; dest_v[id] = 0.0; } } __global__ void Diffusion(float *_chem, float *_lap, float *_boundary, float _difConst, float xLen, float yLen, float dt, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); // have to do this check for non-powers of 2 to work...? //if (checkBounds(_boundary, x, y, w, h)) { if (checkBounds(x, y, w, h)) { // constants //float xLength = (float)x/100.0; //gray-scott //float xLength = (float)x/1.21212; //barkley model float xLength = (float)x/xLen; float yLength = (float)y/yLen; float dx = (float)xLength/(float)x; float dy = (float)yLength/(float)y; float alpha = (float)(_difConst * dt / (float)(dx*dy)); _lap[id] = (float)(-4.0f * _chem[id]) + (float)(_chem[IX(x+1,y)] + _chem[IX(x-1,y)] + _chem[IX(x,y+1)] + _chem[IX(x,y-1)]); _lap[id] = (float)_lap[id]*alpha; } else { return; } } __global__ void AddLaplacian( float *_chem, float *_lap, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); _chem[id] += _lap[id]; } __global__ void React( float *_chemA, float *_chemB, float F, float k, float e, int rdEquation, float *_boundary, float dt, int w, int h) { int x = getX(w); int y = getY(h); int id = IX(x,y); if (checkBounds(_boundary, x, y, w, h)) { //float A = _chemA[id]; //float B = _chemB[id]; // Gray-Scott //float F = 0.05; //float k = 0.0675; //float F = 0.0140; //float k = 0.0490; //float F = 0.0545; //float k = 0.062; //float F = F_input[id]&0xff/255; //F = fitRange(F, 0.0, 1.0, 0.014, 0.066); // //float k = 1.0 - (F_input[id]&0xff/255); //k = fitRange(k, 0.0, 1.0, 0.05, 0.068); //if (A > 1.) A = 1.; if (B > 1.) B = 1.; float A = (_chemA[id] > 1.0) ? 1.0 : _chemA[id]; float B = (_chemB[id] > 1.0) ? 1.0 : _chemB[id]; // Barkley Model if (rdEquation == 1) { float reactionA = A * (1-A) * ((A- (B+k)/F) / e) * dt; float reactionB = (A - B) * dt; _chemA[id] += (dt * reactionA); _chemB[id] += (dt * reactionB); return; } //// Barkley Turbulence //else if (rdEquation == 2) { // float reactionA = A * (1-A) * ((A- (B+k)/F) / e) * dt; // float reactionB = (A*A*A - B) * dt; // _chemA[id] += (dt * reactionA); // _chemB[id] += (dt * reactionB); // return; //} // Gray-Scott else { float reactionA = -A * (B*B) + (F * (1.0-A)); float reactionB = A * (B*B) - (F+k)*B; _chemA[id] += (dt * reactionA); _chemB[id] += (dt * reactionB); return; } } else { _chemA[id] *= -1.0; _chemB[id] *= -1.0; } } //////////////////////////////////////////////////////////////////////// //// Wavelet Turbulence Kernels //////////////////////////////////////////////////////////////////////// //__device__ float //minDx(int x, int y, float *input, int w, int h) { // //const int index = x + y * res[0] + z * res[0] * res[1]; // const int maxx = w-2; // const float FLT_MAX = 1.70141e+38; // // // get grid values // float center = input[IX(x,y)]; // float left = (x <= 1) ? FLT_MAX : input[IX(x-1,y)]; // float right = (x >= maxx) ? FLT_MAX : input[IX(x+1,y)]; // // const float dx = w; // // // get all the derivative estimates // float dLeft = (x <= 1) ? FLT_MAX : (center - left) * dx; // float dRight = (x >= maxx) ? FLT_MAX : (right - center) * dx; // float dCenter = (x <= 1 || x >= maxx) ? FLT_MAX : (right - left) * dx * 0.5f; // // // if it's on a boundary, only one estimate is valid // if (x <= 1) return dRight; // if (x >= maxx) return dLeft; // // // if it's not on a boundary, get the smallest one // float finalD; // finalD = (abs(dCenter) < abs(dRight)) ? dCenter : dRight; // finalD = (abs(finalD) < abs(dLeft)) ? finalD : dLeft; // // return finalD; //} // //__device__ float //minDy(int x, int y, float *input, int w, int h) { // //const int index = x + y * res[0] + z * res[0] * res[1]; // const int maxy = h-2; // const float FLT_MAX = 1.70141e+38; // // // get grid values // float center = input[IX(x,y)]; // float down = (y <= 1) ? FLT_MAX : input[IX(x,y-1)]; // float up = (y >= maxy) ? FLT_MAX : input[IX(x,y+1)]; // // const float dx = h; // only for square domains // // // get all the derivative estimates // float dDown = (y <= 1) ? FLT_MAX : (center - down) * dx; // float dUp = (y >= maxy) ? FLT_MAX : (up - center) * dx; // float dCenter = (y <= 1 || y >= maxy) ? FLT_MAX : (up - down) * dx * 0.5f; // // // if it's on a boundary, only one estimate is valid // if (y <= 1) return dUp; // if (y >= maxy) return dDown; // // // if it's not on a boundary, get the smallest one // float finalD = (abs(dCenter) < abs(dUp)) ? dCenter : dUp; // finalD = (abs(finalD) < abs(dDown)) ? finalD : dDown; // // return finalD; //} // // //__global__ void //ComputeEigenvalues(float *_tcU, float *_tcV, float *_eigMin, float *_eigMax, int w, int h) { // int x = getX(w); // int y = getY(h); // int id = IX(x,y); // // float A[2][2] = { // { minDx(x, y, _tcU, w, h), minDx(x, y, _tcV, w, h) }, // { minDy(x, y, _tcU, w, h), minDy(x, y, _tcV, w, h) } // }; // // const float T = A[0][0] + A[1][1]; // const float D = (A[0][0]*A[1][1]) - (A[0][1]*A[1][0]); // // _eigMin[id] = (T*0.5) - pow((T*T)/(4.0-D), 0.5); // _eigMax[id] = (T*0.5) + pow((T*T)/(4.0-D), 0.5); // //} // //__global__ void //ComputeEnergy(float *_energy, float *_xvel, float *_yvel, int w, int h) { // int x = getX(w); // int y = getY(h); // int id = IX(x,y); // // _energy[id] = 0.5 * (_xvel[id]*_xvel[id] + _yvel[id]*_yvel[id]); // //} // //__global__ void //CopyObstacles(float *inObstacles, unsigned char *_obstacles, int w, int h) { // int x = getX(w); // int y = getY(h); // int id = IX(x,y); // // if (inObstacles[4 * id + 0] > 0) { // _obstacles[id] = 1; // } // else { // _obstacles[id] = 0; // } //} // //__global__ void //IterateEnergy(float *_energy, unsigned char *_obstacles, int w, int h){ // int x = getX(w); // int y = getY(h); // int id = IX(x,y); // // const int MARCHED = 2; // const int RETIRED = 4; // const int slabSizeSm = w * h; // // if (checkBounds(x, y, w, h)) { // if (_obstacles[id] && _obstacles[id] != RETIRED) { // float sum = 0.0f; // int valid = 0; // // if (!_obstacles[id + 1] || _obstacles[id + 1] == RETIRED) { // sum += _energy[id + 1]; // valid++; // } // if (!_obstacles[id - 1] || _obstacles[id - 1] == RETIRED) { // sum += _energy[id - 1]; // valid++; // } // if (!_obstacles[id + w] || _obstacles[id + w] == RETIRED) { // sum += _energy[id + w]; // valid++; // } // if (!_obstacles[id - w] || _obstacles[id - w] == RETIRED) { // sum += _energy[id - w]; // valid++; // } // if (!_obstacles[id + slabSizeSm] || _obstacles[id + slabSizeSm] == RETIRED) { // sum += _energy[id + slabSizeSm]; // valid++; // } // if (!_obstacles[id - slabSizeSm] || _obstacles[id - slabSizeSm] == RETIRED) { // sum += _energy[id - slabSizeSm]; // valid++; // } // if (valid > 0) { // _energy[id] = sum / float(valid); // _obstacles[id] = MARCHED; // } // } // } //} // //__global__ void //CheckMarchedObstacles(unsigned char *_obstacles, int w, int h) { // int x = getX(w); // int y = getY(h); // int id = IX(x,y); // // const int MARCHED = 2; // const int RETIRED = 4; // // if (checkBounds(x, y, w, h)) { // if (_obstacles[id] == MARCHED) { // _obstacles[id] = RETIRED; // } // } // //} // //__global__ void //DownsampleNeumann(float *from, float *to, int n, int stride, int w, int h) { // int x = getX(w); // int y = getY(h); // int id = IX(x,y); // // // if these values are not local incorrect results are generated // float downCoeffs[32] = { 0.000334f,-0.001528f, 0.000410f, 0.003545f,-0.000938f,-0.008233f, 0.002172f, 0.019120f, // -0.005040f,-0.044412f, 0.011655f, 0.103311f,-0.025936f,-0.243780f, 0.033979f, 0.655340f, // 0.655340f, 0.033979f,-0.243780f,-0.025936f, 0.103311f, 0.011655f,-0.044412f,-0.005040f, // 0.019120f, 0.002172f,-0.008233f,-0.000938f, 0.003546f, 0.000410f,-0.001528f, 0.000334f }; // // const float *const aCoCenter= &downCoeffs[16]; // for (int i = 0; i < n / 2; i++) { // to[i * stride] = 0; // for (int k = 2 * i - 16; k < 2 * i + 16; k++) { // // handle boundary // float fromval; // if (k < 0) { // fromval = from[0]; // } else if(k > n - 1) { // fromval = from[(n - 1) * stride]; // } else { // fromval = from[k * stride]; // } // to[i * stride] += aCoCenter[k - 2 * i] * fromval; // } // } //} #endif
19,075
#include <cuda_runtime.h> #include <device_launch_parameters.h> #include <cufft.h> #include <stdio.h> #include <stdlib.h> __global__ void DianCheng(cufftDoubleComplex *a, cufftDoubleComplex *b, cufftDoubleComplex *c,int M, int L)//˵GPU { int tx = threadIdx.x; int by = blockIdx.y; int i=by*L+tx; if (i<=M*L) { c[i].x = a[i].x * b[i].x-a[i].y*b[i].y; c[i].y = a[i].x * b[i].y+a[i].y*b[i].x; } }
19,076
#define t_max 1 #define t 1 /* (w1_a[0][0]=((a[0][0][0][0][1]*(a[0][0][0][0][1]+1.0))*((a[0][0][0][0][1]+2.0)*0.16666666666666666))) (w2_a[0][0]=(((a[0][0][0][0][1]-1.0)*(a[0][0][0][0][1]+1.0))*((a[0][0][0][0][1]+2.0)*-0.5))) (w3_a[0][0]=(((a[0][0][0][0][1]-1.0)*a[0][0][0][0][1])*((a[0][0][0][0][1]+2.0)*0.5))) (w4_a[0][0]=(((a[0][0][0][0][1]-1.0)*a[0][0][0][0][1])*((a[0][0][0][0][1]+1.0)*-0.16666666666666666))) (w1_b[0][0]=((b[0][0][0][0][2]*(b[0][0][0][0][2]+1.0))*((b[0][0][0][0][2]+2.0)*0.16666666666666666))) (w2_b[0][0]=(((b[0][0][0][0][2]-1.0)*(b[0][0][0][0][2]+1.0))*((b[0][0][0][0][2]+2.0)*-0.5))) (w3_b[0][0]=(((b[0][0][0][0][2]-1.0)*b[0][0][0][0][2])*((b[0][0][0][0][2]+2.0)*0.5))) (w4_b[0][0]=(((b[0][0][0][0][2]-1.0)*b[0][0][0][0][2])*((b[0][0][0][0][2]+1.0)*-0.16666666666666666))) (w1_c[0][0]=((c[0][0][0][0][3]*(c[0][0][0][0][3]+1.0))*((c[0][0][0][0][3]+2.0)*0.16666666666666666))) (w2_c[0][0]=(((c[0][0][0][0][3]-1.0)*(c[0][0][0][0][3]+1.0))*((c[0][0][0][0][3]+2.0)*-0.5))) (w3_c[0][0]=(((c[0][0][0][0][3]-1.0)*c[0][0][0][0][3])*((c[0][0][0][0][3]+2.0)*0.5))) (w4_c[0][0]=(((c[0][0][0][0][3]-1.0)*c[0][0][0][0][3])*((c[0][0][0][0][3]+1.0)*-0.16666666666666666))) (u[0][0][0][1][0]=((((((((w1_a*w1_b)*(w1_c*u[-1][-1][-1][0][0]))+((w2_a*w1_b)*(w1_c*u[0][-1][-1][0][0])))+(((w3_a*w1_b)*(w1_c*u[1][-1][-1][0][0]))+((w4_a*w1_b)*(w1_c*u[2][-1][-1][0][0]))))+((((w1_a*w2_b)*(w1_c*u[-1][0][-1][0][0]))+((w2_a*w2_b)*(w1_c*u[0][0][-1][0][0])))+(((w3_a*w2_b)*(w1_c*u[1][0][-1][0][0]))+((w4_a*w2_b)*(w1_c*u[2][0][-1][0][0])))))+(((((w1_a*w3_b)*(w1_c*u[-1][1][-1][0][0]))+((w2_a*w3_b)*(w1_c*u[0][1][-1][0][0])))+(((w3_a*w3_b)*(w1_c*u[1][1][-1][0][0]))+((w4_a*w3_b)*(w1_c*u[2][1][-1][0][0]))))+((((w1_a*w4_b)*(w1_c*u[-1][2][-1][0][0]))+((w2_a*w4_b)*(w1_c*u[0][2][-1][0][0])))+(((w3_a*w4_b)*(w1_c*u[1][2][-1][0][0]))+((w4_a*w4_b)*(w1_c*u[2][2][-1][0][0]))))))+((((((w1_a*w1_b)*(w2_c*u[-1][-1][0][0][0]))+((w2_a*w1_b)*(w2_c*u[0][-1][0][0][0])))+(((w3_a*w1_b)*(w2_c*u[1][-1][0][0][0]))+((w4_a*w1_b)*(w2_c*u[2][-1][0][0][0]))))+((((w1_a*w2_b)*(w2_c*u[-1][0][0][0][0]))+((w2_a*w2_b)*(w2_c*u[0][0][0][0][0])))+(((w3_a*w2_b)*(w2_c*u[1][0][0][0][0]))+((w4_a*w2_b)*(w2_c*u[2][0][0][0][0])))))+(((((w1_a*w3_b)*(w2_c*u[-1][1][0][0][0]))+((w2_a*w3_b)*(w2_c*u[0][1][0][0][0])))+(((w3_a*w3_b)*(w2_c*u[1][1][0][0][0]))+((w4_a*w3_b)*(w2_c*u[2][1][0][0][0]))))+((((w1_a*w4_b)*(w2_c*u[-1][2][0][0][0]))+((w2_a*w4_b)*(w2_c*u[0][2][0][0][0])))+(((w3_a*w4_b)*(w2_c*u[1][2][0][0][0]))+((w4_a*w4_b)*(w2_c*u[2][2][0][0][0])))))))+(((((((w1_a*w1_b)*(w3_c*u[-1][-1][1][0][0]))+((w2_a*w1_b)*(w3_c*u[0][-1][1][0][0])))+(((w3_a*w1_b)*(w3_c*u[1][-1][1][0][0]))+((w4_a*w1_b)*(w3_c*u[2][-1][1][0][0]))))+((((w1_a*w2_b)*(w3_c*u[-1][0][1][0][0]))+((w2_a*w2_b)*(w3_c*u[0][0][1][0][0])))+(((w3_a*w2_b)*(w3_c*u[1][0][1][0][0]))+((w4_a*w2_b)*(w3_c*u[2][0][1][0][0])))))+(((((w1_a*w3_b)*(w3_c*u[-1][1][1][0][0]))+((w2_a*w3_b)*(w3_c*u[0][1][1][0][0])))+(((w3_a*w3_b)*(w3_c*u[1][1][1][0][0]))+((w4_a*w3_b)*(w3_c*u[2][1][1][0][0]))))+((((w1_a*w4_b)*(w3_c*u[-1][2][1][0][0]))+((w2_a*w4_b)*(w3_c*u[0][2][1][0][0])))+(((w3_a*w4_b)*(w3_c*u[1][2][1][0][0]))+((w4_a*w4_b)*(w3_c*u[2][2][1][0][0]))))))+((((((w1_a*w1_b)*(w4_c*u[-1][-1][2][0][0]))+((w2_a*w1_b)*(w4_c*u[0][-1][2][0][0])))+(((w3_a*w1_b)*(w4_c*u[1][-1][2][0][0]))+((w4_a*w1_b)*(w4_c*u[2][-1][2][0][0]))))+((((w1_a*w2_b)*(w4_c*u[-1][0][2][0][0]))+((w2_a*w2_b)*(w4_c*u[0][0][2][0][0])))+(((w3_a*w2_b)*(w4_c*u[1][0][2][0][0]))+((w4_a*w2_b)*(w4_c*u[2][0][2][0][0])))))+(((((w1_a*w3_b)*(w4_c*u[-1][1][2][0][0]))+((w2_a*w3_b)*(w4_c*u[0][1][2][0][0])))+(((w3_a*w3_b)*(w4_c*u[1][1][2][0][0]))+((w4_a*w3_b)*(w4_c*u[2][1][2][0][0]))))+((((w1_a*w4_b)*(w4_c*u[-1][2][2][0][0]))+((w2_a*w4_b)*(w4_c*u[0][2][2][0][0])))+(((w3_a*w4_b)*(w4_c*u[1][2][2][0][0]))+((w4_a*w4_b)*(w4_c*u[2][2][2][0][0]))))))))) */ __global__ void tricubic_interpolation(double * * u_0_1_out, double * u_0_0, double * u_0_1, double * a_1_0, double * b_2_0, double * c_3_0, int x_max, int y_max, int z_max) { /* const double * const u__a_1[16] = { a_1_0 } ; const double * const u__b_2[16] = { b_2_0 } ; const double * const u__c_3[16] = { c_3_0 } ; double * const u__u_0[16] = { u_0_0, u_0_1 } ; */ double w1_a; double w1_b; double w1_c; double w2_a; double w2_b; double w2_c; double w3_a; double w3_b; double w3_c; double w4_a; double w4_b; double w4_c; int _idx0; int _idx1; int _idx10; int _idx11; int _idx12; int _idx13; int _idx14; int _idx15; int _idx16; int _idx17; int _idx18; int _idx19; int _idx2; int _idx20; int _idx21; int _idx22; int _idx23; int _idx24; int _idx25; int _idx26; int _idx27; int _idx28; int _idx29; int _idx3; int _idx30; int _idx31; int _idx32; int _idx33; int _idx34; int _idx35; int _idx36; int _idx37; int _idx38; int _idx39; int _idx4; int _idx40; int _idx41; int _idx42; int _idx43; int _idx44; int _idx45; int _idx46; int _idx47; int _idx48; int _idx49; int _idx5; int _idx50; int _idx51; int _idx52; int _idx53; int _idx54; int _idx55; int _idx56; int _idx57; int _idx58; int _idx59; int _idx6; int _idx60; int _idx61; int _idx62; int _idx63; int _idx64; int _idx7; int _idx8; int _idx9; int idx_1_2; int p_idx_x; int p_idx_x_max; int p_idx_y; int p_idx_y_max; int p_idx_z; int p_idx_z_max; int size_1_1; int size_1_2; //int t; int tmp; /* Initializations */ size_1_1=(y_max/blockDim.y); size_1_2=(z_max/blockDim.z); idx_1_2=(blockIdx.y/size_1_2); tmp=(blockIdx.y-(idx_1_2*size_1_2)); p_idx_x=(threadIdx.x+(blockDim.x*blockIdx.x)); p_idx_x_max=(p_idx_x+1); p_idx_y=(threadIdx.y+(tmp*blockDim.y)); p_idx_y_max=(p_idx_y+1); p_idx_z=(threadIdx.z+(idx_1_2*blockDim.z)); p_idx_z_max=(p_idx_z+1); /* Implementation */ /* for t = 1..t_max by 1 parallel 1 <level 0> schedule { ... } */ //for (t=1; t<=t_max; t+=1) { /* Index bounds calculations for iterators in p[t=t, s=(1, 1, 1)][0] */ /* u[t=(t+1), s=p[t=?, s=?][0]][0]=stencil(u[t=t, s=p[t=?, s=?][0]][0]) */ /* _idx0 = ((((p_idx_z*x_max)*y_max)+(p_idx_y*x_max))+p_idx_x) */ _idx0=((((p_idx_z*x_max)*y_max)+(p_idx_y*x_max))+p_idx_x); w1_a=((a_1_0[_idx0]*(a_1_0[_idx0]+1.0))*((a_1_0[_idx0]+2.0)*0.16666666666666666)); w2_a=(((a_1_0[_idx0]-1.0)*(a_1_0[_idx0]+1.0))*((a_1_0[_idx0]+2.0)*-0.5)); w3_a=(((a_1_0[_idx0]-1.0)*a_1_0[_idx0])*((a_1_0[_idx0]+2.0)*0.5)); w4_a=(((a_1_0[_idx0]-1.0)*a_1_0[_idx0])*((a_1_0[_idx0]+1.0)*-0.16666666666666666)); w1_b=((b_2_0[_idx0]*(b_2_0[_idx0]+1.0))*((b_2_0[_idx0]+2.0)*0.16666666666666666)); w2_b=(((b_2_0[_idx0]-1.0)*(b_2_0[_idx0]+1.0))*((b_2_0[_idx0]+2.0)*-0.5)); w3_b=(((b_2_0[_idx0]-1.0)*b_2_0[_idx0])*((b_2_0[_idx0]+2.0)*0.5)); w4_b=(((b_2_0[_idx0]-1.0)*b_2_0[_idx0])*((b_2_0[_idx0]+1.0)*-0.16666666666666666)); w1_c=((c_3_0[_idx0]*(c_3_0[_idx0]+1.0))*((c_3_0[_idx0]+2.0)*0.16666666666666666)); w2_c=(((c_3_0[_idx0]-1.0)*(c_3_0[_idx0]+1.0))*((c_3_0[_idx0]+2.0)*-0.5)); w3_c=(((c_3_0[_idx0]-1.0)*c_3_0[_idx0])*((c_3_0[_idx0]+2.0)*0.5)); w4_c=(((c_3_0[_idx0]-1.0)*c_3_0[_idx0])*((c_3_0[_idx0]+1.0)*-0.16666666666666666)); /* _idx1 = (((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+((((3*p_idx_z)*t)+p_idx_y)*x_max))+((9*p_idx_z)*(t*t)))+((3*p_idx_y)*t))+p_idx_x) */ _idx1=((((_idx0+(((3*p_idx_z)*t)*y_max))+(((3*p_idx_z)*t)*x_max))+((9*p_idx_z)*(t*t)))+((3*p_idx_y)*t)); /* _idx2 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+((((3*p_idx_z)*t)+p_idx_y)*x_max))+((9*p_idx_z)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+1) */ _idx2=(_idx1+1); /* _idx3 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+((((3*p_idx_z)*t)+p_idx_y)*x_max))+((9*p_idx_z)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+2) */ _idx3=(_idx2+1); /* _idx4 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+((((3*p_idx_z)*t)+p_idx_y)*x_max))+((9*p_idx_z)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+3) */ _idx4=(_idx2+2); /* _idx5 = (((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+1)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x) */ _idx5=((_idx1+x_max)+(3*t)); /* _idx6 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+1)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+1) */ _idx6=(_idx5+1); /* _idx7 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+1)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+2) */ _idx7=(_idx5+2); /* _idx8 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+1)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+3) */ _idx8=(_idx7+1); /* _idx9 = (((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+2)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x) */ _idx9=((_idx5+x_max)+(3*t)); /* _idx10 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+2)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+1) */ _idx10=(_idx9+1); /* _idx11 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+2)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+2) */ _idx11=(_idx10+1); /* _idx12 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+2)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+3) */ _idx12=(_idx11+1); /* _idx13 = (((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+3)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x) */ _idx13=((_idx9+x_max)+(3*t)); /* _idx14 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+3)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+1) */ _idx14=(_idx13+1); /* _idx15 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+3)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+2) */ _idx15=(_idx14+1); /* _idx16 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+3)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+3) */ _idx16=(_idx14+2); /* _idx17 = ((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+(((((3*p_idx_z)+3)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+9)*(t*t)))+((3*p_idx_y)*t))+p_idx_x) */ _idx17=(((_idx1+((x_max+(3*t))*y_max))+((3*t)*x_max))+(9*(t*t))); /* _idx18 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+(((((3*p_idx_z)+3)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+9)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+1) */ _idx18=(_idx17+1); /* _idx19 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+(((((3*p_idx_z)+3)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+9)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+2) */ _idx19=(_idx18+1); /* _idx20 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+(((((3*p_idx_z)+3)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+9)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+3) */ _idx20=(_idx19+1); /* _idx21 = ((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x) */ _idx21=((_idx17+x_max)+(3*t)); /* _idx22 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+1) */ _idx22=(_idx21+1); /* _idx23 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+2) */ _idx23=(_idx22+1); /* _idx24 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+3) */ _idx24=(_idx23+1); /* _idx25 = ((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x) */ _idx25=((_idx21+x_max)+(3*t)); /* _idx26 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+1) */ _idx26=(_idx25+1); /* _idx27 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+2) */ _idx27=(_idx25+2); /* _idx28 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+3) */ _idx28=(_idx27+1); /* _idx29 = ((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x) */ _idx29=((_idx25+x_max)+(3*t)); /* _idx30 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+1) */ _idx30=(_idx29+1); /* _idx31 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+2) */ _idx31=(_idx29+2); /* _idx32 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+3) */ _idx32=(_idx29+3); /* _idx33 = ((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+(((((3*p_idx_z)+6)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+18)*(t*t)))+((3*p_idx_y)*t))+p_idx_x) */ _idx33=(((_idx17+((x_max+(3*t))*y_max))+((3*t)*x_max))+(9*(t*t))); /* _idx34 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+(((((3*p_idx_z)+6)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+18)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+1) */ _idx34=(_idx33+1); /* _idx35 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+(((((3*p_idx_z)+6)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+18)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+2) */ _idx35=(_idx34+1); /* _idx36 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+(((((3*p_idx_z)+6)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+18)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+3) */ _idx36=(_idx34+2); /* _idx37 = ((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x) */ _idx37=((_idx33+x_max)+(3*t)); /* _idx38 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+1) */ _idx38=(_idx37+1); /* _idx39 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+2) */ _idx39=(_idx37+2); /* _idx40 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+3) */ _idx40=(_idx39+1); /* _idx41 = ((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x) */ _idx41=((_idx37+x_max)+(3*t)); /* _idx42 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+1) */ _idx42=(_idx41+1); /* _idx43 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+2) */ _idx43=(_idx41+2); /* _idx44 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+3) */ _idx44=(_idx41+3); /* _idx45 = ((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x) */ _idx45=((_idx41+x_max)+(3*t)); /* _idx46 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+1) */ _idx46=(_idx45+1); /* _idx47 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+2) */ _idx47=(_idx45+2); /* _idx48 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+3) */ _idx48=(_idx45+3); /* _idx49 = ((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+(((((3*p_idx_z)+9)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+27)*(t*t)))+((3*p_idx_y)*t))+p_idx_x) */ _idx49=(((_idx33+((x_max+(3*t))*y_max))+((3*t)*x_max))+(9*(t*t))); /* _idx50 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+(((((3*p_idx_z)+9)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+27)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+1) */ _idx50=(_idx49+1); /* _idx51 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+(((((3*p_idx_z)+9)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+27)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+2) */ _idx51=(_idx50+1); /* _idx52 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+(((((3*p_idx_z)+9)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+27)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+3) */ _idx52=(_idx50+2); /* _idx53 = ((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x) */ _idx53=((_idx49+x_max)+(3*t)); /* _idx54 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+1) */ _idx54=(_idx53+1); /* _idx55 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+2) */ _idx55=(_idx53+2); /* _idx56 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+3) */ _idx56=(_idx53+3); /* _idx57 = ((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x) */ _idx57=((_idx53+x_max)+(3*t)); /* _idx58 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+1) */ _idx58=(_idx57+1); /* _idx59 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+2) */ _idx59=(_idx57+2); /* _idx60 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+3) */ _idx60=(_idx57+3); /* _idx61 = ((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x) */ _idx61=((_idx57+x_max)+(3*t)); /* _idx62 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+1) */ _idx62=(_idx61+1); /* _idx63 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+2) */ _idx63=(_idx61+2); /* _idx64 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+3) */ _idx64=(_idx63+1); u_0_0[_idx22]=((((((((w1_a*w1_b)*(w1_c*u_0_1[_idx1]))+((w2_a*w1_b)*(w1_c*u_0_1[_idx2])))+(((w3_a*w1_b)*(w1_c*u_0_1[_idx3]))+((w4_a*w1_b)*(w1_c*u_0_1[_idx4]))))+((((w1_a*w2_b)*(w1_c*u_0_1[_idx5]))+((w2_a*w2_b)*(w1_c*u_0_1[_idx6])))+(((w3_a*w2_b)*(w1_c*u_0_1[_idx7]))+((w4_a*w2_b)*(w1_c*u_0_1[_idx8])))))+(((((w1_a*w3_b)*(w1_c*u_0_1[_idx9]))+((w2_a*w3_b)*(w1_c*u_0_1[_idx10])))+(((w3_a*w3_b)*(w1_c*u_0_1[_idx11]))+((w4_a*w3_b)*(w1_c*u_0_1[_idx12]))))+((((w1_a*w4_b)*(w1_c*u_0_1[_idx13]))+((w2_a*w4_b)*(w1_c*u_0_1[_idx14])))+(((w3_a*w4_b)*(w1_c*u_0_1[_idx15]))+((w4_a*w4_b)*(w1_c*u_0_1[_idx16]))))))+((((((w1_a*w1_b)*(w2_c*u_0_1[_idx17]))+((w2_a*w1_b)*(w2_c*u_0_1[_idx18])))+(((w3_a*w1_b)*(w2_c*u_0_1[_idx19]))+((w4_a*w1_b)*(w2_c*u_0_1[_idx20]))))+((((w1_a*w2_b)*(w2_c*u_0_1[_idx21]))+((w2_a*w2_b)*(w2_c*u_0_1[_idx22])))+(((w3_a*w2_b)*(w2_c*u_0_1[_idx23]))+((w4_a*w2_b)*(w2_c*u_0_1[_idx24])))))+(((((w1_a*w3_b)*(w2_c*u_0_1[_idx25]))+((w2_a*w3_b)*(w2_c*u_0_1[_idx26])))+(((w3_a*w3_b)*(w2_c*u_0_1[_idx27]))+((w4_a*w3_b)*(w2_c*u_0_1[_idx28]))))+((((w1_a*w4_b)*(w2_c*u_0_1[_idx29]))+((w2_a*w4_b)*(w2_c*u_0_1[_idx30])))+(((w3_a*w4_b)*(w2_c*u_0_1[_idx31]))+((w4_a*w4_b)*(w2_c*u_0_1[_idx32])))))))+(((((((w1_a*w1_b)*(w3_c*u_0_1[_idx33]))+((w2_a*w1_b)*(w3_c*u_0_1[_idx34])))+(((w3_a*w1_b)*(w3_c*u_0_1[_idx35]))+((w4_a*w1_b)*(w3_c*u_0_1[_idx36]))))+((((w1_a*w2_b)*(w3_c*u_0_1[_idx37]))+((w2_a*w2_b)*(w3_c*u_0_1[_idx38])))+(((w3_a*w2_b)*(w3_c*u_0_1[_idx39]))+((w4_a*w2_b)*(w3_c*u_0_1[_idx40])))))+(((((w1_a*w3_b)*(w3_c*u_0_1[_idx41]))+((w2_a*w3_b)*(w3_c*u_0_1[_idx42])))+(((w3_a*w3_b)*(w3_c*u_0_1[_idx43]))+((w4_a*w3_b)*(w3_c*u_0_1[_idx44]))))+((((w1_a*w4_b)*(w3_c*u_0_1[_idx45]))+((w2_a*w4_b)*(w3_c*u_0_1[_idx46])))+(((w3_a*w4_b)*(w3_c*u_0_1[_idx47]))+((w4_a*w4_b)*(w3_c*u_0_1[_idx48]))))))+((((((w1_a*w1_b)*(w4_c*u_0_1[_idx49]))+((w2_a*w1_b)*(w4_c*u_0_1[_idx50])))+(((w3_a*w1_b)*(w4_c*u_0_1[_idx51]))+((w4_a*w1_b)*(w4_c*u_0_1[_idx52]))))+((((w1_a*w2_b)*(w4_c*u_0_1[_idx53]))+((w2_a*w2_b)*(w4_c*u_0_1[_idx54])))+(((w3_a*w2_b)*(w4_c*u_0_1[_idx55]))+((w4_a*w2_b)*(w4_c*u_0_1[_idx56])))))+(((((w1_a*w3_b)*(w4_c*u_0_1[_idx57]))+((w2_a*w3_b)*(w4_c*u_0_1[_idx58])))+(((w3_a*w3_b)*(w4_c*u_0_1[_idx59]))+((w4_a*w3_b)*(w4_c*u_0_1[_idx60]))))+((((w1_a*w4_b)*(w4_c*u_0_1[_idx61]))+((w2_a*w4_b)*(w4_c*u_0_1[_idx62])))+(((w3_a*w4_b)*(w4_c*u_0_1[_idx63]))+((w4_a*w4_b)*(w4_c*u_0_1[_idx64])))))))); } } __global__ void initialize(double * u_0_0, double * u_0_1, double * a_1_0, double * b_2_0, double * c_3_0, int x_max, int y_max, int z_max) { /* const double * const a_1_0[16] = { a_1_0 } ; const double * const b_2_0[16] = { b_2_0 } ; const double * const c_3_0[16] = { c_3_0 } ; double * const u__u_0[16] = { u_0_0, u_0_1 } ; */ double w1_a; double w1_b; double w1_c; double w2_a; double w2_b; double w2_c; double w3_a; double w3_b; double w3_c; double w4_a; double w4_b; double w4_c; int _idx0; int _idx1; int _idx10; int _idx11; int _idx12; int _idx13; int _idx14; int _idx15; int _idx16; int _idx17; int _idx18; int _idx19; int _idx2; int _idx20; int _idx21; int _idx22; int _idx23; int _idx24; int _idx25; int _idx26; int _idx27; int _idx28; int _idx29; int _idx3; int _idx30; int _idx31; int _idx32; int _idx33; int _idx34; int _idx35; int _idx36; int _idx37; int _idx38; int _idx39; int _idx4; int _idx40; int _idx41; int _idx42; int _idx43; int _idx44; int _idx45; int _idx46; int _idx47; int _idx48; int _idx49; int _idx5; int _idx50; int _idx51; int _idx52; int _idx53; int _idx54; int _idx55; int _idx56; int _idx57; int _idx58; int _idx59; int _idx6; int _idx60; int _idx61; int _idx62; int _idx63; int _idx64; int _idx7; int _idx8; int _idx9; int idx_1_2; int p_idx_x; int p_idx_x_max; int p_idx_y; int p_idx_y_max; int p_idx_z; int p_idx_z_max; int size_1_1; int size_1_2; //int t; int tmp; /* Initializations */ size_1_1=(y_max/blockDim.y); size_1_2=(z_max/blockDim.z); idx_1_2=(blockIdx.y/size_1_2); tmp=(blockIdx.y-(idx_1_2*size_1_2)); p_idx_x=(threadIdx.x+(blockDim.x*blockIdx.x)); p_idx_x_max=(p_idx_x+1); p_idx_y=(threadIdx.y+(tmp*blockDim.y)); p_idx_y_max=(p_idx_y+1); p_idx_z=(threadIdx.z+(idx_1_2*blockDim.z)); p_idx_z_max=(p_idx_z+1); /* Implementation */ /* for t = 1..t_max by 1 parallel 1 <level 0> schedule { ... } */ //for (t=1; t<=t_max; t+=1) { /* Index bounds calculations for iterators in p[t=t, s=(1, 1, 1)][0] */ /* u[t=(t+1), s=p[t=?, s=?][0]][0]=stencil(u[t=t, s=p[t=?, s=?][0]][0]) */ /* _idx0 = ((((p_idx_z*x_max)*y_max)+(p_idx_y*x_max))+p_idx_x) */ _idx0=((((p_idx_z*x_max)*y_max)+(p_idx_y*x_max))+p_idx_x); a_1_0[_idx0]=0.2; w1_a=0.1; a_1_0[_idx0]=0.2; w2_a=0.1; a_1_0[_idx0]=0.2; w3_a=0.1; a_1_0[_idx0]=0.2; w4_a=0.1; b_2_0[_idx0]=0.30000000000000004; w1_b=0.1; b_2_0[_idx0]=0.30000000000000004; w2_b=0.1; b_2_0[_idx0]=0.30000000000000004; w3_b=0.1; b_2_0[_idx0]=0.30000000000000004; w4_b=0.1; c_3_0[_idx0]=0.4; w1_c=0.1; c_3_0[_idx0]=0.4; w2_c=0.1; c_3_0[_idx0]=0.4; w3_c=0.1; c_3_0[_idx0]=0.4; w4_c=0.1; /* _idx1 = (((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+((((3*p_idx_z)*t)+p_idx_y)*x_max))+((9*p_idx_z)*(t*t)))+((3*p_idx_y)*t))+p_idx_x) */ _idx1=((((_idx0+(((3*p_idx_z)*t)*y_max))+(((3*p_idx_z)*t)*x_max))+((9*p_idx_z)*(t*t)))+((3*p_idx_y)*t)); u_0_1[_idx1]=0.1; /* _idx2 = ((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+(((((3*p_idx_z)+3)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+9)*(t*t)))+((3*p_idx_y)*t))+p_idx_x) */ _idx2=(((_idx1+((x_max+(3*t))*y_max))+((3*t)*x_max))+(9*(t*t))); u_0_1[_idx2]=0.1; /* _idx3 = ((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+(((((3*p_idx_z)+6)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+18)*(t*t)))+((3*p_idx_y)*t))+p_idx_x) */ _idx3=(((_idx2+((x_max+(3*t))*y_max))+((3*t)*x_max))+(9*(t*t))); u_0_1[_idx3]=0.1; /* _idx4 = ((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+(((((3*p_idx_z)+9)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+27)*(t*t)))+((3*p_idx_y)*t))+p_idx_x) */ _idx4=(((_idx3+((x_max+(3*t))*y_max))+((3*t)*x_max))+(9*(t*t))); u_0_1[_idx4]=0.1; /* _idx5 = (((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+1)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x) */ _idx5=((_idx1+x_max)+(3*t)); u_0_1[_idx5]=0.1; /* _idx6 = ((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x) */ _idx6=((_idx2+x_max)+(3*t)); u_0_1[_idx6]=0.1; /* _idx7 = ((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x) */ _idx7=((_idx3+x_max)+(3*t)); u_0_1[_idx7]=0.1; /* _idx8 = ((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x) */ _idx8=((_idx4+x_max)+(3*t)); u_0_1[_idx8]=0.1; /* _idx9 = (((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+2)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x) */ _idx9=((_idx5+x_max)+(3*t)); u_0_1[_idx9]=0.1; /* _idx10 = ((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x) */ _idx10=((_idx6+x_max)+(3*t)); u_0_1[_idx10]=0.1; /* _idx11 = ((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x) */ _idx11=((_idx7+x_max)+(3*t)); u_0_1[_idx11]=0.1; /* _idx12 = ((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x) */ _idx12=((_idx8+x_max)+(3*t)); u_0_1[_idx12]=0.1; /* _idx13 = (((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+3)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x) */ _idx13=((_idx9+x_max)+(3*t)); u_0_1[_idx13]=0.1; /* _idx14 = ((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x) */ _idx14=((_idx10+x_max)+(3*t)); u_0_1[_idx14]=0.1; /* _idx15 = ((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x) */ _idx15=((_idx11+x_max)+(3*t)); u_0_1[_idx15]=0.1; /* _idx16 = ((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x) */ _idx16=((_idx12+x_max)+(3*t)); u_0_1[_idx16]=0.1; /* _idx17 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+((((3*p_idx_z)*t)+p_idx_y)*x_max))+((9*p_idx_z)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+1) */ _idx17=(_idx1+1); u_0_1[_idx17]=0.1; /* _idx18 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+(((((3*p_idx_z)+3)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+9)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+1) */ _idx18=(_idx2+1); u_0_1[_idx18]=0.1; /* _idx19 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+(((((3*p_idx_z)+6)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+18)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+1) */ _idx19=(_idx3+1); u_0_1[_idx19]=0.1; /* _idx20 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+(((((3*p_idx_z)+9)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+27)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+1) */ _idx20=(_idx4+1); u_0_1[_idx20]=0.1; /* _idx21 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+1)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+1) */ _idx21=(_idx5+1); u_0_1[_idx21]=0.1; /* _idx22 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+1) */ _idx22=(_idx6+1); u_0_1[_idx22]=0.1; /* _idx23 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+1) */ _idx23=(_idx7+1); u_0_1[_idx23]=0.1; /* _idx24 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+1) */ _idx24=(_idx8+1); u_0_1[_idx24]=0.1; /* _idx25 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+2)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+1) */ _idx25=(_idx9+1); u_0_1[_idx25]=0.1; /* _idx26 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+1) */ _idx26=(_idx10+1); u_0_1[_idx26]=0.1; /* _idx27 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+1) */ _idx27=(_idx11+1); u_0_1[_idx27]=0.1; /* _idx28 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+1) */ _idx28=(_idx12+1); u_0_1[_idx28]=0.1; /* _idx29 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+3)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+1) */ _idx29=(_idx13+1); u_0_1[_idx29]=0.1; /* _idx30 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+1) */ _idx30=(_idx14+1); u_0_1[_idx30]=0.1; /* _idx31 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+1) */ _idx31=(_idx15+1); u_0_1[_idx31]=0.1; /* _idx32 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+1) */ _idx32=(_idx16+1); u_0_1[_idx32]=0.1; /* _idx33 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+((((3*p_idx_z)*t)+p_idx_y)*x_max))+((9*p_idx_z)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+2) */ _idx33=(_idx17+1); u_0_1[_idx33]=0.1; /* _idx34 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+(((((3*p_idx_z)+3)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+9)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+2) */ _idx34=(_idx2+2); u_0_1[_idx34]=0.1; /* _idx35 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+(((((3*p_idx_z)+6)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+18)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+2) */ _idx35=(_idx19+1); u_0_1[_idx35]=0.1; /* _idx36 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+(((((3*p_idx_z)+9)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+27)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+2) */ _idx36=(_idx20+1); u_0_1[_idx36]=0.1; /* _idx37 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+1)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+2) */ _idx37=(_idx21+1); u_0_1[_idx37]=0.1; /* _idx38 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+2) */ _idx38=(_idx6+2); u_0_1[_idx38]=0.1; /* _idx39 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+2) */ _idx39=(_idx7+2); u_0_1[_idx39]=0.1; /* _idx40 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+2) */ _idx40=(_idx8+2); u_0_1[_idx40]=0.1; /* _idx41 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+2)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+2) */ _idx41=(_idx25+1); u_0_1[_idx41]=0.1; /* _idx42 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+2) */ _idx42=(_idx10+2); u_0_1[_idx42]=0.1; /* _idx43 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+2) */ _idx43=(_idx27+1); u_0_1[_idx43]=0.1; /* _idx44 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+2) */ _idx44=(_idx12+2); u_0_1[_idx44]=0.1; /* _idx45 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+3)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+2) */ _idx45=(_idx29+1); u_0_1[_idx45]=0.1; /* _idx46 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+2) */ _idx46=(_idx30+1); u_0_1[_idx46]=0.1; /* _idx47 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+2) */ _idx47=(_idx15+2); u_0_1[_idx47]=0.1; /* _idx48 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+2) */ _idx48=(_idx16+2); u_0_1[_idx48]=0.1; /* _idx49 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+((((3*p_idx_z)*t)+p_idx_y)*x_max))+((9*p_idx_z)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+3) */ _idx49=(_idx17+2); u_0_1[_idx49]=0.1; /* _idx50 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+(((((3*p_idx_z)+3)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+9)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+3) */ _idx50=(_idx2+3); u_0_1[_idx50]=0.1; /* _idx51 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+(((((3*p_idx_z)+6)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+18)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+3) */ _idx51=(_idx19+2); u_0_1[_idx51]=0.1; /* _idx52 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+(((((3*p_idx_z)+9)*t)+p_idx_y)*x_max))+(((9*p_idx_z)+27)*(t*t)))+((3*p_idx_y)*t))+p_idx_x)+3) */ _idx52=(_idx20+2); u_0_1[_idx52]=0.1; /* _idx53 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+1)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+3) */ _idx53=(_idx21+2); u_0_1[_idx53]=0.1; /* _idx54 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+3) */ _idx54=(_idx6+3); u_0_1[_idx54]=0.1; /* _idx55 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+3) */ _idx55=(_idx7+3); u_0_1[_idx55]=0.1; /* _idx56 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+1)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+3)*t))+p_idx_x)+3) */ _idx56=(_idx8+3); u_0_1[_idx56]=0.1; /* _idx57 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+2)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+3) */ _idx57=(_idx25+2); u_0_1[_idx57]=0.1; /* _idx58 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+3) */ _idx58=(_idx42+1); u_0_1[_idx58]=0.1; /* _idx59 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+3) */ _idx59=(_idx43+1); u_0_1[_idx59]=0.1; /* _idx60 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+2)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+6)*t))+p_idx_x)+3) */ _idx60=(_idx12+3); u_0_1[_idx60]=0.1; /* _idx61 = ((((((((p_idx_z*x_max)+((3*p_idx_z)*t))*y_max)+(((((3*p_idx_z)*t)+p_idx_y)+3)*x_max))+((9*p_idx_z)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+3) */ _idx61=(_idx29+2); u_0_1[_idx61]=0.1; /* _idx62 = (((((((((p_idx_z+1)*x_max)+(((3*p_idx_z)+3)*t))*y_max)+((((((3*p_idx_z)+3)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+9)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+3) */ _idx62=(_idx46+1); u_0_1[_idx62]=0.1; /* _idx63 = (((((((((p_idx_z+2)*x_max)+(((3*p_idx_z)+6)*t))*y_max)+((((((3*p_idx_z)+6)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+18)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+3) */ _idx63=(_idx15+3); u_0_1[_idx63]=0.1; /* _idx64 = (((((((((p_idx_z+3)*x_max)+(((3*p_idx_z)+9)*t))*y_max)+((((((3*p_idx_z)+9)*t)+p_idx_y)+3)*x_max))+(((9*p_idx_z)+27)*(t*t)))+(((3*p_idx_y)+9)*t))+p_idx_x)+3) */ _idx64=(_idx48+1); u_0_1[_idx64]=0.1; u_0_0[_idx22]=1.1; } }
19,077
#include "includes.h" __global__ void sqr_norm_kernel(const float *in, float *block_res, int total) { extern __shared__ float sdata[]; int in_idx = 2 * (blockIdx.x * blockDim.x + threadIdx.x); int i = threadIdx.x; unsigned ins = blockDim.x; if (in_idx >= total * 2) sdata[i] = 0; else sdata[i] = in[in_idx] * in[in_idx] + in[in_idx + 1] * in[in_idx + 1]; for (unsigned outs = (ins + 1) / 2; ins > 1; ins = outs, outs = (outs + 1) / 2) { __syncthreads(); if (i + outs < ins) sdata[i] += sdata[i + outs]; } if (i == 0) block_res[blockIdx.x] = sdata[0]; }
19,078
#include <iostream> #include <string> #include <fstream> #include <chrono> #include <stdio.h> #include <stdlib.h> using namespace std; const int FILTER_WIDTH = 7; const int BLOCK_SIZE = 256; int FILTER[FILTER_WIDTH*FILTER_WIDTH] = { 1,4,7,10,7,4,1, 4,12,26,33,26,12,4, 7,26,55,71,55,26,7, 10,33,71,91,71,33,10, 7,26,55,71,55,26,7, 4,12,26,33,26,12,4, 1,4,7,10,7,4,1 }; // Display the first and last 10 items void displayResult(const int original[], const int result[], int size) { cout << "Display result: "; cout << "(original -> result)\n"; for (int i = 0; i < 10; i++) { cout << original[i] << " -> " << result[i] << "\n"; } cout << ".\n.\n.\n"; for (int i = size - 10; i < size; i++) { cout << original[i] << " -> " << result[i] << "\n"; } } void initData(string file, int **data, int *sizeX, int *sizeY) { int x; int y; long long i = 0; cout << "Reading "<< file << "... \n"; ifstream myfile(file); if (myfile.is_open()) { myfile >> x; myfile >> y; int *temp = new int[x * y]; for( i=0; i < x * y; i++){ myfile >> temp[(int)i]; } myfile.close(); *data = temp; *sizeX = x; *sizeY = y; } else { cout << "ERROR: File " << file << " not found!\n"; exit(0); } cout << i << " entries imported\n"; } void saveResult(string file, int data[], int sizeX, int sizeY) { long long i = 0; cout << "Saving data to "<< file <<"... \n"; ofstream myfile(file, std::ofstream::out); if (myfile.is_open()) { myfile << sizeX << "\n"; myfile << sizeY << "\n"; for (i = 0; i < sizeX * sizeY; i++){ myfile << data[i] << "\n"; } myfile.close(); } else { cout << "ERROR: Cannot save to " << file << "!\n"; exit(0); } cout << i << " entries saved\n"; } // Kernel function for 2D smoothing in GPU __global__ void calculateResult(int sizeX, int sizeY, int *data, int *result, int *filter){ int halfFilterWidth = FILTER_WIDTH/2; //int indexX = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x; //int indexY = blockIdx.y * blockDim.y + threadIdx.y; int strideY = blockDim.y * gridDim.y; // start from last column in image for(int x = blockIdx.x * blockDim.x + threadIdx.x; x < sizeX ; x += stride){ // start from last row in image for(int y = blockIdx.y * blockDim.y + threadIdx.y; y < sizeY; y += strideY){ // store numberator and denominator for convolution calculation int numerator = 0; int denominator = 0; // traverse the filter in the x-direction for(int filterX = FILTER_WIDTH - 1; filterX >= 0; filterX--){ // traverse the filter in the y-direction for(int filterY = FILTER_WIDTH -1; filterY >= 0; filterY--){ int xPos = x + filterX -halfFilterWidth; int yPos = y + filterY - halfFilterWidth; // adjust xPos to accomodate edges in grid if(xPos < 0){ xPos = 0; } else if(xPos < sizeX){ } else{ xPos = sizeX - 1; } // adjust yPos to accomodate edges in grid if(yPos < 0){ yPos = 0; } else if(yPos < sizeY){ } else{ yPos = sizeY - 1; } // adjust numerator and denominator numerator += data[yPos * sizeX + xPos] * filter[filterY * FILTER_WIDTH + filterX]; denominator += filter[filterY * FILTER_WIDTH + filterX]; } } // store result result[y * sizeX + x] = numerator/denominator; } } } // GPU implementation void GPU_Test(int data[], int result[], int sizeX, int sizeY) { // input: // int data[] - int array holding the flattened original image // int sizeX - the width of the image // int sizeY - the height of the image // output: // int result[] - int array holding the smoothed image int size = sizeX * sizeY; // Allocate device memory for result[], data[] and FILTER[] and copy data onto the device int *r, *d, *f; cudaMalloc((void**)&r, size*sizeof(int)); cudaMalloc((void**)&d, size*sizeof(int)); cudaMalloc((void**)&f, size*sizeof(int)); cudaMemcpy(r, result, size*sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(d, data, size*sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(f, FILTER, FILTER_WIDTH*FILTER_WIDTH*sizeof(int), cudaMemcpyHostToDevice); // Start timer for kernel auto startKernel = chrono::steady_clock::now(); int numBlocks = (size + BLOCK_SIZE - 1) / BLOCK_SIZE; // Call the kernel function calculateResult<<<numBlocks, BLOCK_SIZE>>>(sizeX, sizeY, d, r,f); // End timer for kernel and display kernel time cudaDeviceSynchronize(); // <- DO NOT REMOVE auto endKernel = chrono::steady_clock::now(); cout << "Kernel Elapsed time: " << chrono::duration <double, milli>(endKernel - startKernel).count() << "ms\n"; // Copy reuslt from device to host cudaMemcpy(result, r, size*sizeof(float), cudaMemcpyDeviceToHost); // Free device memory cudaFree(&d); cudaFree(&r); cudaFree(&f); } // CPU implementation void CPU_Test(int data[], int result[], int sizeX, int sizeY) { // input: // int data[] - int array holding the flattened original image // int sizeX - the width of the image // int sizeY - the height of the image // output: // int result[] - int array holding the smoothed image // Smooth the image with filter size = FILTER_WIDTH // apply partial filter for the border int halfFilterWidth = FILTER_WIDTH/2; // start from last column in image for(int x = sizeX -1; x >= 0; x--){ // start from last row in image for(int y = sizeY - 1; y >= 0; y--){ // store numberator and denominator for convolution calculation int numerator = 0; int denominator = 0; // traverse the filter in the x-direction for(int filterX = FILTER_WIDTH - 1; filterX >= 0; filterX--){ // traverse the filter in the y-direction for(int filterY = FILTER_WIDTH -1; filterY >= 0; filterY--){ int xPos = x + filterX -halfFilterWidth; int yPos = y + filterY - halfFilterWidth; // adjust xPos to accomodate edges in grid if(xPos < 0){ xPos = 0; } else if(xPos < sizeX){ } else{ xPos = sizeX - 1; } // adjust yPos to accomodate edges in grid if(yPos < 0){ yPos = 0; } else if(yPos < sizeY){ } else{ yPos = sizeY - 1; } // adjust numerator and denominator numerator += data[yPos * sizeX + xPos] * FILTER[filterY * FILTER_WIDTH + filterX]; denominator += FILTER[filterY * FILTER_WIDTH + filterX]; } } // store result result[y * sizeX + x] = numerator/denominator; } } } // The input is a 2D grayscale image // The image is flattened into a text file of pixel values. int main(int argc, char *argv[]) { string inputFile = (argc == 1) ? "image2D.txt" : argv[1]; int sizeX; int sizeY; int *dataForCPUTest; int *dataForGPUTest; initData(inputFile, &dataForCPUTest, &sizeX, &sizeY); initData(inputFile, &dataForGPUTest, &sizeX, &sizeY); int size = sizeX * sizeY; int *resultForCPUTest = new int[size]; int *resultForGPUTest = new int[size]; cout << "\n"; cout << "CPU Implementation\n"; auto startCPU = chrono::steady_clock::now(); CPU_Test(dataForCPUTest, resultForCPUTest, sizeX, sizeY); auto endCPU = chrono::steady_clock::now(); cout << "Elapsed time: " << chrono::duration <double, milli>(endCPU - startCPU).count() << "ms\n"; displayResult(dataForCPUTest, resultForCPUTest, size); saveResult("2D_result_CPU.txt",resultForCPUTest, sizeX, sizeY); cout << "\n"; cout << "GPU Implementation\n"; auto startGPU = chrono::steady_clock::now(); GPU_Test(dataForGPUTest, resultForGPUTest, sizeX, sizeY); auto endGPU = chrono::steady_clock::now(); cout << "Elapsed time: " << chrono::duration <double, milli>(endGPU - startGPU).count() << "ms\n"; displayResult(dataForGPUTest, resultForGPUTest, size); saveResult("2D_result_GPU.txt",resultForGPUTest, sizeX, sizeY); return 0; }
19,079
#include <cuda_runtime.h> #include <stdio.h> /* * */ __global__ void poly_div1(float* poli, const int N) { int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx < N) { float x = poli[idx]; poli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+x/5.0; } } __global__ void poly_div2(float* poli, const int N) { int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx < N) { float x = poli[idx]; poli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+x*0.2; } } __global__ void poly_div3(float* poli, const int N) { int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx < N) { float x = poli[idx]; poli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+5.0/x; } } __global__ void poly_div4(float* poli, const int N) { int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx < N) { float x = poli[idx]; float y = 5.0/x; poli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+y; } } __global__ void poly_div5(float* poli, const int N) { int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx < N) { float x = poli[idx]; poli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+1.0/x; } } __global__ void poly_div6(float* poli, const int N) { int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx < N) { float x = poli[idx]; float y = 1.0/x; poli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+y; } } __global__ void poly_div7(float* poli, const int N) { int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx < N) { float x = poli[idx]; poli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+1.0f/x; } } __global__ void poly_div8(float* poli, const int N) { int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx < N) { float x = poli[idx]; poli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+5.0f/x; } } cudaEvent_t start, stop; float elapsed_time; void start_event() { cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start, 0); } void end_event() { cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&elapsed_time, start, stop); printf("Execution time = %.6fms\n",elapsed_time); } int main() { int nElem = 1 << 27; // set up device int dev = 0; cudaDeviceProp deviceProp; cudaGetDeviceProperties(&deviceProp, dev); printf("Using Device %d: %s\n", dev, deviceProp.name); cudaSetDevice(dev); size_t nBytes = nElem * sizeof(float); float* h_polinomy = (float*)malloc(nBytes); float* d_polinomy; float* d_results; cudaMalloc((float**)&d_polinomy, nBytes); cudaMalloc((float**)&d_results, nBytes); int iLen = 512; dim3 block (iLen); dim3 grid ((nElem + block.x - 1) / block.x); start_event(); poly_div2<<<grid, block>>>(d_polinomy, nElem); printf("poli2 "); end_event(); cudaDeviceSynchronize(); printf("poli2[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+x*0.2;\n"); start_event(); poly_div1<<<grid, block>>>(d_polinomy, nElem); printf("poli1 "); end_event(); cudaDeviceSynchronize(); printf("poli1[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+x/5.0;\n"); start_event(); poly_div3<<<grid, block>>>(d_polinomy, nElem); printf("poli3 "); end_event(); printf("poli3[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+5.0/x;\n"); start_event(); poly_div4<<<grid, block>>>(d_polinomy, nElem); printf("poli4 "); end_event(); printf("float y = 5.0/x; \n poli4[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+y;\n"); start_event(); poly_div5<<<grid, block>>>(d_polinomy, nElem); printf("poli5 "); end_event(); printf("poli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+1.0/x;\n"); start_event(); poly_div6<<<grid, block>>>(d_polinomy, nElem); printf("poli6 "); end_event(); printf("float y = 1.0/x;\npoli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+y;\n"); start_event(); poly_div7<<<grid, block>>>(d_polinomy, nElem); printf("poli7" ); end_event(); printf("poli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+1.0f/x;\n"); start_event(); poly_div8<<<grid, block>>>(d_polinomy, nElem); printf("poli8 "); end_event(); printf("poli[idx] = 5 + x * ( 7 - x * (9 + x * (5 + x * (5 + x))))+5.0f/x;\n"); cudaMemcpy(h_polinomy, d_polinomy, nBytes, cudaMemcpyDeviceToHost); cudaFree(d_polinomy); free(h_polinomy); }
19,080
/* This version of my sudoku solver will make use of cuda to attemp to gain speedups */ #include <iostream> #include <fstream> //#include <chrono> #define boardSize 81 #define sideSize 9 using namespace std; struct Board{ int values[81]; bool isFixedValue[81]; bool isPossibleValue[81][9]; //int coordinates; }; void getFileName(char fileName[]){ cout << "Please give the file location of your sudoku board."<<endl; cin.getline(fileName, 255); return; } bool getBoard(char fileName[], Board &mainBoard){ ifstream file(fileName);//input can only occur on host this can't be parallelized if(file.fail()) return false; char temp = '0'; for (int i = 0; i < boardSize; i++){ file >> temp; mainBoard.values[i] = (int)((int)temp - (int)'0'); //cout << mainBoard.values[i] << ","; if(mainBoard.values[i] > 0){ mainBoard.isFixedValue[i] = true; } else{ mainBoard.isFixedValue[i] = false; mainBoard.values[i] = (mainBoard.values[i] / -1); } } file.close(); return true; } __global__ void rowCheckKernel(const int board[], const int row, const int value, bool *hasDuplicates) { int i = (row*sideSize) + threadIdx.x; if(value == board[i]) *hasDuplicates = true;//for some reason this doesn't want to pass back the correct value, always the same one } bool anyDuplicates(int board[], int coordinates, int value) { int row = coordinates / sideSize; int column = coordinates % sideSize; //dim3 grid(sideSize); //bool *hasDuplicates = false; //rowCheckKernel<<<1,grid>>>(board,row,value,hasDuplicates); //bool test = &hasDuplicates; /*if(test){ return test; }*/ for(int i = row * sideSize; i < ((row*sideSize)+sideSize); i++) { if(value == board[i]){ return true; } } for(int i = column; i < boardSize; i+=sideSize) { if(value == board[i]) return true; } while(column%3 != 0) column--; while(row%3 != 0) row--; for(int r = row; r < (row+3); r++) { for(int c = column; c < (column+3); c++) { if(value == board[((r*9)+c)]) return true; } } return false; } __global__ void validateBoardKernel(Board board, bool &isValid) { int tempValue = board.values[threadIdx.x]; board.values[threadIdx.x] = 0; if(tempValue != 0 /*&& anyDuplicates(board.values, threadIdx.x, tempValue)*/){ isValid = false; } } bool validateBoard(Board board){//when I attempted to paralellize this it would for(int i = 0; i < boardSize; i++){ int tempValue = board.values[i]; board.values[i] = 0; if(tempValue != 0 && anyDuplicates(board.values, i, tempValue)){ cout<<"ERROR: Duplicate value '" << tempValue; return false; } board.values[i] = tempValue; } return true; } void displayBoard(Board board) {//this cannot be implemented in parallel I don't think, everything I found said input output had to occur on host ofstream results("resultsParallel.txt"); for(int i = 0; i < boardSize; i++){ results<<board.values[i] <<","; if(i%sideSize == 8) results<<endl; } results.close(); } __global__ void isSolvedKernel(const Board board, bool &solved) { if(board.values[threadIdx.x] == 0) solved = true; } bool isSolved(Board board) { bool solved = true; dim3 grid(boardSize); isSolvedKernel<<<1, grid>>>(board,solved); return solved; } bool canChange(Board mainBoard, int coordinates, int value) {//this function has no behavior worth parallelizing if(value==0) return true; if(mainBoard.isFixedValue[coordinates]) return false; if(anyDuplicates(mainBoard.values, coordinates, value)) return false; return true; } __global__ void checkPossiblesKernel(Board &newBoard, bool &noChanges) { int possibles = 0; int value; if(!newBoard.isFixedValue[threadIdx.x]){ for(int guess = 1; guess <= sideSize; guess++){ if(/*canChange(newBoard, threadIdx.x, guess)can't use host function from global*/guess==0 || !newBoard.isFixedValue[threadIdx.x]){ value = guess; newBoard.isPossibleValue[threadIdx.x][guess-1] = true; possibles++; } else newBoard.isPossibleValue[threadIdx.x][guess-1] = false; } } if(possibles==1){ newBoard.values[threadIdx.x] = value; newBoard.isFixedValue[threadIdx.x] = true; noChanges = false; } } bool checkPossibles(Board &newBoard) { bool noChanges; do{ noChanges = true; for(int i = 0; i < boardSize; i++){ int possibles = 0; int value; if(!newBoard.isFixedValue[i]){ for(int guess = 1; guess <= 9; guess++){ if(canChange(newBoard, i, guess)){ value = guess; newBoard.isPossibleValue[i][guess-1] = true; possibles++; } else newBoard.isPossibleValue[i][guess-1] = false; } } if(possibles==1){ newBoard.values[i] = value; newBoard.isFixedValue[i] = true; noChanges = false; } } }while(noChanges == false); if(isSolved(newBoard)) return true; return false; } /*dfs is very difficult to do in parallel I couldn't figure out how to do it __global__ void dfs(Board &board) { } Board parallelBrute(Board board) { } */ Board recursiveBrute(Board board, int startPosition) { while(startPosition < boardSize && board.isFixedValue[startPosition] && board.values[startPosition] != 0) startPosition++; if(startPosition >= boardSize) return board; for(int guess = 1; guess <= 9; guess++){ Board tempBoard = board; if(board.isPossibleValue[startPosition][guess-1] && canChange(board,startPosition,guess)){ board.values[startPosition] = guess; tempBoard = recursiveBrute(board, startPosition+1); } if(isSolved(tempBoard)) return tempBoard; } board.values[startPosition] = 0; return board; } __global__ void solveKernel(Board &newBoard, bool &noChanges) { for(int guess = 1; guess <= sideSize; guess++){ int total = 0; for(int iRow = threadIdx.x; iRow < (threadIdx.x+sideSize); iRow++){ if(!newBoard.isFixedValue[iRow]) total += newBoard.isPossibleValue[iRow][guess-1]; } if(total == 1){ for(int iRow = threadIdx.x; iRow <(threadIdx.x + sideSize); iRow++){ if(newBoard.isPossibleValue[iRow][guess-1] && !newBoard.isFixedValue[iRow]){ newBoard.values[iRow] = guess; newBoard.isFixedValue[iRow] = true; noChanges = false; } } } } } void solve(Board &newBoard) { bool noChanges; do { noChanges = true; if(checkPossibles(newBoard)) return; for(int i = 0; i < boardSize; i += sideSize){ for(int guess = 1; guess <= 9; guess++){ int total = 0; for(int iRow = i; iRow < (i+9); iRow++){ if(!newBoard.isFixedValue[iRow]) total += newBoard.isPossibleValue[iRow][guess-1]; } if(total == 1){ for(int iRow = i; iRow <(i+9); iRow++){ if(newBoard.isPossibleValue[iRow][guess-1] && !newBoard.isFixedValue[iRow]){ newBoard.values[iRow] = guess; newBoard.isFixedValue[iRow] = true; noChanges = false; } } } } } if(checkPossibles(newBoard)) return; for(int i = 0; i < sideSize; i++){ for(int guess = 1; guess <= 9; guess++){ int total = 0; for(int iColumn = i; iColumn < boardSize; iColumn += sideSize){ if(!newBoard.isFixedValue[iColumn]){ total += newBoard.isPossibleValue[iColumn][guess-1]; } } if(total == 1){ for(int iColumn = i; iColumn < boardSize; iColumn += sideSize){ if(newBoard.isPossibleValue[iColumn][guess-1] && !newBoard.isFixedValue[iColumn]){ newBoard.values[iColumn] = guess; newBoard.isFixedValue[iColumn] = true; noChanges = false; } } } } } if(checkPossibles(newBoard)) return; for(int r = 0; r < boardSize; r+= sideSize*3){ for(int c = 0; c < sideSize; c += 3){ int i = r+c; for(int guess = 1; guess <= 9; guess++){ int total = 0; for(int iBlockR = i; iBlockR < (i+sideSize*3); iBlockR += sideSize){ for(int iBlockC = iBlockR; iBlockC < (iBlockR + 3); iBlockC++){ if(!newBoard.isFixedValue[iBlockC]) total += newBoard.isPossibleValue[iBlockC][guess-1]; } } if(total == 1){ for(int iBlockR = i; iBlockR < (i+sideSize*3); iBlockR += sideSize){ for(int iBlockC = iBlockR; iBlockC < (iBlockR + 3); iBlockC++){ if(newBoard.isPossibleValue[iBlockC][guess-1] && !newBoard.isFixedValue[iBlockC]){ newBoard.values[iBlockC] = guess; newBoard.isFixedValue[iBlockC] = true; noChanges = false; } } } } } } } if(checkPossibles(newBoard)) return; }while(noChanges == false); if(!isSolved(newBoard)) newBoard = recursiveBrute(newBoard, 0); } int main(int argc, char *argv[]){ cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start); char fileName[256]; Board mainBoard; getFileName(fileName); if(!getBoard(fileName,mainBoard)) cout <<"Error with filename"; else if(!validateBoard(mainBoard)){ displayBoard(mainBoard); return 0; } solve(mainBoard); displayBoard(mainBoard); cudaEventRecord(stop); cudaEventSynchronize(stop); float milliseconds = 0; cudaEventElapsedTime(&milliseconds, start, stop); cout <<"Solve time took " << milliseconds << "ms"<<endl; return 0; }
19,081
/* Jaitirth Jacob - 13CO125 Vidit Bhargava - 13CO151 */ #include <stdio.h> #include <stdlib.h> #include <time.h> #define ITERATIONS 4 //Repeat the experiment for greater accuracy __global__ void add(int *a, int *b, int *c, int tpb) { //Find the correct thread index in the grid int i = blockIdx.x * tpb + threadIdx.x; c[i] = a[i] + b[i]; } #define N 1000000 //Array Size #define min_threads 16 #define max_threads 1024 int main(void) { int *a,*b,*c; int *d_a, *d_b, *d_c; int size = N * sizeof(int); a = (int *)malloc(size); b = (int *)malloc(size); c = (int *)malloc(size); //Allocate on device cudaMalloc((void **)&d_a, size); cudaMalloc((void **)&d_b, size); cudaMalloc((void **)&d_c, size); srand(time(NULL)); //Populate a and b for (int i = 0; i < N; ++i) { a[i] = rand()%20; b[i] = rand()%37; } int numBlocks; cudaEvent_t start, copy, exec, result; //Events for measuring time //To calculate average over a number of iterations float t1[7], t2[7], t3[7], total[7]; for (int i = 0; i < 7; ++i) { t1[i]=0; t2[i]=0; t3[i]=0; total[i]=0; } printf("t1: time for copying arrays\n"); printf("t2: time for kernel execution\n"); printf("t3: time for copying result back\n\n"); printf("All times in milliseconds\n"); printf("TPB\t\tNB\t\tt1\t\tt2\t\tt3\t\ttotal\t\n"); int count; for (int i = 0; i < ITERATIONS; ++i) { count=0; for (int threadsPerBlock = min_threads; threadsPerBlock <= max_threads; threadsPerBlock*=2) { numBlocks = (N + threadsPerBlock - 1)/threadsPerBlock; cudaEventCreate(&start); cudaEventCreate(&copy); cudaEventCreate(&exec); cudaEventCreate(&result); cudaEventRecord(start); //Copy inputs to device cudaMemcpy(d_a, a, size, cudaMemcpyHostToDevice); cudaMemcpy(d_b, b, size, cudaMemcpyHostToDevice); cudaEventRecord(copy); cudaEventSynchronize(copy); //Launch add() kernel on GPU add<<<numBlocks,threadsPerBlock>>>(d_a, d_b, d_c, threadsPerBlock); cudaEventRecord(exec); cudaEventSynchronize(exec); //Copy result back to host cudaMemcpy(c, d_c, size, cudaMemcpyDeviceToHost); cudaEventRecord(result); cudaEventSynchronize(result); float temp1=0, temp2=0, temp3=0, temptotal; cudaEventElapsedTime(&temp1, start, copy); cudaEventElapsedTime(&temp2, copy, exec); cudaEventElapsedTime(&temp3, exec, result); cudaEventElapsedTime(&temptotal, start, result); t1[count] += temp1; t2[count] += temp2; t3[count] += temp3; total[count] += temptotal; cudaEventDestroy(start); cudaEventDestroy(copy); cudaEventDestroy(exec); cudaEventDestroy(result); count++; } } int threadsPerBlock = min_threads; for (int i = 0; i < 7; ++i) { numBlocks = (N + threadsPerBlock - 1)/threadsPerBlock; t1[i]/=(float)ITERATIONS; t2[i]/=(float)ITERATIONS; t3[i]/=(float)ITERATIONS; total[i]/=(float)ITERATIONS; printf("%d\t\t%d\t\t%.5f\t\t%.5f\t\t%.5f\t\t%.5f\t\t\n", threadsPerBlock, numBlocks, t1[i], t2[i], t3[i], total[i]); threadsPerBlock*=2; } //Cleanup free(a); free(b); free(c); cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); return 0; }
19,082
#include <stdio.h> // Prints info about the device // Takes in the device number and a pointer to the properties void printProperties(int i, cudaDeviceProp *prop){ printf( " --- General Information for device %d ---\n", i ); printf( "Name: %s\n", prop->name ); } int main( void ) { cudaDeviceProp *prop; int count; //Doesn't handle errors cudaGetDeviceCount(&count); printf("Device Count: %d\n", count); while(count > 0){ //Doesn't handle errors cudaGetDeviceProperties( prop, count ); printProperties(count, prop); printf( " --- General Information for device %d ---\n", count ); printf( "Name: %s\n", prop->name ); count--; } return 0; }
19,083
extern "C"{ __global__ void convolution_1D_basic_kernel(int *N, int *M, int *P, int Mask_Width, int Width){ int i = blockIdx.x*blockDim.x + threadIdx.x; float Pvalue = 0; int N_start_point = i - (Mask_Width / 2); for (int j = 0; j < Mask_Width; j++){ if(N_start_point + j >= 0 && N_start_point + j < Width){ Pvalue += N[N_start_point + j] * M[j]; } } P[i] = Pvalue; } }
19,084
#include <cstdio> template<typename T> __device__ __inline__ void add(T& val) { val += 1; } template<typename T> __global__ void func(T* ptr) { add<T>(ptr[blockIdx.x]); } int main() { cudaStream_t stream; cudaStreamCreate(&stream); int *h_ptr, *d_ptr; cudaHostAlloc(&h_ptr, 20, cudaHostAllocDefault); for(int i=0; i<5; i++) { h_ptr[i] = i*2 + 1; } cudaMalloc((void**)&d_ptr, 20); cudaMemcpyAsync(d_ptr, h_ptr, 20, cudaMemcpyHostToDevice, stream); func<<<5, 1, 0, stream>>>(d_ptr); cudaMemcpyAsync(h_ptr, d_ptr, 20, cudaMemcpyDeviceToHost, stream); cudaStreamSynchronize(stream); for(int i=0; i<5; i++) { printf("%d ", h_ptr[i]); } printf("\n"); return 0; }
19,085
#include <stdio.h> #include <stdlib.h> #define MAX_ITER 100 #define MAX 100 //maximum value of the matrix element #define TOL 0.000001 // Generate a random float number with the maximum value of max float rand_float(int max){ return ((float)rand()/(float)(RAND_MAX)) * max; } // Allocate 2D matrix void allocate_init_2Dmatrix(float ***mat, int n, int m){ int i, j; *mat = (float **) malloc(n * sizeof(float *)); for(i = 0; i < n; i++) { (*mat)[i] = (float *)malloc(m * sizeof(float)); for (j = 0; j < m; j++) (*mat)[i][j] = rand_float(MAX); } } // solver void solver(float ***mat, int n, int m){ float diff = 0, temp; int done = 0, cnt_iter = 0, i, j; while (!done && (cnt_iter < MAX_ITER)){ diff = 0; for (i = 1; i < n - 1; i++) for (j = 1; j < m - 1; j++){ temp = (*mat)[i][j]; (*mat)[i][j] = 0.2 * ((*mat)[i][j] + (*mat)[i][j - 1] + (*mat)[i - 1][j] + (*mat)[i][j + 1] + (*mat)[i + 1][j]); diff += abs((*mat)[i][j] - temp); } if (diff/n/n < TOL) done = 1; cnt_iter ++; } if (done) printf("Solver converged after %d iterations\n", cnt_iter); else printf("Solver not converged after %d iterations\n", cnt_iter); } int main(int argc, char *argv[]) { int n, communication; float **a; if (argc < 2) { printf("Call this program with two parameters: matrix_size communication \n"); printf("\t matrix_size: Add 2 to a power of 2 (e.g. : 18, 1026)\n"); exit(1); } n = atoi(argv[1]); printf("Matrix size = %d\n", n); allocate_init_2Dmatrix(&a, n, n); solver(&a, n, n); return 0; }
19,086
// This is the naive implementation of in box check with all matrices squeezed to vector __global__ void inBoxKernel(const float *A, const float *B, int *C, int numElements){ int i = (blockDim.x * blockIdx.x + threadIdx.x)*2; float t11; float t12; float t21; float t22; if (i/2 < numElements) { if ((A[i]<B[0] || (A[i] >B[4]))||((A[i+1]>B[3]) || (A[i+1] <B[7]))) { C[i/2]=-1; } else { t11=(A[i]-B[0])*(B[2]-B[0])+(A[i+1]-B[1])*(B[3]-B[1]); t12=(A[i]-B[4])*(B[6]-B[4])+(A[i+1]-B[5])*(B[7]-B[5]); t21=(A[i]-B[2])*(B[4]-B[2])+(A[i+1]-B[3])*(B[5]-B[3]); t22=(A[i]-B[6])*(B[2]-B[6])+(A[i+1]-B[7])*(B[1]-B[7]); if (t11>=1e-20 && t12>=1e-20 && t21>=1e-20 && t22>=1e-20) { C[i/2]=1; } else { C[i/2]=-1; } } } } void inBoxLauncher(const float *A, const float *B, int *C, int numElements){ int threadsPerBlock = 512; int blocksPerGrid =(numElements - 1) / threadsPerBlock +1; inBoxKernel<<<blocksPerGrid,threadsPerBlock>>>(A, B, C, numElements); }
19,087
// The code which is causing the pointer pointer address space error: // %"struct.Eigen::half_impl::__half" = type { i16 } // %"struct.Eigen::half_impl::half_base" = type { %"struct.Eigen::half_impl::__half" } // %"struct.Eigen::half" = type { %"struct.Eigen::half_impl::half_base" } // %"struct.Eigen::DSizes" = type { %"class.Eigen::array" } // %"class.Eigen::array" = type { [1 x i32] } // %"struct.Eigen::GpuDevice" = type <{ %"class.Eigen::StreamInterface"*, i32, [4 x i8] }> // %"class.Eigen::StreamInterface" = type { i32 (...)** } // %"struct.Eigen::TensorEvaluator.0" = type { %"struct.Eigen::half"*, %"struct.Eigen::DSizes", %"struct.Eigen::GpuDevice"* } // %"struct.Eigen::TensorEvaluator.2" = type { %"struct.Eigen::half"*, %"struct.Eigen::DSizes", %"struct.Eigen::GpuDevice"* } // %"struct.Eigen::TensorEvaluator.7" = type { %"struct.Eigen::internal::scalar_left", %"struct.Eigen::TensorEvaluator.2" } // %"struct.Eigen::TensorEvaluator.6" = type { %"struct.Eigen::TensorEvaluator.0", %"struct.Eigen::TensorEvaluator.7" } // ; Function Attrs: norecurse nounwind // define weak_odr void @_ZN5Eigen8internal15EigenMetaKernelINS_15TensorEvaluatorIKNS_14TensorAssignOpINS_9TensorMapINS_6TensorINS_4halfELi1ELi1EiEELi16EEEKNS_18TensorCwiseUnaryOpINS0_11scalar_leftIS6_S6_NS0_13scalar_sum_opIS6_S6_EEEEKNS4_INS5_IKS6_Li1ELi1EiEELi16EEEEEEENS_9GpuDeviceEEEiEEvT_T0_(%"struct.Eigen::TensorEvaluator.6"* byval nocapture readonly align 8 %memcopied_eval, i32 %size) #1 comdat { // %1 = tail call i32 @llvm.ptx.read.ctaid.x() #6 // %2 = tail call i32 @llvm.ptx.read.ntid.x() #6 // %3 = mul i32 %2, %1 // %4 = tail call i32 @llvm.ptx.read.tid.x() #6 // %5 = add i32 %3, %4 // %6 = tail call i32 @llvm.ptx.read.nctaid.x() #6 // %7 = mul i32 %6, %2 // %8 = getelementptr inbounds %"struct.Eigen::TensorEvaluator.6", %"struct.Eigen::TensorEvaluator.6"* %memcopied_eval, i64 0, i32 0, i32 0 // %9 = load %"struct.Eigen::half"*, %"struct.Eigen::half"** %8, align 8 struct MyStruct { float floatvalue; int intvalue; }; __device__ __host__ float sumStruct(struct MyStruct **p_structs, int N) { float sum = 0; for(int i = 0; i < N; i++) { struct MyStruct *mystruct = p_structs[i]; sum += mystruct->floatvalue + float(mystruct->intvalue) * 3.5f; } return sum; } __global__ void mykernel(float *data, MyStruct *structs, int N) { data[0] = sumStruct(&structs, N); data[3] = sumStruct(&structs, 123); data[4] = sumStruct(&structs, 12300); } class HalfImpl { public: short myshort; }; class HalfBase : public HalfImpl { }; class Half : public HalfBase { }; class StreamInterface { }; class GpuDevice { public: int myint; StreamInterface *pstreamInterface; }; class TensorEvaluator0 { public: Half *phalf; GpuDevice gpuDevice; }; class TensorEvaluator2 { public: Half *phalf; GpuDevice gpuDevice; }; class TensorEvaluator7 { public: Half *phalf; TensorEvaluator2 te2; }; class TensorEvaluator6 { public: class TensorEvaluator0 te0; class TensorEvaluator7 te7; }; // __device__ __host__ float getGpuDeviceValue(GpuDevice *devices, int N) { // float sum = 0; // for(int i = 0; i < N; i++) { // sum += devices[i].myfloat * (5.5f + i); // } // return sum; // } __device__ __host__ float getHalfValue(Half *half_, int a) { return (float)(half_[a].myshort + 123); } __global__ void myte6kernel(TensorEvaluator6 *structs, float *data, GpuDevice *gpudevices, int a, int b, int c) { Half *phalf = structs[a].te0.phalf; data[0] = getHalfValue(phalf, a); //gpudevices[b] = structs[a].te0.gpuDevice; // float sum = 0; // for(int i = 0; i < 1000; i++) { // sum += getGpuDeviceValue(&structs[i].te0.gpuDevice, c + i); // } // data[1] = sum; }
19,088
/****************************************************************************** *cr *cr (C) Copyright 2010 The Board of Trustees of the *cr University of Illinois *cr All Rights Reserved *cr ******************************************************************************/ // Define your kernels in this file you may use more than one kernel if you // need to #define BLOCK_SIZE 512 // INSERT KERNEL(S) HERE __global__ void fillBins(unsigned int* input, unsigned int* bins, unsigned int num_elements, unsigned int num_bins) { extern __shared__ unsigned int private_histogram[]; //Initializing private histogram bins int bin_stride = 0; while((threadIdx.x + bin_stride) < num_bins) { private_histogram[threadIdx.x + bin_stride] = 0; bin_stride += blockDim.x; } __syncthreads(); //Computation of private histogram int i = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x; while(i < num_elements) { atomicAdd(&private_histogram[input[i]], 1); i += stride; } __syncthreads(); //Merging private history bins with global history bins bin_stride = 0; while((threadIdx.x + bin_stride) < num_bins) { atomicAdd(&bins[threadIdx.x + bin_stride], private_histogram[threadIdx.x + bin_stride]); bin_stride += blockDim.x; } __syncthreads(); } /****************************************************************************** Setup and invoke your kernel(s) in this function. You may also allocate more GPU memory if you need to *******************************************************************************/ void histogram(unsigned int* input, unsigned int* bins, unsigned int num_elements, unsigned int num_bins) { // INSERT CODE HERE dim3 dimGrid((num_elements - 1)/BLOCK_SIZE + 1, 1, 1); dim3 dimBlock(BLOCK_SIZE, 1, 1); fillBins<<<dimGrid, dimBlock, num_bins * sizeof(unsigned int)>>>(input, bins, num_elements, num_bins); }
19,089
/* nqueens.cu * Jonathan Lehman * February 26, 2012 * * Compile with: nvcc -o nqueens nqueens.cu * to get default with _N_ = 4 and numBX = 1 numBY = 1 sumOnGPU = 0 * * Or specify _N_ by compiling with: nvcc -o nqueens nqueens.cu -D_N_=x * where x is the board size desired where x must be >= 4 and <= 22 * * and/Or specify numBX by compiling with: nvcc -o nqueens nqueens.cu -DnumBX=y * where y is the number of tuple values to be generated by blockIdx.x * where y must be >= 1 such that N^numBX < maxgridsize (in this case 65535 blocks) * * and/or specify numBY by compiling with nvcc -o nqueens nqueens.cu -DnumBY=z * where z is the number of groups of ((N / 2) + (N % 2)) columns by N^numBX rows that work on the solution * essentially, this evenly divides the work of the tuples being generated iteratively by each thread between each group * where z must be <= N^numBX * * and/or specify whether or not to add the block totals on the GPU or cpu with nvcc -o nqueens nqueens.cu -DsumOnGPU=a * where a is 1 or 0, with 1 doing the sum on the GPU and 0 doing the sum on the CPU * */ #include <cuda.h> #include <stdio.h> #include <math.h> #include <sys/time.h> __global__ void queen(long*, int); __device__ void sumBlocks(long *); void checkArgs(int, char**, int); void checkGPUCapabilities(int, int, int, int, int); double getTime(); //set board size #ifndef _N_ #define _N_ 4 #endif //set the number of values in the tuple BlockIdx.x should be responsible for #ifndef numBX #define numBX 1 #endif #ifndef numBY #define numBY 1 #endif //number of values in tuple to be generated by thread (incrementally) //#ifndef numGen #define numGen _N_ - 3 - numBX //#endif //whether or not the sum of blocksums (solution should be summed on GPU or CPU) //CPU by default //Set to 1 to add on GPU #ifndef sumOnGPU #define sumOnGPU 0 #endif // Keep track of the gpu time. cudaEvent_t start, stop; float elapsedTime; // Keep track of the CPU time. double startTime, stopTime; //array for block sums long *a; int main(int argc, char *argv[]){ /*check errors with macros*/ //ensure number of tuples generated iteratively is not less than 0 if(numGen < 0){ fprintf(stderr, "\nnqeens: The number of values in the tuple generated iteratively cannot be less than 0.\n NumGen = _N_(%d) - 3 - numBX(%d) = %d\n", _N_, numBX, numGen); exit(1); } //ensure N is in the correct range if(_N_ < 4 || _N_ > 22){ fprintf(stderr, "\nnqeens: _N_(%d) must be between 4 and 22 inclusive\n", _N_); exit(1); } //ensure that at least one of the tuple values is generated by the block's X coordinate value if(numBX < 1){ fprintf(stderr, "\nnqeens: The number of tuples generated by each block's X coordinate value (numBX=%d) must be >= 1\n", numBX); exit(1); } //ensure that the number of Y segments that the numGen work is divided into //is at least one per work segment if(numBY > pow(_N_, numGen)){ fprintf(stderr, "\nnqeens: numBY(%d) must be less than or equal to _N_^numGen(%d)\n", numBY, pow(_N_, numGen)); exit(1); } long *dev_a; //check validity of arguments (should be no arguments) checkArgs(argc, argv, 1); int gW, gH, numberBlocks; //calculate grid width based on factor N, gW = pow(_N_, numBX); //depends on if N is even or odd int sizePerYSeg = (_N_ / 2) + (_N_ % 2); gH = sizePerYSeg * numBY; numberBlocks = gW * gH; //check that GPU can handle arguments checkGPUCapabilities(gW, gH, _N_, _N_, numberBlocks); /* Initialize the source arrays here. */ a = new long[numberBlocks]; /* Allocate global device memory. */ cudaMalloc((void **)&dev_a, sizeof(long) * numberBlocks); /* Start the timer. */ cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start, 0); /* Execute the kernel. */ dim3 block(_N_, _N_); //threads w x h dim3 grid(gW, gH); //blocks w x h queen<<<grid, block>>>(dev_a, sizePerYSeg); /* Wait for the kernel to complete. Needed for timing. */ cudaThreadSynchronize(); /* Stop the timer and print the resulting time. */ cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&elapsedTime, start, stop); /* Get result from device. */ cudaMemcpy(a, dev_a, sizeof(long) * numberBlocks, cudaMemcpyDeviceToHost); //print any cuda error messages const char* errorString = cudaGetErrorString(cudaGetLastError()); printf("GPU Error: %s\n", errorString); if(sumOnGPU){ printf("Number of Solutions:%d\n", a[0]); //add cpu time and gpu time and print result printf( "GPU Time/Total Time: %f secs\n", (elapsedTime / 1000.0)); } else{ /* Start the CPU timer. */ startTime = getTime(); int sum = 0; //check if N is even or odd, then calculate sum, which is number of solutions if(_N_ % 2 == 0){ for(int i = 0; i < numberBlocks; i++){ sum+= a[i]; } sum *= 2; } else{ int numBlocksPerSeg = numberBlocks / numBY; int rowSizeOfGrid = pow(_N_, numBX); for(int j = 0; j < numBY; j++){ int start = j * numBlocksPerSeg; for(int i = start; i < start + numBlocksPerSeg - rowSizeOfGrid; i++){ sum+= a[i]; } } sum *= 2; //add last block row of sums for each Y block for(int j = 0; j < numBY; j++){ for(int i = j * numBlocksPerSeg + numBlocksPerSeg - rowSizeOfGrid; i < j * numBlocksPerSeg + numBlocksPerSeg; i++){ sum+= a[i]; } } } /* Stop the CPU timer */ stopTime = getTime(); double totalTime = stopTime - startTime; printf("Number of Solutions: %d\n", sum); //add cpu time and gpu time and print result printf( "GPU Time: %f secs\nCPU Time: %f secs\nTotal Time: %f secs\n", (elapsedTime / 1000.0), totalTime, (elapsedTime / 1000.0) + totalTime ); } //destroy cuda event cudaEventDestroy(start); cudaEventDestroy(stop); /* Free the allocated device memory. */ cudaFree(dev_a); //free allocated host memory free(a); } __global__ void queen(long *a, int sizePerYSeg){ __shared__ long solutions[_N_][_N_]; __shared__ char tuple[_N_][_N_][_N_]; int totalWrong = 0; solutions[threadIdx.x][threadIdx.y] = 0; int totNumGen = powf(_N_, numGen); int bYsegment = blockIdx.y / sizePerYSeg; int workSize = totNumGen / numBY; int extra = totNumGen - workSize * numBY;//extra work to be done by last segment //set tuple by block Y value tuple[threadIdx.x][threadIdx.y][0] = blockIdx.y % sizePerYSeg; //set tuple(s) by block X value int rem = blockIdx.x; for(int i = 1; i <= numBX; i++){ tuple[threadIdx.x][threadIdx.y][i] = rem % _N_; rem = rem / _N_; } int tupCtr = numBX; //set tuples by thread value tuple[threadIdx.x][threadIdx.y][++tupCtr] = threadIdx.x; tuple[threadIdx.x][threadIdx.y][++tupCtr] = threadIdx.y; //check if thread is valid at this point for(int i = tupCtr; i > 0; i--){ for(int j = i - 1, ctr = 1; j >= 0; j--, ctr++){ //same row totalWrong += tuple[threadIdx.x][threadIdx.y][i] == tuple[threadIdx.x][threadIdx.y][j]; //diag upleft totalWrong += (tuple[threadIdx.x][threadIdx.y][i] - ctr) == tuple[threadIdx.x][threadIdx.y][j]; //diag downleft totalWrong += (tuple[threadIdx.x][threadIdx.y][i] + ctr) == tuple[threadIdx.x][threadIdx.y][j]; } } if(totalWrong == 0){ //iterate through all numbers to generate possible solutions thread must check //does not do if thread is already not valid at this point int start = bYsegment * workSize; for(int c = start; c < start + workSize + (bYsegment == numBY - 1) * extra; c++){ //generate last values in tuple, convert to base N and store to tuple array int rem = c; for(int b = 0, k = tupCtr + 1; b < numGen; b++, k++){ tuple[threadIdx.x][threadIdx.y][k] = rem % _N_; rem = rem / _N_; } //checks that the numGen tuple values are indeed unique (saves work overall) for(int x = 0; x < numGen && totalWrong == 0; x++){ for(int y = 0; y < numGen && totalWrong == 0; y++){ totalWrong += tuple[threadIdx.x][threadIdx.y][tupCtr + 1 + x] == tuple[threadIdx.x][threadIdx.y][tupCtr + 1 + y] && x != y; } } //check one solution for(int i = _N_ - 1; i > totalWrong * _N_; i--){ for(int j = i - 1, ctr = 1; j >= 0; j--, ctr++){ //same row totalWrong += tuple[threadIdx.x][threadIdx.y][i] == tuple[threadIdx.x][threadIdx.y][j]; //diag upleft totalWrong += (tuple[threadIdx.x][threadIdx.y][i] - ctr) == tuple[threadIdx.x][threadIdx.y][j]; //diag downleft totalWrong += (tuple[threadIdx.x][threadIdx.y][i] + ctr) == tuple[threadIdx.x][threadIdx.y][j]; } } //add 1 to solution total if nothing wrong solutions[threadIdx.x][threadIdx.y] += !(totalWrong); //reset total wrong totalWrong = 0; } } //sync the threads so that thread 0 can make the calculations __syncthreads(); //have thread 0 sum for all threads in block to get block total if(threadIdx.x == 0 && threadIdx.y == 0){ //ensure that the block total value is 0 initially long sum = 0; //iterate through each threads solution and add it to the block total for(int i =0; i < _N_; i++){ for(int j = 0; j < _N_; j++){ //use local var sum += solutions[i][j]; } } //store to global memory a[gridDim.x * blockIdx.y + blockIdx.x] = sum; } //sync the threads so that calculations can be made __syncthreads(); //have the first thread in the first block sum up the block sums to return to the CPU if(sumOnGPU == 1 && blockIdx.x == 0 && blockIdx.y == 0 && threadIdx.x == 0 && threadIdx.y == 0){ sumBlocks(a); } } __device__ void sumBlocks(long *a){ long sum = 0; int numberBlocks = gridDim.x * gridDim.y; int rowSizeOfGrid = powf(_N_, numBX); //check if N is even or odd, then calculate sum, which is number of solutions if(_N_ % 2 == 0){ for(int i = 0; i < numberBlocks; i++){ sum+= a[i]; } sum *= 2; } else{ int numBlocksPerSeg = numberBlocks / numBY; for(int j = 0; j < numBY; j++){ int start = j * numBlocksPerSeg; for(int i = start; i < start + numBlocksPerSeg - rowSizeOfGrid; i++){ sum+= a[i]; } } sum *= 2; //add last block row of sums for each Y block for(int j = 0; j < numBY; j++){ for(int i = j * numBlocksPerSeg + numBlocksPerSeg - rowSizeOfGrid; i < j * numBlocksPerSeg + numBlocksPerSeg; i++){ sum+= a[i]; } } } //store sum to first index of a a[gridDim.x * blockIdx.y + blockIdx.x] = 0; a[gridDim.x * blockIdx.y + blockIdx.x] = sum; } void checkArgs(int argc, char *argv[], int numArgs){ //check number of arguments if(argc > numArgs){ fprintf(stderr, "\nnqueens: Incorrect number of arguments, %d\nCorrect usage: \"nqueens\"\n", argc - 1); exit(1); } char* invalChar; long arg; //check each argument for(int i = 1; i < numArgs; i++){ //check for overflow of argument if((arg = strtol(argv[i], &invalChar, 10)) >= INT_MAX){ fprintf(stderr, "\nnqueens: Overflow. Invalid argument %d for nqueens, '%s'.\nThe argument must be a valid, positive, non-zero integer less than %d.\n", i, argv[i], INT_MAX); exit(1); } //check that argument is a valid positive integer and check underflow if(!(arg > 0) || (*invalChar)){ fprintf(stderr, "\nnqueens: Invalid argument %d for nqueens, '%s'. The argument must be a valid, positive, non-zero integer.\n", i, argv[i]); exit(1); } } } void checkGPUCapabilities(int gridW, int gridH, int blockW, int blockH, int size){ //check what GPU is being used int devId; cudaGetDevice( &devId ); //get device properties for GPU being used cudaDeviceProp gpuProp; cudaGetDeviceProperties( &gpuProp, devId ); //check if GPU has enough memory if(gpuProp.totalGlobalMem < (size * sizeof(long))){ fprintf(stderr, "\nnqueens: Insufficient GPU. GPU does not have enough memory to handle the data size: %ld. It can only handle data sizes up to %ld.\n", (size * sizeof(float)) * 3, gpuProp.totalGlobalMem); exit(1); } //check if GPU can handle the number of threads per bloc if(gpuProp.maxThreadsPerBlock < (blockW * blockH)){ fprintf(stderr, "\nnqueens: Insufficient GPU. GPU can only handle %d threads per block, not %d.\n", gpuProp.maxThreadsPerBlock, (blockW * blockH)); exit(1); } //check that GPU can handle the number of threads in the block width if(gpuProp.maxThreadsDim[0] < blockW){ fprintf(stderr, "\nnqueens: Insufficient GPU. GPU can only handle %d threads as the block width of each block, not %d.\n", gpuProp.maxThreadsDim[0], blockW ); exit(1); } //check that GPU can handle the number of threads in the block height if(gpuProp.maxThreadsDim[1] < blockH){ fprintf(stderr, "\nnqueens: Insufficient GPU. GPU can only handle %d threads as the block height of each block, not %d.\n", gpuProp.maxThreadsDim[1], blockH ); exit(1); } //check that GPU can handle the number of blocks in the grid width if(gpuProp.maxGridSize[0] < gridW){ fprintf(stderr, "\nnqueens: Insufficient GPU. GPU can only handle %d blocks as the grid width of each grid, not %d.\n", gpuProp.maxGridSize[0], gridW ); exit(1); } //check that GPU can handle the number of blocks in the grid height if(gpuProp.maxGridSize[1] < gridH){ fprintf(stderr, "\nnqueens: Insufficient GPU. GPU can only handle %d blocks as the grid height of each grid, not %d.\n", gpuProp.maxGridSize[1], gridH ); exit(1); } } double getTime(){ timeval thetime; gettimeofday(&thetime, 0); return thetime.tv_sec + thetime.tv_usec / 1000000.0; }
19,090
/* Final project of NVIDIA Fundamentals of CUDA in C/C++ Consits of a simulation of the n-body problem. */ #include <math.h> #include <stdio.h> #include <stdlib.h> #include <time.h> #define SOFTENING 1e-9f /* * Each body contains x, y, and z coordinate positions, * as well as velocities in the x, y, and z directions. */ typedef struct { float x, y, z, vx, vy, vz; } Body; /* * Do not modify this function. A constraint of this exercise is * that it remain a host function. */ void randomizeBodies(float *data, int n) { for (int i = 0; i < n; i++) { data[i] = 2.0f * (rand() / (float)RAND_MAX) - 1.0f; } } /* * This function calculates the gravitational impact of all bodies in the system * on all others, but does not update their positions. */ __global__ void bodyForce(Body *p, float dt, int n) { int tidx = threadIdx.x + blockDim.x*blockIdx.x; int slidex = blockDim.x*gridDim.x; for (int i = tidx; i < n; i+=slidex) { float Fx = 0.0f; float Fy = 0.0f; float Fz = 0.0f; for (int j = 0; j < n; j++) { float dx = p[j].x - p[i].x; float dy = p[j].y - p[i].y; float dz = p[j].z - p[i].z; float distSqr = dx*dx + dy*dy + dz*dz + SOFTENING; float invDist = rsqrtf(distSqr); float invDist3 = invDist * invDist * invDist; Fx += dx * invDist3; Fy += dy * invDist3; Fz += dz * invDist3; } p[i].vx += dt*Fx; p[i].vy += dt*Fy; p[i].vz += dt*Fz; } } __global__ void integratePosition(Body *p, float dt, int n){ int tid = threadIdx.x + blockDim.x*blockIdx.x; int slide = blockDim.x*gridDim.x; for(int i=tid;i<n;i+=slide){ p[i].x += p[i].vx*dt; p[i].y += p[i].vy*dt; p[i].z += p[i].vz*dt; } } int main(const int argc, const char** argv) { /* * Do not change the value for `nBodies` here. If you would like to modify it, * pass values into the command line. */ int nBodies = 2<<11; if (argc > 1) nBodies = 2<<atoi(argv[1]); const float dt = 0.01f; // time step const int nIters = 10; // simulation iterations int bytes = nBodies * sizeof(Body); float *buf; int deviceId; cudaGetDevice(&deviceId); cudaDeviceProp props; cudaGetDeviceProperties(&props, deviceId); int size_warps = props.multiProcessorCount; cudaMallocHost(&buf, bytes); Body *p; cudaMalloc(&p, bytes); /* * As a constraint of this exercise, `randomizeBodies` must remain a host function. */ randomizeBodies(buf, 6 * nBodies); // Init pos / vel data cudaMemcpy(p, buf, bytes, cudaMemcpyHostToDevice); clock_t start = clock(); /* * This simulation will run for 10 cycles of time, calculating gravitational * interaction amongst bodies, and adjusting their positions to reflect. */ /*******************************************************************/ // Do not modify these 2 lines of code. for (int iter = 0; iter < nIters; iter++) { /*******************************************************************/ /* * You will likely wish to refactor the work being done in `bodyForce`, * as well as the work to integrate the positions. */ dim3 num_threads(128); dim3 num_blocks(size_warps*8); bodyForce<<<num_blocks, num_threads>>>(p, dt, nBodies); // compute interbody forces cudaDeviceSynchronize(); /* * This position integration cannot occur until this round of `bodyForce` has completed. * Also, the next round of `bodyForce` cannot begin until the integration is complete. */ integratePosition<<<num_blocks, num_threads>>>(p, dt, nBodies); // integrate position cudaDeviceSynchronize(); } double totalTime = (double)(clock()-start)/CLOCKS_PER_SEC; cudaMemcpy(buf, p, bytes, cudaMemcpyDeviceToHost); double avgTime = totalTime / (double)(nIters); double billionsOfOpsPerSecond = (1e-9 * nBodies * nBodies) / (avgTime); printf("%d Bodies: average %0.3f Billion Interactions / second\n", nBodies, billionsOfOpsPerSecond); /*******************************************************************/ /* * Feel free to modify code below. */ cudaFree(p); cudaFreeHost(buf); }
19,091
#include <stdio.h> #include <stdlib.h> #include <math.h> int row_counter(FILE* fp); int col_counter(FILE* fp); void read_matrix(FILE* fp,int *data); void print_matrix(int *data,int mRr, int mRc); __global__ void matrix_multiplication(int *m1,int *m2, int *mR, int m1r, int m1c, int m2c) { int row = blockIdx.y * blockDim.y + threadIdx.y; int col = blockIdx.x * blockDim.x + threadIdx.x; int sum = 0; int i = 0; if( col < m2c && row < m1r) { for(i = 0; i < m1c; i++) { sum += m1[row * m1c + i] * m2[i * m2c + col]; } mR[row * m2c + col] = sum; } } int main(int argc, char **argv) { if( 3 != argc ){ printf("Incorrect number of params: %d\n",argc - 1); return 1; } //Vars Declaration int m1r,m1c,m2r,m2c,mRr,mRc; //rows & cols int m1s,m2s,mRs; //size int *gpu_m1, *cpu_m1; int *gpu_m2, *cpu_m2; int *gpu_mR, *cpu_mR; cudaError_t err = cudaSuccess; // Open the files FILE* fp1 = fopen(argv[1], "r"); FILE* fp2 = fopen(argv[2], "r"); // Check if files exists if (fp1 == NULL) { printf("Could not open file %s",argv[1] ); return 1; } if (fp2 == NULL) { printf("Could not open file %s",argv[2] ); return 1; } //Read the number of rows and cols of each matrix m1r = row_counter(fp1); m1c = col_counter(fp1); m2r = row_counter(fp2); m2c = col_counter(fp2); mRr = m1r; mRc = m2c; //printf("Matrix 1: %dx%d\n", m1r,m1c); //printf("Matrix 2: %dx%d\n", m2r,m2c); //condition: the number of columns in A must equal the number of rows in B if(m1c != m2r){ printf("Incorrect Matrix Size, cols of Matrix 1 (%d) are different of the rows of Matrix 2 (%d)\n",m1c,m2r ); return 1; } //reserve memory to each matrix m1s = m1r*m1c*sizeof(int); m2s = m2r*m2c*sizeof(int); mRs = mRr*mRc*sizeof(int); cpu_m1 = (int*)malloc(m1s); cpu_m2 = (int*)malloc(m2s); cpu_mR = (int*)malloc(mRs); err = cudaMalloc((void**)&gpu_m1,m1s); if(err != cudaSuccess){printf("Error with Matrix 1\n");exit(1);} err = cudaMalloc((void**)&gpu_m2,m2s); if(err != cudaSuccess){printf("Error with Matrix 2\n");exit(1);} err = cudaMalloc((void**)&gpu_mR,mRs); if(err != cudaSuccess){printf("Error with Matrix R\n");exit(1);} //Read the Files to the CPU memory read_matrix(fp1,cpu_m1); read_matrix(fp2,cpu_m2); //Now the files can be closed fclose(fp1); fclose(fp2); //Copy each matrix to the device err = cudaMemcpy(gpu_m1, cpu_m1, m1s, cudaMemcpyHostToDevice); if(err != cudaSuccess){printf("Error Coping Matrix 1\n");exit(1);} err = cudaMemcpy(gpu_m2, cpu_m2, m2s, cudaMemcpyHostToDevice); if(err != cudaSuccess){printf("Error Coping Matrix 2\n");exit(1);} //Execute the Kernel in the Device int tile_dim = 32; dim3 block_dim(tile_dim,tile_dim,1); dim3 grid_dim(ceil(mRc/float(tile_dim)),ceil(mRr/float(tile_dim)),1); matrix_multiplication<<<grid_dim,block_dim>>>(gpu_m1, gpu_m2, gpu_mR, m1r, m1c, m2c); //copy the result to Host mem err = cudaMemcpy(cpu_mR,gpu_mR,mRs, cudaMemcpyDeviceToHost); if(err != cudaSuccess){printf("Error Coping Matrix R\n");exit(1);} print_matrix(cpu_mR, mRr, mRc); } int row_counter(FILE* fp){ char c; int count = 1; for (c = getc(fp); c != EOF; c = getc(fp)) if (c == '\n') count = count + 1; rewind(fp); return count; } int col_counter(FILE* fp){ char c; int count = 1; for (c = getc(fp); c != '\n'; c = getc(fp)) if (c == ',') count = count + 1; rewind(fp); return count; } void read_matrix(FILE* fp,int *data){ char c; int nc; int temp = 0; int pot = 0; int index = 0; for (c = getc(fp); c != EOF; c = getc(fp)){ nc = (int)c; if (c == ',' || c == '\n') { //save & restart data[index] = (int)temp; temp = 0; pot = 0; index++; }else if(nc >= 48 && nc <= 57){ //add other decimal temp = (temp * (int)pow(10,pot)) + (nc - 48); if(pot == 0){ pot = 1; } } } rewind(fp); return; } void print_matrix(int *data,int mRr, int mRc){ int row, col; for (row=0; row<mRr; row++) { for(col=0; col<mRc; col++) { printf("%d", data[row * mRc + col]); if(col != mRc - 1){ printf(","); } } if(row != mRr - 1){ //if isn't the last line print return printf("\n"); } } return; }
19,092
#include <cuda_runtime.h> #include <stdio.h> __global__ void add_one(int n, float* x) { int i = threadIdx.x; if (i < n) { x[i] = x[i] + 1; printf("thread %d, value=%f\n", i, x[i]); } } void initialize_input(float* h_A, int n) { for (int i = 0; i < n; i++) { h_A[i] = i; } } __constant__ float constData[256]; void test1(int n, float* x) { float data[256]; cudaMemcpyToSymbol(constData, data, sizeof(data)); cudaMemcpyFromSymbol(data, constData, sizeof(data)); } __device__ float devData; void test2(int n, float* x) { float value = 3.14f; cudaMemcpyToSymbol(devData, &value, sizeof(float)); } __device__ float* devPointer; void test3(int n, float* x) { float* ptr; cudaMalloc(&ptr, 256 * sizeof(float)); cudaMemcpyToSymbol(devPointer, &ptr, sizeof(ptr)); } int main(void) { int N = 16; size_t size = N * sizeof(float); // Allocate input vectors h_A in host memory float* h_A = (float*)malloc(size); initialize_input(h_A, N); // Allocate vectors in device memory float* d_A; cudaMalloc(&d_A, size); // Copy vectors from host memory to device memory cudaMemcpy(d_A, h_A, size, cudaMemcpyHostToDevice); add_one<<<1, N>>>(N, d_A); // Copy result from device memory to host memory cudaMemcpy(h_A, d_A, size, cudaMemcpyDeviceToHost); // Free device memory cudaFree(d_A); printf("result: %f,%f\n", h_A[0], h_A[1]); // Free host memory free(h_A); }
19,093
#include "includes.h" __global__ void inverse_kernel(double* d_y, double* d_x) { double x = *d_x; *d_y = 1. / x; }
19,094
#include <sys/time.h> #include <stdio.h> #include <stdlib.h> #include <math.h> #include <curand_kernel.h> #include <fstream> #include <iostream> __device__ __host__ unsigned int bitreverse(unsigned int number) { number = ((0xffff0000 & number) >> 16) | ((0x0000ffff & number) << 16); number = ((0xff00ff00 & number) >> 8) | ((0x00ff00ff & number) << 8); number = ((0xf0f0f0f0 & number) >> 4) | ((0x0f0f0f0f & number) << 4); number = ((0xcccccccc & number) >> 2) | ((0x33333333 & number) << 2); number = ((0xaaaaaaaa & number) >> 1) | ((0x55555555 & number) << 1); return number; } __device__ __host__ int fastexp(int b, int e, int p) { int r = 1; while (e > 0) { if (e & 0x1) r = ((long)r * b) % p; e >>= 1; b = ((long)b * b) % p; } return r; } __device__ __host__ unsigned int getIndex(int m, int j) { unsigned int m1 = (m >> j) << (j + 1); unsigned int m2 = ((1 << j) - 1) & m; return (m1 | m2); } __device__ __host__ unsigned int getShift(unsigned int index, int j, int n) { unsigned int r = index >> (j + 1); r = bitreverse(r); r >>= 32 - ((n >> 1) - j); r <<= j; return r; } __global__ void rand_device_api_kernel(curandState *states, int *out, int n, int p) { int gtid = blockIdx.x * blockDim.x + threadIdx.x; if (gtid < n) { curandState *state = states + gtid; curand_init(1234UL, gtid, 0, state); double rand = curand_uniform_double(state); out[gtid] = (int)floor(p*rand); out[gtid + n] = 0; } } __global__ void montgomeryKernel(int *a, int *b, long *c, int n, int p, int thres) { long temp, temp0; int tid = threadIdx.x; int inc = thres; long *pC = c + (blockIdx.x * n); //int index; for(int i = tid; i < n; i += blockDim.x) pC[i] = 0; __syncthreads(); for (int i = 0; i < n; i++) { for(int j = tid; j < n; j += blockDim.x) { if(j == 0) temp0 = pC[j] + ((long)a[j] * b[i]); else temp = pC[j] + ((long)a[j] * b[i]); if (i == thres) { temp0 %= p; temp %= p; } if(j != 0) pC[j-1] = temp; __syncthreads(); } if (tid == 0) pC[n-1] = -temp0; if(i == thres) thres += inc; __syncthreads(); } for(int i = tid; i < n; i += blockDim.x) pC[i] = ((-pC[i] % p) + p) % p; } //__global__ void nttKernel(int *a, int *b, int *c, int n, int *wn, int *wni, int p, int ni) //{ // extern __shared__ int local[]; // // int *locala = local; // int *localb = local + n; // // int tid = threadIdx.x; // int lgn = (int)roundf(log2f((float)n)); // // // Copy elements in reverse bit order to local memory // locala[2*tid] = a[bitreverse(2*tid) >> (32 - lgn)]; // locala[2*tid+1] = a[bitreverse(2*tid+1) >> (32 - lgn)]; // // localb[2*tid] = b[bitreverse(2*tid) >> (32 - lgn)]; // localb[2*tid+1] = b[bitreverse(2*tid+1) >> (32 - lgn)]; // // __syncthreads(); // // int quotient = tid; // quotient = tid / stride // // // Forward NTT // for(int s = 1; s <= lgn; s++) // { // // Set up what we need for this iteration // int m = 1 << s; // int stride = m >> 1; // // int factor = tid - (quotient << (s-1)); // int t_tid = (quotient << s) + factor; // // int w = wn[(lgn - s + 1) * (n >> 1) + factor]; // // Butterfly operation // long t = (long)w * locala[t_tid + stride]; // int u = locala[t_tid]; // locala[t_tid] = (u + t) % p; // locala[t_tid + stride] = (u - t) % p; // // t = (long)w * localb[t_tid + stride]; // u = localb[t_tid]; // localb[t_tid] = (u + t) % p; // localb[t_tid + stride] = (u - t) % p; // // quotient >>= 1; // // __syncthreads(); // } // // localb[2*tid] = ((long)localb[2*tid] * locala[2*tid]) % p; // localb[2*tid+1] = ((long)localb[2*tid+1] * locala[2*tid+1]) % p; // // __syncthreads(); // // locala[2*tid] = localb[bitreverse(2*tid) >> (32 - lgn)]; // locala[2*tid+1] = localb[bitreverse(2*tid+1) >> (32 - lgn)]; // // __syncthreads(); // // quotient = tid; // // // Backward NTT // for(int s = 1; s <= lgn; s++) // { // int m = 1 << s; // int stride = m >> 1; // int factor = tid - (quotient << (s-1)); // int t_tid = (quotient << s) + factor; // int w = wni[(lgn - s + 1) * (n >> 1) + factor]; // // long t = (long)w * locala[t_tid + stride]; // int u = locala[t_tid]; // locala[t_tid] = (u + t) % p; // locala[t_tid + stride] = (u - t) % p; // // quotient >>= 1; // // __syncthreads(); // } // // // write results to buffer // if(blockIdx.x == 0) // { //// c[2*tid] = locala[2*tid]; //// c[2*tid+1] = locala[2*tid+1]; // locala[2*tid] = (locala[2*tid] * ni) % p; // locala[2*tid+1] = (locala[2*tid+1] * ni) % p; // // __syncthreads(); // // c[tid] = (((locala[tid] - locala[tid + (n/2)]) % p) + p) % p; // } //} __global__ void nttKernel(int *a, int *b, int *c, int n, int *wn, int *wni, int p, int ni) { int tid = threadIdx.x; int bid = blockIdx.x; int lgn = (int)roundf(log2f((float)n)); int *pC1 = c + (bid * n * 2); int *pC2 = pC1 + n; // Copy elements in reverse bit order to scratchpad memory for (int i = tid; i < n/2; i += blockDim.x) { pC1[2*i] = b[bitreverse(2*i) >> (32 - lgn)]; pC1[2*i+1] = b[bitreverse(2*i+1) >> (32 - lgn)]; } __syncthreads(); for (int i = tid; i < n/2; i += blockDim.x) { pC2[2*i] = a[bitreverse(2*i) >> (32 - lgn)]; pC2[2*i+1] = a[bitreverse(2*i+1) >> (32 - lgn)]; } int *locala = pC2; int *localb = pC1; __syncthreads(); // Forward NTT for(int s = 1; s <= lgn; s++) { // Set up what we need for this iteration for (int i = tid; i < n/2; i += blockDim.x) { int m = 1 << s; int stride = m >> 1; int quotient = i / stride; int factor = i - (quotient << (s-1)); int t_tid = (quotient << s) + factor; int w = wn[(lgn - s + 1) * (n >> 1) + factor]; // Butterfly operation long t = ((long)w * locala[t_tid + stride]) % p; long u = locala[t_tid]; locala[t_tid] = (u + t) % p; locala[t_tid + stride] = (((u - t) % p) + p) % p; t = ((long)w * localb[t_tid + stride]) % p; u = localb[t_tid]; localb[t_tid] = (u + t) % p; localb[t_tid + stride] = (((u - t) % p) + p) % p; } __syncthreads(); } for(int i=tid; i < n/2; i+=blockDim.x) { localb[2*i] = ((long)localb[2*i] * locala[2*i]) % p; localb[2*i+1] = ((long)localb[2*i+1] * locala[2*i+1]) % p; } __syncthreads(); for(int i=tid; i < n/2; i+=blockDim.x) { locala[2*i] = localb[bitreverse(2*i) >> (32 - lgn)]; locala[2*i+1] = localb[bitreverse(2*i+1) >> (32 - lgn)]; } __syncthreads(); // Backward NTT for(int s = 1; s <= lgn; s++) { for(int i=tid; i < n/2; i+=blockDim.x) { int m = 1 << s; int stride = m >> 1; int quotient = i / stride; int factor = i - (quotient << (s-1)); int t_tid = (quotient << s) + factor; int w = wni[(lgn - s + 1) * (n >> 1) + factor]; long t = ((long)w * locala[t_tid + stride]) % p; long u = locala[t_tid]; locala[t_tid] = (u + t) % p; locala[t_tid + stride] = (((u - t) % p) + p) % p; } __syncthreads(); } for(int i=tid; i < n/2; i+= blockDim.x) { locala[2*i] = ((long)locala[2*i] * ni) % p; locala[2*i+1] = ((long)locala[2*i+1] * ni) % p; } __syncthreads(); for(int i=tid; i < n/2; i+=blockDim.x) { pC1[i] = (((locala[i] - locala[i + (n/2)]) % p) + p) % p; } } //__global__ void nussKernel(int *rx, int *ry, int *ns, int lgn, int *wx, int *wy, int scratchSize, int *z) //{ // int bid = blockIdx.x; // int tid = threadIdx.x; // // // each block has its reserved scratchpad space // int *brx = bid * scratchSize + rx; // int *bry = bid * scratchSize + ry; // int *bwx = bid * scratchSize + wx; // int *bwy = bid * scratchSize + wy; // // int numTrans = 1; // int numTransPerPass; // int numPasses; // // for (int l = 0; l < lgn; l++) // { // int n = ns[l]; // int TransSize = 1 << n; // int m = 1 << (n >> 1); // int r = (n & 0x1) ? m << 1 : m; // // numTransPerPass = blockDim.x / TransSize; // numPasses = numTrans / numTransPerPass; // // int ptid = tid; // thread id during the different passes // for (int k = 0; k < numPasses; k++) // { // int lbid = ptid / TransSize; // int ltid = ptid - (lbid * TransSize); // // int *lrx = lbid * TransSize + brx; // int *lry = lbid * TransSize + bry; // // int *lwx = lbid * (m << 1) * r + bwx; // int *lwy = lbid * (m << 1) * r + bwy; // // int j = ltid / m; // int i = ltid - (j * m); // // // Initialize // lwx[i * r + j] = lwx[(i + m)*r + j] = lrx[ltid]; // lwy[i * r + j] = lwy[(i + m)*r + j] = lry[ltid]; // // int lm = ltid / r; // int lr = ltid - (lm * r); // // // Transform // for (int j = (n >> 1) - 1; j >= 0; j--) // { // int index = getIndex(lm, j); // int shift = getShift(index, j, n); // shift *= (r == m) ? 1 : 2; // int sindex = (lr >= shift) ? (lr - shift) : (lr - shift + r); // // int tx = lwx[index * r + lr]; // int ux = (lr >= shift) ? lwx[(index + (1 << j)) * r + sindex] : -lwx[(index + (1 << j)) * r + sindex]; // // int ty = lwy[index * r + lr]; // int uy = (lr >= shift) ? lwy[(index + (1 << j)) * r + sindex] : -lwy[(index + (1 << j)) * r + sindex]; // // __syncthreads(); // // lwx[index * r + lr] = tx + ux; // lwx[(index + (1 << j)) * r + lr] = tx - ux; // // lwy[index * r + lr] = ty + uy; // lwy[(index + (1 << j)) * r + lr] = ty - uy; // // __syncthreads(); // } // // ptid += blockDim.x; // } // End of passes // // // update the state // numTrans *= (m << 1); // // int *temp = brx; brx = bwx; bwx = temp; // temp = bry; bry = bwy; bwy = temp; // // __syncthreads(); // } // End of this stage // // // Do all the convolutions of size 2 // for (int i = tid; i < numTrans; i += blockDim.x) // { // // Determine on which data to operate // int *lrx = (i << 1) + brx; // int *lry = (i << 1) + bry; // int *lwx = (i << 1) + bwx; // // // Operate // int x0, x1, y0, y1; // x0 = lrx[0]; x1 = lrx[1]; y0 = lry[0]; y1 = lry[1]; // int t = x0 * (y0 + y1); // lwx[0] = t - (x0 + x1) * y1; // lwx[1] = t + (x1 - x0) * y0; // } // // // Set read and write segments for the last phase // int *brz = bwx; // int *bwz = bwy; // // // Start last phase // for (int l = lgn-1; l >= 0; l--) // { // int n = ns[l]; // int m = 1 << (n >> 1); // int r = (n & 0x1) ? m << 1 : m; // int blocking = r * m; // how many threads per reconstruction // int grouping = m << 1; // how many polys needed for reconstruction // // numTransPerPass = blockDim.x / blocking; // numPasses = (numTrans / grouping) / numTransPerPass; // // int ptid = tid; // for (int k = 0; k < numPasses; k++) // { // int lbid = ptid / blocking; // int ltid = ptid - (lbid * blocking); // //where to read from // int *lrz = grouping * r * lbid + brz; // // //where to write to // int *lwz = lbid * blocking + bwz; // offset is current lbid * size of trans at next level // // int lm = ltid / r; // int lr = ltid - (lm * r); // // // Untransform // for (int j = 0; j <= (n >> 1); j++) // { // int index = getIndex(lm, j); // int shift = getShift(index, j, n); // shift *= (r == m) ? 1 : 2; // //int sindex = (lr < (r - shift)) ? (lr + shift) : (lr + shift - r); // int sindex; // int tp, up; // // int t = lrz[index * r + lr]; // int u = lrz[(index + (1 << j)) * r + lr]; // // if (lr < (r - shift)) // { // sindex = lr + shift; // tp = lrz[index * r + sindex]; // up = lrz[(index + (1 << j)) * r + sindex]; // } // else // { // sindex = lr + shift - r; // tp = -lrz[index * r + sindex]; // up = -lrz[(index + (1 << j)) * r + sindex]; // } // // __syncthreads(); // // lrz[index * r + lr] = (t + u) / 2; // lrz[(index + (1 << j)) * r + lr] = (tp - up) / 2; // // __syncthreads(); // } // // // Repack // int j = ltid / m; // int i = ltid - (j * m); // // lwz[ltid] = (j == 0) ? (lrz[i * r] - lrz[(m + i) * r + (r - 1)]) : (lrz[i * r + j] + lrz[(m + i) * r + (j - 1)]); // // ptid += blockDim.x; // } // // int *temp = brz; brz = bwz; bwz = temp; // numTrans /= grouping; // // __syncthreads(); // } // // z[tid] = brz[tid]; //} __global__ void nussbaumerKernel(int *rx, int *ry, int *ns, int lgn, int *wx, int *wy, int scratchSize, int *z, int p, long inv2) { int bid = blockIdx.x; int tid = threadIdx.x; // each block has its reserved scratchpad space int *brx = bid * scratchSize + rx; int *bry = bid * scratchSize + ry; int *bwx = bid * scratchSize + wx; int *bwy = bid * scratchSize + wy; int *bz = bid * (1 << ns[0]) + z; int numTrans = 1; int numTransPerPass; int numPasses; for (int l = 0; l < lgn; l++) { int n = ns[l]; int TransSize = 1 << n; int m = 1 << (n >> 1); int r = (n & 0x1) ? m << 1 : m; numTransPerPass = blockDim.x / r; numPasses = numTrans / numTransPerPass; int ptid = tid; // thread id during the different passes for (int k = 0; k < numPasses; k++) { int lbid = ptid / r; int ltid = ptid - (lbid * r); int *lrx = lbid * TransSize + brx; int *lry = lbid * TransSize + bry; int *lwx = lbid * (m << 1) * r + bwx; int *lwy = lbid * (m << 1) * r + bwy; for (int o = ltid; o < (1 << n); o += r) { int j = o / m; int i = o - (j * m); // Initialize lwx[i * r + j] = lwx[(i + m)*r + j] = lrx[o]; lwy[i * r + j] = lwy[(i + m)*r + j] = lry[o]; } // Transform for (int j = (n >> 1) - 1; j >= 0; j--) { for (int lm = 0; lm < m; lm++) { int index = getIndex(lm, j); int shift = getShift(index, j, n); shift *= (r == m) ? 1 : 2; int sindex = (ltid >= shift) ? (ltid - shift) : (ltid - shift + r); int tx = lwx[index * r + ltid]; int ux = (ltid >= shift) ? lwx[(index + (1 << j)) * r + sindex] : -lwx[(index + (1 << j)) * r + sindex]; int ty = lwy[index * r + ltid]; int uy = (ltid >= shift) ? lwy[(index + (1 << j)) * r + sindex] : -lwy[(index + (1 << j)) * r + sindex]; __syncthreads(); lwx[index * r + ltid] = ((tx + ux) % p); lwx[(index + (1 << j)) * r + ltid] = (((tx - ux) % p) + p) % p; lwy[index * r + ltid] = ((ty + uy) % p); lwy[(index + (1 << j)) * r + ltid] = (((ty - uy) % p) + p) % p; __syncthreads(); } } ptid += blockDim.x; } // End of passes // update the state numTrans *= (m << 1); //if(tid == 0) printf("%d\n", numTrans); int *temp = brx; brx = bwx; bwx = temp; temp = bry; bry = bwy; bwy = temp; __syncthreads(); } // End of this stage // Determine how many passes to operate on the whole array for (int i = tid; i < numTrans; i += blockDim.x) { // Determine on which data to operate int *lrx = (i << 1) + brx; int *lry = (i << 1) + bry; int *lwx = (i << 1) + bwx; //int *lwy = i * TransSize + bwy; // Operate int x0, x1, y0, y1; x0 = lrx[0]; x1 = lrx[1]; y0 = lry[0]; y1 = lry[1]; long t = (long)x0 * (y0 + y1); lwx[0] = (((t - (x0 + x1) * (long)y1) % p) + p) % p; lwx[1] = ((t + (x1 - x0) * (long)y0) % p); //ptid += blockDim.x; } int *brz = bwx; int *bwz = bwy; // Start last phase for (int l = lgn - 1; l >= 0; l--) { int n = ns[l]; int m = 1 << (n >> 1); int r = (n & 0x1) ? m << 1 : m; int blocking = r; // how many threads per reconstruction int grouping = m << 1; // how many polys needed for reconstruction numTransPerPass = blockDim.x / blocking; numPasses = (numTrans / grouping) / numTransPerPass; int ptid = tid; for (int k = 0; k < numPasses; k++) { int lbid = ptid / blocking; int ltid = ptid - (lbid * blocking); //where to read from int *lrz = grouping * r * lbid + brz; //where to write to int *lwz = lbid * (r * m) + bwz; // offset is current lbid * size of trans at next level // Untransform for (int j = 0; j <= (n >> 1); j++) { for (int lm = 0; lm < m; lm++) { int index = getIndex(lm, j); int shift = getShift(index, j, n); shift *= (r == m) ? 1 : 2; int sindex; int tp, up; int t = lrz[index * r + ltid]; int u = lrz[(index + (1 << j)) * r + ltid]; if (ltid < (r - shift)) { sindex = ltid + shift; tp = lrz[index * r + sindex]; up = lrz[(index + (1 << j)) * r + sindex]; } else { sindex = ltid + shift - r; tp = -lrz[index * r + sindex]; up = -lrz[(index + (1 << j)) * r + sindex]; } __syncthreads(); lrz[index * r + ltid] = (((t + u) * (long)inv2) % p); lrz[(index + (1 << j)) * r + ltid] = ((((tp - up) * (long)inv2) % p) + p) % p; __syncthreads(); } } // Repack for (int o = ltid; o < (1 << n); o += r) { int j = o / m; int i = o - (j * m); lwz[o] = (j == 0) ? (((lrz[i * r] - lrz[(m + i) * r + (r - 1)]) % p) + p) % p : ((lrz[i * r + j] + lrz[(m + i) * r + (j - 1)]) % p); } ptid += blockDim.x; } int *temp = brz; brz = bwz; bwz = temp; numTrans /= grouping; __syncthreads(); } for (int o = tid; o < (1 << ns[0]); o += blockDim.x) { bz[o] = (brz[o] % p); } } void generatePolynomialOnDevice(int n, int p, int **d_poly) { static curandState *states = NULL; int *d_out; cudaMalloc((void**)&d_out, sizeof(int)*n*2); cudaMalloc((void**)&states, sizeof(curandState)*n); dim3 block = 64; dim3 grid = (n + block.x - 1) / block.x; rand_device_api_kernel<<<grid, block>>>(states, d_out, n, p); cudaDeviceSynchronize(); *d_poly = d_out; } void powsOf2(int wn, int *pows, int lgn, int n, int p) { for(int i = 1; i <= lgn; i++) { for(int j = 0; j < n; j++) { pows[i * n + j] = fastexp(fastexp(wn, (1 << (i-1)), p), j , p); } } // for (int i = 1; i <= lgn; i++) // { // for(int j = 0; j < n; j++) // { // printf("[%d, %d %d] ", pows[i * n + j], i, j); // } // printf("\n"); // } } void writeToFile(int *a, int n) { std::ofstream out; out.open("/home/othmane/Dokumente/poly.dat"); for(int i = 0; i < n; i++) out << a[i] << std::endl; out.close(); } void moveToHost(int *d_Buf, int **Buf, int n) { *Buf = (int*)malloc(n*sizeof(int)); cudaMemcpy(*Buf, d_Buf, n*sizeof(int), cudaMemcpyDeviceToHost); } double cpuSecond() { struct timeval tp; gettimeofday(&tp, NULL); return ((double)tp.tv_sec + (double)tp.tv_usec*1.e-6); } void multiplyOnCPU(int *A, int *B, int n, int* rC, int p) { int *C = (int*)calloc(2 * n, sizeof(int)); for (int i = 0; i<n; ++i) for (int j = 0; j < n; ++j) { C[j + i] = (C[j + i] + ((long)B[i] * A[j])) % p; } for (int i = 0; i < n; ++i) C[i] = ((((long)C[i] - C[i + n]) % p) + p) % p; for (int i = 0; i < n; ++i) rC[i] = C[i]; free(C); } void montgomeryMultiply(int *d_a, int *d_b, int *c, int n, int p, int grid) { long *d_c, *h_c; h_c = (long*)malloc(n * sizeof(long)); cudaMalloc((void**)&d_c, n * grid * sizeof(long)); // calculate threshold long square = (long)p * p; long thres = (((1UL << 63) - 1) / square); dim3 block = (n > 1024) ? 1024 : n; double iStart = cpuSecond(); montgomeryKernel<<<grid, block>>>(d_a, d_b, d_c, n, p, (int)thres); cudaDeviceSynchronize(); double iElaps = cpuSecond() - iStart; std::cout << "Time elapsed Montgomery: " << iElaps << std::endl; cudaMemcpy(h_c, d_c, n * sizeof(long), cudaMemcpyDeviceToHost); for(int i = 0; i < n; i++) c[i] = (int)h_c[i]; free(h_c); cudaFree(d_c); } void nttMultiply(int *d_a, int *d_b, int *c, int n, int ni, int wn, int wni, int p, int grid) { int *d_c, *d_wn, *d_wni, *wns, *wnis; int lgn = (int)round(log2((double)n)); wns = (int*)malloc((lgn+1) * (n/2) * sizeof(int)); wnis = (int*)malloc((lgn+1) * (n/2) * sizeof(int)); powsOf2(wn, wns, lgn, n/2, p); powsOf2(wni, wnis, lgn, n/2, p); cudaMalloc((void**)&d_wn, (lgn+1) * (n/2) * sizeof(int)); cudaMalloc((void**)&d_wni, (lgn+1) * (n/2) * sizeof(int)); cudaMalloc((void**)&d_c, n * 2 * sizeof(int)); cudaMemcpy(d_wn, wns, (lgn+1) * (n/2) * sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(d_wni, wnis, (lgn+1) * (n/2) * sizeof(int), cudaMemcpyHostToDevice); dim3 block = ((n / 2) > 1024) ? 1024 : n / 2; double iStart = cpuSecond(); nttKernel<<<1, block>>>(d_a, d_b, d_c, n, d_wn, d_wni, p, ni); cudaDeviceSynchronize(); double iElaps = cpuSecond() - iStart; std::cout << "Time elapsed NTT: " << iElaps << std::endl; cudaMemcpy(c, d_c, n * sizeof(int), cudaMemcpyDeviceToHost); // for(int i=0; i<256; i++) // c[i] = (((c[i] * 256) % 257) + 257) % 257; // // for(int i=0; i<128; i++) // c[i] = (((c[i] - c[i + 128]) % 257) + 257) % 257; cudaFree(d_c); cudaFree(d_wn); cudaFree(d_wni); free(wns); free(wnis); } void nussbaumerMultiply(int *d_a, int *d_b, int *c, int n, int p, int grid, long inv2) { n = (int)round(log2((float)n)); int lgn = (int)ceil(log2((float)n)); int *ns = (int*)malloc(lgn * sizeof(int)); int size = 1 << n; int scalableSize = (n & 0x1) ? (1 << ((n >> 1) + 1)) : (1 << (n >> 1)); int scratchSize = size * (1 << lgn); ns[0] = n; for (int i = 1; i < lgn; i++) ns[i] = (ns[i - 1] & 0x1) ? (ns[i-1] >> 1) + 1 : (ns[i-1] >> 1); int *d_rx, *d_ry, *d_wx, *d_wy, *d_ns, *d_z, *z; cudaMalloc((void**)&d_rx, scratchSize * grid * sizeof(int)); cudaMalloc((void**)&d_ry, scratchSize * grid * sizeof(int)); cudaMalloc((void**)&d_wx, scratchSize * grid * sizeof(int)); cudaMalloc((void**)&d_wy, scratchSize * grid * sizeof(int)); cudaMalloc((void**)&d_ns, lgn * sizeof(int)); cudaMalloc((void**)&d_z, size * grid * sizeof(int)); z = (int*)malloc(size * sizeof(int)); cudaMemcpy(d_rx, d_a, size * sizeof(int), cudaMemcpyDeviceToDevice); cudaMemcpy(d_ry, d_b, size * sizeof(int), cudaMemcpyDeviceToDevice); cudaMemcpy(d_ns, ns, lgn * sizeof(int), cudaMemcpyHostToDevice); double iStart = cpuSecond(); nussbaumerKernel<<<1, scalableSize>>>(d_rx, d_ry, d_ns, lgn, d_wx, d_wy, scratchSize, d_z, p, inv2); cudaDeviceSynchronize(); double iElaps = cpuSecond() - iStart; std::cout << "Time elapsed Nussbaumer: " << iElaps << std::endl; cudaMemcpy(z, d_z , size * sizeof(int), cudaMemcpyDeviceToHost); for (int i = 0; i < size; i++) c[i] = z[i]; free(z); free(ns); cudaFree(d_rx); cudaFree(d_ry); cudaFree(d_wx); cudaFree(d_wy); cudaFree(d_z); } void initialize(int **d_a, int **d_b, int **c, int **c_ref, int n, int p) { generatePolynomialOnDevice(n, p, d_a); generatePolynomialOnDevice(n, p, d_b); *c = (int*)malloc(n * 2 * sizeof(int)); *c_ref = (int*)malloc(n * sizeof(int)); } int main() { // int n = 128; // int p = 257; // // int ni = 256; // int wn = 3; // int wni = 86; // int n = 16; // int p = 786433; // // int ni = 761857; // int wn = 41596; // int wni = 116823; // int n = 256; // int p = 7681; // // int ni = 7666; // int wn = 62; // int wni = 1115; // // long inv2 = 3841; /* params for n = 1024 */ int n = 1024; int p = 786433; int ni = 786049; int wn = 19; int wni = 579477; long inv2 = 393217; /* params for n = 2048*/ // int n = 2048; // int p = 786433; // int ni = 786241; // int wn = 14; // int wni = 280869; // long inv2 = 393217; /* params for n = 4096 */ // int n = 4096; // int p = 786433; // int ni = 786337; // int wn = 804; // int wni = 292467; // long inv2 = 393217; int grid = 1; int *d_a, *d_b, *a, *b, *c, *c_ref; initialize(&d_a, &d_b, &c, &c_ref, n, p); moveToHost(d_a, &a, n); moveToHost(d_b, &b, n); montgomeryMultiply(d_a, d_b, c, n, p, grid); nussbaumerMultiply(d_a, d_b, c, n, p, grid, inv2); nttMultiply(d_a, d_b, c, 2*n, ni, wn, wni, p, grid); multiplyOnCPU(a, b, n, c_ref, p); for(int i = 0; i < n; i++) if(c[i] != c_ref[i]) printf("Error at %d, %d %d\n", i, c_ref[i], c[i]); free(a); free(b); free(c); free(c_ref); cudaFree(d_a); cudaFree(d_b); return 0; }
19,095
__global__ void hadamardProductKernel(float *a, float *b, int l) { int index = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x; for (int i = index; i < l; i += stride) a[i] = a[i] * b[i]; } extern "C" void hadamard_wrapper(float *a, float *b, int l) { int blockSize = 256; int numBlocks = (l + blockSize - 1) / blockSize; hadamardProductKernel<<<numBlocks, blockSize>>>(a, b, l); }
19,096
#include <stdio.h> #include <iostream> #include <sys/time.h> #define CHECK(call) \ { \ const cudaError_t error = call; \ if(error!=cudaSuccess) { \ printf("Error: %s:%d, ", __FILE__, __LINE__); \ printf("code: %d, reason: %s\n", error, cudaGetErrorString(error)); \ } \ } \ double cpuSecond() { struct timeval tp; gettimeofday(&tp,NULL); return ((double)tp.tv_sec + (double)tp.tv_usec*1.e-6); } void checkResult(float *hostRef, float *gpuRef, const int N) { double epsilon = 1.0E-8; bool match = true; for(int i=0; i<N; i++){ if (abs(hostRef[i]-gpuRef[i])>epsilon){ match = false; std::cout << "Arrays do not match!" << std::endl; printf("host %5.2f gpu %5.2f at current %d\n", hostRef[i], gpuRef[i],i); break; } } if (match) std::cout << "Arrays match. " << std::endl; } void initialData(float *ip, int size) { // generate different seed for random number time_t t; srand((unsigned int) time(&t)); for(int i=0; i<size; i++) { ip[i] = (float)(rand()&0xFF)/10.f; } } void sumMatrixOnHost(float *A, float *B, float *C, const int nx, const int ny) { float *ia = A; float *ib = B; float *ic = C; for(int iy=0; iy<ny; iy++) { for(int ix=0; ix<nx; ix++) { ic[ix]=ia[ix]+ib[ix]; } ia += nx; ib += nx; ic += nx; } } __global__ void sumMatrixOnGPU2D(float *MatA, float *MatB, float *MatC, int nx, int ny) { unsigned int ix = threadIdx.x+blockIdx.x*blockDim.x; unsigned int iy = threadIdx.y+blockIdx.y*blockDim.y; unsigned int idx = iy*nx+ix; if (ix<nx && iy<ny) { MatC[idx] = MatA[idx] + MatB[idx]; } } int main (int argc, char **argv) { std::cout << argv[0] << " Starting..." << std::endl; // set up device int dev = 0; cudaDeviceProp deviceProp; CHECK(cudaGetDeviceProperties(&deviceProp, dev)); std::cout << "Using Device " << dev << ": " << deviceProp.name << std::endl; CHECK(cudaSetDevice(dev)); // set up data size of matrix int nx = 1<<12; int ny = 1<<12; int nxy=nx*ny; int nBytes=nxy*sizeof(float); std::cout << "Matrix size: " << nx << ", " << ny << std::endl; // malloc host memory float *h_A, *h_B, *hostRef, *gpuRef; h_A =(float *)malloc(nBytes); h_B =(float *)malloc(nBytes); hostRef=(float *)malloc(nBytes); gpuRef =(float *)malloc(nBytes); // cpuSecond() double iStart, iElaps; // initial dataset at host site iStart=cpuSecond(); initialData(h_A, nxy); initialData(h_B, nxy); iElaps=cpuSecond()-iStart; memset(hostRef, 0, nBytes); memset(gpuRef, 0, nBytes); // add vector at host side for result checks iStart=cpuSecond(); sumMatrixOnHost(h_A, h_B, hostRef, nx, ny); iElaps=cpuSecond()-iStart; // malloc device global memory float *d_A, *d_B, *d_C; cudaMalloc((void **)&d_A, nBytes); cudaMalloc((void **)&d_B, nBytes); cudaMalloc((void **)&d_C, nBytes); // transfer data from host to device cudaMemcpy(d_A, h_A, nBytes, cudaMemcpyHostToDevice); cudaMemcpy(d_B, h_B, nBytes, cudaMemcpyHostToDevice); // invoke kernel at host side int dimx=32; int dimy=32; dim3 block(dimx, dimy); dim3 grid((nx+block.x-1)/block.x, (ny+block.y-1)/block.y); iStart=cpuSecond(); sumMatrixOnGPU2D <<<grid, block>>> (d_A, d_B, d_C, nx, ny); cudaDeviceSynchronize(); iElaps=cpuSecond()-iStart; std::cout << "sumMatrixOnGPU2D <<< (" << grid.x << ", " << grid.y << "), (" << block.x << ", " << block.y << ") >>>" << "Time elapsed " << iElaps << std::endl; // copy kernel result back to host side cudaMemcpy(gpuRef, d_C, nBytes, cudaMemcpyDeviceToHost); // check device results checkResult(hostRef, gpuRef, nxy); // free device global memory cudaFree(d_A); cudaFree(d_B); cudaFree(d_C); // free host memory free(h_A); free(h_B); free(hostRef); free(gpuRef); cudaDeviceReset(); return 0; }
19,097
#include "includes.h" __global__ void CudaPermuteWeightsPVToCudnn( float *dest, float *src, int outFeatures, int ny, int nx, int inFeatures, int manyScaleX, int manyScaleY) { // Parameter dimensions are PV source dimensions int kSrc = (blockIdx.x * blockDim.x) + threadIdx.x; if (kSrc < outFeatures * manyScaleX * manyScaleY * ny * nx * inFeatures) { int kOF = kSrc / (ny * nx * inFeatures); int kY = (kSrc % (ny * nx * inFeatures)) / (nx * inFeatures); int kX = (kSrc % (nx * inFeatures)) / inFeatures; int kIF = (kSrc % inFeatures); int sOF = inFeatures * ny * nx; int sIF = ny * nx; int sY = nx; int kDest = kOF * sOF + kIF * sIF + (ny - kY - 1) * sY + (nx - kX - 1); dest[kDest] = src[kSrc]; } }
19,098
/************************************************************************************\ * * * Copyright � 2014 Advanced Micro Devices, Inc. * * Copyright (c) 2015 Mark D. Hill and David A. Wood * * All rights reserved. * * * * Redistribution and use in source and binary forms, with or without * * modification, are permitted provided that the following are met: * * * * You must reproduce the above copyright notice. * * * * Neither the name of the copyright holder nor the names of its contributors * * may be used to endorse or promote products derived from this software * * without specific, prior, written permission from at least the copyright holder. * * * * You must include the following terms in your license and/or other materials * * provided with the software. * * * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" * * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * * IMPLIED WARRANTIES OF MERCHANTABILITY, NON-INFRINGEMENT, AND FITNESS FOR A * * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER * * OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, * * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT * * OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING * * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY * * OF SUCH DAMAGE. * * * * Without limiting the foregoing, the software may implement third party * * technologies for which you must obtain licenses from parties other than AMD. * * You agree that AMD has not obtained or conveyed to you, and that you shall * * be responsible for obtaining the rights to use and/or distribute the applicable * * underlying intellectual property rights related to the third party technologies. * * These third party technologies are not licensed hereunder. * * * * If you use the software (in whole or in part), you shall adhere to all * * applicable U.S., European, and other export laws, including but not limited to * * the U.S. Export Administration Regulations ("EAR") (15 C.F.R Sections 730-774), * * and E.U. Council Regulation (EC) No 428/2009 of 5 May 2009. Further, pursuant * * to Section 740.6 of the EAR, you hereby certify that, except pursuant to a * * license granted by the United States Department of Commerce Bureau of Industry * * and Security or as otherwise permitted pursuant to a License Exception under * * the U.S. Export Administration Regulations ("EAR"), you will not (1) export, * * re-export or release to a national of a country in Country Groups D:1, E:1 or * * E:2 any restricted technology, software, or source code you receive hereunder, * * or (2) export to Country Groups D:1, E:1 or E:2 the direct product of such * * technology or software, if such foreign produced direct product is subject to * * national security controls as identified on the Commerce Control List (currently * * found in Supplement 1 to Part 774 of EAR). For the most current Country Group * * listings, or for additional information about the EAR or your obligations under * * those regulations, please refer to the U.S. Bureau of Industry and Security's * * website at http://www.bis.doc.gov/. * * * \************************************************************************************/ #include <cuda.h> #include <stdio.h> #include <stdlib.h> #include <string.h> #include <sys/time.h> #include <algorithm> #include "../graph_parser/parse.h" #include "../graph_parser/util.h" #include "kernel.cu" #ifdef GEM5_FUSION #include <stdint.h> extern "C" { void m5_work_begin(uint64_t workid, uint64_t threadid); void m5_work_end(uint64_t workid, uint64_t threadid); } #endif void print_vector(int *vector, int num); int main(int argc, char **argv) { char *tmpchar; bool directed = 1; int num_nodes; int num_edges; int file_format = 1; cudaError_t err = cudaSuccess; if (argc == 3) { tmpchar = argv[1]; // Graph inputfile file_format = atoi(argv[2]); } else { fprintf(stderr, "You did something wrong!\n"); exit(1); } // Allocate the csr structure csr_array *csr; // Parse the graph and store it into the CSR structure if (file_format == 1) { csr = parseMetis_transpose(tmpchar, &num_nodes, &num_edges, directed); } else if (file_format == 0) { csr = parseCOO_transpose(tmpchar, &num_nodes, &num_edges, directed); } else { printf("reserve for future"); exit(1); } // Allocate the cost array int *cost_array = (int *)malloc(num_nodes * sizeof(int)); if (!cost_array) fprintf(stderr, "malloc failed cost_array\n"); // Set the cost array to zero for (int i = 0; i < num_nodes; i++) { cost_array[i] = 0; } // Create device-side buffers int *row_d; int *col_d; int *data_d; int *vector_d1; int *vector_d2; int *stop_d; // Create the device-side graph structure err = cudaMalloc(&row_d, (num_nodes + 1) * sizeof(int)); if (err != cudaSuccess) { fprintf(stderr, "ERROR: cudaMalloc row_d (size:%d) => %s\n", num_nodes, cudaGetErrorString(err)); return -1; } err = cudaMalloc(&col_d, num_edges * sizeof(int)); if (err != cudaSuccess) { fprintf(stderr, "ERROR: cudaMalloc col_d (size:%d) => %s\n", num_edges, cudaGetErrorString(err)); return -1; } err = cudaMalloc(&data_d, num_edges * sizeof(int)); if (err != cudaSuccess) { fprintf(stderr, "ERROR: cudaMalloc data_d (size:%d) => %s\n", num_edges, cudaGetErrorString(err)); return -1; } // Termination variable err = cudaMalloc(&stop_d, sizeof(int)); if (err != cudaSuccess) { fprintf(stderr, "ERROR: cudaMalloc stop_d (size:%d) => %s\n", 1, cudaGetErrorString(err)); return -1; } // Create the device-side buffers for sssp err = cudaMalloc(&vector_d1, num_nodes * sizeof(int)); if (err != cudaSuccess) { fprintf(stderr, "ERROR: cudaMalloc vector_d1 (size:%d) => %s\n", num_nodes, cudaGetErrorString(err)); return -1; } err = cudaMalloc(&vector_d2, num_nodes * sizeof(int)); if (err != cudaSuccess) { fprintf(stderr, "ERROR: cudaMalloc vector_d2 (size:%d) => %s\n", num_nodes, cudaGetErrorString(err)); return -1; } double timer1 = gettime(); #ifdef GEM5_FUSION m5_work_begin(0, 0); #endif // Copy data to device side buffers err = cudaMemcpy(row_d, csr->row_array, (num_nodes + 1) * sizeof(int), cudaMemcpyHostToDevice); if (err != cudaSuccess) { fprintf(stderr, "ERROR: cudaMemcpy row_d (size:%d) => %s\n", num_nodes, cudaGetErrorString(err)); return -1; } err = cudaMemcpy(col_d, csr->col_array, num_edges * sizeof(int), cudaMemcpyHostToDevice); if (err != cudaSuccess) { fprintf(stderr, "ERROR: cudaMemcpy col_d (size:%d) => %s\n", num_nodes, cudaGetErrorString(err)); return -1; } err = cudaMemcpy(data_d, csr->data_array, num_edges * sizeof(int), cudaMemcpyHostToDevice); if (err != cudaSuccess) { fprintf(stderr, "ERROR: cudaMemcpy data_d (size:%d) => %s\n", num_nodes, cudaGetErrorString(err)); return -1; } double timer3 = gettime(); // Work dimensions int block_size = 64; int num_blocks = (num_nodes + block_size - 1) / block_size; dim3 threads(block_size, 1, 1); dim3 grid(num_blocks, 1, 1); // Source vertex 0 int sourceVertex = 0; // Launch the initialization kernel vector_init <<<grid, threads>>>(vector_d1, vector_d2, sourceVertex, num_nodes); cudaThreadSynchronize(); err = cudaGetLastError(); if (err != cudaSuccess) { fprintf(stderr, "ERROR: vector_init failed (%s)\n", cudaGetErrorString(err)); return -1; } int stop = 1; int cnt = 0; // Main computation loop for (int i = 1; i < num_nodes; i++) { // Reset the termination variable stop = 0; // Copy the termination variable to the device err = cudaMemcpy(stop_d, &stop, sizeof(int), cudaMemcpyHostToDevice); if (err != cudaSuccess) { fprintf(stderr, "ERROR: write stop_d (%s)\n", cudaGetErrorString(err)); return -1; } // Launch the assignment kernel vector_assign <<<grid, threads>>>(vector_d1, vector_d2, num_nodes); // Launch the min.+ kernel spmv_min_dot_plus_kernel <<<grid, threads>>>(num_nodes, row_d, col_d, data_d, vector_d1, vector_d2); // Launch the check kernel vector_diff <<<grid, threads>>>(vector_d1, vector_d2, stop_d, num_nodes); // Read the termination variable back err = cudaMemcpy(&stop, stop_d, sizeof(int), cudaMemcpyDeviceToHost); if (err != cudaSuccess) { fprintf(stderr, "ERROR: read stop_d (%s)\n", cudaGetErrorString(err)); return -1; } // Exit the loop if (stop == 0) { break; } cnt++; } cudaThreadSynchronize(); double timer4 = gettime(); // Read the cost_array back err = cudaMemcpy(cost_array, vector_d1, num_nodes * sizeof(int), cudaMemcpyDeviceToHost); if (err != cudaSuccess) { fprintf(stderr, "ERROR: read vector_d1 (%s)\n", cudaGetErrorString(err)); return -1; } #ifdef GEM5_FUSION m5_work_end(0, 0); #endif double timer2 = gettime(); // Print the timing statistics printf("kernel + memcpy time = %lf ms\n", (timer2 - timer1) * 1000); printf("kernel time = %lf ms\n", (timer4 - timer3) * 1000); printf("number iterations = %d\n", cnt); #if 1 // Print cost_array print_vector(cost_array, num_nodes); #endif // Clean up the host arrays free(cost_array); csr->freeArrays(); free(csr); // Clean up the device-side buffers cudaFree(row_d); cudaFree(col_d); cudaFree(data_d); cudaFree(stop_d); cudaFree(vector_d1); cudaFree(vector_d2); return 0; } void print_vector(int *vector, int num) { FILE * fp = fopen("result.out", "w"); if (!fp) { printf("ERROR: unable to open result.txt\n"); } for (int i = 0; i < num; i++) fprintf(fp, "%d: %d\n", i + 1, vector[i]); fclose(fp); }
19,099
#include <thrust/device_vector.h> #include <thrust/host_vector.h> #include <iostream> #include <chrono> int main() { std::vector<double> stocks; int n = 0; while (std::cin){ n = n + 1; double stock_day; std::cin >> stock_day; stocks.push_back(stock_day); } auto start = std::chrono::steady_clock::now(); thrust::device_vector<double> dev(stocks); auto end = std::chrono::steady_clock::now(); std::chrono::duration<double> time_seconds = end-start; double preco_medio = thrust::reduce(dev.begin(), dev.end(), 0, thrust::plus<double>()) / n; double preco_medio_ano = thrust::reduce(dev.begin() + n - 365, dev.end(), 0, thrust::plus<double>()) / 360; double maior_preco = thrust::reduce(dev.begin(), dev.end(), 0, thrust::maximum<double>());; double menor_preco = thrust::reduce(dev.begin(), dev.end(), maior_preco, thrust::minimum<double>());; double maior_ano = thrust::reduce(dev.begin() + n - 365, dev.end(), 0, thrust::maximum<double>()); double menor_ano = thrust::reduce(dev.begin() + n - 365, dev.end(), maior_ano, thrust::minimum<double>()); std::cout << "media: " << preco_medio << " media ano: " << preco_medio_ano << " maior: " << maior_preco << " menor: " << menor_preco << " maior ano: " << maior_ano << " menor ano: " << menor_ano << "\n"; }
19,100
// Copyright 2013 Google Inc. All Rights Reserved. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. // You may obtain a copy of the License at // // http://www.apache.org/licenses/LICENSE-2.0 // // Unless required by applicable law or agreed to in writing, software // distributed under the License is distributed on an "AS IS" BASIS, // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. // See the License for the specific language governing permissions and // limitations under the License. #include <stdio.h> #include <stdlib.h> #include <string.h> #include <math.h> #include <pthread.h> #include <cuda_runtime.h> #define MAX_STRING 100 #define EXP_TABLE_SIZE 1000 #define MAX_EXP 6 #define MAX_SENTENCE_LENGTH 1000 #define MAX_CODE_LENGTH 40 // Commented version of Mikolov's original code // Cloned from https://github.com/chrisjmccormick/word2vec_commented.git /* Report cuda error */ #define checkCUDAerr(err) {\ cudaError_t ce = err;\ if (cudaSuccess != ce) {\ printf("%s %d : %s\n", __FILE__, __LINE__, cudaGetErrorString(ce));\ exit(0);\ }\ } /* * The size of the hash table for the vocabulary. * The vocabulary won't be allowed to grow beyond 70% of this number. * For instance, if the hash table has 30M entries, then the maximum * vocab size is 21M. This is to minimize the occurrence (and performance * impact) of hash collisions. */ const int vocab_hash_size = 30000000; // Maximum 30 * 0.7 = 21M words in the vocabulary typedef float real; // Precision of float numbers /** * ======== vocab_word ======== * Properties: * cn - The word frequency (number of times it appears). * word - The actual string word. */ struct vocab_word { long long cn; int *point; char *word, *code, codelen; }; //TODO Allocate cuda structure first without incuring error /* * ======== Global Variables ======== * */ char train_file[MAX_STRING], output_file[MAX_STRING]; char save_vocab_file[MAX_STRING], read_vocab_file[MAX_STRING]; /* * ======== vocab ======== * This array will hold all of the words in the vocabulary. * This is internal state. */ struct vocab_word *vocab; int binary = 0, cbow = 1, debug_mode = 2, window = 5, min_count = 5, num_threads = 12, min_reduce = 1; /* * ======== vocab_hash ======== * This array is the hash table for the vocabulary. Word strings are hashed * to a hash code (an integer), then the hash code is used as the index into * 'vocab_hash', to retrieve the index of the word within the 'vocab' array. */ int *vocab_hash; /* * ======== vocab_max_size ======== * This is not a limit on the number of words in the vocabulary, but rather * a chunk size for allocating the vocabulary table. The vocabulary table will * be expanded as necessary, and is allocated, e.g., 1,000 words at a time. * * ======== vocab_size ======== * Stores the number of unique words in the vocabulary. * This is not a parameter, but rather internal state. * * ======== layer1_size ======== * This is the number of features in the word vectors. * It is the number of neurons in the hidden layer of the model. */ long long vocab_max_size = 1000, vocab_size = 0, layer1_size = 100; /* * */ long long train_words = 0, word_count_actual = 0, iter = 5, file_size = 0, classes = 0; /* * ======== alpha ======== * TODO - This is a learning rate parameter. * * ======== starting_alpha ======== * * ======== sample ======== * This parameter controls the subsampling of frequent words. * Smaller values of 'sample' mean words are less likely to be kept. * Set 'sample' to 0 to disable subsampling. * See the comments in the subsampling section for more details. */ float alpha = 0.025, starting_alpha, sample = 1e-3; /* * IMPORTANT - Note that the weight matrices are stored as 1D arrays, not * 2D matrices, so to access row 'i' of syn0, the index is (i * layer1_size). * * ======== syn0 ======== * This is the hidden layer weights (which is also the word vectors!) * * ======== syn1 ======== * This is the output layer weights *if using heirarchical softmax* * * ======== syn1neg ======== * This is the output layer weights *if using negative sampling* * * ======== expTable ======== * Stores precalcultaed activations for the output layer. */ float *syn0, *syn1, *syn1neg, *expTable; clock_t start; int hs = 0, negative = 5; const int table_size = 1e8; int *table; // CUDA Init int *vocab_codelen, *vocab_point, *d_vocab_codelen, *d_vocab_point; int *d_table; char *vocab_code, *d_vocab_code; float *d_syn0, *d_syn1, *d_expTable; /** * ======== InitUnigramTable ======== * This table is used to implement negative sampling. * Each word is given a weight equal to it's frequency (word count) raised to * the 3/4 power. The probability for a selecting a word is just its weight * divided by the sum of weights for all words. * * Note that the vocabulary has been sorted by word count, descending, such * that we will go through the vocabulary from most frequent to least. */ void InitUnigramTable() { int a, i; double train_words_pow = 0; double d1, power = 0.75; // Allocate the table. It's bigger than the vocabulary, because words will // appear in it multiple times based on their frequency. // Every vocab word appears at least once in the table. // The size of the table relative to the size of the vocab dictates the // resolution of the sampling. A larger unigram table means the negative // samples will be selected with a probability that more closely matches the // probability calculated by the equation. table = (int *)malloc(table_size * sizeof(int)); // Calculate the denominator, which is the sum of weights for all words. for (a = 0; a < vocab_size; a++) train_words_pow += pow(vocab[a].cn, power); // 'i' is the vocabulary index of the current word, whereas 'a' will be // the index into the unigram table. i = 0; // Calculate the probability that we choose word 'i'. This is a fraction // betwee 0 and 1. d1 = pow(vocab[i].cn, power) / train_words_pow; // Loop over all positions in the table. for (a = 0; a < table_size; a++) { // Store word 'i' in this position. Word 'i' will appear multiple times // in the table, based on its frequency in the training data. table[a] = i; // If the fraction of the table we have filled is greater than the // probability of choosing this word, then move to the next word. if (a / (double)table_size > d1) { // Move to the next word. i++; // Calculate the probability for the new word, and accumulate it with // the probabilities of all previous words, so that we can compare d1 to // the percentage of the table that we have filled. d1 += pow(vocab[i].cn, power) / train_words_pow; } // Don't go past the end of the vocab. // The total weights for all words should sum up to 1, so there shouldn't // be any extra space at the end of the table. Maybe it's possible to be // off by 1, though? Or maybe this is just precautionary. if (i >= vocab_size) i = vocab_size - 1; } } /** * ======== ReadWord ======== * Reads a single word from a file, assuming space + tab + EOL to be word * boundaries. * * Parameters: * word - A char array allocated to hold the maximum length string. * fin - The training file. */ void ReadWord(char *word, FILE *fin) { // 'a' will be the index into 'word'. int a = 0, ch; // Read until the end of the word or the end of the file. while (!feof(fin)) { // Get the next character. ch = fgetc(fin); // ASCII Character 13 is a carriage return 'CR' whereas character 10 is // newline or line feed 'LF'. if (ch == 13) continue; // Check for word boundaries... if ((ch == ' ') || (ch == '\t') || (ch == '\n')) { // If the word has at least one character, we're done. if (a > 0) { // Put the newline back before returning so that we find it next time. if (ch == '\n') ungetc(ch, fin); break; } // If the word is empty and the character is newline, treat this as the // end of a "sentence" and mark it with the token </s>. if (ch == '\n') { strcpy(word, (char *)"</s>"); return; // If the word is empty and the character is tab or space, just continue // on to the next character. } else continue; } // If the character wasn't space, tab, CR, or newline, add it to the word. word[a] = ch; a++; // If the word's too long, truncate it, but keep going till we find the end // of it. if (a >= MAX_STRING - 1) a--; } // Terminate the string with null. word[a] = 0; } /** * ======== GetWordHash ======== * Returns hash value of a word. The hash is an integer between 0 and * vocab_hash_size (default is 30E6). * * For example, the word 'hat': * hash = ((((h * 257) + a) * 257) + t) % 30E6 */ int GetWordHash(char *word) { unsigned long long a, hash = 0; for (a = 0; a < strlen(word); a++) hash = hash * 257 + word[a]; hash = hash % vocab_hash_size; return hash; } /** * ======== SearchVocab ======== * Lookup the index in the 'vocab' table of the given 'word'. * Returns -1 if the word is not found. * This function uses a hash table for fast lookup. */ int SearchVocab(char *word) { // Compute the hash value for 'word'. unsigned int hash = GetWordHash(word); // Lookup the index in the hash table, handling collisions as needed. // See 'AddWordToVocab' to see how collisions are handled. while (1) { // If the word isn't in the hash table, it's not in the vocab. if (vocab_hash[hash] == -1) return -1; // If the input word matches the word stored at the index, we're good, // return the index. if (!strcmp(word, vocab[vocab_hash[hash]].word)) return vocab_hash[hash]; // Otherwise, we need to scan through the hash table until we find it. hash = (hash + 1) % vocab_hash_size; } // This will never be reached. return -1; } /** * ======== ReadWordIndex ======== * Reads the next word from the training file, and returns its index into the * 'vocab' table. */ int ReadWordIndex(FILE *fin) { char word[MAX_STRING]; ReadWord(word, fin); if (feof(fin)) return -1; return SearchVocab(word); } /** * ======== AddWordToVocab ======== * Adds a new word to the vocabulary (one that hasn't been seen yet). */ int AddWordToVocab(char *word) { // Measure word length. unsigned int hash, length = strlen(word) + 1; // Limit string length (default limit is 100 characters). if (length > MAX_STRING) length = MAX_STRING; // Allocate and store the word string. vocab[vocab_size].word = (char *)calloc(length, sizeof(char)); strcpy(vocab[vocab_size].word, word); // Initialize the word frequency to 0. vocab[vocab_size].cn = 0; // Increment the vocabulary size. vocab_size++; // Reallocate memory if needed if (vocab_size + 2 >= vocab_max_size) { vocab_max_size += 1000; vocab = (struct vocab_word *)realloc(vocab, vocab_max_size * sizeof(struct vocab_word)); } // Add the word to the 'vocab_hash' table so that we can map quickly from the // string to its vocab_word structure. // Hash the word to an integer between 0 and 30E6. hash = GetWordHash(word); // If the spot is already taken in the hash table, find the next empty spot. while (vocab_hash[hash] != -1) hash = (hash + 1) % vocab_hash_size; // Map the hash code to the index of the word in the 'vocab' array. vocab_hash[hash] = vocab_size - 1; // Return the index of the word in the 'vocab' array. return vocab_size - 1; } // Used later for sorting by word counts int VocabCompare(const void *a, const void *b) { return ((struct vocab_word *)b)->cn - ((struct vocab_word *)a)->cn; } /** * ======== SortVocab ======== * Sorts the vocabulary by frequency using word counts, and removes words that * occur fewer than 'min_count' times in the training text. * * Removing words from the vocabulary requires recomputing the hash table. */ void SortVocab() { int a, size; unsigned int hash; /* * Sort the vocabulary by number of occurrences, in descending order. * * Keep </s> at the first position by sorting starting from index 1. * * Sorting the vocabulary this way causes the words with the fewest * occurrences to be at the end of the vocabulary table. This will allow us * to free the memory associated with the words that get filtered out. */ qsort(&vocab[1], vocab_size - 1, sizeof(struct vocab_word), VocabCompare); // Clear the vocabulary hash table. for (a = 0; a < vocab_hash_size; a++) vocab_hash[a] = -1; // Store the initial vocab size to use in the for loop condition. size = vocab_size; // Recompute the number of training words. train_words = 0; // For every word currently in the vocab... for (a = 0; a < size; a++) { // If it occurs fewer than 'min_count' times, remove it from the vocabulary. if ((vocab[a].cn < min_count) && (a != 0)) { // Decrease the size of the new vocabulary. vocab_size--; // Free the memory associated with the word string. free(vocab[a].word); } else { // Hash will be re-computed, as after the sorting it is not actual hash=GetWordHash(vocab[a].word); while (vocab_hash[hash] != -1) hash = (hash + 1) % vocab_hash_size; vocab_hash[hash] = a; train_words += vocab[a].cn; } } // Reallocate the vocab array, chopping off all of the low-frequency words at // the end of the table. vocab = (struct vocab_word *)realloc(vocab, (vocab_size + 1) * sizeof(struct vocab_word)); // Allocate memory for the binary tree construction for (a = 0; a < vocab_size; a++) { vocab[a].code = (char *)calloc(MAX_CODE_LENGTH, sizeof(char)); vocab[a].point = (int *)calloc(MAX_CODE_LENGTH, sizeof(int)); } } // Reduces the vocabulary by removing infrequent tokens void ReduceVocab() { int a, b = 0; unsigned int hash; for (a = 0; a < vocab_size; a++) if (vocab[a].cn > min_reduce) { vocab[b].cn = vocab[a].cn; vocab[b].word = vocab[a].word; b++; } else free(vocab[a].word); vocab_size = b; for (a = 0; a < vocab_hash_size; a++) vocab_hash[a] = -1; for (a = 0; a < vocab_size; a++) { // Hash will be re-computed, as it is not actual hash = GetWordHash(vocab[a].word); while (vocab_hash[hash] != -1) hash = (hash + 1) % vocab_hash_size; vocab_hash[hash] = a; } fflush(stdout); min_reduce++; } // Main execution for cbow model __global__ void cbow_exec(int window, int layer1_size, int negative, int hs, int table_size, int vocab_size, float alpha, const float* expTable, const int* table, const int* vocab_codelen, const int* vocab_point, const char* vocab_code, const int* sen, const int* sentence_length, float* syn1, float* syn0) { __shared__ float f, g; // init for sentence index using blockIdx int sent_idx_s = sentence_length[blockIdx.x]; int sent_idx_e = sentence_length[blockIdx.x + 1]; unsigned long next_random = blockIdx.x; if(threadIdx.x < layer1_size) for(int sentence_position = sent_idx_s; sentence_position < sent_idx_e; sentence_position++){ int word = sen[sentence_position]; if(word == -1) continue; float neu1 = 0; float neu1e = 0; next_random = next_random * (unsigned long)2514903917 + 11; int b = next_random % window; int cw = 0; for(int a = b; a < window * 2 + 1 - b; a++) if (a != window) { int c = sentence_position - window + a; // Verify c isn't outisde the bounds of the sentence. if (c < sent_idx_s) continue; if (c >= sent_idx_e) continue; int last_word = sen[c]; if (last_word == -1) continue; // TODO Not sure here neu1 += syn0[threadIdx.x + last_word * layer1_size]; cw++; } if(cw){ // neu1 was the sum of the context word vectors, and now becomes // their average. neu1 /= cw; // HS Phase if(hs) for(int d = vocab_codelen[word]; d < vocab_codelen[word+1]; d++){ int l2 = vocab_point[d] * layer1_size; f = 0; f += neu1 * syn1[threadIdx.x + l2]; __syncthreads(); // Apply the sigmoid activation to the current output neuron. if (f <= -MAX_EXP) continue; else if (f >= MAX_EXP) continue; else f = expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))]; g = (1 - vocab_code[d] - f) * alpha; neu1e += g * syn1[threadIdx.x + l2]; // Atomic addition for threads atomicAdd(&syn1[threadIdx.x + l2], g * neu1); } if(negative > 0) for (int d = 0; d < negative + 1; d++) { int target , label; if(d == 0){ target = word; label = 1; }else{ next_random = next_random * (unsigned long long)25214903917 + 11; target = table[(next_random >> 16) % table_size]; if (target == 0) target = next_random % (vocab_size - 1) + 1; if (target == word) continue; label = 0; } int l2 = target * layer1_size; f = 0; // not sure f += neu1 * syn1[threadIdx.x + l2]; // should split with thead or not? __syncthreads(); if (f > MAX_EXP) g = (label - 1) * alpha; else if (f < -MAX_EXP) g = (label - 0) * alpha; else g = (label - expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))]) * alpha; neu1e += g * syn1[l2 + threadIdx.x]; // Atomic addition for threads atomicAdd(&syn1[threadIdx.x + l2], g * neu1); } for(int a = b; a < window * 2 + 1 - b; a++) if (a != window) { int c = sentence_position - window + a; if(c < sent_idx_s) continue; if(c >= sent_idx_e) continue; int last_word = sen[c]; if (last_word == -1) continue; // Atomic addition for threads atomicAdd(&syn0[threadIdx.x + last_word * layer1_size], g * neu1e); } } } } // Skip gram execution, not sure if should create f array for each thread or not __global__ void skipgram_exec(int window, int layer1_size, int negative, int hs, int table_size, int vocab_size, float alpha, const float* expTable, const int* table, const int* vocab_codelen, const int* vocab_point, const char* vocab_code, const int* sen, const int* sentence_length, float* syn1, float* syn0) { __shared__ float f, g; // init for sentence index using blockIdx int sent_idx_s = sentence_length[blockIdx.x]; int sent_idx_e = sentence_length[blockIdx.x + 1]; unsigned long next_random = blockIdx.x; if(threadIdx.x < layer1_size) for(int sentence_position = sent_idx_s; sentence_position < sent_idx_e; sentence_position++){ int word = sen[sentence_position]; if(word == -1) continue; float neu1e = 0; next_random = next_random * (unsigned long)2514903917 + 11; int b = next_random % window; for (int a = b; a < window * 2 + 1 - b; a++) if (a != window) { int c = sentence_position - window + a; if (c < sent_idx_s) continue; if (c >= sent_idx_e) continue; int last_word = sen[c]; if (last_word == -1) continue; // TODO Not sure here int l1 = last_word * layer1_size; neu1e = 0; //HS if(hs) for(int d = vocab_codelen[word]; d < vocab_codelen[word+1]; d++){ f = 0; int l2 = vocab_point[d] * layer1_size; f = syn0[threadIdx.x + l1] * syn1[threadIdx.x + l2]; __syncthreads(); if (f <= -MAX_EXP) continue; else if (f >= MAX_EXP) continue; else f = expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))]; neu1e += g * syn1[threadIdx.x + l2]; atomicAdd(&syn1[threadIdx.x + l2], g * syn0[threadIdx.x + l1]); } // Negative Sampling if(negative > 0) for (int d = 0; d < negative + 1; d++) { int target, label; if(d == 0){ target = word; label = 1; }else{ next_random = next_random * (unsigned long long)25214903917 + 11; target = table[(next_random >> 16) % table_size]; if (target == 0) target = next_random % (vocab_size - 1) + 1; if (target == word) continue; label = 0; } int l2 = target * layer1_size; f = 0; f = syn0[threadIdx.x +l1] * syn1[threadIdx.x + l2];//should use syn1neg? __syncthreads(); if (f > MAX_EXP) g = (label - 1) * alpha; else if (f < -MAX_EXP) g = (label - 0) * alpha; else g = (label - expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))]) * alpha; neu1e += g * syn1[threadIdx.x + l2]; atomicAdd(&syn1[threadIdx.x + l2], g * syn0[threadIdx.x + l1]); } atomicAdd(&syn0[threadIdx.x + l1], neu1e); } } } // CUDA Structure allocation void initVocabCuda(){ vocab_codelen = (int*)malloc((vocab_size+1) * sizeof(int)); vocab_codelen[0] = 0; for(int i = 1; i <= vocab_size; i++){ vocab_codelen[i] = vocab_codelen[i-1]+vocab[i-1].codelen; // global codelen struct getting vocab information for cuda sync } vocab_point = (int*)malloc(vocab_codelen[vocab_size] * sizeof(int)); vocab_code = (char*)malloc(vocab_codelen[vocab_size] * sizeof(char)); cudaMalloc((void **)&d_vocab_codelen, (vocab_size+1) * sizeof(int)); cudaMalloc((void **)&d_vocab_point, vocab_codelen[vocab_size] * sizeof(int)); cudaMalloc((void **)&d_vocab_code, vocab_codelen[vocab_size] * sizeof(char)); for(int i=0; i < vocab_size; i++){ for(int j = 0; j < vocab_size; j++){ vocab_point[vocab_codelen[i]+j] = vocab[i].point[j]; vocab_code[vocab_codelen[i]+j] = vocab[i].code[j]; } } cudaMemcpy(d_vocab_codelen, vocab_codelen, (vocab_size+1) * sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(d_vocab_point, vocab_point, vocab_codelen[vocab_size] * sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(d_vocab_code, vocab_code, vocab_codelen[vocab_size] * sizeof(char), cudaMemcpyHostToDevice); } /** * ======== CreateBinaryTree ======== * Create binary Huffman tree using the word counts. * Frequent words will have short unique binary codes. * Huffman encoding is used for lossless compression. * For each vocabulary word, the vocab_word structure includes a `point` array, * which is the list of internal tree nodes which: * 1. Define the path from the root to the leaf node for the word. * 2. Each correspond to a row of the output matrix. * The `code` array is a list of 0s and 1s which specifies whether each output * in `point` should be trained to output 0 or 1. */ void CreateBinaryTree() { long long a, b, i, min1i, min2i, pos1, pos2, point[MAX_CODE_LENGTH]; char code[MAX_CODE_LENGTH]; // Default is 40 // Note that calloc initializes these arrays to 0. long long *count = (long long *)calloc(vocab_size * 2 + 1, sizeof(long long)); long long *binary = (long long *)calloc(vocab_size * 2 + 1, sizeof(long long)); long long *parent_node = (long long *)calloc(vocab_size * 2 + 1, sizeof(long long)); // The count array is twice the size of the vocabulary, plus one. // - The first half of `count` becomes a list of the word counts // for each word in the vocabulary. We do not modify this part of the // list. // - The second half of `count` is set to a large positive integer (1 // quadrillion). When we combine two trees under a word (e.g., word_id // 13), then we place the total weight of those subtrees into the word's // position in the second half (e.g., count[vocab_size + 13]). // for (a = 0; a < vocab_size; a++) count[a] = vocab[a].cn; for (a = vocab_size; a < vocab_size * 2; a++) count[a] = 1e15; // `pos1` and `pos2` are indeces into the `count` array. // - `pos1` starts at the middle of `count` (the end of the list of word // counts) and moves left. // - `pos2` starts at the beginning of the list of large integers and moves // right. pos1 = vocab_size - 1; pos2 = vocab_size; /* =============================== * Step 1: Create Huffman Tree * =============================== * [Original Comment] Following algorithm constructs the Huffman tree by * adding one node at a time * * The Huffman coding algorithm starts with every node as its own tree, and * then combines the two smallest trees on each step. The weight of a tree is * the sum of the word counts for the words it contains. * * Once the tree is constructed, you can use the `parent_node` array to * navigate it. For the word at index 13, for example, you would look at * parent_node[13], and then parent_node[parent_node[13]], and so on, till * you reach the root. * * A Huffman tree stores all of the words in the vocabulary at the leaves. * Frequent words have short paths, and infrequent words have long paths. * Here, we are also associating each internal node of the tree with a * row of the output matrix. Every time we combine two trees and create a * new node, we give it a row in the output matrix. */ // The number of tree combinations needed is equal to the size of the vocab, // minus 1. for (a = 0; a < vocab_size - 1; a++) { // First, find two smallest nodes 'min1, min2' // Find min1 (at index `min1i`) if (pos1 >= 0) { if (count[pos1] < count[pos2]) { min1i = pos1; pos1--; } else { min1i = pos2; pos2++; } } else { min1i = pos2; pos2++; } // Find min2 (at index `min2i`). if (pos1 >= 0) { if (count[pos1] < count[pos2]) { min2i = pos1; pos1--; } else { min2i = pos2; pos2++; } } else { min2i = pos2; pos2++; } // Calculate the combined weight. We could be combining two words, a word // and a tree, or two trees. count[vocab_size + a] = count[min1i] + count[min2i]; // Store the path for working back up the tree. parent_node[min1i] = vocab_size + a; parent_node[min2i] = vocab_size + a; // binary[min1i] = 0; // This is implied. // min1 is the (left?) node and is labeled '0', min2 is the (right?) node // and is labeled '1'. binary[min2i] = 1; } /* ========================================== * Step 2: Define Samples for Each Word * ========================================== * [Original Comment] Now assign binary code to each vocabulary word * * vocab[word] * .code - A variable-length string of 0s and 1s. * .point - A variable-length array of output row indeces. * .codelen - The length of the `code` array. * The point array has length `codelen + 1`. * */ // For each word in the vocabulary... for (a = 0; a < vocab_size; a++) { b = a; i = 0; // `i` stores the code length. // Construct the binary code... // `code` stores 1s and 0s. // `point` stores indeces. // This loop works backwards from the leaf, so the `code` and `point` // lists end up in reverse order. while (1) { // Lookup whether this is on the left or right of its parent node. code[i] = binary[b]; // Note: point[0] always holds the word iteself... point[i] = b; // Increment the code length. i++; // This will always return an index in the second half of the array. b = parent_node[b]; // We've reached the root when... if (b == vocab_size * 2 - 2) break; } // Record the code length (the length of the `point` list). vocab[a].codelen = i; // The root node is at row `vocab_size - 2` of the output matrix. vocab[a].point[0] = vocab_size - 2; // For each bit in this word's code... for (b = 0; b < i; b++) { // Reverse the code in `code` and store it in `vocab[a].code` vocab[a].code[i - b - 1] = code[b]; // Store the row indeces of the internal nodes leading to this word. // These are the set of outputs which will be trained every time // this word is encountered in the training data as an output word. vocab[a].point[i - b] = point[b] - vocab_size; } } free(count); free(binary); free(parent_node); } /** * ======== LearnVocabFromTrainFile ======== * Builds a vocabulary from the words found in the training file. * * This function will also build a hash table which allows for fast lookup * from the word string to the corresponding vocab_word object. * * Words that occur fewer than 'min_count' times will be filtered out of * vocabulary. */ void LearnVocabFromTrainFile() { char word[MAX_STRING]; FILE *fin; long long a, i; // Populate the vocab table with -1s. for (a = 0; a < vocab_hash_size; a++) vocab_hash[a] = -1; // Open the training file. fin = fopen(train_file, "rb"); if (fin == NULL) { printf("ERROR: training data file not found!\n"); exit(1); } vocab_size = 0; // The special token </s> is used to mark the end of a sentence. In training, // the context window does not go beyond the ends of a sentence. // // Add </s> explicitly here so that it occurs at position 0 in the vocab. AddWordToVocab((char *)"</s>"); while (1) { // Read the next word from the file into the string 'word'. ReadWord(word, fin); // Stop when we've reached the end of the file. if (feof(fin)) break; // Count the total number of tokens in the training text. train_words++; // Print progress at every 100,000 words. if ((debug_mode > 1) && (train_words % 100000 == 0)) { printf("%lldK%c", train_words / 1000, 13); fflush(stdout); } // Look up this word in the vocab to see if we've already added it. i = SearchVocab(word); // If it's not in the vocab... if (i == -1) { // ...add it. a = AddWordToVocab(word); // Initialize the word frequency to 1. vocab[a].cn = 1; // If it's already in the vocab, just increment the word count. } else vocab[i].cn++; // If the vocabulary has grown too large, trim out the most infrequent // words. The vocabulary is considered "too large" when it's filled more // than 70% of the hash table (this is to try and keep hash collisions // down). if (vocab_size > vocab_hash_size * 0.7) ReduceVocab(); } // Sort the vocabulary in descending order by number of word occurrences. // Remove (and free the associated memory) for all the words that occur // fewer than 'min_count' times. SortVocab(); // Report the final vocabulary size, and the total number of words // (excluding those filtered from the vocabulary) in the training set. if (debug_mode > 0) { printf("Vocab size: %lld\n", vocab_size); printf("Words in train file: %lld\n", train_words); } file_size = ftell(fin); fclose(fin); } void SaveVocab() { long long i; FILE *fo = fopen(save_vocab_file, "wb"); for (i = 0; i < vocab_size; i++) fprintf(fo, "%s %lld\n", vocab[i].word, vocab[i].cn); fclose(fo); } void ReadVocab() { long long a, i = 0; char c; char word[MAX_STRING]; FILE *fin = fopen(read_vocab_file, "rb"); if (fin == NULL) { printf("Vocabulary file not found\n"); exit(1); } for (a = 0; a < vocab_hash_size; a++) vocab_hash[a] = -1; vocab_size = 0; while (1) { ReadWord(word, fin); if (feof(fin)) break; a = AddWordToVocab(word); fscanf(fin, "%lld%c", &vocab[a].cn, &c); i++; } SortVocab(); if (debug_mode > 0) { printf("Vocab size: %lld\n", vocab_size); printf("Words in train file: %lld\n", train_words); } fin = fopen(train_file, "rb"); if (fin == NULL) { printf("ERROR: training data file not found!\n"); exit(1); } fseek(fin, 0, SEEK_END); file_size = ftell(fin); fclose(fin); } /** * ======== InitNet ======== * */ void InitNet() { long long a, b; unsigned long long next_random = 1; // Allocate the hidden layer of the network, which is what becomes the word vectors. // The variable for this layer is 'syn0'. a = posix_memalign((void **)&syn0, 128, (long long)vocab_size * layer1_size * sizeof(real)); if (syn0 == NULL) {printf("Memory allocation failed\n"); exit(1);} // If we're using hierarchical softmax for training... if (hs) { a = posix_memalign((void **)&syn1, 128, (long long)vocab_size * layer1_size * sizeof(real)); if (syn1 == NULL) {printf("Memory allocation failed\n"); exit(1);} for (a = 0; a < vocab_size; a++) for (b = 0; b < layer1_size; b++) syn1[a * layer1_size + b] = 0; } // If we're using negative sampling for training... if (negative>0) { // Allocate the output layer of the network. // The variable for this layer is 'syn1neg'. // This layer has the same size as the hidden layer, but is the transpose. a = posix_memalign((void **)&syn1neg, 128, (long long)vocab_size * layer1_size * sizeof(real)); if (syn1neg == NULL) {printf("Memory allocation failed\n"); exit(1);} // Set all of the weights in the output layer to 0. for (a = 0; a < vocab_size; a++) for (b = 0; b < layer1_size; b++) syn1neg[a * layer1_size + b] = 0; } // Randomly initialize the weights for the hidden layer (word vector layer). // TODO - What's the equation here? for (a = 0; a < vocab_size; a++) for (b = 0; b < layer1_size; b++) { next_random = next_random * (unsigned long long)25214903917 + 11; syn0[a * layer1_size + b] = (((next_random & 0xFFFF) / (real)65536) - 0.5) / layer1_size; } // Create a binary tree for Huffman coding. // TODO - As best I can tell, this is only used for hierarchical softmax training... CreateBinaryTree(); } /** * ======== TrainModelThread ======== * This function performs the training of the model. */ void *TrainModelThread(void *id) { /* * word - Stores the index of a word in the vocab table. * word_count - Stores the total number of training words processed. */ long long a, b, d, cw, word, last_word, sentence_length = 0, sentence_position = 0; long long word_count = 0, last_word_count = 0, sen[MAX_SENTENCE_LENGTH + 1]; long long l1, l2, c, target, label, local_iter = iter; unsigned long long next_random = (long long)id; real f, g; clock_t now; // neu1 is only used by the CBOW architecture. real *neu1 = (real *)calloc(layer1_size, sizeof(real)); // neu1e is used by both architectures. real *neu1e = (real *)calloc(layer1_size, sizeof(real)); // Open the training file and seek to the portion of the file that this // thread is responsible for. FILE *fi = fopen(train_file, "rb"); fseek(fi, file_size / (long long)num_threads * (long long)id, SEEK_SET); // This loop covers the whole training operation... while (1) { /* * ======== Variables ======== * iter - This is the number of training epochs to run; default is 5. * word_count - The number of input words processed. * train_words - The total number of words in the training text (not * including words removed from the vocabuly by ReduceVocab). */ // This block prints a progress update, and also adjusts the training // 'alpha' parameter. if (word_count - last_word_count > 10000) { word_count_actual += word_count - last_word_count; last_word_count = word_count; // The percentage complete is based on the total number of passes we are // doing and not just the current pass. if ((debug_mode > 1)) { now=clock(); printf("%cAlpha: %f Progress: %.2f%% Words/thread/sec: %.2fk ", 13, alpha, // Percent complete = [# of input words processed] / // ([# of passes] * [# of words in a pass]) word_count_actual / (real)(iter * train_words + 1) * 100, word_count_actual / ((real)(now - start + 1) / (real)CLOCKS_PER_SEC * 1000)); fflush(stdout); } // Update alpha to: [initial alpha] * [percent of training remaining] // This means that alpha will gradually decrease as we progress through // the training text. alpha = starting_alpha * (1 - word_count_actual / (real)(iter * train_words + 1)); // Don't let alpha go below [initial alpha] * 0.0001. if (alpha < starting_alpha * 0.0001) alpha = starting_alpha * 0.0001; } // This 'if' block retrieves the next sentence from the training text and // stores it in 'sen'. // TODO - Under what condition would sentence_length not be zero? if (sentence_length == 0) { while (1) { // Read the next word from the training data and lookup its index in // the vocab table. 'word' is the word's vocab index. word = ReadWordIndex(fi); if (feof(fi)) break; // If the word doesn't exist in the vocabulary, skip it. if (word == -1) continue; // Track the total number of training words processed. word_count++; // 'vocab' word 0 is a special token "</s>" which indicates the end of // a sentence. if (word == 0) break; /* * ================================= * Subsampling of Frequent Words * ================================= * This code randomly discards training words, but is designed to * keep the relative frequencies the same. That is, less frequent * words will be discarded less often. * * We first calculate the probability that we want to *keep* the word; * this is the value 'ran'. Then, to decide whether to keep the word, * we generate a random fraction (0.0 - 1.0), and if 'ran' is smaller * than this number, we discard the word. This means that the smaller * 'ran' is, the more likely it is that we'll discard this word. * * The quantity (vocab[word].cn / train_words) is the fraction of all * the training words which are 'word'. Let's represent this fraction * by x. * * Using the default 'sample' value of 0.001, the equation for ran is: * ran = (sqrt(x / 0.001) + 1) * (0.001 / x) * * You can plot this function to see it's behavior; it has a curved * L shape. * * Here are some interesting points in this function (again this is * using the default sample value of 0.001). * - ran = 1 (100% chance of being kept) when x <= 0.0026. * - That is, any word which is 0.0026 of the words *or fewer* * will be kept 100% of the time. Only words which represent * more than 0.26% of the total words will be subsampled. * - ran = 0.5 (50% chance of being kept) when x = 0.00746. * - ran = 0.033 (3.3% chance of being kept) when x = 1. * - That is, if a word represented 100% of the training set * (which of course would never happen), it would only be * kept 3.3% of the time. * * NOTE: Seems like it would be more efficient to pre-calculate this * probability for each word and store it in the vocab table... * * Words that are discarded by subsampling aren't added to our training * 'sentence'. This means the discarded word is neither used as an * input word or a context word for other inputs. */ if (sample > 0) { // Calculate the probability of keeping 'word'. real ran = (sqrt(vocab[word].cn / (sample * train_words)) + 1) * (sample * train_words) / vocab[word].cn; // Generate a random number. // The multiplier is 25.xxx billion, so 'next_random' is a 64-bit integer. next_random = next_random * (unsigned long long)25214903917 + 11; // If the probability is less than a random fraction, discard the word. // // (next_random & 0xFFFF) extracts just the lower 16 bits of the // random number. Dividing this by 65536 (2^16) gives us a fraction // between 0 and 1. So the code is just generating a random fraction. if (ran < (next_random & 0xFFFF) / (real)65536) continue; } // If we kept the word, add it to the sentence. sen[sentence_length] = word; sentence_length++; // Verify the sentence isn't too long. if (sentence_length >= MAX_SENTENCE_LENGTH) break; } sentence_position = 0; } if (feof(fi) || (word_count > train_words / num_threads)) { word_count_actual += word_count - last_word_count; local_iter--; if (local_iter == 0) break; word_count = 0; last_word_count = 0; sentence_length = 0; fseek(fi, file_size / (long long)num_threads * (long long)id, SEEK_SET); continue; } // Get the next word in the sentence. The word is represented by its index // into the vocab table. word = sen[sentence_position]; if (word == -1) continue; for (c = 0; c < layer1_size; c++) neu1[c] = 0; for (c = 0; c < layer1_size; c++) neu1e[c] = 0; // This is a standard random integer generator, as seen here: // https://en.wikipedia.org/wiki/Linear_congruential_generator next_random = next_random * (unsigned long long)25214903917 + 11; // 'b' becomes a random integer between 0 and 'window' - 1. // This is the amount we will shrink the window size by. b = next_random % window; /* * ==================================== * CBOW Architecture * ==================================== * sen - This is the array of words in the sentence. Subsampling has * already been applied. Words are represented by their ids. * * sentence_position - This is the index of the current input word. * * a - Offset into the current window, relative to the window start. * a will range from 0 to (window * 2) * * b - The amount to shrink the context window by. * * c - 'c' is a scratch variable used in two unrelated ways: * 1. It's first used as the index of the current context word * within the sentence (the `sen` array). * 2. It's then used as the for-loop variable for calculating * vector dot-products and other arithmetic. * * syn0 - The hidden layer weights. Note that the weights are stored as a * 1D array, so word 'i' is found at (i * layer1_size). * * target - The output word we're working on. If it's the positive sample * then `label` is 1. `label` is 0 for negative samples. * Note: `target` and `label` are only used in negative sampling, * and not HS. * * neu1 - This vector will hold the *average* of all of the context word * vectors. This is the output of the hidden layer for CBOW. * * neu1e - Holds the gradient for updating the hidden layer weights. * It's a vector of length 300, not a matrix. * This same gradient update is applied to all context word * vectors. */ if (cbow) { //train the cbow architecture // in -> hidden cw = 0; // This loop will sum together the word vectors for all of the context // words. // // Loop over the positions in the context window (skipping the word at // the center). 'a' is just the offset within the window, it's not the // index relative to the beginning of the sentence. for (a = b; a < window * 2 + 1 - b; a++) if (a != window) { // Convert the window offset 'a' into an index 'c' into the sentence // array. c = sentence_position - window + a; // Verify c isn't outisde the bounds of the sentence. if (c < 0) continue; if (c >= sentence_length) continue; // Get the context word. That is, get the id of the word (its index in // the vocab table). last_word = sen[c]; // At this point we have two words identified: // 'word' - The word (word ID) at our current position in the // sentence (in the center of a context window). // 'last_word' - The word (word ID) at a position within the context // window. // Verify that the word exists in the vocab if (last_word == -1) continue; // Add the word vector for this context word to the running sum in // neur1. // `layer1_size` is 300, `neu1` is length 300 for (c = 0; c < layer1_size; c++) neu1[c] += syn0[c + last_word * layer1_size]; // Count the number of context words. cw++; } // Skip if there were somehow no context words. if (cw) { // neu1 was the sum of the context word vectors, and now becomes // their average. for (c = 0; c < layer1_size; c++) neu1[c] /= cw; // // HIERARCHICAL SOFTMAX // vocab[word] // .point - A variable-length list of row ids, which are the output // rows to train on. // .code - A variable-length list of 0s and 1s, which are the desired // labels for the outputs in `point`. // .codelen - The length of the `code` array for this // word. // if (hs) for (d = 0; d < vocab[word].codelen; d++) { f = 0; // point[d] is the index of a row of the ouput matrix. // l2 is the index of that word in the output layer weights (syn1). l2 = vocab[word].point[d] * layer1_size; // Propagate hidden -> output // neu1 is the average of the context words from the hidden layer. // This loop computes the dot product between neu1 and the output // weights for the output word at point[d]. for (c = 0; c < layer1_size; c++) f += neu1[c] * syn1[c + l2]; // Apply the sigmoid activation to the current output neuron. if (f <= -MAX_EXP) continue; else if (f >= MAX_EXP) continue; else f = expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))]; // 'g' is the error multiplied by the learning rate. // The error is (label - f), so label = (1 - code), meaning if // code is 0, then this is a positive sample and vice versa. g = (1 - vocab[word].code[d] - f) * alpha; // Propagate errors output -> hidden for (c = 0; c < layer1_size; c++) neu1e[c] += g * syn1[c + l2]; // Learn weights hidden -> output for (c = 0; c < layer1_size; c++) syn1[c + l2] += g * neu1[c]; } // NEGATIVE SAMPLING // Rather than performing backpropagation for every word in our // vocabulary, we only perform it for the positive sample and a few // negative samples (the number of words is given by 'negative'). // These negative words are selected using a "unigram" distribution, // which is generated in the function InitUnigramTable. if (negative > 0) for (d = 0; d < negative + 1; d++) { // On the first iteration, we're going to train the positive sample. if (d == 0) { target = word; label = 1; // On the other iterations, we'll train the negative samples. } else { // Pick a random word to use as a 'negative sample'; do this using // the unigram table. // Get a random integer. next_random = next_random * (unsigned long long)25214903917 + 11; // 'target' becomes the index of the word in the vocab to use as // the negative sample. target = table[(next_random >> 16) % table_size]; // If the target is the special end of sentence token, then just // pick a random word from the vocabulary instead. if (target == 0) target = next_random % (vocab_size - 1) + 1; // Don't use the positive sample as a negative sample! if (target == word) continue; // Mark this as a negative example. label = 0; } // At this point, target might either be the positive sample or a // negative sample, depending on the value of `label`. // Get the index of the target word in the output layer. l2 = target * layer1_size; // Calculate the dot product between: // neu1 - The average of the context word vectors. // syn1neg[l2] - The output weights for the target word. f = 0; for (c = 0; c < layer1_size; c++) f += neu1[c] * syn1neg[c + l2]; // This block does two things: // 1. Calculates the output of the network for this training // pair, using the expTable to evaluate the output layer // activation function. // 2. Calculate the error at the output, stored in 'g', by // subtracting the network output from the desired output, // and finally multiply this by the learning rate. if (f > MAX_EXP) g = (label - 1) * alpha; else if (f < -MAX_EXP) g = (label - 0) * alpha; else g = (label - expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))]) * alpha; // Multiply the error by the output layer weights. // (I think this is the gradient calculation?) // Accumulate these gradients over all of the negative samples. for (c = 0; c < layer1_size; c++) neu1e[c] += g * syn1neg[c + l2]; // Update the output layer weights by multiplying the output error // by the average of the context word vectors. for (c = 0; c < layer1_size; c++) syn1neg[c + l2] += g * neu1[c]; } // hidden -> in // Backpropagate the error to the hidden layer (the word vectors). // This code is used both for heirarchical softmax and for negative // sampling. // // Loop over the positions in the context window (skipping the word at // the center). 'a' is just the offset within the window, it's not // the index relative to the beginning of the sentence. for (a = b; a < window * 2 + 1 - b; a++) if (a != window) { // Convert the window offset 'a' into an index 'c' into the sentence // array. c = sentence_position - window + a; // Verify c isn't outisde the bounds of the sentence. if (c < 0) continue; if (c >= sentence_length) continue; // Get the context word. That is, get the id of the word (its index in // the vocab table). last_word = sen[c]; // Verify word exists in vocab. if (last_word == -1) continue; // Note that `c` is no longer the sentence position, it's just a // for-loop index. // Add the gradient in the vector `neu1e` to the word vector for // the current context word. // syn0[last_word * layer1_size] <-- Accesses the word vector. for (c = 0; c < layer1_size; c++) syn0[c + last_word * layer1_size] += neu1e[c]; } } } /* * ==================================== * Skip-gram Architecture * ==================================== * sen - This is the array of words in the sentence. Subsampling has * already been applied. Words are represented by their ids. * * sentence_position - This is the index of the current input word. * * a - Offset into the current window, relative to the window start. * a will range from 0 to (window * 2) * * b - The amount to shrink the context window by. * * c - 'c' is a scratch variable used in two unrelated ways: * 1. It's first used as the index of the current context word * within the sentence (the `sen` array). * 2. It's then used as the for-loop variable for calculating * vector dot-products and other arithmetic. * * syn0 - The hidden layer weights. Note that the weights are stored as a * 1D array, so word 'i' is found at (i * layer1_size). * * l1 - Index into the hidden layer (syn0). Index of the start of the * weights for the current input word. * * target - The output word we're working on. If it's the positive sample * then `label` is 1. `label` is 0 for negative samples. * Note: `target` and `label` are only used in negative sampling, * and not HS. */ else { // Loop over the positions in the context window (skipping the word at // the center). 'a' is just the offset within the window, it's not // the index relative to the beginning of the sentence. for (a = b; a < window * 2 + 1 - b; a++) if (a != window) { // Convert the window offset 'a' into an index 'c' into the sentence // array. c = sentence_position - window + a; // Verify c isn't outisde the bounds of the sentence. if (c < 0) continue; if (c >= sentence_length) continue; // Get the context word. That is, get the id of the word (its index in // the vocab table). last_word = sen[c]; // At this point we have two words identified: // 'word' - The word at our current position in the sentence (in the // center of a context window). // 'last_word' - The word at a position within the context window. // Verify that the word exists in the vocab (I don't think this should // ever be the case?) if (last_word == -1) continue; // Calculate the index of the start of the weights for 'last_word'. l1 = last_word * layer1_size; for (c = 0; c < layer1_size; c++) neu1e[c] = 0; // HIERARCHICAL SOFTMAX if (hs) for (d = 0; d < vocab[word].codelen; d++) { f = 0; l2 = vocab[word].point[d] * layer1_size; // Propagate hidden -> output for (c = 0; c < layer1_size; c++) f += syn0[c + l1] * syn1[c + l2]; if (f <= -MAX_EXP) continue; else if (f >= MAX_EXP) continue; else f = expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))]; // 'g' is the gradient multiplied by the learning rate g = (1 - vocab[word].code[d] - f) * alpha; // Propagate errors output -> hidden for (c = 0; c < layer1_size; c++) neu1e[c] += g * syn1[c + l2]; // Learn weights hidden -> output for (c = 0; c < layer1_size; c++) syn1[c + l2] += g * syn0[c + l1]; } // NEGATIVE SAMPLING // Rather than performing backpropagation for every word in our // vocabulary, we only perform it for a few words (the number of words // is given by 'negative'). // These words are selected using a "unigram" distribution, which is generated // in the function InitUnigramTable if (negative > 0) for (d = 0; d < negative + 1; d++) { // On the first iteration, we're going to train the positive sample. if (d == 0) { target = word; label = 1; // On the other iterations, we'll train the negative samples. } else { // Pick a random word to use as a 'negative sample'; do this using // the unigram table. // Get a random integer. next_random = next_random * (unsigned long long)25214903917 + 11; // 'target' becomes the index of the word in the vocab to use as // the negative sample. target = table[(next_random >> 16) % table_size]; // If the target is the special end of sentence token, then just // pick a random word from the vocabulary instead. if (target == 0) target = next_random % (vocab_size - 1) + 1; // Don't use the positive sample as a negative sample! if (target == word) continue; // Mark this as a negative example. label = 0; } // Get the index of the target word in the output layer. l2 = target * layer1_size; // At this point, our two words are represented by their index into // the layer weights. // l1 - The index of our input word within the hidden layer weights. // l2 - The index of our output word within the output layer weights. // label - Whether this is a positive (1) or negative (0) example. // Calculate the dot-product between the input words weights (in // syn0) and the output word's weights (in syn1neg). // Note that this calculates the dot-product manually using a for // loop over the vector elements! f = 0; for (c = 0; c < layer1_size; c++) f += syn0[c + l1] * syn1neg[c + l2]; // This block does two things: // 1. Calculates the output of the network for this training // pair, using the expTable to evaluate the output layer // activation function. // 2. Calculate the error at the output, stored in 'g', by // subtracting the network output from the desired output, // and finally multiply this by the learning rate. if (f > MAX_EXP) g = (label - 1) * alpha; else if (f < -MAX_EXP) g = (label - 0) * alpha; else g = (label - expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))]) * alpha; // Multiply the error by the output layer weights. // Accumulate these gradients over the negative samples and the one // positive sample. for (c = 0; c < layer1_size; c++) neu1e[c] += g * syn1neg[c + l2]; // Update the output layer weights by multiplying the output error // by the hidden layer weights. for (c = 0; c < layer1_size; c++) syn1neg[c + l2] += g * syn0[c + l1]; } // Once the hidden layer gradients for the negative samples plus the // one positive sample have been accumulated, update the hidden layer // weights. // Note that we do not average the gradient before applying it. for (c = 0; c < layer1_size; c++) syn0[c + l1] += neu1e[c]; } } // Advance to the next word in the sentence. sentence_position++; // Check if we've reached the end of the sentence. // If so, set sentence_length to 0 and we'll read a new sentence at the // beginning of this loop. if (sentence_position >= sentence_length) { sentence_length = 0; continue; } } fclose(fi); free(neu1); free(neu1e); pthread_exit(NULL); } /** * ======== TrainModel ======== * Main entry point to the training process. */ void TrainModel() { long a, b, c, d; FILE *fo; pthread_t *pt = (pthread_t *)malloc(num_threads * sizeof(pthread_t)); printf("Starting training using file %s\n", train_file); starting_alpha = alpha; // Either load a pre-existing vocabulary, or learn the vocabulary from // the training file. if (read_vocab_file[0] != 0) ReadVocab(); else LearnVocabFromTrainFile(); // Save the vocabulary. if (save_vocab_file[0] != 0) SaveVocab(); // Stop here if no output_file was specified. if (output_file[0] == 0) return; // Allocate the weight matrices and initialize them. InitNet(); // If we're using negative sampling, initialize the unigram table, which // is used to pick words to use as "negative samples" (with more frequent // words being picked more often). if (negative > 0) InitUnigramTable(); // Record the start time of training. start = clock(); // Run training, which occurs in the 'TrainModelThread' function. for (a = 0; a < num_threads; a++) pthread_create(&pt[a], NULL, TrainModelThread, (void *)a); for (a = 0; a < num_threads; a++) pthread_join(pt[a], NULL); fo = fopen(output_file, "wb"); if (classes == 0) { // Save the word vectors fprintf(fo, "%lld %lld\n", vocab_size, layer1_size); for (a = 0; a < vocab_size; a++) { fprintf(fo, "%s ", vocab[a].word); if (binary) for (b = 0; b < layer1_size; b++) fwrite(&syn0[a * layer1_size + b], sizeof(real), 1, fo); else for (b = 0; b < layer1_size; b++) fprintf(fo, "%lf ", syn0[a * layer1_size + b]); fprintf(fo, "\n"); } } else { // Run K-means on the word vectors int clcn = classes, iter = 10, closeid; int *centcn = (int *)malloc(classes * sizeof(int)); int *cl = (int *)calloc(vocab_size, sizeof(int)); real closev, x; real *cent = (real *)calloc(classes * layer1_size, sizeof(real)); for (a = 0; a < vocab_size; a++) cl[a] = a % clcn; for (a = 0; a < iter; a++) { for (b = 0; b < clcn * layer1_size; b++) cent[b] = 0; for (b = 0; b < clcn; b++) centcn[b] = 1; for (c = 0; c < vocab_size; c++) { for (d = 0; d < layer1_size; d++) cent[layer1_size * cl[c] + d] += syn0[c * layer1_size + d]; centcn[cl[c]]++; } for (b = 0; b < clcn; b++) { closev = 0; for (c = 0; c < layer1_size; c++) { cent[layer1_size * b + c] /= centcn[b]; closev += cent[layer1_size * b + c] * cent[layer1_size * b + c]; } closev = sqrt(closev); for (c = 0; c < layer1_size; c++) cent[layer1_size * b + c] /= closev; } for (c = 0; c < vocab_size; c++) { closev = -10; closeid = 0; for (d = 0; d < clcn; d++) { x = 0; for (b = 0; b < layer1_size; b++) x += cent[layer1_size * d + b] * syn0[c * layer1_size + b]; if (x > closev) { closev = x; closeid = d; } } cl[c] = closeid; } } // Save the K-means classes for (a = 0; a < vocab_size; a++) fprintf(fo, "%s %d\n", vocab[a].word, cl[a]); free(centcn); free(cent); free(cl); } fclose(fo); } int ArgPos(char *str, int argc, char **argv) { int a; for (a = 1; a < argc; a++) if (!strcmp(str, argv[a])) { if (a == argc - 1) { printf("Argument missing for %s\n", str); exit(1); } return a; } return -1; } int main(int argc, char **argv) { int i; if (argc == 1) { printf("WORD VECTOR estimation toolkit v 0.1c\n\n"); printf("Options:\n"); printf("Parameters for training:\n"); printf("\t-train <file>\n"); printf("\t\tUse text data from <file> to train the model\n"); printf("\t-output <file>\n"); printf("\t\tUse <file> to save the resulting word vectors / word clusters\n"); printf("\t-size <int>\n"); printf("\t\tSet size of word vectors; default is 100\n"); printf("\t-window <int>\n"); printf("\t\tSet max skip length between words; default is 5\n"); printf("\t-sample <float>\n"); printf("\t\tSet threshold for occurrence of words. Those that appear with higher frequency in the training data\n"); printf("\t\twill be randomly down-sampled; default is 1e-3, useful range is (0, 1e-5)\n"); printf("\t-hs <int>\n"); printf("\t\tUse Hierarchical Softmax; default is 0 (not used)\n"); printf("\t-negative <int>\n"); printf("\t\tNumber of negative examples; default is 5, common values are 3 - 10 (0 = not used)\n"); printf("\t-threads <int>\n"); printf("\t\tUse <int> threads (default 12)\n"); printf("\t-iter <int>\n"); printf("\t\tRun more training iterations (default 5)\n"); printf("\t-min-count <int>\n"); printf("\t\tThis will discard words that appear less than <int> times; default is 5\n"); printf("\t-alpha <float>\n"); printf("\t\tSet the starting learning rate; default is 0.025 for skip-gram and 0.05 for CBOW\n"); printf("\t-classes <int>\n"); printf("\t\tOutput word classes rather than word vectors; default number of classes is 0 (vectors are written)\n"); printf("\t-debug <int>\n"); printf("\t\tSet the debug mode (default = 2 = more info during training)\n"); printf("\t-binary <int>\n"); printf("\t\tSave the resulting vectors in binary moded; default is 0 (off)\n"); printf("\t-save-vocab <file>\n"); printf("\t\tThe vocabulary will be saved to <file>\n"); printf("\t-read-vocab <file>\n"); printf("\t\tThe vocabulary will be read from <file>, not constructed from the training data\n"); printf("\t-cbow <int>\n"); printf("\t\tUse the continuous bag of words model; default is 1 (use 0 for skip-gram model)\n"); printf("\nExamples:\n"); printf("./word2vec -train data.txt -output vec.txt -size 200 -window 5 -sample 1e-4 -negative 5 -hs 0 -binary 0 -cbow 1 -iter 3\n\n"); return 0; } output_file[0] = 0; save_vocab_file[0] = 0; read_vocab_file[0] = 0; if ((i = ArgPos((char *)"-size", argc, argv)) > 0) layer1_size = atoi(argv[i + 1]); if ((i = ArgPos((char *)"-train", argc, argv)) > 0) strcpy(train_file, argv[i + 1]); if ((i = ArgPos((char *)"-save-vocab", argc, argv)) > 0) strcpy(save_vocab_file, argv[i + 1]); if ((i = ArgPos((char *)"-read-vocab", argc, argv)) > 0) strcpy(read_vocab_file, argv[i + 1]); if ((i = ArgPos((char *)"-debug", argc, argv)) > 0) debug_mode = atoi(argv[i + 1]); if ((i = ArgPos((char *)"-binary", argc, argv)) > 0) binary = atoi(argv[i + 1]); if ((i = ArgPos((char *)"-cbow", argc, argv)) > 0) cbow = atoi(argv[i + 1]); if (cbow) alpha = 0.05; if ((i = ArgPos((char *)"-alpha", argc, argv)) > 0) alpha = atof(argv[i + 1]); if ((i = ArgPos((char *)"-output", argc, argv)) > 0) strcpy(output_file, argv[i + 1]); if ((i = ArgPos((char *)"-window", argc, argv)) > 0) window = atoi(argv[i + 1]); if ((i = ArgPos((char *)"-sample", argc, argv)) > 0) sample = atof(argv[i + 1]); if ((i = ArgPos((char *)"-hs", argc, argv)) > 0) hs = atoi(argv[i + 1]); if ((i = ArgPos((char *)"-negative", argc, argv)) > 0) negative = atoi(argv[i + 1]); if ((i = ArgPos((char *)"-threads", argc, argv)) > 0) num_threads = atoi(argv[i + 1]); if ((i = ArgPos((char *)"-iter", argc, argv)) > 0) iter = atoi(argv[i + 1]); if ((i = ArgPos((char *)"-min-count", argc, argv)) > 0) min_count = atoi(argv[i + 1]); if ((i = ArgPos((char *)"-classes", argc, argv)) > 0) classes = atoi(argv[i + 1]); // Allocate the vocabulary table. vocab = (struct vocab_word *)calloc(vocab_max_size, sizeof(struct vocab_word)); // Allocate the hash table for mapping word strings to word entries. vocab_hash = (int *)calloc(vocab_hash_size, sizeof(int)); /* * ======== Precomputed Exp Table ======== * To calculate the softmax output, they use a table of values which are * pre-computed here. * * From the top of this file: * #define EXP_TABLE_SIZE 1000 * #define MAX_EXP 6 * * First, let's look at this inner term: * i / (real)EXP_TABLE_SIZE * 2 - 1 * This is just a straight line that goes from -1 to +1. * (0, -1.0), (1, -0.998), (2, -0.996), ... (999, 0.998), (1000, 1.0). * * Next, multiplying this by MAX_EXP = 6, it causes the output to range * from -6 to +6 instead of -1 to +1. * (0, -6.0), (1, -5.988), (2, -5.976), ... (999, 5.988), (1000, 6.0). * * So the total input range of the table is * Range = MAX_EXP * 2 = 12 * And the increment on the inputs is * Increment = Range / EXP_TABLE_SIZE = 0.012 * * Let's say we want to compute the output for the value x = 0.25. How do * we calculate the position in the table? * index = (x - -MAX_EXP) / increment * Which we can re-write as: * index = (x + MAX_EXP) / (range / EXP_TABLE_SIZE) * = (x + MAX_EXP) / ((2 * MAX_EXP) / EXP_TABLE_SIZE) * = (x + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2) * * The last form is what we find in the code elsewhere for using the table: * expTable[(int)((f + MAX_EXP) * (EXP_TABLE_SIZE / MAX_EXP / 2))] * */ // Allocate the table, 1000 floats. expTable = (real *)malloc((EXP_TABLE_SIZE + 1) * sizeof(real)); // For each position in the table... for (i = 0; i < EXP_TABLE_SIZE; i++) { // Calculate the output of e^x for values in the range -6.0 to +6.0. expTable[i] = exp((i / (real)EXP_TABLE_SIZE * 2 - 1) * MAX_EXP); // Precompute the exp() table // Currently the table contains the function exp(x). // We are going to replace this with exp(x) / (exp(x) + 1), which is // just the sigmoid activation function! // Note that // exp(x) / (exp(x) + 1) // is equivalent to // 1 / (1 + exp(-x)) expTable[i] = expTable[i] / (expTable[i] + 1); // Precompute f(x) = x / (x + 1) } TrainModel(); return 0; }