serial_no
int64
1
24.2k
cuda_source
stringlengths
11
9.01M
2,301
#include "includes.h" /** * Programma che simula il comportamento del gpdt per * la risoluzione di un kernel di una serie di * valori di dimensione variabile utilizzando la * tecnologia cuda. * compilare con: * nvcc -o simil_gpdt_si_cuda simil_gpdt_si_cuda.cu * lanciare con: * ./simil_gpdt_si_cuda [numero vettori] [numero componenti] [numero di righe da calcolare] [tipo di kernel] [grado(int)/sigma(float)] **/ using namespace std; /** * Funzione che riempie i vettori con numeri * casuali compresi tra 0 e 99. **/ __global__ void Kernel_lineare(float *Vd, float *Ris, int N, int C, int dim_indici, int *ind, int *Vp, int *Vnp, int nr_max_val) { int x = blockIdx.x * blockDim.x + threadIdx.x; int y = blockIdx.y * blockDim.y + threadIdx.y; int j; int pos; int tmp_ind; float lin; for ( ; x < N ; x+=blockDim.x * gridDim.x) { for( ; y < dim_indici; y+=blockDim.y * gridDim.y) { tmp_ind = ind[y]; lin = 0.0; int Nr_val = Vnp[x]; for(j = 0; j < Nr_val; j++) { pos = Vp[x * nr_max_val + j]; lin = lin + (Vd[x * C + pos] * Vd[tmp_ind * C + pos]); } //Ris[x * dim_indici + y] = lin; Ris[y * N + x ] = lin; } } }
2,302
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include <stdlib.h> #include <stdio.h> cudaError_t cudaDotProduct(int *c, const int *a, const int *b, unsigned int size); int* allocAndAssignMat(int size); __global__ void dot(int *c, const int *a, const int *b) { int i = threadIdx.x + blockIdx.x * blockDim.x; c[i] += a[i] * b[i]; } int main() { cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); const int N = 10000; // this is the sqrt of the total elements or the len of one side of the square matrix const int* a = allocAndAssignMat(N * N); const int* b = allocAndAssignMat(N * N); int* c = (int*)malloc((N * N) * sizeof(int)); for (int i = 0; i < N * N; i++) { c[i] = 0; } int mySum = 0; cudaEventRecord(start); // Add vectors in parallel. cudaError_t cudaStatus = cudaDotProduct(c, a, b, N*N); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaDotProduct failed!"); return 1; } cudaEventRecord(stop); cudaEventSynchronize(stop); float milliseconds = 0; cudaEventElapsedTime(&milliseconds, start, stop); for (int i = 0; i < N*N; i++) { //printf("%d ", c[i]); mySum += c[i]; } //Results printf("Size of N*N: %d \nResult: %d \nTime in kernel %f \n", N * N, mySum, milliseconds); // cudaDeviceReset must be called before exiting in order for profiling and // tracing tools such as Nsight and Visual Profiler to show complete traces. cudaStatus = cudaDeviceReset(); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaDeviceReset failed!"); return 1; } return 0; } // Helper function for using CUDA to add vectors in parallel. cudaError_t cudaDotProduct(int *c, const int *a, const int *b, unsigned int size) { int *dev_a = 0; int *dev_b = 0; int *dev_c = 0; cudaError_t cudaStatus; // Choose which GPU to run on, change this on a multi-GPU system. cudaStatus = cudaSetDevice(0); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaSetDevice failed! Do you have a CUDA-capable GPU installed?"); goto Error; } // Allocate GPU buffers for three vectors (two input, one output) . cudaStatus = cudaMalloc((void**)&dev_c, size * sizeof(int)); // allocating the space on the gpu if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaMalloc failed!"); goto Error; } cudaStatus = cudaMalloc((void**)&dev_a, size * sizeof(int)); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaMalloc failed!"); goto Error; } cudaStatus = cudaMalloc((void**)&dev_b, size * sizeof(int)); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaMalloc failed!"); goto Error; } // Copy input vectors from host memory to GPU buffers. cudaStatus = cudaMemcpy(dev_a, a, size * sizeof(int), cudaMemcpyHostToDevice); // moving the data to the gpu counterpart not c as that is results if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaMemcpy failed!"); goto Error; } cudaStatus = cudaMemcpy(dev_b, b, size * sizeof(int), cudaMemcpyHostToDevice); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaMemcpy failed!"); goto Error; } // Launch a kernel on the GPU with one thread for each element. dot<<<(size+255)/256, 256>>>(dev_c, dev_a, dev_b); // execution configuration - // Check for any errors launching the kernel cudaStatus = cudaGetLastError(); if (cudaStatus != cudaSuccess) { fprintf(stderr, "addKernel launch failed: %s\n", cudaGetErrorString(cudaStatus)); goto Error; } // cudaDeviceSynchronize waits for the kernel to finish, and returns // any errors encountered during the launch. cudaStatus = cudaDeviceSynchronize(); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaDeviceSynchronize returned error code %d after launching addKernel!\n", cudaStatus); goto Error; } // Copy output vector from GPU buffer to host memory. cudaStatus = cudaMemcpy(c, dev_c, size * sizeof(int), cudaMemcpyDeviceToHost); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaMemcpy failed!"); goto Error; } Error: cudaFree(dev_c); cudaFree(dev_a); cudaFree(dev_b); return cudaStatus; } int* allocAndAssignMat(int size) { /* This function takes in the size of the matrix (N*N) and returns a pointer with appropriate memory allocated as well as filled with values @params: int size @returns: int* ptr */ int* ptr = (int*)malloc(size * sizeof(int)); for (int i = 0; i < size; i++) { ptr[i] = 2; } return ptr; }
2,303
#include <math.h> #include <stdlib.h> #include <stdio.h> #include <string.h> #include "time.h" #define NUM_THREADS 256 int nsamp = 65536, ndms = 1024, detrendLen = 32768; // --------------------------- Detrend and Normalisation kernel ---------------------------- __global__ void detrend_normalise(float *input, int detrendLen) { // Store temporary least-square fit values __shared__ float3 shared[NUM_THREADS]; // First pass, each thread computes its part of the buffer { float sy = 0, sxy = 0, sxx = 0; for (unsigned i = threadIdx.x; i < detrendLen; i += blockDim.x) { float x = - detrendLen * 0.5 + 0.5 + i; int index = blockIdx.y * gridDim.x * detrendLen + blockIdx.x * detrendLen + i; float y = input[index]; sy += y; sxy += x * y; sxx += x * x; } // Initialise shared memory shared[threadIdx.x].x = sy; shared[threadIdx.x].y = sxy; shared[threadIdx.x].z = sxx; } __syncthreads(); // Perform the rest of the computation through reduction for (unsigned i = NUM_THREADS / 2; i >= 1; i /= 2) { if (threadIdx.x < i) { shared[threadIdx.x].x += shared[threadIdx.x + i].x; shared[threadIdx.x].y += shared[threadIdx.x + i].y; shared[threadIdx.x].z += shared[threadIdx.x + i].z; } __syncthreads(); } if (threadIdx.x == 0) { shared[0].y /= shared[0].z; shared[0].x /= detrendLen; } __syncthreads(); // Detrend and compute partial standard deviation { float a = shared[0].x; float b = shared[0].y; float stddev = 0; for (unsigned i = threadIdx.x; i < detrendLen ; i += blockDim.x) { float x = - detrendLen / 2.0 + 0.5 + i; int index = blockIdx.y * gridDim.x * detrendLen + blockIdx.x * detrendLen + i; float val = input[index] - (a + b * x); input[index] = val; stddev += val * val; } shared[threadIdx.x].z = stddev; } __syncthreads(); // Compute the full standard deviation through reduction for (unsigned i = NUM_THREADS / 2; i >= 1; i /= 2) if (threadIdx.x < i) shared[threadIdx.x].z += shared[threadIdx.x + i].z; __syncthreads(); if (threadIdx.x == 0) shared[0].z = sqrt(shared[0].z / detrendLen); __syncthreads(); // Normalise Data float stddev = shared[0].z; for (unsigned i = threadIdx.x; i < detrendLen ; i += blockDim.x) input[blockIdx.y * gridDim.x * detrendLen + blockIdx.x * detrendLen + i] /= stddev; } // --------------------------- Main processing ---------------------------- int main(int argc, char *argv[]) { float *input, *d_input; int i, j; // Allocate and generate buffers input = (float *) malloc(nsamp * ndms * sizeof(float)); srand ( time(NULL) ); for(i = 0; i < ndms; i++) for(j = 0; j < nsamp; j++) input[i * nsamp + j] = ((float)rand() / (float)RAND_MAX) + j * 0.001 + i * j * 0.001; printf("nsamp: %d, ndms: %d\n", nsamp, ndms); // FILE *fp = fopen("/home/lessju/Code/MDSM/release/pelican-mdsm/pipelines/output.dat", "rb"); // fread(input, sizeof(float), nsamp, fp); // fclose(fp); // Initialise cudaSetDevice(0); cudaThreadSetCacheConfig(cudaFuncCachePreferL1); cudaEvent_t event_start, event_stop; float timestamp; cudaEventCreate(&event_start); cudaEventCreate(&event_stop); // Allocate GPU memory and copy data cudaMalloc((void **) &d_input, nsamp * ndms * sizeof(float) ); cudaMemcpy(d_input, input, nsamp * ndms * sizeof(float), cudaMemcpyHostToDevice); // Call kernel cudaEventRecord(event_start, 0); detrend_normalise<<<dim3(nsamp / detrendLen, ndms), NUM_THREADS>>>(d_input, detrendLen); cudaEventRecord(event_stop, 0); cudaEventSynchronize(event_stop); cudaEventElapsedTime(&timestamp, event_start, event_stop); printf("Performed detrending in: %lf\n", timestamp); // Get output cudaMemcpy(input, d_input, ndms * nsamp * sizeof(float), cudaMemcpyDeviceToHost); FILE *fp = fopen("testDetrend.dat", "wb"); fwrite(input, sizeof(float), nsamp * ndms, fp); fclose(fp); }
2,304
//#include "cudacpp\DeviceVector.h" template<typename type, int size> __global__ void setKernel(type* c, type val) { auto idx = threadIdx.x * size; #pragma unroll(size) for (auto i = 0; i < size; i++) { c[idx] = val; idx++; } }
2,305
template<typename TF> __global__ void doublify(TF* a, const int itot, const int jtot) { const int i = blockIdx.x*blockDim.x + threadIdx.x; const int j = blockIdx.y*blockDim.y + threadIdx.y; const int k = blockIdx.z; int ijk = i + j*itot + k*itot*jtot; a[ijk] += TF(ijk); } template<typename TF> void launch_doublify(TF* a, const int itot, const int jtot, const int ktot) { dim3 grid_gpu (1, 1, ktot); dim3 block_gpu(itot, jtot, 1); doublify<<<grid_gpu, block_gpu>>>(a, itot, jtot); }
2,306
#include "includes.h" __global__ void rowDiv(float* a, float* b, float* c, int M, int N){ int i = blockIdx.x*blockDim.x + threadIdx.x; c[i] = a[i]/b[blockIdx.x]; }
2,307
#ifndef __HELPER__ #define __HELPER__ #include <bits/stdc++.h> using namespace std; #define __HD__ __host__ __device__ #define SQR(x) ((x) * (x)) __inline__ __HD__ int rnd(int x, int n) { return ((x + n - 1) / n) * n; } #define LG_BLOCK_N 64 #define LG_BLOCK_M 64 #define LG_BLOCK_K 8 #define LG_THREAD_N 8 #define LG_THREAD_M 4 #define LG_GRID_SIZE ((LG_BLOCK_N * LG_BLOCK_M) / (LG_THREAD_N * LG_THREAD_M)) #define SM_BLOCK_N 32 #define SM_BLOCK_M 32 #define SM_BLOCK_K 4 #define SM_THREAD_N 4 #define SM_THREAD_M 2 #define SM_GRID_SIZE ((SM_BLOCK_N * SM_BLOCK_M) / (SM_THREAD_N * SM_THREAD_M)) #endif void setGrid(int n, dim3 &blockDim, dim3 &gridDim) { if (n > 384) { blockDim.x = LG_GRID_SIZE; blockDim.y = blockDim.z = 1; int nn = rnd(n, LG_BLOCK_N); gridDim.x = nn / LG_BLOCK_N; gridDim.y = nn / LG_BLOCK_M; gridDim.z = 1; } else { blockDim.x = SM_GRID_SIZE; blockDim.y = blockDim.z = 1; int nn = rnd(n, SM_BLOCK_N); gridDim.x = nn / SM_BLOCK_N; gridDim.y = nn / SM_BLOCK_M; gridDim.z = 1; } }
2,308
#include <cuda.h> #include <cuda_runtime.h> #include <stdio.h> #include <stdlib.h> #include <time.h> #define BLOCK_SIZE 32 #define WA 64 #define HA 64 #define HC 3 #define WC 3 #define PAD 1 #define WB (WA+2*PAD - WC + 1) #define HB (HA+2*PAD - HC + 1) #define CHANNEL_SIZE 3 __device__ void flat_conv(float* Input, float* Kernel, float* Output,int* image_size, int* kernel_size, int* pad,int* out_w) { //__shared__ float kernel_part[kernel_size[2]][kernel_size[3]][kernel_size[1]]; //__shared__ float kernel_part[3][3][3]; extern __shared__ float kernel_part[]; int col_idx = blockIdx.x - pad[0] + threadIdx.x; int row_idx = blockIdx.y - pad[0] + threadIdx.y; int img_flat_size = image_size[1]*image_size[2]; int kernel_flat_size = kernel_size[2]*kernel_size[3]; if( image_size[2]>col_idx && col_idx >=0 && image_size[1]>row_idx && row_idx >=0) { kernel_part[(threadIdx.y * kernel_size[3]+threadIdx.x)*kernel_size[1]+threadIdx.z] = Input[(col_idx * image_size[2] +row_idx) + img_flat_size*threadIdx.z] * Kernel[threadIdx.y*kernel_size[3] + threadIdx.x + kernel_flat_size*threadIdx.z]; } else { kernel_part[(threadIdx.y * kernel_size[3]+threadIdx.x)*kernel_size[1]+threadIdx.z] = 0; } //__syncthreads; atomicAdd(&(Output[blockIdx.x * out_w[0] +blockIdx.y]), kernel_part[(threadIdx.y * kernel_size[3]+threadIdx.x)*kernel_size[1]+threadIdx.z]); } __global__ void conv(float* Input, float* Kernel, float* Output,int* image_size, int* kernel_size,int* pad) { int out_w = image_size[2]+2*pad[0] - kernel_size[3] + 1; int out_h = image_size[1]+2*pad[0] - kernel_size[2] + 1; int flat_kernel_size = kernel_size[3]*kernel_size[2]*kernel_size[1]; int flat_img_size = out_w*out_h; flat_conv(Input, Kernel + flat_kernel_size*blockIdx.z , Output + flat_img_size*blockIdx.z, image_size, kernel_size, pad,&out_w); } void randomInit(float* data, int size) { for (int i = 0; i < size; ++i) data[i] = (rand() / (float)RAND_MAX); } __host__ int main(void) { // float h_a[3][64][64] ={0.0}; // h_a[0][0][0] = 2.1; // h_a[1][0][0] = 2.1; // h_a[2][0][0] = 2.1; // float h_b[2][3][3][3] ={1.0,1.0,1.0,1.0,1.0,1.0,1.0,1.0,1.0, // 1.0,1.0,1.0,1.0,1.0,1.0,1.0,1.0,1.0, // 1.0,1.0,1.0,1.0,1.0,1.0,1.0,1.0,1.0, // 2.0,2.0,2.0,2.0,2.0,2.0,2.0,2.0,2.0, // 2.0,2.0,2.0,2.0,2.0,2.0,2.0,2.0,2.0, // 2.0,2.0,2.0,2.0,2.0,2.0,2.0,2.0,2.0}; // float h_c[2][64][64] ={0.0}; int kernel_size[4] ={2,3,3,3}; //O I H W; int image_size[3] = {3,64,64}; // O H W; float* h_a; float *h_b; float* h_c; int h_a_size = sizeof(float)*3*64*64; int h_b_size = sizeof(float)*3*64*64; int h_c_size = sizeof(float)*3*64*64; h_a = (float*)malloc(h_a_size); h_b = (float*)malloc(h_b_size); h_c = (float*)malloc(h_c_size); randomInit(h_a,3*64*64); randomInit(h_b,2*3*3*3); randomInit(h_c,2*64*64); int pad = 1; float *cimg; float *coimg; float *ckernel; int * cimg_size; int * ckernel_size; int * cpad; cudaMalloc((void***)&cimg,h_a_size); cudaMalloc((void***)&ckernel,h_b_size); cudaMalloc((void***)&coimg,h_c_size); cudaMalloc(&cimg_size,sizeof(image_size)); cudaMalloc(&ckernel_size,sizeof(kernel_size)); cudaMalloc(&cpad,sizeof(int)); cudaMemcpy(cimg,h_a,h_a_size,cudaMemcpyHostToDevice); cudaMemcpy(ckernel,h_b,h_b_size,cudaMemcpyHostToDevice); cudaMemcpy(cimg_size,image_size,sizeof(image_size),cudaMemcpyHostToDevice); cudaMemcpy(ckernel_size,kernel_size,sizeof(kernel_size),cudaMemcpyHostToDevice); cudaMemcpy(cpad,&pad,sizeof(int),cudaMemcpyHostToDevice); dim3 threads(kernel_size[3], kernel_size[2], kernel_size[1]); dim3 grid(image_size[2],image_size[1],kernel_size[0]); clock_t start = clock(); int flat_kernel_size = kernel_size[3]* kernel_size[2]* kernel_size[1]*sizeof(float); conv <<< grid,threads,flat_kernel_size>>>(cimg,ckernel,coimg,cimg_size,ckernel_size,cpad); //Convolution <<< grid,threads>>>(cimg,ckernel,coimg,cimg_size,ckernel_size); clock_t end = clock(); cudaMemcpy(h_c,coimg,h_c_size,cudaMemcpyDeviceToHost); int cnt = 0; // for(int i = 0;i < 2; i++) // { // for(int j =0; j < WB;j ++) // { // for(int k =0; k < WB;k ++) // { // printf("%.0f ",h_c[cnt]); // cnt +=1; // } // printf("\n"); // } // printf("\n"); // } cudaFree(cimg); cudaFree(ckernel); cudaFree(coimg); cudaFree(cimg_size); cudaFree(ckernel_size); cudaFree(cpad); printf("%f",(float)(end - start)/CLOCKS_PER_SEC); }
2,309
__global__ void vectorAddition(const float* a, const float* b, float* result, const float scalar) { int index = blockIdx.x * blockDim.x + threadIdx.x; result[index] = a[index] + b[index] + scalar; }
2,310
#include<cuda.h> #include<stdio.h> #include<math.h> #define TILEWIDTH 32 __global__ void vecMulMatrixKernel(float* A, float* B, float* C, int n){ //each block loads the corresponding row of blocks of A matrix and column of blocks of B matrix, one block at a time and then clculates the product for that part then product of a the parts is added. // each thread loads 2 elements one from A and one from B in each phase // there are total gridDim.x phases // the element loaded is the element at the same position as this thread but in a different block //if run more thread than max then not run int tx=threadIdx.x; int ty=threadIdx.y; int bx=blockIdx.x; int by=blockIdx.y; int row=by*blockDim.y+ty; int col=bx*blockDim.x+tx; __shared__ float Ads[TILEWIDTH][TILEWIDTH]; __shared__ float Bds[TILEWIDTH][TILEWIDTH]; if(row<n && col <n){ int i; float val=0.0; for(i=0;i<gridDim.x-1;i++){ Ads[ty][tx] = A[ row*n + i*TILEWIDTH + tx]; Bds[ty][tx] = B[ (i*TILEWIDTH + ty)*n + col]; __syncthreads(); for(int k=0;k<TILEWIDTH;k++){ val+= Ads[ty][k]*Bds[tx][k]; } __syncthreads(); } if(i*TILEWIDTH + tx <n ) //if n was a multiple of blockDim then this was not required Ads[ty][tx] = A[ row*n + i*TILEWIDTH + tx]; if(i*TILEWIDTH + ty <n ) Bds[ty][tx] = B[ (i*TILEWIDTH + ty)*n + col]; __syncthreads(); int m =n%TILEWIDTH; if(m==0) m=TILEWIDTH; for(int k=0;k<m;k++){//printf("add"); val+= Ads[ty][k]*Bds[tx][k]; } __syncthreads(); C[row*n + col]= val; } } int min2Power(int x){ int res=1; while(res<x){ res*=2; } return res/2; } __host__ void vecMulMatrix(float* A,float* B,float* C, int n){ int size = n * n * sizeof(float); float *d_A, *d_B, *d_C; //Allocate device memory for A,B,C cudaMalloc((void**)&d_A, size); cudaMalloc((void**)&d_B, size); cudaMalloc((void**)&d_C, size); //copy A,B to device memory cudaMemcpy(d_A, A, size, cudaMemcpyHostToDevice); cudaMemcpy(d_B, B, size, cudaMemcpyHostToDevice); //call kernal function that the calculates the product and stores it in C dim3 dimBlock(TILEWIDTH,TILEWIDTH,1); dim3 dimGrid(ceil(n/(float)TILEWIDTH),ceil(n/(float)TILEWIDTH),1); vecMulMatrixKernel<<<dimGrid,dimBlock >>>(d_A,d_B,d_C,n); //copy C from devce memory cudaMemcpy(C, d_C, size, cudaMemcpyDeviceToHost); //free device memories cudaFree(d_A); cudaFree(d_B); cudaFree(d_C); } //Kernal function that runs in each thread int main(){ int n=10; int i,j; float A[n][n],C[n][n],B[n][n]; for(i=0;i<n;i++){ for(j=0;j<n;j++){ A[i][j]=i+j; B[i][j]=i*j; } } vecMulMatrix(&A[0][0],&B[0][0],&C[0][0],n); for(i=0;i<n;i++){ for(j=0;j<n;j++){ printf("%.3f ",A[i][j]); } printf("\n"); } printf("---\n"); for(i=0;i<n;i++){ for(j=0;j<n;j++){ printf("%.3f ",B[i][j]); } printf("\n"); } printf("---\n"); for(i=0;i<n;i++){ for(j=0;j<n;j++){ printf("%.3f ",C[i][j]); } printf("\n"); } return 0; }
2,311
/******************************************************************************* * PROGRAM: canny_edge_detector * FILE: non_maximal_supp.cu * PURPOSE: Apply non maximal suppression. * NAME: Vuong Pham-Duy * Faculty of Computer Science and Technology * Ho Chi Minh University of Technology, Viet Nam * vuongpd95@gmail.com * DATE: 11/10/2016 *******************************************************************************/ #include "cuda_runtime.h" #include "device_launch_parameters.h" #define _USE_MATH_DEFINES #include <stdio.h> #include <stdlib.h> #include <math.h> #include <string.h> #define NOEDGE 255 #define POSSIBLE_EDGE 128 #define EDGE 0 #define VERBOSE 1 __global__ void non_max_supp_kernel(int rows, int cols, int blockSize, short int *d_magnitude, short int *d_delta_x, short int *d_delta_y, unsigned char *d_nms); #define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); } void gpuAssert(cudaError_t code, const char *file, int line, bool abort = true); /******************************************************************************* * PROCEDURE: non_maximal_supp * PURPOSE: perform non-maximal suppression * NAME: Vuong Pham-duy * DATE: 10/11/2016 *******************************************************************************/ void non_maximal_supp(int rows, int cols, int blockSize, int gridSize, short int **d_magnitude, short int **d_delta_x, short int **d_delta_y, unsigned char **d_nms) { /**************************************************************************** * Perform non-maximal suppression. ****************************************************************************/ if (VERBOSE) printf("Doing the non-maximal suppression.\n"); gpuErrchk(cudaMemset((*d_nms), 0, rows * cols * sizeof(unsigned char))); non_max_supp_kernel<<<gridSize, blockSize>>>(rows, cols, blockSize, (*d_magnitude), (*d_delta_x), (*d_delta_y), (*d_nms)); } /******************************************************************************* * PROCEDURE: non_max_supp_kernel * PURPOSE: This routine applies non-maximal suppression to the magnitude of * every pixel of the gradient image * NAME: Vuong Pham-duy * DATE: 10/11/2016 *******************************************************************************/ __global__ void non_max_supp_kernel(int rows, int cols, int blockSize, short int *d_magnitude, short int *d_delta_x, short int *d_delta_y, unsigned char *d_nms) { /* This thread process the number img_idx element of image */ int img_idx = blockIdx.x * blockSize + threadIdx.x; if (img_idx >= (rows * cols)) return; int r = img_idx / cols; /* row position of the pixel, range [0, rows - 1] */ int c = img_idx - r * cols; /* col position of the pixel, range [0, cols - 1] */ if ((r != rows - 1) && (r != 0) && (c != 0) && (c != cols - 1)) { short m00; m00 = d_magnitude[img_idx]; if (m00 == 0) d_nms[img_idx] = (unsigned char)NOEDGE; else { short gx, gy, z1, z2; float mag1, mag2, xperp, yperp; gx = d_delta_x[img_idx]; gy = d_delta_y[img_idx]; xperp = -gx / ((float)m00); yperp = gy / ((float)m00); if (gx >= 0){ if (gy >= 0){ if (gx >= gy) { /* 111 */ /* Left point */ z1 = d_magnitude[img_idx - 1]; z2 = d_magnitude[img_idx - cols - 1]; mag1 = (m00 - z1)*xperp + (z2 - z1)*yperp; /* Right point */ z1 = d_magnitude[img_idx + 1]; z2 = d_magnitude[img_idx + cols + 1]; mag2 = (m00 - z1)*xperp + (z2 - z1)*yperp; } else { /* 110 */ /* Left point */ z1 = d_magnitude[img_idx - cols]; z2 = d_magnitude[img_idx - cols - 1]; mag1 = (z1 - z2)*xperp + (z1 - m00)*yperp; /* Right point */ z1 = d_magnitude[img_idx + cols]; z2 = d_magnitude[img_idx + cols + 1]; mag2 = (z1 - z2)*xperp + (z1 - m00)*yperp; } } else { if (gx >= -gy) { /* 101 */ /* Left point */ z1 = d_magnitude[img_idx - 1]; z2 = d_magnitude[img_idx + cols - 1]; mag1 = (m00 - z1)*xperp + (z1 - z2)*yperp; /* Right point */ z1 = d_magnitude[img_idx + 1]; z2 = d_magnitude[img_idx - cols + 1]; mag2 = (m00 - z1)*xperp + (z1 - z2)*yperp; } else { /* 100 */ /* Left point */ z1 = d_magnitude[img_idx + cols]; z2 = d_magnitude[img_idx + cols - 1]; mag1 = (z1 - z2)*xperp + (m00 - z1)*yperp; /* Right point */ z1 = d_magnitude[img_idx - cols]; z2 = d_magnitude[img_idx - cols + 1]; mag2 = (z1 - z2)*xperp + (m00 - z1)*yperp; } } } else { if (gy >= 0) { if (-gx >= gy) { /* 011 */ /* Left point */ z1 = d_magnitude[img_idx + 1]; z2 = d_magnitude[img_idx - cols + 1]; mag1 = (z1 - m00)*xperp + (z2 - z1)*yperp; /* Right point */ z1 = d_magnitude[img_idx - 1]; z2 = d_magnitude[img_idx + cols - 1]; mag2 = (z1 - m00)*xperp + (z2 - z1)*yperp; } else { /* 010 */ /* Left point */ z1 = d_magnitude[img_idx - cols]; z2 = d_magnitude[img_idx - cols + 1]; mag1 = (z2 - z1)*xperp + (z1 - m00)*yperp; /* Right point */ z1 = d_magnitude[img_idx + cols]; z2 = d_magnitude[img_idx + cols - 1]; mag2 = (z2 - z1)*xperp + (z1 - m00)*yperp; } } else { if (-gx > -gy) { /* 001 */ /* Left point */ z1 = d_magnitude[img_idx + 1]; z2 = d_magnitude[img_idx + cols + 1]; mag1 = (z1 - m00)*xperp + (z1 - z2)*yperp; /* Right point */ z1 = d_magnitude[img_idx - 1]; z2 = d_magnitude[img_idx - cols - 1]; mag2 = (z1 - m00)*xperp + (z1 - z2)*yperp; } else { /* 000 */ /* Left point */ z1 = d_magnitude[img_idx + cols]; z2 = d_magnitude[img_idx + cols + 1]; mag1 = (z2 - z1)*xperp + (m00 - z1)*yperp; /* Right point */ z1 = d_magnitude[img_idx - cols]; z2 = d_magnitude[img_idx - cols - 1]; mag2 = (z2 - z1)*xperp + (m00 - z1)*yperp; } } } /* Now determine if the current point is a maximum point */ if ((mag1 > 0.0) || (mag2 > 0.0)) { d_nms[img_idx] = (unsigned char)NOEDGE; } else { if (mag2 == 0.0) d_nms[img_idx] = (unsigned char)NOEDGE; else d_nms[img_idx] = (unsigned char)POSSIBLE_EDGE; } } } }
2,312
#include <cuda.h> #include <thrust/device_vector.h> #include <thrust/inner_product.h> #include <iostream> #include <stdio.h> //NOTE: COMPILE WITH -arch=sm_20 #define CUDA_CHECK {cudaThreadSynchronize(); \ cudaError_t err = cudaGetLastError();\ if(err){\ std::cout << "Error: " << cudaGetErrorString(err) << " line " << __LINE__ << std::endl; \ exit(1);\ }} double calcDotProductThrust(double* x, double* y, int N){ thrust::device_ptr<double> xThStart(x); thrust::device_ptr<double> yThStart(y); thrust::device_ptr<double> xThEnd(x + N); thrust::device_ptr<double> yThEnd(y + N); return thrust::inner_product(xThStart, xThEnd, yThStart, 0.0); } double timeDotProduct(double (*kernel)(double*, double*, int), double *x, double *y, int N, double ans) { CUDA_CHECK; cudaEvent_t start, end; cudaEventCreate(&start); cudaEventCreate(&end); CUDA_CHECK; cudaEventRecord(start, 0); double prod = kernel(x, y, N); cudaEventRecord(end, 0); CUDA_CHECK; cudaEventSynchronize(end); CUDA_CHECK; cudaError_t err = cudaGetLastError(); if(err){ std::cout << "Error: " << cudaGetErrorString(err) << std::endl; } if( fabs(prod - ans) / fabs(ans) > 1e-4 ) { std::cout << "Multiplication wrong! " << prod << " != " << ans << std::endl; } else { std::cout << "Multiplication right " << prod << " = " << ans << std::endl; } float timeInMs; cudaEventElapsedTime(&timeInMs, start, end); std::cout << "Time: " << timeInMs << "ms" << std::endl << std::endl; CUDA_CHECK; cudaEventDestroy(start); cudaEventDestroy(end); CUDA_CHECK; return prod; } //int main(void) double inner_product(int N, std::vector<double>& first_vector, std::vector<double>& second_vector) { //const int N = 20000000; /* float *x_host = new float[N]; float *y_host = new float[N]; */ double *x_host = new double[N]; double *y_host = new double[N]; std::copy(first_vector.begin(), first_vector.end(), x_host); std::copy(second_vector.begin(), second_vector.end(), y_host); // Fill matrix and vector on host /*for(int i=0 ; i < N ; i++) { //x_host[i]=first_vector[i]; //y_host[i]=second_vector[i]; x_host[i] = sin(i*0.013); y_host[i] = cos(i*0.019); }*/ /* float *x; float *y; */ double *x; double *y; cudaMalloc(&x, N*sizeof(double)); cudaMalloc(&y, N*sizeof(double)); CUDA_CHECK; // Copy x and y to device cudaMemcpy(x, x_host, N*sizeof(double), cudaMemcpyHostToDevice); cudaMemcpy(y, y_host, N*sizeof(double), cudaMemcpyHostToDevice); CUDA_CHECK; //cudaMemcpy(y_host, y, N*sizeof(float), cudaMemcpyDeviceToHost); // Check result /* clock_t st = clock(); double prod = 0; for(int i=0 ; i < N ; i++) { prod += y_host[i] * x_host[i]; } clock_t end = clock(); */ // std::cout << "CPU time = " << (end - st) / (float)CLOCKS_PER_SEC * 1000 << " ms" << std::endl; /* std::cout << "Naive approach - wrong" << std::endl; timeDotProduct(calcDotProduct1, x, y, N, prod); std::cout << "Using atomic operations" << std::endl; timeDotProduct(calcDotProduct2, x, y, N, prod); std::cout << "Reduction across one thread block only" << std::endl; timeDotProduct(calcDotProduct3, x, y, N, prod); std::cout << "Repeated reduction" << std::endl; timeDotProduct(calcDotProduct3Reduce, x, y, N, prod); */ //std::cout << "Thrust" << std::endl; //double dot = timeDotProduct(calcDotProductThrust, x, y, N, prod); double dot = calcDotProductThrust(x, y, N); cudaFree(x); cudaFree(y); delete[] x_host; delete[] y_host; return dot; }
2,313
#include "includes.h" __device__ int translate_idx(int ii, int d1, int d2, int d3, int scale_factor) { int x, y, z, w; w = ii % d3; ii = ii/d3; z = ii % d2; ii = ii/d2; y = ii % d1; ii = ii/d1; x = ii; w = w/scale_factor; z = z/scale_factor; d2 /= scale_factor; d3 /= scale_factor; return (((x*d1+y)*d2)+z)*d3+w; } __global__ void upscale(float *input, float *output, long no_elements, int scale_factor, int d1, int d2, int d3) { // output offset: long ii = threadIdx.x + blockDim.x * blockIdx.x; ii += threadIdx.y + blockDim.y * (blockDim.x * gridDim.x) * blockIdx.y; if (ii >= no_elements) return; int ipidx = translate_idx(ii, d1, d2, d3, scale_factor); output[ii]=input[ipidx]; }
2,314
#include "includes.h" __global__ void convolutionLayers3DKernel( float *d_Dst, float *d_Src, int imageW, int imageH, int imageD, int kernel_index, int kernel_radius ) { __shared__ float s_Data[LAYERS_BLOCKDIM_X][LAYERS_BLOCKDIM_Y][(LAYERS_RESULT_STEPS + 2 * LAYERS_HALO_STEPS) * LAYERS_BLOCKDIM_Z + 1]; //Offset to the upper halo edge const int baseX = blockIdx.x * LAYERS_BLOCKDIM_X + threadIdx.x; const int baseY = blockIdx.y * LAYERS_BLOCKDIM_Y + threadIdx.y; const int baseZ = (blockIdx.z * LAYERS_RESULT_STEPS - LAYERS_HALO_STEPS) * LAYERS_BLOCKDIM_Z + threadIdx.z; d_Src += (baseZ * imageH + baseY) * imageW + baseX; d_Dst += (baseZ * imageH + baseY) * imageW + baseX; const int pitch = imageW*imageH; const float* kernel = &c_Kernel[kernel_index*MAX_KERNEL_LENGTH]; //Main data #pragma unroll for (int i = LAYERS_HALO_STEPS; i < LAYERS_HALO_STEPS + LAYERS_RESULT_STEPS; i++) { s_Data[threadIdx.x][threadIdx.y][threadIdx.z + i * LAYERS_BLOCKDIM_Z] = d_Src[i * LAYERS_BLOCKDIM_Z * pitch]; } //Upper halo #pragma unroll for (int i = 0; i < LAYERS_HALO_STEPS; i++) { s_Data[threadIdx.x][threadIdx.y][threadIdx.z + i * LAYERS_BLOCKDIM_Z] = (baseZ + i * LAYERS_BLOCKDIM_Z >= 0) ? d_Src[i * LAYERS_BLOCKDIM_Z * pitch] : 0; } //Lower halo #pragma unroll for (int i = LAYERS_HALO_STEPS + LAYERS_RESULT_STEPS; i < LAYERS_HALO_STEPS + LAYERS_RESULT_STEPS + LAYERS_HALO_STEPS; i++) { s_Data[threadIdx.x][threadIdx.y][threadIdx.z + i * LAYERS_BLOCKDIM_Z]= (baseZ + i * LAYERS_BLOCKDIM_Z < imageD) ? d_Src[i * LAYERS_BLOCKDIM_Z * pitch] : 0; } //Compute and store results __syncthreads(); #pragma unroll for (int i = LAYERS_HALO_STEPS; i < LAYERS_HALO_STEPS + LAYERS_RESULT_STEPS; i++) { float sum = 0; //#pragma unroll for (int j = -kernel_radius; j <= kernel_radius; j++) { sum += kernel[kernel_radius - j] * s_Data[threadIdx.x][threadIdx.y][threadIdx.z + i * LAYERS_BLOCKDIM_Z + j]; } d_Dst[i * LAYERS_BLOCKDIM_Z * pitch] = sum; } }
2,315
#include <stdio.h> #include <stdlib.h> #include <math.h> #include <iostream> void mean_filter_h(int **img, int **res, int N, int M, int k) { int count; float temp; for(int n = 0; n < N; n++) { for(int m = 0; m < M; m++) { count = 0; temp = 0.0; for(int i = N - k; i <= N + k; i++) { for(int j = M - k; j <= M + k; j++) { if(i >= 0 && i < N && j >= 0 && j < M) { count = count + 1; temp = res[n][m] + img[i][j]; } } temp = temp / count; res[n][m] = (int)temp; } } } } int main() { const int N = 4; const int M = 4; int k = 3; int *a[N], *b[N]; for(int i = 0; i < N; i++ ) { a[i] = (int *)malloc(M * sizeof(int)); b[i] = (int *)malloc(M * sizeof(int)); } for(int i = 0; i < N; i++ ) { for(int j = 0; j < M; j++ ) { a[i][j] = 1; b[i][j] = 0; } } //clock_t start_h = clock(); mean_filter_h(a, b, N, M, k / 2); //clock_t end_h = clock(); //double time_h = (double)(end_h-start_h)/CLOCKS_PER_SEC; //printf("CPU Time: %f\n", time_h); for(int i = 0; i < N; i++ ) { for(int j = 0; j < M; j++ ) { printf("%d ", b[i][j]); } } return 0; }
2,316
//#include <algorithm> //#include <vector> // //#include "caffe/layers/clusters_triplet_loss_layer.hpp" //#include "caffe/util/math_functions.hpp" // //namespace caffe { // // template <typename Dtype> // __global__ void ClustersTripletForward(const int nthreads, const int batch_size, // const int dim, const Dtype margin, const Dtype* pos_data,const Dtype* neg_data, // Dtype* vec_loss) { // // CUDA_KERNEL_LOOP(index, nthreads) { // const int ind_a = index / (batch_size * batch_size); // const int ind_p = (index % (batch_size * batch_size)) / batch_size; // const int ind_n = index % batch_size; // // Dtype dpa(0.0), dna(0.0), t; // for (int i = 0; i < dim; i++) // { // t = pos_data[ind_p * dim + i] - pos_data[ind_a * dim + i]; // dpa += t * t; // t = neg_data[ind_n * dim + i] - pos_data[ind_a * dim + i]; // dna += t * t; // } // vec_loss[index] = max(Dtype(0), dpa + margin - dna); // } // } // // template <typename Dtype> // __global__ void ClustersTripletBackward(const int nthreads, const int batch_size, // const int dim, const Dtype* pos_data, const Dtype* neg_data, const Dtype* vec_loss, // Dtype* pos_diff, Dtype* neg_diff) { // // CUDA_KERNEL_LOOP(index, nthreads) { // const int item = index / dim; // const int k = index % dim; // // Dtype diff; // for (int i = 0; i < batch_size; i++) // { // for (int j = 0; j < batch_size; j++) // { // if (vec_loss[(item * batch_size + i) * batch_size + j] > 0) // { // pos_diff[item * dim + k] += neg_data[j * dim + k] - pos_data[i * dim + k]; // } // // if (vec_loss[(i * batch_size + item) * batch_size + j] > 0) // { // pos_diff[item * dim + k] += pos_data[item * dim + k] - pos_data[i * dim + k]; // } // // if (vec_loss[(i * batch_size + j) * batch_size + item] > 0) // { // neg_diff[item * dim + k] += pos_data[i * dim + k] - neg_data[item * dim + k]; // } // } // } // } // } // // template <typename Dtype> // void ClustersTripletLossLayer<Dtype>::Forward_gpu( // const vector<Blob<Dtype>*>& bottom, // const vector<Blob<Dtype>*>& top) { // // //ClustersTripletLossLayer<Dtype>::Forward_cpu(bottom, top); // // Dtype margin = this->layer_param_.contrastive_loss_param().margin(); // int batch_size = bottom[0]->num() / 2; // int dim = bottom[0]->channels(); // const Dtype* pos_data = bottom[0]->gpu_data(); // const Dtype* neg_data = bottom[0]->gpu_data() + batch_size * dim; // Dtype* vec_loss = vec_loss_.mutable_gpu_data(); // Dtype nthreads = batch_size * batch_size * batch_size; // caffe_gpu_set(vec_loss_.count(), Dtype(0), vec_loss); // ClustersTripletForward<Dtype> << <CAFFE_GET_BLOCKS(nthreads), CAFFE_CUDA_NUM_THREADS >> >( // nthreads, batch_size, dim, margin, pos_data, neg_data, vec_loss); // // Dtype loss(0.0); // for (int i = 0; i < vec_loss_.count(); i++) // { // loss += vec_loss_.cpu_data()[i]; // } // top[0]->mutable_cpu_data()[0] = loss / batch_size / 2; // } // // template <typename Dtype> // void ClustersTripletLossLayer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top, // const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) { // // //ClustersTripletLossLayer<Dtype>::Backward_cpu(top, propagate_down, bottom); // // int batch_size = bottom[0]->num() / 2; // int dim = bottom[0]->channels(); // const Dtype* pos_data = bottom[0]->gpu_data(); // const Dtype* neg_data = bottom[0]->gpu_data() + batch_size * dim; // const Dtype* vec_loss = vec_loss_.gpu_data(); // Dtype* pos_diff = bottom[0]->mutable_gpu_diff(); // Dtype* neg_diff = bottom[0]->mutable_gpu_diff() + batch_size * dim; // Dtype nthreads = batch_size * dim; // // if (propagate_down[0]) { // caffe_gpu_set(bottom[0]->count(), Dtype(0), bottom[0]->mutable_gpu_diff()); // ClustersTripletBackward<Dtype> << <CAFFE_GET_BLOCKS(nthreads), CAFFE_CUDA_NUM_THREADS >> >( // nthreads, batch_size, dim, pos_data, neg_data, vec_loss, pos_diff, neg_diff); // const Dtype alpha = top[0]->cpu_diff()[0] / batch_size; // caffe_gpu_scal(bottom[0]->count(), alpha, bottom[0]->mutable_gpu_diff()); // } // // } // //INSTANTIATE_LAYER_GPU_FUNCS(ClustersTripletLossLayer); // //} // namespace caffe
2,317
__device__ float magnitude(const float *r) { return pow((r[0] * r[0]) + (r[1] * r[1]) + (r[2] * r[2]), 0.5); } __device__ float magnitude_withId(const float *r) { const int i = blockIdx.x; return pow((r[3 * i] * r[3 * i]) + (r[3 * i + 1] * r[3 * i + 1]) + (r[3 * i + 2] * r[3 * i + 2]), 0.5); } __device__ float clamp(float x, float max_v, float min_v) { return max(min_v, min(max_v, x)); } __device__ float WPoly6(const float mag, const float *h, const float *WPoly6_const) { if (mag < *h && mag > 0) { float inner_val = ((*h * *h) - (mag * mag)); return inner_val * inner_val * inner_val * *WPoly6_const; } else return 0; } __device__ void grad_WPoly6(float *grad, float *r, const float mag, const float *h, const float *grad_WPoly6_const) { float inner_val = ((*h * *h) - (mag * mag)); inner_val = inner_val * inner_val * *grad_WPoly6_const * (1 / (mag + 0.000001f)); grad[0] = inner_val * r[0]; grad[1] = inner_val * r[1]; grad[2] = inner_val * r[2]; } __device__ float lap_WPoly6(const float *h, const float mag, const float *lap_WPoly6_const) { float inner_val = ((*h * *h) - (mag * mag)); return *lap_WPoly6_const * inner_val * ((3 * *h * *h) - (7 * mag * mag)); } __device__ float Wspiky(const float *h, const float mag, const float *Wpiky_const) { if (mag < *h && mag > 0) { return *Wpiky_const * (*h - mag) * (*h - mag) * (*h - mag); } else { return 0; } } __device__ void grad_WSpiky(float *grad, const float mag, const float *r, const float *h, const float *grad_Wspiky_const) { float inner_val = *grad_Wspiky_const * (*h - mag) * (*h - mag) * (1 / (mag + 0.0000001f)); grad[0] = r[0] * inner_val; grad[1] = r[1] * inner_val; grad[2] = r[2] * inner_val; } __device__ float Wviscosity(const float *h, const float mag, const float *Wviscosity_const) { float inner_val = ((-mag * mag * mag) / (2 * *h * *h * *h)) + ((mag * mag) / (*h * *h)) + (*h / (2 * mag + 0.0000001f)) - 1; return inner_val * *Wviscosity_const; } __device__ float lap_Wviscosity(const float mag, const float *h, const float *lap_Wviscosity_const) { return *lap_Wviscosity_const * (*h - mag); } __device__ void pressure_force(float *press_force, const float *mass, const float mag, const float *r, const float *h, const float *density_n, const int *neighbors, const float *rest_density, const float *k, const float *grad_Wspiky_const) { const int i = neighbors[blockIdx.x * blockDim.x + threadIdx.x]; float grad[3]; grad_WSpiky(grad, mag, r, h, grad_Wspiky_const); float const_val = *mass * *k * (density_n[blockIdx.x] + density_n[i] - 2 * *rest_density) * 1 / (2 * density_n[i]); press_force[0] = const_val * grad[0]; press_force[1] = const_val * grad[1]; press_force[2] = const_val * grad[2]; } __device__ void viscosity_force(float *visc_force, const float *eta, const float *mass, const float mag, const float *h, const float *density_n, const float *vel_n, const int *neighbors, const float *lap_Wviscosity_const) { const int i = neighbors[blockIdx.x * blockDim.x + threadIdx.x]; float const_val = *eta * *mass * (1 / (density_n[i] + 0.0000001f)) * lap_Wviscosity(mag, h, lap_Wviscosity_const); visc_force[0] = const_val * (vel_n[3 * i] - vel_n[3 * blockIdx.x]); visc_force[1] = const_val * (vel_n[3 * i + 1] - vel_n[3 * blockIdx.x + 1]); visc_force[2] = const_val * (vel_n[3 * i + 2] - vel_n[3 * blockIdx.x + 2]); } __device__ void color_field_grad(float *color_field_grad_val, const float *mass, const float *density_n, float *r_dash, const int *neighbors, const float *h, const float mag, const float *grad_WPoly6_const) { const int i = neighbors[blockIdx.x * blockDim.x + threadIdx.x]; float inner_val = *mass * 1 / (density_n[i] + 0.0000001f); float grad_Wpoly[3]; grad_WPoly6(grad_Wpoly, r_dash, mag, h, grad_WPoly6_const); color_field_grad_val[blockIdx.x * blockDim.x * 3 + 3 * threadIdx.x] = grad_Wpoly[0] * inner_val; color_field_grad_val[blockIdx.x * blockDim.x * 3 + 3 * threadIdx.x + 1] = grad_Wpoly[1] * inner_val; color_field_grad_val[blockIdx.x * blockDim.x * 3 + 3 * threadIdx.x + 2] = grad_Wpoly[2] * inner_val; } __device__ void color_field_lap(float *color_field_lap_val, const float *mass, const float *density_n, const float *r, const int *neighbors, const float *h, const float mag, const float *lap_WPoly6_const) { const int i = blockIdx.x * blockDim.x + threadIdx.x; color_field_lap_val[i] = *mass * 1 / (density_n[neighbors[i]] + 0.0000001f) * lap_WPoly6(h, mag, lap_WPoly6_const); } __global__ void calc_density(float *density, const float *r1, const int *neighbors, const float *mass, const float *h, const float *WPoly6_const) { float r_dash[3]; const int i = threadIdx.x; const int j = blockIdx.x; const int n = j * blockDim.x + i; r_dash[0] = r1[3 * j] - r1[3 * neighbors[n]]; r_dash[1] = r1[3 * j + 1] - r1[3 * neighbors[n] + 1]; r_dash[2] = r1[3 * j + 2] - r1[3 * neighbors[n] + 2]; const float mag = magnitude(r_dash); density[n] = WPoly6(mag, h, WPoly6_const) * *mass; } __global__ void calc_forces(float *force, float *color_field_lap_val, float *color_field_grad_val, const float *r1, const float *density_n, float *vel_n, int *neighbors, const float *h, const float *eta, const float *mass, const float *rest_density, const float *k, const float *grad_WPoly6_const, const float *lap_WPoly6_const, const float *Wspiky_const, const float *grad_Wspiky_const, const float *Wviscosity_const, const float *lap_Wviscosity_const) { const int i = threadIdx.x; const int j = blockIdx.x; const int n = j * blockDim.x + i; float r_dash[3], press_force[3], visc_force[3]; r_dash[0] = r1[3 * j] - r1[3 * neighbors[n]]; r_dash[1] = r1[3 * j + 1] - r1[3 * neighbors[n] + 1]; r_dash[2] = r1[3 * j + 1] - r1[3 * neighbors[n] + 2]; float mag = magnitude(r_dash); if (mag > 0.000001f && mag < *h) { pressure_force(press_force, mass, mag, r_dash, h, density_n, neighbors, rest_density, k, grad_Wspiky_const); viscosity_force(visc_force, eta, mass, mag, h, density_n, vel_n, neighbors, lap_Wviscosity_const); color_field_grad(color_field_grad_val, mass, density_n, r_dash, neighbors, h, mag, grad_WPoly6_const); color_field_lap(color_field_lap_val, mass, density_n, r_dash, neighbors, h, mag, lap_WPoly6_const); force[blockIdx.x * blockDim.x * 3 + 3 * i] = clamp(press_force[0] + visc_force[0], 200, -200); force[blockIdx.x * blockDim.x * 3 + 3 * i + 1] = clamp(press_force[1] + visc_force[1], 200, -200); force[blockIdx.x * blockDim.x * 3 + 3 * i + 2] = clamp(press_force[2] + visc_force[2], 200, -200); } else { force[blockIdx.x * blockDim.x * 3 + 3 * i] = 0; force[blockIdx.x * blockDim.x * 3 + 3 * i + 1] = 0; force[blockIdx.x * blockDim.x * 3 + 3 * i + 2] = 0; color_field_grad_val[blockIdx.x * blockDim.x * 3 + 3 * i] = 0; color_field_grad_val[blockIdx.x * blockDim.x * 3 + 3 * i + 1] = 0; color_field_grad_val[blockIdx.x * blockDim.x * 3 + 3 * i + 2] = 0; color_field_lap_val[blockIdx.x * blockDim.x + threadIdx.x] = 0; } } __global__ void update_pos(float *r, float *vel_p, float *force, const float *threshold, const float *mass, const float *time, const float *sigma, const float *Width, const float *damping, const float *eps, const float *color_field_lap_val, const float *color_field_grad_val) { const int i = blockIdx.x; float gradient_length = magnitude_withId(color_field_grad_val); float force_surface[3] = {0, 0, 0}; if (gradient_length >= *threshold) { float const_val = -1 * *sigma * color_field_lap_val[i] * (1 / (gradient_length + 0.0000001f)); force_surface[0] = const_val * color_field_grad_val[3 * i]; force_surface[1] = const_val * color_field_grad_val[3 * i + 1]; force_surface[2] = const_val * color_field_grad_val[3 * i + 2]; } vel_p[3 * i] += clamp( (force[3 * i] + force_surface[0]) * (1 / (*mass)) * *time, 200, -200); vel_p[3 * i + 1] += clamp( (force[3 * i + 1] + force_surface[1]) * (1 / (*mass)) * *time, 200, -200); vel_p[3 * i + 2] += clamp( (((force[3 * i + 2] + force_surface[2]) * 1 / (*mass)) - 10 * 12000) * *time, 200, -200); r[3 * i] += vel_p[3 * i] * *time; r[3 * i + 1] += vel_p[3 * i + 1] * *time; r[3 * i + 2] += vel_p[3 * i + 2] * *time; if (r[3 * i] < -*Width) { r[3 * i] = -*Width + *eps; vel_p[3 * i] = *damping * vel_p[3 * i]; } if (r[3 * i + 1] < -*Width) { r[3 * i + 1] = -*Width + *eps; vel_p[3 * i + 1] = *damping * vel_p[3 * i + 1]; } if (r[3 * i + 2] < -*Width) { r[3 * i + 2] = -*Width + *eps; vel_p[3 * i + 2] = *damping * vel_p[3 * i + 2]; } if (r[3 * i] > *Width) { r[3 * i] = *Width - *eps; vel_p[3 * i] = *damping * vel_p[3 * i]; } if (r[3 * i + 1] > *Width) { r[3 * i + 1] = *Width - *eps; vel_p[3 * i + 1] = *damping * vel_p[3 * i + 1]; } if (r[3 * i + 2] > *Width) { r[3 * i + 2] = *Width * 10 - *eps; vel_p[3 * i + 2] = *damping * vel_p[3 * i + 2]; } }
2,318
// #include <algorithm> // #include <vector> // #include "omp.h" // #include <iostream> // using namespace std; // #include "caffe/layers/set_loss2_layer.hpp" // #include "caffe/util/math_functions.hpp" // #include "caffe/util/io.hpp" // namespace caffe // { // template <typename Dtype> // void SetLoss2Layer<Dtype>::Forward_gpu( // const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top) // { // const Dtype* label = bottom[1]->cpu_data(); // const Dtype* data = bottom[0]->cpu_data(); // const Dtype* data_gpu = bottom[0]->gpu_data(); // // initialization of cat_per_iter_, im_per_cat_ and cat_mean_. // if ( im_per_cat_ < 0 ) // { // int j; // for ( j = 0; j < batch_size_; ++j) // { // if (label[0] != label[j]) // { // break; // } // } // im_per_cat_ = j; // cat_per_iter_ = batch_size_ / j; // temp_loss_.Reshape(1, 1, cat_per_iter_, code_length_); // cat_mean_.Reshape(1, 1, cat_per_iter_, code_length_); // cat_grad_.Reshape(1, 1, cat_per_iter_, code_length_); // } // // compute cat_mean_ // memset(cat_mean_.mutable_cpu_data(), 0, sizeof(Dtype) * code_length_ * cat_per_iter_); // for ( int i = 0; i < batch_size_; i += im_per_cat_) // { // for (int j = 0; j < im_per_cat_; ++j) // { // caffe_gpu_axpby<Dtype>(code_length_, 1. / im_per_cat_, data_gpu + (i + j) * code_length_, 1, cat_mean_.mutable_gpu_data() + int(i / im_per_cat_) * code_length_); // } // } // // compute loss and cat grad // memset(cat_grad_.mutable_cpu_data(), 0, sizeof(Dtype) * code_length_ * cat_per_iter_); // float loss(0);// = new float* [cat_per_iter_]; // for ( int i = 0; i < cat_per_iter_; ++i) // { // // loss[i] = new float [cat_per_iter_]; // for (int j = i + 1; j < cat_per_iter_; ++j) // { // caffe_gpu_sub<Dtype>(code_length_, cat_mean_.gpu_data() + i * code_length_, cat_mean_.gpu_data() + j * code_length_, temp_loss_.mutable_gpu_data()+i*code_length_); // Dtype sub_loss; // caffe_gpu_dot<Dtype>(code_length_, temp_loss_.gpu_data()+i*code_length_, temp_loss_.gpu_data()+i*code_length_, &sub_loss); // sub_loss = alpha_ - 0.25 * sub_loss; // if (sub_loss > 0) // { // loss += sub_loss; // caffe_gpu_axpby<Dtype>(code_length_, 1., temp_loss_.gpu_data()+i*code_length_, 1., cat_grad_.mutable_gpu_data() + i * code_length_); // caffe_gpu_axpby<Dtype>(code_length_, -1., temp_loss_.gpu_data()+i*code_length_, 1., cat_grad_.mutable_gpu_data() + j * code_length_); // } // // else // // { // // loss[i][j] = 0; // // } // } // } // // float loss_sum(0.); // // for (int i = 0; i < cat_per_iter_; ++i) // // { // // for (int j = i + 1; j < cat_per_iter_; ++j) // // { // // loss_sum += loss[i][j]; // // } // // } // top[0]->mutable_cpu_data()[0] = 2 * loss / cat_per_iter_ / (cat_per_iter_ - 1); // } // template <typename Dtype> // void SetLoss2Layer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top, // const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) // { // if (propagate_down[1]) // { // LOG(FATAL) << this->type() << " Layer cannot backpropagate to label inputs. "; // } // Dtype* diff = bottom[0]->mutable_gpu_diff(); // loss_weight_ = top[0]->cpu_diff()[0]; // caffe_gpu_scal<Dtype>(code_length_ * cat_per_iter_, - loss_weight_ / cat_per_iter_ / (cat_per_iter_ - 1) / im_per_cat_, cat_grad_.mutable_gpu_data()); // // copy gradient to each sample // for (int i = 0; i < batch_size_; ++i) // { // cudaMemcpy(diff + i * code_length_, cat_grad_.gpu_data() + int(i / im_per_cat_) * code_length_, sizeof(Dtype) * code_length_, cudaMemcpyDefault); // } // } // INSTANTIATE_LAYER_GPU_FUNCS(SetLoss2Layer); // }
2,319
#include "includes.h" __device__ void find_index(short *vec, const int vec_length, int *value, int *index) { for (int i = threadIdx.x; i < vec_length; i = i + blockDim.x) { if (vec[i] == *value) { atomicMax(index, i); } } } __global__ void find_index(int *vec, const int vec_length, int *value, int *index){ for (int i = threadIdx.x; i < vec_length; i = i + blockDim.x) { if(vec[i]==*value){ atomicMax(index, i); } } }
2,320
#include <cuda.h> #include <stdio.h> #include <stdlib.h> /* __global__ static inline int mandel(float c_re, float c_im, int count) { float z_re = c_re, z_im = c_im; int i; for (i = 0; i < count; ++i) { if (z_re * z_re + z_im * z_im > 4.f) break; float new_re = z_re * z_re - z_im * z_im; float new_im = 2.f * z_re * z_im; z_re = c_re + new_re; z_im = c_im + new_im; } return i; } */ __global__ void mandelKernel( int* d_data,float stepX, float stepY,float lowerX,float lowerY ,int width, int maxIterations, size_t pitch) { // To avoid error caused by the floating number, use the following pseudo code // int thisX= blockIdx.x * blockDim.x + threadIdx.x; int thisY= blockIdx.y * blockDim.y + threadIdx.y; float x = lowerX + thisX * stepX; float y = lowerY + thisY * stepY; float c_re = x, c_im = y; float z_re = c_re, z_im = c_im; int i; for (i = 0; i < maxIterations; ++i) { if (z_re * z_re + z_im * z_im > 4.f) break; float new_re = z_re * z_re - z_im * z_im; float new_im = 2.f * z_re * z_im; z_re = c_re + new_re; z_im = c_im + new_im; } int *ptr = (int *)((char*)d_data+thisY*pitch); ptr[thisX] = i; //d_data[ thisX + thisY * width ] = i; } // Host front-end function that allocates the memory and launches the GPU kernel void hostFE (float upperX, float upperY, float lowerX, float lowerY, int* img, int resX, int resY, int maxIterations) { float stepX = (upperX - lowerX) / resX; float stepY = (upperY - lowerY) / resY; int *data, *d_data; dim3 threadPerBlock(25,25); dim3 numBlocks(resX/threadPerBlock.x,resY/threadPerBlock.y); size_t pitch; //data = (int*)malloc( sizeof(int)*resX*resY ); cudaHostAlloc(&data, sizeof(int) * resX*resY, cudaHostAllocMapped); cudaMallocPitch((void **)&d_data, &pitch, sizeof(int)*resX, resY); // cudaMalloc((void**)&d_data, sizeof(int)*resX*resY ); mandelKernel<<<numBlocks,threadPerBlock>>>(d_data,stepX,stepY,lowerX,lowerY,resX,maxIterations,pitch); cudaMemcpy2D(data, sizeof(int)*resX, d_data, pitch, sizeof(int)*resX, resY, cudaMemcpyDeviceToHost); memcpy(img,data,sizeof(int)*resX*resY); cudaFreeHost(data); cudaFree(d_data); }
2,321
#include<stdio.h> #include<cuda_runtime.h> bool init_cuda(){ int count; cudaGetDeviceCount(&count); if(0 == count) { fprintf(stderr, "There is no device \n"); return false; } int i; for(i = 0; i < count; i++) { cudaDeviceProp prop; if(cudaSuccess == cudaGetDeviceProperties(&prop, i)) { if(prop.major >= 1){ break; } } } if(i == count){ fprintf(stderr, "There is no device supporting CUDA 1.x.\n"); return false; } cudaSetDevice(i); return true; } int main() { if(!init_cuda()) { return 0; } printf("CUDA initialize.\n"); return 0; }
2,322
#include <stdio.h> #include <cuda.h> // CPU code to do matrix ADdition void matrixAdd(int *a, int *b, int *c, int N) { int index; for(int col=0; col<N; col++) { for(int row=0; row<N; row++) { index = row * N + col; c[index] = a[index] + b[index]; } } } // GPU code to do matrix addition __global__ void matrixAddKernel(int *a, int *b, int *c, int N) { int col = threadIdx.x + blockDim.x * blockIdx.x; int row = threadIdx.y + blockDim.y * blockIdx.y; int index = row * N + col; if(col < N && row < N) { c[index] = a[index] + b[index]; } } int main(void) { cudaSetDevice(3); //size of the matrix (the matrix will have NxN elements) int N = 2000; dim3 grid(16,1,1); dim3 block(1024,1,1); // pointers to host memory int *a_h; int *b_h; int *c_h; int *d_h; // pointers to device memory int *a_d; int *b_d; int *c_d; // this variable holds the number of bytes required by arrays int size; // use CUDA events to measure time cudaEvent_t start; cudaEvent_t stop; float elapsedTime; //print out the information about number of blocks and threads printf("Number of threads: %i (%ix%i)\n", block.x*block.y, block.x, block.y); printf("Number of blocks: %i (%ix%i)\n", grid.x*grid.y, grid.x, grid.y); //dynamically alocate host memory and load the arrays with some data size = N * N * sizeof(int); a_h = (int*) malloc(size); b_h = (int*) malloc(size); c_h = (int*) malloc(size); d_h = (int*) malloc(size); for(int i=0; i<N; i++) { for(int j=0; j<N; j++) { a_h[i * N + j] = i; b_h[i * N + j] = i; } } //allocate memory on the device cudaMalloc((void**)&a_d, size); cudaMalloc((void**)&b_d, size); cudaMalloc((void**)&c_d, size); //copy the host memory to the device cudaMemcpy(a_d, a_h, size, cudaMemcpyHostToDevice); cudaMemcpy(b_d, b_h, size, cudaMemcpyHostToDevice); cudaMemcpy(c_d, c_h, size, cudaMemcpyHostToDevice); //start the timers cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start, 0); //launch the kernel matrixAddKernel<<<grid,block>>>(a_d, b_d, c_d, N); //stop the timer and print out the execution time cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&elapsedTime, start, stop); printf("Time to calculate results on GPU: %f ms.\n", elapsedTime); //copy the results to host cudaMemcpy(c_h, c_d, size ,cudaMemcpyDeviceToHost); //time to measure CPU performance cudaEventRecord(start,0); //launch the matrixAdd function that executes on the CPU matrixAdd(a_h, b_h, d_h, N); //strop the timer and print out the results cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&elapsedTime, start, stop ); printf("Time to calculate results on CPU: %f ms.\n", elapsedTime); //check that GPU and CPU results match for(int i=0; i<N*N; i++) { if (c_h[i] != d_h[i]) printf("Error: CPU and GPU results do not match\n"); break; } // clean up free(a_h); free(b_h); free(c_h); free(d_h); cudaFree(a_d); cudaFree(b_d); cudaFree(c_d); cudaEventDestroy(start); cudaEventDestroy(stop); return 0; }
2,323
/* ============================================================================ Filename : implementation.cu Author : Martino Milani / Sébastien Gachoud SCIPER : 286204 / 250083 ============================================================================ */ #include <iostream> #include <iomanip> #include <sys/time.h> #include <cuda_runtime.h> using namespace std; // CPU Baseline void array_process(double *input, double *output, int length, int iterations) { double *temp; for(int n=0; n<(int) iterations; n++) { for(int i=1; i<length-1; i++) { for(int j=1; j<length-1; j++) { output[(i)*(length)+(j)] = (input[(i-1)*(length)+(j-1)] + input[(i-1)*(length)+(j)] + input[(i-1)*(length)+(j+1)] + input[(i)*(length)+(j-1)] + input[(i)*(length)+(j)] + input[(i)*(length)+(j+1)] + input[(i+1)*(length)+(j-1)] + input[(i+1)*(length)+(j)] + input[(i+1)*(length)+(j+1)] ) / 9; } } output[(length/2-1)*length+(length/2-1)] = 1000; output[(length/2)*length+(length/2-1)] = 1000; output[(length/2-1)*length+(length/2)] = 1000; output[(length/2)*length+(length/2)] = 1000; temp = input; input = output; output = temp; } } __global__ void gpu_computation(double* input, double* output, int length); // GPU Optimized function void GPU_array_process(double *input, double *output, int length, int iterations) { //Cuda events for calculating elapsed time cudaEvent_t cpy_H2D_start, cpy_H2D_end, comp_start, comp_end, cpy_D2H_start, cpy_D2H_end; cudaEventCreate(&cpy_H2D_start); cudaEventCreate(&cpy_H2D_end); cudaEventCreate(&cpy_D2H_start); cudaEventCreate(&cpy_D2H_end); cudaEventCreate(&comp_start); cudaEventCreate(&comp_end); /* Preprocessing goes here */ /*----- What I did -----*/ const long SIZE = length * length * sizeof(double); double* gpu_input; double* gpu_output; dim3 threadsPerBlock(32,32); dim3 nbBlocks(length / threadsPerBlock.x + 1, length / threadsPerBlock.y + 1); const long PADDED_SIZE = (nbBlocks.x+1) * threadsPerBlock.x * (nbBlocks.y+1) * threadsPerBlock.y * sizeof(double); //+1 to avoid going out of the input cudaSetDevice(0); if(cudaMalloc((void**)&gpu_input, PADDED_SIZE) != cudaSuccess){ cerr << "Error allocating input" << endl; } if(cudaMalloc((void**)&gpu_output, PADDED_SIZE) != cudaSuccess){ cerr << "Error allocating output" << endl; } /*----------------------*/ cudaEventRecord(cpy_H2D_start); /* Copying array from host to device goes here */ /*----- What I did -----*/ if(cudaMemcpy(gpu_input, input, SIZE, cudaMemcpyHostToDevice) != cudaSuccess){ cerr << "Error copying input to gpu" << endl; } if(cudaMemcpy(gpu_output, output, SIZE, cudaMemcpyHostToDevice) != cudaSuccess){ cerr << "Error copying output to gpu" << endl; } /*----------------------*/ cudaEventRecord(cpy_H2D_end); cudaEventSynchronize(cpy_H2D_end); //Copy array from host to device cudaEventRecord(comp_start); /* GPU calculation goes here */ /*----- What I did -----*/ for(int iter(0); iter < iterations; iter++){ if(iter%2){ gpu_computation <<< nbBlocks, threadsPerBlock >>> (gpu_output, gpu_input, length); } else{ gpu_computation <<< nbBlocks, threadsPerBlock >>> (gpu_input, gpu_output, length); } cudaThreadSynchronize(); } /*----------------------*/ cudaEventRecord(comp_end); cudaEventSynchronize(comp_end); cudaEventRecord(cpy_D2H_start); /* Copying array from device to host goes here */ /*----- What I did -----*/ if(iterations%2==0) { if(cudaMemcpy(output, gpu_input, SIZE, cudaMemcpyDeviceToHost) != cudaSuccess){ cerr << "failed to retrieve gpu_input" << endl; } } else{ if(cudaMemcpy(output, gpu_output, SIZE, cudaMemcpyDeviceToHost) != cudaSuccess){ cerr << "failed to retrieve gpu_output" << endl; } } /*----------------------*/ cudaEventRecord(cpy_D2H_end); cudaEventSynchronize(cpy_D2H_end); /* Postprocessing goes here */ /*----- What I did -----*/ cudaFree(&gpu_input); cudaFree(&gpu_output); /*----------------------*/ float time; cudaEventElapsedTime(&time, cpy_H2D_start, cpy_H2D_end); cout<<"Host to Device MemCpy takes "<<setprecision(4)<<time/1000<<"s"<<endl; cudaEventElapsedTime(&time, comp_start, comp_end); cout<<"Computation takes "<<setprecision(4)<<time/1000<<"s"<<endl; cudaEventElapsedTime(&time, cpy_D2H_start, cpy_D2H_end); cout<<"Device to Host MemCpy takes "<<setprecision(4)<<time/1000<<"s"<<endl; } __global__ void gpu_computation(double* input, double* output, int length){ int x_glob = (blockIdx.x * blockDim.x) + threadIdx.x + 1; //+1 to avoid first column int y_glob = (blockIdx.y * blockDim.y) + threadIdx.y + 1; //+1 to avoid first row int element_id = (y_glob * length) + x_glob; if ( ((x_glob == length/2-1) || (x_glob == length/2)) && ((y_glob == length/2-1) || (y_glob == length/2)) || x_glob >= length - 1 || y_glob >= length-1) { return; } output[element_id] = (input[(y_glob-1)*(length)+(x_glob-1)] + input[(y_glob-1)*(length)+(x_glob)] + input[(y_glob-1)*(length)+(x_glob+1)] + input[(y_glob)*(length)+(x_glob-1)] + input[(y_glob)*(length)+(x_glob)] + input[(y_glob)*(length)+(x_glob+1)] + input[(y_glob+1)*(length)+(x_glob-1)] + input[(y_glob+1)*(length)+(x_glob)] + input[(y_glob+1)*(length)+(x_glob+1)] ) / 9; }
2,324
#include <stdio.h> #define TILE_SIZE 32 #define KERNEL_RADIUS 8 #define KERNEL_LENGTH (2 * KERNEL_RADIUS + 1) __constant__ float c_M[KERNEL_LENGTH][KERNEL_LENGTH]; const int Width = 3072; const int Height = 3072; const int nIter = 300; float * h_Kernel,*h_Input,*h_Output; float * d_Input, *d_Output; //this optimization is not good. __global__ void convolution_2D_tiled_kernel(float* P, float* N, int height, int width, int pitch, int Mask_Width){ int tx = threadIdx.x; int ty = threadIdx.y; int row_o = blockIdx.y * TILE_SIZE + ty; int col_o = blockIdx.x * TILE_SIZE + tx; int row_i = row_o - Mask_Width/2; int col_i = col_o - Mask_Width/2; __shared__ float N_ds[TILE_SIZE + KERNEL_LENGTH - 1][TILE_SIZE + KERNEL_LENGTH - 1]; if((row_i >= 0) && (row_i < height) && (col_i >= 0) && (col_i < width)){ N_ds[ty][tx] = N[row_i * pitch + col_i]; }else{ N_ds[ty][tx] = 0.0f; } __syncthreads(); float output = 0.0f; if(ty < TILE_SIZE && tx < TILE_SIZE){ for(int i = 0;i<Mask_Width;i++){ for(int j = 0;j<Mask_Width;j++){ output += c_M[i][j] * N_ds[i + ty][j + tx]; } } if(row_o < height && col_o < width){ P[row_o * width + col_o] = output; } } } void gpuRunTiledKernel(){ dim3 blockDim(TILE_SIZE,TILE_SIZE); dim3 gridDim((Width + TILE_SIZE - 1)/TILE_SIZE,(Height + TILE_SIZE - 1)/TILE_SIZE); convolution_2D_tiled_kernel<<<gridDim,blockDim>>>(d_Output,d_Input,Height,Width,Width,KERNEL_LENGTH); cudaDeviceSynchronize(); cudaEvent_t start; cudaEventCreate(&start); cudaEvent_t stop; cudaEventCreate(&stop); cudaEventRecord(start,NULL); for(int i = 0;i<nIter;i++){ convolution_2D_tiled_kernel<<<gridDim,blockDim>>>(d_Output,d_Input,Height,Width,Width,KERNEL_LENGTH); } cudaEventRecord(stop, NULL); cudaDeviceSynchronize(); float msecTotal = 0.0f; cudaEventElapsedTime(&msecTotal,start,stop); float gpuTime = (msecTotal / nIter) * 0.001; printf("Tiled Kernel Throughput = %0.4f MPixels/sec, Time= %.5f sec, Size= %u Pixels\n", (1.0e-6 * (Width*Height)/gpuTime), gpuTime, Width); cudaMemcpy(h_Output,d_Output,Width*Height* sizeof(float),cudaMemcpyDeviceToHost); bool correct = true; double eps = 1.e-6; for(int i = 0;i<Width*Height;i++){ double abs_err = fabs(h_Output[i] - KERNEL_LENGTH * KERNEL_LENGTH * 1.0); if (abs_err > eps) { //printf("Error! Index = %d,h_Output = %f,true value = %d\n", // i, h_Output[i], KERNEL_LENGTH*KERNEL_LENGTH); correct = false; } } printf("%s\n", correct ? "Result = PASS" : "Result = FAIL"); } int main(){ h_Kernel = (float*)malloc(KERNEL_LENGTH * KERNEL_LENGTH * sizeof(float)); h_Input = (float*)malloc(Width * Height * sizeof(float)); h_Output = (float*)malloc(Width * Height * sizeof(float)); for(unsigned int i = 0;i<KERNEL_LENGTH*KERNEL_LENGTH;i++){ h_Kernel[i] = 1.0f; } for(unsigned int i = 0;i<Width*Height;i++){ h_Input[i] = 1.0f; } cudaMalloc((void**)&d_Input, Width * Height * sizeof(float)); cudaMalloc((void**)&d_Output,Width * Height * sizeof(float)); cudaMemcpy(d_Input,h_Input,Width * Height * sizeof(float),cudaMemcpyHostToDevice); cudaMemcpyToSymbol(c_M,h_Kernel,KERNEL_LENGTH * KERNEL_LENGTH * sizeof(float)); printf("Running GPU convlution 2D ....\n"); gpuRunTiledKernel(); free(h_Input); free(h_Kernel); free(h_Output); cudaFree(d_Input); cudaFree(d_Output); }
2,325
#include "includes.h" #define DOUBLE #ifdef DOUBLE #define Complex cufftDoubleComplex #define Real double #define Transform CUFFT_Z2Z #define TransformExec cufftExecZ2Z #else #define Complex cufftComplex #define Real float #define Transform CUFFT_C2C #define TransformExec cufftExecC2C #endif #define TILE_DIM 8 // synchronize blocks __global__ void spread_i(Real* src, unsigned int spitch, Real* dst, unsigned int dpitch) { unsigned int bid = blockIdx.y * gridDim.x + blockIdx.x; unsigned int tid = threadIdx.x; Real res = src[bid * dpitch + tid]; if( tid < dpitch) dst[bid * spitch + tid] = res; }
2,326
#include "includes.h" __global__ void hyst_kernel(unsigned char *data, unsigned char *out, int rows, int cols) { // Establish our high and low thresholds as floats float lowThresh = 10; float highThresh = 70; // These variables are offset by one to avoid seg. fault errors // As such, this kernel ignores the outside ring of pixels const int row = blockIdx.y * blockDim.y + threadIdx.y + 1; const int col = blockIdx.x * blockDim.x + threadIdx.x + 1; const int pos = row * cols + col; const unsigned char EDGE = 255; unsigned char magnitude = data[pos]; if(magnitude >= highThresh) out[pos] = EDGE; else if(magnitude <= lowThresh) out[pos] = 0; else { float med = (highThresh + lowThresh) / 2; if(magnitude >= med) out[pos] = EDGE; else out[pos] = 0; } }
2,327
#include "includes.h" __device__ u_char clamp(float t) { if (t < 0) { return 0; } else if (t > 255){ return 255; } return t; } __global__ void kernel_colorSpaceYUV420PToRGBA(dev_t *src, dev_t *dst, int pitch_src, int pitch_dst, int w, int h) { unsigned int dim_x = blockDim.x * blockIdx.x + threadIdx.x; unsigned int dim_y = blockDim.y * blockIdx.y + threadIdx.y; int r,g,b,y,u,v; if (dim_x < w && dim_y < h) { y = *((u_char*)src + dim_y * pitch_src + dim_x); u = *((u_char*)src + (h + dim_y / 4) * pitch_src + dim_x / 2); v = *((u_char*)src + (h * 5 + dim_y) / 4 * pitch_src + dim_x / 2); r = clamp(y + 1.402 * (v - 128) + 0.5); g = clamp(y - 0.34414 * (u - 128) - 0.71414 * (v - 128) + 0.5); b = clamp(y + 1.772 * (u - 128) + 0.5); // *((uint32_t*)dst + dim_y * pitch_dst / 4 + dim_x) = (r << 24) + (g << 16) + (b << 8); *((u_char*)dst + dim_y * pitch_dst + dim_x * 4) = r; *((u_char*)dst + dim_y * pitch_dst + dim_x * 4 + 1) = g; *((u_char*)dst + dim_y * pitch_dst + dim_x * 4 + 2) = b; *((u_char*)dst + dim_y * pitch_dst + dim_x * 4 + 3) = 255; } }
2,328
/* #include "dpCudaFFT.hpp" #include "errorCheck.hpp" #define BEGIN cudaEventRecord(begin, 0); #define END cudaEventRecord(end, 0); cudaEventSynchronize(end); cudaEventElapsedTime(&delTime, begin, end); #define cudaErrChk(ans) { gpuAssert((ans), __FILE__, __LINE__); } //code from stackexchange to print cuda return messages inline void gpuAssert(cudaError_t code, char *file, int line, bool abort=false){ if (code != cudaSuccess){ fprintf(stderr,"%s %s %d\n", cudaGetErrorString(code), file, line); if (abort) exit(code); } } dpCudaFFT::dpCudaFFT(cl_context ctx, cl_command_queue q){ context = ctx; queue = q; workDimension = ONE_D; name = "FFT"; cudaEventCreate(&begin); cudaEventCreate(&end); } void dpCudaFFT::setup(int dataMB, int xLocal, int yLocal, int zLocal){ localSize[0] =localSize[1]=localSize[2]=1; Asize = (dataMB*1048576)/(sizeof(cufftComplex)); if (Asize%2 != 0) Asize++; MB = Asize * sizeof(cufftComplex)/1048576; } void dpCudaFFT::init(){ dataParameters.push_back(Asize); dataNames.push_back("nVectors"); Ain = (cufftComplex*) malloc(Asize*sizeof(cufftComplex)); Aout = (cufftComplex*) malloc(Asize * sizeof(cufftComplex)); if (!Aout || !Ain) fprintf(stderr,"error in malloc"); generate(Ain, Asize); } void dpCudaFFT::memoryCopyOut(){ BEGIN cudaErrChk(cudaMalloc((void**)&A_d, Asize * sizeof(cufftComplex))); cudaErrChk(cudaMemcpy(A_d, Ain, Asize * sizeof(cufftComplex), cudaMemcpyHostToDevice)); END //printf("%0.3f,", delTime); } void dpCudaFFT::plan(){ int ret = 0; BEGIN ret = cufftPlan1d(&plancufft, Asize, CUFFT_C2C, 1); if (ret != 0) fprintf(stderr, "%s %d", "cufftplan1d fail err: ", ret); END //printf("%0.3f,", delTime); } int dpCudaFFT::execute(){ int ret = 0; BEGIN ret = cufftExecC2C(plancufft, A_d, A_d, CUFFT_FORWARD); if (ret != 0){ fprintf(stderr, "%s %d", "cufftexecc2c fail err: ", ret); return -1; } END //printf("%0.3f,", delTime); return 0; } void dpCudaFFT::memoryCopyIn(){ BEGIN cudaErrChk(cudaMemcpy(Aout, A_d, Asize * sizeof(cufftComplex), cudaMemcpyDeviceToHost)); END //printf("%0.3f,\n", delTime); } void dpCudaFFT::cleanUp(){ cudaErrChk(cudaFree(A_d)); cufftDestroy(plancufft); free(Aout); free(Ain); } void dpCudaFFT::generate(cufftComplex *A, int N){ int i; srand(time(NULL)); for (i=0; i < N; i++){ A[i].x = rand() / (RAND_MAX/99999.9 + 1); A[i].y = rand() / (RAND_MAX/99999.9 + 1); } } */
2,329
#include<iostream> #include<fstream> using namespace std; const int N = 4096; const int BLOCKSIZE = 1024; __global__ void add_me(int *a, int *b, int *c) { int i = blockIdx.x * blockDim.x + threadIdx.x; c[i] = a[i] + b[i]; } int main() { ofstream outfile; outfile.open("output.txt"); int a[N] = {0}; int b[N]; int c[N] = {0}; int sum = 0; // Load b with 1s for (int i = 0; i < N; i++) b[i] = 1; int *ad; int *bd; int *cd; const int isize = N*sizeof(int); cudaMalloc( (void**)&ad, isize ); cudaMalloc( (void**)&bd, isize ); cudaMalloc( (void**)&cd, isize ); cudaMemcpy( ad, a, isize, cudaMemcpyHostToDevice ); cudaMemcpy( bd, b, isize, cudaMemcpyHostToDevice ); cudaMemcpy( cd, c, isize, cudaMemcpyHostToDevice ); dim3 dimGrid( 4, 1 ); dim3 dimBlock( BLOCKSIZE, 1 ); add_me<<<dimGrid, dimBlock>>>(ad, bd, cd); cudaMemcpy( c, cd, isize, cudaMemcpyDeviceToHost ); cudaFree( ad ); cudaFree( bd ); cudaFree( cd ); for (int i = 0; i < N; i++) sum += c[i]; cout << "The sum is: " << sum << '\n'; return EXIT_SUCCESS; }
2,330
#include <stdio.h> #include <stdlib.h> #define DSIZE 256 #define nTPB 64 #define cudaCheckErrors(msg) \ do { \ cudaError_t __err = cudaGetLastError(); \ if (__err != cudaSuccess) { \ fprintf(stderr, "Fatal error: %s (%s at %s:%d)\n", \ msg, cudaGetErrorString(__err), \ __FILE__, __LINE__); \ fprintf(stderr, "*** FAILED - ABORTING\n"); \ exit(1); \ } \ } while (0) __global__ void vector_add_kernel(float *c, const float *a, const float *b){ int idx = threadIdx.x+blockDim.x*blockIdx.x; if (idx < DSIZE) c[idx] = a[idx] + b[idx]; } int main(){ float *h_a, *h_b, *h_c, *d_a, *d_b, *d_c; h_a = (float *)malloc(DSIZE * sizeof(float)); h_b = (float *)malloc(DSIZE * sizeof(float)); h_c = (float *)malloc(DSIZE * sizeof(float)); if (h_a == NULL) {printf("malloc fail\n"); return 1;} if (h_b == NULL) {printf("malloc fail\n"); return 1;} if (h_c == NULL) {printf("malloc fail\n"); return 1;} for (int i = 0; i < DSIZE; i++){ h_c[i] = 0.0f; h_a[i] = rand()/(float)RAND_MAX; h_b[i] = rand()/(float)RAND_MAX;} cudaMalloc(&d_a, DSIZE*sizeof(float)); cudaMalloc(&d_b, DSIZE*sizeof(float)); cudaMalloc(&d_c, DSIZE*sizeof(float)); cudaCheckErrors("cudaMalloc fail"); cudaMemcpy(d_a, h_a, DSIZE*sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(d_b, h_b, DSIZE*sizeof(float), cudaMemcpyHostToDevice); cudaMemset(d_c, 0, DSIZE*sizeof(float)); cudaCheckErrors("cudaMemcpy H2D fail"); dim3 threads(nTPB, 1, 1); dim3 blocks((DSIZE+threads.x-1)/threads.x, 1, 1); vector_add_kernel<<<blocks, threads>>>(d_c, d_a, d_b); cudaDeviceSynchronize(); cudaCheckErrors("kernel fail"); cudaMemcpy(h_c, d_c, DSIZE*sizeof(float), cudaMemcpyDeviceToHost); cudaCheckErrors("cudaMemcpy D2H fail"); printf("h_a[0] = %f\n", h_a[0]); printf("h_b[0] = %f\n", h_b[0]); printf("h_c[0] = %f\n", h_c[0]); return 0; }
2,331
#include "includes.h" __global__ void matrixMul(int *a, int *b, int *c){ int my_x, my_y; my_x = blockIdx.x*blockDim.x + threadIdx.x; my_y = blockIdx.y*blockDim.y + threadIdx.y; int local_c = 0; for(int i = 0 ; i < size; i++) local_c += a[my_x * size + i] * b[i * size + my_y]; c[my_x * size + my_y ] = local_c; }
2,332
#include <stdio.h> /* * This file is an attempt at producing what the generated target code * should look like for the multiplyMatrixMatrix routine. */ /* Prototype matrix representation. */ struct dag_array_t{ size_t rows; size_t cols; int* matrix; }; /* DAG Primitive. Here, we leverage the NVIDIA developer examples to obtain a high-bandwith operation. They make use of shared memory to avoid strided global memory accesses, and instead perform the strided access in the shared block, which is roughly a ~3x improvement. TILE_DIM = 32 BLOCK_ROWS = 8 https://devblogs.nvidia.com/efficient-matrix-transpose-cuda-cc/ */ const int tp_TILE_DIM = 32; const int tp_BLOCK_ROWS = 8; __global__ void transposeCoalesced(int *result, const int *in) { const int TILE_DIM = tp_TILE_DIM; const int BLOCK_ROWS = tp_BLOCK_ROWS; __shared__ int tile[TILE_DIM][TILE_DIM]; int x = blockIdx.x * TILE_DIM + threadIdx.x; int y = blockIdx.y * TILE_DIM + threadIdx.y; int width = gridDim.x * TILE_DIM; for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS) tile[threadIdx.y+j][threadIdx.x] = in[(y+j)*width + x]; __syncthreads(); x = blockIdx.y * TILE_DIM + threadIdx.x; // transpose block offset y = blockIdx.x * TILE_DIM + threadIdx.y; for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS) result[(y+j)*width + x] = tile[threadIdx.x][threadIdx.y + j]; } __global__ void multiplyMatrixVector(int* result, int* matrix, int* vector, int cols) { __shared__ int reduce_array[256]; // Within a block int vector_slice_offset = blockIdx.x * cols + threadIdx.x; int matrix_slice_offset = blockIdx.y * cols + threadIdx.x; reduce_array[threadIdx.x] = matrix[matrix_slice_offset] * vector[vector_slice_offset]; __syncthreads(); // Sequential reduce. if (threadIdx.x == 0){ int accumulator = 0; for (int i = 0; i < blockDim.x; i++) { accumulator += reduce_array[i]; } result[blockIdx.x * cols + blockIdx.y] = accumulator; } } // We use single-dimensional lists. void matrixMultiply(dag_array_t* result, dag_array_t* m1, dag_array_t* m2){ // Precompute size information size_t size_m1 = m1->rows * m1->cols; size_t size_m2 = m2->rows * m2->cols; size_t size_result = m1->rows * m2->cols; // Copy onto device int* d_m1; cudaMalloc(&d_m1,size_m1); cudaMemcpy(d_m1,m1->matrix,size_m1,cudaMemcpyHostToDevice); int* d_m2; cudaMalloc(&d_m2,size_m2); cudaMemcpy(d_m2,m2->matrix,size_m2,cudaMemcpyHostToDevice); int* d_col; // We know that transpose will return same # of elem. cudaMalloc(&d_col,size_m2); int* d_result; // Allocate our result. cudaMalloc(&d_result,size_result); // A cruical optimization involves removing extraneous cudaMemcpy and cudaMallocs. dim3 dimGrid(m2->rows/tp_TILE_DIM, m2->cols/tp_TILE_DIM, 1); dim3 dimBlock(tp_TILE_DIM, tp_BLOCK_ROWS, 1); transposeCoalesced<<<dimGrid,dimBlock>>>(d_col,d_m2); const int threadsPerBlock = 256; dim3 dimBlock2(threadsPerBlock,1,1); // 256 threads per row dim3 dimGrid2((m1->rows + threadsPerBlock - 1) / threadsPerBlock, (m2->rows + threadsPerBlock - 1) / threadsPerBlock,1); multiplyMatrixVector<<<dimGrid2,dimBlock2>>>(d_result,d_m1,d_col,m1->cols); cudaMemcpy(result->matrix,d_result,size_result,cudaMemcpyDeviceToHost); result->rows = m1->rows; result->cols = m2->cols; cudaFree(d_m1); cudaFree(d_m2); cudaFree(d_result); cudaFree(d_col); } int main(){ int* a = (int*) malloc(100*sizeof(int)); int* b = (int*) malloc(100*sizeof(int)); dag_array_t A; A.rows = 10; A.cols = 10; A.matrix = a; dag_array_t B; B.rows = 10; B.cols = 10; B.matrix = b; dag_array_t C; C.rows = 10; C.cols = 10; C.matrix = (int*) malloc(100*sizeof(int)); for (int i = 0; i < 10; i++) { for (int j = 0; j < 10; j++) { a[i*10+j] = 1; b[i*10+j] = 2; } } matrixMultiply(&C,&A,&B); for (int i = 0; i < 10; i++) { for (int j = 0; j < 10; j++) { printf(" %d ",C.matrix[i*10+j]); } } }
2,333
#include <cstdio> #include <cstring> #include<iostream> using namespace std; class Myclass { public: Myclass(int a=0,int b=0) {_a=a;_b=b;} virtual __host__ __device__ void printValues() { printf("a = %d, b = %d\n", _a, _b); } private: int _a; int _b; }; __global__ void virtualFunctions(Myclass *vf) { Myclass vf_local = Myclass(*vf); memcpy(vf, &vf_local, sizeof(Myclass)); vf->printValues(); } __global__ void callVFunc(Myclass *vf) { vf->printValues(); } int main() { //CPU Myclass vf_host(4,5); //GPU Myclass *vf; cudaMalloc(&vf, sizeof(Myclass)); // CPU --> GPU cudaMemcpy(vf, &vf_host, sizeof(Myclass), cudaMemcpyHostToDevice); virtualFunctions<<<1, 1>>>(vf); cudaDeviceSynchronize(); callVFunc<<<1, 1>>>(vf); cudaDeviceSynchronize(); return 0; }
2,334
#include <iostream> #include <cmath> #include <vector> #include <random> #include <cassert> #define GPU_CHECK(ans) { gpuAssert((ans), __FILE__, __LINE__); } constexpr auto VECTOR_LENGTH = 1024u * 2; constexpr auto EPS = 1e-4f; inline void gpuAssert(cudaError_t code, const char *file, int line, bool abort = true) { if (code != cudaSuccess) { std::cout << "GPUassert: " << cudaGetErrorString(code) << " " << file << ":" << line << "\n"; if (abort) { std::exit(code); } } } void addHost(const std::vector<float> &A, const std::vector<float> &B, std::vector<float> &C) { assert(A.size() == B.size() && B.size() == C.size()); for (auto i = 0u; i < A.size(); ++i) { C[i] = A[i] + B[i]; } } __global__ void saxpy(const float *A, const float *B, float *C) { int i = blockIdx.x * blockDim.x + threadIdx.x; C[i] = A[i] + B[i]; } void addGPU(const std::vector<float> &A, const std::vector<float> &B, std::vector<float> &C) { float *A_gpu, *B_gpu, *C_gpu; auto byte_size = VECTOR_LENGTH * sizeof(float); GPU_CHECK(cudaMalloc(&A_gpu, byte_size)); GPU_CHECK(cudaMalloc(&B_gpu, byte_size)); GPU_CHECK(cudaMalloc(&C_gpu, byte_size)); GPU_CHECK(cudaMemcpy(A_gpu, A.data(), byte_size, cudaMemcpyHostToDevice)); GPU_CHECK(cudaMemcpy(B_gpu, B.data(), byte_size, cudaMemcpyHostToDevice)); GPU_CHECK(cudaMemcpy(C_gpu, C.data(), byte_size, cudaMemcpyHostToDevice)); saxpy<<<2, VECTOR_LENGTH / 2>>>(A_gpu, B_gpu, C_gpu); GPU_CHECK(cudaMemcpy(C.data(), C_gpu, byte_size, cudaMemcpyDeviceToHost)); } bool verify(const std::vector<float> &A, const std::vector<float> &B, const std::vector<float> &C) { for (auto i = 0u; i < VECTOR_LENGTH; ++i) { if (A[i] + B[i] - C[i] > EPS) { std::cout << "ERROR! Index " << i << "\n"; std::cout << A[i] + B[i] << " " << C[i] << "\n"; return false; } } return true; } int main() { std::random_device rd{}; std::mt19937 gen{rd()}; std::normal_distribution<float> dist{5, 2}; std::vector<float> A(VECTOR_LENGTH), B(VECTOR_LENGTH), C(VECTOR_LENGTH); for (auto i = 0u; i < VECTOR_LENGTH; ++i) { A[i] = dist(gen); B[i] = dist(gen); C[i] = M_PI; } addHost(A, B, C); if (!verify(A, B, C)) { return 1; } std::cout << "Host verified\n"; std::fill(C.begin(), C.end(), M_PI); addGPU(A, B, C); if (!verify(A, B, C)) { return 1; } std::cout << "GPU verified\n"; return 0; }
2,335
/* This is a automatically generated test. Do not modify */ #include <stdio.h> #include <stdlib.h> #include <math.h> __global__ void compute(float comp, float var_1,float var_2,float var_3,float var_4,float var_5,float var_6,float var_7,float var_8,float var_9,float var_10,float var_11,float var_12,float var_13,float var_14,float var_15,float var_16,float var_17,float var_18,float var_19,float var_20,float var_21,float var_22,float var_23,float var_24,float var_25,float var_26,float var_27,float var_28,float var_29,float var_30,float var_31,float var_32,float var_33) { comp = sinf((var_1 * var_2 / coshf(+1.6899E-36f - var_3 - var_4 + var_5))); comp = +1.2287E-43f + -1.8636E-35f; comp += (var_6 - (var_7 - -1.9095E5f * (+1.9059E-37f - logf((-1.5088E-36f / fmodf(sinf(var_8 + sqrtf((var_9 + (+0.0f + var_10)))), (var_11 + ldexpf((-1.6471E25f * sqrtf((var_12 + -1.3755E-37f + var_13 * powf((var_14 - -1.9217E35f / var_15 / sinf(-1.4114E-42f)), +0.0f / (-1.5518E13f / var_16 / (var_17 / atanf(+1.0853E29f * +1.2548E-3f))))))), 2)))))))); comp += (+1.3202E-44f - +1.1656E-36f / var_18 - +1.7676E-36f * var_19); if (comp == var_20 + fmodf((var_21 + +1.1755E-37f + var_22 * (var_23 - acosf(cosf((+0.0f - (-1.1293E36f + -0.0f / +1.7082E-42f)))))), +1.7465E-15f / var_24 * (+1.1625E36f + var_25))) { comp = (-1.7123E15f - +1.0857E-44f / var_26 * -1.0677E-41f); comp += (+1.1663E-20f * (+1.0656E-43f - ceilf(ldexpf((+1.8071E-25f * var_27 * var_28 - (var_29 + -1.1838E-28f * +1.2957E-36f)), 2)))); float tmp_1 = floorf(atan2f(+1.3948E-44f + (var_30 + var_31 / +1.9812E36f), (-0.0f + (-1.9448E-36f * atanf(+1.9803E-35f))))); comp += tmp_1 - (-1.1810E-43f / (var_32 + sqrtf(-1.3258E35f / -1.0834E36f + (+0.0f - (-1.0444E-35f / var_33))))); } printf("%.17g\n", comp); } float* initPointer(float v) { float *ret = (float*) malloc(sizeof(float)*10); for(int i=0; i < 10; ++i) ret[i] = v; return ret; } int main(int argc, char** argv) { /* Program variables */ float tmp_1 = atof(argv[1]); float tmp_2 = atof(argv[2]); float tmp_3 = atof(argv[3]); float tmp_4 = atof(argv[4]); float tmp_5 = atof(argv[5]); float tmp_6 = atof(argv[6]); float tmp_7 = atof(argv[7]); float tmp_8 = atof(argv[8]); float tmp_9 = atof(argv[9]); float tmp_10 = atof(argv[10]); float tmp_11 = atof(argv[11]); float tmp_12 = atof(argv[12]); float tmp_13 = atof(argv[13]); float tmp_14 = atof(argv[14]); float tmp_15 = atof(argv[15]); float tmp_16 = atof(argv[16]); float tmp_17 = atof(argv[17]); float tmp_18 = atof(argv[18]); float tmp_19 = atof(argv[19]); float tmp_20 = atof(argv[20]); float tmp_21 = atof(argv[21]); float tmp_22 = atof(argv[22]); float tmp_23 = atof(argv[23]); float tmp_24 = atof(argv[24]); float tmp_25 = atof(argv[25]); float tmp_26 = atof(argv[26]); float tmp_27 = atof(argv[27]); float tmp_28 = atof(argv[28]); float tmp_29 = atof(argv[29]); float tmp_30 = atof(argv[30]); float tmp_31 = atof(argv[31]); float tmp_32 = atof(argv[32]); float tmp_33 = atof(argv[33]); float tmp_34 = atof(argv[34]); compute<<<1,1>>>(tmp_1,tmp_2,tmp_3,tmp_4,tmp_5,tmp_6,tmp_7,tmp_8,tmp_9,tmp_10,tmp_11,tmp_12,tmp_13,tmp_14,tmp_15,tmp_16,tmp_17,tmp_18,tmp_19,tmp_20,tmp_21,tmp_22,tmp_23,tmp_24,tmp_25,tmp_26,tmp_27,tmp_28,tmp_29,tmp_30,tmp_31,tmp_32,tmp_33,tmp_34); cudaDeviceSynchronize(); return 0; }
2,336
/* #include "LennardJones.h" //------------------------Lennard Jones Potential -----------------------------// __host__ __device__ double lennardJonesForce(double dist, double sig, double eps) { double sigsq = sig*sig; double con = 24.0*eps/sigsq; double dist2 = dist * dist; dist2 /= sigsq; double dist4 = dist2*dist2; double dist8 = dist4*dist4; double dist14 = dist2*dist4*dist8; double invdist8= 1.0/dist8; double invdist14= 1.0/dist14; double s = 2.0*invdist14-invdist8; return s * con; } //----------------------------------------------------------------------------// __host__ __device__ double lennardJonesPotential(double dist, double sig, double eps) { double expr = sig/dist; double expr2 = expr*expr; double expr4 = expr2*expr2; double expr6 = expr4*expr2; double expr12 = expr6*expr6; return 4.0*eps*(expr12-expr6); } */
2,337
#include "includes.h" __global__ void ForwardCrossEntropy(float *output, float *labels, int nColsOutput, float *loss) { int col = blockIdx.x; float temp = -(labels[col] * logf(output[col]) + logf(1 - output[col]) * (1 - labels[col])); atomicAdd(loss, temp); }
2,338
#include "includes.h" __global__ void kHingeLinearRowMajor(float* mat, float* labels, float* target, unsigned int width, unsigned int height, float margin) { int image_id = blockIdx.x * blockDim.x + threadIdx.x; if (image_id < height) { mat += image_id; target += image_id; const int correct_label = (int)labels[image_id]; const float correct_label_score = mat[correct_label * height]; float sum = 0; for (unsigned int i = 0; i < width; i++) { float diff = margin + mat[i*height] - correct_label_score; float grad = (diff > 0) ? 1 : 0; target[i*height] = (i == correct_label) ? 0 : grad; sum += (i == correct_label) ? 0 : grad; } target[correct_label * height] = -sum; } }
2,339
#include <iostream> #include <stdio.h> #include <math.h> #include <sys/time.h> #define MATRIX_SIZE 256 #define BLOCK_SIZE 16 using namespace std; __global__ void matMul(float *x, float *y, float *z, int matrixSize){ float zTemp = 0.0f; __shared__ float xblkMat[BLOCK_SIZE*BLOCK_SIZE], yblkMat[BLOCK_SIZE*BLOCK_SIZE]; const int global_x = threadIdx.x + blockIdx.x * blockDim.x; const int global_y = threadIdx.y + blockIdx.y * blockDim.y; const int blocked_x = blockIdx.x; const int blocked_y = blockIdx.y; const int blocked_x_id = threadIdx.x; const int blocked_y_id = threadIdx.y; const int numBlocks = matrixSize / BLOCK_SIZE; int xStart = blocked_y * matrixSize * BLOCK_SIZE; int yStart = blocked_x * BLOCK_SIZE; for (int block = 0; block < numBlocks; block++) { xblkMat[blocked_x_id + (blocked_y_id*BLOCK_SIZE)] = x[xStart + ((blocked_y_id*matrixSize) + blocked_x_id)]; yblkMat[blocked_x_id + (blocked_y_id*BLOCK_SIZE)] = y[yStart + ((blocked_y_id*matrixSize) + blocked_x_id)]; __syncthreads(); for (int k = 0;k < BLOCK_SIZE;k++) { zTemp += xblkMat[k + (blocked_y_id * BLOCK_SIZE)] * yblkMat[blocked_x_id + (k * BLOCK_SIZE)]; } __syncthreads(); xStart += BLOCK_SIZE; yStart += BLOCK_SIZE; } z[global_x + (global_y * matrixSize)] = zTemp; } int main(){ float *x,*y,*z; struct timeval start; struct timeval end; double elapsedTime; double numOps; float gFLOPS; cudaMallocManaged(&x, MATRIX_SIZE * MATRIX_SIZE * sizeof(float)); cudaMallocManaged(&y, MATRIX_SIZE * MATRIX_SIZE * sizeof(float)); cudaMallocManaged(&z, MATRIX_SIZE * MATRIX_SIZE * sizeof(float)); for(int i=0;i<MATRIX_SIZE;i++){ for(int j=0;j<MATRIX_SIZE;j++){ *(x + i*MATRIX_SIZE + j) = 1.0f; *(y + i*MATRIX_SIZE + j) = 1.0f; *(z + i*MATRIX_SIZE + j) = 0.0f; } } // Keep track of when we start doing work gettimeofday(&start, NULL); dim3 threads(BLOCK_SIZE,BLOCK_SIZE); dim3 grid(MATRIX_SIZE/threads.x,MATRIX_SIZE/threads.y); matMul<<<grid,threads>>>(x,y,z,MATRIX_SIZE); cudaDeviceSynchronize(); // Keep track of when we finish our work gettimeofday(&end, NULL); // Calculate the time it took to do the above task elapsedTime = (end.tv_sec - start.tv_sec) * 1000.0; elapsedTime += (end.tv_usec - start.tv_usec) / 1000.0; elapsedTime /= 1000; //Calculate the GFLOPS obtained and print it along with the execution time numOps = 2 * pow(MATRIX_SIZE, 3); gFLOPS = float(1.0e-9 * numOps / elapsedTime); printf("CUDA : %.3f seconds ( %f GFLOPS )\n",elapsedTime,gFLOPS); /*cout << "X[23][65] : " << *(x + 23*MATRIX_SIZE + 65) << endl; cout << "Y[23][65] : " << *(y + 23*MATRIX_SIZE + 65) << endl; cout << "Z[23][65] : " << *(z + 23*MATRIX_SIZE + 65) << endl;*/ cudaFree(x); cudaFree(y); cudaFree(z); return 0; }
2,340
#include "includes.h" __global__ void kernel_push_atomic2(int *g_terminate, int *g_push_reser, int *s_push_reser, int *g_block_num, int width1) { int x = __umul24(blockIdx.x, blockDim.x) + threadIdx.x; int y = __umul24(blockIdx.y, blockDim.y) + threadIdx.y; int thid = __umul24(y, width1) + x; if (s_push_reser[thid] - g_push_reser[thid] != 0) { g_terminate[blockIdx.y * (*g_block_num) + blockIdx.x] = 1; } }
2,341
#include <stdlib.h> #include <stdio.h> #include <string.h> #include <cuda.h> #define BLOCKS 8 #define THREADS 16 #define WIDTH 128 #define HEIGHT 64 __global__ void add(int* a, int* b, int* c) { int idx = threadIdx.x + blockIdx.x * blockDim.x; int idy = threadIdx.y + blockIdx.y * blockDim.y; if (idx > WIDTH || idy > HEIGHT) return; c[idy * WIDTH + idx] = a[idy * WIDTH + idx] + b[idy * WIDTH + idx]; } void init(int* h_v, int numb) { for (int i = 0; i < HEIGHT; i++) { for (int j = 0; j < WIDTH; ++j) { h_v[i * HEIGHT + j] = numb; } } } // N int main( void ) { int *result, *h_a, *h_b; int *dev_a, *dev_b, *dev_c; int size = HEIGHT * WIDTH * sizeof(int); size_t size_pitch; result = (int*) malloc( size ); h_a = (int*) malloc( size ); h_b = (int*) malloc( size ); init(h_a, 1); init(h_b, 2); // cudaMallocPitch cudaMemcpy2D cudaError_t error = cudaMallocPitch( &dev_a, &size_pitch, WIDTH * sizeof(int), HEIGHT ); // Warning: el pitch varia por el tamaño que reservas si a b c fueren diferente tamaño.. error = cudaMallocPitch( &dev_b, &size_pitch, WIDTH * sizeof(int), HEIGHT ); error = cudaMallocPitch( &dev_c, &size_pitch, WIDTH * sizeof(int), HEIGHT ); // cudaError_t error = cudaMalloc( &dev_a, WIDTH * sizeof(int) * HEIGHT ); // error = cudaMalloc( &dev_b, size ); // error = cudaMalloc( &dev_c, size ); // fprintf(stderr, "Size %lu\n", WIDTH * sizeof(int) * HEIGHT); // se transfieren los datos a memoria de dispositivo. // cudaMemcpy( dev_a, h_a, size , cudaMemcpyHostToDevice); // cudaMemcpy( dev_b, h_b, size , cudaMemcpyHostToDevice); // cudaMemset(dev_c, 0, size); // cudaMemcpy2D (void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind) error = cudaMemcpy2D( dev_a, size_pitch, h_a, WIDTH * sizeof(int), WIDTH * sizeof(int), HEIGHT, cudaMemcpyHostToDevice); error = cudaMemcpy2D( dev_b, size_pitch, h_b, WIDTH * sizeof(int), WIDTH * sizeof(int), HEIGHT, cudaMemcpyHostToDevice); // cudaError_t cudaMemset2D (void *devPtr, size_t pitch, int value, size_t width, size_t height) error = cudaMemset2D(dev_c, size_pitch, 0, sizeof(int) * WIDTH, HEIGHT); dim3 t(THREADS, THREADS); dim3 b( (WIDTH + t.x - 1) / t.x , (HEIGHT + t.y - 1) / t.y); add<<<b, t>>>( dev_a, dev_b, dev_c); // se transfieren los datos del dispositivo a memoria. // cudaMemcpy(result, dev_c, size, cudaMemcpyDeviceToHost); // cudaMemcpy2D (void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind) cudaMemcpy2D(result, WIDTH * sizeof(int), dev_c, size_pitch, WIDTH * sizeof(int), HEIGHT, cudaMemcpyDeviceToHost); fprintf(stderr, "%s\n", "Result..."); for (int i = 0; i < HEIGHT; ++i) { for (int j = 0; j < WIDTH; ++j) { fprintf(stderr, " %d ", result[i * HEIGHT + j]); } fprintf(stderr, "%s\n", ""); } fprintf(stderr, "%s\n", ""); free(h_a), free(h_b), free(result); cudaFree(dev_a); cudaFree(dev_b); cudaFree(dev_c); return 0; }
2,342
#include <iostream> #include <chrono> #include <cassert> #include <cmath> #include <cstdlib> #include <vector> #include <algorithm> #define BLOCKSIZE 128 #define LOG_BLOCKSIZE 7 // MUST BE ASSOCIATIVE __device__ inline int f(int a, int b){ return a + b; } /** * In this variant, several optimizations have been applied: * * - Since only half the threads were doing work in parallel_main.cu, only BLOCKSIZE / 2 threads * now make up a block, reducing the number of warps scheduled. * * - In parallel_cu, a warp would always have divergence in that some threads would do work and * others would idle. In this variant, if the current downsweep iteration needs 8 workers, then * only the first 8 threads of the block will actually do work, reducing dependence. **/ __global__ void scan(const int n, int *x, int *out){ __shared__ int scan_v[2 * BLOCKSIZE]; int tid = threadIdx.x; int tcount = blockDim.x; int i = blockIdx.x * 2 * blockDim.x + threadIdx.x * 2; scan_v[2 * tid] = x[i]; scan_v[(2 * tid) ^ 1] = x[i ^ 1]; __syncthreads(); for(int i = 1, j = 1; j <= LOG_BLOCKSIZE; i = (i << 1) + 1, j++){ int curr_tid = (tid << j) | i; int oth = curr_tid - (1 << (j - 1)); if(curr_tid < 2 * tcount){ scan_v[curr_tid] = f(scan_v[curr_tid], scan_v[oth]); } __syncthreads(); } for(int i = BLOCKSIZE / 2, j = LOG_BLOCKSIZE - 1; i >= 1; i /= 2, j--){ int curr_tid = ((tid + 1) << (j + 1)) | (i - 1); int oth = curr_tid - i; if(curr_tid < BLOCKSIZE){ scan_v[curr_tid] = f(scan_v[curr_tid], scan_v[oth]); } __syncthreads(); } x[i] = scan_v[2 * tid]; x[i ^ 1] = scan_v[(2 * tid) ^ 1]; if(tid == blockDim.x - 1){ out[blockIdx.x] = scan_v[(2 * tid) ^ 1]; } } /** * One slight optimization was done here: * - The first block in parallel_main.cu doesn't do any work. We shift each block by +1, resulting * in one less thread block scheduled, and no need for the condition anymore. **/ __global__ void propagate(const int n, int *in, int *out){ int bid = blockIdx.x + 1; int tcount = blockDim.x; int tid = threadIdx.x; int i = bid * tcount + tid; out[i] = f(out[i], in[bid - 1]); } std::vector<int> get_levels(const int n, int block_size){ std::vector<int> res; int x = n; while(x > 1){ res.push_back(x); x = (x + block_size - 1) / block_size; } res.push_back(1); return res; } int main(){ const int n = (1 << 28); const int block_size = BLOCKSIZE; assert(n % block_size == 0); std::vector<int> levels = get_levels(n, block_size); for(int i : levels){ std::cout << i << ' '; } std::cout << std::endl; int *x = (int *) malloc(n * sizeof(int)); assert(x != NULL); for(int i = 0; i < n; i++){ x[i] = 1; } int *d_arrays[levels.size()]; for(int i = 0; i < levels.size(); i++){ cudaMalloc(&d_arrays[i], levels[i] * sizeof(int)); assert(d_arrays[i] != NULL); } cudaMemcpy(d_arrays[0], x, levels[0] * sizeof(int), cudaMemcpyHostToDevice); for(int i = 1; i < levels.size(); i++){ int block_count = levels[i]; scan<<<block_count, block_size / 2>>>(levels[i - 1], d_arrays[i - 1], d_arrays[i]); } for(int i = levels.size() - 2; i >= 1; i--){ int block_count = levels[i]; propagate<<<block_count - 1, block_size>>>(levels[i - 1], d_arrays[i], d_arrays[i - 1]); } int *result = (int *) malloc(n * sizeof(int)); cudaMemcpy(result, d_arrays[0], n * sizeof(int), cudaMemcpyDeviceToHost); for(int i = 0; i < n; i++){ if(result[i] != i + 1){ std::cerr << i << ' ' << i + 1 << ' ' << result[i] << '\n'; return -1; } } std::cout << "memory usage: " << n * sizeof(int) << " bytes" << std::endl; }
2,343
/* This is a automatically generated test. Do not modify */ #include <stdio.h> #include <stdlib.h> #include <math.h> __global__ void compute(float comp, float var_1,float var_2,float var_3,float var_4,float var_5,int var_6,float var_7,float var_8,float var_9,float var_10,float var_11,float var_12,float var_13,float var_14,float var_15,float var_16,float var_17,float var_18,float var_19,float var_20,float var_21,float var_22,float var_23,float var_24,float var_25,float var_26,float var_27,float var_28,float var_29,float var_30,float var_31,float var_32,float var_33,float var_34,float var_35,float var_36,float var_37,float var_38,float var_39,float var_40,float var_41) { if (comp == var_1 / +0.0f - var_2 - (var_3 * var_4 - var_5)) { comp = +1.4329E-43f * (+0.0f / (var_7 / -1.1450E36f)); comp += asinf((+1.5663E-22f / var_8 + logf(ceilf((var_9 + +1.8261E-44f * fmodf((var_10 + sqrtf(+1.6147E-36f)), (+1.2292E35f * (-0.0f - +1.5430E-41f)))))))); comp += (var_11 * sinf((+1.5481E36f / var_12 - atanf(tanhf(-1.8254E-44f))))); if (comp == (+1.6499E35f * atanf(atanf(var_13 / var_14)))) { comp += var_15 - +0.0f / var_16; float tmp_1 = var_17 * +0.0f - var_18 / tanhf(-1.6246E-36f); comp += tmp_1 - log10f((+1.3728E-44f / fmodf((var_19 / (var_20 * +0.0f / -1.4205E-35f / +1.7331E-41f)), var_21 + var_22 * (var_23 / +1.4503E34f)))); comp = (+1.9415E34f * (var_24 * +1.0316E-29f * powf(+1.4698E-42f, -1.9273E35f - +1.8795E34f))); } if (comp < -1.5516E21f / (+1.6774E35f * (var_25 + -1.9158E36f))) { comp += (var_26 * (-1.3676E-37f / (var_27 / var_28))); float tmp_2 = log10f((var_29 * (+1.6247E-42f / var_30))); comp += tmp_2 * (-1.5496E-42f / var_31 + (var_32 + var_33 - var_34)); } for (int i=0; i < var_6; ++i) { comp += (+1.8339E-43f + -0.0f * (-1.8858E-41f + (var_35 + expf((var_36 + var_37 / -0.0f / var_38))))); comp = (-1.1668E-37f + fabsf((+1.9394E34f + var_39 / var_40 + (var_41 * -1.3973E-21f)))); } } printf("%.17g\n", comp); } float* initPointer(float v) { float *ret = (float*) malloc(sizeof(float)*10); for(int i=0; i < 10; ++i) ret[i] = v; return ret; } int main(int argc, char** argv) { /* Program variables */ float tmp_1 = atof(argv[1]); float tmp_2 = atof(argv[2]); float tmp_3 = atof(argv[3]); float tmp_4 = atof(argv[4]); float tmp_5 = atof(argv[5]); float tmp_6 = atof(argv[6]); int tmp_7 = atoi(argv[7]); float tmp_8 = atof(argv[8]); float tmp_9 = atof(argv[9]); float tmp_10 = atof(argv[10]); float tmp_11 = atof(argv[11]); float tmp_12 = atof(argv[12]); float tmp_13 = atof(argv[13]); float tmp_14 = atof(argv[14]); float tmp_15 = atof(argv[15]); float tmp_16 = atof(argv[16]); float tmp_17 = atof(argv[17]); float tmp_18 = atof(argv[18]); float tmp_19 = atof(argv[19]); float tmp_20 = atof(argv[20]); float tmp_21 = atof(argv[21]); float tmp_22 = atof(argv[22]); float tmp_23 = atof(argv[23]); float tmp_24 = atof(argv[24]); float tmp_25 = atof(argv[25]); float tmp_26 = atof(argv[26]); float tmp_27 = atof(argv[27]); float tmp_28 = atof(argv[28]); float tmp_29 = atof(argv[29]); float tmp_30 = atof(argv[30]); float tmp_31 = atof(argv[31]); float tmp_32 = atof(argv[32]); float tmp_33 = atof(argv[33]); float tmp_34 = atof(argv[34]); float tmp_35 = atof(argv[35]); float tmp_36 = atof(argv[36]); float tmp_37 = atof(argv[37]); float tmp_38 = atof(argv[38]); float tmp_39 = atof(argv[39]); float tmp_40 = atof(argv[40]); float tmp_41 = atof(argv[41]); float tmp_42 = atof(argv[42]); compute<<<1,1>>>(tmp_1,tmp_2,tmp_3,tmp_4,tmp_5,tmp_6,tmp_7,tmp_8,tmp_9,tmp_10,tmp_11,tmp_12,tmp_13,tmp_14,tmp_15,tmp_16,tmp_17,tmp_18,tmp_19,tmp_20,tmp_21,tmp_22,tmp_23,tmp_24,tmp_25,tmp_26,tmp_27,tmp_28,tmp_29,tmp_30,tmp_31,tmp_32,tmp_33,tmp_34,tmp_35,tmp_36,tmp_37,tmp_38,tmp_39,tmp_40,tmp_41,tmp_42); cudaDeviceSynchronize(); return 0; }
2,344
#include <stdio.h> const int TILE_DIM = 32; const int BLOCK_ROWS = 8; const int NUM_REPS = 100; __global__ void copy(float *odata, const float *idata) { int x = blockIdx.x * TILE_DIM + threadIdx.x; int y = blockIdx.y * TILE_DIM + threadIdx.y; int width = gridDim.x * TILE_DIM; for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS) { // row is (tile y + offset into tile j)*width of matrix + col x odata[(y+j)*width + x] = idata[(y+j)*width + x]; } } __global__ void transposeNaive(float *odata, const float *idata) { int x = blockIdx.x * TILE_DIM + threadIdx.x; int y = blockIdx.y * TILE_DIM + threadIdx.y; int width = gridDim.x * TILE_DIM; for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS) { odata[x*width + (y+j)] = idata[(y+j)*width + x]; } } __global__ void transposeCoalesced(float *odata, float *idata) { __shared__ float tile[TILE_DIM][TILE_DIM + 1]; int x = blockIdx.x * TILE_DIM + threadIdx.x; int y = blockIdx.y * TILE_DIM + threadIdx.y; int width = gridDim.x * TILE_DIM; for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS) { tile[threadIdx.y+j][threadIdx.x] = idata[(y+j)*width + x]; } __syncthreads(); // Now these are the offsets into the transposed tile x = blockIdx.y * TILE_DIM + threadIdx.x; y = blockIdx.x * TILE_DIM + threadIdx.y; for (int j = 0; j < TILE_DIM; j += BLOCK_ROWS) { odata[(y+j)*width + x] = tile[threadIdx.x][threadIdx.y + j]; } }
2,345
#include "includes.h" __global__ void copy(float *odata, const float *idata) { int x = blockIdx.x * TILE_DIM + threadIdx.x; int y = blockIdx.y * TILE_DIM + threadIdx.y; int width = gridDim.x * TILE_DIM; for (int j = 0; j < TILE_DIM; j+= BLOCK_ROWS) odata[(y+j)*width + x] = idata[(y+j)*width + x]; }
2,346
#include <iostream> #include <complex> #include <math.h> #include <thrust/complex.h> #include <sys/time.h> using namespace std; __global__ void fft(thrust::complex<double> *g_odata, thrust::complex<double> *g_idata, int n) { extern __shared__ thrust::complex<double> temp[]; // allocated on invocation int thid = threadIdx.x; int pout = 0, pin = 1; temp[pout*n + thid] = g_idata[thid]; __syncthreads(); int thid1=0; int b = __log2f(n+1); for(int i=0; i<b;i++){ if(thid & (1<<i)) thid1 |= (1<<(b-1-i)); } pout = 1 - pout; pin = 1 - pin; temp[pout*n + thid] = temp[pin*n + thid1]; __syncthreads(); for(int i=1; i<n; i*=2){ pout = 1 - pout; pin = 1 - pin; thid1 = thid ^ i; thrust::complex<double> factor(cos(-M_PI*thid/i), sin(-M_PI*thid/i)); if(thid1 > thid){ temp[pout*n + thid] = temp[pin*n + thid] + factor * temp[pin*n + thid1]; } else{ temp[pout*n + thid] = temp[pin*n + thid1] + factor * temp[pin*n + thid]; } __syncthreads(); } g_odata[thid] = temp[pout*n + thid]; } void checkError(){ cudaError_t errSync = cudaGetLastError(); cudaError_t errAsync = cudaDeviceSynchronize(); if (errSync != cudaSuccess) printf("Sync kernel error: %s\n", cudaGetErrorString(errSync)); if (errAsync != cudaSuccess) printf("Async kernel error: %s\n", cudaGetErrorString(errAsync)); } int main(void) { int N; cin>>N; thrust::complex<double> *x, *y; cudaMallocManaged(&x, N*sizeof(thrust::complex<double>)); cudaMallocManaged(&y, N*sizeof(thrust::complex<double>)); for(int i=0; i<N;i++){ int t,u; cin>>t>>u; x[i] = complex<double>(t, u); } int blockSize = N; int numBlocks = (N + blockSize - 1) / blockSize; cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start); fft<<<numBlocks, blockSize, 2*N*sizeof(thrust::complex<double>)>>>(y, x, N); checkError(); cudaEventRecord(stop); cudaEventSynchronize(stop); float milliseconds = 0; cudaEventElapsedTime(&milliseconds, start, stop); cout<<milliseconds; // for (int i = 0; i < N; i++) // cout<<y[i]<<"\n"; // cout<<endl; cudaFree(x); cudaFree(y); }
2,347
#include <iostream> #include <math.h> __global__ void add(int n, float *x, float *y){ int index = threadIdx.x; int stride = blockDim.x; for (int i = index; i < n; i+=stride) y[i] = x[i] + y[i]; } int main(void){ int N = 1<<20; // 1M elemenets std::cout << N << std::endl; float *x, *y; cudaMallocManaged(&x, N*sizeof(float)); cudaMallocManaged(&y, N*sizeof(float)); for (int i = 0; i < N; i++){ x[i] = 1.0f; y[i] = 2.0f; } ///Run kernel add <<< 1, 256 >>> (N,x,y); //wait for synchro cudaDeviceSynchronize(); float maxError = 0.0f; for (int i = 0; i < N; i++) maxError = fmax(maxError, fabs(y[i]-3.0f)); std::cout << "max error: " << maxError << std::endl; cudaFree(x); cudaFree(y); return 0; }
2,348
extern "C" __global__ void mtranReference( float *output, float *input, const int width, const int height) { int x = blockIdx.x*blockDim.x + threadIdx.x; int y = blockIdx.y*blockDim.y + threadIdx.y; output[y*width + x] = input[x*height + y]; }
2,349
#include <iostream> #include <sys/time.h> #include <cuda.h> using namespace std; #define CUDA_CHECK_RETURN(value) {\ cudaError_t _m_cudaStat = value;\ if (_m_cudaStat != cudaSuccess) {\ fprintf(stderr, "Error %s at line %d in file %s\n", cudaGetErrorString(_m_cudaStat), __LINE__, __FILE__);\ exit(1);\ }} __global__ void VecSum(float *A, float *B, float *C, int size) { int i = threadIdx.x + blockIdx.x * blockDim.x; if (i < size) C[i] = A[i] + B[i]; } void printVec(float *C, int size) { for (int i = 0; i < size; i++) cout << C[i] << "\t"; cout << endl; } int main(int argc, char* argv[]) { if (argc != 3) { cout << "launch parametrs: [vector size] [threads per block]" << endl; return 1; } int size = atoi(argv[1]); int threads_per_block = atoi(argv[2]); srand(time(NULL)); float *A = new float[size]; float *B = new float[size]; float *C = new float[size]; float *dev_A, *dev_B, *dev_C; cudaMalloc((void**)&dev_A, size * sizeof(float)); cudaMalloc((void**)&dev_B, size * sizeof(float)); cudaMalloc((void**)&dev_C, size * sizeof(float)); for (int i = 0; i < size; i++) { A[i] = rand(); B[i] = rand(); } cudaMemcpy(dev_A, A, size * sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(dev_B, B, size * sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(dev_C, C, size * sizeof(float), cudaMemcpyHostToDevice); int blockTotal = ceilf(float(size) / float(threads_per_block)); cout << "Block total: " << blockTotal << endl; cout << "Threads per block : " << threads_per_block << endl; cout << "Threads total: " << blockTotal * threads_per_block << endl; float elapsedTime; cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start, 0); VecSum <<< blockTotal, threads_per_block >>> (dev_A, dev_B, dev_C, size); cudaEventRecord(stop, 0); cudaEventSynchronize(stop); CUDA_CHECK_RETURN(cudaDeviceSynchronize()); CUDA_CHECK_RETURN(cudaGetLastError()); cudaEventElapsedTime(&elapsedTime, start, stop); cudaMemcpy(C, dev_C, size * sizeof(float), cudaMemcpyDeviceToHost); cout << "time: " << elapsedTime << " ms" << endl; //printVec(C, size); delete [] A; delete [] B; delete [] C; cudaEventDestroy(start); cudaEventDestroy(stop); cudaFree(dev_A); cudaFree(dev_B); cudaFree(dev_C); return 0; }
2,350
#include <iostream> using namespace std; __global__ void square(float *d_out, float *d_in){ int idx = blockDim.x*blockIdx.x + threadIdx.x; float f = d_in[idx]; d_out[idx] = f*f; } int main(){ const int ARRAY_SIZE = 10000; 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]; float *d_in; float *d_out; cudaMalloc((void**) &d_in, ARRAY_BYTES); cudaMalloc((void**) &d_out, ARRAY_BYTES); cudaMemcpy(d_in, h_in, ARRAY_BYTES, cudaMemcpyHostToDevice); int NUM_THREADS = 512; int NUM_BLOCKS = ARRAY_SIZE / NUM_THREADS + 1; square<<<NUM_BLOCKS, NUM_THREADS>>>(d_out, d_in); cudaMemcpy(h_out, d_out, ARRAY_BYTES, cudaMemcpyDeviceToHost); for(int i=0; i< ARRAY_SIZE; i++){ cout << h_out[i]; if(i%10!=9) cout << "\t"; else cout << endl; } }
2,351
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include <stdlib.h> #include <stdio.h> __global__ void Read_texture_obj_kernel(float *iptr, cudaTextureObject_t tex) { int x = threadIdx.x + blockIdx.x * blockDim.x; int y = threadIdx.y + blockIdx.y * blockDim.y; int offset = x + y * blockDim.x * gridDim.x; float c = tex1Dfetch<float>(tex, offset); iptr[offset] = c; } cudaTextureObject_t *TexObjFloat1D(float *devPtr, int length) { // create texture object cudaResourceDesc resDesc; memset(&resDesc, 0, sizeof(resDesc)); resDesc.resType = cudaResourceTypeLinear; resDesc.res.linear.devPtr = devPtr; resDesc.res.linear.desc.f = cudaChannelFormatKindFloat; resDesc.res.linear.desc.x = 32; // bits per channel resDesc.res.linear.sizeInBytes = length * sizeof(float); cudaTextureDesc texDesc; memset(&texDesc, 0, sizeof(texDesc)); texDesc.readMode = cudaReadModeElementType; cudaTextureObject_t *tex = (cudaTextureObject_t *)malloc(sizeof(cudaTextureObject_t)); cudaCreateTextureObject(tex, &resDesc, &texDesc, NULL); return tex; }
2,352
#include <array> // CUDA kernel. Each thread takes care of one element of c template<class T> __global__ void vecAdd(T *a, T *b, T *c, int n) { // Get our global thread ID int id = blockIdx.x*blockDim.x+threadIdx.x; // Make sure we do not go out of bounds if (id < n) c[id] = a[id] + b[id]; } template <typename T, size_t N> void simple_vadd_cuda(const std::array<T, N>& VA, const std::array<T, N>& VB, std::array<T, N>& VC) { // Device input vectors T *d_a; T *d_b; //Device output vector T *d_c; // Size, in bytes, of each vector const size_t bytes = N*sizeof(T); // Allocate memory for each vector on GPU cudaMalloc(&d_a, bytes); cudaMalloc(&d_b, bytes); cudaMalloc(&d_c, bytes); // Copy host vectors to device cudaMemcpy( d_a, VA.data(), bytes, cudaMemcpyHostToDevice); cudaMemcpy( d_b, VB.data(), bytes, cudaMemcpyHostToDevice); int blockSize, gridSize; // Number of threads in each thread block blockSize = 1024; // Number of thread blocks in grid gridSize = (int)ceil((float)N/blockSize); // Execute the kernel vecAdd<<<gridSize, blockSize>>>(d_a, d_b, d_c, N); // Copy array back to host cudaMemcpy( VC.data(), d_c, bytes, cudaMemcpyDeviceToHost ); // Release device memory cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); } template void simple_vadd_cuda<float, 4>(const std::array<float, 4>& VA, const std::array<float, 4>& VB, std::array<float, 4>& VC); template void simple_vadd_cuda<int, 4>(const std::array<int, 4>& VA, const std::array<int, 4>& VB, std::array<int, 4>& VC);
2,353
/***************************************************************************//** * \file projectVelocity.cu * \author Anush Krishnan (anush@bu.edu), * \author Christopher Minar (minarc@oregonstate.edu) * \brief kernels to update the velocity field */ #include "projectVelocity.h" namespace kernels { __global__ void project_velocity_X(double *u, double *uhat, double *uold, double *pressure, int *ghostTagsP, double *dx, double dt, int nx, int ny) { int i = threadIdx.x + (blockDim.x * blockIdx.x), I = i % (nx-1), J = i / (nx-1), ip = J*nx + I, numU = (nx-1)*ny; if (i >= numU) return; u[i] = uhat[i] - (ghostTagsP[ip+1] ==-1 && ghostTagsP[ip] == -1) * dt*(pressure[ip+1]-pressure[ip]) / (0.5*dx[I+1]+0.5*dx[I]); } __global__ void project_velocity_Y(double *u, double *uhat, double *uold, double *pressure, int *ghostTagsP, double *dy, double dt, int nx, int ny) { int numU= (nx-1)*ny, i = threadIdx.x + (blockDim.x * blockIdx.x), I = i % nx, J = i / nx, ip = J*nx + I, numUV = (ny-1)*nx + numU; i += numU; if (i >= numUV) return; u[i] = uhat[i] - (ghostTagsP[ip+nx] == -1 && ghostTagsP[ip] == -1) * dt*(pressure[ip+nx]-pressure[ip]) / (0.5*dy[J+1]+0.5*dy[J]); } }//end namespace kernels
2,354
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include <time.h> #include <stdio.h> #include <stdlib.h> //#include <atlimage.h> enum color_transform_t { grayscale, sRGB, LAB }; enum transform_t { Gaussian }; #define SIZE 1000 //typedef struct //{ // int r; // int g; // int b; //} rgb_t; // //typedef rgb_t* rgb_ptr; //typedef rgb_ptr* rgb_list; //typedef rgb_list* rgb_mat; // cudaError_t transform(uchar3 *dst_img, uchar3 *src_img, int img_size, int block_size, int grid_size, color_transform_t type); cudaError_t transform(); // convert one scanline to grayscale in parallel __global__ void grayscale_transform(uchar3 *dst_img, uchar3 *src_img, int img_size) { unsigned int y = blockIdx.y * blockDim.y + threadIdx.y; unsigned int x = blockIdx.x * blockDim.x + threadIdx.x; unsigned int idx = y * SIZE + x; uchar3 rgb = src_img[idx]; int average = (rgb.x + rgb.y + rgb.z) / 3; dst_img[idx].x = average; dst_img[idx].y = average; dst_img[idx].z = average; } void host_grayscale(uchar3 *dst_img, uchar3 *src_img, int img_size) { for (int i = 0; i < SIZE * SIZE; i++) { uchar3 rgb = src_img[i]; int average = (rgb.x + rgb.y + rgb.z) / 3; dst_img[i].x = average; dst_img[i].y = average; dst_img[i].z = average; } } int main() { // genreate a dummy image int size = SIZE * SIZE; int img_size = size * sizeof(uchar3); int block_size = size / SIZE; int grid_size = size / block_size; //CImage img; uchar3 *src_img, *gray_img, srgb; src_img = (uchar3*)malloc(img_size); gray_img = (uchar3*)malloc(img_size); for (int i = 0; i < SIZE * SIZE; i++) { uchar3 src, gray; src.x = 128; src.y = 64; src.x = 256; gray.x = 0; gray.y = 0; gray.z = 0; src_img[i] = src; gray_img[i] = gray; } cudaError_t cudaStatus = transform(gray_img, src_img, img_size, block_size, grid_size, grayscale); if (cudaStatus != cudaSuccess) { fprintf(stderr, "addWithCuda failed!"); return 1; } cudaStatus = cudaDeviceReset(); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudadevicereset failed!"); return 1; } clock_t begin = clock(); host_grayscale(gray_img, src_img, img_size); clock_t end = clock(); double time_spent = 1000 * (double)(end - begin) / CLOCKS_PER_SEC; printf("CPU Execution Time: %32fms", time_spent); free(gray_img); free(src_img); return 0; system("pause"); return 0; system("pause"); } // transform an image cudaError_t transform(uchar3 *dst_img, uchar3 *src_img, int img_size, int block_size, int grid_size, color_transform_t type) { cudaError_t cudaStatus; uchar3 *t_src, *gpu_output; cudaStatus = cudaMalloc((void**)&t_src, img_size); if (cudaStatus != cudaSuccess) fprintf(stderr, "cudaMalloc failed!"); cudaStatus = cudaMalloc((void**)&gpu_output, img_size); if (cudaStatus != cudaSuccess) fprintf(stderr, "cudaMalloc failed!"); cudaStatus = cudaMemcpy(t_src, src_img, img_size, cudaMemcpyHostToDevice); if (cudaStatus != cudaSuccess) fprintf(stderr, "cudaMemcpy failed!"); float et; cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start); if (type == grayscale) grayscale_transform<<<grid_size, block_size>>>(gpu_output, t_src, img_size); cudaEventRecord(stop); cudaEventSynchronize(stop); cudaEventElapsedTime(&et, start, stop); printf("GPU Execution Time: %32fms\n", et); //// Check for any errors launching the kernel cudaStatus = cudaGetLastError(); if (cudaStatus != cudaSuccess) fprintf(stderr, "addKernel launch failed: %s\n", cudaGetErrorString(cudaStatus)); // cudaDeviceSynchronize waits for the kernel to finish, and returns // any errors encountered during the launch. cudaStatus = cudaDeviceSynchronize(); if (cudaStatus != cudaSuccess) fprintf(stderr, "cudaDeviceSynchronize returned error code %d after launching addKernel!\n", cudaStatus); //// Copy output vector from GPU buffer to host memory. cudaStatus = cudaMemcpy(dst_img, gpu_output, img_size, cudaMemcpyDeviceToHost); if (cudaStatus != cudaSuccess) fprintf(stderr, "cudaMemcpy failed!"); return cudaStatus; } //cudaError_t transform(rgb_mat dst_img, rgb_mat src_img, unsigned int img_size, color_transform_t type); //cudaError_t transform(rgb_t ***image, unsigned int line, transform_t type); // //// convert one scanline to grayscale in parallel //__global__ void grayscale_transform(rgb_list gpu_output, rgb_mat t_src, unsigned int line) //{ // unsigned int y = blockIdx.y * blockDim.y + threadIdx.y; // unsigned int x = blockIdx.x * blockDim.x + threadIdx.x; // unsigned int idx = y * 100 + x; // // rgb_ptr src_rgb = t_src[y][x]; // // rgb_ptr gpu_rgb = nullptr; // gpu_rgb->r = 0; // gpu_rgb->g = 0; // gpu_rgb->b = 0; // // double average = (src_rgb->r + src_rgb->g + src_rgb->b) / 3; // // gpu_rgb->r = average; // gpu_rgb->g = average; // gpu_rgb->b = average; // // gpu_output[idx] = gpu_rgb; //} // //int main() //{ // // genreate a dummy image // int img_size = 100 * 100 * sizeof(rgb_ptr); // int block_size = 4; // int grid_size = 100000 / block_size; // // rgb_mat src_img = (rgb_list*)malloc(100 * sizeof(rgb_list)); // rgb_mat gray_img = (rgb_list*)malloc(100 * sizeof(rgb_list)); // // for (int y = 0; y < 100; y++) // { // src_img[y] = (rgb_list)malloc(100 * sizeof(rgb_ptr)); // gray_img[y] = (rgb_list)malloc(100 * sizeof(rgb_ptr)); // for (int x = 0; x < 100; x++) // { // src_img[y][x] = (rgb_ptr)malloc(sizeof(rgb_ptr)); // gray_img[y][x] = (rgb_ptr)malloc(sizeof(rgb_ptr)); // } // } // // // int i, j; // for (i = 0; i < 100; i++) // { // for (j = 0; j < 100; j++) // { // rgb_ptr s_p = src_img[i][j]; // s_p->r = 128.0; // s_p->g = 76.0; // s_p->b = 256.0; // src_img[i][j] = s_p; // // rgb_ptr g_p = gray_img[i][j]; // g_p->r = 0.0; // g_p->g = 0.0; // g_p->b = 0.0; // gray_img[i][j] = g_p; // } // } // // cudaError_t cudaStatus = transform(gray_img, src_img, img_size, block_size, grid_size, grayscale); // if (cudaStatus != cudaSuccess) // { // fprintf(stderr, "addWithCuda failed!"); // return 1; // } // // //cudaStatus = cudaDeviceReset(); // //if (cudaStatus != cudaSuccess) // //{ // // fprintf(stderr, "cudadevicereset failed!"); // // return 1; // //} // // for (i = 0; i < 100; i++) // { // for (int j = 0; j < 100; j++) // { // rgb_ptr rgb = gray_img[i][j]; // printf("%.3f : %.3f : %.3f\n", rgb->r, rgb->g, rgb->b); // } // } // // free(gray_img); // free(src_img); // // return 0; // system("pause"); //} // //// transform an image //cudaError_t transform(rgb_mat dst_img, rgb_mat src_img, unsigned int img_size, int block_size, int grid_size, color_transform_t type) //{ // rgb_mat t_src; // rgb_list gpu_output; // // cudaError_t cudaStatus; // // cudaStatus = cudaMalloc((void**)&t_src, img_size); // if (cudaStatus != cudaSuccess) // fprintf(stderr, "cudaMalloc failed!"); // // cudaStatus = cudaMalloc((void**)&gpu_output, img_size); // if (cudaStatus != cudaSuccess) // fprintf(stderr, "cudaMalloc failed!"); // // cudaStatus = cudaMemcpy(t_src, src_img, img_size, cudaMemcpyHostToDevice); // if (cudaStatus != cudaSuccess) // fprintf(stderr, "cudaMemcpy failed!"); // // if (type == grayscale) // grayscale_transform<<<grid_size, block_size>>>(gpu_output, t_src, img_size); // // //// Check for any errors launching the kernel // cudaStatus = cudaGetLastError(); // if (cudaStatus != cudaSuccess) // fprintf(stderr, "addKernel launch failed: %s\n", cudaGetErrorString(cudaStatus)); // // // cudaDeviceSynchronize waits for the kernel to finish, and returns // // any errors encountered during the launch. // cudaStatus = cudaDeviceSynchronize(); // if (cudaStatus != cudaSuccess) // fprintf(stderr, "cudaDeviceSynchronize returned error code %d after launching addKernel!\n", cudaStatus); // // //// Copy output vector from GPU buffer to host memory. // cudaStatus = cudaMemcpy(dst_img, gpu_output, img_size, cudaMemcpyDeviceToHost); // if (cudaStatus != cudaSuccess) // fprintf(stderr, "cudaMemcpy failed!"); // // //cudaStatus = cudaFree(gpu_output); // //if (cudaStatus != cudaSuccess) // // fprintf(stderr, "cudaFree failed!"); // // return cudaStatus; //} /// HERE IS A WORKING EXAMPLE //cudaError_t addWithCuda(int *c, const int *a, const int *b, unsigned int size); // //__global__ void addKernel(int *c, const int *a, const int *b) //{ // int i = threadIdx.x; // c[i] = a[i] + b[i]; //} // //int main() //{ // const int arraySize = 5; // const int a[arraySize] = { 1, 2, 3, 4, 5 }; // const int b[arraySize] = { 10, 20, 30, 40, 50 }; // int c[arraySize] = { 0 }; // // // Add vectors in parallel. // cudaError_t cudaStatus = addWithCuda(c, a, b, arraySize); // if (cudaStatus != cudaSuccess) { // fprintf(stderr, "addWithCuda failed!"); // return 1; // } // // printf("{1,2,3,4,5} + {10,20,30,40,50} = {%d,%d,%d,%d,%d}\n", // c[0], c[1], c[2], c[3], c[4]); // // // cudaDeviceReset must be called before exiting in order for profiling and // // tracing tools such as Nsight and Visual Profiler to show complete traces. // cudaStatus = cudaDeviceReset(); // if (cudaStatus != cudaSuccess) { // fprintf(stderr, "cudaDeviceReset failed!"); // return 1; // } // // return 0; //} // //// Helper function for using CUDA to add vectors in parallel. //cudaError_t addWithCuda(int *c, const int *a, const int *b, unsigned int size) //{ // int *dev_a = 0; // int *dev_b = 0; // int *dev_c = 0; // cudaError_t cudaStatus; // // // Choose which GPU to run on, change this on a multi-GPU system. // cudaStatus = cudaSetDevice(0); // if (cudaStatus != cudaSuccess) { // fprintf(stderr, "cudaSetDevice failed! Do you have a CUDA-capable GPU installed?"); // goto Error; // } // // // Allocate GPU buffers for three vectors (two input, one output) . // cudaStatus = cudaMalloc((void**)&dev_c, size * sizeof(int)); // if (cudaStatus != cudaSuccess) { // fprintf(stderr, "cudaMalloc failed!"); // goto Error; // } // // cudaStatus = cudaMalloc((void**)&dev_a, size * sizeof(int)); // if (cudaStatus != cudaSuccess) { // fprintf(stderr, "cudaMalloc failed!"); // goto Error; // } // // cudaStatus = cudaMalloc((void**)&dev_b, size * sizeof(int)); // if (cudaStatus != cudaSuccess) { // fprintf(stderr, "cudaMalloc failed!"); // goto Error; // } // // // Copy input vectors from host memory to GPU buffers. // cudaStatus = cudaMemcpy(dev_a, a, size * sizeof(int), cudaMemcpyHostToDevice); // if (cudaStatus != cudaSuccess) { // fprintf(stderr, "cudaMemcpy failed!"); // goto Error; // } // // cudaStatus = cudaMemcpy(dev_b, b, size * sizeof(int), cudaMemcpyHostToDevice); // if (cudaStatus != cudaSuccess) { // fprintf(stderr, "cudaMemcpy failed!"); // goto Error; // } // // // Launch a kernel on the GPU with one thread for each element. // addKernel<<<1, size>>>(dev_c, dev_a, dev_b); // // // Check for any errors launching the kernel // cudaStatus = cudaGetLastError(); // if (cudaStatus != cudaSuccess) { // fprintf(stderr, "addKernel launch failed: %s\n", cudaGetErrorString(cudaStatus)); // goto Error; // } // // // cudaDeviceSynchronize waits for the kernel to finish, and returns // // any errors encountered during the launch. // cudaStatus = cudaDeviceSynchronize(); // if (cudaStatus != cudaSuccess) { // fprintf(stderr, "cudaDeviceSynchronize returned error code %d after launching addKernel!\n", cudaStatus); // goto Error; // } // // // Copy output vector from GPU buffer to host memory. // cudaStatus = cudaMemcpy(c, dev_c, size * sizeof(int), cudaMemcpyDeviceToHost); // if (cudaStatus != cudaSuccess) { // fprintf(stderr, "cudaMemcpy failed!"); // goto Error; // } // //Error: // cudaFree(dev_c); // cudaFree(dev_a); // cudaFree(dev_b); // // return cudaStatus; //}
2,355
#include <stdio.h> #include <iostream> #include <cstdlib> #include <limits.h> #include <algorithm> #include <sys/time.h> #include <cuda_runtime.h> using namespace std; #define INF INT_MAX-1 __global__ void FloydWarshall(int k, int i, float *matrix, int n) { int col = blockIdx.x * blockDim.x + threadIdx.x; /* This thread’s matrix column */ if(col >= n) return; int arrayIndex = n * i + col; // Calculating D[i][j] float matrix_value_i_k = matrix[n * i + k]; // Calculating D[i][k] float matrix_value_k_j = matrix[k*n + col]; /* this column, kth row */ matrix[arrayIndex] = fmin(matrix[arrayIndex], matrix_value_i_k + matrix_value_k_j); } int main(int argc, char *argv[]) { char *arg_vertices = getenv("N_VERTICES"); char *arg_threads_per_block = getenv("N_THREADS"); size_t vertices = atoi(arg_vertices); int threads_per_block = atoi(arg_threads_per_block); float *host_matrix =(float *)malloc( vertices * vertices * sizeof(float)); for(int i = 0 ; i < vertices ; i++ ) { for(int j = 0 ; j< vertices; j++ ) { if( i == j ) host_matrix[i * vertices + j] = 0; else { int num = i + j; if (num % 3 == 0) host_matrix[i * vertices + j] = num / 2; else if (num % 2 == 0) host_matrix[i * vertices + j] = num * 2; else host_matrix[i * vertices + j] = num; } } } size_t tot = vertices * vertices * sizeof(float); float *device_matrix = NULL; cudaMalloc((float **)&device_matrix, tot); cudaMemcpy(device_matrix, host_matrix, tot, cudaMemcpyHostToDevice); int blocks_per_grid = vertices + (threads_per_block - 1) /threads_per_block; struct timeval tvalBefore, tvalAfter; gettimeofday (&tvalBefore, NULL); for(int via = 0; via < vertices; via++) { for(int j = 0; j < vertices; j++){ FloydWarshall<<<blocks_per_grid, threads_per_block>>>(via, j, device_matrix, vertices); cudaThreadSynchronize(); } } gettimeofday (&tvalAfter, NULL); printf("Time: %ld microseconds\n", ((tvalAfter.tv_sec - tvalBefore.tv_sec)*1000000L +tvalAfter.tv_usec) - tvalBefore.tv_usec ); float *result_matrix =(float *)malloc( vertices * vertices * sizeof(float)); cudaMemcpy(result_matrix, device_matrix, tot, cudaMemcpyDeviceToHost); return 0; }
2,356
#include <stdio.h> #include <stdlib.h> #include "cuda_runtime.h" #include "device_launch_parameters.h" __global__ void hello_kernel() { printf("Hello World from Thread %d\n", threadIdx.x); } int main(int argc, char *argv[]) { //set the CUDA device to the default CUDA GPU (device 0) cudaError result = cudaSetDevice(0); if (result != cudaSuccess ){ printf("Error setting default GPU device.\n"); } //call a CUDA kernel dim3 blocksPerGrid(1, 1, 1); dim3 threadsPerBlock(10, 1, 1); hello_kernel<<<blocksPerGrid, threadsPerBlock>>>(); //synchronise cudaDeviceSynchronize(); return 0; }
2,357
#include <iostream> #include <stdlib.h> #include <stdio.h> #include <time.h> #include <unistd.h> #include <cuda.h> #include <cuda_runtime.h> //extern __device__ int testxyz[1000]; //int localtrace[10000]; //__device__ float* tracehandle; __device__ float foo_CC(float a) { return a*0.9; } __device__ int foo_DD(float a) { if (threadIdx.x < 2 || threadIdx.y > 2) return (int) a; else return a+2; } __device__ float foo_BB(float a) { if (threadIdx.x > 3 || threadIdx.y > 11) return a + foo_CC(a); else return a + (float)foo_DD(a) /2; } __device__ float foo_AA( float a, float b) { if (threadIdx.x < 8 || threadIdx.y > 4) return a*3.1415+1; else return (b*a)*0.5 + foo_BB(b); } __global__ void axpy_kernel2(float a, float* x, float* y) { //tracehandle = newbu; int blockId = blockIdx.x + blockIdx.y * gridDim.x; int threadId = blockId * (blockDim.x * blockDim.y) + (threadIdx.y * blockDim.x) + threadIdx.x; int index = threadId; float aa = y[index] + x[index] + 1.1; float b = 0.5*y[index] + 0.25* x[index] + 1.0; y[index] += ( x[index]*1.67 + foo_AA(aa, b) ); // y[index] += ( x[index]*1.67 + aa + b ); } __global__ void axpy_kernel1(float a, float* x, float* y) { int blockId = blockIdx.x + blockIdx.y * gridDim.x; int threadId = blockId * (blockDim.x * blockDim.y) + (threadIdx.y * blockDim.x) + threadIdx.x; int index = threadId; y[index] = x[index]*0.3; if (index>2) y[index] += 99; else y[index] += 999 + foo_CC(a); } int main(int argc, char* argv[]) { //const int kDataLen2 = 128; float a = 2.0f; //int blocks2 = 600; cudaSetDevice(0); if (argc != 5) { printf("usage: ./axpy [blocks_x] [blocks_y] [threads_x] [threads_y]\n"); exit(1); } int blocksx = atoi(argv[1]) ; int blocksy = atoi(argv[2]) ; int kDataLenx = atoi(argv[3]); int kDataLeny = atoi(argv[4]); int sizen = blocksx *blocksy *kDataLenx *kDataLeny; // cudaThreadSetLimit(cudaLimitMallocHeapSize, 1024*1024); //sderek cudaDeviceSetLimit(cudaLimitMallocHeapSize, 1024*1024*500); //sderek // tracetest = (int*)malloc( 1234); // float host_y[blocks*kDataLen]; // float host_y[blocks*kDataLen]; float* host_x = (float*) malloc( sizen* sizeof(float)); float* host_y = (float*) malloc( sizen* sizeof(float)); void* host_newbu = (void*) malloc( 1000 ); int ii; for( ii=0; ii<sizen; ii++) host_x[ii] = ii%8; for( ii=0; ii<sizen; ii++) host_y[ii] = ii%5; /* int x[5]; x[0] = 13; printf("%p\n",x); printf("%p\n",&x); printf("%d\n",*x); printf("%d\n",*(x+1)); */ // Copy input data to device. float* device_x; float* device_y; // void* newbu; // printf(" %p\n", device_x); cudaMalloc((void**)&device_x, sizen * sizeof(float)); // printf(" %p\n", device_x); // printf(" %p\n", device_y); cudaMalloc((void**)&device_y, sizen * sizeof(float) + 18); // printf(" %p\n", device_y); // printf(" %p\n", newbu); // cudaMalloc(&newbu, 1000); // printf(" %p\n", newbu); /* std::cout << &(device_x) << "\n"; std::cout << &(device_y) << "\n"; std::cout << &(*device_x) << "\n"; std::cout << &(*device_y) << "\n"; std::cout << (device_x) << "\n"; std::cout << (device_y) << "\n"; */ cudaMemcpy(device_x, host_x, sizen * sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(device_y, host_y, sizen * sizeof(float), cudaMemcpyHostToDevice); dim3 CTAs(blocksx, blocksy); dim3 Threads(kDataLenx, kDataLeny); std::cout << "launching kernel...\n"; axpy_kernel1<<<CTAs, Threads>>>(a, device_x, device_y); cudaDeviceSynchronize(); cudaError_t error = cudaGetLastError(); if (error != cudaSuccess) { printf ("CUDA error: %s\n", cudaGetErrorString(error)); exit(-1); } axpy_kernel2<<<CTAs, Threads>>>(a, device_x, device_y); cudaDeviceSynchronize(); error = cudaGetLastError(); if (error != cudaSuccess) { printf ("CUDA error: %s\n", cudaGetErrorString(error)); exit(-1); } cudaMemcpy(host_y, device_y, sizen* sizeof(float), cudaMemcpyDeviceToHost); // cudaMemcpy(host_newbu, newbu, 1000, cudaMemcpyDeviceToHost); free(host_newbu); // cudaFree(newbu); int verify = 0; for (int ii = 0; ii < 8; ii++) std::cout << "y[" << ii << "] = " << host_y[ii] << "\n"; for (int ii = 0; ii < sizen; ii++) { if( host_y[ii] == ii%5) verify ++; // std::cout << "y[" << i << "] = " << host_y[i] << "\n"; } std::cout << "\n\n[TOOL verify] There are a total of\t" << verify << " incorrect numbers." << std::endl; if (verify==0) std::cout << "[TOOL verify] passed!" << std::endl << std::endl; // for (int i = 0; i < 20; ++i) { // std::cout << "newtrace [" << i << "] = " << host_newbu[i] << "\n"; // std::cout << & (host_y[i] )<< "\n"; // } /* cudaMemcpyFromSymbol(localtrace, testxyz, 40*sizeof(int), 0, cudaMemcpyDeviceToHost); for (int i = 0; i < 20; ++i) printf("%d\t", localtrace[i] ); std::cout << std::endl; cudaMemcpyFromSymbol(localtrace+8, testxyz, 40*sizeof(int), 0, cudaMemcpyDeviceToHost); for (int i = 0; i < 20; ++i) printf("%d\t", localtrace[i] ); std::cout << std::endl; */ // int* show_h; // cudaMemcpyFromSymbol(show_h, show, sizeof(int), 0, cudaMemcpyDeviceToHost); // msg = cudaGetSymbolAddress((void **)&d_G, test); // printf("the address is %p\n", d_G); // if (msg == cudaSuccess) // { // int tmp[4]; // printf("before %d %d %d %d@ %p\n", *tmp, *(tmp+1), *(tmp+2), *(tmp+3), tmp); // cudaMemcpyFromSymbol(tracetest, test1, 4*sizeof(int), 0, cudaMemcpyDeviceToHost); // cudaMemcpyFromSymbol(tmp, test2, 4*sizeof(int), 0, cudaMemcpyDeviceToHost); // printf("copy %d %d %d %d@ %p\n", *tmp, *(tmp+1), *(tmp+2), *(tmp+3), tmp); // cudaMemcpyFromSymbol(tmp, test2, 4*sizeof(int), 0, cudaMemcpyDeviceToHost); // printf("after %d %d %d %d@ %p\n", tmp[0], tmp[1], tmp[2], tmp[3], tmp); // } //else // std::cout << cudaGetErrorString(msg) << "\n\n"; cudaFree(device_x); cudaFree(device_y); cudaDeviceReset(); return 0; }
2,358
//xfail:BOOGIE_ERROR //--blockDim=1024 --gridDim=1 --no-inline //error: possible null pointer access #include <stdio.h> #include <assert.h> #include <cuda.h> #define N 2//8 #define tid (blockIdx.x * blockDim.x + threadIdx.x) __device__ float multiplyByTwo(float *v, unsigned int index) { return v[index] * 2.0f; } __device__ float divideByTwo(float *v, unsigned int index) { return v[index] * 0.5f; } typedef float(*funcType)(float*, unsigned int); __global__ void foo(float *v) { funcType f = (funcType)3; // it's a null pointer v[tid] = (*f)(v, tid); }
2,359
#include<stdio.h> #include<stdlib.h> #include<curand_kernel.h> #include<curand.h> #include<sys/time.h> unsigned int NUM_PARTICLES = 1000000; unsigned int NUM_ITERATIONS = 10; unsigned int BLOCK_SIZE = 192; //unsigned int GRID_SIZE = ((NUM_PARTICLES/BLOCK_SIZE) + 1); unsigned int NUM_STREAMS = 10; typedef struct { float3 posId; }position; typedef struct { float3 velId; }velocity; typedef struct { position pos; velocity vel; }Particle; void fill_data(Particle *p) { for(int i=0; i< NUM_PARTICLES; i++) { p[i].pos.posId.x = 10*((float)rand()/RAND_MAX); p[i].pos.posId.y = 10*((float)rand()/RAND_MAX); p[i].pos.posId.z = 10*((float)rand()/RAND_MAX); p[i].vel.velId.x = 100*((float)rand()/RAND_MAX); p[i].vel.velId.y = 100*((float)rand()/RAND_MAX); p[i].vel.velId.z = 100*((float)rand()/RAND_MAX); } } __global__ void particle_kernel_per_iteration(Particle *p, int offset, int streamSize) { int i = (blockIdx.x*blockDim.x)+threadIdx.x; if(i < streamSize) { p[offset + i].pos.posId.x += p[offset + i].vel.velId.x; p[offset + i].pos.posId.y += p[offset + i].vel.velId.y; p[offset + i].pos.posId.z += p[offset + i].vel.velId.z; } __syncthreads(); } void update_velocity_position_in_gpu(Particle *p) { struct timeval start_time; struct timeval stop_time; Particle *gPar = NULL; cudaMalloc(&gPar, NUM_PARTICLES * sizeof(Particle)); unsigned long streamSize = NUM_PARTICLES/NUM_STREAMS; unsigned long streamBytes = streamSize * sizeof(Particle); cudaStream_t stream[NUM_STREAMS]; for(int i=0; i<NUM_STREAMS; i++) cudaStreamCreate(&stream[i]); //Start time gettimeofday(&start_time, NULL); #ifdef TYPE1 for(int i=0; i<NUM_ITERATIONS; i++) { for(int s=0; s<NUM_STREAMS; s++) { unsigned long offset = s * streamSize; // Copy Data to GPU Memory Asynchronously cudaMemcpyAsync(&gPar[offset], &p[offset], streamBytes, cudaMemcpyHostToDevice, stream[s]); //Launch kernel particle_kernel_per_iteration<<<((streamSize/BLOCK_SIZE) + 1), BLOCK_SIZE, 0, stream[s]>>>(gPar, offset, streamSize); //cudaDeviceSynchronize(); //Copy Data back to Host cudaMemcpyAsync(&p[offset], &gPar[offset], streamBytes, cudaMemcpyDeviceToHost, stream[s]); } cudaDeviceSynchronize(); //Update Velocity in Host before copying data to GPU Memory for(int j=0; j<NUM_PARTICLES;j++) { p[j].vel.velId.x = 100*((float)rand()/RAND_MAX); p[j].vel.velId.y = 100*((float)rand()/RAND_MAX); p[j].vel.velId.z = 100*((float)rand()/RAND_MAX); } } #else for(int i=0; i<NUM_ITERATIONS; i++) { for(int s=0; s<NUM_STREAMS; s++) { unsigned long offset = s * streamSize; // Copy Data to GPU Memory Asynchronously cudaMemcpyAsync(&gPar[offset], &p[offset], streamBytes, cudaMemcpyHostToDevice, stream[s]); } for(int s=0; s<NUM_STREAMS; s++) { unsigned long offset = s * streamSize; //Launch kernel particle_kernel_per_iteration<<<((streamSize/BLOCK_SIZE) + 1), BLOCK_SIZE, 0, stream[s]>>>(gPar, offset, streamSize); } for(int s=0; s<NUM_STREAMS; s++) { unsigned long offset = s * streamSize; //Copy Data back to Host cudaMemcpyAsync(&p[offset], &gPar[offset], streamBytes, cudaMemcpyDeviceToHost, stream[s]); } cudaDeviceSynchronize(); //Update Velocity in Host before copying data to GPU Memory for(int j=0; j<NUM_PARTICLES;j++) { p[j].vel.velId.x = 100*((float)rand()/RAND_MAX); p[j].vel.velId.y = 100*((float)rand()/RAND_MAX); p[j].vel.velId.z = 100*((float)rand()/RAND_MAX); } } #endif //Stop time gettimeofday(&stop_time, NULL); for(int i=0; i<NUM_STREAMS; i++) cudaStreamDestroy(stream[i]); cudaFree(gPar); printf("Total time of Execution in GPU: %ld msec\n\n", ((stop_time.tv_sec*1000000 + stop_time.tv_usec)-(start_time.tv_sec*1000000 + start_time.tv_usec))/1000); } int main(int argc, char *argv[]) { if(argc != 3) { printf("No. of arguments to be passed should be 2 i.e. 1st as NUM_PARTICLES and 2nd as NUM_STREAMS\n"); exit(1); } NUM_PARTICLES = atoi(argv[1]); NUM_STREAMS = atoi(argv[2]); Particle *par = NULL; cudaMallocHost(&par, NUM_PARTICLES*sizeof(Particle)); fill_data(par); update_velocity_position_in_gpu(par); cudaFree(par); return 0; }
2,360
#include "includes.h" __global__ void kernel_1(int columns, int rows, float* mat1, float* matanswer) { int columna = threadIdx.x; //En que columna operamos (no filas) float temp_value = 0; for (int k = 0; k < rows; k++) { temp_value = temp_value + mat1[(k * columns) + columna]; } matanswer[columna] = temp_value; }
2,361
#include <iostream> #include <sys/time.h> #define N 16 __global__ void add(int *a, int *b, int *c) { int i = blockIdx.x; if(i < N) c[i] = a[i] + b[i]; } void add_host(int *a, int *b, int *c) { for(int i = 0; i < N; i++) { c[i] = a[i] + b[i]; } } int main (void) { // variables to store host and device data int a[N], b[N], c[N]; int *dev_a, *dev_b, *dev_c; // allocate memory on the GPU cudaMalloc( (void **) &dev_a, N * sizeof(int) ); cudaMalloc( (void **) &dev_b, N * sizeof(int) ); cudaMalloc( (void **) &dev_c, N * sizeof(int) ); // fill the arrays with data for(int i = 0; i < N; i++) { a[i] = -i; b[i] = i * i; } // copy arrays a and b to the device cudaMemcpy( dev_a, a, N * sizeof(int), cudaMemcpyHostToDevice ); cudaMemcpy( dev_b, b, N * sizeof(int), cudaMemcpyHostToDevice ); // print the current time struct timeval tv; struct timezone tz; struct tm *tm; gettimeofday(&tv, &tz); tm = localtime(&tv.tv_sec); printf(" %d:%02d:%02d %d \n", tm->tm_hour, tm->tm_min, tm->tm_sec, tv.tv_usec); // Do the addition operation add<<<N,1>>>( dev_a, dev_b, dev_c ); //add_host((int *)a, (int *)b, (int *)c); // print current time gettimeofday(&tv, &tz); tm = localtime(&tv.tv_sec); printf(" %d:%02d:%02d %d \n", tm->tm_hour, tm->tm_min, tm->tm_sec, tv.tv_usec); // copy the array 'c' back from the device cudaMemcpy(c, dev_c, N * sizeof(int), cudaMemcpyDeviceToHost ); // display the results for(int i = 0; i<N; i++) { printf("%d + %d = %d\n", a[i], b[i], c[i]); } // free the memory used on the device cudaFree( dev_a ); cudaFree( dev_b ); cudaFree( dev_c ); return 0; }
2,362
#include <stdio.h> #define FIRST_CHAR 32 #define LAST_CHAR 128 #define NBR 96 __global__ void histo_kernel(unsigned char *buffer,long size, unsigned int *histo){ __shared__ unsigned int temp[256]; temp[threadIdx.x]=0; int i = threadIdx.x + blockIdx.x *blockDim.x; int offset = blockDim.x *gridDim.x; while(i<size){ atomicAdd(&(histo[buffer[i]]),1); i+=offset; } __syncthreads(); atomicAdd( &(histo[threadIdx.x]), temp[threadIdx.x] ); } int main(int argc, char *argv[]){ // unsigned char *buffer = (unsigned char *) big_random_block(SIZE); if(argc <= 2){ fprintf(stderr, "Arguments non valide"); return 1; } FILE *f_input; FILE *f_output; long lSize; char *buffer; f_input = fopen ( argv[1] , "r" ); f_output = fopen( argv[2],"w"); if( !f_input ) perror(argv[1]),exit(1); fseek( f_input , 0L , SEEK_END); lSize = ftell( f_input ); rewind( f_input ); printf("The size is : %li", lSize); //buffer = calloc( 1, lSize+1 ); buffer =(char*) malloc(lSize); if( !buffer ) fclose(f_input),fputs("memory alloc fails",stderr),exit(1); if( 1!=fread( buffer , lSize, 1 , f_input) ) fclose(f_input),free(buffer),fputs("entire read fails",stderr),exit(1); /*Create event for co;pute running time*/ cudaEvent_t start, stop; cudaEventCreate( &start ); cudaEventCreate( &stop ); /*Launch event to specify the start of running*/ cudaEventRecord( start, 0); /*allocate device memory*/ unsigned char *dev_buffer; unsigned int *dev_histo; /*Give space in Global memory of GPU to store different variable*/ cudaMalloc( (void**)&dev_buffer, lSize); /*Copy from CPU Global memory to GPU Global memory*/ cudaMemcpy( dev_buffer, buffer, lSize, cudaMemcpyHostToDevice ); /*Create space for histo variable and initialize at 0 each slopt*/ cudaMalloc( (void**)&dev_histo, 256 * sizeof( long)); cudaMemset( dev_histo, 0, 256 * sizeof( int )); cudaDeviceProp prop; cudaGetDeviceProperties( &prop, 0 ); int blocks = prop.multiProcessorCount; histo_kernel<<<blocks*2,256>>>( dev_buffer, lSize, dev_histo ); /*Define histo vqriqble and copy on GPU global memory*/ unsigned int histo[256]; cudaMemcpy( histo, dev_histo,256 * sizeof( int ),cudaMemcpyDeviceToHost); for(int i =FIRST_CHAR;i< LAST_CHAR;i++){ printf("%c:%d\n",i,histo[i]); fprintf(f_output, "%c:%d\n",i,histo[i]); } /*Get event at the end of loop*/ cudaEventRecord( stop, 0 ); cudaEventSynchronize( stop ); float elapsedTime; cudaEventElapsedTime( &elapsedTime, start, stop ); printf( "Time to generate: %3.1f ms\n", elapsedTime ); /*Destroy event for running time*/ cudaEventDestroy( start ); cudaEventDestroy( stop ); /*Free memory and close the files**/ cudaFree( dev_histo ); cudaFree( dev_buffer ); fclose(f_input); fclose(f_output); free( buffer ); return 0; }
2,363
#include <cuda_runtime.h> #include <cuda.h> #include <stdio.h> #include <stdlib.h> #include <math.h> // needed for the function sqrtf() #define TILE_SIZE 32 // NB // Block SIZE /* * Function to perform rank-k update * half of the threads working */ __device__ void ssyrk_tile(float* rA1, float* rA2) { int row = blockIdx.y * TILE_SIZE + threadIdx.y; int column = blockIdx.x * TILE_SIZE + threadIdx.x; if(column <= row) { float updatedValue = rA2[row * TILE_SIZE + column]; for(int k=0; k<TILE_SIZE; k++) { updatedValue -= rA1[row * TILE_SIZE + k] * rA1[column * TILE_SIZE + k]; } rA2[row * TILE_SIZE + column] = updatedValue; } } /* * Function to perform general matrix multiplication * DOUBT: I think calculation is given wrong in paper it should be rA2[k][n] */ __device__ void sgemm_tile(const float* rA1, const float* rA2, float* rA3) { int row = blockIdx.y * TILE_SIZE + threadIdx.y; int column = blockIdx.x * TILE_SIZE + threadIdx.x; float updatedValue = rA3[row * TILE_SIZE + column]; for(int i=0; i<TILE_SIZE; i++) { updatedValue -= rA1[row * TILE_SIZE + i] * rA2[i * TILE_SIZE + column]; } rA3[row * TILE_SIZE + column] = updatedValue; } /* * Function to store full tile from shared memory to global memory */ __device__ void store_full(const float* s_data, float* g_data) { int g_row = blockIdx.y * TILE_SIZE + threadIdx.y; int g_column = blockIdx.x * TILE_SIZE + threadIdx.x; int l_row = threadIdx.y; int l_column = threadIdx.x; g_data[g_row * TILE_SIZE + g_column] = s_data[l_row * TILE_SIZE + l_column]; __syncthreads(); } /* * Function to store lower triangular tile from shared memory to global memory */ __device__ void store_lower(const float* s_data, float* g_data) { int g_row = blockIdx.y * TILE_SIZE + threadIdx.y; int g_column = blockIdx.x * TILE_SIZE + threadIdx.x; int l_row = threadIdx.y; int l_column = threadIdx.x; if(l_column <= l_row) g_data[g_row * TILE_SIZE + g_column] = s_data[l_row * TILE_SIZE + l_column]; else g_data[g_row * TILE_SIZE + g_column] = 0; __syncthreads(); } /* * Function to perform Choleshky Factorization for a tile */ __device__ void spotrf_tile(float* t_A) { int ty = blockIdx.x*blockDim.x + threadIdx.x; // col int tx = blockIdx.y*blockDim.y + threadIdx.y; // row for(int k{0};k<TILE_SIZE;k++){ // square root of diagonal elements if(tx==0 && ty==0) t_A[k*TILE_SIZE + k] = sqrtf(t_A[k*TILE_SIZE + k]); __syncthreads(); // division step done parallaly if(ty<=tx && tx<TILE_SIZE - 1 && ty<TILE_SIZE - 1 && ty == k) { t_A[(tx+1)*TILE_SIZE + k]/= t_A[k*TILE_SIZE + k]; } __syncthreads(); if(ty<=tx && tx<TILE_SIZE - 1 && ty<TILE_SIZE - 1 && ty >= k) { t_A[(tx+1)*TILE_SIZE + (ty+1)]-= t_A[(tx+1)*TILE_SIZE + k]*t_A[(ty+1)*TILE_SIZE + k]; } __syncthreads(); } } /* * Function to perform triangular solve for a tile */ __device__ void strsm_tile(float *t_A1, float *t_A2) { // t_A2 is current unkonown int ty = blockIdx.x*TILE_SIZE + threadIdx.x; int tx = blockIdx.y*TILE_SIZE + threadIdx.y; for(int i{0};i<TILE_SIZE;i++){ if(ty==0){ t_A2[tx*TILE_SIZE + i]/= t_A1[i*TILE_SIZE + i]; } __syncthreads(); if(ty>i && i<TILE_SIZE-1) { t_A2[tx*TILE_SIZE+ty]-= t_A2[tx*TILE_SIZE + i]*t_A1[ty*TILE_SIZE + i]; } __syncthreads(); } } /* * Function to load a full tile from global memory to shared memory */ __device__ void load_full(float *t_A,float * S_A) { // assigning a 2-D array in shared memory int g_ty = blockIdx.x*blockDim.x + threadIdx.x; // col int g_tx = blockIdx.y*blockDim.y + threadIdx.y; // row int l_tx = threadIdx.x; int l_ty = threadIdx.y; if(l_tx<TILE_SIZE && l_ty<TILE_SIZE) S_A[l_tx * TILE_SIZE + l_ty] = t_A[g_tx*TILE_SIZE + g_ty]; __syncthreads(); }
2,364
#include <cuda.h> #include <stdio.h> #include <time.h> #include <stdlib.h> // kernel __global__ void convolution_1D_Kernel(float* d_m, float* d_mask, float* d_n, size_t length, size_t maskLength) { // indexing variables int i = blockIdx.x * blockDim.x + threadIdx.x; int m_index = i - maskLength / 2; // thread boundary check if(i < length) { for(int j = 0; j < maskLength; ++j) { if(m_index + j >= 0 && m_index + j < length) { d_n[i] += d_m[m_index + j] * d_mask[j]; } } } } // CUDA error checking void errorCheck(unsigned int line) { cudaError_t cudaError = cudaGetLastError(); if(cudaError != cudaSuccess) { printf("CUDA error in line %u in file %s: %s\n", line - 1, __FILE__, cudaGetErrorString(cudaError)); exit(EXIT_FAILURE); } } // host function containing kernel call void convolution_1D(float* m, float* mask, float* n, size_t length, size_t maskLength) { dim3 numOfBlocks(ceil(length / 1024.0), 1, 1); dim3 numOfThreads(1024, 1, 1); size_t bytes_m = length * sizeof(float); size_t bytes_mask = maskLength * sizeof(float); size_t bytes_n = length * sizeof(float); float* d_m; float* d_mask; float* d_n; cudaMalloc((void**) &d_m, bytes_m); errorCheck(__LINE__); cudaMalloc((void**) &d_mask, bytes_mask); errorCheck(__LINE__); cudaMalloc((void**) &d_n, bytes_n); errorCheck(__LINE__); cudaMemcpy(d_m, m, bytes_m, cudaMemcpyHostToDevice); errorCheck(__LINE__); cudaMemcpy(d_mask, mask, bytes_mask, cudaMemcpyHostToDevice); errorCheck(__LINE__); convolution_1D_Kernel<<<numOfBlocks, numOfThreads>>>(d_m, d_mask, d_n, length, maskLength); errorCheck(__LINE__); cudaMemcpy(n, d_n, bytes_n, cudaMemcpyDeviceToHost); errorCheck(__LINE__); cudaFree(d_m); errorCheck(__LINE__); cudaFree(d_mask); errorCheck(__LINE__); cudaFree(d_n); errorCheck(__LINE__); } int main() { struct timespec start, end; srand(time(NULL)); size_t length = rand() % 1048577 + 15728640; size_t maskLength = 121; float* m = (float*) malloc(length * sizeof(float)); float* mask = (float*) malloc(maskLength * sizeof(float)); float* n = (float*) malloc(length * sizeof(float)); for(int i = 0; i < length; ++i) { m[i] = rand() % 129 - 64; } for(int j = 0; j < maskLength; ++j) { mask[j] = rand() % 1001 / 1000.0; } clock_gettime(CLOCK_REALTIME, &start); // do convolution convolution_1D(m, mask, n, length, maskLength); clock_gettime(CLOCK_REALTIME, &end); time_t execTime = (end.tv_sec - start.tv_sec) * 1000000 + (end.tv_nsec - start.tv_nsec) / 1000; printf("Execution time: %d microseconds.", execTime); return 0; }
2,365
#include <stdio.h> //compilar: nvcc matrizMultiplicacao.cu -o matrizMultiplicacao //for i in `seq 1 10`; do ./matrizMultiplicacao; done #define N 2048 #define B 32 __global__ void matrix_multi(float *a, float *b, float *c) { int y = blockIdx.x * blockDim.x + threadIdx.x; int x = blockIdx.y * blockDim.y + threadIdx.y; int i; float soma = 0.0; if(x < N && y < N){ for(i=0; i<N; i++){ soma += a[x * N + i ] * b[y + N * i]; } c[x + y * N] = soma; } } int main() { float *a, *b, *c; float *d_a, *d_b, *d_c; int size = N; dim3 dimen (B, B); cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaMalloc( (void **) &d_a, size*size*sizeof(float) ); cudaMalloc( (void **) &d_b, size*size*sizeof(float) ); cudaMalloc( (void **) &d_c, size*size*sizeof(float) ); a = (float *)malloc( size*size*sizeof(float) ); b = (float *)malloc( size*size*sizeof(float) ); c = (float *)malloc( size*size*sizeof(float) ); for( int i = 0; i < N*N; i++ ) { a[i] = b[i] = i; c[i] = 0; } cudaEventRecord(start); cudaMemcpy( d_a, a, size*size*sizeof(float), cudaMemcpyHostToDevice ); cudaMemcpy( d_b, b, size*size*sizeof(float), cudaMemcpyHostToDevice ); dim3 grade ((N + B-1)/B, (N + B-1)/B); matrix_multi<<<grade, dimen>>>( d_a, d_b, d_c ); cudaMemcpy( c, d_c, size*size*sizeof(float), cudaMemcpyDeviceToHost ); cudaEventRecord(stop); cudaEventSynchronize(stop); float milliseconds = 0; cudaEventElapsedTime(&milliseconds, start, stop); printf("%f\n", milliseconds/1000.0); //printf( "c[0] = %lf\n", c[0] ); //printf( "c[%d] = %lf\n",N*N, c[N*N-1] ); /* int i; for(i=0; i<N*N; i++){ printf( "c[%d] = %lf\n",i, c[i] ); } */ free(a); free(b); free(c); cudaFree( d_a ); cudaFree( d_b ); cudaFree( d_c ); return 0; } /* end main */
2,366
#include<stdio.h> #include<cuda_runtime.h> #include "device_launch_parameters.h" #include <ctime> #include <iostream> using namespace std; #define BLOCK 16 #define WIDTH 1024 float* d_A, * d_B, * d_C; __global__ void d_multiply0(float* A, float* B, float* C) { unsigned int r = blockDim.y * blockIdx.y + threadIdx.y; unsigned int c = blockDim.x * blockIdx.x + threadIdx.x; unsigned int i; float tmp; tmp = 0.0f; for (i = 0; i < WIDTH; i++) { tmp += A[WIDTH * r + i] * B[WIDTH * i + c]; } C[WIDTH * r + c] = tmp; } float ha_A[WIDTH * WIDTH]; float ha_B[WIDTH * WIDTH]; float ha_C[WIDTH * WIDTH]; void d_multiply() { clock_t start = clock(); size_t size = sizeof(float) * WIDTH * WIDTH; cudaMemcpy(d_A, ha_A, size, cudaMemcpyHostToDevice); cudaMemcpy(d_B, ha_B, size, cudaMemcpyHostToDevice); dim3 grid(WIDTH / BLOCK, WIDTH / BLOCK); dim3 block(BLOCK, BLOCK); d_multiply0 << < grid, block >> > (d_A, d_B, d_C); cudaMemcpy(ha_C, d_C, size, cudaMemcpyDeviceToHost); cudaFree(d_A); cudaFree(d_B); cudaFree(d_C); clock_t end = clock(); cout << "CUDA計算時間:" << (double)(end - start) << endl; } float* d_process() { unsigned int i; cudaMalloc((void**)&d_A, sizeof(float) * WIDTH * WIDTH); cudaMalloc((void**)&d_B, sizeof(float) * WIDTH * WIDTH); cudaMalloc((void**)&d_C, sizeof(float) * WIDTH * WIDTH); for (i = 0; i < WIDTH * WIDTH; i++) { ha_A[i] = (float)i; ha_B[i] = (float)i; } d_multiply(); return &ha_C[0]; }
2,367
#include "includes.h" __global__ void cunn_MSECriterion_updateGradInput_kernel(float *gradInput, float *input, float *target, float norm, int nframe, int dim) { int k = blockIdx.x; float *gradInput_k = gradInput + k*dim; float *input_k = input + k*dim; float *target_k = target + k*dim; int i_start = threadIdx.x; int i_end = dim; int i_step = blockDim.x; // gradInput for (int i=i_start; i<i_end; i+=i_step) gradInput_k[i] = norm*(input_k[i] - target_k[i]); }
2,368
__global__ void doublify(float *a) { int idx = threadIdx.x + threadIdx.y*4; a[idx] *= 2; }
2,369
#include <stdio.h> #include <stdio.h> #include <stdlib.h> #include <math.h> #include <time.h> // problem parameters const double a = 1.; const double b = 1.; const int nx = 1024; //number of node points along y const int ny = 1024; //number of node points along x //convergence parameters double tol = 1e-4; int iter_max = 1000; double sol_ref(double x, double y) { return sin(M_PI*x)*exp(-M_PI*y); } void discretisation(double *x, double *y) { double dx = a/nx; // spatial step along x double dy = b/ny; // spatial step along y for (int i=0; i<=nx; i++) { x[i] = i*dx; y[i] = i*dy; } } void boundary(double *T, double *x, double *y) { /*Boundary conditions along the x axis for all processes */ for ( int i=0; i<=nx; i++) { T[i*nx] = sol_ref(x[i], y[0]); T[i*ny+ny] = sol_ref(x[i], y[ny]); } /*Boundary conditions along the y axis for all processes */ for ( int j=0; j<=ny; j++) { T[j] = sol_ref(x[0], y[j]); T[nx*ny+j] = sol_ref(x[nx], y[j]); } } void laplace2d(double *T, double *Tnew, double *error) { for( int j = 1; j <= nx-1; j++) { for( int i = 1; i <= ny-1; i++) { Tnew[j*nx + i] = 0.25 * ( T[j*nx + (i+1)] + T[j*nx + (i-1)] + T[(j-1)*nx + i] + T[(j+1)*nx + i] ); *error = fmax(*error, fabs(Tnew[j*nx + i] - T[j*nx + i])); } } for( int j = 1; j <= nx-1; j++) { for( int i = 1; i <= ny-1; i++) { T[j*nx + i] = Tnew[j*nx + i]; } } } int main(int argc,char **argv) { double tol = 1e-4; int iter = 0; double error; double *T = (double*) malloc(sizeof(double) * (nx+1) * (ny+1)); double *Tnew = (double*) malloc(sizeof(double) * (nx+1) * (ny+1)); double *x = (double*) malloc(nx * sizeof(double)); double *y = (double*) malloc(ny * sizeof(double)); if(!x || !y || !T || !Tnew ) return 0; discretisation(x, y); boundary(T, x, y); while (iter < iter_max ) { error = 0.0; laplace2d(T, Tnew, &error); if (iter % 100 == 0 ) printf("%d, %0.6f\n", iter, error); if (error < tol) break; iter++; } free(T); free(Tnew); free(x); free(y); return 0; }
2,370
#include "includes.h" __device__ unsigned int getGid3d3d(){ int blockId = blockIdx.x + blockIdx.y * gridDim.x + gridDim.x * gridDim.y * blockIdx.z; int threadId = blockId * (blockDim.x * blockDim.y * blockDim.z) + (threadIdx.y * blockDim.x) + (threadIdx.z * (blockDim.x * blockDim.y)) + threadIdx.x; return threadId; } __global__ void derive(double *data, double *out, int stride, int gsize, double dx){ int gid = getGid3d3d(); if (gid < gsize){ if (gid + stride < gsize){ out[gid] = (data[gid+stride] - data[gid])/dx; } else{ out[gid] = data[gid]/dx; } } }
2,371
#include "includes.h" __global__ void RBMDropoutMaskKernel( float *maskPtr, float dropout, int thisLayerSize ) { int index = blockDim.x * blockIdx.y * gridDim.x //rows preceeding current row in grid + blockDim.x * blockIdx.x //blocks preceeding current block + threadIdx.x; if (index < thisLayerSize) { maskPtr[index] = dropout < maskPtr[index]; } }
2,372
#include "cuda_runtime.h" #include <stdlib.h> #include <stdio.h> #define NUM_ELEMENTS (1024 * 1024 * 10) #define RUN (100) void globalReduceSum(float *out, const float *in, int numElements); void sharedReduceSum(float *out, const float *in, int numElements); void warpReduceSum(float *out, const float *in, int numElements); void runTest(const char *label, void(*fptr) (float *, const float *, int), float *out, const float *in, int numElements) { cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); // warm up for (int i = 0; i < 10; i++) { fptr(out, in, numElements); cudaDeviceSynchronize(); } cudaEventRecord(start); for (int i = 0; i < RUN; i++) { fptr(out, in, numElements); cudaDeviceSynchronize(); } cudaEventRecord(stop); cudaDeviceSynchronize(); float time_ms; cudaEventElapsedTime(&time_ms, start, stop); float time_s = time_ms / (float)1e3; float GB = (float) numElements * sizeof(float) * RUN; float GBs = GB / time_s / (float)1e9; printf("%s. Time: %f, GB/s: %f\n", label, time_s, GBs); } int main(int argc, char **argv) { float *out_d, *in_d, *out_h, *in_h; int numElements = NUM_ELEMENTS; in_h = (float *)malloc(sizeof(float) * numElements); out_h = (float *)malloc(sizeof(float) * numElements); for (int i = 0; i < numElements; ++i) { in_h[i] = 1.f; } cudaMalloc(&out_d, sizeof(float)); cudaMalloc(&in_d, sizeof(float) * numElements); cudaMemcpy(in_d, in_h, sizeof(float) * numElements, cudaMemcpyHostToDevice); cudaMemset(out_d, 0, sizeof(float)); runTest("Reduce with global memory", globalReduceSum, out_d, in_d, numElements); cudaMemcpy(out_h, out_d, sizeof(float), cudaMemcpyDeviceToHost); printf("sum: %.4f\n", *out_h); cudaMemset(out_d, 0, sizeof(float)); runTest("Reduce with shared memory", sharedReduceSum, out_d, in_d, numElements); cudaMemcpy(out_h, out_d, sizeof(float), cudaMemcpyDeviceToHost); printf("sum: %.4f\n", *out_h); cudaMemset(out_d, 0, sizeof(float)); runTest("Reduce with warp shuffle", warpReduceSum, out_d, in_d, numElements); cudaMemcpy(out_h, out_d, sizeof(float), cudaMemcpyDeviceToHost); printf("sum: %.4f\n", *out_h); double sum = 0; for (int i = 0; i < numElements; ++i) { sum += in_h[i]; } printf("num: %d, reference sum: %f\n", numElements, sum); cudaFree(out_d); cudaFree(in_d); free(out_h); free(in_h); return 0; }
2,373
// CUDA libraries. #include <cuda.h> #include <cuda_runtime.h> //dynamical // Include associated header file. #include "cuda_kernel.cuh" #include <stdio.h> __global__ void dotProductKernel(double* A, double* B, double* result) { // number of elements per thread block const int N = 256; // shared memory for storing partial sums __shared__ double sums[N]; // thread index int i = threadIdx.x; // partial dot product double temp = 0; for (int j = i; j < N; j += N) { temp += A[j] * B[j]; } // store partial sum in shared memory sums[i] = temp; // synchronize threads in block __syncthreads(); // reduction in shared memory for (int offset = 1; offset < N; offset *= 2) { if (i % (2*offset) == 0) { sums[i] += sums[i + offset]; } __syncthreads(); } // result[i] = sums[i]; // write result for block to global memory if (i == 0) { result[blockIdx.x] = sums[0]; } } void kernel(double *A, double *B, double *Result, int arraySize) { // Initialize device pointers. double* gpu_A; double* gpu_B; double* gpu_result; // Always Allocate new device memory for A every time I call this fuction cudaMalloc((void**) &gpu_A, arraySize * sizeof(double)); cudaMalloc((void**) &gpu_B, arraySize * sizeof(double)); cudaMalloc((void**) &gpu_result, sizeof(double)); // Transfer arrays A and B to device. cudaMemcpy(gpu_A, A, arraySize * sizeof(double), cudaMemcpyHostToDevice); cudaMemcpy(gpu_B, B, arraySize * sizeof(double), cudaMemcpyHostToDevice); // cudaMemcpy(gpu_result, Result, arraySize * sizeof(double), cudaMemcpyHostToDevice); dim3 threads(256); dim3 blocks(1); dotProductKernel<<<blocks,threads>>>(gpu_A,gpu_B, gpu_result); cudaMemcpy(Result, gpu_result, sizeof(double), cudaMemcpyDeviceToHost); // Result[1] = 44; // Free device memory cudaFree(gpu_A); cudaFree(gpu_B); cudaFree(gpu_result); }
2,374
//Test 1 #include<cuda.h> #include <bits/stdc++.h> using namespace std; //Sequential //Vector addition kernel void vecAdd(float *h_A, float *h_B, float *h_C, int n){ int i; for (i = 0; i < n; i++) h_C[i] = h_A[i] + h_B[i]; } //Parallel __global__ void vecAddP (float *A, float *B, float *C, int n){ int i = threadIdx.x + blockDim.x * blockIdx.x; if(i < n) C[i] = A[i] + B[i]; } void vectorAdd(float *A, float *B, float *C, int n){ int size = n * sizeof(float); float *d_A, *d_B, *d_C; //1. Allocate memory for d_A, etc. on the device (cudaMalloc) cudaMalloc(&d_A, size); cudaMalloc(&d_B, size); cudaMalloc(&d_C, size); //2. Copy Data from host to d_A, etc. (cudaMemcpy) cudaMemcpy(d_A, A, size, cudaMemcpyHostToDevice); cudaMemcpy(d_B, B, size, cudaMemcpyHostToDevice); //3. Kernel Launch Code vecAddP<<<ceil(n/256.0), 256>>> (d_A, d_B, d_C, n); //4. Copy d_C to C from device, free device memory (cusdaFree), sync if neccessary cudaMemcpy (C, d_C, size, cudaMemcpyDeviceToHost); cudaFree(d_A); cudaFree(d_B); cudaFree(d_C); } int main(){ //Memalloc A, B, C //I/O to read A, B, C float A[5]; float B[5]; float C[5]; float j = 0.0; int n = 5; for (int i = 0; i < 5; i = i + 1, j = j + 1.0){ A[i] = j; B[i] = j +1.5; } vectorAdd(A, B, C, n); for (int i = 0 ; i < 5; i++){ cout << C[i] << endl; } }
2,375
/* * Team - Suraj Singh and Mahir Jain * Roll Numbers - 16CO146 and 16CO123 respectively. */ #include <stdio.h> #include <cuda.h> #include <time.h> #define r_size 10 #define c_size 28 // Only one black is used. #define BLOCKS 1 // Depends on the GPU used. #define THREADS 1024 // Function for calculating execution time void print_elapsed(clock_t start, clock_t stop) { double elapsed = ((double) (stop - start)) / CLOCKS_PER_SEC; printf("Elapsed time: %fs\n", elapsed); } // Main CUDA kernel // most of the parameters here are too big to be stored in shared memory // and hence we have used global memory instead __global__ void brandes(int s, int *R, int *C, int *S, int *d, float *sigma, float *delta, int *Q, int *Q2, int * ends, float *bc) { int idx = threadIdx.x; // Initialise values for BFS for(int k=idx; k < r_size; k+= blockDim.x) { //printf("hi"); if( k == s ) { d[k] = 0; sigma[k] =1; } else { d[k] = -1; sigma[k] = 0; } } // initialize variables common to all threads in the block that fit in shared memory __shared__ int Q_len; __shared__ int Q2_len; __shared__ int ends_len; __shared__ int depth; __shared__ int S_len; int count; if( idx == 0 ) { Q[0] = s; Q_len = 1; S[0] = s; S_len=1; Q2_len = 0; ends[0] =0; ends[1] = 1; ends_len = 2; } __syncthreads(); count =0; while(1) { for(int k=idx; k < ends[count+1] - ends[count]; k+=blockDim.x) { int v = Q[k]; // Same logic as ocean kernel! __syncthreads(); for(int r = R[v]; r< R[v+1]; r++) { int w = C[r]; int t; // Adding neighbours to our 'stack' implemented as a queue if (atomicCAS(&d[w], -1, d[v]+1) == -1) { //printf("%d\n", w); t = atomicAdd(&Q2_len,1); //f =1; Q2[t] = w; } // if v was the shortest path to w, update sigma if(d[w] == (d[v]+1)) { atomicAdd(&sigma[w],sigma[v]); } } } __syncthreads(); if(Q2_len==0) { if(idx==0) { // calculate depth for next section of code depth = d[S[S_len-1]]; } break; } else { // swap Q with Q2 for(int k =idx; k < Q2_len; k+=blockDim.x) { Q[k] = Q2[k]; S[k+S_len] = Q2[k]; } __syncthreads(); if(idx==0) { ends[ends_len] = ends[ends_len-1] + Q2_len; ends_len = ends_len +1; Q_len = Q2_len; S_len = S_len + Q2_len; Q2_len = 0; } __syncthreads(); } count++; __syncthreads(); } // everyone needs to stop after breaking out of while loop __syncthreads(); while(depth > 0) { // all threads execute in parallel if(idx >= ends[depth] && idx <= ends[depth+1] -1) { int w = S[idx]; float dsw = 0; float sw = sigma[w]; // update delta for a vertex by traversing its neighbours for(int r = R[w]; r< R[w+1]; r++) { int v = C[r]; if(d[v] == d[w] + 1) { dsw += (sw/sigma[v])*(1 + delta[v]); } } delta[w] = dsw; __syncthreads(); // add to BC value of the vertex! if(w!=s) { atomicAdd(&bc[w],delta[w]/2); //bc[w] += delta[w]/2; } } depth--; } } int main(int argc, char const *argv[]) { FILE *R = fopen("R.txt", "r"); FILE *C = fopen("C.txt", "r"); clock_t start, stop; int r[r_size]; int c[c_size]; for(int i=0;i< r_size; i++) { fscanf(R, "%d\n", &r[i]); } for(int i=0;i< c_size; i++) { fscanf(C, "%d\n", &c[i]); } int *dev_r, *dev_c, *dev_d, *dev_Q, *dev_Q2,*ends, *dev_S; float *dev_sigma, *dev_delta, *dev_bc; float *bc_val = (float*)malloc(r_size*sizeof(float)); cudaMalloc((void**) &dev_r, r_size*sizeof(int)); cudaMalloc((void**) &dev_c, c_size*sizeof(int)); cudaMalloc((void**) &dev_bc, r_size*sizeof(float)); cudaMalloc((void**) &dev_d, r_size*sizeof(int)); cudaMalloc((void**) &dev_sigma, r_size*sizeof(float)); cudaMalloc((void**) &dev_delta, r_size*sizeof(float)); cudaMalloc((void**) &dev_Q, r_size*sizeof(int)); cudaMalloc((void**) &dev_Q2, r_size*sizeof(int)); cudaMalloc((void**) &dev_S, r_size*sizeof(int)); cudaMalloc((void**) &ends, (r_size+1)*sizeof(int)); cudaMemcpy(dev_r, r, r_size*sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(dev_c, c, c_size*sizeof(int), cudaMemcpyHostToDevice); dim3 blocks(BLOCKS,1); dim3 threads(THREADS,1); start = clock(); for(int s=0; s < r_size; s++) { brandes<<<blocks, threads>>>(s, dev_r, dev_c, dev_S, dev_d , dev_sigma, dev_delta,dev_Q,dev_Q2, ends, dev_bc); } stop=clock(); print_elapsed(start,stop); cudaMemcpy(bc_val, dev_bc, r_size*(sizeof(float)), cudaMemcpyDeviceToHost); return 0; }
2,376
#include <cuda_runtime.h> #include <stdio.h> __global__ void checkIndex(void){ printf("threadIdx: (%d, %d, %d) blockIdx: (%d, %d, %d) \ blockDim: (%d, %d, %d) gridDim: (%d, %d, %d)\n", threadIdx.x, threadIdx.y, threadIdx.z, blockIdx.x, blockIdx.y, blockIdx.z, blockDim.x, blockDim.y, blockDim.z, gridDim.x, gridDim.y, gridDim.z); } int main(int argc, char **argv){ int n = 6; dim3 block(3); dim3 grid((n-1)/block.x+1); printf("grid: (%d, %d, %d)\n", grid.x, grid.y, grid.z); printf("block: (%d, %d, %d)\n", block.x, block.y, block.z); checkIndex<<<grid, block>>>(); cudaDeviceReset(); return 0; }
2,377
#include <thrust/device_vector.h> #include <thrust/host_vector.h> #include <iostream> #include <chrono> #include <thrust/extrema.h> #include <thrust/execution_policy.h> #include <thrust/functional.h> int main() { thrust::host_vector<double> host(10, 0); for(int i = 0; i<10;i++){ double s; std::cin >> s; host[i] = s; } // for(int i =0; i < 2517; i++){ // std::cin >> stocks; // host.push_back(stocks); // } auto start = std::chrono::steady_clock::now(); thrust::device_vector<double> dev(host); auto end = std::chrono::steady_clock::now(); std::chrono::duration<double> elapsed_seconds = end-start; printf("Host vector: "); for (auto i = host.begin(); i != host.end(); i++) { std::cout << *i << " "; // este acesso é rápido -- CPU } printf("\n"); printf("Device vector: "); for (auto i = dev.begin(); i != dev.end(); i++) { std::cout << *i << " "; // este acesso é lento! -- GPU } printf("\n"); std::cerr << "elapsed time: " << elapsed_seconds.count() << "s\n"; // val = thrust::reduce(dev.begin(), dev.end(), 0, thrust::plus<double>())); }
2,378
#include "includes.h" __global__ void mat_mult_transposed_kernel(int *mat_a, int *mat_b, int *res) { int B_TRANS_ROWS = B_COLS; int B_TRANS_COLS = B_ROWS; // El for each thread, shared per block __shared__ int smem[128]; for (int row_block = 0; row_block * gridDim.x < A_ROWS; row_block++) { int a_row = blockIdx.x + (row_block * gridDim.x); for (int b_row = 0; b_row < B_TRANS_ROWS; b_row++) { int total = 0; for (int thread_i = 0; thread_i * blockDim.x < A_COLS; thread_i++) { int thread_col = threadIdx.x + (thread_i * blockDim.x); // Need to check because 240 not even multiple of 128 if (thread_col >= A_COLS) smem[threadIdx.x] = 0; else smem[threadIdx.x] = mat_a[a_row * A_COLS + thread_col] * mat_b[b_row * B_TRANS_COLS + thread_col]; __syncthreads(); //Parallel reduction for (int i = blockDim.x / 2; i > 0; i /= 2) { if (threadIdx.x < i) { int temp = smem[threadIdx.x] + smem[threadIdx.x + i]; smem[threadIdx.x] = temp; } __syncthreads(); } if (threadIdx.x == 0) { total += smem[threadIdx.x]; } } if (threadIdx.x == 0) { res[a_row * C_COLS + b_row] = total; } } } }
2,379
#include "includes.h" #define FALSE 0 #define TRUE !FALSE #define NUMTHREADS 16 #define THREADWORK 32 __global__ void gpuPMCCNoTest(const float * vectsa, size_t na, const float * vectsb, size_t nb, size_t dim, const float * numPairs, const float * means, const float * sds, float * correlations) { size_t offset, stride, x = blockIdx.x, y = blockIdx.y, tx = threadIdx.x; float a, b, n, scoreA, scoreB; __shared__ float meanA, meanB, sdA, sdB, threadSums[NUMTHREADS]; if((x >= na) || (y >= nb)) return; if(tx == 0) { meanA = means[x*nb*2+y*2]; meanB = means[x*nb*2+y*2+1]; sdA = sds[x*nb*2+y*2]; sdB = sds[x*nb*2+y*2+1]; n = numPairs[x*nb+y]; } __syncthreads(); threadSums[tx] = 0.f; for(offset = tx; offset < dim; offset += NUMTHREADS) { a = vectsa[x * dim + offset]; b = vectsb[y * dim + offset]; scoreA = (a - meanA) / sdA; scoreB = (b - meanB) / sdB; threadSums[tx] += scoreA * scoreB; } __syncthreads(); for(stride = NUMTHREADS >> 1; stride > 0; stride >>= 1) { if(tx < stride) threadSums[tx] += threadSums[tx + stride]; __syncthreads(); } if(tx == 0) correlations[x*nb+y] = threadSums[0] / (n - 1.f); }
2,380
// https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#miscellaneous-instructions-pmevent extern "C" __global__ void run_atomics( const float *A, const float *B, float *OUT, int numElements) { int id = blockDim.x * blockIdx.x + threadIdx.x; asm("pmevent 1;"); asm("pmevent 7;"); asm("pmevent.mask 0xFF;"); OUT[id] = A[id] + B[id]; }
2,381
#ifndef KERNEL_REDUCE #define KERNEL_REDUCE #include <math.h> #include <float.h> #include <cuda.h> __global__ void gpu_Heat(float *h, float *g, float *residuals, int N) { float diff = 0.0; int tidx = blockIdx.x * blockDim.x + threadIdx.x; int tidy = blockIdx.y * blockDim.y + threadIdx.y; if (tidx > 0 && tidx < N - 1 && tidy > 0 && tidy < N - 1) { g[tidx * N + tidy] = 0.25 * (h[tidx * N + (tidy - 1)] + // left h[tidx * N + (tidy + 1)] + // right h[(tidx - 1) * N + tidy] + // top h[(tidx + 1) * N + tidy]); // bottom diff = g[tidx * N + tidy] - h[tidx * N + tidy]; residuals[tidx * N + tidy] = diff * diff; } } template<size_t blockSize, typename T> __device__ void warpReduce(volatile T *sdata, size_t tid) { if (blockSize >= 64) sdata[tid] += sdata[tid + 32]; if (blockSize >= 32) sdata[tid] += sdata[tid + 16]; if (blockSize >= 16) sdata[tid] += sdata[tid + 8]; if (blockSize >= 8) sdata[tid] += sdata[tid + 4]; if (blockSize >= 4) sdata[tid] += sdata[tid + 2]; if (blockSize >= 2) sdata[tid] += sdata[tid + 1]; } template<int blockSize, class T> __global__ void reduce1(T *g_idata, T *g_odata, int n) { extern __shared__ T sdata[]; unsigned int tid = threadIdx.x; unsigned int i = blockIdx.x * blockDim.x + threadIdx.x; sdata[tid] = 0; if (i < n) { sdata[tid] = g_idata[i]; } __syncthreads(); for (unsigned int s = 1; s < blockDim.x; s *= 2) { if (tid % (2 * s) == 0) { sdata[tid] += sdata[tid + s]; } __syncthreads(); } if (tid == 0) g_odata[blockIdx.x] = sdata[0]; } template<int blockSize, class T> __global__ void reduce2(T *g_idata, T *g_odata, int n) { extern __shared__ T sdata[]; unsigned int tid = threadIdx.x; unsigned int i = blockIdx.x * blockDim.x + threadIdx.x; sdata[tid] = 0; if (i < n) { sdata[tid] = g_idata[i]; } __syncthreads(); for (unsigned int s = 1; s < blockDim.x; s *= 2) { int index = 2 * s * tid; if (index < blockDim.x) { sdata[index] += sdata[index + s]; } __syncthreads(); } if (tid == 0) g_odata[blockIdx.x] = sdata[0]; } template<int blockSize, class T> __global__ void reduce3(T *g_idata, T *g_odata, int n) { extern __shared__ T sdata[]; unsigned int tid = threadIdx.x; unsigned int i = blockIdx.x * blockDim.x + threadIdx.x; sdata[tid] = 0; if (i < n) { sdata[tid] = g_idata[i]; } __syncthreads(); for (unsigned int s = blockDim.x / 2; s > 0; s >>= 1) { if (tid < s) { sdata[tid] += sdata[tid + s]; } __syncthreads(); } if (tid == 0) g_odata[blockIdx.x] = sdata[0]; } template<int blockSize, class T> __global__ void reduce4(T *g_idata, T *g_odata, int n) { extern __shared__ T sdata[]; unsigned int tid = threadIdx.x; unsigned int i = blockIdx.x * (blockDim.x * 2) + threadIdx.x; sdata[tid] = 0; if (i < n) { sdata[tid] = g_idata[i] + g_idata[i + blockDim.x]; } __syncthreads(); for (unsigned int s = blockDim.x / 2; s > 0; s >>= 1) { if (tid < s) { sdata[tid] += sdata[tid + s]; } __syncthreads(); } if (tid == 0) g_odata[blockIdx.x] = sdata[0]; } template<int blockSize, class T> __global__ void reduce5(T *g_idata, T *g_odata, int n) { extern __shared__ T sdata[]; unsigned int tid = threadIdx.x; unsigned int i = blockIdx.x * (blockDim.x * 2) + threadIdx.x; sdata[tid] = 0; if (i < n) { sdata[tid] = g_idata[i] + g_idata[i + blockDim.x]; } __syncthreads(); for (unsigned int s = blockDim.x / 2; s > 32; s >>= 1) { if (tid < s) { sdata[tid] += sdata[tid + s]; } __syncthreads(); } if (tid < 32) warpReduce<blockSize>(sdata, tid); if (tid == 0) g_odata[blockIdx.x] = sdata[0]; } template<int blockSize, class T> __global__ void reduce6(T *g_idata, T *g_odata, int n) { extern __shared__ T sdata[]; unsigned int tid = threadIdx.x; unsigned int i = blockIdx.x * (blockDim.x * 2) + threadIdx.x; sdata[tid] = 0; if (i < n) { sdata[tid] = g_idata[i] + g_idata[i + blockDim.x]; } __syncthreads(); if (blockSize >= 512 && tid < 256) { sdata[tid] += sdata[tid + 256]; } __syncthreads(); if (blockSize >= 256 && tid < 128) { sdata[tid] += sdata[tid + 128]; } __syncthreads(); if (blockSize >= 128 && tid < 64) { sdata[tid] += sdata[tid + 64]; } __syncthreads(); if (tid < 32) warpReduce<blockSize>(sdata, tid); if (tid == 0) g_odata[blockIdx.x] = sdata[0]; } template<size_t blockSize, typename T> __global__ void finalReduce(T *g_idata, T *g_odata, size_t n) { __shared__ T sdata[blockSize]; size_t tid = threadIdx.x; size_t i = blockIdx.x * (blockSize) + tid; size_t gridSize = blockSize * gridDim.x; sdata[tid] = 0; while (i < n) { sdata[tid] += g_idata[i]; i += gridSize; } __syncthreads(); if (blockSize >= 1024) { if (tid < 512) { sdata[tid] += sdata[tid + 512]; } __syncthreads(); } if (blockSize >= 512) { if (tid < 256) { sdata[tid] += sdata[tid + 256]; } __syncthreads(); } if (blockSize >= 256) { if (tid < 128) { sdata[tid] += sdata[tid + 128]; } __syncthreads(); } if (blockSize >= 128) { if (tid < 64) { sdata[tid] += sdata[tid + 64]; } __syncthreads(); } if (tid < 32) warpReduce<blockSize>(sdata, tid); if (tid == 0) g_odata[blockIdx.x] = sdata[0]; } template<size_t blockSize, typename T> T GPUReductionOrg(T *dA, size_t N) { T tot = 0.; size_t n = N; size_t blocksPerGrid = std::ceil((1. * n) / blockSize); T *tmp; cudaMalloc(&tmp, sizeof(T) * blocksPerGrid); T *from = dA; do { blocksPerGrid = std::ceil((1. * n) / blockSize); finalReduce<blockSize><<<blocksPerGrid, blockSize, blockSize*sizeof(T)>>>(from, tmp, n); from = tmp; n = blocksPerGrid; } while (n > blockSize); if (n > 1) finalReduce<blockSize><<<1, blockSize, blockSize*sizeof(T)>>>(tmp, tmp, n); cudaDeviceSynchronize(); cudaMemcpy(&tot, tmp, sizeof(T), cudaMemcpyDeviceToHost); cudaFree(tmp); return tot; } #endif
2,382
#include "includes.h" __global__ void ComputeCubesKernel( float *pointsCoordinates, float *vertexData, int quadOffset, float cubeSide, int *activityFlag, int textureWidth, int maxCells ) { int threadId = blockDim.x*blockIdx.y*gridDim.x //rows preceeding current row in grid + blockDim.x*blockIdx.x //blocks preceeding current block + threadIdx.x; if(threadId < maxCells) { float x = pointsCoordinates[threadId * 3]; float y = pointsCoordinates[threadId * 3 + 1]; float z = pointsCoordinates[threadId * 3 + 2]; float halfSide = 0.50f * cubeSide; if(activityFlag[threadId] == 0) { halfSide = 0.00f; } int textureOffset = quadOffset + maxCells * 4 * 6 * 3; float textureAbsLength = (float)(maxCells * textureWidth); // BOTTOM SIDE vertexData[quadOffset + threadId * 72] = x - halfSide; vertexData[quadOffset + threadId * 72 + 1] = y - halfSide; vertexData[quadOffset + threadId * 72 + 2] = z + halfSide; vertexData[quadOffset + threadId * 72 + 3] = x - halfSide; vertexData[quadOffset + threadId * 72 + 4] = y - halfSide; vertexData[quadOffset + threadId * 72 + 5] = z - halfSide; vertexData[quadOffset + threadId * 72 + 6] = x + halfSide; vertexData[quadOffset + threadId * 72 + 7] = y - halfSide; vertexData[quadOffset + threadId * 72 + 8] = z - halfSide; vertexData[quadOffset + threadId * 72 + 9] = x + halfSide; vertexData[quadOffset + threadId * 72 + 10] = y - halfSide; vertexData[quadOffset + threadId * 72 + 11] = z + halfSide; vertexData[textureOffset + threadId * 48] = (float)((threadId) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 1] = 0.00f; vertexData[textureOffset + threadId * 48 + 2] = (float)((threadId) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 3] = 1.00f; vertexData[textureOffset + threadId * 48 + 4] = (float)((threadId + 1) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 5] = 1.00f; vertexData[textureOffset + threadId * 48 + 6] = (float)((threadId + 1) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 7] = 0.00f; // FRONT SIDE vertexData[quadOffset + threadId * 72 + 12] = x - halfSide; vertexData[quadOffset + threadId * 72 + 13] = y + halfSide; vertexData[quadOffset + threadId * 72 + 14] = z + halfSide; vertexData[quadOffset + threadId * 72 + 15] = x - halfSide; vertexData[quadOffset + threadId * 72 + 16] = y - halfSide; vertexData[quadOffset + threadId * 72 + 17] = z + halfSide; vertexData[quadOffset + threadId * 72 + 18] = x + halfSide; vertexData[quadOffset + threadId * 72 + 19] = y - halfSide; vertexData[quadOffset + threadId * 72 + 20] = z + halfSide; vertexData[quadOffset + threadId * 72 + 21] = x + halfSide; vertexData[quadOffset + threadId * 72 + 22] = y + halfSide; vertexData[quadOffset + threadId * 72 + 23] = z + halfSide; vertexData[textureOffset + threadId * 48 + 8] = (float)((threadId) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 9] = 0.00f; vertexData[textureOffset + threadId * 48 + 10] = (float)((threadId) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 11] = 1.00f; vertexData[textureOffset + threadId * 48 + 12] = (float)((threadId + 1) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 13] = 1.00f; vertexData[textureOffset + threadId * 48 + 14] = (float)((threadId + 1) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 15] = 0.00f; // LEFT SIDE vertexData[quadOffset + threadId * 72 + 24] = x - halfSide; vertexData[quadOffset + threadId * 72 + 25] = y + halfSide; vertexData[quadOffset + threadId * 72 + 26] = z - halfSide; vertexData[quadOffset + threadId * 72 + 27] = x - halfSide; vertexData[quadOffset + threadId * 72 + 28] = y - halfSide; vertexData[quadOffset + threadId * 72 + 29] = z - halfSide; vertexData[quadOffset + threadId * 72 + 30] = x - halfSide; vertexData[quadOffset + threadId * 72 + 31] = y - halfSide; vertexData[quadOffset + threadId * 72 + 32] = z + halfSide; vertexData[quadOffset + threadId * 72 + 33] = x - halfSide; vertexData[quadOffset + threadId * 72 + 34] = y + halfSide; vertexData[quadOffset + threadId * 72 + 35] = z + halfSide; vertexData[textureOffset + threadId * 48 + 16] = (float)((threadId) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 17] = 0.00f; vertexData[textureOffset + threadId * 48 + 18] = (float)((threadId) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 19] = 1.00f; vertexData[textureOffset + threadId * 48 + 20] = (float)((threadId + 1) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 21] = 1.00f; vertexData[textureOffset + threadId * 48 + 22] = (float)((threadId + 1) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 23] = 0.00f; // BACK SIDE vertexData[quadOffset + threadId * 72 + 36] = x - halfSide; vertexData[quadOffset + threadId * 72 + 37] = y + halfSide; vertexData[quadOffset + threadId * 72 + 38] = z - halfSide; vertexData[quadOffset + threadId * 72 + 39] = x - halfSide; vertexData[quadOffset + threadId * 72 + 40] = y - halfSide; vertexData[quadOffset + threadId * 72 + 41] = z - halfSide; vertexData[quadOffset + threadId * 72 + 42] = x + halfSide; vertexData[quadOffset + threadId * 72 + 43] = y - halfSide; vertexData[quadOffset + threadId * 72 + 44] = z - halfSide; vertexData[quadOffset + threadId * 72 + 45] = x + halfSide; vertexData[quadOffset + threadId * 72 + 46] = y + halfSide; vertexData[quadOffset + threadId * 72 + 47] = z - halfSide; vertexData[textureOffset + threadId * 48 + 24] = (float)((threadId + 1) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 25] = 0.00f; vertexData[textureOffset + threadId * 48 + 26] = (float)((threadId + 1) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 27] = 1.00f; vertexData[textureOffset + threadId * 48 + 28] = (float)((threadId) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 29] = 1.00f; vertexData[textureOffset + threadId * 48 + 30] = (float)((threadId) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 31] = 0.00f; // RIGHT SIDE vertexData[quadOffset + threadId * 72 + 48] = x + halfSide; vertexData[quadOffset + threadId * 72 + 49] = y + halfSide; vertexData[quadOffset + threadId * 72 + 50] = z - halfSide; vertexData[quadOffset + threadId * 72 + 51] = x + halfSide; vertexData[quadOffset + threadId * 72 + 52] = y - halfSide; vertexData[quadOffset + threadId * 72 + 53] = z - halfSide; vertexData[quadOffset + threadId * 72 + 54] = x + halfSide; vertexData[quadOffset + threadId * 72 + 55] = y - halfSide; vertexData[quadOffset + threadId * 72 + 56] = z + halfSide; vertexData[quadOffset + threadId * 72 + 57] = x + halfSide; vertexData[quadOffset + threadId * 72 + 58] = y + halfSide; vertexData[quadOffset + threadId * 72 + 59] = z + halfSide; vertexData[textureOffset + threadId * 48 + 32] = (float)((threadId + 1) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 33] = 0.00f; vertexData[textureOffset + threadId * 48 + 34] = (float)((threadId + 1) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 35] = 1.00f; vertexData[textureOffset + threadId * 48 + 36] = (float)((threadId) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 37] = 1.00f; vertexData[textureOffset + threadId * 48 + 38] = (float)((threadId) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 39] = 0.00f; // UPPER SIDE vertexData[quadOffset + threadId * 72 + 60] = x - halfSide; vertexData[quadOffset + threadId * 72 + 61] = y + halfSide; vertexData[quadOffset + threadId * 72 + 62] = z + halfSide; vertexData[quadOffset + threadId * 72 + 63] = x - halfSide; vertexData[quadOffset + threadId * 72 + 64] = y + halfSide; vertexData[quadOffset + threadId * 72 + 65] = z - halfSide; vertexData[quadOffset + threadId * 72 + 66] = x + halfSide; vertexData[quadOffset + threadId * 72 + 67] = y + halfSide; vertexData[quadOffset + threadId * 72 + 68] = z - halfSide; vertexData[quadOffset + threadId * 72 + 69] = x + halfSide; vertexData[quadOffset + threadId * 72 + 70] = y + halfSide; vertexData[quadOffset + threadId * 72 + 71] = z + halfSide; vertexData[textureOffset + threadId * 48 + 40] = (float)((threadId) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 41] = 1.00f; vertexData[textureOffset + threadId * 48 + 42] = (float)((threadId) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 43] = 0.00f; vertexData[textureOffset + threadId * 48 + 44] = (float)((threadId + 1) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 45] = 0.00f; vertexData[textureOffset + threadId * 48 + 46] = (float)((threadId + 1) * textureWidth) / textureAbsLength; vertexData[textureOffset + threadId * 48 + 47] = 1.00f; } }
2,383
#include <stdio.h> #include <stdlib.h> #include <cmath> using namespace std; void zzz2(int c){ if(c<10){ char a=c+'0'; printf("%c",a); }else{ char a=c-10+'a'; printf("%c",a); } } void zzz1(uchar3 a){ int x=a.x; int y=a.y; int z=a.z; zzz2(x/16); zzz2(x%16); zzz2(y/16); zzz2(y%16); zzz2(z/16); zzz2(z%16); } void zzz3(uchar4 a){ int x=a.x; int y=a.y; int z=a.z; zzz2(x/16); zzz2(x%16); zzz2(y/16); zzz2(y%16); zzz2(z/16); zzz2(z%16); } __device__ double SQRT(double A){ return sqrt(A); } __device__ double POW(double a){ return a*a; } #define CSC(call) do { \ cudaError_t e = call; \ if (e != cudaSuccess) { \ fprintf(stderr, "CUDA Error in %s:%d: %s\n"\ , __FILE__, __LINE__, cudaGetErrorString(e)); \ exit(0); \ } \ } while(0) #define RGBToWB(r, g, b) 0.299*r + 0.587*g + 0.114*b #define chek(a) a>255.1?255.1:a //texture<uchar4, 2, cudaReadModeElementType> tex; __constant__ double3 conTest[33]; __global__ void kernel(uchar4 *dst, int w, int h, int n) { int idx = threadIdx.x + blockDim.x * blockIdx.x; int idy = threadIdx.y + blockDim.y * blockIdx.y; int offsetx = blockDim.x * gridDim.x; int offsety = blockDim.y * gridDim.y; int i, j,k; uchar4 a; double3 b; int ans; double max,tmp; for(i = idx; i < w; i += offsetx){ for(j = idy; j < h; j += offsety){ max=1000000; a = dst[j * w + i]; for(k=0;k<n;k++){ b=conTest[k]; tmp=SQRT(POW(a.x-b.x)+ POW(a.y-b.y)+ POW(a.z-b.z)); //zzz1(b); //printf(" %d %d %f %f\n",j * w + i,k,tmp,max); if(max>tmp){ max=tmp; ans=k; } } dst[j * w + i] = make_uchar4(a.x,a.y,a.z, ans); } } } int main() { int h, w; char path_in[257]; char path_out[257]; scanf("%s", path_in); scanf("%s", path_out); FILE* in = fopen(path_in, "rb"); fread(&w, sizeof(int), 1, in); fread(&h, sizeof(int), 1, in); uchar4 *img = (uchar4 *)malloc(sizeof(uchar4) * h * w); fread(img, sizeof(uchar4), h * w, in); fclose(in); int n,x,y; scanf("%d",&n); double3 *test = (double3 *)malloc(sizeof(double3) * n); double3 *testNext = (double3 *)malloc(sizeof(double3) * n); double4 *res = (double4 *)malloc(sizeof(double4) * n); for(int i=0;i<n;i++){ scanf("%d%d",&x,&y); test[i].x=img[y*w+x].x; test[i].y=img[y*w+x].y; test[i].z=img[y*w+x].z; /*zzz1(test[i]); printf("\n");*/ } /*cudaArray *dev_arr; cudaChannelFormatDesc ch = cudaCreateChannelDesc<uchar4>(); CSC(cudaMallocArray(&dev_arr, &ch, w, h)); CSC(cudaMemcpyToArray(dev_arr, 0, 0, img, sizeof(uchar4) * w * h, cudaMemcpyHostToDevice)); tex.addressMode[0] = cudaAddressModeClamp; tex.addressMode[1] = cudaAddressModeClamp; tex.channelDesc = ch; tex.filterMode = cudaFilterModePoint; tex.normalized = false; CSC(cudaBindTextureToArray(tex, dev_arr, ch));*/ uchar4 *dev_img; CSC(cudaMalloc(&dev_img, sizeof(uchar4) * w * h)); CSC(cudaMemcpy ( dev_img, img, sizeof(uchar4) * w * h, cudaMemcpyHostToDevice )); y=1; while (y) { /*for(int i=0;i<n;i++){ zzz1(test[i]); }*/ CSC( cudaMemcpyToSymbol(conTest, test, n*sizeof(double3)) ); kernel<<< dim3(32, 32), dim3(32, 32) >>>(dev_img, w, h,n); CSC(cudaMemcpy(img, dev_img, sizeof(uchar4) * w * h, cudaMemcpyDeviceToHost)); for(int i=0;i<n;i++){ res[i].x = res[i].y = res[i].z = res[i].w = 0; } for(int i = 0; i < w*h; i++){ //zzz3(img[i]); //printf("%d %d \n",i,img[i].w); res[img[i].w].x+=img[i].x; res[img[i].w].y+=img[i].y; res[img[i].w].z+=img[i].z; res[img[i].w].w+=1; //printf("%f %f ",res[img[i].w].x,res[img[i].w].w); } //printf("\n"); for(int i=0;i<n;i++){ testNext[i].x=res[i].x/res[i].w; testNext[i].y=res[i].y/res[i].w; testNext[i].z=res[i].z/res[i].w; //zzz1(testNext[i]); } //printf("NEXT\n"); y=0; for(int i=0;i<n;i++){ if(test[i].x!=testNext[i].x){ y=1; } if(test[i].y!=testNext[i].y){ y=1; } if(test[i].z!=testNext[i].z){ y=1; } } for(int i=0;i<n;i++){ test[i].x=testNext[i].x; test[i].y=testNext[i].y; test[i].z=testNext[i].z; } } FILE* out = fopen(path_out, "wb"); fwrite(&w, sizeof(int), 1, out); fwrite(&h, sizeof(int), 1, out); fwrite(img, sizeof(uchar4), h * w, out); fclose(out); CSC(cudaFree(dev_img)); free(img); free(test); free(testNext); free(res); return 0; }
2,384
#include "cuda_runtime.h" #include "device_launch_parameters.h" #include <cstdlib> #include <cstdio> #include <cassert> #include <iostream> #include <cmath> __global__ void fibonacci_kernel(double* a, int n) { unsigned int index = threadIdx.x; if (index < n) a[index] = (pow((1 + sqrt(5.0)) / 2, index) - pow((1 - sqrt(5.0)) / 2, index)) / sqrt(5.0); } class Fibonacci { public: int arrayLength; explicit Fibonacci(int arrayLength); void run(int numGrids, int numThreads) const; void displayResult(double* array, double* resultArray) const; }; Fibonacci::Fibonacci(int arrayLength) { this->arrayLength = arrayLength; } void Fibonacci::displayResult(double *array, double* resultArray) const { for(int i = 0; i < this->arrayLength; i++) printf("Index %d: %f\n", i, resultArray[i]); } void Fibonacci::run(int numGrids, int numThreads) const { int deviceId = cudaGetDevice(&deviceId); printf("GPU Device ID: %d\n", deviceId); printf("CPU Device ID: %d\n\n", cudaCpuDeviceId); double * hostArray, * resultArray, * deviceArray; size_t arrayBytes = sizeof(int) * this->arrayLength; cudaMallocHost(&hostArray, arrayBytes); cudaMallocHost(&resultArray, arrayBytes); cudaMalloc(&deviceArray, arrayBytes); cudaMemcpy(deviceArray, hostArray, arrayBytes, cudaMemcpyHostToDevice); fibonacci_kernel<<<numGrids, numThreads>>>(deviceArray, arrayLength); cudaDeviceSynchronize(); cudaMemcpy(resultArray, deviceArray, arrayBytes, cudaMemcpyDeviceToHost); displayResult(hostArray, resultArray); cudaFreeHost(hostArray); cudaFreeHost(resultArray); cudaFree(deviceArray); } int main() { Fibonacci program(16); program.run(1, 256); }
2,385
extern "C" __global__ __launch_bounds__(256) void sgemm_nn_128x128( const float *param_A, const float *param_B, float *param_C, float param_alpha, float param_beta, int param_lda8, int param_ldb8, int param_ldc, int param_m, int param_n, int param_k) { __shared__ float share[128 * 8 * 4 + 32]; int tid = threadIdx.x; share[tid] = 1; }
2,386
#include <cuda_runtime_api.h> #include <stdio.h> int main(){ cudaStream_t s; cudaError_t res; res = cudaStreamCreate(&s); printf("res : %d\n", res); res = cudaStreamDestroy(s); printf("res : %d\n", res); return 0; }
2,387
#include "CCubicDomain.cuh" namespace NBody { //ds ctor/dtor //ds default constructor requires environmental parameters: N number of bodies, dT time step, T number of time steps CCubicDomain::CCubicDomain( const unsigned int& p_uNumberOfParticles ): m_arrPositions( 0 ), m_arrVelocities( 0 ), m_arrAccelerations( 0 ), m_arrMasses( 0 ), m_uNumberOfParticles( p_uNumberOfParticles ), m_strParticleInformation( "" ), m_strIntegralsInformation( "" ) { //ds nothing to do } //ds default destructor CCubicDomain::~CCubicDomain( ) { //ds deallocate memory delete[] m_arrPositions; delete[] m_arrVelocities; delete[] m_arrAccelerations; delete[] m_arrMasses; } //ds accessors void CCubicDomain::createParticlesUniformFromNormalDistribution( const float& p_dTargetKineticEnergy, const float& p_fParticleMass ) { //ds allocate arrays (linear since we're using CUDA) m_arrPositions = new float[m_uNumberOfParticles*3]; m_arrVelocities = new float[m_uNumberOfParticles*3]; m_arrAccelerations = new float[m_uNumberOfParticles*3]; m_arrMasses = new float[m_uNumberOfParticles]; //ds kinetic energy to derive from initial situation float dKineticEnergy( 0.0 ); //ds set particle information for each for( unsigned int u = 0; u < m_uNumberOfParticles; ++u ) { //ds set the particle mass (same for all particles in this case) m_arrMasses[u] = p_fParticleMass; //ds set the position: uniformly distributed between boundaries in this case m_arrPositions[3*u+0] = _getUniformlyDistributedNumber( ); m_arrPositions[3*u+1] = _getUniformlyDistributedNumber( ); m_arrPositions[3*u+2] = _getUniformlyDistributedNumber( ); //ds set velocity values: from normal distribution m_arrVelocities[3*u+0] = _getNormallyDistributedNumber( ); m_arrVelocities[3*u+1] = _getNormallyDistributedNumber( ); m_arrVelocities[3*u+2] = _getNormallyDistributedNumber( ); //ds set acceleration values: 0 m_arrAccelerations[3*u+0] = 0; m_arrAccelerations[3*u+1] = 0; m_arrAccelerations[3*u+2] = 0; //ds add the resulting kinetic component (needed below) dKineticEnergy += m_arrMasses[u]/2*pow( sqrt( pow( m_arrVelocities[3*u+0], 2 ) + pow( m_arrVelocities[3*u+1], 2 ) + pow( m_arrVelocities[3*u+2], 2 ) ), 2 ); } //ds calculate the rescaling factor const float dRescalingFactor( sqrt( p_dTargetKineticEnergy/dKineticEnergy ) ); //ds for each particle for( unsigned int u = 0; u < m_uNumberOfParticles; ++u ) { //ds rescale the velocity component m_arrVelocities[3*u+0] = dRescalingFactor*m_arrVelocities[3*u+0]; m_arrVelocities[3*u+1] = dRescalingFactor*m_arrVelocities[3*u+1]; m_arrVelocities[3*u+2] = dRescalingFactor*m_arrVelocities[3*u+2]; } } void CCubicDomain::saveParticlesToStream( ) { //ds format: X Y Z U V W //ds for each particle for( unsigned int u = 0; u < m_uNumberOfParticles; ++u ) { //ds get a buffer for snprintf (max possible buffer size checked) char chBuffer[256]; //ds get the particle stream std::snprintf( chBuffer, 256, "%f %f %f %f %f %f", m_arrPositions[3*u+0], m_arrPositions[3*u+1], m_arrPositions[3*u+2], m_arrVelocities[3*u+0], m_arrVelocities[3*u+1], m_arrVelocities[3*u+2] ); //ds append the buffer to our string m_strParticleInformation += chBuffer; m_strParticleInformation += "\n"; } } void CCubicDomain::saveIntegralsToStream( const float& p_fTotalEnergy ) { //ds format: E X Y Z X Y Z X Y Z //ds buffer for snprintf char chBuffer[256]; //ds get information - caution, memory gets allocated const float* vecCenterOfMass = getCenterOfMass( ); const float* vecAngularMomentum = getAngularMomentum( ); const float* vecLinearMomentum = getLinearMomentum( ); //ds get the integrals stream std::snprintf( chBuffer, 100, "%f %f %f %f %f %f %f %f %f %f", p_fTotalEnergy, vecCenterOfMass[0], vecCenterOfMass[1], vecCenterOfMass[2], vecAngularMomentum[0], vecAngularMomentum[1], vecAngularMomentum[2], vecLinearMomentum[0], vecLinearMomentum[1], vecLinearMomentum[2] ); //ds free memory delete vecCenterOfMass; delete vecAngularMomentum; delete vecLinearMomentum; //ds append the buffer to our string m_strIntegralsInformation += chBuffer; m_strIntegralsInformation += "\n"; } void CCubicDomain::saveIntegralsToStream( const float& p_fTotalEnergy, const float p_vecCenterOfMass[3] ) { //ds format: E X Y Z X Y Z X Y Z //ds buffer for snprintf char chBuffer[256]; //ds get information - caution, memory gets allocated //const float* vecCenterOfMass = getCenterOfMass( ); const float* vecAngularMomentum = getAngularMomentum( ); const float* vecLinearMomentum = getLinearMomentum( ); //ds get the integrals stream std::snprintf( chBuffer, 100, "%f %f %f %f %f %f %f %f %f %f", p_fTotalEnergy, p_vecCenterOfMass[0], p_vecCenterOfMass[1], p_vecCenterOfMass[2], vecAngularMomentum[0], vecAngularMomentum[1], vecAngularMomentum[2], vecLinearMomentum[0], vecLinearMomentum[1], vecLinearMomentum[2] ); //ds free memory //delete vecCenterOfMass; delete vecAngularMomentum; delete vecLinearMomentum; //ds append the buffer to our string m_strIntegralsInformation += chBuffer; m_strIntegralsInformation += "\n"; } void CCubicDomain::writeParticlesToFile( const std::string& p_strFilename, const unsigned int& p_uNumberOfTimeSteps ) { //ds ofstream object std::ofstream ofsFile; //ds open the file for writing ofsFile.open( p_strFilename.c_str( ), std::ofstream::out ); //ds if it worked if( ofsFile.is_open( ) ) { //ds first dump setup information number of particles and timesteps ofsFile << m_uNumberOfParticles << " " << p_uNumberOfTimeSteps << "\n" << m_strParticleInformation; } //ds close the file ofsFile.close( ); } void CCubicDomain::writeIntegralsToFile( const std::string& p_strFilename, const unsigned int& p_uNumberOfTimeSteps, const double& p_dTimeStepSize ) { //ds ofstream object std::ofstream ofsFile; //ds open the file for writing ofsFile.open( p_strFilename.c_str( ), std::ofstream::out ); //ds if it worked if( ofsFile.is_open( ) ) { //ds dump first integrals information ofsFile << p_uNumberOfTimeSteps << " " << p_dTimeStepSize << "\n" << m_strIntegralsInformation; } //ds close the file ofsFile.close( ); } float* CCubicDomain::getPositions( ) { return m_arrPositions; } float* CCubicDomain::getVelocities( ) { return m_arrVelocities; } float* CCubicDomain::getAccelerations( ) { return m_arrAccelerations; } float* CCubicDomain::getMasses( ) { return m_arrMasses; } const float* CCubicDomain::getCenterOfMass( ) const { //ds center to find float* vecCenter = new float[3]; //ds set it to zero for sure vecCenter[0] = 0.0; vecCenter[1] = 0.0; vecCenter[2] = 0.0; //ds total mass float fMassTotal( 0.0 ); //ds for each particle for( unsigned int u = 0; u < m_uNumberOfParticles; ++u ) { //ds mass instance const float fCurrentMass( m_arrMasses[u] ); //ds add the current relative mass vecCenter[0] += fCurrentMass*m_arrPositions[3*u+0]; vecCenter[1] += fCurrentMass*m_arrPositions[3*u+1]; vecCenter[2] += fCurrentMass*m_arrPositions[3*u+2]; //ds add the current mass fMassTotal += fCurrentMass; } //ds divide by total mass vecCenter[0] /= fMassTotal; vecCenter[1] /= fMassTotal; vecCenter[2] /= fMassTotal; return vecCenter; } const float* CCubicDomain::getAngularMomentum( ) const { //ds momentum float* vecMomentum = new float[3]; //ds set it to zero for sure vecMomentum[0] = 0.0; vecMomentum[1] = 0.0; vecMomentum[2] = 0.0; //ds for each particle for( unsigned int u = 0; u < m_uNumberOfParticles; ++u ) { //ds mass instance const float fCurrentMass( m_arrMasses[u] ); //ds add the current momentum vecMomentum[0] += fCurrentMass*( m_arrPositions[3*u+1]*m_arrVelocities[3*u+2] - m_arrPositions[3*u+2]*m_arrVelocities[3*u+1] ); vecMomentum[1] += fCurrentMass*( m_arrPositions[3*u+2]*m_arrVelocities[3*u+0] - m_arrPositions[3*u+0]*m_arrVelocities[3*u+2] ); vecMomentum[2] += fCurrentMass*( m_arrPositions[3*u+0]*m_arrVelocities[3*u+1] - m_arrPositions[3*u+1]*m_arrVelocities[3*u+0] ); } return vecMomentum; } const float* CCubicDomain::getLinearMomentum( ) const { //ds momentum float* vecMomentum = new float[3]; //ds set it to zero for sure vecMomentum[0] = 0.0; vecMomentum[1] = 0.0; vecMomentum[2] = 0.0; //ds for each particle for( unsigned int u = 0; u < m_uNumberOfParticles; ++u ) { //ds mass instance const float fCurrentMass( m_arrMasses[u] ); //ds add the current momentum vecMomentum[0] += fCurrentMass*m_arrVelocities[3*u+0]; vecMomentum[1] += fCurrentMass*m_arrVelocities[3*u+1]; vecMomentum[2] += fCurrentMass*m_arrVelocities[3*u+2]; } return vecMomentum; } float CCubicDomain::_getUniformlyDistributedNumber( ) const { //ds drand48 returns [0,1], we need [-1,1] -> therefore 2x[0,1] -> [0,2] -> -1 ->[0-1,2-1] = [-1,1] return static_cast< float >( 2*drand48( )-1 ); } float CCubicDomain::_getNormallyDistributedNumber( ) const { //ds calculate the uniform number first [0,1] const float dUniformNumber( static_cast< float >( drand48( ) ) ); //ds return the normal one return static_cast< float >( sqrt( -2*log( dUniformNumber ) )*cos( 2*static_cast< float >( M_PI )*dUniformNumber ) ); } } //ds namespace NBody
2,388
#include "includes.h" __global__ void vecAdd_kernel(float *c, const float* a, const float* b) { int idx = blockIdx.x * blockDim.x + threadIdx.x; for (int i = 0; i < 500; i++) c[idx] = a[idx] + b[idx]; }
2,389
#include <stdio.h> #include <stdlib.h> #include <cuda.h> #include <cuda_runtime.h> //device (1) __global__ void suma_2_enteros(int *d1, int *d2, int *sum){ *sum = *d1 + *d2; } //HOST int main(int argc, char **argv){ int DeviceCount = 0; int h_d1, h_d2, h_sum; //HOST int *d_d1, *d_d2, *d_sum; //DEVICE (2) h_d1 = atoi(argv[1]); h_d2 = atoi(argv[2]); //inicializamos CUDA if(cuInit(0) != 0){ printf("ERROR en la inicializacion de CUDA\n"); exit(0); } cuDeviceGetCount(&DeviceCount); if(DeviceCount == 0){ printf("ERROR, ningun dispositivo compatible con CUDA\n"); exit(0); } //reservamos memoria en DEVICE (2.1) cudaMalloc((void**)&d_d1, sizeof(d_d1)); cudaMalloc((void**)&d_d2, sizeof(d_d2)); cudaMalloc((void**)&d_sum, sizeof(d_sum)); //copiamos desde HOST --> DEVICE (3) cudaMemcpy(d_d1, &h_d1, sizeof(h_d1), cudaMemcpyHostToDevice); cudaMemcpy(d_d2, &h_d2, sizeof(h_d2), cudaMemcpyHostToDevice); //llamamos al KERNEL suma_2_enteros<<<1,1,0,0>>>(d_d1, d_d2, d_sum); //recogemos los resultados del DEVICE (DEVICE --> HOST) (4) cudaMemcpy(&h_sum, d_sum, sizeof(h_sum), cudaMemcpyDeviceToHost); printf("Resultado: %d\n", h_sum); cudaFree(d_d1); cudaFree(d_d2); cudaFree(d_sum); }
2,390
#include <chrono> #include <iostream> #include <random> #include <cuda.h> using std::cout; using std::endl; __global__ void multiply_me_GPU(int *a, int *b, int *c, int width) { int row = blockIdx.y * gridDim.y + threadIdx.y; int column = blockIdx.x * gridDim.x + threadIdx.x; int sum = 0; for (int ii = 0; ii < width; ii++) { sum += a[row * width + ii] * b[column + ii * width]; } c[row * width + column] = sum; } void multiply_me_CPU(int *a, int *b, int *c, int width) { int sum; for (int row = 0; row < width; row++) { for (int column = 0; column < width; column++) { sum = 0; for (int kk = 0; kk < width; kk++) { sum += a[row * width + kk] * b[column + kk * width]; } c[row * width + column] = sum; } } } int main(int argc, char *argv[]) { int N = 1024; int *h_a = new int[N * N]; int *h_b = new int[N * N]; int *h_c = new int[N * N]; int *h_c2 = new int [N * N]; unsigned int seed = std::chrono::system_clock::now().time_since_epoch().count(); std::mt19937 engine(seed); std::uniform_int_distribution<int> distribution(0, 50); for (int ii = 0; ii < N * N; ii++) { h_a[ii] = distribution(engine); h_b[ii] = distribution(engine); } int *d_a; int *d_b; int *d_c; dim3 nblocks (32,32,1); dim3 nthreads(N/nblocks.x, N/nblocks.y, 1); float copy_elapsed; cudaEvent_t copy_start; cudaEvent_t copy_stop; cudaEventCreate(&copy_start); cudaEventCreate(&copy_stop); cudaEventRecord(copy_start, 0); cudaMalloc((void**)&d_a, N * N * sizeof(int)); cudaMalloc((void**)&d_b, N * N * sizeof(int)); cudaMalloc((void**)&d_c, N * N * sizeof(int)); cudaMemcpy(d_a, h_a, N * N * sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(d_b, h_b, N * N * sizeof(int), cudaMemcpyHostToDevice); cudaEventRecord(copy_stop, 0); cudaEventSynchronize(copy_stop); cudaEventElapsedTime(&copy_elapsed, copy_start, copy_stop); float GPU_elapsed; cudaEvent_t GPU_start; cudaEvent_t GPU_stop; cudaEventCreate(&GPU_start); cudaEventCreate(&GPU_stop); cudaEventRecord(GPU_start, 0); multiply_me_GPU<<<nblocks, nthreads>>>(d_a, d_b, d_c, N); cudaEventRecord(GPU_stop, 0); cudaEventSynchronize(GPU_stop); cudaEventElapsedTime(&GPU_elapsed, GPU_start, GPU_stop); cudaMemcpy(h_c, d_c, N * N * sizeof(int), cudaMemcpyDeviceToHost); std::chrono::time_point<std::chrono::system_clock> CPU_start, CPU_stop; std::chrono::duration<double> CPU_elapsed; CPU_start = std::chrono::system_clock::now(); multiply_me_CPU(h_a, h_b, h_c2, N); CPU_stop = std::chrono::system_clock::now(); CPU_elapsed = CPU_stop - CPU_start; cout << "It tool " << copy_elapsed / 1000.0f << "s to copy to data to the GPU" << endl; cout << "It took " << GPU_elapsed / 1000.0f << "s to multiply the matrix on the GPU" << endl; cout << "It took " << CPU_elapsed.count() << "s to multiply the matrix on the CPU" << endl; delete[] h_a; delete[] h_b; delete[] h_c; delete[] h_c2; cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); return 0; }
2,391
#include <stdio.h> #define CSC(call) { \ cudaError err = call; \ if(err != cudaSuccess) { \ fprintf(stderr, "CUDA error in file '%s' in line %i: %s.\n", \ __FILE__, __LINE__, cudaGetErrorString(err)); \ exit(1); \ } \ } while (0) __global__ void kernel(int *a, int n, int k) { int idx = threadIdx.x + blockIdx.x * blockDim.x; int offset = blockDim.x * gridDim.x; for(; idx < n; idx += offset) a[idx] *= k; } int main() { int i, n = 10000; int *a = (int *)malloc(sizeof(int) * n); int *dev_a; for(i = 0; i < n; i++) a[i] = 1; CSC(cudaMalloc(&dev_a, sizeof(int) * n)); CSC(cudaMemcpy(dev_a, a, sizeof(int) * n, cudaMemcpyHostToDevice)); kernel<<<dim3(2), dim3(32)>>>(dev_a, n, 2); CSC(cudaGetLastError()); CSC(cudaMemcpy(a, dev_a, sizeof(int) * n, cudaMemcpyDeviceToHost)); for(i = 0; i < n; i++) printf("%d ", a[i]); printf("\n"); CSC(cudaFree(dev_a)); free(a); return 0; }
2,392
#include "includes.h" /* * CudaOperations.cu * * Created on: Feb 6, 2019 * Author: alexander */ __global__ void allocHamiltonian(float* devMat, float* devSpins, int index, int size, double* energyTempor) { int i; int j; int wIndex = threadIdx.x + blockIdx.x * blockDim.x; while (wIndex < size * size) { i = wIndex % size; j = (int) (wIndex / size); energyTempor[wIndex] = (double) (devSpins[i + index * size] * devSpins[j + index * size] * devMat[wIndex]); wIndex = wIndex + blockDim.x * gridDim.x; } }
2,393
#include <stdio.h> #include <stdlib.h> #include <string.h> #include <math.h> using namespace std; __global__ void hello(){ printf("hello?\n"); return; } __global__ void mtxAddKernel(int* d_a, int* d_b, int m, int n, int* d_c){ int x = blockIdx.x * blockDim.x + threadIdx.x; int y = blockIdx.y * blockDim.y + threadIdx.y; //printf("%d,%d\n", x, y); if(y < m && x <n){ d_c[y*n+x] = d_a[y*n+x] + d_b[y*n+x]; } } void mtxAdd(int * h_a, int* h_b, int m, int n, int* h_c){ int size = m*n*sizeof(int); int* d_a, *d_b, *d_c; cudaMalloc((void**)&d_a, size); cudaMemcpy(d_a, h_a, size, cudaMemcpyHostToDevice); cudaMalloc((void**)&d_b, size); cudaMemcpy(d_b, h_b, size, cudaMemcpyHostToDevice); cudaMalloc((void**)&d_c, size); dim3 dimGrid(int(ceil(n/32.0)), int(ceil(m/32.0)), 1); dim3 dimBlock(32, 32, 1); //printf("%d, %d, %d, %d\n",n, m, ceil(n/32.0), ceil(m/32.0)); mtxAddKernel<<<dimGrid, dimBlock>>>(d_a, d_b, m, n, d_c); cudaError_t err = cudaDeviceSynchronize(); if(err != cudaSuccess){ printf("%s in %s at line %d\n", cudaGetErrorString(err), __FILE__, __LINE__); exit(EXIT_FAILURE); } cudaMemcpy(h_c, d_c, size, cudaMemcpyDeviceToHost); cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); } int main(){ int m, n; scanf("%d%d", &m, &n); printf("%d,%d\n",m,n); int* h_a = (int*)malloc(m*n*sizeof(int)); int* h_b = (int*)malloc(m*n*sizeof(int)); int* h_c = (int*)malloc(m*n*sizeof(int)); //memset(h_a, -1, sizeof(m*n*sizeof(int))); //memset(h_b, -1, sizeof(m*n*sizeof(int))); for(int i = 0; i < m; ++i){ for(int j = 0; j < n; ++j){ h_a[i*n+j] = -1; h_b[i*n+j] = -1; } } mtxAdd(h_a, h_b, m, n, h_c); for(int i = 0; i < m; ++i){ for(int j = 0; j < n; ++j){ if(h_c[i*n+j]!=-2){ printf("%d %d %d false\n",i, j, h_c[i*n+j]); } //printf("%d ", h_c[i*m+j]); } } return 0; }
2,394
#include "includes.h" __global__ void Correlation_backward_input1(int item, float *gradInput1, int nInputChannels, int inputHeight, int inputWidth, float *gradOutput, int nOutputChannels, int outputHeight, int outputWidth, float *rInput2, int pad_size, int kernel_size, int max_displacement, int stride1, int stride2) { // n (batch size), c (num of channels), y (height), x (width) int n = item; int y = blockIdx.x * stride1 + pad_size; int x = blockIdx.y * stride1 + pad_size; int c = blockIdx.z; int tch_off = threadIdx.x; int kernel_rad = (kernel_size - 1) / 2; int displacement_rad = max_displacement / stride2; int displacement_size = 2 * displacement_rad + 1; int xmin = (x - kernel_rad - max_displacement) / stride1; int ymin = (y - kernel_rad - max_displacement) / stride1; int xmax = (x + kernel_rad - max_displacement) / stride1; int ymax = (y + kernel_rad - max_displacement) / stride1; if (xmax < 0 || ymax < 0 || xmin >= outputWidth || ymin >= outputHeight) { // assumes gradInput1 is pre-allocated and zero filled return; } if (xmin > xmax || ymin > ymax) { // assumes gradInput1 is pre-allocated and zero filled return; } xmin = max(0,xmin); xmax = min(outputWidth-1,xmax); ymin = max(0,ymin); ymax = min(outputHeight-1,ymax); int pInputWidth = inputWidth + 2 * pad_size; int pInputHeight = inputHeight + 2 * pad_size; int pdimyxc = pInputHeight * pInputWidth * nInputChannels; int pdimxc = pInputWidth * nInputChannels; int pdimc = nInputChannels; int tdimcyx = nOutputChannels * outputHeight * outputWidth; int tdimyx = outputHeight * outputWidth; int tdimx = outputWidth; int odimcyx = nInputChannels * inputHeight* inputWidth; int odimyx = inputHeight * inputWidth; int odimx = inputWidth; float nelems = kernel_size * kernel_size * nInputChannels; __shared__ float prod_sum[CUDA_NUM_THREADS]; prod_sum[tch_off] = 0; for (int tc = tch_off; tc < nOutputChannels; tc += CUDA_NUM_THREADS) { int i2 = (tc % displacement_size - displacement_rad) * stride2; int j2 = (tc / displacement_size - displacement_rad) * stride2; int indx2 = n * pdimyxc + (y + j2)* pdimxc + (x + i2) * pdimc + c; float val2 = rInput2[indx2]; for (int j = ymin; j <= ymax; ++j) { for (int i = xmin; i <= xmax; ++i) { int tindx = n * tdimcyx + tc * tdimyx + j * tdimx + i; prod_sum[tch_off] += gradOutput[tindx] * val2; } } } __syncthreads(); if(tch_off == 0) { float reduce_sum = 0; for(int idx = 0; idx < CUDA_NUM_THREADS; idx++) { reduce_sum += prod_sum[idx]; } const int indx1 = n * odimcyx + c * odimyx + (y - pad_size) * odimx + (x - pad_size); gradInput1[indx1] = reduce_sum / nelems; } }
2,395
#include<stdio.h> #include<iostream> using namespace std; /* a sum reduction on the array of floats 'in'. * The reduction result is written to the * address 'result'. The number of elements to * be reduced is given by 'size' * * The example contains data races because barrier * synchronisation statements, of the form: * __syncthreads(); * are missing. * * Can you add them to eliminate all data races? */ #define N 8 /* Same as blockDim */ #define tid threadIdx.x __global__ void reduce(int *in, int *result, int size) { __shared__ int partial_sums[N]; /* Each thread sums elements in[tid], in[tid + N], in[tid + 2*N], ... */ partial_sums[tid] = in[tid]; for(int i = tid + N; i < size; i += N) { partial_sums[i] += in[i]; } /* Tree reduction computes final sum into partial_sums[0] */ for(int d = N/2; d > 0; d >>= 1) { if(tid < d) { partial_sums[tid] += partial_sums[tid + d]; } } /* Master thread writes out result */ if(tid == 0) { *result = partial_sums[0]; } }
2,396
//#include "cuda_runtime.h" //#include <cuda.h> //#include <cuda_runtime_api.h> //#include "device_launch_parameters.h" #include<stdio.h> //#include<stdlib.h> //#include<string.h> //#include<math.h> //#include<cutil.h> #include<iostream> #define NUM 2048 ////////////////////////////////////////////////////////////////// __global__ void caculate(unsigned char * g_x) { __shared__ char s[512]; const unsigned int bid=blockIdx.x*blockDim.x+threadIdx.x; if(bid>=NUM) return; s[bid]=g_x[bid]; s[bid]*=2; g_x[bid]=s[bid]; } /////////////////////////////////////////////////////////////// int main(int argc,char**argv) { //重定向到文件 //freopen("1.txt", "w", stdout); int SIZE=sizeof(unsigned char); //---------------------------------------- unsigned char *h_x=(unsigned char*)malloc(SIZE*NUM); for(int i=0;i<NUM;i++) h_x[i]=100; //--------------------------- unsigned char *d_x; cudaMalloc((void**)&d_x,SIZE*NUM); //输入数组从内存拷贝到显存 cudaMemcpy(d_x,h_x,SIZE*NUM,cudaMemcpyHostToDevice); //调用核心函数 dim3 grid; dim3 block; block.x=512; grid.x=(NUM+block.x-1)/block.x; cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); double sum=0; for(int i=0;i<1;i++){ cudaEventRecord(start, 0); float runTime; //==================================== caculate<<<grid,block>>>(d_x); cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&runTime, start, stop); printf("kernel error no =[%d]",cudaGetLastError()); printf("time= %f\n ",runTime); sum+=runTime; } printf("aver time= %f\n ",sum); //两个同步语句cudaThreadSynchronize、cudaDeviceSynchronize必须有一个才能让nsight中显示内核函数。 //cudaThreadSynchronize(); //===================================== //CUT_CHECK_ERROR("Kernel execution failed"); //输出数组从显存拷贝到内存 cudaMemcpy(h_x,d_x,SIZE*NUM,cudaMemcpyDeviceToHost); //在主机端打印 //for(int i=0;i<NUM;i++) printf("h_x[0]=[%c]\n",h_x[0]); printf("h_x[num-1]=[%c]\n",h_x[NUM-1]); //释放内存、显存 free(h_x); cudaFree(d_x); printf("press enter to quit:"); getchar(); }
2,397
#include <stdio.h> #include <stdlib.h> #include <cuda_runtime.h> #include <time.h> __global__ void vAdd(float* A, int num_elements, int factor_hilos, float* s){ //__local__ float a = 0.0; __shared__ float a; if(threadIdx.x == 0) a = 0.0; __syncthreads(); //Posicion del thread int i = (blockIdx.x * blockDim.x + threadIdx.x); //printf("Hola desde el hilo %d, en el bloque %d y el hilo %d\n", i, blockIdx.x, threadIdx.x); if(i < factor_hilos*num_elements){ atomicAdd(&a, A[i%num_elements]); //atomicAdd(&a, 2); //A[i%num_elements] = A[i%num_elements] + 1; } //A[i%num_elements] = a; s[0] = a; //printf("%d", s[0]); } void fError(cudaError_t err, int i){ if(err != cudaSuccess){ printf("%d Ha ocurrido un error con codigo: %s\n", i, cudaGetErrorString(err)); } } int main(){ //cudaSetDevice(1); int num_elements = 1024; int factor_hilos = 1; //Reservar espacio en memoria HOST float * h_A = (float*)malloc(num_elements * sizeof(float)); if(h_A == NULL ){ printf("Error al reservar memoria para los vectores HOST"); exit(1); } float * h_sum = (float*)malloc(sizeof(float)); h_sum[0] = 0; //Inicializar elementos de los vectores for(int i=0; i<num_elements; i++){ h_A[i] = (float)i; } cudaError_t err; int size = num_elements * sizeof(float); float * d_A = NULL; err = cudaMalloc((void **)&d_A, size); fError(err,1); float * d_sum = NULL; err = cudaMalloc((void **)&d_sum, sizeof(float)); fError(err, 3); //Copiamos a GPU DEVICE err = cudaMemcpy(d_A, h_A, size, cudaMemcpyHostToDevice); cudaMemcpy(d_sum, h_sum, sizeof(float), cudaMemcpyHostToDevice); int HilosPorBloque = 256; int BloquesPorGrid = (factor_hilos*num_elements + HilosPorBloque -1) / HilosPorBloque; cudaError_t Err; //Lanzamos el kernel y medimos tiempos cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start, 0); vAdd<<<BloquesPorGrid, HilosPorBloque>>>(d_A, num_elements, factor_hilos, d_sum); Err = cudaGetLastError(); fError(Err,2); cudaEventRecord(stop,0); cudaEventSynchronize(stop); float tiempo_reserva_host; cudaEventElapsedTime(&tiempo_reserva_host, start, stop); printf("Tiempo de suma vectores DEVICE: %f\n", tiempo_reserva_host); cudaEventDestroy(start); cudaEventDestroy(stop); //Copiamos a CPU el vector C err = cudaMemcpy(h_A, d_A, size, cudaMemcpyDeviceToHost); cudaMemcpy(h_sum, d_sum, sizeof(float), cudaMemcpyDeviceToHost); /*for(int i=0; i<20; i++){ printf("%f ", h_A[i]); //printf("\n"); }*/ printf("La suma es: %f", h_sum[0]); }
2,398
//System header #include <stdio.h> #include <stdlib.h> #include <time.h> //CUDA header #include "cuda_runtime.h" #include "device_launch_parameters.h" __global__ void CUParaSgemv(const float *a, float *b, float *c,unsigned int size)//valid { unsigned int id = blockIdx.x * blockDim.x + threadIdx.x; //int i = threadIdx.x; if(size<=id) return; float temp = 0.0; for(unsigned int k = 0;k<size; k++) { if(id < size) //Column access - coalesced access temp += a[k*size+id] * b[k]; //Row access //temp += a[id*size+k] * b[k]; } c[id] += temp; } //__global__ void transpose(float *odata, float *idata, int width, int height) //{ // __shared__ float block[BLOCK_DIM][BLOCK_DIM+1]; // // unsigned int xIndex = blockIdx.x * BLOCK_DIM + threadIdx.x; // unsigned int yIndex = blockIdx.y * BLOCK_DIM + threadIdx.y; // if((xIndex < width) && (yIndex < height)) // { // unsigned int index_in = yIndex * width + xIndex; // block[threadIdx.y][threadIdx.x] = idata[index_in]; // } // // __syncthreads(); // // xIndex = blockIdx.y * BLOCK_DIM + threadIdx.x; // yIndex = blockIdx.x * BLOCK_DIM + threadIdx.y; // if((xIndex < height) && (yIndex < width)) // { // unsigned int index_out = yIndex * height + xIndex; // odata[index_out] = block[threadIdx.x][threadIdx.y]; // } // // float temp = 0.0; // unsigned int idx = blockIdx.x * BLOCK_DIM + threadIdx.x; // // if(idx<height){ // for(int i=0;i<width;i++) // { // temp = idata[idx*width+i]; // odata[i*] // } // } // //} //void easyTranspose(float o_a[],float i_a[],int size) //{ // int col = size*size; // for(int i = 0;i<col;i++) // { // for(int j=0;j<col;j++) // o_a[j*col+i]=i_a[i*col+j]; // } //} void simple_sgemv(float *A, float *B, float *C,unsigned int size) //valid { unsigned int i,j; for(i = 0;i < size; i++) { float prod = 0; for(j = 0;j < size; j++) { prod += A[i * size + j] * B[j]; } C[i] = prod; } } int main() { //# of nodes(equations) //each node has 3-direction displacement unsigned int Nodes = 100; //threashold 3500-old/4500-new unsigned int ARRAY_SIZE = 3*Nodes; //Vector Scale; unsigned int ARRAY_SIZE2 = ARRAY_SIZE*ARRAY_SIZE; //Matrix Scale; //CPU timing clock_t start, finish; //CPU_sgemv time elapse clock_t malloc_start,malloc_fin; //CPU malloc time clock_t init_start,init_fin; //CPU inital time clock_t trans_start,trans_fin; //CPU time on transpose Matrix float duration; float malloc_duration; float init_duration; float trans_duration; //GPU timing float cal_time; float cudamalloctime; float cudamemcpytime; float cudamemcpyout; cudaEvent_t d_start, d_stop; cudaEvent_t cuda_mallocstart,cuda_mallocfin; cudaEvent_t cuda_memcpystart,cuda_memcpyfin; cudaEvent_t cuda_memcpyout_start,cuda_memcpyout_fin; //Host float *h_a; float *h_b; float *h_c; float *h_cpu; float *h_check; float *h_atr; //Device float *d_a; //float *d_atr; float *d_b; float *d_c; //cuda status record cudaError_t cudaStatus; printf("The nodes number is: %d\n",Nodes); printf("The total equations number is : %d\n",ARRAY_SIZE); printf("Total bytes will be transfered\n"); printf("\tMatrix A: %d MB\n",ARRAY_SIZE2*4/1000000); printf("\tVector b: %d KB\n",ARRAY_SIZE*4/1000); printf("Pre-processing in CPU...\n"); /******Malloc on CPU*******/ //start the clock malloc_start = clock(); //generate the input array on the host h_a=(float*)malloc(sizeof(float)*ARRAY_SIZE2); h_b=(float*)malloc(sizeof(float)*ARRAY_SIZE); h_c=(float*)malloc(sizeof(float)*ARRAY_SIZE); h_cpu=(float*)malloc(sizeof(float)*ARRAY_SIZE); h_atr = (float*)malloc(sizeof(float)*ARRAY_SIZE2); //h_check=(float*)malloc(sizeof(float)*ARRAY_SIZE2); //finish time malloc_fin = clock(); //Processing Time in CPU malloc_duration = (float)(malloc_fin - malloc_start) / CLOCKS_PER_SEC; printf( "\n%f seconds passed in mallocation\n", malloc_duration); printf("\n"); /****************************/ /******Initalization on CPU*******/ //use h_ = float(i) to standard the value for the initialization init_start = clock(); srand((int)time(0)); //inital the h_a, h_b for(unsigned int i = 0;i<ARRAY_SIZE2;i++){ h_a[i] = float(i);//rand();//float(i);//rand(); } for(unsigned int i = 0;i<ARRAY_SIZE;i++){ h_b[i] = float(i);//rand();//float(i);//rand(); } for(unsigned int i = 0;i<ARRAY_SIZE;i++){ h_c[i] = float(0); } for(unsigned int i = 0;i<ARRAY_SIZE;i++){ h_cpu[i] = float(0); } //time on transpose trans_start = clock(); for(unsigned int i = 0;i<ARRAY_SIZE;i++){ //h_atr[i] = float(0); for(unsigned int j=0;j<ARRAY_SIZE;j++) h_atr[j*ARRAY_SIZE+i]=h_a[i*ARRAY_SIZE+j]; } trans_fin = clock(); trans_duration = (float)(trans_fin - trans_start) / CLOCKS_PER_SEC; printf( "\n%f seconds passed in transpose..\n", trans_duration); init_fin = clock(); //Processing Time on CPU init_duration = (float)(init_fin - init_start) / CLOCKS_PER_SEC; printf( "\n%f seconds passed in initalizaton\n", init_duration); printf("\n"); printf("******************End Pre-processing.**************\n"); /**********************************/ /**************CPU sgemv calculation time********************/ start = clock(); //kernel function on CPU simple_sgemv(h_a,h_b,h_cpu,ARRAY_SIZE); finish = clock(); //Processing Time in CPU duration = (float)(finish - start) ;// CLOCKS_PER_SEC; printf( "\n%f milliseconds passed in CPU_sgemv\n", duration); printf("\n"); /**********************************/ //system("pause"); ////Print Result //printf("\nThe result Matrix C is:\n"); //for(unsigned int i=0;i<ARRAY_SIZE;i++){ // printf("%f\n", h_cpu[i]); //} printf("Pre-processing in GPU...\n"); /**************GPU malloc********************/ cudaEventCreate(&cuda_mallocstart); cudaEventCreate(&cuda_mallocfin); cudaEventRecord(cuda_mallocstart,0); //mark event //allocate GPU memory //Malloc the memory for matrix and check cudaStatus = cudaMalloc((void**)&d_a, sizeof(float)*ARRAY_SIZE2); cudaStatus = cudaGetLastError(); if(cudaStatus != cudaSuccess) { printf("\nCuda Error(cudaMalloc Matrix):%s\n",cudaGetErrorString(cudaStatus)); system("pause\n"); return 0; } //Malloc the memory for vector and check cudaStatus = cudaMalloc((void**)&d_b, sizeof(float)*ARRAY_SIZE); cudaStatus = cudaGetLastError(); if(cudaStatus != cudaSuccess) { printf("\nCuda Error(cudaMalloc Vector):%s\n",cudaGetErrorString(cudaStatus)); system("pause\n"); return 0; } //Malloc the memory for storing result and check cudaStatus = cudaMalloc((void**)&d_c, sizeof(float)*ARRAY_SIZE); cudaStatus = cudaGetLastError(); if(cudaStatus != cudaSuccess) { printf("\nCuda Error(cudaMalloc result):%s\n",cudaGetErrorString(cudaStatus)); system("pause\n"); return 0; } cudaThreadSynchronize(); cudaEventRecord(cuda_mallocfin,0); cudaEventSynchronize(cuda_mallocfin); cudaEventElapsedTime(&cudamalloctime,cuda_mallocstart,cuda_mallocfin); printf( "\n%f milliseconds passed in GPU malloc\n", cudamalloctime ); /*********************************************/ /**************GPU Memcpy time********************/ //Timer cudaEventCreate(&cuda_memcpystart); cudaEventCreate(&cuda_memcpyfin); cudaEventRecord(cuda_memcpystart,0); //mark event //transfer the array from Host to device(CPU->GPU) and check the cudaStatus //Column access cudaStatus = cudaMemcpy(d_a, h_atr, sizeof(float)*ARRAY_SIZE2, cudaMemcpyHostToDevice); //Row access //cudaStatus = cudaMemcpy(d_a, h_a, sizeof(float)*ARRAY_SIZE2, cudaMemcpyHostToDevice); if(cudaStatus != cudaSuccess) { printf("\nCuda Error(cudaMemcpy matrix):%s\n",cudaGetErrorString(cudaStatus)); system("pause\n"); return 0; } cudaStatus = cudaMemcpy(d_b, h_b, sizeof(float)*ARRAY_SIZE, cudaMemcpyHostToDevice); if(cudaStatus != cudaSuccess) { printf("\nCuda Error(cudaMemcpy vector):%s\n",cudaGetErrorString(cudaStatus)); system("pause\n"); return 0; } cudaStatus = cudaMemcpy(d_c, h_c, sizeof(float)*ARRAY_SIZE, cudaMemcpyHostToDevice); if(cudaStatus != cudaSuccess) { printf("\nCuda Error(cudaMemcpy result):%s\n",cudaGetErrorString(cudaStatus)); system("pause\n"); return 0; } cudaThreadSynchronize(); cudaEventRecord(cuda_memcpyfin,0); cudaEventSynchronize(cuda_memcpyfin); cudaEventElapsedTime(&cudamemcpytime,cuda_memcpystart,cuda_memcpyfin); printf( "\n%f milliseconds passed in cuda memory copy\n", cudamemcpytime ); /*********************************************/ printf("*****************End Pre-processing in GPU********************\n"); /**************GPU Caculation time********************/ printf("\n*****************A transpose before the calculation********************\n"); //easyTranspose(h_atr,h_a,ARRAY_SIZE); ////A transpose Before the calculation... //cudaStatus = cudaMalloc((void**)&d_atr, sizeof(float)*ARRAY_SIZE2); ////Get malloc error //cudaStatus = cudaGetLastError(); //if(cudaStatus != cudaSuccess) //{ // printf("\nCuda Error(cudaMalloc Matrix):%s\n",cudaGetErrorString(cudaStatus)); // system("pause\n"); // return 0; //} ////Memory copy //cudaStatus = cudaMemcpy(d_atr, h_a, sizeof(float)*ARRAY_SIZE2, cudaMemcpyHostToDevice); //if(cudaStatus != cudaSuccess) //{ // printf("\nCuda Error(cudaMemcpy transpose matrix):%s\n",cudaGetErrorString(cudaStatus)); // system("pause\n"); // return 0; //} ////Run transpose kernel //transpose<<<1, 128>>>(d_atr,d_a,ARRAY_SIZE,ARRAY_SIZE);//addKernel ////transfer the array from Device to Host(GPU->CPU) & check //cudaStatus = cudaMemcpy(h_check, d_atr, sizeof(float)*ARRAY_SIZE2, cudaMemcpyDeviceToHost); //cudaStatus = cudaGetLastError(); //if(cudaStatus != cudaSuccess) //{ // printf("\nCuda Error:%s\n",cudaGetErrorString(cudaStatus)); // system("pause\n"); // return 0; //} ////print out the transpose result //printf("\After transpose...\n"); //for(long i = 0; i<ARRAY_SIZE2;i++){ // printf("%f\n", h_atr[i]); //} ////print out the original Matrix A //printf("\nBefore transpose...\n"); //for(long i = 0; i<ARRAY_SIZE2;i++){ // printf("%f\n", h_a[i]); //} printf("\n*****************End of transpose********************\n"); //Run kernel function calculate the matrix-vector multiplication printf("\n\nRunning Kernel...\n\n"); //Timer cudaEventCreate(&d_start); cudaEventCreate(&d_stop); cudaEventRecord(d_start,0); //mark event //Check //cudaError_t cudaState = cudaSuccess; cudaStatus = cudaSuccess; //MVKernel int nblocks= ARRAY_SIZE/512+1; CUParaSgemv<<<nblocks, 512>>>(d_a,d_b,d_c,ARRAY_SIZE);//addKernel //addKernel<<<1, ARRAY_SIZE>>>(d_a,d_b,d_c,ARRAY_SIZE); cudaThreadSynchronize(); cudaEventRecord(d_stop,0); cudaEventSynchronize(d_stop); cudaEventElapsedTime(&cal_time,d_start,d_stop); printf( "\n%f milliseconds passed in GPU_CUParaSgemv\n", cal_time ); cudaStatus = cudaGetLastError(); if(cudaStatus != cudaSuccess) { printf("\nCuda Error(GPU calculation):%s\n",cudaGetErrorString(cudaStatus)); system("pause\n"); return 0; } //printf( "\n%f milliseconds passed in calculation\n", time ); /*********************************************/ printf("\n*********Copy Data to Host*********\n"); /**************GPU Memory copy out time********************/ //Timer cudaEventCreate(&cuda_memcpyout_start); cudaEventCreate(&cuda_memcpyout_fin); cudaEventRecord(cuda_memcpyout_start,0); //mark event //transfer the array from Device to Host(GPU->CPU) & check cudaStatus = cudaMemcpy(h_c, d_c, sizeof(float)*ARRAY_SIZE, cudaMemcpyDeviceToHost); cudaThreadSynchronize(); cudaEventRecord(cuda_memcpyout_fin,0); cudaEventSynchronize(cuda_memcpyout_fin); cudaEventElapsedTime(&cudamemcpyout,cuda_memcpyout_start,cuda_memcpyout_fin); printf( "\n%f milliseconds passed in cuda memory copy out\n", cudamemcpyout ); cudaStatus = cudaGetLastError(); if(cudaStatus != cudaSuccess) { printf("\nCuda Error:%s\n",cudaGetErrorString(cudaStatus)); system("pause\n"); return 0; } /***********************************************/ //system("pause"); /**************Print out the result********************/ ////print out the result array //for(long i = 0; i<ARRAY_SIZE;i++){ // printf("%f\n", h_c[i]); //} /***********************************************/ //free GPU memory allocation cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); //free Host memory allocation free(h_atr); free(h_a); free(h_b); free(h_c); system("pause"); return 0; }
2,399
#include "includes.h" __global__ void matrix_2d_mul_float_gpu(float *A, float *B, float *C, int num_rows_A, int num_cols_A, int num_cols_B) { // Create shared variables (Available to all threads on the same block) __shared__ float A_tile[N_THREADS][N_THREADS]; __shared__ float B_tile[N_THREADS][N_THREADS]; // Block index int bx = blockIdx.x; int by = blockIdx.y; // Thread index int tx = threadIdx.x; int ty = threadIdx.y; // Index of the first sub-matrix of A processed by the block int aBegin = num_cols_A * N_THREADS * by; // Index of the last sub-matrix of A processed by the block int aEnd = aBegin + num_cols_A - 1; // Index of the first sub-matrix of B processed by the block int bBegin = N_THREADS * bx; int bStep = N_THREADS * num_cols_B; int aStep = N_THREADS; float sum = 0; for (int a = aBegin, b = bBegin;a <= aEnd;a += aStep, b += bStep) { A_tile[ty][tx] = A[a + num_cols_A * ty + tx]; B_tile[tx][ty] = B[b + num_cols_B * tx + ty]; // Synchronize to make sure the matrices are loaded __syncthreads(); for (int k = 0; k < N_THREADS; ++k) sum += A_tile[ty][k] * B_tile[k][tx]; // Wait other threads to finish their sub-matrices __syncthreads(); } // Write the block sub-matrix to device memory; // each thread writes one element int c = num_cols_B * N_THREADS * by + N_THREADS * bx; C[c + num_cols_B * ty + tx] = sum; }
2,400
#include <emmintrin.h> #include <sys/time.h> #include <stdio.h> const long N = 1000000; // Change array size (may need a long) /////////////////////////////////////////////////////////////////////////////////////////////////////////// // HELPER CODE TO INITIALIZE, PRINT AND TIME struct timeval start, end; void starttime() { gettimeofday( &start, 0 ); } void endtime(const char* c) { gettimeofday( &end, 0 ); double elapsed = ( end.tv_sec - start.tv_sec ) * 1000.0 + ( end.tv_usec - start.tv_usec ) / 1000.0; printf("%s: %f ms\n", c, elapsed); } void init(const char* c) { printf("***************** %s **********************\n", c); // TMC Commenting Out for Class printf("Running %s...\n", c); starttime(); } void finish(int a, long N, const char* c) { endtime(c); printf("Done.\n"); printf("\nThere are %ld Prime numbers between 1 and %ld.", a, N); printf("***************************************************\n"); } ////////////////////////////////////////////////////////////////////////////////////////////////////////// /* __global__ void prime(long* a, long high) { // Prime algorithm bool check = false; for(int i = 2; i <= high/2; ++i) { if(high % i == 0) { check = true; break; } } if(check) ++a; } */ // Normal C function to square root values int normal(int a, long N) { long low = 2, high = N, i, check; // printf("Prime numbers between 1 and %d are: ",high); while (low < high) { check = 0; for(i = 2; i <= low/2; ++i) { if(low % i == 0) { check = 1; break; } } if (check == 0) ++a; //printf("%d ", low); ++low; } return a; } // GPU function to square root values // Every thread on every core runs this function __global__ void gpu_prime(int* a, long N) { // One element per thread on each core // blockIdx.x = Core # // blockDim.x = Threads per core // threadIdx.x = Thread # // The formula below makes sure the value of element // is different on every thread on every core long element = blockIdx.x*blockDim.x + threadIdx.x; // If there is not an event split, some threads will be // out of bounds // We just let those do nothing // The rest square root their elements if (element <= N && element >= 2) { /* if (element % 2 != 0) element = N - element; //printf("%d\n", element); */ //printf("%d\n", element); int check = 0; for(int i = 2; i <= element/2; ++i) { if(element % i == 0) { check = 1; break; } } if (check == 0){ atomicAdd(a,1); } } } void gpu(int* a, long N) { int threadsPerCore = 512; // This can vary, up to 1024 long numCores = N / threadsPerCore + 1; // This division will work. If the split is uneven, we overshoot // Budget memory for counter // Memory must be on the graphics card (use cudaMalloc for this) int* gpuA; cudaMalloc(&gpuA, sizeof(int)); // Allocate enough memory on the GPU // Copy array of floats a from CPU memory to gpuA on the graphics card // Note: This operation is SLOW. You will have to offset this cost with the parallelism below cudaMemcpy(gpuA, a, sizeof(int), cudaMemcpyHostToDevice); //printf("%ld\n", *gpuA); // Call parallel function with specified number of cores and threads per core gpu_prime<<<numCores, threadsPerCore>>>(gpuA, N); // Copy square rooted array of floats gpuA from graphics card to a in CPU memory // Again, this operation is SLOW. cudaMemcpy(a, gpuA, sizeof(int), cudaMemcpyDeviceToHost); // Release the memory for gpuA cudaFree(&gpuA); // Free the memory on the GPU } int main() { ///////////////////////////////////////////////////////////////////////// // GPUs will likely have large N // Budget memory on the heap, prevent a stack overflow int a = 1; ///////////////////////////////////////////////////////////////////////// // Test 1: Sequential For Loop init ("Normal"); a = normal(a, N); finish(a, N, "Normal"); // Test 2: GPU a = 1; init("GPU"); gpu(&a, N); finish(a, N, "GPU"); // Memory on the heap must be freed manually //free(&a); return 0; }