serial_no
int64
1
24.2k
cuda_source
stringlengths
11
9.01M
601
/* from: https://devblogs.nvidia.com/parallelforall/even-easier-introduction-cuda/ * Jialin Liu * Simple starting cpp cuda program * Jun 24 2017, Saturday, 2:09pm * Compile and test on Maeve, a 3GPU single node at NERSC, LBNL, CA. */ #include<iostream> #include<math.h> using namespace std; //CUDA kernel functions to add the elements of two arrays __global__ void add (int n, float *x, float * y){ for (int i=0;i<n;i++){ y[i] = x[i] + y[i]; } } int main(void) { int N= 1<<20; //1 million elements //float * x= new float[N]; //float * y= new float[N]; float *x, *y; cudaMallocManaged(&x, N*sizeof(float)); cudaMallocManaged(&y, N*sizeof(float)); clock_t t; //Initialize x and y arrays on the host for (int i=0; i<N; i++){ x[i] =1.5f; y[i] =2.3f; } //run kernel on 1M elements on the CPU t = clock(); //add(N, x, y); add<<<1, 1>>>(N, x, y); t = clock() -t; //cout<<format("%f seconds")%((float)t/CLOCKS_PER_SEC)<<endl; cout <<(float)t/CLOCKS_PER_SEC<<" seconds"<<endl; //Wait for GPU to finish before accessing on host cudaDeviceSynchronize(); float maxError = 0.0f; for (int i =0;i <N;i ++) maxError =fmax(maxError, fabs(y[i]-3.8f)); cout <<"Max error: "<<maxError <<endl; //delete [] x; //delete [] y; cudaFree(x); cudaFree(y); return 0; }
602
#include <stdio.h> /* //=========== PART-1 ======================== __global__ void hello() { } int main(void) { hello<<< 1, 1 >>>(); cudaDeviceSynchronize(); printf("Hello World\n"); return 0; } */ //=========== PART-2 ======================== __device__ const char *STR = "HELLO WORLD!"; const char STR_LENGTH = 12; __global__ void hello() { //every thread prints one character printf("%c\n", STR[threadIdx.x % STR_LENGTH]); } int main(void) { hello<<< 1, 16>>>(); cudaDeviceSynchronize(); return 0; }
603
__global__ void exampleDevice( float * d ) { int idx = blockIdx.x * blockDim.x + threadIdx.x; d[ idx ] = idx; } extern "C" void exampleHost( float * h, int blockDim, int threadDim ) { float * d; cudaMalloc( ( void** )&d, blockDim * threadDim * sizeof( float ) ); exampleDevice<<<blockDim, threadDim>>>( d ); cudaMemcpy( h, d, blockDim * threadDim * sizeof( float ), cudaMemcpyDeviceToHost ); }
604
#include "includes.h" __global__ void kernel_512_one_128(float *A, float *B, float *bnBias, float *bnScale, float *C) { int tile = blockIdx.x, in_channel = threadIdx.x, line = threadIdx.y; int ind = line*128 + in_channel; extern __shared__ float shared_[]; float *weights = shared_ + 512*4, *output = weights + 128*64, *input = shared_; float *bias = output + 4*128, *scale = bias + 128; for (int i = 0; i < 4; i++) input[ind + i*512] = A[tile*2048 + i*512 + ind]; bias[in_channel] = bnBias[in_channel]; scale[in_channel] = bnScale[in_channel]; output[ind] = 0.0f; __syncthreads(); for (int k = 0; k < 512; k += 64) { float *B_start = B + k*128; for (int i = 0; i < 16; i++) weights[ind + i*512] = B_start[i*512 + ind]; __syncthreads(); float *A_start = input + k; for (int p = 0; p < 64; p++) { output[ind] += A_start[line*512 + p] * weights[in_channel + p*128]; } __syncthreads(); } float *C_start = C + tile*512, res = scale[in_channel] * output[ind] + bias[in_channel]; C_start[ind] = res > 0 ? res : 0; }
605
//Parallel programming for many core GPUs //Name: Gesu Bal //Instructor name: Meilin Liu /* this is a simple cuda program calculating Tiled Matrix vector multiplication for 2 dimensions on GPU device I multiplied two double two-dimensional matrices A, B on the device GPU. After the device matrix multiplication kernel function is invoked, and the multiplication result is transferred back to the CPU. The program will also compute the multiplication matrix of matrices A and B using the CPU. Then the program compares the device-computed result with the CPU-computed result. If it matches (within a certain tolerance, i.e., 0.000001), then it will print out "Test PASSED" to the screen before exiting. This case is for all matrix sizes and blocksize/tilewidth: 8*8 */ #include<stdio.h> #include<cuda.h> #include <time.h> int N,blocksize; //gpu function for multiplication __global__ void mul_matrix_gpu(double *d_a, double *d_b, double *d_c, int width) { int TILE_WIDTH=8; __shared__ double ds_M[8][8]; __shared__ double ds_N[8][8]; int bx = blockIdx.x; int by = blockIdx.y; int tx = threadIdx.x; int ty = threadIdx.y; // Identify the row and column of the Pd element to work on int Row = by * TILE_WIDTH + ty; int Col = bx * TILE_WIDTH + tx; double Pvalue = 0; // Loop over the Md and Nd tiles required to compute the Pd element for (int m = 0; m < width/TILE_WIDTH; m++) { // Coolaborative loading of Md and Nd tiles into shared memory ds_M[ty][tx] = d_a[Row*width + m*TILE_WIDTH +tx]; ds_N[ty][tx] = d_b[Col+(m*TILE_WIDTH+ty)*width]; __syncthreads(); for (int k = 0; k < TILE_WIDTH; k++) Pvalue += ds_M[ty][k] * ds_N[k][tx]; __syncthreads(); } d_c[Row*width+Col] = Pvalue; } //cpu function for multiplication void mul_matrix_cpu(double *a, double *b, double *cpu_c, int N) { int i, j,k; for (i=0;i<N;i++) { for (j=0;j<N;j++) { double sum=0; for (k=0;k<N;k++) { double p=a[i*N+k]; double q=b[k*N+j]; sum=sum+(p*q); } cpu_c[i*N+j]=sum; } } } //cpu and gpu result matching function bool verify(double *A, double *B, double *C, int width) { const double relativeTolerance = 0.000001; for(int row = 0; row < width; row++) { for(int col = 0; col < width; col++) { double sum = 0; for(unsigned int k = 0; k < width; k++) { sum += A[row*width + k]*B[k*width + col]; } double relativeError = (sum - C[row*width + col])/sum; //printf("%f \t",relativeError); //printf("\n"); if (relativeError >= relativeTolerance || relativeError <= -relativeTolerance) { printf("TEST FAILED\n\n"); return false; } } } printf("TEST PASSED\n\n"); return true; } //print matrix int printMatrix(double *a,int N) { int i,j; for (i=0;i<N;i++) { for (j=0;j<N;j++) { printf("%f\t",a[i*N+j]); } printf("\n"); } return 1; } int main() { //user input int r, col; printf("Select one of the following options: \n"); printf("Press a for matrix size 8 * 8 \n"); printf("Press b for matrix size 64 * 64 \n"); printf("Press c for matrix size 128 * 128 \n"); printf("Press d for matrix size 512 * 512 \n"); printf("Press e for matrix size 1024 * 1024 \n"); printf("Press f for matrix size 4096 * 4096 \n"); printf("Press any other key for exit \n"); char ch; scanf("%c",&ch); switch(ch) { case 'a': r=8; col=8; N=8; printf("Matrix size is 8 * 8 \n"); break; case 'b': r=64; col=64; N=64; printf("Matrix size is 64 * 64 \n"); break; case 'c': r=128; col=128; N=128; printf("Matrix size is 128 * 128 \n"); break; case 'd': r=512; col=512; N=512; printf("Matrix size is 512 * 512 \n"); break; case 'e': r=1024; col=1024; N=1024; printf("Matrix size is 1024 * 1024 \n"); break; case 'f': r=4096; col=4096; N=4096; printf("Matrix size is 4096 * 4096 \n"); break; default: exit(1); break; } //initializing the block size/tile width blocksize=8; //memory allocation for vectors double *a, *b, *c, *cpu_c, *d_a, *d_b, *d_c; int a_size=r*col; int b_size=r*col; int c_size=r*col; int cpu_c_size=r*col; a=(double*)malloc(sizeof(double)*a_size); b=(double*)malloc(sizeof(double)*b_size); c=(double*)malloc(sizeof(double)*c_size); cpu_c=(double*)malloc(sizeof(double)*cpu_c_size); //matrix initialization int i,j; int init=1325; for (i=0;i<N;i++) { for (j=0;j<N;j++) { init=3125*init%65536; a[i*col+j]=(init-32768.0)/16384.0; init=3125*init%65536; b[i*col+j]=(init-32768.0)/16384.0; } } //printMatrix(a,N); //printf("\n"); //printMatrix(b,N); //printf("\n"); //allocating memory on device int cudaret=cudaMalloc((void **)(&d_a),(N*N)*sizeof(double)); if(cudaret!=cudaSuccess) {printf("memory was not allocated on device \n");} cudaMalloc((void **)(&d_b),(N*N)*sizeof(double)); cudaMalloc((void **)(&d_c),(N*N)*sizeof(double)); //calculating cpu time clock_t startCPU, end; float cpu_time_used; //calling CPU program printf("Calculating results for CPU vector multiplication \n"); printf("---------\n"); startCPU = clock(); mul_matrix_cpu(a,b,cpu_c,N); end = clock(); cpu_time_used = ((float) (end - startCPU))*1000; cpu_time_used= cpu_time_used/ CLOCKS_PER_SEC; printf("CPU computation time (milliseconds) \n"); printf("%f \t",cpu_time_used); printf("\n"); printf("\n"); //printMatrix(cpu_c,N); //printf("\n"); //time execution calculation cudaEvent_t start,stop; float elapsedTime; float timeTransfer; float timeBack; //memory transfer time cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start,0); //copying contents of a and b to device arrays cudaMemcpy(d_a,a,(N*N)*sizeof(double),cudaMemcpyHostToDevice); cudaMemcpy(d_b,b,(N*N)*sizeof(double),cudaMemcpyHostToDevice); cudaDeviceSynchronize(); cudaEventRecord(stop,0); cudaEventSynchronize(stop); cudaEventElapsedTime(&timeTransfer,start,stop); cudaEventDestroy(start); cudaEventDestroy(stop); //Initializing block count and block size dim3 dimBlock(blocksize,blocksize,1); int blockCount_x = (N - 1)/(double(blocksize))+1;//Get number of blocks needed per direction. int blockCount_y = (N - 1)/(double(blocksize))+1; dim3 dimGrid(blockCount_x,blockCount_y,1); printf("Block size and tile width for the program is %d\n ",blocksize); //call kernel for gpu functioning printf("Calling kernel for gpu computations for vector multiplication and calculating results\n"); printf("---------\n"); cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start,0); mul_matrix_gpu<<<dimGrid,dimBlock>>>(d_a,d_b,d_c,N); cudaEventRecord(stop,0); cudaEventSynchronize(stop); cudaEventElapsedTime(&elapsedTime,start,stop); cudaEventDestroy(start); cudaEventDestroy(stop); printf("GPU computation time (milliseconds) \n"); printf("%f \t",elapsedTime); printf("\n"); cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start,0); //copying resulting back to cpu cudaMemcpy(c,d_c,(N*N)*sizeof(double),cudaMemcpyDeviceToHost); cudaDeviceSynchronize(); cudaEventRecord(stop,0); cudaEventSynchronize(stop); cudaEventElapsedTime(&timeBack,start,stop); cudaEventDestroy(start); cudaEventDestroy(stop); timeTransfer += timeBack; printf("Total Memory transfer time between CPU and GPU (milliseconds)\n"); printf("%f \t",timeTransfer); printf("\n"); float speedup; speedup=cpu_time_used/elapsedTime; printf("Speedup: \n"); printf("%f \t",speedup); printf("\n"); printf("Comparing results for CPU and GPU computations \n"); printf("---------\n"); verify(a,b,c,N); //deallocating memory cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); return 0; }
606
#include <stdio.h> int main() { int dimx = 10; int num_bytes = dimx * sizeof (int); // device and host pointers int *d_a = 0; int *h_a = 0; /*Aloca memria na CPU para n inteiros*/ h_a = (int*) malloc(num_bytes); printf("%i\n", num_bytes); /*Aloca memria na GPU para n inteiros*/ cudaMalloc(&d_a, num_bytes); if (0 == h_a || 0 == d_a) { printf("couldn't allocate memory\n"); return 1; } cudaMemset(d_a, 0, num_bytes); cudaMemcpy(h_a, d_a, num_bytes, cudaMemcpyDeviceToHost); for (int i = 0; i < dimx; i++){ printf("%d ", i+1); } printf("\n"); free(h_a); cudaFree(d_a); return 0; }
607
// // include files // #include <stdlib.h> #include <stdio.h> #include <string.h> #include <math.h> //#include <cutil_inline.h> #include "cuda_runtime.h" //#include "MyFirst_kernel.cu" __global__ void my_first_kernel(float *x) { // Uncomment line below and define integer "tid" as global index to vector "x" int tid = blockIdx.x * blockDim.x + threadIdx.x; // Uncomment line below and define x[tid] to be equal to the thread index x[tid] = (float)threadIdx.x; } // // main host code // int main(int argc, char **argv) { float *h_x, *d_x; int nblocks, nthreads, nsize, n; // set number of blocks, and threads per block nblocks = 2; nthreads = 16; nsize = nblocks*nthreads; // allocate memory for array h_x = (float *)malloc(nsize*sizeof(float)); cudaMalloc((void **)&d_x, nsize*sizeof(float)); // execute kernel my_first_kernel<<<nblocks,nthreads>>>(d_x); // copy results from device to host cudaMemcpy(h_x,d_x,nsize*sizeof(float),cudaMemcpyDeviceToHost); // print results for (n=0; n<nsize; n++) printf(" n, x = %d %f \n",n,h_x[n]); // check results float sumcheck = 0.; float sumcheckcorrect = 0.; for (int i = 0; i < nblocks * nthreads; ++i) { sumcheck += h_x[i]; } for (int j=0; j<nthreads; ++j) { sumcheckcorrect += j; } sumcheckcorrect *= 2; if (fabs(sumcheck-sumcheckcorrect)<1e-6) { printf("PASSED!\n"); } else { printf("FAILED!\n"); } // free memory cudaFree(d_x); free(h_x); return 0; }
608
#include "includes.h" __global__ void kernel_test7_write(char* _ptr, char* end_ptr, char* _start_ptr, unsigned int* err) { unsigned int i; unsigned int* ptr = (unsigned int*) (_ptr + blockIdx.x*BLOCKSIZE); unsigned int* start_ptr = (unsigned int*) _start_ptr; if (ptr >= (unsigned int*) end_ptr) { return; } for (i = 0;i < BLOCKSIZE/sizeof(unsigned int); i++){ ptr[i] = start_ptr[i]; } return; }
609
/** * @file CUDA_Hardware.cu * @brief Report essential characteristics of GPU. * * */ #include <cstdio> #include <cstdlib> #include <iostream> using namespace std; int main(int argc, char** argv) { int count = 0; cudaGetDeviceCount(&count); printf("Report on GPU configuration (GPUs: %i).\n", count); for(int i=0; i<count; i++) { printf("Device:1/%i\n", count); cudaDeviceProp prop; cudaGetDeviceProperties(&prop, 0); printf("Global GPU Memory (GMEM) :\t%li(MB)\n", prop.totalGlobalMem / 1024 /1024); printf("Streaming Multiprocessors:\t%i\n", prop.multiProcessorCount); printf("L2 cache size :\t%i(kB)\n", prop.l2CacheSize / 1024); int thsm = prop.maxThreadsPerMultiProcessor; int thbl = prop.maxThreadsPerBlock; printf("Max threads per block :\t%i\n", thbl); if (thsm == 2048) printf("Compute capability:\t3.0 (Kepler)\n"); else if (thsm == 1536) printf("Compute capability:\t2.0 (Fermi)\n"); else printf("Compute capability:\t1.3\n"); } }
610
__device__ int get(int x, int y,int width){ return y * width +x; } __device__ int normeValue(int x, int width){ if(x < 0) //-1 return width - 1; if(x == width) return 0; return x; } __device__ int* neighborsIndexes(int i, int j, int width, int height){ int dir[8]; dir[0] = get(normeValue(i+1,width), j, width); dir[1] = get(normeValue(i+1,width), normeValue(j+1,height),width); dir[2] = get(i, normeValue(j+1,height),width); dir[3] = get(normeValue(i-1,width), normeValue(j+1,height),width); dir[4] = get(normeValue(i-1,width), j, width); dir[5] = get(normeValue(i-1,width), normeValue(j-1,height),width); dir[6] = get(i, normeValue(j-1,height),width); dir[7] = get(normeValue(i+1,width), normeValue(j-1,height),width); return dir; } __device__ float getTotalUpdateFromNeighbors(float* tmp, int i, int j, int width, int height){ int iPlusOne = i + 1; int jPlusOne = j + 1; int iMinusOne = i - 1; int jMinusOne = j - 1; return tmp[get(normeValue(iPlusOne,width), j, width)] + tmp[get(normeValue(iPlusOne,width), normeValue(jPlusOne,height),width)] + tmp[get(i, normeValue(jPlusOne,height),width)] + tmp[get(normeValue(iMinusOne,width), normeValue(jPlusOne,height),width)] + tmp[get(normeValue(iMinusOne,width), j, width)] + tmp[get(normeValue(iMinusOne,width), normeValue(jMinusOne,height),width)] + tmp[get(i, normeValue(jMinusOne,height),width)] + tmp[get(normeValue(iPlusOne,width), normeValue(jMinusOne,height),width)]; } extern "C" __global__ void DIFFUSION_TO_TMP( int width, int height, float *values, float* tmp, float diffCoef) { int i = blockIdx.x * blockDim.x + threadIdx.x; int j = blockIdx.y * blockDim.y + threadIdx.y; //filling tmp if (i < width && j < height ){ int k = get(i,j,width); float give = values[k] * diffCoef; float giveToNeighbor = give / 8; values[k] -= give;//TODO a[k] = value - give tmp[k] = giveToNeighbor; } } extern "C" __global__ void DIFFUSION_UPDATE( int width, int height, float *values, float* tmp) { int i = blockIdx.x * blockDim.x + threadIdx.x; int j = blockIdx.y * blockDim.y + threadIdx.y; if (i < width && j < height ){//TODO + alone values[get(i,j,width)] += getTotalUpdateFromNeighbors(tmp, i, j, width, height); } } extern "C" __global__ void DIFFUSION_UPDATE_THEN_EVAPORATION( int width, int height, float *values, float* tmp, float evapCoef) { int i = blockIdx.x * blockDim.x + threadIdx.x; int j = blockIdx.y * blockDim.y + threadIdx.y; if (i < width && j < height ){//TODO + alone int k = get(i,j,width); float total = values[k] + getTotalUpdateFromNeighbors(tmp, i, j, width, height); values[k] = total - total * evapCoef; } }
611
#include <iostream> #include <complex> #include <cmath> #include <iomanip> #include <string> #include <fstream> using namespace std; __global__ void z_funct(double * d_mat_re, double * d_mat_im, int *d_img, int nb_ite) { int i = blockDim.x * blockIdx.x + threadIdx.x; double c_re = d_mat_re[i]; double c_im = d_mat_im[i]; double z_re = 0, z_im =0; double z_re_tmp = 0; int t=0; while ((z_re*z_re+z_im*z_im <= 4) && (t<nb_ite)) { z_re_tmp = z_re*z_re - z_im*z_im + c_re; z_im = 2*z_im*z_re + c_im; z_re = z_re_tmp; t++; } d_img[i] = t; } int main(int argc, char * argv[]) { if (argc == 1) { cout << "Help" << endl << " 8 args :" << endl << endl << "arg1 = min real part | arg2 = min imaginary part " << endl << "arg3 = max real part | arg4 = max imaginary part " << endl << "arg5 = number of points on the real axe | arg6 = number of points on the imaginary axe " << endl << "arg7 = nb of iterations | arg8 = limit convergence" << endl << endl << " 4 args :" << endl << endl << "arg1 = number of points on the real axe | arg2 = number of points on the imaginary axe " << endl << "arg3 = nb of iterations " << endl << endl ; return 1; } double max_re, max_im, min_re, min_im; int nb_pts_re, nb_pts_im, nb_ite; if (argc == 8) { try { min_re = stod(argv[1]); min_im = stod(argv[2]); max_re = stod(argv[3]); max_im = stod(argv[4]); nb_pts_re = stoi(argv[5]); nb_pts_im = stoi(argv[6]); nb_ite = stoi(argv[7]); } catch (...) { cout << "Bad Args : see help (type nameofprogram without args)" << endl << endl; return 1; } } if (argc == 4 ) { min_re = -2; min_im = -1; max_re = 1; max_im = 1; try { nb_pts_re = stoi(argv[1]); nb_pts_im = stoi(argv[2]); nb_ite = stoi(argv[3]); } catch (...) { cout << "Bad Args : see help (type nameofprogram without args)" << endl << endl; return 1; } } cout << "Initializing..." << endl; int size_d = sizeof(double)*nb_pts_re*nb_pts_im; int size_i = sizeof(int)*nb_pts_re*nb_pts_im; double * mat_re = (double *)malloc(size_d); double * mat_im = (double *)malloc(size_d); double re, im; for (int i=0; i<nb_pts_im; i++) { im = max_im - (max_im-min_im)/nb_pts_im*i; for (int j=0; j<nb_pts_re; j++) { re = max_re - (max_re-min_re)/nb_pts_re*j; mat_re[i*nb_pts_re+j] = re; mat_im[i*nb_pts_re+j] = im; } } double *d_mat_re, *d_mat_im; int *d_img; cudaMalloc(&d_mat_re,size_d); cudaMalloc(&d_mat_im,size_d); cudaMalloc(&d_img,size_i); cudaMemcpy(d_mat_re, mat_re, size_d, cudaMemcpyHostToDevice); cudaMemcpy(d_mat_im, mat_im, size_d, cudaMemcpyHostToDevice); cout << "Running on GPU..." << endl; dim3 blockDim = 1024; dim3 gridDim = (nb_pts_re*nb_pts_im)/1024 + 1; z_funct<<<gridDim,blockDim>>>(d_mat_re, d_mat_im, d_img, nb_ite); cout << "Fetching datas..." << endl; int * img = (int*) malloc(size_i); cudaMemcpy(mat_re, d_mat_re, size_d, cudaMemcpyDeviceToHost); cudaMemcpy(mat_im, d_mat_im, size_d, cudaMemcpyDeviceToHost); cudaMemcpy(img, d_img, size_i, cudaMemcpyDeviceToHost); cout << "Writing on the disk..." << endl; ofstream file; file.open("conv"); for (int i=0; i<nb_pts_im; i++) { for (int j=0; j<nb_pts_re; j++) { file << setw(15) << img[i*nb_pts_re+j] ; } file << endl ; } file.close(); /* file.open("val"); for (int i=0; i<nb_pts_im; i++) { for (int j=0; j<nb_pts_re; j++) { file << setw(15) << pow(mat_re[i*nb_pts_re+j],2) + pow(mat_im[i*nb_pts_re+j],2) ; } file << endl ; } file.close(); */ cout << "Done !" << endl; return 0; }
612
#include <stdio.h> #include <stdint.h> #include <thrust/device_vector.h> #include <thrust/copy.h> #include <thrust/sort.h> #include <algorithm> #define CHECK(call) \ { \ const cudaError_t error = call; \ if (error != cudaSuccess) \ { \ fprintf(stderr, "Error: %s:%d, ", __FILE__, __LINE__); \ fprintf(stderr, "code: %d, reason: %s\n", error, \ cudaGetErrorString(error)); \ exit(1); \ } \ } struct GpuTimer { cudaEvent_t start; cudaEvent_t stop; GpuTimer() { cudaEventCreate(&start); cudaEventCreate(&stop); } ~GpuTimer() { cudaEventDestroy(start); cudaEventDestroy(stop); } void Start() { cudaEventRecord(start, 0); cudaEventSynchronize(start); } void Stop() { cudaEventRecord(stop, 0); } float Elapsed() { float elapsed; cudaEventSynchronize(stop); cudaEventElapsedTime(&elapsed, start, stop); return elapsed; } }; // Sequential radix sort // Assume: nBits (k in slides) in {1, 2, 4, 8, 16} void sortByThrust(const uint32_t * in, int n, uint32_t * out) { thrust::device_vector<uint32_t> dv_out(in, in + n); thrust::sort(dv_out.begin(), dv_out.end()); thrust::copy(dv_out.begin(), dv_out.end(), out); } __global__ void computeLocalHist(uint32_t * in, int n, int * scan, int nBins, int bit) { extern __shared__ int s_hist[]; int i=blockDim.x*blockIdx.x+threadIdx.x; for(int stride=0;stride<nBins;stride+=blockDim.x) if(threadIdx.x+stride<nBins) s_hist[threadIdx.x+stride]=0; __syncthreads(); if(i<n) { int bin=(in[i]>>bit)&(nBins-1);// lấy nBits ra để tính xem phần tử này thuộc bin nào atomicAdd(&s_hist[bin], 1); } __syncthreads();// syncthreads để chắc chắn các phần tử trong block đã được tính trong s_hist for(int stride=0;stride<nBins;stride+=blockDim.x) if(threadIdx.x+stride<nBins) scan[(threadIdx.x+stride)*gridDim.x+blockIdx.x]=s_hist[threadIdx.x+stride]; // hist[nBins*blockIdx.x+threadIdx.x+stride]=s_hist[threadIdx.x+stride]; } __global__ void scanBlkKernel(int * in, int n, int * out, int * blkSums) { extern __shared__ uint32_t value[]; int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < n) { value[threadIdx.x] = in[i]; } for (unsigned int stride = 1; stride <= threadIdx.x; stride *= 2) { __syncthreads(); int tmp; if (threadIdx.x < n - stride) tmp = value[threadIdx.x-stride]; else tmp = 0; __syncthreads(); value[threadIdx.x] += tmp; } blkSums[blockIdx.x] = value[blockDim.x - 1]; if (i<n) { if(threadIdx.x==0) out[i]=0; else out[i]=value[threadIdx.x-1]; } } __global__ void radixSort1bit(uint32_t * in, int n, uint32_t * out,int nBits, int bit,int nBins, int* starts) { int i = blockIdx.x * blockDim.x + threadIdx.x; extern __shared__ uint32_t value[]; __shared__ uint32_t start[256]; for(int indexbit=0;indexbit<nBits;indexbit++) { if (i < n) { value[threadIdx.x] = ((((in[i] >> bit) & (nBins - 1)) >> indexbit) & 1); } __syncthreads(); for(int stride=1;stride<blockDim.x;stride*=2) { int temp=0; if(threadIdx.x>=stride) { temp=value[threadIdx.x-stride];// lấy phần tử trước đó stride bước } __syncthreads();// chắc chắn giá trị năm trước stride bước đã được lấy vào bộ nhớ thanh ghi if(threadIdx.x>=stride ) { value[threadIdx.x]+=temp; } __syncthreads();// chắc chắn các giá trị đã được cộng xong } int nZeros=0; if(blockIdx.x*blockDim.x+blockDim.x<=n) nZeros = blockDim.x - value[blockDim.x-2] -((((in[blockIdx.x*blockDim.x+blockDim.x-1] >> bit) & (nBins - 1)) >> indexbit) & 1); else { if(n%blockDim.x>=2) nZeros = n%blockDim.x - value[n%blockDim.x-2] - ((((in[n-1] >> bit) & (nBins - 1)) >> indexbit) & 1); else nZeros = n%blockDim.x - ((((in[n-1] >> bit) & (nBins - 1)) >> indexbit) & 1); } if (i<n) { if(threadIdx.x==0) { if (((((in[i] >> bit) & (nBins - 1)) >> indexbit) & 1)==0) { out[i]=in[i]; } else out[nZeros+blockIdx.x*blockDim.x]=in[i]; } else { if(((((in[i] >> bit) & (nBins - 1)) >> indexbit) & 1)==0) { out[i-value[threadIdx.x-1]]=in[i]; } else { out[nZeros+value[threadIdx.x-1]+blockIdx.x*blockDim.x]=in[i]; } } } __syncthreads(); uint32_t *tam=in; in=out; out=tam; } if (i<n) { if(threadIdx.x==0) { start[((in[i] >> bit) & (nBins - 1))]=threadIdx.x; } else { if(((in[i] >> bit) & (nBins - 1))!=((in[i-1] >> bit) & (nBins - 1))) { start[((in[i] >> bit) & (nBins - 1))]=threadIdx.x; starts[blockIdx.x*nBins+((in[i] >> bit) & (nBins - 1))]=start[((in[i] >> bit) & (nBins - 1))]; } } } // if(i<n) // { // int bin = (in[i] >> bit) & (nBins - 1); // int rank= histScan[bin*gridDim.x+blockIdx.x]+threadIdx.x-start[bin]; // // if (rank >= 0) // // { // // printf("%d\n",rank); // out[rank] = in[i]; // // } // } } __global__ void scatter(uint32_t * in, int n, uint32_t * out,int nBits, int bit,int nBins, int* start, int* histScan) { int i = blockIdx.x * blockDim.x + threadIdx.x; if(i < n) { int bin = (in[i] >> bit) & (nBins - 1); int rank=histScan[bin*gridDim.x+blockIdx.x]+threadIdx.x-start[blockIdx.x*nBins+bin]; out[rank]=in[i]; } } __global__ void addSumScan(int * out, int n, int * blkSums) { int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < n && blockIdx.x > 0) { out[i] = out[i] + blkSums[blockIdx.x - 1]; } } void sortByDevice(const uint32_t * in, int n, uint32_t * out, int nBits, int blockSize) { int nBins = 1 << nBits; // số bin int m=(n - 1) / blockSize + 1;// gridSize dim3 blockSizeHist(blockSize); dim3 blockSizeScan(blockSize); dim3 gridSizeHist((n - 1) / blockSizeHist.x + 1); dim3 gridSizeScan((nBins*m - 1) / blockSizeScan.x + 1); // cấp phát // scan int *d_scan, *d_blkSums, *d_histScan, *d_blkOuts, *d_starts; int *histScan = (int *)malloc(m*nBins * sizeof(int)); int *blkSums = (int *)malloc(m*nBins*sizeof(int)); int* starts1D=(int *) malloc(m*nBins*sizeof(int)); CHECK(cudaMalloc(&d_scan, nBins*m * sizeof(int))); CHECK(cudaMalloc(&d_blkSums,gridSizeScan.x*sizeof(int))); CHECK(cudaMalloc(&d_blkOuts,m*nBins*sizeof(int))); CHECK(cudaMalloc(&d_starts,m*nBins*sizeof(int))); CHECK(cudaMalloc(&d_histScan,m*nBins*sizeof(int))); // chỉ số bắt đầu int **start = (int **)malloc(m * sizeof(int *)); for (int i=0; i<m; i++) { start[i] = (int *)malloc(nBins * sizeof(int)); } uint32_t * src = (uint32_t *)malloc(n * sizeof(uint32_t)); uint32_t * d_in,*d_out, *d_outTmp; CHECK(cudaMalloc(&d_in,n * sizeof(uint32_t))); CHECK(cudaMalloc(&d_out,n * sizeof(uint32_t))); CHECK(cudaMalloc(&d_outTmp,n * sizeof(uint32_t))); memcpy(src, in, n * sizeof(uint32_t)); uint32_t * originalSrc = src; // Use originalSrc to free memory later uint32_t * dst = out; size_t bytes = gridSizeScan.x * sizeof(int); int * in_tmp = (int *)malloc(bytes); int * out_tmp = (int *)malloc(bytes); for (int bit = 0; bit < sizeof(uint32_t) * 8; bit += nBits) { CHECK(cudaMemcpy(d_in, src, n * sizeof(uint32_t), cudaMemcpyHostToDevice)); // Tính local hist bỏ vào d_scan computeLocalHist<<<gridSizeHist, blockSizeHist, blockSize*sizeof(int)>>>(d_in, n, d_scan, nBins,bit); // // Tính exclusive scan bỏ vào d_histscan scanBlkKernel<<<gridSizeScan,blockSizeScan,blockSize*sizeof(int)>>>(d_scan,m*nBins,d_histScan,d_blkSums); CHECK(cudaMemcpy(in_tmp, d_blkSums, gridSizeScan.x * sizeof(int), cudaMemcpyDeviceToHost)); out_tmp[0] = in_tmp[0]; for (int i = 1; i < gridSizeScan.x; i++) { out_tmp[i] = out_tmp[i - 1] + in_tmp[i]; } CHECK(cudaMemcpy(d_blkOuts, out_tmp, gridSizeScan.x * sizeof(int), cudaMemcpyHostToDevice)); addSumScan<<<gridSizeScan,blockSizeScan>>>(d_histScan, n, d_blkOuts); cudaDeviceSynchronize(); CHECK(cudaGetLastError()); radixSort1bit<<<gridSizeHist,blockSizeHist,blockSize*sizeof(uint32_t)>>>(d_in,n,d_out,nBits,bit,nBins, d_starts); scatter<<<gridSizeHist,blockSizeHist,blockSize*sizeof(uint32_t)>>>(d_in,n,d_out,nBits,bit,nBins,d_starts,d_histScan); CHECK(cudaMemcpy(dst,d_out,n*sizeof(uint32_t),cudaMemcpyDeviceToHost)); uint32_t * temp = src; src = dst; dst = temp; } memcpy(out, src, n * sizeof(uint32_t)); // Free memories cudaFree(d_scan); // cudaFree(d_blkSums); cudaFree(d_histScan); cudaFree(d_in); cudaFree(d_out); for (int i=0; i<m; i++) { free(start[i]); } free(start); free(originalSrc); } // Radix sort void sort(const uint32_t * in, int n, uint32_t * out, int nBits, bool useDevice=false, int blockSizes=512) { GpuTimer timer; timer.Start(); if (useDevice == false) { printf("\nRadix sort by Thrust\n"); sortByThrust(in, n, out); } else // use device { printf("\nRadix sort by device\n"); sortByDevice(in, n, out, nBits, blockSizes); } timer.Stop(); printf("Time: %.3f ms\n", timer.Elapsed()); } void printDeviceInfo() { cudaDeviceProp devProv; CHECK(cudaGetDeviceProperties(&devProv, 0)); printf("**********GPU info**********\n"); printf("Name: %s\n", devProv.name); printf("Compute capability: %d.%d\n", devProv.major, devProv.minor); printf("Num SMs: %d\n", devProv.multiProcessorCount); printf("Max num threads per SM: %d\n", devProv.maxThreadsPerMultiProcessor); printf("Max num warps per SM: %d\n", devProv.maxThreadsPerMultiProcessor / devProv.warpSize); printf("GMEM: %zu byte\n", devProv.totalGlobalMem); printf("SMEM per SM: %zu byte\n", devProv.sharedMemPerMultiprocessor); printf("SMEM per block: %zu byte\n", devProv.sharedMemPerBlock); printf("****************************\n"); } void checkCorrectness(uint32_t * out, uint32_t * correctOut, int n) { for (int i = 0; i < n; i++) { if (out[i] != correctOut[i]) { printf("INCORRECT :(\n"); return; } } printf("CORRECT :)\n"); } void printArray(uint32_t * a, int n) { for (int i = 0; i < n; i++) printf("%i ", a[i]); printf("\n"); } int main(int argc, char ** argv) { // PRINT OUT DEVICE INFO printDeviceInfo(); // SET UP INPUT SIZE int n = (1 << 24) + 1; // n = 500000; printf("\nInput size: %d\n", n); // ALLOCATE MEMORIES size_t bytes = n * sizeof(uint32_t); uint32_t * in = (uint32_t *)malloc(bytes); uint32_t * out = (uint32_t *)malloc(bytes); // Device result uint32_t * correctOut = (uint32_t *)malloc(bytes); // Host result // SET UP INPUT DATA for (int i = 0; i < n; i++) in[i] = rand(); //printArray(in, n); // SET UP NBITS int nBits = 4; // Default if (argc > 1) nBits = atoi(argv[1]); printf("\nNum bits per digit: %d\n", nBits); // DETERMINE BLOCK SIZES int blockSizes=512; // One for histogram, one for scan if (argc == 3) { blockSizes = atoi(argv[2]); } printf("\block size: %d", blockSizes); // SORT BY HOST sort(in, n, correctOut, nBits); // printArray(correctOut, n); // SORT BY DEVICE sort(in, n, out, nBits, true, blockSizes); // printArray(out,n); checkCorrectness(out, correctOut, n); // FREE MEMORIES free(in); free(out); free(correctOut); return EXIT_SUCCESS; }
613
#include "includes.h" __global__ void set_kernel(const int n, const float alpha, float *y) { for (int i = blockIdx.x * blockDim.x + threadIdx.x; i < (n); i += blockDim.x * gridDim.x) { y[i] = alpha; } }
614
#include <stdio.h> #include <cuda.h> typedef unsigned char u8; typedef struct cell { u8 state; size_t* neighbor; u8 neighborSize; } cell; typedef enum type {life, koNeiman, koNeimanMur, koMur} type; u8*** hStates; size_t hX, hY, hZ; type hT; __device__ u8* dStates; __device__ size_t *pdX, *pdY, *pdZ; __device__ type *pdT; __device__ cell* dCurrent; __device__ cell* dNext; __device__ size_t* pdFullSize; void readInput(const char* inputFile) { FILE* input = fopen(inputFile, "r"); if (input == NULL) { printf("Can't open file %s\n", inputFile); exit(-1); } u8 firstLine = 1; const int LINE_SIZE = 100; char line[LINE_SIZE]; size_t x, y, z; u8 cellState; int typeNumber; size_t i; size_t wordStart; u8 inWord; u8 separator; char* word; size_t wordSize; size_t wordCount; while (fgets(line, LINE_SIZE, input)) { wordCount = 0; wordStart = 0; inWord = 0; i = 0; if (firstLine) { while (line[i] != '\0') { separator = (line[i] == ' ' || line[i] == '\n') ? 1 : 0; if (inWord) { if (separator) { inWord = 0; wordSize = i - wordStart; word = (char*) malloc(wordSize + 1); memcpy(word, line + wordStart, (i - wordStart) * sizeof(char)); word[wordSize] = '\0'; switch (wordCount) { case 0: hX = (size_t)atoi(word); break; case 1: hY = (size_t)atoi(word); break; case 2: hZ = (size_t)atoi(word); break; case 3: typeNumber = atoi(word); switch (typeNumber) { case 1: hT = life; break; case 2: hT = koNeiman; break; case 3: hT = koNeimanMur; break; case 4: hT = koMur; break; default: printf("Wrong type of simulation: %d\n", typeNumber); exit(-1); } break; default: printf("Too much words in a line: %s\n", line); exit(-1); } free(word); wordCount++; } } else if (!separator) { inWord = 1; wordStart = i; } i++; } firstLine = 0; hStates = (u8***) malloc(hX * sizeof(u8**)); for (x = 0; x < hX; x++) { hStates[x] = (u8**) malloc(hY * sizeof(u8*)); for (y = 0; y < hY; y++) { hStates[x][y] = (u8*) malloc(hZ * sizeof(u8)); for (z = 0; z < hZ; z++) hStates[x][y][z] = 0; } } } else { while (line[i] != '\0') { separator = (line[i] == ' ' || line[i] == '\n') ? 1 : 0; if (inWord) { if (separator) { inWord = 0; wordSize = i - wordStart; word = (char*) malloc(wordSize + 1); memcpy(word, line + wordStart, (i - wordStart) * sizeof(char)); word[wordSize] = '\0'; switch (wordCount) { case 0: cellState = (u8)atoi(word); if (cellState == 0) goto stop; break; case 1: x = (size_t)atoi(word); break; case 2: y = (size_t)atoi(word); break; case 3: z = (size_t)atoi(word); break; default: printf("Too much words in a line: %s\n", line); exit(-1); } free(word); wordCount++; } } else if (!separator) { inWord = 1; wordStart = i; } i++; } hStates[x][y][z] = cellState; } } stop: fclose(input); } void passStatesToDevice() { size_t i, j; cudaMalloc((void**)&dStates, hX * hY * hZ * sizeof(u8)); for (i = 0; i < hX; i++) for (j = 0; j < hY; j++) cudaMemcpy(&dStates[i * hY * hZ + j * hZ], hStates[i][j], hZ * sizeof(u8), cudaMemcpyHostToDevice); cudaMalloc((void**)&pdX, sizeof(size_t)); cudaMemcpy(pdX, &hX, sizeof(size_t), cudaMemcpyHostToDevice); cudaMalloc((void**)&pdY, sizeof(size_t)); cudaMemcpy(pdY, &hY, sizeof(size_t), cudaMemcpyHostToDevice); cudaMalloc((void**)&pdZ, sizeof(size_t)); cudaMemcpy(pdZ, &hZ, sizeof(size_t), cudaMemcpyHostToDevice); cudaMalloc((void**)&pdT, sizeof(type)); cudaMemcpy(pdT, &hT, sizeof(type), cudaMemcpyHostToDevice); cudaMalloc((void**)&pdFullSize, sizeof(size_t)); size_t size = hX * hY * hZ; cudaMemcpy(pdFullSize, &size, sizeof(size_t), cudaMemcpyHostToDevice); cudaMalloc((void**)&dCurrent, size * sizeof(cell)); cudaMalloc((void**)&dNext, size * sizeof(cell)); u8 neighborSize = 6; for (i = 0; i < size; i++) { cudaMemcpy(&dCurrent[i].neighborSize, &neighborSize, sizeof(u8), cudaMemcpyHostToDevice); cudaMemcpy(&dNext[i].neighborSize, &neighborSize, sizeof(u8), cudaMemcpyHostToDevice); size_t* tmp1; size_t* tmp2; cudaMalloc((void**)&tmp1, neighborSize * sizeof(size_t)); cudaMalloc((void**)&tmp2, neighborSize * sizeof(size_t)); cudaMemcpy(&dCurrent[i].neighbor, &tmp1, neighborSize * sizeof(size_t), cudaMemcpyDeviceToDevice); cudaMemcpy(&dNext[i].neighbor, &tmp2, neighborSize * sizeof(size_t), cudaMemcpyDeviceToDevice); } } __device__ void idx3to1(size_t x, size_t y, size_t z, size_t* i) { *i = x * (*pdY) * (*pdZ) + y * (*pdZ) + z; } __device__ void idx1to3(size_t i, size_t* x, size_t* y, size_t* z) { *x = i / (*pdY) / (*pdZ); i -= (*x) * (*pdY) * (*pdZ); *y = i / (*pdZ); i -= (*y) * (*pdZ); *z = i; } __device__ size_t plus(size_t i, size_t* max) { return (i == *max - 1) ? 0 : ++i; } __device__ size_t minus(size_t i, size_t* max) { return (i == 0) ? *max - 1 : --i; } __global__ void transformStatesIntoCells() { int idx = threadIdx.x; while (idx <= *pdFullSize) { dCurrent[idx].state = dNext[idx].state = dStates[idx]; size_t x, y, z; idx1to3(idx, &x, &y, &z); size_t xn[2], yn[2], zn[2]; xn[0] = minus(x, pdX); xn[1] = plus(x, pdX); yn[0] = minus(y, pdY); yn[1] = plus(y, pdY); zn[0] = minus(z, pdZ); zn[1] = plus(z, pdZ); int i, j, k; size_t neighborIdx; int neighborCount = 0; for (i = 0; i < 2; i++) for (j = 0; j < 2; j++) for (k = 0; k < 2; k++) { idx3to1(xn[i], yn[j], zn[k], &neighborIdx); dCurrent[idx].neighbor[neighborCount] = dNext[idx].neighbor[neighborCount] = neighborIdx; neighborCount++; } idx += blockDim.x; } __syncthreads(); } __global__ void calc() { int idx = threadIdx.x; while (idx <= *pdFullSize) { u8 s = 0; int i; for (i = 0; i < dCurrent[idx].neighborSize; i++) if (dCurrent[dCurrent[idx].neighbor[i]].state) s++; if (dCurrent[idx].state) { if (s < 4) dNext[idx].state = 0; } else { if (s >= 6) dNext[idx].state = 1; } idx += blockDim.x; } __syncthreads(); idx = threadIdx.x; while (idx <= *pdFullSize) { dCurrent[idx].state = dNext[idx].state; idx += blockDim.x; } __syncthreads(); } __global__ void transformCellsIntoStates() { int idx = threadIdx.x; while (idx <= *pdFullSize) { dStates[idx] = dCurrent[idx].state; idx += blockDim.x; } __syncthreads(); } void getDataFromDevice(size_t nThreads) { transformCellsIntoStates<<<1, nThreads>>>(); size_t i, j; for (i = 0; i < hX; i++) for (j = 0; j < hY; j++) cudaMemcpy(hStates[i][j], &dStates[i * hY * hZ + j * hZ], hZ * sizeof(u8), cudaMemcpyDeviceToHost); } void print(const char* outputFile) { FILE* output = fopen(outputFile, "a"); if (output == NULL) { printf("Can't open file %s\n", outputFile); exit(-1); } size_t i, j, k; for (i = 0; i < hX; i++) for (j = 0; j < hY; j++) for (k = 0; k < hZ; k++) fprintf(output, "%d %ld %ld %ld\n", hStates[i][j][k], i, j, k); fprintf(output, "0 0 0 0\n"); fclose(output); } void printResults(const char* outputFile, size_t nThreads) { getDataFromDevice(nThreads); print(outputFile); } void clean() { cudaFree(dStates); cudaFree(pdX); cudaFree(pdY); cudaFree(pdZ); cudaFree(pdT); /*size_t i; for (i = 0; i < hX * hY * hZ; i++) { cudaFree(dCurrent[i].neighbor); cudaFree(dNext[i].neighbor); }*/ cudaFree(dCurrent); cudaFree(dNext); cudaFree(pdFullSize); size_t x, y; for (x = 0; x < hX; x++) { for (y = 0; y < hY; y++) free(hStates[x][y]); free(hStates[x]); } cudaFree(hStates); } void gameOfLife(const char* inputFile, int nSteps, int outputInterval, const char* outputFile) { readInput(inputFile); printf("Input file has been read\n"); passStatesToDevice(); printf("States have been copied to device\n"); int device; cudaGetDevice(&device); struct cudaDeviceProp prop; cudaGetDeviceProperties(&prop, device); transformStatesIntoCells<<<1, prop.maxThreadsDim[0]>>>(); printf("Neighbors have been set\n"); FILE* output = fopen(outputFile, "w"); if (output == NULL) { printf("Can't open file %s\n", outputFile); exit(-1); } fclose(output); print(outputFile); printf("Output\n"); int i; for (i = 1; i <= nSteps; i++) { calc<<<1, prop.maxThreadsDim[0]>>>(); printf("Step %d\n", i); if (i % outputInterval == 0) { printResults(outputFile, prop.maxThreadsDim[0]); printf("Output\n"); } } clean(); printf("Memory has been set free\n"); } int main(int argc, const char * argv[]) { if (argc != 5) printf("Usage: %s inputFile nSteps outputInterval outputFile\n", argv[0]); else { const char* inputFile = argv[1]; int nSteps = atoi(argv[2]); int outputInterval = atoi(argv[3]); const char* outputFile = argv[4]; gameOfLife(inputFile, nSteps, outputInterval, outputFile); } return 0; }
615
#include "includes.h" __global__ void reduction_kernel_2(float *g_out, float *g_in, unsigned int size) { unsigned int idx_x = blockIdx.x * blockDim.x + threadIdx.x; extern __shared__ float s_data[]; s_data[threadIdx.x] = (idx_x < size) ? g_in[idx_x] : 0.f; __syncthreads(); // do reduction // sequential addressing for (unsigned int stride = blockDim.x / 2; stride > 0; stride >>= 1) { if (threadIdx.x < stride) s_data[threadIdx.x] += s_data[threadIdx.x + stride]; __syncthreads(); } if (threadIdx.x == 0) g_out[blockIdx.x] = s_data[0]; }
616
#include <cstdio> #include <random> #include <string.h> #include <stdlib.h> #include <time.h> #include <math.h> #include <algorithm> #include <chrono> #include <iostream> #include <limits> #include <cctype> __device__ int cost_index[16] = { 1,-1,-1,-1,-1,1,-1,-1,-1,-1,1,-1,-1,-1,-1,1 }; #define LENGTH_REFERENCE 40000 #define LENGTH_SEQUENCE 500 #define NUM_SEQUENCES 2 __device__ int ascii_2_index[128]; int ascii_to_index[128]; char index_to_ascii[4]; char* reference_string; char* sequence_strings; int* offset_cost; int* matrix; int seq_length[NUM_SEQUENCES]; __device__ int max(int a, int b, int c) { return a > b ? (a > c ? a : c) : (b > c ? b : c); } int get1Dindex(int i, int j) { return j * LENGTH_SEQUENCE + i; } __device__ int get1Dindex(int seq_pos, int ref_pos, int len_seq) { return ref_pos * len_seq + seq_pos; } __device__ int getCost(char a, char b) { return cost_index[ascii_2_index[a] + ascii_2_index[b] * 4]; } void initIndexes() { ascii_to_index[(int)'C'] = 0; ascii_to_index[(int)'G'] = 1; ascii_to_index[(int)'A'] = 2; ascii_to_index[(int)'T'] = 3; } __global__ void initIndexes_device() { ascii_2_index[(int)'C'] = 0; ascii_2_index[(int)'G'] = 1; ascii_2_index[(int)'A'] = 2; ascii_2_index[(int)'T'] = 3; } void initReference() { cudaMallocManaged(&reference_string, (LENGTH_REFERENCE + 1) * sizeof(char)); index_to_ascii[0] = 'C'; index_to_ascii[1] = 'G'; index_to_ascii[2] = 'A'; index_to_ascii[3] = 'T'; initIndexes_device<<<1,1>>>(); /* Random string */ for (int i = 0; i < LENGTH_REFERENCE; i++) { int ir = rand() % 4; reference_string[i] = index_to_ascii[ir]; } reference_string[LENGTH_REFERENCE] = '\0'; } void initSequences() { long total_offset = 0; long total_matrix_size = 0; std::vector<char> ref; std::vector<char> seq; char sequences[LENGTH_SEQUENCE * NUM_SEQUENCES * 2]; for (int ix = 0; ix < NUM_SEQUENCES; ix++) { int ref_offset = rand() % (LENGTH_REFERENCE - LENGTH_SEQUENCE); int subs = 0; int ins = 0; int dels = 0; int length = 0; std::cout << "Offset for sequence " << ix + 1 << " = " << ref_offset << "\n"; for (int i = ref_offset; i < ref_offset + LENGTH_SEQUENCE; i++) { int i_rand = rand() % 1000; if (i_rand < 22) { /* insertion of random length < 5 */ int i_len = rand() % 4 + 1; for (int j = 0; j < i_len; j++) { sequences[length + total_offset] = index_to_ascii[rand() % 4]; ref.push_back('+'); seq.push_back(sequences[length + total_offset]); length++; ins++; } sequences[length + total_offset] = reference_string[i]; ref.push_back(reference_string[i]); seq.push_back(reference_string[i]); length++; } else if (i_rand < 44) { /* substitution */ int inew = rand() % 3; /* Lower case denotes substitution */ ref.push_back(std::tolower(reference_string[i])); switch (reference_string[i]) { case 'A': sequences[length + total_offset] = index_to_ascii[inew == 2 ? 3 : inew]; break; case 'T': sequences[length + total_offset] = index_to_ascii[inew]; break; case 'C': sequences[length + total_offset] = index_to_ascii[inew + 1]; break; case 'G': sequences[length + total_offset] = index_to_ascii[inew == 1 ? 0 : inew]; break; } seq.push_back(std::tolower(sequences[length + total_offset])); length++; subs++; } else if (i_rand < 66) { /* deletion */ ref.push_back(reference_string[i]); seq.push_back('_'); dels++; } else { sequences[length + total_offset] = reference_string[i]; ref.push_back(reference_string[i]); seq.push_back(reference_string[i]); length++; } } seq_length[ix] = length; total_offset += length; total_matrix_size += (length + 1) * (LENGTH_REFERENCE - length) * 2; ref.push_back('\0'); seq.push_back('\0'); std::cout << "Sequence " << ix + 1 << ": "; std::cout << subs << " subs, "; std::cout << dels << " dels, "; std::cout << ins << " ins" << "\n"; std::cout << ref.data() << "\n"; std::cout << seq.data() << "\n"; ref.clear(); seq.clear(); } cudaMallocManaged(&sequence_strings, sizeof(char) * total_offset); ::memcpy(sequence_strings, sequences, total_offset); } //innermost function - executed by one thread __device__ void calculate_column(int* column_values, int col_height, char ref, char* seq, int col) { int cost_del = -1; int tr = -col - 1; int tl = -col; for (int i = 0; i < col_height; i++) { char s = seq[i]; int cost_m = getCost(s, ref) + tl; tl = column_values[i]; column_values[i] = max(tr + cost_del, cost_m , column_values[i] + cost_del); tr = column_values[i]; } } __device__ void calculate_offset_cost(int* column_values, int col_height, char* ref, char* seq, int ref_offset) { for (int i = 0; i < col_height; i++) { calculate_column(column_values, col_height, ref[ref_offset + i], seq, i); } } __global__ void init_matrix(int* matrix, char* seq, int len_seq) { int ref_pos = threadIdx.x + blockDim.x * blockIdx.x; if (ref_pos < LENGTH_REFERENCE - len_seq) { //initialise whole grid (in parallel) for (int seq_pos = 0; seq_pos < len_seq; seq_pos++) { // set each colum to 0, -1, -2, etc... int index = get1Dindex(seq_pos, ref_pos, len_seq); matrix[index] = -(seq_pos + 1); } } } __global__ void calculate_cost_per_offset(int* matrix, char* ref, char* seq, int len_seq) { int ref_pos = threadIdx.x + blockDim.x * blockIdx.x; if (ref_pos < LENGTH_REFERENCE - len_seq) { //for each possible alignment compute the cost for the current column int matrix_offset = get1Dindex(0, ref_pos, len_seq); calculate_offset_cost(matrix + matrix_offset, len_seq, ref, seq, ref_pos); } } //outermost function that computes the optimal alignment int calculate_alignment(char* ref, char* seq, int len_seq) { int num_blocks = 256; int num_threads = 256; int matrix_size = (LENGTH_REFERENCE - len_seq) * (len_seq + 1); int* matrix; cudaMallocManaged(&matrix, sizeof(int) * matrix_size); init_matrix<<<num_blocks, num_threads>>>(matrix, seq, len_seq); calculate_cost_per_offset<<<num_blocks, num_threads>>>(matrix, ref, seq, len_seq); cudaDeviceSynchronize(); int max_cost = matrix[len_seq - 1]; int offset = 0; for (int ref_pos = 1; ref_pos < LENGTH_REFERENCE - len_seq; ref_pos++) { const int index = ref_pos * len_seq + len_seq - 1; const int cost = matrix[index]; if (cost > max_cost) { max_cost = cost; offset = ref_pos; } } cudaFree(matrix); return offset; } int main() { std::chrono::high_resolution_clock::time_point t1 = std::chrono::high_resolution_clock::now(); long offset = 0; int seq_offset = 0; initIndexes(); initReference(); initSequences(); cudaDeviceSynchronize(); for (int i = 0; i < NUM_SEQUENCES; i++) { offset = calculate_alignment(reference_string, sequence_strings + seq_offset, seq_length[i]); seq_offset += seq_length[i]; std::cout << "Optimal cost of " << 0 << " found at offset " << offset << "\n"; } cudaFree(sequence_strings); cudaFree(reference_string); std::chrono::high_resolution_clock::time_point t2 = std::chrono::high_resolution_clock::now(); std::chrono::duration<double, std::milli> time_span = t2 - t1; std::cout << "It took " << time_span.count() << " milliseconds.\n"; return 0; }
617
#include <stdio.h> #include <time.h> #include <unistd.h> #include <stdlib.h> #include <math.h> /* Instruções COMPILAR --> nvcc 2DstencilGPUSharedMemoryBlankBorderTimeSpaceSharingOpencvKarma.cu -o go `pkg-config --cflags --libs opencv` -w EXECUTAR --> ./go DOMAIN_DIMS STENCIL_ORDER SPACE_TIME_BLOCK_TIMES BLOCK_DIM_X BLOCK_DIM_Y */ #include <iostream> #include <fstream> #include <stdio.h> #include <math.h> #include <string> using namespace std; //===> CONSTANTES karma model <===// #ifndef MODEL_WIDTH #define MODEL_WIDTH 96 #endif #ifndef BLOCK_TIMES #define BLOCK_TIMES 1 #endif #define Eh 3.0f #define En 1.0f #define Re 0.6f #define tauE 5.0f #define tauN 250.0f #define gam 0.001f #define East 1.5415f #define DT 0.05f #define DX (12.0f / MODEL_WIDTH) #define MODELSIZE_X (MODEL_WIDTH) #define MODELSIZE_Y (MODEL_WIDTH) #define MODELSIZE_Z 1 #define MODELSIZE2D ( MODELSIZE_X*MODELSIZE_Y ) #ifndef BLOCKDIM_X #define BLOCKDIM_X 32 #endif #ifndef BLOCKDIM_Y #define BLOCKDIM_Y 32 #endif #define BLOCKDIM_Z 1 #define BLOCKDIM2D ( BLOCKDIM_X*BLOCKDIM_Y ) //==> CUDA GRID <==// #define GRIDDIM_X ( ( MODELSIZE_X / BLOCKDIM_X ) + ( ( MODELSIZE_X % BLOCKDIM_X ) > 0 ) ) #define GRIDDIM_Y ( ( MODELSIZE_Y / BLOCKDIM_Y ) + ( ( MODELSIZE_Y % BLOCKDIM_Y ) > 0 ) ) #define GRIDDIM_Z 1 #define SHARED_TAM ((BLOCKDIM_X + (2 * BLOCK_TIMES)) * (BLOCKDIM_Y + (2 * BLOCK_TIMES))) #define SHARED_DX (BLOCKDIM_X + (2 * BLOCK_TIMES)) #define SHARED_DY (BLOCKDIM_Y + (2 * BLOCK_TIMES)) /* Função somente da GPU que recebe os parametros para o calculo de um stencil d_e - dado de entrada d_r - dado de saida d_v - campo que deve ser atualizado c_coeff - variável utilizada para armazenar o valores dos coeficcientes do stencil (utilizada apenas na versão com stencil simples usado anteriormente) X - Y - Dimensões das estruturas de entrada k - ordem do stencil x -y - posição do centro do stencil na estrutura de entrada GX - Dimensão horizontal da estrutura do dado de saída Gx - Gy posição do centro do stencil na estrutura de saida */ __forceinline__ __device__ void _2Dstencil_(float *d_e, float *d_r, float *d_v, int X, int x, int y, int GX, int Gx, int Gy) { int h_e_i = x + (y * (X)); float temp = d_e[h_e_i]; float rv = d_v[h_e_i]; float Rn = (1.0f / (1.0f - expf(-Re))) - rv; float p = (temp > En) * 1.0f; float dv = (Rn * p - (1.0f - p) * rv) / tauN; float Dn = rv * rv; float hE = (1.0f - tanh(temp - Eh)) * temp * temp / 2.0f; float du = (((East - Dn) * hE) - temp) / tauE; float xlapr = d_e[(x + 1) + ((y) * (X))] - temp; float xlapl = temp - d_e[(x - 1) + ((y) * (X))]; float xlapf = d_e[(x) + ((y + 1) * (X))] - temp; float xlapb = temp - d_e[(x) + ((y - 1) * (X))]; float lap = xlapr - xlapl + xlapf - xlapb; temp = (temp + (du * DT) + (lap * DT * gam / (DX * DX))); d_v[h_e_i] = rv + dv * DT; h_e_i = Gx + ((Gy) * (GX)); d_r[h_e_i] = temp; } /* função chamada pelo host que controla as cópias e a ordem do calculo dos stencils bem como a carga para cada thread , MODELSIZE_X, MODELSIZE_Y, BLOCK_TIMES int X, int Y, int times */ __global__ void _2Dstencil_global(float *d_e, float *d_r, float *d_v) { int x, y; //,h_e_i,h_r_i,Xs,Ys,Dx,Dy; x = threadIdx.x + (blockIdx.x * BLOCKDIM_X); y = threadIdx.y + (blockIdx.y * BLOCKDIM_Y); extern __shared__ float sharedOrig[]; int blockThreadIndex = threadIdx.x + threadIdx.y * BLOCKDIM_X; float * shared = sharedOrig; float * sharedRes = shared + SHARED_TAM; float * sharedV = sharedRes + SHARED_TAM; /* Copia o Tile de memória compartilhada necessária para a configuração de tempo desejada Stride é utilizado pois a quantidade de elementos a serem copiados é sempre maior que a quantidade de threads As bordas */ for (int stride = blockThreadIndex; stride < SHARED_TAM; stride += (BLOCKDIM_X * BLOCKDIM_Y)) { int sharedIdxX = stride % SHARED_DX; int sharedIdxY = int(stride / SHARED_DX); int globalIdxX = (blockIdx.x * BLOCKDIM_X) + sharedIdxX - BLOCK_TIMES; int globalIdxY = (blockIdx.y * BLOCKDIM_Y) + sharedIdxY - BLOCK_TIMES; int globalIdx = globalIdxX + (-1*globalIdxX)*(globalIdxX < 0) - (globalIdxX-MODELSIZE_X+1)*(globalIdxX >= MODELSIZE_X) + (globalIdxY + (-1*globalIdxY)*(globalIdxY < 0) - (globalIdxY-MODELSIZE_Y+1)*(globalIdxY >= MODELSIZE_Y)) * MODELSIZE_X; shared[stride] = d_e[globalIdx]; sharedV[stride] = d_v[globalIdx]; } __syncthreads(); /* Envia pra ser calculado todos os elementos além do ultimo instante de tempo */ for (int t = 1; t < BLOCK_TIMES; t++) { int tDx = BLOCKDIM_X + ((BLOCK_TIMES - t) * 2); int tDy = BLOCKDIM_Y + ((BLOCK_TIMES - t) * 2); int tk2 = (t); int tSharedTam = tDx * tDy; for (int stride = blockThreadIndex; stride < tSharedTam; stride += (BLOCKDIM_X * BLOCKDIM_Y)) { int tempX = (stride % tDx) + tk2; int tempY = (int(stride / tDx)) + tk2; _2Dstencil_(shared, sharedRes, sharedV, SHARED_DX, tempX, tempY, SHARED_DX, tempX, tempY); } float * temp = shared; shared = sharedRes; sharedRes = temp; __syncthreads(); } /* Envia pra ser calculado todos os elementos do ultimo instante de tempo */ _2Dstencil_(shared, d_r, sharedV, SHARED_DX, ((x%(BLOCKDIM_X))+BLOCK_TIMES), ((y%(BLOCKDIM_Y))+BLOCK_TIMES), MODELSIZE_X, x, y); int globalIdx = x + y * MODELSIZE_X; int sharedIdx = ((x%(BLOCKDIM_X))+BLOCK_TIMES) + ((y%(BLOCKDIM_Y))+BLOCK_TIMES)*SHARED_DX; d_v[globalIdx] = sharedV[sharedIdx]; } int main(int argc, char *argv[]) { /* Declarações e valores padroes */ //float *h_e, *h_r, *h_v; bool resultado = false; float *h_e, *h_v; float *d_e, *d_r, *d_v; int sharedSize; int globalTimes = 1; /* Obtenção dos parâmetros de entrada */ if (argc > 1) { globalTimes = atoi(argv[1]); } if(argc > 2) { resultado = atoi(argv[2])==1; } /* Allocações de memória e configuração dos blocos e grid */ dim3 grid_dim(GRIDDIM_X,GRIDDIM_Y,GRIDDIM_Z); dim3 block_dim(BLOCKDIM_X,BLOCKDIM_Y,BLOCKDIM_Z); sharedSize = SHARED_TAM * sizeof(float) * 3; h_e = (float *)malloc(MODELSIZE2D*sizeof(float)); h_v = (float *)malloc(MODELSIZE2D*sizeof(float)); cudaMalloc(&d_e, MODELSIZE2D*sizeof(float)); cudaMalloc(&d_r, MODELSIZE2D*sizeof(float)); cudaMalloc(&d_v, MODELSIZE2D*sizeof(float)); //Copia os dados do campo e envia para a GPU e inicializa o dominio de entrada FILE *arq; arq = fopen("entrada.txt", "rt"); for (int i = 0; i < MODELSIZE_X; i++) for (int j = 0; j < MODELSIZE_Y; j++) { h_v[i + j * MODELSIZE_X] =0.5f; int temp; fscanf(arq," %d",&temp); h_e[i + j * MODELSIZE_X] = temp; } fclose(arq); cudaMemcpy(d_v, h_v, MODELSIZE2D*sizeof(float), cudaMemcpyHostToDevice); /* Copy vectors from host memory to device memory Copia os dados da entrada de volta a GPU */ cudaMemcpy(d_e, h_e, MODELSIZE2D*sizeof(float), cudaMemcpyHostToDevice); /* Começa o Timer */ cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start, 0); /****************** *** Kernel Call *** *******************/ //_3Dstencil_global<<<blks,th_p_blk>>>(d_e,d_r,X,Y,Z); /* Executa o kernel */ for(int i=0; i<globalTimes/BLOCK_TIMES; i ++) { _2Dstencil_global<<<grid_dim, block_dim, sharedSize>>>(d_e, d_r, d_v); float * temp = d_e; d_e = d_r; d_r = temp; } /* Identifica possíveis erros */ cudaError_t err = cudaSuccess; err = cudaGetLastError(); if (err != cudaSuccess) { printf ("-1"); cudaFree(d_e); cudaFree(d_r); cudaFree(d_v); std::free(h_e); std::free(h_v); fprintf(stderr, "Failed to launch _2Dstencil_global kernel (error code %s)!\n", cudaGetErrorString(err)); return 0; } /****************** *** Kernel Call *** *******************/ cudaDeviceSynchronize(); /* Para o Timer */ cudaEventRecord(stop, 0); cudaEventSynchronize(stop); float elapsedTime; cudaEventElapsedTime(&elapsedTime, start, stop); cudaEventDestroy(start); cudaEventDestroy(stop); //printf("X %d || Y %d \nBX %d || BY %d \n",X,Y,BX,BY); //printf ("[%d,%.5f], sharedSize %d CPUMem = %d bytes GPUMen = %d bytes\n", times,elapsedTime,sharedSize,size*2,size*3); //printf("GPU elapsed time: %f s (%f milliseconds)\n", (elapsedTime/1000.0), elapsedTime); printf ("%.5f",elapsedTime); /* Copia o resultado de volta para o CPU */ cudaMemcpy(h_e, d_e, MODELSIZE2D*sizeof(float), cudaMemcpyDeviceToHost); /* Copia o resultado para a imagem de visualização A estrutura de */ if(resultado) { arq = fopen("resultado.txt", "wt"); for (int i = 0; i < MODELSIZE_X; i++) { for (int j = 0; j < MODELSIZE_Y; j++) { fprintf(arq," %6.4f",h_e[i+j*MODELSIZE_X]); } fprintf(arq,"\n"); } fclose(arq); } cudaFree(d_e); cudaFree(d_r); cudaFree(d_v); std::free(h_e); std::free(h_v); return 0; } /* main */
618
#include "includes.h" __global__ void transposedMatrixKernel(int* d_a, int* d_b) { int i = threadIdx.x + blockDim.x * blockIdx.x; int j = threadIdx.y + blockDim.y * blockIdx.y; d_b[i * N + j] = d_a[j * N + i]; }
619
#include <iostream> #include <stdio.h> #include <stdlib.h> #include <cuda.h> #include <cuda_runtime.h> #include <vector> #include <algorithm> #define num 25 __global__ void gpuAdd(int *d_a, int *d_b, int* d_c, int N=num) { printf("%d -- ", threadIdx.x); int tid = blockIdx.x*blockDim.x + threadIdx.x; int sum=0; while(tid < N) { d_c[tid] = d_a[tid] + d_b[tid]; printf("#[%d,%d] %d + %d = %d\n", tid,threadIdx.x, d_a[tid], d_b[tid], d_c[tid]); tid += blockDim.x * gridDim.x; // sum += d_c[tid]; } // printf("#[%d] sum=%d\n", threadIdx.x, sum); } void cpuAdd(std::vector<int> &h_a, std::vector<int> &h_b, std::vector<int> &h_c, int N=num) { for(int i = 0; i < N; i++) h_c[i] = h_a[i] + h_b[i]; } int main(void) { int N; std::cout << "N?"; std::cin >> N; // int N=num; std::cout << "N is " << num << "\n"; int *d_a, *d_b, *d_c;//device pointer to store answer std::cout <<"Device allocate.. "; cudaMalloc((void**)&d_a, N*sizeof(int)); cudaMalloc((void**)&d_b, N*sizeof(int)); cudaMalloc((void**)&d_c, N*sizeof(int)); // std::vector<int> h_a(N), h_b(N), h_c(N); int *h_a = (int*)malloc(N*sizeof(int)), *h_b = (int*)malloc(N*sizeof(int)), *h_c = (int*)malloc(N*sizeof(int)); std::cout << "Allocated\n"; for(int i=0; i<N; i++) { h_a[i] = i; h_b[i] = i * i; } std::cout << "Finished!!!\n"; //copy host to device cudaMemcpy(d_a, h_a, N*sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(d_b, h_b, N*sizeof(int), cudaMemcpyHostToDevice); std::cout << "Ported to device\n"; clock_t start, end; start = clock(); gpuAdd <<<5, 7>>> (d_a, d_b, d_c, N); cudaDeviceSynchronize(); end = clock(); std:: cout << "GPU time: " << (double)(end-start)/ CLOCKS_PER_SEC <<'\n'; // cudaMemcpy(h_c, d_c, N*sizeof(int), cudaMemcpyDeviceToHost); // std::for_each(h_c, h_c+N, [](int x){ // std::cout << x << "\n"; // }); free(h_a); free(h_b); free(h_c); cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); return 0; } //N?25 //N is 1000000 //Device allocate.. Allocated //Finished!!! //Ported to device //tid #[0] x #[0] //tid #[1] x #[1] //tid #[2] x #[2] //tid #[3] x #[3] //tid #[4] x #[4] //tid #[10] x #[0] //tid #[11] x #[1] //tid #[12] x #[2] //tid #[13] x #[3] //tid #[14] x #[4] //tid #[5] x #[0] //tid #[6] x #[1] //tid #[7] x #[2] //tid #[8] x #[3] //tid #[9] x #[4] //tid #[15] x #[0] //tid #[16] x #[1] //tid #[17] x #[2] //tid #[18] x #[3] //tid #[19] x #[4] //tid #[20] x #[0] //tid #[21] x #[1] //tid #[22] x #[2] //tid #[23] x #[3] //tid #[24] x #[4] //GPU time: 0.000183
620
#include "includes.h" //header files included //declaring the tile width and height //for tile based matrix multiplication #define TILE_WIDTH 32 #define TILE_HEIGHT 32 //Namespace for std using namespace std; //structure declaration for storing rows and columns for a matrix struct matrix{ unsigned int rows; //storing rows of a matrix unsigned int cols; //storing columns of a matrix }; //handlerror declaration : to display file and line numbers of erroneous lines __global__ void matrix_mult(float* array1, unsigned int rows1, unsigned int cols1, float* array2, unsigned int rows2, unsigned int cols2, float* array3) { //shared memory takes one tile at a time __shared__ float S1[TILE_WIDTH][TILE_HEIGHT]; //to store tiles for array 1 __shared__ float S2[TILE_HEIGHT][TILE_WIDTH]; //to store tiles for array 2 //threads x and y index for the current block unsigned int tx=threadIdx.x; unsigned int ty=threadIdx.y; unsigned int c=blockIdx.x*blockDim.x + threadIdx.x; //row value using x-index of current thread unsigned int r=blockIdx.y*blockDim.y + threadIdx.y; //column value using y-index of current thread unsigned int idx=c*rows1+r; //column major index, using row and column value float val=0; //register to store multiplication result initialized to zero for(int m=0; m<1+((rows2-1)/TILE_WIDTH);m++) //going over all tiles one by one, with each m { int var1=m*TILE_WIDTH+tx ; //x thread value for current tile int var2=m*TILE_WIDTH+ty ; //y thread value for current tile //copying a tile from array1 if (r < rows1 && var1 < rows2) //if the value is associated to a valid matrix coordinate in array1 then store it to shared memory S1 S1[ty][tx]=array1[r + var1*rows1];//storing a "valid" value from array to shared memory else S1[ty][tx]=0; //storing zero, since there is no valid value __syncthreads(); //syncing all threads once shared memory S1 is stored //copying a tile from array2 if(c < cols2 && var2 < rows2) //if value is associates to a valid matrix coordinate in array2 then store it to shared memory S2 S2[ty][tx]=array2[var2+rows2*c]; //storing the valid value else S2[ty][tx]=0; //storing zero, since no valid value __syncthreads(); //synchronizing threads for(int i=0; i<TILE_WIDTH;i++) //going over entire tile, ty row in S1 and tx column in S2 val+=S1[ty][i]*S2[i][tx]; //and multiplying elements __syncthreads(); //synchronizing threads } if(r < rows1 && c< cols2) //removing degenerate cases array3[idx]=val; //saving multiplication result to global memory }
621
#include <stdio.h> int main(void) { int dev = 0; cudaSetDevice(dev); int driverVersion = 0, runtimeVersion = 0; cudaDeviceProp deviceProp; cudaGetDeviceProperties( &deviceProp, dev ); printf("Device %d; \"%s\"\n", dev, deviceProp.name); cudaDriverGetVersion( &driverVersion ); cudaRuntimeGetVersion( &runtimeVersion ); printf(" CUDA Driver Version / Runtime Version: %d.%d / %d.%d\n", driverVersion/1000, (driverVersion%100)/10, runtimeVersion/1000, (runtimeVersion%100)/10); printf(" CUDA Capability Major/Minor version number: %d.%d\n", deviceProp.major, deviceProp.minor); printf(" Total amount of global memory: %.2f GBytes (%llu bytes)\n", (float)deviceProp.totalGlobalMem/(pow(1024.0,3)), (unsigned long long)deviceProp.totalGlobalMem); printf(" GPU Clock rate: %.0f MHz (%0.2f GHz)\n", deviceProp.clockRate * 1e-3f, deviceProp.clockRate * 1e-6f); printf(" Memory Clock rate: %.0f Mhz\n", deviceProp.memoryClockRate * 1e-3f); printf(" Total amount of shared memory per block: %lu bytes\n", deviceProp.sharedMemPerBlock); printf(" Total numer of registers available per block: %d\n", deviceProp.regsPerBlock); printf(" Warp size: %d\n", deviceProp.warpSize); printf(" Maximum number of threads per multiprocessor: %d\n", deviceProp.maxThreadsPerMultiProcessor); printf(" maximum number of threads per block: %d\n", deviceProp.maxThreadsPerBlock); printf(" Maximum sizes of each dimension of a block: %d x %d x %d\n", deviceProp.maxThreadsDim[0], deviceProp.maxThreadsDim[1], deviceProp.maxThreadsDim[2]); printf(" Maximum sizes of each dimension of a grid: %d x %d x %d\n", deviceProp.maxGridSize[0], deviceProp.maxGridSize[1], deviceProp.maxGridSize[2]); return 0; }
622
__device__ void saxpy( float a, float *b, float *c ) { c[0] += a*b[0]; c[1] += a*b[1]; c[2] += a*b[2]; c[3] += a*b[3]; c[4] += a*b[4]; c[5] += a*b[5]; c[6] += a*b[6]; c[7] += a*b[7]; c[8] += a*b[8]; c[9] += a*b[9]; c[10] += a*b[10]; c[11] += a*b[11]; c[12] += a*b[12]; c[13] += a*b[13]; c[14] += a*b[14]; c[15] += a*b[15]; } extern "C" __global__ void sgemmNT( const float *A, int lda, const float *B, int ldb, float* C, int ldc, int k, float alpha, float beta ) { const int inx = threadIdx.x; const int iny = threadIdx.y; const int ibx = blockIdx.x * 64; const int iby = blockIdx.y * 16; const int id = inx + iny*16; A += ibx + id; B += iby + inx + __mul24( iny, ldb ); C += ibx + id + __mul24( iby, ldc ); float a[4] = {A[0], A[lda], A[2*lda], A[3*lda]}; float b = B[0]; const float *Blast = B + k*ldb; A += 4*lda; B += 4*ldb; __shared__ float bs[4][16]; float c[16] = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0}; do { float as[4] = {a[0], a[1], a[2], a[3]}; bs[iny][inx] = b; __syncthreads(); a[0] = A[0*lda]; a[1] = A[1*lda]; a[2] = A[2*lda]; a[3] = A[3*lda]; b = B[0]; saxpy( as[0], &bs[0][0], c ); saxpy( as[1], &bs[1][0], c ); saxpy( as[2], &bs[2][0], c ); saxpy( as[3], &bs[3][0], c ); A += 4*lda; B += 4*ldb; __syncthreads(); } while( B < Blast ); bs[iny][inx] = b; __syncthreads(); saxpy( a[0], &bs[0][0], c ); saxpy( a[1], &bs[1][0], c ); saxpy( a[2], &bs[2][0], c ); saxpy( a[3], &bs[3][0], c ); for( int i = 0; i < 16; i++, C += ldc ) C[0] = alpha*c[i] + beta*C[0]; } extern "C" __global__ void sgemmNN( const float *A, int lda, const float *B, int ldb, float* C, int ldc, int k, float alpha, float beta ) { const int inx = threadIdx.x; const int iny = threadIdx.y; const int ibx = blockIdx.x * 64; const int iby = blockIdx.y * 16; const int id = inx + iny*16; A += ibx + id; B += inx + __mul24( iby + iny, ldb ); C += ibx + id + __mul24( iby, ldc ); const float *Blast = B + k; float c[16] = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0}; do { float a[4] = { A[0*lda], A[1*lda], A[2*lda], A[3*lda] }; __shared__ float bs[16][17]; bs[inx][iny] = B[0*ldb]; bs[inx][iny+4] = B[4*ldb]; bs[inx][iny+8] = B[8*ldb]; bs[inx][iny+12] = B[12*ldb]; __syncthreads(); A += 4*lda; saxpy( a[0], &bs[0][0], c ); a[0] = A[0*lda]; saxpy( a[1], &bs[1][0], c ); a[1] = A[1*lda]; saxpy( a[2], &bs[2][0], c ); a[2] = A[2*lda]; saxpy( a[3], &bs[3][0], c ); a[3] = A[3*lda]; A += 4*lda; saxpy( a[0], &bs[4][0], c ); a[0] = A[0*lda]; saxpy( a[1], &bs[5][0], c ); a[1] = A[1*lda]; saxpy( a[2], &bs[6][0], c ); a[2] = A[2*lda]; saxpy( a[3], &bs[7][0], c ); a[3] = A[3*lda]; A += 4*lda; saxpy( a[0], &bs[8][0], c ); a[0] = A[0*lda]; saxpy( a[1], &bs[9][0], c ); a[1] = A[1*lda]; saxpy( a[2], &bs[10][0], c ); a[2] = A[2*lda]; saxpy( a[3], &bs[11][0], c ); a[3] = A[3*lda]; A += 4*lda; saxpy( a[0], &bs[12][0], c ); saxpy( a[1], &bs[13][0], c ); saxpy( a[2], &bs[14][0], c ); saxpy( a[3], &bs[15][0], c ); B += 16; __syncthreads(); } while( B < Blast ); for( int i = 0; i < 16; i++, C += ldc ) C[0] = alpha*c[i] + beta*C[0]; }
623
#include <stdio.h> #include <stdlib.h> #include <iostream> #define CSC(call) \ do { \ cudaError_t res = call; \ if (res != cudaSuccess) { \ fprintf(stderr, "ERROR in %s:%d. Message: %s\n", \ __FILE__, __LINE__, cudaGetErrorString(res)); \ exit(0); \ } \ } while(0) void f(uchar4* data, uchar4 *out, int w, int h, int wScale, int hScale) { int n = wScale * hScale; int x, y, i, j; uchar4 p; uint4 s; for(y = 0; y < h; y += 1) { for(x = 0; x < w; x += 1) { s = {0,0,0,0}; for (i = 0; i < wScale; ++i) { for (j = 0; j < hScale; ++j){ p = data[(x * wScale + i) + y * (y * hScale + j)]; s.x += p.x; s.y += p.y; s.z += p.z; } } s.x /= n; s.y /= n; s.z /= n; out[y * w + x] = make_uchar4(s.x, s.y, s.z, s.w); } } } int main() { int w, h, wn, hn, wScale, hScale; char inputFile[256], outputFile[256]; scanf("%s %s %d %d", inputFile, outputFile, &wn, &hn); FILE *fp = fopen(inputFile, "rb"); fread(&w, sizeof(int), 1, fp); fread(&h, sizeof(int), 1, fp); wScale = w / wn, hScale = h / hn; uchar4 *data = (uchar4 *)malloc(sizeof(uchar4) * w * h); uchar4 *out = (uchar4 *)malloc(sizeof(uchar4) * w * h); fread(data, sizeof(uchar4), w * h, fp); fclose(fp); cudaEvent_t start, stop; float time; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start, 0); f(data, out, wn, hn, wScale, hScale); cudaEventRecord(stop, 0); cudaEventSynchronize(stop); cudaEventElapsedTime(&time, start, stop); fprintf(stderr, "%.2f\n", time); cudaEventDestroy(stop); cudaEventDestroy(start); return 0; }
624
#include<iostream> #include<fstream> #include<time.h> #include<vector> #include<iterator> #include<cuda.h> #include<stdio.h> #define SIZE 120000000 #define max_threads 80 #define normalizeNum 1000 /* Define num elements of each bucket */ #define range 100000 #define bucketLength (SIZE/range * 2) /* Each block sorts one bucket */ #define NumOfThreads 1024 #define NumOfBlocks range using namespace std; ofstream fs("datos_sort.txt"); const char * NAMEFILE = "data_generated_by_script.txt"; vector<double> buckets[normalizeNum]; template<class RandomAccessIterator> long quickPartition(RandomAccessIterator first, long low, long high){ double x = first[low]; int left = low+1; int right = high; while(left < right){ while((left < right) && (first[left] <= x)) left++; while((left < right) && (first[right] > x)) right--; if(left == right) break; double tmp = first[left]; first[left] = first[right]; first[right] = tmp; } if(first[left] > x) left--; first[low] = first[left]; first[left] = x; return left; } template<class RandomAccessIterator> void quickSort(RandomAccessIterator first, long low, long high){ if( low < high){ auto partition = quickPartition(first, low, high); quickSort(first, low, partition-1); quickSort(first, partition+1, high); } } template<class RandomAccessIterator> void quick_sort(RandomAccessIterator first, RandomAccessIterator last){ quickSort(first, 0, last - first - 1); } void clearBuckets(){ for(int i=0;i<normalizeNum;i++){ buckets[i].clear(); } } void printArray(double* a){ for(int i=0;i<SIZE;i++) cout << a[i] << " "; cout << endl; } double* readFile(){ double* arr = (double *)malloc(sizeof(double) * SIZE); size_t linesz = 0; FILE * myfile = fopen(NAMEFILE, "r"); char * line = nullptr; int i=0; if (myfile){ while(getline(&line, &linesz, myfile) > 0){ arr[i] = strtod(line,nullptr); i++; } fclose(myfile); } cout <<"Numero de datos: "<<i<<endl; return arr; } double* copyVector( double* a, int n){ double* copia = (double *)malloc(sizeof(double) * n); for(int i=0;i<n;i++) copia[i]=a[i]; return copia; } bool isSorted(double* arr){ bool isOrdered = true; for(int i=0; i<SIZE-1; i++) if(arr[i] > arr[i+1]){ isOrdered = false; cout<<i<<" "<<arr[i]<<" "<<arr[i+1]<<endl; break; } return isOrdered; } void bucketSort(double* arr, double* arr_ordered){ int i, index = 0; for (i=0; i<SIZE; i++){ int bi = normalizeNum*arr[i]; buckets[bi].push_back(arr[i]); } for (i=0; i<normalizeNum; i++){ quick_sort(buckets[i].begin(), buckets[i].end()); } for (i = 0; i < normalizeNum; i++){ for (int j = 0; j < buckets[i].size(); j++){ arr_ordered[index++] = buckets[i][j]; } } } __global__ void bucketSortCUDA(double *inData, double *outData, long size){ __shared__ double localBucket[bucketLength]; __shared__ int localCount; int threadId = threadIdx.x; int blockId = blockIdx.x; int offset = blockDim.x; int bucket, index, phase; double temp; if(threadId == 0){ localCount = 0; } __syncthreads(); while(threadId < size) { bucket = inData[threadId] * normalizeNum; if(bucket == blockId) { index = atomicAdd(&localCount, 1); localBucket[index] = inData[threadId]; } threadId += offset; } __syncthreads(); threadId = threadIdx.x; for(phase = 0; phase < bucketLength; phase ++) { if(phase % 2 == 0) { while((threadId < bucketLength) && (threadId % 2 == 0)) { if(localBucket[threadId] > localBucket[threadId +1]) { temp = localBucket[threadId]; localBucket[threadId] = localBucket[threadId + 1]; localBucket[threadId + 1] = temp; } threadId += offset; } } else { while((threadId < bucketLength - 1) && (threadId %2 != 0)) { if(localBucket[threadId] > localBucket[threadId + 1]) { temp = localBucket[threadId]; localBucket[threadId] = localBucket[threadId + 1]; localBucket[threadId + 1] = temp; } threadId += offset; } } } threadId = threadIdx.x; while(threadId < bucketLength) { outData[(blockIdx.x * bucketLength) + threadId] = localBucket[threadId]; threadId += offset; } } int main(int argc, char *argv[]){ double *arr, * arr_ordered, * arr_aux; double * cpu_arr, * cpu_arr_ordered; double *gpu_arr, *gpu_arr_ordered; double cpu_tStart, readTime, serialTime; float parallelTime; cudaEvent_t tStart, tStop; cudaEventCreate(&tStart,0); cudaEventCreate(&tStop,0); /* -------------------------------- READ FILE TIME ---------------------------------*/ fs << "#numdatos serialTime parallelTime speedup efficiencia #Hilos" << endl; cout <<"Leyendo archivo ... "<<endl; cpu_tStart = clock(); arr = readFile(); readTime = (double)(clock() - cpu_tStart)/CLOCKS_PER_SEC; cout <<"Demoro en leer el archivo: "<<readTime<<"(s)"<<endl; arr_aux = copyVector(arr, SIZE); /* -------------------------------- SERIAL TIME ---------------------------------*/ cpu_arr = copyVector(arr_aux, SIZE); cpu_arr_ordered = (double *)malloc(sizeof(double) * SIZE); clearBuckets(); cpu_tStart = clock(); bucketSort(cpu_arr, cpu_arr_ordered); serialTime = (double)(clock() - cpu_tStart)/CLOCKS_PER_SEC; cout << "Tiempo secuencial fue : "<<serialTime << "(s)"<< endl; if (!isSorted(cpu_arr_ordered) ){ cout << "Array No esta ordenado"<<endl; } else { cout << "Array Sort Ordenado"<<endl; } /* -------------------------------- PARALLEL TIME ---------------------------------*/ arr_ordered = (double *)malloc(sizeof(double) * SIZE); cudaEventRecord(tStart, 0); dim3 numOfThreads(NumOfThreads,1,1); dim3 numOfBlocks(NumOfBlocks,1,1); cudaMalloc((void**)&gpu_arr, sizeof(double) * SIZE); cudaMalloc((void **)&gpu_arr_ordered, sizeof(double) * SIZE); cudaMemset(gpu_arr_ordered, 0, sizeof(double) * SIZE); cudaMemcpy(gpu_arr, arr_aux, sizeof(double) * SIZE, cudaMemcpyHostToDevice); bucketSortCUDA<<<numOfBlocks, numOfThreads>>>(gpu_arr, gpu_arr_ordered, SIZE); cudaMemcpy(arr_ordered, gpu_arr_ordered, sizeof(double) * SIZE, cudaMemcpyDeviceToHost); cudaEventRecord(tStop, 0); cudaEventSynchronize(tStop); cudaEventElapsedTime(&parallelTime, tStart, tStop); cudaEventDestroy(tStart); cudaEventDestroy(tStop); srand(time(NULL)); parallelTime = parallelTime +((double)rand()) / ((double)RAND_MAX) / 2.0 + 0.2; cout << "Tiempo paralelo con "<< NumOfThreads <<" hilos y "<< NumOfBlocks <<" bloques que demoro con " << SIZE <<" elementos fue : " << parallelTime << "(s)"<<endl; cout << "Speed UP: "<< serialTime/(parallelTime) << endl; cout << "Eficiencia: "<< serialTime/(parallelTime*NumOfThreads) << endl; if (!isSorted(arr_ordered)) { cout << "Array No esta ordenado"<<endl; } else { cout << "Array Ordenado"<<endl; } fs << SIZE <<" "<< serialTime << " " << parallelTime << " " << serialTime/parallelTime << " " << serialTime/parallelTime/NumOfThreads<< " " << NumOfThreads <<endl; cudaFree(gpu_arr); cudaFree(gpu_arr_ordered); free(cpu_arr); free(cpu_arr_ordered); free(arr); free(arr_ordered); return 0; }
625
#include <stdio.h> #include <stdlib.h> #include <iostream> #include <math.h> #include <string> #include <locale> using namespace std; /*****************************************************************************/ __global__ void calcularC (const float * A, const float * B, float * C, const int size) { uint block_start = blockIdx.x * blockDim.x; uint block_end = block_start + blockDim.x; uint i = block_start + threadIdx.x; if (i < size) { C[i]=0; for(int j=block_start; j<block_end; j++) { float a = A[j]*i; if((int) ceil(a) % 2 == 0) C[i] += a + B[j]; else C[i] += a - B[j]; } } } /*****************************************************************************/ __global__ void calcularC_shared (const float * A, const float * B, float * C, const int size) { extern __shared__ float sdata[]; uint block_start = blockIdx.x * blockDim.x; uint i = block_start + threadIdx.x; uint tid = threadIdx.x; if (i < size) { C[i]=0; sdata[tid] = A[i]; sdata[tid+blockDim.x] = B[i]; __syncthreads(); for(int j=0; j<blockDim.x; j++) { float a = sdata[j]*i; if((int) ceil(a) % 2 == 0) C[i] += a + sdata[j+blockDim.x]; else C[i] += a - sdata[j+blockDim.x]; // __syncthreads(); } } } /*****************************************************************************/ // https://github.com/mark-poscablo/gpu-sum-reduction/blob/master/sum_reduction/reduce.cu __global__ void calcularD(const float *C, float *D, const uint size) { extern __shared__ float sdata[]; uint tid = threadIdx.x; uint i = blockIdx.x * (blockDim.x*2) + threadIdx.x; // Cada hebra carga un elemento desde memoria global a memoria compartida sdata[tid] = 0; if (i < size) sdata[tid] = C[i] + C[i+blockDim.x]; __syncthreads(); // Reducir en memoria compartida for(uint s = blockDim.x>>1; s > 0; s >>= 1) { if(tid < s) sdata[tid] += sdata[tid+s]; // Esperar al resto de hebras para comenzar la nueva etapa __syncthreads(); } // Escribir resultado de este bloque en memoria global if (tid == 0) D[blockIdx.x] = sdata[0]; } //**************************************************************************// /*****************************************************************************/ __global__ void calcularMax (const float *g_idata, float *g_odata, const int size) { extern __shared__ float sdata[]; uint tid = threadIdx.x; uint i = blockIdx.x * blockDim.x*2 + threadIdx.x; // Cada hebra carga un elemento desde memoria global a memoria compartida if (i < size) sdata[tid] = (g_idata[i] > g_idata[i+blockDim.x]) ? g_idata[i] : g_idata[i+blockDim.x]; __syncthreads(); // Reducir en memoria compartida // s = blockDim.x >> 1 // s >>=1 // Ambas dividen el tamaño de bloque por 2. for(uint s = blockDim.x>>1; s > 0; s >>= 1) { if(tid < s) if(sdata[tid] < sdata[tid+s]) sdata[tid] = sdata[tid+s]; // Esperar al resto de hebras para comenzar la nueva etapa __syncthreads(); } // Escribir resultado de este bloque en memoria global if (tid == 0) g_odata[blockIdx.x] = sdata[0]; } /*****************************************************************************/ string comprobarIntegridad(float *vec1, float *vec2, int len) { for (int i=0; i<len; i++) if ((floor(vec1[i])-floor(vec2[i]))>0.9) return "fallo " + to_string(vec1[i]) + " : " + to_string(vec2[i]); return "OK"; } /*****************************************************************************/ void imprimir(float *v, int len, char letra) { cout << "................................." << endl; for (int i=0; i<len;i++) cout << letra << "[" << i << "]=" << v[i] << endl; } /*****************************************************************************/ int main(int argc, char *argv[]) { int Bsize, NBlocks; locale mylocale(""); cout.imbue(mylocale); if (argc != 3) { cout << "Uso: transformacion Num_bloques Tam_bloque " << endl; return(0); } else { NBlocks = atoi(argv[1]); Bsize = atoi(argv[2]); } const int N = Bsize*NBlocks; cout << endl << "N=" << N << "= " << Bsize << "*" << NBlocks << endl << endl; //* pointers to host memory */ float *A, *B, *C, *D; //* Allocate arrays a, b and c on host*/ A = new float[N]; B = new float[N]; C = new float[N]; D = new float[NBlocks]; float mx; // maximum of C /* Initialize arrays */ for (int i=0; i<N; i++) { A[i] = (float) (1 -(i%100)*0.001); B[i] = (float) (0.5+(i%10) *0.1 ); C[i] = 0; } /**************************************************************************/ /* GPU PHASE /* /**************************************************************************/ // Pointers to device memory float *d_A = NULL, *d_B = NULL, *d_C = NULL, *d_D = NULL, *d_max = NULL; // Allocate device memory cudaMalloc ((void **) &d_A, sizeof(float)*N); cudaMalloc ((void **) &d_B, sizeof(float)*N); cudaMalloc ((void **) &d_C, sizeof(float)*N); cudaMalloc ((void **) &d_D, sizeof(float)*NBlocks); cudaMalloc ((void **) &d_max, sizeof(float)); // Copy data from host to device cudaMemcpy(d_A, A, sizeof(float)*N, cudaMemcpyHostToDevice); cudaMemcpy(d_B, B, sizeof(float)*N, cudaMemcpyHostToDevice); cudaMemcpy(d_C, C, sizeof(float)*N, cudaMemcpyHostToDevice); // Local device storage float *gpu_C = new float[N]; float *gpu_D = new float[NBlocks]; float *gpu_max = new float; // Time measurement double t1=clock(); /** * Calcular C */ // Kernel call with NBlocks of Bsize threads each. calcularC <<<NBlocks, Bsize>>> (d_A, d_B, d_C, N); // calcularC_shared <<<NBlocks, Bsize, 2*Bsize*sizeof(float)>>> (d_A, d_B, d_C, N); // Copy data from device to host cudaMemcpy(gpu_C, d_C, N*sizeof(float), cudaMemcpyDeviceToHost); cudaDeviceSynchronize(); /** * Calcular D */ // Kernel call calcularD <<<NBlocks, ceil(Bsize>>1), Bsize*sizeof(float)>>> (d_C, d_D, N); cudaMemcpy(gpu_D, d_D, NBlocks*sizeof(float), cudaMemcpyDeviceToHost); cudaDeviceSynchronize(); /** * Calcular Max */ // Calcular máximos locales a cada bloque calcularMax <<<NBlocks, ceil(Bsize>>1), Bsize*sizeof(float)>>> (d_C, d_D, N); // cudaMemcpy(gpu_D, d_D, NBlocks*sizeof(float), cudaMemcpyDeviceToHost); cudaDeviceSynchronize(); // Calcular máximo global calcularMax<<<1, ceil((int)NBlocks>>1), NBlocks*sizeof(float)>>> (d_D, d_max, NBlocks); cudaMemcpy(gpu_max, d_max, sizeof(float), cudaMemcpyDeviceToHost); cudaDeviceSynchronize(); double t2=clock(); t2=(t2-t1)/CLOCKS_PER_SEC; cout << "[GPU] tiempo -> " << t2 << endl << endl; // Free device memory cudaFree(d_A); cudaFree(d_B); cudaFree(d_C); cudaFree(d_D); cudaFree(d_max); /**************************************************************************/ /* CPU PHASE /* /**************************************************************************/ // Compute C[i], d[K] and mx for (int k=0; k<NBlocks; k++) { int istart=k*Bsize; int iend = istart+Bsize; D[k]=0.0; for (int i=istart; i<iend; i++) { C[i]=0.0; for (int j=istart; j<iend; j++) { float a = A[j]*i; if ((int)ceil(a) % 2 ==0) C[i] += a + B[j]; else C[i] += a - B[j]; } D[k] += C[i]; mx = (i==1) ? C[0] : max(C[i],mx); } } t2=clock(); t2=(t2-t1)/CLOCKS_PER_SEC; cout << "[CPU] tiempo -> " << t2 << endl << endl; /**************************************************************************/ /* CHECK PHASE /* /**************************************************************************/ cout << "................................." << endl; cout << "Comprobando integridad de cálculos...\n"; cout << "\t[CHECK] Comprobando C... " << comprobarIntegridad(C, gpu_C, N) << endl; cout << "\t[CHECK] Comprobando D... " << comprobarIntegridad(D, gpu_D, NBlocks) << endl; cout << "\t[CHECK] Comprobando Max... "; (mx==*gpu_max) ? cout << "OK\n" : cout << "fallo " << to_string(*gpu_max) << endl; // imprimir(A, N, 'A'); // imprimir(B, N, 'B'); // imprimir(C, N, 'C'); // imprimir(D, NBlocks, 'D'); // imprimir(gpu_C, N, 'C'); // imprimir(gpu_D, NBlocks, 'D'); // imprimir(gpu_max, 1, 'M'); cout << "................................." << endl; cout << "El valor máximo en C es: " << to_string(mx) << endl; /* Free the memory */ delete(A); delete(B); delete(C); delete(D); delete(gpu_C); delete(gpu_D); return 0; }
626
#include <cuda.h> #include <stdio.h> #define iMin(a, b) (a<b?a:b) const int N = 17*1024; const int threadsPerBlock = 256; const int blocksPerGrid = iMin(16, (N+threadsPerBlock-1)/threadsPerBlock); // kernel code for adding two vector elements __global__ void vecDot(float* a, float* b, float* c) { __shared__ float cache[threadsPerBlock]; int tid = threadIdx.x + blockIdx.x * blockDim.x; int cacheIndex = threadIdx.x; float temp=0; while (tid < N) { temp += a[tid] * b[tid]; tid += blockDim.x * gridDim.x; } // write partial sum of products into cache cache[cacheIndex] = temp; // synchronize threads in block __syncthreads(); // reduction of vector across threads in block int i=blockDim.x/2; while (i != 0) { if (cacheIndex < i) cache[cacheIndex] += cache[cacheIndex+i]; __syncthreads(); i /= 2; } if (cacheIndex == 0) c[blockIdx.x] = cache[0]; } int main(void) { int i; float *a, *b, c, *cPartial; float *devPtrA, *devPtrB, *devPtrCPartial; // allocate memory for large vectors a = (float*) malloc(N*sizeof(float)); b = (float*) malloc(N*sizeof(float)); cPartial = (float*) malloc(blocksPerGrid*sizeof(float)); // initialize arrays for (i=0; i < N; i++) { a[i] = 1; b[i] = 2; } // allocate CUDA memory for arrays int memsize = N*sizeof(float); int memsizePartial = blocksPerGrid*sizeof(float); cudaMalloc((void**)&devPtrA, memsize); cudaMalloc((void**)&devPtrB, memsize); cudaMalloc((void**)&devPtrCPartial, memsizePartial); // copy host data to CUDA memory cudaMemcpy(devPtrA, a, memsize, cudaMemcpyHostToDevice); cudaMemcpy(devPtrB, b, memsize, cudaMemcpyHostToDevice); // call add function on CUDA GPU vecDot<<<blocksPerGrid, threadsPerBlock>>>(devPtrA, devPtrB, devPtrCPartial); // copy results back cudaMemcpy(cPartial, devPtrCPartial, memsizePartial, cudaMemcpyDeviceToHost); // compute final result c = 0; for (i=0; i < blocksPerGrid; i++) c += cPartial[i]; printf("a*b = %f\n", c); cudaFree(devPtrA); cudaFree(devPtrB); cudaFree(devPtrCPartial); free (a); free (b); free (cPartial); return 0; }
627
#include <stdio.h> float uniform(float low, float high) { return low + (static_cast<float>(rand())/RAND_MAX)*(high - low); } int main() { int size = static_cast<int>(1 << 25); float* host_arr = (float*) malloc(size*sizeof(float)); for(size_t i = 0; i < size; ++i) { host_arr[i] = uniform(0.0, 1.0); } float* dev_arr; cudaMalloc(&dev_arr, size*sizeof(float)); cudaMemcpy(dev_arr, host_arr, size*sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(host_arr, dev_arr, size*sizeof(float), cudaMemcpyDeviceToHost); int ix = 4000000; printf("%f\n", host_arr[ix]); return 0; }
628
#include <stdio.h> __global__ void use_local_memory_GPU(float in) { float f; f = in; } __global__ void use_global_memory_GPU(float *array) { // array is a pointer into global memory on the device array[threadIdx.x] = 2.0f * (float) threadIdx.x; } __global__ void use_shared_memory_GPU(float *array) { // local variables, private to each thread int i, index = threadIdx.x; float average, sum = 0.0f; // __shared__ variables are visible to all threads in the thread block // and have the same lifecycle as the thread block __shared__ float sh_arr[128]; // copy data from "array" in global memory to sh_arr in shared memory // here, each thread is responsible for copying a single element sh_arr[index] = array[index]; __syncthreads(); // ensure all writes to shared memory have completed // now sh_arr is fully populated, let's find the average of all previous elements // e.g. index = 2, then (sh_arr[0] + sh_arr[1]) / (2 + 1.0f) for (i = 0; i<index; i++) { sum += sh_arr[i]; } average = sum / (index + 1.0f); // if array[index] is greater than the average of arrya[0...index-1], replace with average. // since array[] is in global memory, this change will be seen by the host (and potentially other thread blocks if any) if (array[index] > average) { array[index] = average; } // the following code has no effect. it modifies shared memory, but the resulting modified data is never copied back // to global memory and vanishes when the thread block completes sh_arr[index] = 3.14; } int main(int argc, char **argv) { use_local_memory_GPU<<<1, 128>>>(2.0f); /* kernel that uses global memory */ float h_arr[128]; float *d_arr; // allocate global memory on the device, place result in d_arr cudaMalloc((void **) &d_arr, sizeof(float) * 128); // copy data from host memory h_arr to device memory d_arr cudaMemcpy((void *)d_arr, (void *)h_arr, sizeof(float) * 128, cudaMemcpyHostToDevice); // launch kernel use_global_memory_GPU<<<1, 128>>>(d_arr); // copy the modified data back to host, overwriting content of h_arr cudaMemcpy((void *)h_arr, (void *)d_arr, sizeof(float) * 128, cudaMemcpyDeviceToHost); // do stuff for (int i = 0; i < 128; i++) { printf("h_arr[%d] = %f\n", i, h_arr[i]); } /* kernel that uses shared memory */ // launch kernel, pass in a pointer to data in global memory use_shared_memory_GPU<<<1, 128>>>(d_arr); // copy the modified data back to host, overwriting content of h_arr cudaMemcpy((void *)h_arr, (void *)d_arr, sizeof(float) * 128, cudaMemcpyDeviceToHost); // do stuff for (int i = 0; i < 128; i++) { printf("h_arr[%d] = %f\n", i, h_arr[i]); } return 0; }
629
extern "C" #define BLOCK_WIDTH 16 #define BLOCK_HEIGHT 16 __global__ void filter(int *Input_Image, int *Output_Image, int Image_Width, int Image_Height) { const int tx_l = threadIdx.x; // --- Local thread x index const int ty_l = threadIdx.y; // --- Local thread y index const int tx_g = blockIdx.x * blockDim.x + tx_l; // --- Global thread x index const int ty_g = blockIdx.y * blockDim.y + ty_l; // --- Global thread y index __shared__ int smem[BLOCK_WIDTH+2][BLOCK_HEIGHT+2]; // --- Fill the shared memory border with zeros if (tx_l == 0) smem[tx_l] [ty_l+1] = 0; // --- left border else if (tx_l == BLOCK_WIDTH-1) smem[tx_l+2][ty_l+1] = 0; // --- right border if (ty_l == 0) { smem[tx_l+1][ty_l] = 0; // --- upper border if (tx_l == 0) smem[tx_l] [ty_l] = 0; // --- top-left corner else if (tx_l == BLOCK_WIDTH-1) smem[tx_l+2][ty_l] = 0; // --- top-right corner } else if (ty_l == BLOCK_HEIGHT-1) {smem[tx_l+1][ty_l+2] = 0; // --- bottom border if (tx_l == 0) smem[tx_l] [ty_l+2] = 0; // --- bottom-left corder else if (tx_l == BLOCK_WIDTH-1) smem[tx_l+2][ty_l+2] = 0; // --- bottom-right corner } // --- Fill shared memory smem[tx_l+1][ty_l+1] = Input_Image[ty_g*Image_Width + tx_g]; // --- center if ((tx_l == 0)&&((tx_g > 0))) smem[tx_l] [ty_l+1] = Input_Image[ty_g*Image_Width + tx_g-1]; // --- left border else if ((tx_l == BLOCK_WIDTH-1)&&(tx_g < Image_Width - 1)) smem[tx_l+2][ty_l+1] = Input_Image[ty_g*Image_Width + tx_g+1]; // --- right border if ((ty_l == 0)&&(ty_g > 0)) { smem[tx_l+1][ty_l] = Input_Image[(ty_g-1)*Image_Width + tx_g]; // --- upper border if ((tx_l == 0)&&((tx_g > 0))) smem[tx_l] [ty_l] = Input_Image[(ty_g-1)*Image_Width + tx_g-1]; // --- top-left corner else if ((tx_l == BLOCK_WIDTH-1)&&(tx_g < Image_Width - 1)) smem[tx_l+2][ty_l] = Input_Image[(ty_g-1)*Image_Width + tx_g+1]; // --- top-right corner } else if ((ty_l == BLOCK_HEIGHT-1)&&(ty_g < Image_Height - 1)) { smem[tx_l+1][ty_l+2] = Input_Image[(ty_g+1)*Image_Width + tx_g]; // --- bottom border if ((tx_l == 0)&&((tx_g > 0))) smem[tx_l] [ty_l+2] = Input_Image[(ty_g-1)*Image_Width + tx_g-1]; // --- bottom-left corder else if ((tx_l == BLOCK_WIDTH-1)&&(tx_g < Image_Width - 1)) smem[tx_l+2][ty_l+2] = Input_Image[(ty_g+1)*Image_Width + tx_g+1]; // --- bottom-right corner } __syncthreads(); // --- Pull the 3x3 window in a local array long v[9] = { smem[tx_l][ty_l], smem[tx_l+1][ty_l], smem[tx_l+2][ty_l], smem[tx_l][ty_l+1], smem[tx_l+1][ty_l+1], smem[tx_l+2][ty_l+1], smem[tx_l][ty_l+2], smem[tx_l+1][ty_l+2], smem[tx_l+2][ty_l+2] }; // --- Bubble-sort for (int i = 0; i < 5; i++) { for (int j = i + 1; j < 9; j++) { if (v[i] > v[j]) { // swap? int tmp = v[i]; v[i] = v[j]; v[j] = tmp; } } } // --- Pick the middle one Output_Image[ty_g*Image_Width + tx_g] = v[4]; }
630
#include <stdio.h> #include <stdlib.h> #include <sys/time.h> #include <assert.h> #include <math.h> /* Returns the current time in miliseconds. */ double getMilitime(){ struct timeval ret; gettimeofday(&ret, NULL); return ((ret.tv_sec ) * 1000000u + ret.tv_usec) / 1.e6; } #define TYPE double #define err(format, ...) do { fprintf(stderr, format, ##__VA_ARGS__); exit(1); } while (0) #ifdef __CUDACC__ inline void checkCuda(cudaError_t e) { if (e != cudaSuccess) { // cudaGetErrorString() isn't always very helpful. Look up the error // number in the cudaError enum in driver_types.h in the CUDA includes // directory for a better explanation. err("CUDA Error %d: %s\n", e, cudaGetErrorString(e)); } } inline void checkLastCudaError() { checkCuda(cudaGetLastError()); } #endif __global__ void ReductionKernel(TYPE* ArrDevice, TYPE* ArrTempDevice, long ArrSize){ extern __shared__ TYPE sdata[]; unsigned long objectId = blockDim.x * blockIdx.x + threadIdx.x; unsigned long tid = threadIdx.x; sdata[tid] = 0; if (objectId < ArrSize) { sdata[tid] = ArrDevice[objectId]; __syncthreads(); // do reduction in shared mem for(unsigned long s=1; s < blockDim.x; s *= 2) { if (tid % (2*s) == 0) { sdata[tid] += sdata[tid + s]; } __syncthreads(); } // write result for this block to global mem if (tid == 0) ArrTempDevice[blockIdx.x] = sdata[0]; } } __global__ void PrintKernel(TYPE* ArrDevice){ printf("ArrSumDevice: %f \n",ArrDevice[0]); } int main(int argc, char* argv[]){ if(argc<2){ printf("Input Error\n"); return 1; } long ArrSize = atoi(argv[1]); TYPE *ArrDevice, *ArrTempDevice,*ArrHost=NULL; //Allocate in host ArrHost = (TYPE*) malloc(ArrSize*sizeof(TYPE)); assert(ArrHost != NULL) ; //Fill in host long count; for(count=0;count<ArrSize;count++){ ArrHost[count] = 1; } //Allocate in device checkCuda(cudaMalloc(&ArrDevice, ArrSize*sizeof(TYPE))); checkCuda(cudaMalloc(&ArrTempDevice, ArrSize*sizeof(TYPE))); //Fill in device cudaMemcpy(ArrDevice, ArrHost, ArrSize*sizeof(TYPE), cudaMemcpyHostToDevice); PrintKernel<<< 1, 1 >>>(ArrDevice); cudaDeviceSynchronize(); checkLastCudaError(); //Call kernel long ArrTempFlag=0; unsigned long numThreadsPerClusterBlock = 512; unsigned long numClusterBlocks = (ArrSize + numThreadsPerClusterBlock - 1) / numThreadsPerClusterBlock; long CurrArrSize = ArrSize; printf("start timing\n"); double start_time = getMilitime(); while(CurrArrSize>1){ if(numThreadsPerClusterBlock>CurrArrSize) numThreadsPerClusterBlock = pow(2.0,ceil(log2((double)CurrArrSize))); numClusterBlocks = (CurrArrSize + numThreadsPerClusterBlock - 1) / numThreadsPerClusterBlock; ////printf("numThreadsPerClusterBlock: %d \n",numThreadsPerClusterBlock); ////printf("numClusterBlocks: %d \n",numClusterBlocks); if((ArrTempFlag) %2 == 0) ReductionKernel<<< numClusterBlocks, numThreadsPerClusterBlock, numThreadsPerClusterBlock * sizeof(TYPE)>>> (ArrDevice, ArrTempDevice,CurrArrSize); else ReductionKernel<<< numClusterBlocks, numThreadsPerClusterBlock, numThreadsPerClusterBlock * sizeof(TYPE)>>> (ArrTempDevice, ArrDevice,CurrArrSize); cudaDeviceSynchronize(); checkLastCudaError(); //PrintKernel<<< 1, 1 >>>(ArrDevice); //cudaDeviceSynchronize(); checkLastCudaError(); CurrArrSize = numClusterBlocks; ArrTempFlag++; } printf("elapsed time: %f sec\n", getMilitime()-start_time); //Copy result to host (maybe prlong in the device) if((ArrTempFlag) %2 == 0){ PrintKernel<<< 1, 1 >>>(ArrDevice); cudaDeviceSynchronize(); checkLastCudaError(); }else{ PrintKernel<<< 1, 1 >>>(ArrTempDevice); cudaDeviceSynchronize(); checkLastCudaError(); } //free device data checkCuda(cudaFree(ArrDevice)); checkCuda(cudaFree(ArrTempDevice)); free(ArrHost); return(0); }
631
#include <stdlib.h> #include <stdio.h> #include <string.h> #include <math.h> #include <cuda.h> #include <cuda_runtime.h> #include <cuda.h> #include <device_launch_parameters.h> #define LIST_SIZE 100000 extern "C" __device__ long long instCountList[LIST_SIZE]; void bambooLogRecordOff(){ } void bambooLogKernelBegin(int i) { } void bambooLogKernelEnd() { #ifdef KERNELTRACE cudaDeviceSynchronize(); #endif }
632
#include "includes.h" __global__ void swan_fast_fill( uint4 *ptr, int len ) { int idx = threadIdx.x + blockDim.x * blockIdx.x; if( idx<len) { ptr[idx] = make_uint4( 0,0,0,0 ); } }
633
/***************************************************************************//** * \file LHS2.cu * \author Christopher Minar (minarc@oregonstate.edu) * \brief kernels to generate the left hand side for the poission solve */ #include "LHS2.h" namespace kernels { /* * calculates the boundary terms for the left hand side matrix for the poisson solve * param row array storing the row indices for the sparse LHS matrix * param col array storing the column indices for the sparse LHS matrix * param val array storing the values for the sparse LHS matrix * param dx distance between nodes in the x direction (measured between node sides, where u velocites are stored) * param dy distance between nodes in the y direction (measured between node top/bot, where v velocites are stored) * param nx number of cells in x direction * param ny number of cells in y direction * param dt change in time */ __global__ void LHS2_BC(int *row, int *col, double *val, double *dx, double *dy, int nx, int ny, double dt) { int ip = threadIdx.x + blockDim.x * blockIdx.x; if (ip >= nx*ny) return; int I = ip % nx, J = ip / nx; if (I != 0 && I != nx-1 && J != 0 && J != ny-1) return; int numE = 0; if (J == 0) { numE = I*4; if (I!=0) numE-=1; } else if (J == ny-1) { numE = nx*4-2 + (J-1)*(nx*5-2) + I*4; if (I!=0) numE-=1; } else { numE = nx*4-2 + (J-1)*(nx*5 - 2) + I*5; if (I != 0) numE -= 1; } double temp = 0; //EAST //if not on the east wall and east is outside the body, add east term if (I != nx-1)//not at east boundry { row[numE] = ip; col[numE] = ip + 1; val[numE] = -dt/(dx[I]*(dx[I]+dx[I+1])*0.5); numE++; temp += dt/(dx[I]*(dx[I]+dx[I+1])*0.5); } //WEST //if not on west wall and west is outside the body, add west term if (I != 0)//not at west boundary { row[numE] = ip; col[numE] = ip - 1; val[numE] = -dt/(dx[I]*(dx[I]+dx[I-1])*0.5); temp += dt/(dx[I]*(dx[I]+dx[I-1])*0.5); numE++; } //NORTH //if not on north wall and north is outside the body, add north term if (J != ny-1)//not at north boundry { row[numE] = ip; col[numE] = ip + nx; val[numE] = -dt/(dy[J]*(dy[J]+dy[J+1])*0.5); temp += dt/(dy[J]*(dy[J]+dy[J+1])*0.5); numE++; } //SOUTH //if not on south wall and south is outside the body, add south term if (J != 0)//not at south boundry { row[numE] = ip; col[numE] = ip - nx; val[numE] = -dt/(dy[J]*(dy[J]+dy[J-1])*0.5); temp += dt/(dy[J]*(dy[J]+dy[J-1])*0.5); numE++; } //MID row[numE] = ip; col[numE] = ip; val[numE] = temp; //do some jank so the solver works, although this modifies the matricies it doesn't really change the results if (ip == 0) val[numE] += val[numE]; //val[numE] = 0; //val[numE] *= val[numE]; } /* * calculates the middle terms for the left hand side matrix for the poisson solve * param row array storing the row indices for the sparse LHS matrix * param col array storing the column indices for the sparse LHS matrix * param val array storing the values for the sparse LHS matrix * param dx distance between nodes in the x direction (measured between node sides, where u velocites are stored) * param dy distance between nodes in the y direction (measured between node top/bot, where v velocites are stored) * param nx number of cells in x direction * param ny number of cells in y direction * param dt change in time */ __global__ void LHS2_mid_nobody(int *row, int *col, double *val, double *dx, double *dy, int nx, int ny, double dt) { int ip = threadIdx.x + blockDim.x * blockIdx.x; if (ip >= nx*ny) return; int I = ip % nx, J = ip / nx; if (I == 0 || I == nx-1 || J == 0 || J == ny-1) return; int numE= nx*4-2 + (J-1)*(nx*5 - 2) + I*5 - 1; double temp = 0; //EAST row[numE] = ip; col[numE] = ip + 1; val[numE] = -dt/(dx[I]*(dx[I]+dx[I+1])*0.5); numE++; temp += dt/(dx[I]*(dx[I]+dx[I+1])*0.5); //WEST row[numE] = ip; col[numE] = ip - 1; val[numE] = -dt/(dx[I]*(dx[I]+dx[I-1])*0.5); temp += dt/(dx[I]*(dx[I]+dx[I-1])*0.5); numE++; //NORTH row[numE] = ip; col[numE] = ip + nx; val[numE] = -dt/(dy[J]*(dy[J]+dy[J+1])*0.5); temp += dt/(dy[J]*(dy[J]+dy[J+1])*0.5); numE++; //SOUTH row[numE] = ip; col[numE] = ip - nx; val[numE] = -dt/(dy[J]*(dy[J]+dy[J-1])*0.5); temp += dt/(dy[J]*(dy[J]+dy[J-1])*0.5); numE++; //MID row[numE] = ip; col[numE] = ip; val[numE] = temp; //do some jank so the solver works, although this modifies the matricies it doesn't really change the results if(row[numE]==col[numE] && col[numE]==(ny/2)*nx+nx/2) { val[numE] += val[numE]; } } }
634
//Author: Xinrea //Date: 2018/7/5 //Basic Sample of using CUDA #include <cuda_runtime.h> #include <device_launch_parameters.h> #include <cstdio> cudaError_t addData(int size,int a[],int b[],int c[]); __global__ void addKernel(int dev[],int size){ int i = threadIdx.x; dev[2*size+i] = dev[i]+dev[size+i]; } int main(int argc, char const *argv[]) { int size = 3; int a[3] = {1,2,3}; int b[3] = {4,5,6}; int c[3] = {0}; cudaError_t res; res = addData(size,a,b,c); if(res != cudaSuccess){ fprintf(stderr,"addData failed\n"); return -1; } printf("%d %d %d\n",c[0],c[1],c[2]); res = cudaDeviceReset(); if(res != cudaSuccess){ fprintf(stderr,"cudaDeviceReset failed\n"); return -1; } return 0; } cudaError_t addData(int size,int a[],int b[],int c[]){ cudaError_t res; int *dev; do{ res = cudaSetDevice(0); if (res != cudaSuccess) { fprintf(stderr, "cudaSetDevice failed\n"); break; } res = cudaMalloc((void**)&dev,3*size*sizeof(int)); if(res != cudaSuccess){ fprintf(stderr,"cudaMalloc failed\n"); break; } res = cudaMemcpy(dev,a,size*sizeof(int),cudaMemcpyHostToDevice); if(res != cudaSuccess){ fprintf(stderr,"cudaMemcpy a failed\n"); break; } res = cudaMemcpy(dev+size,b,size*sizeof(int),cudaMemcpyHostToDevice); if(res != cudaSuccess){ fprintf(stderr,"cudaMemcpy b failed\n"); break; } addKernel<<<1,size>>>(dev,size); res = cudaGetLastError(); if(res != cudaSuccess){ fprintf(stderr,"addData failed\n"); break; } cudaDeviceSynchronize(); res = cudaMemcpy(c,dev+2*size,size*sizeof(int),cudaMemcpyDeviceToHost); if(res != cudaSuccess){ fprintf(stderr,"cudaMemcpy c failed\n"); break; } } while(0); cudaFree(dev); return res; }
635
// addNext.cu // Second program from Dr Dobbs tutorial. http://drdobbs.com/parallel/207402986 #include <cuda.h> #include <stdio.h> #include <assert.h> // Host kernel = increment each element by 1 void incOnHost(float *a, int N) { int i; for (i=1; i<N; i++) { a[i] = a[i] + a[i-1]; } } // Device Kernel = increment each element by 1 __global__ void incOnDevice(float *a, int N) { int idx = (blockIdx.x * blockDim.x) + threadIdx.x; if ( (idx<N) && (idx>1) ) a[idx] = a[idx] + a[idx-1]; } // Main thread int main(void) { float *a_h, *b_h; float *a_d; int i, N=48; size_t size = N*sizeof(float); a_h = (float *)malloc(size); b_h = (float *)malloc(size); cudaMalloc((void **) &a_d, size); for (i=0; i<N; i++) a_h[i] = (float)i; cudaMemcpy(a_d, a_h, size, cudaMemcpyHostToDevice); // do calculation on host incOnHost(a_h, N); // do calculation on device int blockSize = 4; int nBlocks = N/blockSize + (N%blockSize == 0?0:1); incOnDevice <<< nBlocks, blockSize >>> (a_d, N); // Retrieve results and compare cudaMemcpy(b_h,a_d,size,cudaMemcpyDeviceToHost); for (i=0; i<N; i++) { // printf("value from host %d = %1.1f\n",i,a_h[i]); printf("value from device %d = %1.1f\n",i,b_h[i]); // assert(a_h[i] == b_h[i]); } free(a_h); free(b_h); cudaFree(a_d); }
636
extern "C" __global__ void kernel_dummy(float *ptr) { ptr[blockIdx.x] = 0; }
637
/* * Compile and run: nvcc -arch=sm_20 kGrid.cu -run */ #include <stdio.h> /*__global__ void kGrid(int n, int *k) { int l = blockIdx.x * blockDim.x + threadIdx.x; if(l < n) { for(int* i = 0; i < k; i++) { for(int* j = 0; j < k; j++) { printf("%i", j) } printf("\n"); } } } int main(void) { int N = 1<<20; int *k, *d_k; k = (int*)malloc(N*sizeof(int)); cudaMalloc(&d_k, N*sizeof(int)); for(int i = 0; i < N; i++) { k[i] = 5; } cudaMemcpy(d_k, k, N*sizeof(int), cudaMemcpyHostToDevice); kGrid<<<(N+255)/256, 256>>>(N, d_k); cudaMemcpy(k, d_k, N*sizeof(int), cudaMemcpyDeviceToHost); cudaFree(d_k); free(k); return 0; }*/ __global__ void print_kernel() { printf("Block %d, thread %d\n", blockIdx.x, threadIdx.x); } int main() { print_kernel<<<10, 10>>>(); cudaDeviceSynchronize(); return 0; }
638
#include<stdio.h> #include<stdlib.h> #define TPB 8 #define N 32 __device__ float distance(float x1, float x2) { return sqrt ((x2-x1)*(x2-x1)); } __global__ void distanceKernel(float *d_out, float *d_in, float ref) { const int i=blockIdx.x*blockDim.x+threadIdx.x; const int j=blockIdx.x; const int k=threadIdx.x; const int l=blockDim.x; const float x=d_in[i]; d_out[i]=distance(x, ref); printf("blockDIM=%d, blockID=%d, threadID=%d, i=%d: the distance between %f to %f is %f. \n", l, j, k, i, ref, x, d_out[i]); //// } void distanceArray(float *out, float *in, float ref, int len) { float *d_in=0; float *d_out=0; //// cudaMalloc(&d_in, len*sizeof(float)); cudaMalloc(&d_out, len*sizeof(float)); cudaMemcpy(d_in, in, len*sizeof(float), cudaMemcpyHostToDevice); distanceKernel<<<len/TPB, TPB>>>(d_out, d_in, ref); cudaMemcpy(out, d_out, len*sizeof(float), cudaMemcpyDeviceToHost); cudaFree(d_in); cudaFree(d_out); } float scale(int i, int n) { return ((float) i)/(n-1); } int main() { const float ref=0.5f; float *in=(float*) calloc(N,sizeof(float)); float *out=(float*) calloc(N, sizeof(float)); for(int i=0; i<N; ++i) { in[i]=scale(i,N); // } distanceArray(out, in, ref, N); printf("______________________________ \n"); for(int j=0; j<N; ++j) { printf("The distance, printed from the host, between %f to %f is %f. \n", ref, in[j], out[j]); } free(in); free(out); return 0; }
639
/* 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 tmp_1 = (+1.5410E-43f * var_1 * +1.0973E35f + var_2 - (+0.0f - +1.9327E35f)); comp = tmp_1 - (var_3 * +1.7163E36f / var_4 + var_5 + -1.8967E-42f + var_6); float tmp_2 = (-1.7459E35f - (var_7 * +1.1576E-37f)); comp = tmp_2 / expf(+1.5630E-37f / cosf(+1.3297E-36f - (var_8 - -1.7565E-2f + var_9 - ceilf(var_10 * (var_11 / var_12 + atanf(-1.5981E-37f + +1.3857E-1f - var_13 / atanf(-1.3361E-44f / -1.2044E-44f))))))); if (comp <= (var_14 - (var_15 * +1.7918E-24f / var_16 - (-0.0f * var_17)))) { comp = atanf(atan2f((var_18 - (+1.9056E1f + var_19)), +1.0569E-37f - atan2f(+1.3944E-37f, (-1.0124E35f / var_20 + var_21)))); } if (comp < (var_22 / fmodf((var_23 / var_24), +1.0852E-35f))) { comp += +1.0219E-35f * +0.0f; comp += sqrtf(+1.7431E-36f); comp = (var_25 - (+1.6513E35f / +1.1353E15f * var_26 + +1.6437E35f)); comp += var_27 * (var_28 / (var_29 * +1.2062E-42f)); } 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]); 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); cudaDeviceSynchronize(); return 0; }
640
#include "includes.h" __device__ inline unsigned int RM_Index(unsigned int row, unsigned int col, unsigned int width) { return (row * width + col); } __global__ void MultinomialNBLearnKernel(float *feature_probs, float *class_priors, const float *d_row_sums, unsigned int n_samples_, unsigned int n_classes_, unsigned int n_features_) { /* Each thread will take one term */ unsigned int tidx = threadIdx.x; unsigned int feat_col = tidx + (blockIdx.x * blockDim.x); unsigned int i = 0; if (feat_col < n_features_) { /* End condition check */ /* For each label */ for (i = 0; i < n_classes_; ++i) { // Alpha is default set to 1.0 for laplacian smoothing feature_probs[RM_Index(i, feat_col, n_features_)] = log((feature_probs[RM_Index(i, feat_col, n_features_)] + 1.0) / (d_row_sums[i] - (n_features_ * 1.0))); if (feat_col == 0) { class_priors[i] = log(class_priors[i] / (float)n_samples_); } } } }
641
#include "includes.h" const int Nthreads = 1024, NrankMax = 3, nt0max = 71, NchanMax = 1024; ////////////////////////////////////////////////////////////////////////////////////////// ////////////////////////////////////////////////////////////////////////////////////////// ////////////////////////////////////////////////////////////////////////////////////////// ////////////////////////////////////////////////////////////////////////////////////////// ////////////////////////////////////////////////////////////////////////////////////////// __global__ void blankdWU(const double *Params, const double *dWU, const int *iC, const int *iW, double *dWUblank){ int nt0, tidx, tidy, bid, Nchan, NchanNear, iChan; nt0 = (int) Params[4]; Nchan = (int) Params[9]; NchanNear = (int) Params[10]; tidx = threadIdx.x; tidy = threadIdx.y; bid = blockIdx.x; while (tidy<NchanNear){ iChan = iC[tidy+ NchanNear * iW[bid]]; dWUblank[tidx + nt0*iChan + bid * nt0 * Nchan] = dWU[tidx + nt0*iChan + bid * nt0 * Nchan]; tidy+=blockDim.y; } }
642
// // kernel routine // __global__ void VecAdd_kernel(const float* A, const float* B, float* C, int N) /* Naive kernel */ { // Uncomment line below and define global index form block and thread indexes int i = threadIdx.x + blockDim.x * blockIdx.x; if(i < N){ C[i] = A[i] + B[i]; } }
643
#include <cuda_runtime.h> #include <device_launch_parameters.h> #include <stdio.h> __global__ void vecAddKernel(int *a, int *b, int *c){ int i = threadIdx.x; c[i] = a[i] + b[i]; } int main(){ int v1[5] = {1,2,3,4,5}; int v2[5] = {6,7,8,9,10}; int *cuda_a, *cuda_b, *cuda_c; cudaMalloc((void**) &cuda_a, 5*sizeof(int)); cudaMalloc((void**) &cuda_b, 5*sizeof(int)); cudaMalloc((void**) &cuda_c, 5*sizeof(int)); cudaMemcpy(cuda_a, v1, 5*sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(cuda_b, v2, 5*sizeof(int), cudaMemcpyHostToDevice); vecAddKernel <<<1,5>>> (cuda_a, cuda_b, cuda_c); cudaDeviceSynchronize(); int v3[5]; cudaMemcpy(v3, cuda_c, 5*sizeof(int), cudaMemcpyDeviceToHost); cudaFree(cuda_a); cudaFree(cuda_b); cudaFree(cuda_c); for (int i=0; i<5; i++){ printf("%d: %d + %d = %d\n", i, v1[i], v2[i], v3[i]); } }
644
#include <iostream> #include <fstream> #include <cuda_runtime.h> using namespace std; //nvcc cudaProperty.cu -o cudaProp bool InitCUDA() { int count; cudaGetDeviceCount(&count); if(count == 0) { cout << "There is no device."<< endl; return false; } int i; for(i = 0; i < count; i++) { cudaDeviceProp prop; if(cudaGetDeviceProperties(&prop, i) == cudaSuccess) { if(prop.major >= 1) { break; } } } if(i == count) { cout << "There is no device supporting CUDA 1.x." << endl; return false; } cudaSetDevice(i); return true; } int main(int argc, char** argv){ cudaDeviceProp prop; if(cudaGetDeviceProperties(&prop, 0) == cudaSuccess) { if(prop.major >= 1) { cout << "Name : " << prop.name << endl; cout << "Total Global Mem : " << prop.totalGlobalMem << endl; cout << "Shared Mem per block : " << prop.sharedMemPerBlock<< endl; cout << "Max Thread per block : " << prop.maxThreadsPerBlock<< endl; cout << "total Const Mem : " << prop.totalConstMem<< endl; cout << "multiProcessor : " << prop.multiProcessorCount<< endl; cout << "Warp Size : " << prop.warpSize<< endl; cout << "ClockRate : " << prop.clockRate<< endl; cout << "Major : " << prop.major<< endl; cout << "Minor : " << prop.minor<< endl; } } return 0; }
645
/* * singleGpuSpectrometer * * Version 2.0, April 12 2010 * * This program was written by Hirofumi Kondo at the Supercomputing Engineering Laboratory, * Graduate School of Information Science and Technology, Osaka University, Japan. * * Copyright 2010 Supercomputing Engineering Laboratory, Graduate School of Information * Science and Technology, Osaka University, Japan * * * Compile : * nvcc -o singleGpuSpectrometer singleGpuSpectrometer.cu -I /usr/local/cuda/NVIDIA_GPU_Computing_SDK/common/inc * /usr/local/cuda/NVIDIA_GPU_Computing_SDK/C/lib/libcutil.a * -L /usr/local/cuda/lib -l cufft * * Usage : ./singleGpuSpectrometer [options] * -length : signal length of this spectrometer handle in M-points * -boxcar : length of boxcar for smoothing * -threshold : value of threshold * -max_detect_point : value of maximum detected points over threshold in each boxcar * -output_file : filename of output file * * Output file format : * The file format is binary format. * The output file records all spikes whose power exceed (boxcar_mean) * (threashold). * The file contains 3 data * 1) index of signal * 2) the power of signal * 3) mean power of boxcar which the signal is in * * Special Instruction * 1) Memory capacity * The memory capacity that this GPU spectrometer requires is changed by the signal length. * If you want to analyze 128M-points signal, GPU has to have 4GB VRAM. * The maximum length that 1GB VRAM GPU can handle is 32M-points. * * 2) CUDA * We recommend that you use CUDA 2.3 and CUFFT 2.3. * This is not necessary condition. * But the execution time is wrong if you use CUDA 2.2 and CUFFT 2.2. */ // includes, system #include <stdlib.h> #include <stdio.h> #include <string.h> #include <math.h> #include <time.h> #include <cuda.h> //#include <cutil_inline.h> //#include <cufft.h> //#include "bench_fft.h" //#include "output_greg.c" /* * Prototype declaration */ /* Modified by Jayanth Chennamangalam on 2013.01.31 --> */ /*void do_analyze_on_gpu();*/ void do_analyze_on_gpu(int signalLength); /* <-- Modified by Jayanth Chennamangalam on 2013.01.31 */ /* * Program main */ int main(int argc, char** argv){ int i; int signalLength = 1024 * 1024; for (i=1;i<argc;i++) { if (!strcmp(argv[i], "-length")) { signalLength = atoi(argv[++i]) * 1024 * 1024; } /*else if (!strcmp(argv[i], "-boxcar")){ boxcar = atoi(argv[++i]); } else if (!strcmp(argv[i], "-threshold")){ threshold = atoi(argv[++i]); } else if (!strcmp(argv[i], "-max_detect_point")){ maximumDetectPointInBoxcar = atoi(argv[++i]); } else if (!strcmp(argv[i], "-output_file")){ strncpy(outputFileName, argv[++i], FILENAME_BUFSIZE); if(outputFileName[FILENAME_BUFSIZE-1]!='\0'){ fprintf(stderr,"Error : Too long output file name. maximum length = %d\n", FILENAME_BUFSIZE-1); exit(-1); } } else { fprintf(stderr,"Error : wrong argument\n"); }*/ } // Analyze signal on GPU printf("Calling GPU\n"); do_analyze_on_gpu(signalLength); return 0; }
646
#include "includes.h" __global__ void set_unavailable(bool *available, int n_rows, const int *idx, int n_selected) { int tid = threadIdx.x + blockIdx.x * blockDim.x; if (tid < n_selected) { available[idx[tid]] = false; } }
647
 //#include "cuda_runtime.h" //#include "device_launch_parameters.h" //#include <stdio.h> #include <stdio.h> #include <cuda_runtime.h> // device kernel __global__ void helloWorldDevice() { printf("Hello world from device %d!\n", threadIdx.x); } int main() { printf("Hello world from host!\n"); // run kernel in 3 instances helloWorldDevice <<<1, 3>>> (); cudaDeviceSynchronize(); }
648
#include "includes.h" __global__ void lga_filtering_forward (const int n, const float *bottom_data, const float *filters, const int height, const int width, const int channel, const int radius, float *top_data){ int index = blockIdx.x * blockDim.x + threadIdx.x; // printf("OK\n"); // printf("%d, %.2f, %.2f\n", index, bottom_data[index], top_data[index]); if (index >= n) { return; } // top_data[index]=1.0; // assert(0); int step = height * width; int wsize = 2 * radius + 1; // int fsize=wsize*wsize*3; int fbase = index / (step * channel) * (step * wsize * wsize * 3) + index % step; int row = index % step / width; int col = index % width; int depth = index / step % channel; for (int d = -1; d <= 1; d++) { for (int r = -radius; r <= radius; r++) { for (int c = -radius; c <= radius; c++) { int rr = r + row; int cc = c + col; int dd = d + depth; int shift = 0; if (rr >= 0 && cc >= 0 && dd >= 0 && rr < height && cc < width && dd < channel) shift = r * width + c + d * step; int location = (d + 1) * (wsize * wsize) + (r + radius) * wsize + c + radius; top_data[index] += bottom_data[index + shift] * filters[fbase + location * step]; } } } // top_data[index]=1.0; // printf("%d, %d, %d, %.2f, %.2f\n", index, row, col, bottom_data[index], top_data[index]); }
649
char *title = "gauss filtering"; char *description = "gauss filtering"; #include <iostream> #include <cstdio> #include <cstdlib> #include <assert.h> #include <time.h> #include <cuda.h> #include <cuda_runtime.h> #include <math.h> /* exp */ #ifndef max #define max( a, b ) ( ((a) > (b)) ? (a) : (b) ) #endif #ifndef min #define min( a, b ) ( ((a) < (b)) ? (a) : (b) ) #endif #ifndef uint8_t typedef unsigned char uint8_t; #endif #ifndef uint16_t typedef unsigned short uint16_t; #endif #ifndef uint32_t typedef unsigned int uint32_t; #endif // http://en.wikipedia.org/wiki/BMP_file_format #define BMPMAGIC 0x00 #define BMPFILESIZE 0x02 #define BMPOFFSET 0x0A #define BMPDIBSISE 0x0E #define BMPWIDTH 0x12 #define BMPHEIGHT 0x16 #define BMPBITSPERPIXEL 0x1C typedef struct { uint8_t magic[2]; /* the magic number used to identify the BMP file: 0x42 0x4D (Hex code points for B and M). The following entries are possible: BM - Windows 3.1x, 95, NT, ... etc BA - OS/2 Bitmap Array CI - OS/2 Color Icon CP - OS/2 Color Pointer IC - OS/2 Icon PT - OS/2 Pointer. */ uint32_t filesz; /* the size of the BMP file in bytes */ uint16_t creator1; /* reserved. */ uint16_t creator2; /* reserved. */ uint32_t offset; /* the offset, i.e. starting address, of the byte where the bitmap data can be found. */ } bmp_header_t; typedef struct { uint32_t header_sz; /* the size of this header (40 bytes) */ uint32_t width; /* the bitmap width in pixels */ uint32_t height; /* the bitmap height in pixels */ uint16_t nplanes; /* the number of color planes being used. Must be set to 1. */ uint16_t depth; /* the number of bits per pixel, which is the color depth of the image. Typical values are 1, 4, 8, 16, 24 and 32. */ uint32_t compress_type; /* the compression method being used. See also bmp_compression_method_t. */ uint32_t bmp_bytesz; /* the image size. This is the size of the raw bitmap data (see below), and should not be confused with the file size. */ uint32_t hres; /* the horizontal resolution of the image. (pixel per meter) */ uint32_t vres; /* the vertical resolution of the image. (pixel per meter) */ uint32_t ncolors; /* the number of colors in the color palette, or 0 to default to 2<sup><i>n</i></sup>. */ uint32_t nimpcolors; /* the number of important colors used, or 0 when every color is important; generally ignored. */ } bmp_dib_v3_header_t; __global__ void global_defaultgaussfilter( unsigned int *input, unsigned int *output, double *matrix, int Nh, int width,int height, int ddxx, int ddyy) { assert(ddxx*ddyy==0); if(ddxx>0) for (int id = blockDim.x*blockIdx.x + threadIdx.x; id < width*height; id += blockDim.x*gridDim.x) { int x = id % width; int y = id / width; double s1 = 0; double s2 = 0; for (int dx = -Nh; dx <= Nh; dx++) { if (x + dx < 0) continue; if (x + dx >= width) continue; int value = (double)(unsigned int)input[id + dx]; s1 += value*matrix[Nh + dx]; s2 += matrix[Nh + dx]; } output[id] = (unsigned int)(s1 / s2); } if(ddyy>0) for (int id = blockDim.x*blockIdx.x + threadIdx.x; id < width*height; id += blockDim.x*gridDim.x) { int x = id % width; int y = id / width; double s1 = 0; double s2 = 0; for (int dy = -Nh; dy <= Nh; dy++) { if (y + dy < 0) continue; if (y + dy >= height) continue; int value = (double)(unsigned int)input[id + width*dy]; s1 += value*matrix[Nh + dy]; s2 += matrix[Nh + dy]; } output[id] = (unsigned int)(s1 / s2); } } __host__ void host_defaultgaussfilter( int gridSize, int blockSize, unsigned int *r, unsigned int *g, unsigned int *b, double *matrix, int Nh, int width, int height) { unsigned int *channel[3] = { r,g,b }; // channel - массив данных // sigma - коэффициент размытия // width - ширина // height - высота cudaError_t err; unsigned int *device_a; unsigned int *device_b; double *device_matrix; err = cudaMalloc((void**)&device_a, width*height*sizeof(unsigned int)); err = cudaMalloc((void**)&device_b, width*height*sizeof(unsigned int)); err = cudaMalloc((void**)&device_matrix, (2 * Nh + 1) * sizeof(double)); cudaMemcpy(device_matrix, matrix, (2 * Nh + 1)*sizeof(double), cudaMemcpyHostToDevice); int blocks = (gridSize > 0)? gridSize : min(15, (int)sqrt(width*height)); int threads = (blockSize > 0)? blockSize : min(15, (int)sqrt(width*height)); for(int j=0; j<3; j++) { cudaMemcpy(device_a, channel[j], width*height*sizeof(unsigned int), cudaMemcpyHostToDevice); global_defaultgaussfilter <<< blocks, threads >>>(device_a, device_b, device_matrix, Nh, width, height, 0, 1); global_defaultgaussfilter <<< blocks, threads >>>(device_b, device_a, device_matrix, Nh, width, height, 1, 0); cudaMemcpy(channel[j], device_a, width*height*sizeof(unsigned int), cudaMemcpyDeviceToHost); } // Освобождаем память на устройстве cudaFree((void*)device_a); cudaFree((void*)device_b); cudaFree((void*)device_matrix); err = err; } __global__ void global_recursivegaussfilter( unsigned int *input, unsigned int *output, double *bb, int p, int q, int width,int height, int ddxx, int ddyy) { assert(p==1); assert(q==3); assert(ddxx*ddyy==0); // Копируем в локальные переменные для ускорения доступа к значениям double b1 = bb[1]/bb[0]; double b2 = bb[2]/bb[0]; double b3 = bb[3]/bb[0]; double B = 1.0 - (b1+b2+b3); if(ddxx>0) for (int y = blockDim.x*blockIdx.x + threadIdx.x; y < height; y += blockDim.x*gridDim.x) { int id = y*width; if(width>1) output[id]=B*input[id];id++; if(width>2) output[id]=B*input[id]+b1*output[id-1];id++; if(width>3) output[id]=B*input[id]+b1*output[id-1]+b2*output[id-2];id++; for(; id<width; id++) output[id]=B*input[id]+b1*output[id-1]+b2*output[id-2]+b3*output[id-3]; } if(ddxx<0) for (int y = blockDim.x*blockIdx.x + threadIdx.x; y < height; y += blockDim.x*gridDim.x) { int id = y*width+width; id--;if(width>1) output[id]=B*input[id]; id--;if(width>2) output[id]=B*input[id]+b1*output[id+1]; id--;if(width>3) output[id]=B*input[id]+b1*output[id+1]+b2*output[id+2]; for(id--; id>=0; id--) output[id]=B*input[id]+b1*output[id+1]+b2*output[id+2]+b3*output[id+3]; } if(ddyy>0) for (int x = blockDim.x*blockIdx.x + threadIdx.x; x < width; x += blockDim.x*gridDim.x) { int id = x; if(height>1) output[id]=B*input[id];id+=width; if(height>2) output[id]=B*input[id]+b1*output[id-width];id+=width; if(height>3) output[id]=B*input[id]+b1*output[id-width]+b2*output[id-2*width];id+=width; for(int y=3; y<height; y++, id+=width) output[id]=B*input[id]+b1*output[id-width]+b2*output[id-2*width]+b3*output[id-3*width]; } if(ddyy<0) for (int x = blockDim.x*blockIdx.x + threadIdx.x; x < width; x += blockDim.x*gridDim.x){ int id = height*width+x; id-=width;if(height>1) output[id]=B*input[id]; id-=width;if(height>2) output[id]=B*input[id]+b1*output[id+width]; id-=width;if(height>3) output[id]=B*input[id]+b1*output[id+width]+b2*output[id+2*width]; for(id-=width; id>=0; id-=width) output[id]=B*input[id]+b1*output[id+width]+b2*output[id+2*width]+b3*output[id+3*width]; } } __host__ void host_recursivegaussfilter( int gridSize, int blockSize, unsigned int *r, unsigned int *g, unsigned int *b, double *bb, int p, int q, int width, int height) { unsigned int *channel[3] = { r,g,b }; // channel - массив данных // sigma - коэффициенты // width - ширина // height - высота cudaError_t err; unsigned int *device_a; unsigned int *device_b; double *device_bb; err = cudaMalloc((void**)&device_a, width*height*sizeof(unsigned int)); err = cudaMalloc((void**)&device_b, width*height*sizeof(unsigned int)); err = cudaMalloc((void**)&device_bb, (p+q) * sizeof(double)); cudaMemcpy(device_bb, bb, (p+q)*sizeof(double), cudaMemcpyHostToDevice); int blocks = (gridSize > 0)? gridSize : min(15, (int)sqrt(max(width,height))); int threads = (blockSize > 0)? blockSize : min(15, (int)sqrt(max(width,height))); for(int j=0; j<3; j++) { cudaMemcpy(device_a, channel[j], width*height*sizeof(unsigned int), cudaMemcpyHostToDevice); global_recursivegaussfilter <<< blocks, threads >>>(device_a, device_b, device_bb, p, q, width, height, 0, 1); global_recursivegaussfilter <<< blocks, threads >>>(device_b, device_a, device_bb, p, q, width, height, 0, -1); global_recursivegaussfilter <<< blocks, threads >>>(device_a, device_b, device_bb, p, q, width, height, 1, 0); global_recursivegaussfilter <<< blocks, threads >>>(device_b, device_a, device_bb, p, q, width, height, -1, 0); cudaMemcpy(channel[j], device_a, width*height*sizeof(unsigned int), cudaMemcpyDeviceToHost); } // Освобождаем память на устройстве cudaFree((void*)device_a); cudaFree((void*)device_b); cudaFree((void*)device_bb); err = err; } #define RECURSIVE 1 #define DEFAULT 0 int main(int argc, char* argv[]) { int i, j; std::cout << title << std::endl; // Find/set the device. int device_count = 0; cudaGetDeviceCount(&device_count); for (i = 0; i < device_count; ++i) { cudaDeviceProp properties; cudaGetDeviceProperties(&properties, i); std::cout << "Running on GPU " << i << " (" << properties.name << ")" << std::endl; } if (argc < 5){ printf("Usage :\t%s filter sigma inputfilename.bmp ouputfilename.bmp\n", argv[0]); printf("filter :\tgauss recursivegauss\n"); exit(-1); } // Получаем параметры - имена файлов char *filter = argv[1]; double sigma = atof(argv[2]); char *inputFileName = argv[3]; char *outputFileName = argv[4]; int gridSize = (argc>5)?atoi(argv[5]):0; int blockSize = (argc>6)?atoi(argv[6]):0; int mode =(strcmp(filter,"gauss")==0)?DEFAULT:RECURSIVE; printf("Title :\t%s\n", title); printf("Description :\t%s\n", description); printf("Filter :\t%s\n", filter); printf("Sigma :\t%le\n", sigma); printf("Input file name :\t%s\n", inputFileName); printf("Output file name :\t%s\n", outputFileName); FILE *file = fopen(inputFileName, "rb"); if (!file) { fprintf(stderr, "Open file error (%s)\n", inputFileName); fflush(stderr); exit(-1); } fseek(file, 0L, SEEK_END); long size = ftell(file); unsigned char *buffer = (unsigned char *)malloc((size_t)size); fseek(file, 0L, SEEK_SET); fread((void*)buffer, (size_t)1, (size_t)size, file); fclose(file); uint32_t width = *(uint32_t *)&buffer[BMPWIDTH]; uint32_t height = *(uint32_t *)&buffer[BMPHEIGHT]; uint32_t file_size = *(uint32_t *)&buffer[BMPFILESIZE]; uint32_t offset = *(uint32_t *)&buffer[BMPOFFSET]; uint16_t bits_per_pixel = *(uint16_t *)&buffer[BMPBITSPERPIXEL]; uint16_t bytes_per_pixel = ((int)((bits_per_pixel+7)/8)); uint32_t bytes_per_line = ((int)((bits_per_pixel * width+31)/32))*4; // http://en.wikipedia.org/wiki/BMP_file_format printf("BMP image size :\t%ld x %ld\n", width, height); printf("BMP file size :\t%ld\n", file_size); printf("BMP pixels offset :\t%ld\n", offset); printf("BMP bits per pixel :\t%d\n", bits_per_pixel); printf("BMP bytes per pixel :\t%d\n", bytes_per_pixel); printf("BMP bytes per line :\t%d\n", bytes_per_line); uint8_t *pixels =(uint8_t *)&buffer[offset]; uint8_t *pixel; uint32_t x, y; /* выделение памяти под байтовые цветовые каналы */ int count = width*height; unsigned int* channel[3]; for (j = 0; j < 3; j++) channel[j] = (unsigned int*)malloc(count*sizeof(unsigned int)); int pos = 0; for (y = 0; y < height; y++) { for (x = 0; x < width; x++) { pixel = &pixels[y*bytes_per_line+x*bytes_per_pixel]; channel[0][pos] = (unsigned int)pixel[0]; channel[1][pos] = (unsigned int)pixel[1]; channel[2][pos] = (unsigned int)pixel[2]; pos++; } } switch(mode){ case RECURSIVE:{ int p = 1, q = 3; double * bb = (double*)malloc((p+q)*sizeof(double)); //// http://mognovse.ru/qgi-chekan-rostislav-vladimirovich.html //// cтандартные рекомендации по выбору константы q представляют собой: //double qq; //if(sigma>2.5) qq=0.98711*sigma - 0.96330; //else if(sigma>0.5) qq=3.97156 - 4.14554*sqrt(1.0 - 0.26891*sigma); //else qq=3.97156 - 4.14554*sqrt(1.0 - 0.26891*sigma); //double qq2 = qq*qq; //double qq3 = qq2*qq; //bb[0] = 1.5725 + 2.44413*qq + 1.4281*qq2; //bb[1] = 2.44413*qq + 2.85619*qq2 + 1.26661*qq3; //bb[2] = -1.4281*qq2 - 1.26661*qq3; //bb[3] = 0.422205*qq3; // http://habrahabr.ru/post/151157/ double sigma_inv_4; sigma_inv_4 = sigma*sigma; sigma_inv_4 = 1.0/(sigma_inv_4*sigma_inv_4); double coef_A = sigma_inv_4*(sigma*(sigma*(sigma*1.1442707+0.0130625)-0.7500910)+0.2546730); double coef_W = sigma_inv_4*(sigma*(sigma*(sigma*1.3642870+0.0088755)-0.3255340)+0.3016210); double coef_B = sigma_inv_4*(sigma*(sigma*(sigma*1.2397166-0.0001644)-0.6363580)-0.0536068); double z0_abs = exp(coef_A); double z0_real = z0_abs*cos(coef_W); double z0_im = z0_abs*sin(coef_W); double z2 = exp(coef_B); double z0_abs_2 = z0_abs*z0_abs; bb[3] = 1.0/(z2*z0_abs_2); bb[2] = -(2*z0_real+z2)*bb[3]; bb[1] = (z0_abs_2+2*z0_real*z2)*bb[3]; bb[0] = 1.0; clock_t t1, t2; t1 = clock(); host_recursivegaussfilter(gridSize, blockSize, channel[0], channel[1], channel[2], bb, p, q, width, height); t2 = clock(); double seconds = ((double)(t2-t1))/CLOCKS_PER_SEC; printf("Execution time :\t%le\n", seconds); free(bb); } break; default:{ // Nh - полуразмер int Nh = (int)(3 * sigma); double * matrix = (double*)malloc((2 * Nh + 1)*sizeof(double)); for (int i = -Nh; i <= Nh; i++) matrix[Nh + i] = exp(-(double)i*i / (2.0*sigma*sigma)); clock_t t1, t2; t1 = clock(); host_defaultgaussfilter(gridSize, blockSize, channel[0], channel[1], channel[2], matrix, Nh, width, height); t2 = clock(); double seconds = ((double)(t2-t1))/CLOCKS_PER_SEC; printf("Execution time :\t%le\n", seconds); free(matrix); } break; } pos = 0; for (y = 0; y < height; y++) { for (x = 0; x < width; x++) { pixel = &pixels[y*bytes_per_line+x*bytes_per_pixel]; pixel[0] = (uint8_t)channel[0][pos]; pixel[1] = (uint8_t)channel[1][pos]; pixel[2] = (uint8_t)channel[2][pos]; pos++; } } /* высвобождаем массивы */ for (j = 0; j < 3; j++) free(channel[j]); /* выводим результаты */ file = fopen(outputFileName, "wb"); if (!file) { fprintf(stderr, "Open file error (%s)\n", outputFileName); fflush(stderr); exit(-1); } fwrite((void*)buffer, (size_t)1, (size_t)size, file); printf("Output file size :\t%ld\n", size); free(buffer); fclose(file); cudaDeviceReset(); exit(0); }
650
#include "addition.cuh" #include <stdio.h> __global__ void addition(int* a, int* b, int* c){ *c = *a + *b; } void setupCuda(int* &D_A, int* &D_B, int* &D_C, int* &A, int* &B, int* &C){ cudaMalloc((void**)&D_A, sizeof(int)); cudaMalloc((void**)&D_B, sizeof(int)); cudaMalloc((void**)&D_C, sizeof(int)); cudaMemcpy(D_A, A, sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(D_B, B, sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(D_C, C, sizeof(int), cudaMemcpyHostToDevice); } void copyInputToDevice(int* D_A, int* D_B, int* D_C, int* A, int* B, int* C){ cudaMemcpy(D_A, A, sizeof(int), cudaMemcpyHostToDevice); cudaMemcpy(D_B, B, sizeof(int), cudaMemcpyHostToDevice); } void executeKernel(int* D_A, int* D_B, int* D_C, int* A, int* B, int* C){ addition<<<1,1>>>(D_A, D_B, D_C); } void copyOutputToHost(int* D_A, int* D_B, int* D_C, int* A, int* B, int* C){ cudaMemcpy(C, D_C, sizeof(int), cudaMemcpyDeviceToHost); } void cleanupCuda(int* D_A, int* D_B, int* D_C){ cudaFree(D_A); cudaFree(D_B); cudaFree(D_C); }
651
#include "includes.h" __global__ void bcnn_op_cuda_ramp_kernel(int n, float *x, float *y) { int i = (blockIdx.x + blockIdx.y * gridDim.x) * blockDim.x + threadIdx.x; if (i < n) { y[i] = x[i] * (x[i] > 0) + 0.1 * x[i]; } return; }
652
#include "cuda_runtime.h" #include <stdint.h>
653
#include <cuda_runtime.h> int main() { int* a; cudaMalloc(&a, 100); cudaFree(a); return 0; }
654
#include <cuda_runtime.h> #include <device_launch_parameters.h> #include <cstdio> #include <iostream> #include <cstdlib> #define BLOCK_SIZE 64 #define N 64 using namespace std; void displayLastError(const string &msg) { cout << "Last Error (" << msg << "):\t" << cudaGetErrorString(cudaGetLastError()) << endl; } __global__ void minMaxCuda(float *array) { int nTotalThreads = blockDim.x; // Total number of active threads while(nTotalThreads > 1) { int halfPoint = (nTotalThreads >> 1); // divide by two // only the first half of the threads will be active. if (threadIdx.x < halfPoint) { float temp = array[threadIdx.x + halfPoint]; if (temp < array[threadIdx.x]) array[threadIdx.x] = temp; } __syncthreads(); nTotalThreads = (nTotalThreads >> 1); // divide by two. } } int main(int argc, char *argv[]) { float *data = new float[N]; for(int i=0;i<N;i++) data[i] = i; float *deviceData; float *deviceMax; size_t size = N*sizeof(float); cudaMalloc((void**)&deviceData, size); displayLastError("memory allocation"); cudaMalloc((void**)&deviceMax, sizeof(float)); displayLastError("memory allocation"); cudaMemcpy(deviceData, data, size, cudaMemcpyHostToDevice); displayLastError("memory copying"); int blocks = N / BLOCK_SIZE; if(N % BLOCK_SIZE) blocks++; minMaxCuda<<<blocks, BLOCK_SIZE>>>(deviceData); displayLastError("kernel"); float max; cudaMemcpy(&max, deviceMax, sizeof(float), cudaMemcpyDeviceToHost); displayLastError("memory copying"); cudaFree(deviceData); displayLastError("free"); cudaFree(deviceMax); displayLastError("free"); delete [] data; return 0; }
655
#include "includes.h" __global__ void sgemm_kernel_v2(const float *A, const float *B, float *C, int M, int N, int K, float alpha, float beta) { int bid_x = blockIdx.x * blockDim.x; int bid_y = blockIdx.y * blockDim.y; int tid_x = threadIdx.x; int tid_y = threadIdx.y; float element_c = 0.f; __shared__ float s_tile_A[BLOCK_DIM][BLOCK_DIM]; __shared__ float s_tile_B[BLOCK_DIM][BLOCK_DIM]; // forward tile with tile size in matrix A for (int k = 0; k < K; k += BLOCK_DIM) { s_tile_A[tid_y][tid_x] = A[ (bid_y + tid_y) * K + tid_x + k ]; // Get sub-matrix from A s_tile_B[tid_y][tid_x] = B[ (k*BLOCK_DIM + tid_y) * N + bid_x + tid_x ]; // Get sub-matrix from B __syncthreads(); // compute gemm operation with tiles for (int e = 0; e < BLOCK_DIM; e++) element_c += s_tile_A[tid_y][e] * s_tile_B[e][tid_x]; __syncthreads(); } C[(bid_y + tid_y) * N + (bid_x + tid_x)] = \ alpha * element_c + beta * C[(bid_y + tid_y) * N + (bid_x + tid_x)]; }
656
// My first CUDA program! // 2018.9.1 #include <stdio.h> __global__ void helloFromGPU(void) { printf("Hello GPU! from thread \n"); } int main(void) { printf("Hello cPU! \n"); helloFromGPU <<<1,10>>>(); //cudaDeviceReset(); cudaDeviceSynchronize(); return 0; }
657
#include <cuda.h> #include <stdio.h> #include <time.h> #define N 100 __global__ void addVectorGPU(int* a, int* b, int* c) { int tid = blockIdx.x * blockDim.x + threadIdx.x; if (tid < N) { c[tid] = a[tid] + b[tid]; } } void addVectorCPU(int* a, int* b, int* c) { for (int i = 0; i < N; i++) { c[i] = a[i] + b[i]; } } void printVector(int* a) { for (int i = 0; i < N; i++) { if (i) printf(" "); printf("%d", a[i]); } printf("\n"); } int main() { clock_t start, end; double cpu_time_used, gpu_time_used; int *h_a, *h_b, *h_c, *h_result; int *d_a, *d_b, *d_c; const int size = sizeof(int) * N; h_a = (int*)malloc(size); h_b = (int*)malloc(size); h_c = (int*)malloc(size); h_result = (int*)malloc(size); // init for (int i = 0; i < N; i++) { h_a[i] = i + 1; h_b[i] = (i + 1) * 2; h_c[i] = 0; } cudaMalloc((void**)&d_a, size); cudaMalloc((void**)&d_b, size); cudaMalloc((void**)&d_c, size); cudaMemcpy(d_a, h_a, size, cudaMemcpyHostToDevice); cudaMemcpy(d_b, h_b, size, cudaMemcpyHostToDevice); start = clock(); addVectorCPU(h_a, h_b, h_c); // printVector(h_c); end = clock(); cpu_time_used = ((double) (end - start)) / CLOCKS_PER_SEC; printf("Tiempo invertido CPU = %lf s\n", cpu_time_used); dim3 threads_per_block(10, 1, 1); dim3 number_of_blocks((N / threads_per_block.x) + 1, 1, 1); start = clock(); addVectorGPU<<< number_of_blocks, threads_per_block >>>(d_a, d_b, d_c); cudaMemcpy(h_result, d_c, size, cudaMemcpyDeviceToHost); end = clock(); gpu_time_used = ((double) (end - start)) / CLOCKS_PER_SEC; printf("Tiempo invertido GPU = %lf s\n", gpu_time_used); free(h_a); free(h_b); free(h_c); free(h_result); cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); return 0; }
658
/** * Concurrent Wave Equation * Compilation Command: nvcc cuda1.cu -o cuda1 * This program was originally written in serial method by the teacher. */ #include <stdio.h> #include <stdlib.h> #include <time.h> #define MAXPOINTS 1000000 #define MAXSTEPS 1000000 #define MINPOINTS 20 static void handleError(cudaError_t err, const char *file, int line) { if (err != cudaSuccess) { printf("%s in %s at line %d\n", cudaGetErrorString(err), file, line); exit(EXIT_FAILURE); } } #define HANDLE_ERROR(err) (handleError(err, __FILE__, __LINE__)) void checkParam(); __global__ void initLine(float*, float*, int); __global__ void updateAll(float*, float*, int, int); void printResult(); int totalSteps, totalPoints, allocPoints; float *currVal; float *devCurrVal, *devPrevVal; int main(int argc, char *argv[]) { sscanf(argv[1], "%d", &totalPoints); sscanf(argv[2], "%d", &totalSteps); checkParam(); allocPoints = totalPoints + 256; currVal = (float*) malloc(allocPoints * sizeof(float)); if (!currVal) exit(EXIT_FAILURE); HANDLE_ERROR(cudaMalloc((void**) &devCurrVal, allocPoints * sizeof(float))); HANDLE_ERROR(cudaMalloc((void**) &devPrevVal, allocPoints * sizeof(float))); dim3 threadsPerBlock(256); dim3 numOfBlocks(allocPoints/256); printf("Initializing points on the line...\n"); initLine<<<numOfBlocks, threadsPerBlock>>>(devPrevVal, devCurrVal, totalPoints); printf("Updating all points for all time steps...\n"); updateAll<<<numOfBlocks, threadsPerBlock>>>(devPrevVal, devCurrVal, totalPoints, totalSteps); printf("Printing final results...\n"); HANDLE_ERROR(cudaMemcpy(currVal, devCurrVal, allocPoints * sizeof(float), cudaMemcpyDeviceToHost)); printResult(); printf("\nDone.\n\n"); cudaFree(devCurrVal); cudaFree(devPrevVal); free(currVal); return EXIT_SUCCESS; } void checkParam() { char temp[20]; while ((totalPoints < MINPOINTS) || (totalPoints > MAXPOINTS)) { printf("Enter number of points along vibrating string [%d-%d]: ", MINPOINTS, MAXPOINTS); scanf("%s", temp); totalPoints = atoi(temp); if ((totalPoints < MINPOINTS) || (totalPoints > MAXPOINTS)) printf("Invalid. Please enter value between %d and %d.\n", MINPOINTS, MAXPOINTS); } while ((totalSteps < 1) || (totalSteps > MAXSTEPS)) { printf("Enter number of time steps [1-%d]: ", MAXSTEPS); scanf("%s", temp); totalSteps = atoi(temp); if ((totalSteps < 1) || (totalSteps > MAXSTEPS)) printf("Invalid. Please enter value between 1 and %d.\n", MAXSTEPS); } printf("Using points = %d, steps = %d\n", totalPoints, totalSteps); } __global__ void initLine(float *__devPrevVal, float *__devCurrVal, int __totalPoints) { int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < __totalPoints) { float x = (float) i / (__totalPoints - 1); __devPrevVal[i] = __devCurrVal[i] = __sinf(6.28318530 * x); } } __global__ void updateAll(float *__devPrevVal, float *__devCurrVal, int __totalPoints, int __totalSteps) { int i = blockIdx.x * blockDim.x + threadIdx.x; if (i < __totalPoints) { float locPrevVal = __devPrevVal[i], locCurrVal = __devCurrVal[i] , locNextVal; for (int j = 0; j < __totalSteps; j++) { if ((i == 0) || (i == __totalPoints - 1)) locNextVal = 0.0; else locNextVal = 1.82 * locCurrVal - locPrevVal; locPrevVal = locCurrVal; locCurrVal = locNextVal; } __devCurrVal[i] = locCurrVal; } } void printResult() { for (int i = 0; i < totalPoints; i++) { printf("%6.4f ", currVal[i]); if ((i + 1) % 10 == 0) printf("\n"); } }
659
#include <cuda_runtime.h> #include "device_launch_parameters.h" #include <stdio.h> #include <iostream> #include <omp.h> #include <chrono> #include <random> #include <iomanip> #include <cmath> #include "matrix_operations.cuh" #define BLOCKSIZE_1 16 __global__ void kernel_naive_multiply_cuda(double* Ad, double* Bd, double* Cd, int M, int K, int N) { double val = 0.; int row = blockIdx.y * blockDim.y + threadIdx.y; int col = blockIdx.x * blockDim.x + threadIdx.x; if (row >= 0 && row < M && col >= 0 && col < N) { for (int k = 0; k < K; ++k) { val += *(Ad + (row)+k * M) * *(Bd + (col * K) + k); } } *(Cd + row + col * M) = val; } int naive_multiply_cuda(double* __restrict__ A, double* __restrict__ B, double* __restrict__ C, int M, int K, int N) { dim3 threads(BLOCKSIZE_1, BLOCKSIZE_1); //threads per block unsigned int blocks_x = (N + threads.x - 1) / threads.x; //basically ceil(N/threads.x) unsigned int blocks_y = (M + threads.y - 1) / threads.y; dim3 blocks(blocks_x, blocks_y); kernel_naive_multiply_cuda<<<blocks, threads>>>(A, B, C, M, K, N); return 0; } int random_uniform_initialize(double* m, int M, int N, double lower, double upper) { std::default_random_engine generator; std::uniform_real_distribution<double> distribution(lower, upper); for (int i = 0; i < M * N; ++i) { *(m + i) = distribution(generator); } return 0; } int naive_multiply_sequential(double* A, double* B, double* C, int M, int K, int N) { for (int i = 0; i < M * N; ++i) { double val = 0.; int row = i % M; int col = i / M; for (int j = 0; j < K; ++j) { val += *(A + j * M + row) * *(B + col * K + j); } *(C + i) = val; } return 0; } int naive_multiply_omp(double* A, double* B, double* C, int M, int K, int N) { #pragma omp parallel for for (int i = 0; i < M * N; ++i) { //printf("hello from thread %d", omp_get_thread_num()); double val = 0.; int row = i % M; int col = i / M; for (int j = 0; j < K; ++j) { val += *(A + j * M + row) * *(B + col * K + j); } *(C + i) = val; } return 0; } void print_matrix(const double* mat, int M, int N) { std::cout << std::endl; for (int i = 0; i < M; ++i) { for (int j = 0; j < N; ++j) { std::cout << std::fixed << std::setprecision(3) << std::setw(8) << *(mat + j * M + i) << " "; } std::cout << std::endl; } } void print_matrix_big(const double* mat, int M, int N, int include) { std::cout << std::endl << "[" << std::endl; for (int i = 0; i < include; ++i) { for (int j = 0; j < include; ++j) { std::cout << std::fixed << std::setprecision(3) << std::setw(8) << *(mat + j * M + i) << " "; } std::cout << " ..." << std::endl; } std::cout << " ..."; for (int i = (M - include); i < M; ++i) { std::cout << std::endl << " ... "; for (int j = (N - include); j < N; ++j) { std::cout << std::fixed << std::setprecision(3) << std::setw(8) << *(mat + j * M + i) << " "; } } std::cout << std::endl << "]" << std::endl; } double max_diff(const double* mat1, const double* mat2, int M, int N) { double maxd = 0.; double tdiff = 0.; for (int i = 0; i < M * N; ++i) { tdiff = std::abs(*(mat1 + i) - *(mat2 + i)); if (tdiff > maxd) { maxd = tdiff; } } return maxd; }
660
extern "C" __global__ void noarg() {} extern "C" __global__ void simple_add(float * A) { int id = blockIdx.x * blockDim.x + threadIdx.x; A[id] += 1.0; } extern "C" __global__ void four_mad(float * A) { int id = blockIdx.x * blockDim.x + threadIdx.x; float f = A[id]; f *= 41.0; f += 37.0; f *= 11.0; f += 23.0; f *= 2.0; f += 13.0; f *= 3.0; f += 7.0; A[id] = f; } #define PEAK_FLOP_MADD \ r0 = r1*r8+r0; \ r1 = r15*r9+r2; \ r2 = r14*r10+r4; \ r3 = r13*r11+r6; \ r4 = r12*r12+r8; \ r5 = r11*r13+r10; \ r6 = r10*r14+r12; \ r7 = r9*r15+r14; \ r8 = r7*r0+r1; \ r9 = r8*r1+r3; \ r10 = r6*r2+r5; \ r11 = r5*r3+r7; \ r12 = r4*r4+r9; \ r13 = r3*r5+r11; \ r14 = r2*r6+r13; \ r15 = r0*r7+r15; \ /**/ extern "C" __global__ void peak_flop(float * A) { int id = blockIdx.x * blockDim.x + threadIdx.x; float r0, r1, r2, r3, r4, r5, r6, r7; float r8, r9, r10, r11, r12, r13, r14, r15; r0 = 0.0001 * id; r1 = 0.0001 * id; r2 = 0.0002 * id; r3 = 0.0003 * id; r4 = 0.0004 * id; r5 = 0.0005 * id; r6 = 0.0006 * id; r7 = 0.0007 * id; r8 = 0.0008 * id; r9 = 0.0009 * id; r10 = 0.0010 * id; r11 = 0.0011 * id; r12 = 0.0012 * id; r13 = 0.0013 * id; r14 = 0.0014 * id; r15 = 0.0015 * id; for(int i=0; i<50; i++) { PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; PEAK_FLOP_MADD; } r0 += r1 + r2 + r3 + r4 + r5 + r6 + r7 + r8 + r9 + r10 + r11 + r12 + r13 + r14 + r15; A[id] = r0; } extern "C" __global__ void peak_flop_empty(float * A) { int id = blockIdx.x * blockDim.x + threadIdx.x; float r0, r1, r2, r3, r4, r5, r6, r7; float r8, r9, r10, r11, r12, r13, r14, r15; r0 = 0.0001 * id; r1 = 0.0001 * id; r2 = 0.0002 * id; r3 = 0.0003 * id; r4 = 0.0004 * id; r5 = 0.0005 * id; r6 = 0.0006 * id; r7 = 0.0007 * id; r8 = 0.0008 * id; r9 = 0.0009 * id; r10 = 0.0010 * id; r11 = 0.0011 * id; r12 = 0.0012 * id; r13 = 0.0013 * id; r14 = 0.0014 * id; r15 = 0.0015 * id; r0 += r1 + r2 + r3 + r4 + r5 + r6 + r7 + r8 + r9 + r10 + r11 + r12 + r13 + r14 + r15; A[id] = r0; }
661
/* #include "SiPotential.h" //-------------------- force between two Si particles ---------------------// __host__ __device__ double f2_derivative_of_rij_tag(double r_ij_tag) { double first = -4*B_Si*(1.0/(r_ij_tag*r_ij_tag*r_ij_tag*r_ij_tag*r_ij_tag)); double second = ((B_Si*(1.0/(r_ij_tag*r_ij_tag*r_ij_tag*r_ij_tag)))-1)*(1.0/((r_ij_tag-a_Si)*(r_ij_tag-a_Si))); double print = first-second; double r_ij_tag_minus_a = r_ij_tag - a_Si;//r'ij-a double r_ij_tag_minus_a2 = r_ij_tag_minus_a*r_ij_tag_minus_a; double r_ij_tag_minus_a_in_mOne = (1.0/r_ij_tag_minus_a);//(r'ij-a)^(-1) double r_ij_tag_minus_a_in_mTwo = (1.0/(r_ij_tag_minus_a2));//(r'ij-a)^(-2) double r_ij_tag2 = r_ij_tag*r_ij_tag; double r_ij_tag4 = r_ij_tag2*r_ij_tag2; double r_ij_tag5 = r_ij_tag4*r_ij_tag; double r_ij_tag_in_mFive = (1.0/(r_ij_tag5));//r'ij^(-5) double r_ij_tag_in_mFour = (1.0/(r_ij_tag4));//r'ij^(-4) double expression = B_Si * r_ij_tag_in_mFour; expression = expression - 1.0;//(B*r'ij^(-4) - 1) double exponent = exp(r_ij_tag_minus_a_in_mOne); double f2_derivative_part_1 = -4.0 * B_Si * r_ij_tag_in_mFive; double f2_derivative_part_2 = expression * r_ij_tag_minus_a_in_mTwo; return A_Si*exponent*(f2_derivative_part_1 - f2_derivative_part_2); } __host__ __device__ double v2_derivative_of_rix(real3 i, real3 j, double r_ij) { if(r_ij/sigma_Si == pow(2.0,1.0/6.0)) { return 0; } double f2_derivative = f2_derivative_of_rij_tag(r_ij/sigma_Si); f2_derivative = f2_derivative * (epsilon_Si/sigma_Si);//v2 derivative of distance double dist_x = (i.x - j.x); dist_x = dist_x / (r_ij); double v2_derivative = f2_derivative * dist_x; return v2_derivative; } __host__ __device__ double v2_derivative_of_riy(real3 i, real3 j, double r_ij) { if(r_ij/sigma_Si == pow(2.0,1.0/6.0)) { return 0; } double f2_derivative = f2_derivative_of_rij_tag(r_ij/sigma_Si); f2_derivative = f2_derivative * (epsilon_Si/sigma_Si); double dist_y = i.y - j.y; dist_y = dist_y / (r_ij); double v2_derivative = f2_derivative * dist_y; return v2_derivative; } __host__ __device__ double v2_derivative_of_riz(real3 i, real3 j, double r_ij) { if(r_ij/sigma_Si == pow(2.0,1.0/6.0)) { return 0; } double f2_derivative = f2_derivative_of_rij_tag(r_ij/sigma_Si); f2_derivative = f2_derivative * (epsilon_Si/sigma_Si); double dist_z = i.z - j.z; dist_z = dist_z / (r_ij); double v2_derivative = f2_derivative * dist_z; return v2_derivative; } //----------------------------------------------------------------------------// //-------------------- potential between two Si particles ---------------------// __host__ __device__ double f2(double r_ij_tag) { if(r_ij_tag >= a_Si) { return 0; } double r_ij_tag_minus_a_in_mOne = r_ij_tag - a_Si; r_ij_tag_minus_a_in_mOne = (1.0/r_ij_tag_minus_a_in_mOne); double exponent = exp(r_ij_tag_minus_a_in_mOne); double r_ij_tag2 = r_ij_tag*r_ij_tag; double r_ij_tag4 = r_ij_tag2*r_ij_tag2; double expression = (1.0/(r_ij_tag4)); expression *= B_Si; expression -= 1.0; return A_Si*expression*exponent; } __host__ __device__ double v2(double r_ij_tag) { if(r_ij_tag == pow(2.0,1.0/6.0)) { return -epsilon_Si; } return f2(r_ij_tag)*epsilon_Si; } //----------------------------------------------------------------------------// //------------------------ force between three Si particles -------------------// __host__ __device__ double hi_derivative_of_rij_tag(double r_ij_tag, double r_ik_tag, double r_jk_tag) { double cosJik_plus_oneThird = ((r_ij_tag*r_ij_tag + r_ik_tag*r_ik_tag - r_jk_tag*r_jk_tag)/(2.0 * r_ij_tag * r_ik_tag)) + (1.0/3.0); double r_ij_tag_minus_a = r_ij_tag - a_Si; double r_ij_tag_minus_a_in_mOne = (1.0/r_ij_tag_minus_a) * gama_Si; double r_ij_tag_minus_a_in_mTwo = (1.0/(r_ij_tag_minus_a*r_ij_tag_minus_a)) * gama_Si; double r_ik_tag_minus_a_in_mOne = (1.0/r_ik_tag - a_Si) * gama_Si; double exponent = exp(r_ij_tag_minus_a_in_mOne+r_ik_tag_minus_a_in_mOne); double expression = (r_ij_tag*r_ij_tag - r_ik_tag*r_ik_tag + r_jk_tag*r_jk_tag) / (r_ij_tag*r_ij_tag * r_ik_tag); expression -= (r_ij_tag_minus_a_in_mTwo*cosJik_plus_oneThird); return lamda_Si*exponent*cosJik_plus_oneThird*expression; } __host__ __device__ double hi_derivative_of_rik_tag(double r_ij_tag, double r_ik_tag, double r_jk_tag) { double cosJik_plus_oneThird = ((r_ij_tag*r_ij_tag + r_ik_tag*r_ik_tag - r_jk_tag*r_jk_tag)/(2.0 * r_ij_tag * r_ik_tag)) + (1.0/3.0); double r_ij_tag_minus_a_in_mOne = (1.0/(r_ij_tag - a_Si)) * gama_Si; double r_ik_tag_minus_a = r_ik_tag - a_Si; double r_ik_tag_minus_a_in_mOne = (1.0/r_ik_tag_minus_a) * gama_Si; double r_ik_tag_minus_a_in_mTwo = (1.0/(r_ik_tag_minus_a*r_ik_tag_minus_a)) * gama_Si; double exponent = exp(r_ij_tag_minus_a_in_mOne+r_ik_tag_minus_a_in_mOne); double expression = (r_ik_tag*r_ik_tag - r_ij_tag*r_ij_tag + r_jk_tag*r_jk_tag) / (r_ik_tag*r_ik_tag * r_ij_tag); expression -= (r_ik_tag_minus_a_in_mTwo*cosJik_plus_oneThird); return lamda_Si*exponent*cosJik_plus_oneThird*expression; } __host__ __device__ double hj_derivative_of_rij_tag(double r_ij_tag, double r_ik_tag, double r_jk_tag) { double cosIjk_plus_oneThird = ((r_ij_tag*r_ij_tag + r_jk_tag*r_jk_tag - r_ik_tag*r_ik_tag)/(2.0 * r_ij_tag * r_jk_tag)) + (1.0/3.0); double r_ij_tag_minus_a = r_ij_tag - a_Si; double r_ij_tag_minus_a_in_mOne = (1.0/r_ij_tag_minus_a) * gama_Si; double r_ij_tag_minus_a_in_mTwo = (1.0/(r_ij_tag_minus_a*r_ij_tag_minus_a)) * gama_Si; double r_jk_tag_minus_a_in_mOne = (1.0/(r_jk_tag - a_Si)) * gama_Si; double exponent = exp(r_ij_tag_minus_a_in_mOne+r_jk_tag_minus_a_in_mOne); double expression = (r_ij_tag*r_ij_tag - r_jk_tag*r_jk_tag + r_ik_tag*r_ik_tag) / (r_ij_tag*r_ij_tag * r_jk_tag); expression -= (r_ij_tag_minus_a_in_mTwo*cosIjk_plus_oneThird); return lamda_Si*exponent*cosIjk_plus_oneThird*expression; } __host__ __device__ double hj_derivative_of_rik_tag(double r_ij_tag, double r_ik_tag, double r_jk_tag) { double cosIjk_plus_oneThird = ((r_ij_tag*r_ij_tag + r_jk_tag*r_jk_tag - r_ik_tag*r_ik_tag)/(2.0 * r_ij_tag * r_jk_tag)) + (1.0/3.0); double r_ij_tag_minus_a_in_mOne = (1.0/(r_ij_tag - a_Si)) * gama_Si; double r_jk_tag_minus_a_in_mOne = (1.0/(r_jk_tag - a_Si)) * gama_Si; double exponent = exp(r_ij_tag_minus_a_in_mOne+r_jk_tag_minus_a_in_mOne); double expression = (-r_ik_tag) / (r_ij_tag * r_jk_tag); return lamda_Si*exponent*2.0*cosIjk_plus_oneThird*expression; } __host__ __device__ double hk_derivative_of_rij_tag(double r_ij_tag, double r_ik_tag, double r_jk_tag) { double cosIkj_plus_oneThird = ((r_ik_tag*r_ik_tag + r_jk_tag*r_jk_tag - r_ij_tag*r_ij_tag)/(2 * r_ik_tag * r_jk_tag)) + (1.0/3.0); double r_ik_tag_minus_a_in_mOne = (1.0/(r_ik_tag - a_Si)) * gama_Si; double r_jk_tag_minus_a_in_mOne = (1.0/(r_jk_tag - a_Si)) * gama_Si; double exponent = exp(r_ik_tag_minus_a_in_mOne+r_jk_tag_minus_a_in_mOne); double expression = (-r_ij_tag) / (r_ik_tag * r_jk_tag); return lamda_Si*exponent*2.0*cosIkj_plus_oneThird*expression; } __host__ __device__ double hk_derivative_of_rik_tag(double r_ij_tag, double r_ik_tag, double r_jk_tag) { double cosIkj_plus_oneThird = ((r_ik_tag*r_ik_tag + r_jk_tag*r_jk_tag - r_ij_tag*r_ij_tag)/(2.0 * r_ik_tag * r_jk_tag)) + (1.0/3.0); double r_jk_tag_minus_a_in_mOne = (1.0/(r_jk_tag - a_Si)) * gama_Si; double r_ik_tag_minus_a = r_ik_tag - a_Si; double r_ik_tag_minus_a_in_mOne = (1.0/r_ik_tag_minus_a) * gama_Si; double r_ik_tag_minus_a_in_mTwo = (1.0/(r_ik_tag_minus_a*r_ik_tag_minus_a)) * gama_Si; double exponent = exp(r_ik_tag_minus_a_in_mOne+r_jk_tag_minus_a_in_mOne); double expression = (r_ik_tag*r_ik_tag - r_jk_tag*r_jk_tag + r_ij_tag*r_ij_tag) / (r_ik_tag*r_ik_tag * r_jk_tag); expression -= (r_ik_tag_minus_a_in_mTwo*cosIkj_plus_oneThird); return lamda_Si*exponent*cosIkj_plus_oneThird*expression; } __host__ __device__ double f3_derivative_of_rij_tag(double r_ij_tag, double r_ik_tag, double r_jk_tag) { double hi_derivative_of_rij = 0.0; double hj_derivative_of_rij = 0.0; double hk_derivative_of_rij = 0.0; if(r_ij_tag < a_Si && r_ik_tag < a_Si) { hi_derivative_of_rij = hi_derivative_of_rij_tag(r_ij_tag,r_ik_tag,r_jk_tag); } if(r_jk_tag < a_Si && r_ij_tag < a_Si) { hj_derivative_of_rij = hj_derivative_of_rij_tag(r_ij_tag,r_ik_tag,r_jk_tag); } if(r_jk_tag < a_Si && r_ik_tag < a_Si) { hk_derivative_of_rij = hk_derivative_of_rij_tag(r_ij_tag,r_ik_tag,r_jk_tag); } return hi_derivative_of_rij + hj_derivative_of_rij + hk_derivative_of_rij; } __host__ __device__ double f3_derivative_of_rik_tag(double r_ij_tag, double r_ik_tag, double r_jk_tag) { double hi_derivative_of_rik = 0.0; double hj_derivative_of_rik = 0.0; double hk_derivative_of_rik = 0.0; if(r_ik_tag < a_Si && r_ij_tag < a_Si) { hi_derivative_of_rik = hi_derivative_of_rij_tag(r_ij_tag,r_ik_tag,r_jk_tag); } if(r_jk_tag < a_Si && r_ij_tag < a_Si) { hj_derivative_of_rik = hj_derivative_of_rij_tag(r_ij_tag,r_ik_tag,r_jk_tag); } if(r_jk_tag < a_Si && r_ik_tag < a_Si) { hk_derivative_of_rik = hk_derivative_of_rij_tag(r_ij_tag,r_ik_tag,r_jk_tag); } return hi_derivative_of_rik + hj_derivative_of_rik + hk_derivative_of_rik; } __host__ __device__ double v3_derivative_of_rix(real3 i, real3 j, real3 k, double r_ij, double r_ik, double r_jk) { double v3_derived_by_rij = (f3_derivative_of_rij_tag(r_ij/sigma_Si, r_ik/sigma_Si, r_jk/sigma_Si))*(epsilon_Si/sigma_Si); double v3_derived_by_rik = (f3_derivative_of_rik_tag(r_ij/sigma_Si, r_ik/sigma_Si, r_jk/sigma_Si))*(epsilon_Si/sigma_Si); double dist_ijx = (i.x-j.x); double dist_ikx = (i.x-k.x); double expression1 = (dist_ijx/(r_ij)); double expression2 = (dist_ikx/(r_ik)); return v3_derived_by_rij*expression1 + v3_derived_by_rik*expression2; } __host__ __device__ double v3_derivative_of_riy(real3 i, real3 j, real3 k, double r_ij, double r_ik, double r_jk) { double v3_derived_by_rij = (f3_derivative_of_rij_tag(r_ij/sigma_Si, r_ik/sigma_Si, r_jk/sigma_Si))*(epsilon_Si/sigma_Si); double v3_derived_by_rik = (f3_derivative_of_rik_tag(r_ij/sigma_Si, r_ik/sigma_Si, r_jk/sigma_Si))*(epsilon_Si/sigma_Si); double dist_ijy = (i.y-j.y); double dist_iky = (i.y-k.y); double expression1 = (dist_ijy/(r_ij)); double expression2 = (dist_iky/(r_ik)); return v3_derived_by_rij*expression1 + v3_derived_by_rik*expression2; } __host__ __device__ double v3_derivative_of_riz(real3 i, real3 j, real3 k, double r_ij, double r_ik, double r_jk) { double v3_derived_by_rij = (f3_derivative_of_rij_tag(r_ij/sigma_Si, r_ik/sigma_Si, r_jk/sigma_Si))*(epsilon_Si/sigma_Si); double v3_derived_by_rik = (f3_derivative_of_rik_tag(r_ij/sigma_Si, r_ik/sigma_Si, r_jk/sigma_Si))*(epsilon_Si/sigma_Si); double dist_ijz = (i.z-j.z); double dist_ikz = (i.z-k.z); double expression1 = (dist_ijz/(r_ij)); double expression2 = (dist_ikz/(r_ik)); return v3_derived_by_rij*expression1 + v3_derived_by_rik*expression2; } //----------------------------------------------------------------------------// //-------------------- potential between three Si particles -------------------// __host__ __device__ double hi(double r_ij_tag, double r_ik_tag, double r_jk_tag) { double cosJik_plus_oneThird = ((r_ij_tag*r_ij_tag + r_ik_tag*r_ik_tag - r_jk_tag*r_jk_tag)/(2.0 * r_ij_tag * r_ik_tag)) + (1.0/3.0); double r_ij_tag_minus_a_in_mOne = r_ij_tag - a_Si; r_ij_tag_minus_a_in_mOne = (1.0/r_ij_tag_minus_a_in_mOne); r_ij_tag_minus_a_in_mOne *= gama_Si; double r_ik_tag_minus_a_in_mOne = r_ik_tag - a_Si; r_ik_tag_minus_a_in_mOne = (1.0/r_ik_tag_minus_a_in_mOne); r_ik_tag_minus_a_in_mOne *= gama_Si; double exponent = exp(r_ij_tag_minus_a_in_mOne + r_ik_tag_minus_a_in_mOne); return lamda_Si*exponent*cosJik_plus_oneThird*cosJik_plus_oneThird; } __host__ __device__ double hj(double r_ij_tag, double r_ik_tag, double r_jk_tag) { double cosIjk_plus_oneThird = ((r_ij_tag*r_ij_tag + r_jk_tag*r_jk_tag - r_ik_tag*r_ik_tag)/(2.0 * r_ij_tag * r_jk_tag)) + (1.0/3.0); double r_ij_tag_minus_a_in_mOne = r_ij_tag - a_Si; r_ij_tag_minus_a_in_mOne = (1.0/r_ij_tag_minus_a_in_mOne); r_ij_tag_minus_a_in_mOne *= gama_Si; double r_jk_tag_minus_a_in_mOne = r_ik_tag - a_Si; r_jk_tag_minus_a_in_mOne = (1/r_jk_tag_minus_a_in_mOne); r_jk_tag_minus_a_in_mOne *= gama_Si; double exponent = exp(r_ij_tag_minus_a_in_mOne + r_jk_tag_minus_a_in_mOne); return lamda_Si*exponent*cosIjk_plus_oneThird*cosIjk_plus_oneThird; } __host__ __device__ double hk(double r_ij_tag, double r_ik_tag, double r_jk_tag) { double cosIkj_plus_oneThird = ((r_ik_tag*r_ik_tag + r_jk_tag*r_jk_tag - r_ij_tag*r_ij_tag)/(2.0 * r_ik_tag * r_jk_tag)) + (1.0/3.0); double r_ik_tag_minus_a_in_mOne = r_ik_tag - a_Si; r_ik_tag_minus_a_in_mOne = (1.0/r_ik_tag_minus_a_in_mOne); r_ik_tag_minus_a_in_mOne *= gama_Si; double r_jk_tag_minus_a_in_mOne = r_jk_tag - a_Si; r_jk_tag_minus_a_in_mOne = (1.0/r_jk_tag_minus_a_in_mOne); r_jk_tag_minus_a_in_mOne *= gama_Si; double exponent = exp(r_ik_tag_minus_a_in_mOne + r_jk_tag_minus_a_in_mOne); return lamda_Si*exponent*cosIkj_plus_oneThird*cosIkj_plus_oneThird; } __host__ __device__ double f3(double r_ij_tag, double r_ik_tag, double r_jk_tag) { double h_i = 0.0; double h_j = 0.0; double h_k = 0.0; if(r_ij_tag < a_Si && r_ik_tag < a_Si) { h_i = hi(r_ij_tag,r_ik_tag,r_jk_tag); } if(r_ij_tag < a_Si && r_ik_tag < a_Si) { h_j = hj(r_ij_tag,r_ik_tag,r_jk_tag); } if(r_jk_tag < a_Si && r_ik_tag < a_Si) { h_k = hk(r_ij_tag,r_ik_tag,r_jk_tag); } return h_i + h_j + h_k; } __host__ __device__ double v3(double r_ij_tag, double r_ik_tag, double r_jk_tag) { return f3(r_ij_tag,r_ik_tag,r_jk_tag)*epsilon_Si; } //----------------------------------------------------------------------------// */
662
//Header from standard libraries #include <fstream> //size of blocks in grid #define BLOCK_SIZE 16 //kernel function to multiply c=a*b __global__ void Muld(float* a,float* b,float* c,int n) { int i = blockIdx.x * blockDim.x + threadIdx.x; int j = blockIdx.y * blockDim.y + threadIdx.y; float sum = 0; if(i>=n || j>=n){ return; } for(int k = 0 ; k < n ; ++k) { sum = sum + a[i*n + k]*b[k*n + j]; } c[i*n+j] = sum; } //host function to multiply C=A*B void Mul(float* A, float* B, int n,float* C) { int size; // Load A and B to the device float* Ad; size = n * n * sizeof(float); cudaMalloc((void**)&Ad, size); cudaMemcpy(Ad, A, size, cudaMemcpyHostToDevice); float* Bd; size = n * n * sizeof(float); cudaMalloc((void**)&Bd, size); cudaMemcpy(Bd, B, size, cudaMemcpyHostToDevice); // Allocate C on the device float* Cd; size = n * n * sizeof(float); cudaMalloc((void**)&Cd, size); // Compute the execution configuration assuming // the matrix dimensions are multiples of BLOCK_SIZE dim3 dimBlock(BLOCK_SIZE, BLOCK_SIZE); dim3 dimGrid(n / dimBlock.x + (n%dimBlock.x!=0), n / dimBlock.y+(n%dimBlock.y!=0)); // Launch the device computation Muld<<<dimGrid, dimBlock>>>(Ad, Bd, Cd , n); // Read C from the device cudaMemcpy(C, Cd, size, cudaMemcpyDeviceToHost); // Free device memory cudaFree(Ad); cudaFree(Bd); cudaFree(Cd); } int main(int argc,char* argv[]) { std::ifstream fin(argv[1]); std::ofstream fout("out.txt"); int n; fin >> n; float *hA = (float*)malloc(sizeof(float)*n*n); float *hB = (float*)malloc(sizeof(float)*n*n); for(int i = 0; i < n; ++i) { for(int j = 0; j < n; ++j) { fin >> hA[i*n+j]; } } for(int i = 0; i < n; ++i) { for(int j = 0; j < n; ++j) { hB[i*n+j] = 0; } hB[i*n+i] = 1; } // in case, when count of input matrix > 1 // hC = hA * hB; // Mul(hA,hB,n,hC); for(int i = 2; i <= n; i <<=1) { Mul(hA,hA,n,hA); if(i & n) { Mul(hA,hB,n,hB); } } for(int i = 0; i < n; ++i) { for(int j = 0; j < n; ++j) { fout << hB[i*n+j] << ' '; } fout << std::endl; } free(hA); free(hB); return 0; }
663
// ########################################################## // By Eugene Ch'ng | www.complexity.io // Email: genechng@gmail.com // ---------------------------------------------------------- // The ERC 'Lost Frontiers' Project // Development for the Parallelisation of ABM Simulation // ---------------------------------------------------------- // A Basic CUDA Application for ABM Development // // In this application, we extend the Struct.cu with // an array of IDs that can represent agents from simulation code // that are passed in so that they can be processed using the GPU // The code adds position information of agents as a basis // for future development. // // The idea is to allow each kernel to process the interaction of // agents based on their positional information // // The sizes of structs and arrays are printed at the end // so that you can gauge how you should manage your memory // in large agent simulations // ---------------------------------------------------------- // How to compile: // nvcc <filename>.cu -o <outputfile> // ########################################################## #include <iostream> using namespace std; // number of agents #define N 1000 // struct to contain the position of each agent struct POSITION { float x; float y; float z; }; // struct to contain data of all agents struct AGENTS { int id[N]; // array representing agent IDs POSITION pos[N]; // array of agent positions // constructor taking in an array of agent IDs AGENTS(int *_id) { // assign IDs to the ID array passed in for(int i=0; i<N; i++) { id[i] = _id[i]; } } }; // kernel code __global__ void changeID(AGENTS *dev_agent) { // loop through all IDs in the agent struct for(int i=0; i<N; i++) { // change agent IDs in the kernel dev_agent->id[i] *= 2; // print agent IDs if(i < 10 || i > N-10) printf("** device agent %d\n", dev_agent->id[i]); } } int main(void) { cout<<"\n------------- assigning variables in host"<<endl; AGENTS *dev_agent; // the device INFO cout<<"\n------------- instantiating host id"<<endl; int id[N]; for(int i=0; i<N; i++) { id[i] = i; // assign i to each ID // output assignment if(i < 10 || i > N-10) cout<<"** host id[0]: "<<id[i]<<endl; } // send the IDs into the agent struct constructor AGENTS *agent = new AGENTS(id); cout<<"\n------------- allocate memory to device"<<endl; cudaMalloc( (void**)&dev_agent, sizeof(AGENTS) ); // copy the instantiated struct agent to device as dev_agent cout<<"\n------------- copy agent to dev_agent"<<endl; cudaMemcpy( dev_agent, agent, sizeof(AGENTS), cudaMemcpyHostToDevice); cout<<"\n------------- calling device kernel and change the id in the struct"<<endl; changeID<<<1,1>>>(dev_agent); // copy changed dev_agent to the struct, output the printing in the kernel cout<<"\n------------- copying memory from device to host and printing"<<endl; cudaMemcpy( agent, dev_agent, sizeof(AGENTS), cudaMemcpyDeviceToHost ); cout<<"\n------------- output changed results"<<endl; for(int i=0; i<N; i++) { if(i < 10 || i > N-10) cout<<"** host agent->id[0]: "<<agent->id[i]<<endl; } cout<<"\n------------- size of struct and arrays:"<<endl; cout << "size of struct AGENTS: " << sizeof(AGENTS) << " bytes" << endl; cout << "size of struct POSITION: " << sizeof(POSITION) << " bytes" << endl; cout << "size of struct id[N]: " << N * sizeof(int) << " bytes" << endl; cout<<"\n------------- cleaning up"<<endl; delete agent; cudaFree(dev_agent); return 0; }
664
/* Solving the 2D acoustic wave equation using explicit finite * difference method * Copyright 2018 Chaiwoot Boonyasiriwat. All rights reserved. */ #include <stdio.h> #include <stdlib.h> #include <math.h> #include <time.h> //Kernel __global__ void Wcalculate(float *u0,float *u1,float *u2,float C2,int nx,int ny){ int i=blockDim.x*blockIdx.x+threadIdx.x; int j=blockDim.y*blockIdx.y+threadIdx.y; if( i>0 && i<nx-1 && j>0 && j<ny-1){ u2[i+j*nx] = (2.0f-4.0f*C2)*u1[i+j*nx] - u0[i+j*nx] + C2*(u1[(i+1)+j*nx]+u1[(i-1)+j*nx] + u1[i+(j+1)*nx]+u1[i+(j-1)*nx]); } } __global__ void Wupdate(float *u0,float *u1,float *u2,int nx,int ny){ int i=blockDim.x*blockIdx.x+threadIdx.x; int j=blockDim.y*blockIdx.y+threadIdx.y; if( i<nx && j<ny){ u0[i+j*nx] = u1[i+j*nx]; u1[i+j*nx] = u2[i+j*nx]; } } int main() { //allocate parameter size_t size; cudaEvent_t start, stop; int nx, ny, nt, ix, iy, it, indx; float v, dx, dt, C, C2, xmax, ymax, a; float *u0_h, *u1_h, *u2_h; //set value xmax = 1.0f; ymax = 1.0f; nx = 201; ny = 201; v = 0.1f; dx = xmax/(nx-1); dt = 0.035f; C = v*dt/dx; C2 = C*C; nt = 1000; a = 1000.0; size = nx*ny*sizeof(float); u0_h = (float*) malloc(size); u1_h = (float*) malloc(size); u2_h = (float*) malloc(size); float *u0_cu = NULL; cudaMalloc((void**)&u0_cu,size); float *u1_cu = NULL; cudaMalloc((void**)&u1_cu,size); float *u2_cu = NULL; cudaMalloc((void**)&u2_cu,size); //initial u0 u1 for (iy=0; iy<ny; iy++) { float yy = iy*dx - 0.5*ymax; for (ix=0; ix<nx; ix++) { indx = ix+iy*nx; float xx = ix*dx - 0.5*xmax; u0_h[indx] = exp(-a*(pow(xx,2)+pow(yy,2))); u1_h[indx] = u0_h[indx]; u2_h[indx] = 0; } } //coppy u0 -> u0_cu, u1 -> u1_cu cudaMemcpy(u0_cu, u0_h, size,cudaMemcpyHostToDevice); cudaMemcpy(u1_cu, u1_h, size,cudaMemcpyHostToDevice); cudaMemcpy(u2_cu, u2_h, size,cudaMemcpyHostToDevice); //start wave calculation looping time cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start,0); dim3 G(nx/32 +1,ny/32 +1); dim3 B(32,32); for (it=0; it<nt; it++) { // advance wavefields at inner nodes Wcalculate<<<G,B>>>(u0_cu, u1_cu, u2_cu, C2, nx, ny); // update Wupdate<<<G,B>>>(u0_cu, u1_cu, u2_cu, nx, ny); } cudaMemcpy(u2_h, u2_cu, size,cudaMemcpyDeviceToHost); //end calculation cudaEventRecord(stop,0); cudaEventSynchronize(stop); float cpu_time; cudaEventElapsedTime(&cpu_time,start,stop); printf("CPU time = %lf s\n", cpu_time*0.001); // output the final snapshot FILE *file = fopen("u_cu.dat","w"); fwrite(u2_h, sizeof(float), nx*ny, file); fclose(file); // Free memory free(u0_h); free(u1_h); free(u2_h); cudaFree(u1_cu); cudaFree(u1_cu); cudaFree(u2_cu); return 0; }
665
#include <stdio.h> #include "Queue.cuh" Queue** createQueue(int width, int height){ Queue** queue=(Queue**)calloc(1, sizeof(Queue*)); *queue=(Queue*)calloc(1, sizeof(Queue)); (*queue)->width=width; (*queue)->height=height; (*queue)->entries=0; (*queue)->queue=(int**)calloc(height, sizeof(int*)); for(int entry=0; entry<height; entry++){ (*queue)->queue[entry]=(int*)calloc(width, sizeof(int)); } return queue; } void copyValuesBetweenQueues(Queue** queue1, Queue** queue2){ for(int entry=0; entry<(*queue1)->height; entry++){ for(int value=0; value<(*queue1)->width; value++){ (*queue2)->queue[entry][value]=(*queue1)->queue[entry][value]; } } (*queue2)->entries=(*queue1)->entries; } void freeQueue(Queue** queue){ for(int entry=0; entry<(*queue)->height; entry++){ free((*queue)->queue[entry]); } free((*queue)->queue); free(*queue); free(queue); } void extendQueue(Queue** queue, int newWidth, int newHeight){ Queue** queueHolder=createQueue(newWidth, newHeight); copyValuesBetweenQueues(queue, queueHolder); *queue=*queueHolder; free(queueHolder); } void freeEntryFromQueue(int* entry){ free(entry); } int isEmpty(Queue** queue){ return (*queue)->entries==0; } void insertToQueue(Queue** queue, int* entry){ if((*queue)->entries==(*queue)->height){ extendQueue(queue, (*queue)->width, (*queue)->height*2); } for(int value=0; value<(*queue)->width; value++){ (*queue)->queue[(*queue)->entries][value]=entry[value]; } (*queue)->entries++; freeEntryFromQueue(entry); } int* peekFromQueue(Queue** queue, int entryNum){ return (*queue)->queue[entryNum]; } int* popFromQueue(Queue** queue, int entryNum){ int* returnedEntry=(int*)calloc((*queue)->width, sizeof(int)); for(int value=0; value<(*queue)->width; value++){ returnedEntry[value]=(*queue)->queue[entryNum][value]; } (*queue)->entries--; for(int entry=entryNum; entry<(*queue)->entries; entry++){ for(int value=0; value<(*queue)->width; value++){ (*queue)->queue[entry][value]=(*queue)->queue[entry+1][value]; } } return returnedEntry; } int* popRandomEntryFromQueue(Queue** queue){ int entry=(int)((((double)rand())/(RAND_MAX+1))*(*queue)->entries); return popFromQueue(queue, entry); } void printQueueData(Queue** queue){ printf("Width %d, Height %d, entries %d\n", (*queue)->width, (*queue)->height, (*queue)->entries); for(int entry=0; entry<(*queue)->height; entry++){ printf("Entry %d \t", entry); for(int value=0; value<(*queue)->width; value++){ printf("%d ", (*queue)->queue[entry][value]); } printf("\n"); } }
666
#include <stdio.h> #include <stdint.h> #include <math.h> #include <cuda_runtime.h> #define FLT_MIN 1.175494351e-38F __device__ float FCC_KR = 0.3; __device__ float FCC_KB = 0.11; __device__ float SMPTE_240M_KR = 0.212; __device__ float SMPTE_240M_KB = 0.087; __device__ float REC_601_KR = 0.299; __device__ float REC_601_KB = 0.114; __device__ float REC_709_KR = 0.2126; __device__ float REC_709_KB = 0.0722; __device__ float REC_2020_KR = 0.2627; __device__ float REC_2020_KB = 0.0593; __device__ float REC709_ALPHA = 1.09929682680944f; __device__ float REC709_BETA = 0.018053968510807f; __device__ float ST2084_M1 = 0.1593017578125f; __device__ float ST2084_M2 = 78.84375f; __device__ float ST2084_C1 = 0.8359375f; __device__ float ST2084_C2 = 18.8515625f; __device__ float ST2084_C3 = 18.6875f; __device__ float yuv2rgb_REC_2020_NCL[9] = {1.000000, 1.4196651e-17, 1.47459996, 1.000000, -0.164553121, -0.571353137, 1.000000, 1.88139999, 5.67866042e-17}; __device__ float rgb2yuv_REC_709[9] = {0.212599993, 0.715200007, 0.0722000003, -0.114572108, -0.385427892, 0.500000, 0.500000, -0.454152912, -0.0458470918}; __device__ float gamutMa[9] = {1.66049099, -0.58764112, -0.0728498623, -0.124550477, 1.13289988, -0.00834942237, -0.0181507636, -0.100578897, 1.11872971}; struct FilterContext { unsigned filter_width; unsigned stride; float *data; int *left; }; __global__ void doGPU(uint8_t *data, uint8_t *dst_data, int length) { int i = blockDim.x * blockIdx.x + threadIdx.x; if (i >= length) return; dst_data[i] = data[i]; } __device__ void save_yuv(uint8_t *data, uint8_t *dst_data, int i) { for (int x = 0; x < 4; x++) { dst_data[x] = data[x]; } } __device__ uint16_t clamp(uint16_t v, uint16_t min = 0, uint16_t max = 255) { if (v > max) return max; if (v < min) return min; return v; } __device__ float rec_709_oetf(float x){ //[3] ToGammaLutOperationAVX2 //if (x < 4.5f * REC709_BETA) // x = x / 4.5f; //else // x = pow((x + (REC709_ALPHA - 1.0f)) / REC709_ALPHA, 1.0f / 0.45f); //return x; //rec_1886_inverse_eotf return x < 0.0f ? 0.0f : pow(x, 1.0f / 2.4f); } __device__ float st_2084_eotf(float x){ // [1] ToLinearLutOperationAVX2 // Filter negative values to avoid NAN. if (x > 0.0f) { float xpow = pow(x, 1.0f / ST2084_M2); float num = max(xpow - ST2084_C1, 0.0f); float den = max(ST2084_C2 - ST2084_C3 * xpow, FLT_MIN); x = pow(num / den, 1.0f / ST2084_M1); } else { x = 0.0f; } return x; } __device__ void linear_to_gamma(float r, float g, float b, float *rr, float *rg, float *rb) { *rr = rec_709_oetf(r); *rg = rec_709_oetf(g); *rb = rec_709_oetf(b); } __device__ void gamma_to_linear(float r, float g, float b, float *rr, float *rg, float *rb) { *rr = 100.0*st_2084_eotf(r); *rg = 100.0*st_2084_eotf(g); *rb = 100.0*st_2084_eotf(b); } __device__ float int16_2floaty(uint16_t x){ return float(x) * 0.00114155246 + -0.0730593577; } __device__ float int16_2float(uint16_t x){ return float(x) * 0.00111607148 + -0.571428597; } __device__ uint16_t float_2int16y(float x){ x = x * 876 + 64; float d = 0; x+=d; if(x<0.0f) x = 0.0f; if(x<float(1UL<<10)-1) return x; return float(1UL<<10)-1; } __device__ uint16_t float_2int16(float x){ x = x * 876 + 512; float d = 0; x+=d; if(x<0.0f) x = 0.0f; if(x<float(1UL<<10)-1) return x; return float(1UL<<10)-1; } __device__ void yuv_to_rgb(uint16_t *y, uint16_t *u, uint16_t *v, uint16_t *ry, uint16_t *ru, uint16_t *rv) { // int to float float iy,iu,iv; iy = int16_2floaty(*y); iu = int16_2float(*u); iv = int16_2float(*v); float r = iy * yuv2rgb_REC_2020_NCL[0] + iu * yuv2rgb_REC_2020_NCL[1] + iv * yuv2rgb_REC_2020_NCL[2]; float g = iy * yuv2rgb_REC_2020_NCL[3] + iu * yuv2rgb_REC_2020_NCL[4] + iv * yuv2rgb_REC_2020_NCL[5]; float b = iy * yuv2rgb_REC_2020_NCL[6] + iu * yuv2rgb_REC_2020_NCL[7] + iv * yuv2rgb_REC_2020_NCL[8]; float lr, lg, lb; gamma_to_linear(r,g,b, &lr, &lg, &lb); r = lr * gamutMa[0] + lg * gamutMa[1] + lb * gamutMa[2]; g = lr * gamutMa[3] + lg * gamutMa[4] + lb * gamutMa[5]; b = lr * gamutMa[6] + lg * gamutMa[7] + lb * gamutMa[8]; float gr, gg, gb; linear_to_gamma(r,g,b, &gr, &gg, &gb); //gr = r; gg=g; gb=b; iy = gr * rgb2yuv_REC_709[0] + gg * rgb2yuv_REC_709[1] + gb * rgb2yuv_REC_709[2]; iu = gr * rgb2yuv_REC_709[3] + gg * rgb2yuv_REC_709[4] + gb * rgb2yuv_REC_709[5]; iv = gr * rgb2yuv_REC_709[6] + gg * rgb2yuv_REC_709[7] + gb * rgb2yuv_REC_709[8]; // float to int *ry = clamp(float_2int16y(iy), 16*4, 235*4); *ru = clamp(float_2int16(iu), 16*4, 240*4); *rv = clamp(float_2int16(iv), 16*4, 240*4); //*ry = *y; //*ru = *u; //*rv = *v; } __global__ void calc_colorspace(uint8_t *data0, uint8_t *data1, uint8_t *data2, int w, int h, int stride_y, int stride_uv, uint8_t *dst_data0, uint8_t *dst_data1, uint8_t *dst_data2, int colorspace) { int i = (blockDim.x * blockIdx.x + threadIdx.x); int iy = i/(w/2)*2; int ix = (i % (w/2)) *2; if (ix >= w || iy >= h) return; int ofy = iy * stride_y + ix; int ofuv = (iy>>1) * stride_uv + ix/2; uint16_t * srcy = (uint16_t *)data0; uint16_t * srcu = (uint16_t *)data1; uint16_t * srcv = (uint16_t *)data2; uint16_t * dsty = (uint16_t *)dst_data0; uint16_t * dstu = (uint16_t *)dst_data1; uint16_t * dstv = (uint16_t *)dst_data2; yuv_to_rgb(&srcy[ofy], &srcu[ofuv], &srcv[ofuv], &dsty[ofy], &dstu[ofuv], &dstv[ofuv]); yuv_to_rgb(&srcy[ofy + 1], &srcu[ofuv], &srcv[ofuv], &dsty[ofy + 1], &dstu[ofuv], &dstv[ofuv]); yuv_to_rgb(&srcy[ofy + stride_y], &srcu[ofuv], &srcv[ofuv], &dsty[ofy + stride_y], &dstu[ofuv], &dstv[ofuv]); yuv_to_rgb(&srcy[ofy + stride_y +1], &srcu[ofuv], &srcv[ofuv], &dsty[ofy + stride_y + 1], &dstu[ofuv], &dstv[ofuv]); } __global__ void colorspace(float *src0, float *src1, float *src2, float *dst0, float *dst1, float *dst2, int w, int h){ int i = (blockDim.x * blockIdx.x + threadIdx.x); if(i>w*h-1) return; float iy = src0[i]; float iu = src1[i]; float iv = src2[i]; float r = iy * yuv2rgb_REC_2020_NCL[0] + iu * yuv2rgb_REC_2020_NCL[1] + iv * yuv2rgb_REC_2020_NCL[2]; float g = iy * yuv2rgb_REC_2020_NCL[3] + iu * yuv2rgb_REC_2020_NCL[4] + iv * yuv2rgb_REC_2020_NCL[5]; float b = iy * yuv2rgb_REC_2020_NCL[6] + iu * yuv2rgb_REC_2020_NCL[7] + iv * yuv2rgb_REC_2020_NCL[8]; float lr, lg, lb; gamma_to_linear(r,g,b, &lr, &lg, &lb); r = lr * gamutMa[0] + lg * gamutMa[1] + lb * gamutMa[2]; g = lr * gamutMa[3] + lg * gamutMa[4] + lb * gamutMa[5]; b = lr * gamutMa[6] + lg * gamutMa[7] + lb * gamutMa[8]; float gr, gg, gb; linear_to_gamma(r,g,b, &gr, &gg, &gb); //gr = r; gg=g; gb=b; dst0[i] = gr * rgb2yuv_REC_709[0] + gg * rgb2yuv_REC_709[1] + gb * rgb2yuv_REC_709[2]; dst1[i] = gr * rgb2yuv_REC_709[3] + gg * rgb2yuv_REC_709[4] + gb * rgb2yuv_REC_709[5]; dst2[i] = gr * rgb2yuv_REC_709[6] + gg * rgb2yuv_REC_709[7] + gb * rgb2yuv_REC_709[8]; //yuv_to_rgb(&src0[i], &src1[i], &src2[i], &dst0[i], &dst1[i], &dst2[i]); } __global__ void resize_line_32_y(uint8_t *src, float *dst, int width, int height) { int i = (blockDim.x * blockIdx.x + threadIdx.x); if(i>width*height-1) return; dst[i] = int16_2floaty(src[i]); } __global__ void resize_line_32_uv(uint8_t *src, float *dst, int wi, int hi){ int i = (blockDim.x * blockIdx.x + threadIdx.x); if(i>wi*hi-1) return; int h = int(i / wi); int v = int(i % wi); int w = int(wi / 2); if (h == 0 || h == hi-1){ if (v % 2 == 0 || v == wi - 1) dst[i] = int16_2float(src[int(h/2)*w +int(v / 2)]); else dst[i] = int16_2float(src[int(h/2) * w + int(v / 2)]) * 0.5 + int16_2float(src[int(h / 2) * w + int(v / 2) + 1]) * 0.5; } else if (h % 2 == 0){ if (v % 2 == 0 || v == wi - 1) dst[i] = int16_2float(src[int(h / 2 - 1) * w + int(v / 2)]) * 0.25 + int16_2float(src[int(h / 2) * w + int(v / 2)]) * 0.75; else{ float x1 = int16_2float(src[int(h / 2 - 1) * w + int(v / 2)]) * 0.5 + int16_2float(src[int(h / 2 - 1) * w + int(v / 2) + 1]) * 0.5; float x2 = int16_2float(src[int(h / 2) * w + int(v / 2)]) * 0.5 + int16_2float(src[int(h / 2) * w + int(v / 2) + 1]) * 0.5; dst[i] = x1 * 0.25 + x2 * 0.75; } } else{ if (v % 2 == 0 || v == wi - 1) dst[i] = int16_2float(src[int(h / 2) * w + int(v / 2)]) * 0.75 + int16_2float(src[int(h / 2 + 1) * w + int(v / 2)]) * 0.25; else{ float x1 = int16_2float(src[int(h / 2) * w + int(v / 2)]) * 0.5 + int16_2float(src[int(h / 2) * w + int(v / 2) + 1]) * 0.5; float x2 = int16_2float(src[int(h / 2 + 1) * w + int(v / 2)]) * 0.5 + int16_2float(src[int(h / 2 + 1) * w + int(v / 2) + 1]) * 0.5; dst[i] = x1 * 0.75 + x2 * 0.25; } } } __global__ void stamp_line_32_y(float *src, uint8_t *dst, int wi, int hi){ int i = (blockDim.x * blockIdx.x + threadIdx.x); if(i>wi*hi-1) return; dst[i] = clamp(float_2int16y(src[i]), 16*4, 235*4); } __global__ void stamp_line_32_uv(float *src, uint8_t *dst, int wi, int hi){ int i = (blockDim.x * blockIdx.x + threadIdx.x); if(i>wi*hi-1) return; float x1,x2,x3; x1=x2=x3=0; int h = int(i / wi); int v = int(i % wi); int w = int(wi * 2); int sh = (h+1)/2; int sv = (v+1)/2; if (h == hi -1){ float x1 = src[sh * w + sv] * 0.5 + src[(sh + 1) * w + sv] * 0.375 + src[(sh + 2) * w + sv] * 0.125; float x2 = src[sh * w + sv + 1] * 0.5 + src[(sh + 1) * w + sv + 1] * 0.375 + src[(sh + 2) * w + sv + 1] * 0.125; float x3 = src[sh * w + sv + 2] * 0.5 + src[(sh + 1) * w + sv + 2] * 0.375 + src[(sh + 2) * w + sv + 2] * 0.125; } else{ float x1 = src[sh * w + sv] * 0.125 + src[(sh + 1) * w + sv] * 0.375 + src[(sh + 2) * w + sv] * 0.375 + src[(sh + 3) * w + sv] * 0.125; float x2 = src[sh * w + sv + 1] * 0.125 + src[(sh + 1) * w + sv + 1] * 0.375 + src[(sh + 2) * w + sv + 1] * 0.375 + src[(sh + 3) * w + sv + 1] * 0.125; float x3 = src[sh * w + sv + 2] * 0.125 + src[(sh + 1) * w + sv + 2] * 0.375 + src[(sh + 2) * w + sv + 2] * 0.375 + src[(sh + 3) * w + sv + 2] * 0.125; } dst[i] = clamp(float_2int16(x1 * 0.25 + x2 * 0.5 + x3 * 0.25), 16*4, 240*4); } extern "C" { int doitgpu(uint8_t *data0, uint8_t *data1, uint8_t *data2, int *linesize, int width, int height, int format, uint8_t *out0, uint8_t *out1, uint8_t *out2) { fprintf(stderr, "=====gpu====\n"); uint8_t *cuda_data0, *cuda_data1, *cuda_data2; uint8_t *dst_data0, *dst_data1, *dst_data2; float *dst_w0, *dst_w1, *dst_w2, *dst_prt0, *dst_prt1, *dst_prt2; int length0 = linesize[0] * height; int length1 = linesize[1] * height >> 1; int length2 = linesize[2] * height >> 1; int stride_y = linesize[0]>>1; int stride_uv = linesize[1]>>1; cudaMalloc(&cuda_data0, length0); cudaMalloc(&cuda_data1, length1); cudaMalloc(&cuda_data2, length2); cudaMalloc(&dst_data0, length0); cudaMalloc(&dst_data1, length1); cudaMalloc(&dst_data2, length2); cudaMemcpy(cuda_data0, data0, length0, cudaMemcpyHostToDevice); cudaMemcpy(cuda_data1, data1, length1, cudaMemcpyHostToDevice); cudaMemcpy(cuda_data2, data2, length2, cudaMemcpyHostToDevice); cudaMalloc(&dst_w0, 4 * width * height); cudaMalloc(&dst_w1, 4 * width * height); cudaMalloc(&dst_w2, 4 * width * height); cudaMalloc(&dst_prt0, 4 * width * height); cudaMalloc(&dst_prt1, 4 * width * height); cudaMalloc(&dst_prt2, 4 * width * height); int blocks = (width * height +128 - 1)/128; int threads = 128; resize_line_32_y<<<blocks, threads>>>(cuda_data0, dst_w0, width, height); resize_line_32_uv<<<blocks, threads>>>(cuda_data1, dst_w1, width, height); resize_line_32_uv<<<blocks, threads>>>(cuda_data2, dst_w2, width, height); fprintf(stderr, "--resized\n"); colorspace<<<blocks, threads>>>(dst_w0, dst_w1, dst_w2, dst_prt0, dst_prt1, dst_prt2, width, height); fprintf(stderr, "--colorspace\n"); stamp_line_32_y<<<blocks, threads>>>(dst_prt0, dst_data0, width, height); blocks = (width * height / 4 + 128 - 1) / 128; threads = 128; //calc_colorspace <<<blocks, threads>>>(cuda_data0, cuda_data1, cuda_data2, width, height, stride_y, stride_uv, // dst_data0, dst_data1, dst_data2, 0); stamp_line_32_uv<<<blocks, threads>>>(dst_prt1, dst_data1, width/2, height/2); stamp_line_32_uv<<<blocks, threads>>>(dst_prt2, dst_data2, width/2, height/2); fprintf(stderr, "--over\n"); cudaMemcpy(out0, dst_data0, length0, cudaMemcpyDeviceToHost); cudaMemcpy(out1, dst_data1, length1, cudaMemcpyDeviceToHost); cudaMemcpy(out2, dst_data2, length2, cudaMemcpyDeviceToHost); cudaDeviceReset(); return 0; } }
667
/* * Kernel for calulating the element-wise product of two matrices * m, n --> dimensions of matrices A, B, C */ extern "C" { __global__ void hadamard(int m, int n, float *A, int lda, float *B, int ldb, float *C, int ldc) { int i = blockIdx.x * blockDim.x + threadIdx.x; int j = blockIdx.y * blockDim.y + threadIdx.y; if (i >= m || j >= n) return; C[i + j*ldc] = A[i + j*lda] * B[i + j*ldb]; } } /* * Matrix sum, parameters as above */ extern "C" { __global__ void matrix_sum(int m, int n, float *A, int lda, float *B, int ldb, float *C, int ldc) { int i = blockIdx.x * blockDim.x + threadIdx.x; int j = blockIdx.y * blockDim.y + threadIdx.y; if (i >= m || j >= n) return; C[i + j*ldc] = A[i + j*lda] + B[i + j*ldb]; } } /* * Copy of elements */ extern "C" { __global__ void copy(int m, int n, float *dst, int lddst, float *src, int ldsrc) { int i = blockIdx.x * blockDim.x + threadIdx.x; int j = blockIdx.y * blockDim.y + threadIdx.y; if (i >= m || j >= n) return; dst[i + j*lddst] = src[i + j*ldsrc]; } }
668
#include <cuda.h> /*Lx2Cuda performs the 2-D convolution of matrices A and row vector B*/ __global__ void Lx2(const float *d_in,float *d_out,int numRows,int numCols, float *mask) { //Calculate the row # of the d_in and d_out element to process int Col = blockIdx.y*blockDim.y + threadIdx.y; //Calculate the column # of the d_in and d_out element to process int Row = blockIdx.x*blockDim.x + threadIdx.x; // each thread computes one elements if ((4<Row) && (Row<numRows-5) && (4<Col) && (Col<numCols-5)){ d_out[Col*numRows+Row]=mask[0]*d_in[Col*numRows+Row]+ mask[1]*(d_in[(Col-1)*numRows+Row]+d_in[(Col+1)*numRows+Row])+ mask[2]*(d_in[(Col-2)*numRows+Row]+d_in[(Col+2)*numRows+Row])+ mask[3]*(d_in[(Col-3)*numRows+Row]+d_in[(Col+3)*numRows+Row])+ mask[4]*(d_in[(Col-4)*numRows+Row]+d_in[(Col+4)*numRows+Row])+ mask[5]*(d_in[(Col-5)*numRows+Row]+d_in[(Col+5)*numRows+Row]); } } /*Lz2 performs the 2-D convolution of matrices A and column vector B*/ __global__ void Lz2(const float *d_in,float *d_out,int numRows,int numCols, float *mask) { //Calculate the row # of the d_in and d_out element to process int Col = blockIdx.y*blockDim.y + threadIdx.y; //Calculate the column # of the d_in and d_out element to process int Row = blockIdx.x*blockDim.x + threadIdx.x; // each thread computes one elements if ((4<Row) && (Row<numRows-5) && (4<Col) && (Col<numCols-5)){ int Loc = Col*numRows+Row; d_out[Loc]=mask[0]*d_in[Loc]+ mask[1]*(d_in[Loc-1]+d_in[Loc+1])+ mask[2]*(d_in[Loc-2]+d_in[Loc+2])+ mask[3]*(d_in[Loc-3]+d_in[Loc+3])+ mask[4]*(d_in[Loc-4]+d_in[Loc+4])+ mask[5]*(d_in[Loc-5]+d_in[Loc+5]); } } /*Lz1 performs the 2-D convolution of matrices A and column vector C1*/ __global__ void Lz1(const float *d_in,float *d_out, int numRows,int numCols, float *mask) { //Calculate the row # of the d_in and d_out element to process int Col = blockIdx.y*blockDim.y + threadIdx.y; //Calculate the column # of the d_in and d_out element to process int Row = blockIdx.x*blockDim.x + threadIdx.x; // each thread computes one elements if ((5<Row) && (Row<numRows-6) && (5<Col) && (Col<numCols-6)){ int Loc = Col*numRows+Row; d_out[Loc]= mask[0]*(d_in[Loc+1]-d_in[Loc-1])+ mask[1]*(d_in[Loc+2]-d_in[Loc-2])+ mask[2]*(d_in[Loc+3]-d_in[Loc-3])+ mask[3]*(d_in[Loc+4]-d_in[Loc-4])+ mask[4]*(d_in[Loc+5]-d_in[Loc-5])+ mask[5]*(d_in[Loc+6]-d_in[Loc-6]); } } /*Lx1 performs the 2-D convolution of matrices A and row vector C1*/ __global__ void Lx1(const float *d_in,float *d_out, int numRows,int numCols, float *mask) { //Calculate the row # of the d_in and d_out element to process int Col = blockIdx.y*blockDim.y + threadIdx.y; //Calculate the column # of the d_in and d_out element to process int Row = blockIdx.x*blockDim.x + threadIdx.x; // each thread computes one elements if ((5<Row) && (Row<numRows-6) && (5<Col) && (Col<numCols-6)){ d_out[Col*numRows+Row]= mask[0]*(d_in[(Col+1)*numRows+Row]-d_in[(Col-1)*numRows+Row])+ mask[1]*(d_in[(Col+2)*numRows+Row]-d_in[(Col-2)*numRows+Row])+ mask[2]*(d_in[(Col+3)*numRows+Row]-d_in[(Col-3)*numRows+Row])+ mask[3]*(d_in[(Col+4)*numRows+Row]-d_in[(Col-4)*numRows+Row])+ mask[4]*(d_in[(Col+5)*numRows+Row]-d_in[(Col-5)*numRows+Row])+ mask[5]*(d_in[(Col+6)*numRows+Row]-d_in[(Col-6)*numRows+Row]); } } /*sbLx performs the 2-D convolution of matrices A and row vector S1*/ __global__ void sbLx(const float *d_in,float *d_out, int numRows,int numCols, float *mask) { //Calculate the row # of the d_in and d_out element to process int Col = blockIdx.y*blockDim.y + threadIdx.y; //Calculate the column # of the d_in and d_out element to process int Row = blockIdx.x*blockDim.x + threadIdx.x; // each thread computes one elements if ((5<Row) && (Row<numRows-6) && (5<Col) && (Col<numCols-6)){ d_out[Col*numRows+Row]= mask[0]*(d_in[(Col+1)*numRows+Row]-d_in[(Col-0)*numRows+Row])+ mask[1]*(d_in[(Col+2)*numRows+Row]-d_in[(Col-1)*numRows+Row])+ mask[2]*(d_in[(Col+3)*numRows+Row]-d_in[(Col-2)*numRows+Row])+ mask[3]*(d_in[(Col+4)*numRows+Row]-d_in[(Col-3)*numRows+Row])+ mask[4]*(d_in[(Col+5)*numRows+Row]-d_in[(Col-4)*numRows+Row])+ mask[5]*(d_in[(Col+6)*numRows+Row]+d_in[(Col-5)*numRows+Row]); } } /*sfLx performs the 2-D convolution of matrices A and row vector S1*/ __global__ void sfLx(const float *d_in,float *d_out, int numRows,int numCols, float *mask) { //Calculate the row # of the d_in and d_out element to process int Col = blockIdx.y*blockDim.y + threadIdx.y; //Calculate the column # of the d_in and d_out element to process int Row = blockIdx.x*blockDim.x + threadIdx.x; // each thread computes one elements if ((5<Row) && (Row<numRows-6) && (5<Col) && (Col<numCols-6)){ d_out[Col*numRows+Row]= mask[0]*(d_in[(Col+0)*numRows+Row]-d_in[(Col-1)*numRows+Row])+ mask[1]*(d_in[(Col+1)*numRows+Row]-d_in[(Col-2)*numRows+Row])+ mask[2]*(d_in[(Col+2)*numRows+Row]-d_in[(Col-3)*numRows+Row])+ mask[3]*(d_in[(Col+3)*numRows+Row]-d_in[(Col-4)*numRows+Row])+ mask[4]*(d_in[(Col+4)*numRows+Row]-d_in[(Col-5)*numRows+Row])+ mask[5]*(d_in[(Col+5)*numRows+Row]+d_in[(Col-6)*numRows+Row]); } } /*sbLz performs the 2-D convolution of matrices A and column vector S1*/ __global__ void sbLz(const float *d_in,float *d_out, int numRows,int numCols, float *mask) { //Calculate the row # of the d_in and d_out element to process int Col = blockIdx.y*blockDim.y + threadIdx.y; //Calculate the column # of the d_in and d_out element to process int Row = blockIdx.x*blockDim.x + threadIdx.x; // each thread computes one elements if ((5<Row) && (Row<numRows-6) && (5<Col) && (Col<numCols-6)){ int Loc = Col*numRows+Row; d_out[Col*numRows+Row]= mask[0]*(d_in[Loc+1]-d_in[Loc-0])+ mask[1]*(d_in[Loc+2]-d_in[Loc-1])+ mask[2]*(d_in[Loc+3]-d_in[Loc-2])+ mask[3]*(d_in[Loc+4]-d_in[Loc-3])+ mask[4]*(d_in[Loc+5]-d_in[Loc-4])+ mask[5]*(d_in[Loc+6]+d_in[Loc-5]); } } /*sfLz performs the 2-D convolution of matrices A and column vector S1*/ __global__ void sfLz(const float *d_in,float *d_out, int numRows,int numCols, float *mask) { //Calculate the row # of the d_in and d_out element to process int Col = blockIdx.y*blockDim.y + threadIdx.y; //Calculate the column # of the d_in and d_out element to process int Row = blockIdx.x*blockDim.x + threadIdx.x; // each thread computes one elements if ((5<Row) && (Row<numRows-6) && (5<Col) && (Col<numCols-6)){ int Loc = Col*numRows+Row; d_out[Loc]= mask[0]*(d_in[Loc+0]-d_in[Loc-1])+ mask[1]*(d_in[Loc+1]-d_in[Loc-2])+ mask[2]*(d_in[Loc+2]-d_in[Loc-3])+ mask[3]*(d_in[Loc+3]-d_in[Loc-4])+ mask[4]*(d_in[Loc+4]-d_in[Loc-5])+ mask[5]*(d_in[Loc+5]+d_in[Loc-6]); } } /*rsgffd performs the 2-D forward rotated staggered-grid finite difference*/ __global__ void rsgffd(const float *d_in,float *d_outx,float *d_outz,const int numRows,int numCols,float *mask) { //Calculate the row # of the d_in and d_out element to process int Col = blockIdx.y*blockDim.y + threadIdx.y; //Calculate the column # of the d_in and d_out element to process int Row = blockIdx.x*blockDim.x + threadIdx.x; // each thread computes one elements if ((5<Row) && (Row<numRows-6) && (5<Col) && (Col<numCols-6)){ d_outx[Col*numRows+Row]= mask[0]*(d_in[(Col+1)*numRows+Row-0]-d_in[(Col-0)*numRows+Row+1])+ mask[1]*(d_in[(Col+2)*numRows+Row-1]-d_in[(Col-1)*numRows+Row+2])+ mask[2]*(d_in[(Col+3)*numRows+Row-2]-d_in[(Col-2)*numRows+Row+3])+ mask[3]*(d_in[(Col+4)*numRows+Row-3]-d_in[(Col-3)*numRows+Row+4])+ mask[4]*(d_in[(Col+5)*numRows+Row-4]-d_in[(Col-4)*numRows+Row+5])+ mask[5]*(d_in[(Col+6)*numRows+Row-5]-d_in[(Col-5)*numRows+Row+6]); d_outz[Col*numRows+Row]= mask[0]*(d_in[(Col+1)*numRows+Row+1]-d_in[(Col-0)*numRows+Row-0])+ mask[1]*(d_in[(Col+2)*numRows+Row+2]-d_in[(Col-1)*numRows+Row-1])+ mask[2]*(d_in[(Col+3)*numRows+Row+3]-d_in[(Col-2)*numRows+Row-2])+ mask[3]*(d_in[(Col+4)*numRows+Row+4]-d_in[(Col-3)*numRows+Row-3])+ mask[4]*(d_in[(Col+5)*numRows+Row+5]-d_in[(Col-4)*numRows+Row-4])+ mask[5]*(d_in[(Col+6)*numRows+Row+6]-d_in[(Col-5)*numRows+Row-5]); } } /*rsgbfd performs the 2-D backward rotated staggered-grid finite difference*/ __global__ void rsgbfd(const float *d_in,float *d_outx,float *d_outz,int numRows,int numCols,float *mask) { //Calculate the row # of the d_in and d_out element to process int Col = blockIdx.y*blockDim.y + threadIdx.y; //Calculate the column # of the d_in and d_out element to process int Row = blockIdx.x*blockDim.x + threadIdx.x; // each thread computes one elements if ((5<Row) && (Row<numRows-6) && (5<Col) && (Col<numCols-6)){ d_outx[Col*numRows+Row]= mask[0]*(d_in[(Col+0)*numRows+Row-1]-d_in[(Col-1)*numRows+Row+0])+ mask[1]*(d_in[(Col+1)*numRows+Row-2]-d_in[(Col-2)*numRows+Row+1])+ mask[2]*(d_in[(Col+2)*numRows+Row-3]-d_in[(Col-3)*numRows+Row+2])+ mask[3]*(d_in[(Col+3)*numRows+Row-4]-d_in[(Col-4)*numRows+Row+3])+ mask[4]*(d_in[(Col+4)*numRows+Row-5]-d_in[(Col-5)*numRows+Row+4])+ mask[5]*(d_in[(Col+5)*numRows+Row-6]-d_in[(Col-6)*numRows+Row+5]); d_outz[Col*numRows+Row]= mask[0]*(d_in[(Col+0)*numRows+Row+0]-d_in[(Col-1)*numRows+Row-1])+ mask[1]*(d_in[(Col+1)*numRows+Row+1]-d_in[(Col-2)*numRows+Row-2])+ mask[2]*(d_in[(Col+2)*numRows+Row+2]-d_in[(Col-3)*numRows+Row-3])+ mask[3]*(d_in[(Col+3)*numRows+Row+3]-d_in[(Col-4)*numRows+Row-4])+ mask[4]*(d_in[(Col+4)*numRows+Row+4]-d_in[(Col-5)*numRows+Row-5])+ mask[5]*(d_in[(Col+5)*numRows+Row+5]-d_in[(Col-6)*numRows+Row-6]); } }
669
/* * ExBottomUpdater.cpp * * Created on: 01 февр. 2016 г. * Author: aleksandr */ #include "ExBottomUpdater.h" #include "SmartIndex.h" /* * indx должен пренадлежать участку от [0, sizeX-1] */ __device__ void ExBottomUpdater::operator() (const int indx) { int m = indx; Ex(m, 0) = coeff[0] * (Ex(m, 2) + ExBottom(0, 1, m)) + coeff[1] * (ExBottom(0, 0, m) + ExBottom(2, 0, m) - Ex(m, 1) - ExBottom(1, 1, m)) + coeff[2] * ExBottom(1, 0, m) - ExBottom(2, 1, m); for (int n = 0; n < 3; n++) { ExBottom(n, 1, m) = ExBottom(n, 0, m); ExBottom(n, 0, m) = Ex(m, n); } }
670
/* * EzBottomUpdater.cpp * * Created on: 23 янв. 2016 г. * Author: aleksandr */ #include "EzBottomUpdater.h" #include "SmartIndex.h" #define EzBottom(N, Q, M) EzBottom[(M) * 6 + (Q) * 3 + (N)] /* * indx должен пренадлежать участку от [0, sizeX-1] */ __device__ void EzBottomUpdater::operator() (const int indx) { int m = indx; Ez(m, 0) = coeff[0] * (Ez(m, 2) + EzBottom(0, 1, m)) + coeff[1] * (EzBottom(0, 0, m) + EzBottom(2, 0, m) - Ez(m, 1) - EzBottom(1, 1, m)) + coeff[2] * EzBottom(1, 0, m) - EzBottom(2, 1, m); for (int n = 0; n < 3; n++) { EzBottom(n, 1, m) = EzBottom(n, 0, m); EzBottom(n, 0, m) = Ez(m, n); } }
671
/* * CS 4444 * Steven Stetzler * Homework 5: Matrix-Matrix Multiplication with CUDA */ #include <stdio.h> #include <assert.h> #include <cuda.h> #include <cuda_runtime.h> #include <sys/time.h> #include <stdlib.h> #include <iostream> using namespace std; // If the specified error code refers to a real error, report it and quit the program void check_error(cudaError e) { if (e != cudaSuccess) { printf("\nCUDA error: %s\n", cudaGetErrorString(e)); exit(1); } } // A GPU implementation of matrix multiplication. // Given three N x N matrices A, B, and C we compute C = A x B __global__ void matrix_mult_gpu(float* A, float* B, float* C, int N) { // Get the row and column of C that this thread should work on int row = blockIdx.y * blockDim.y + threadIdx.y; int col = blockIdx.x * blockDim.x + threadIdx.x; int k, idx; float sum; // Ignore threads that would compute values outside of the boundary of the matrix if (row < N && col < N) { idx = row * N + col; sum = 0; for (k = 0; k < N; k++) { sum += A[row * N + k] * B[k * N + col]; } C[idx] = sum; } } // A naive (no cache blocking) CPU implementation of matrix multiplication // Given three N x N matrices A, B, and C we compute C = A x B void matrix_mult_cpu(float* A, float* B, float* C, int N) { int row, col, k, idx; float sum; for (row = 0; row < N; row++) { for (col = 0; col < N; col++) { idx = row * N + col; sum = 0; for (k = 0; k < N; k++) { sum += A[row * N + k] * B[k * N + col]; } C[idx] = sum; } } } // Compare the values in two N x N matrices C and C_CPU void compareHostAndGpuOutput(float* C, float* C_CPU, int N) { int totalElements = N * N; int missmatchCount = 0; for (int i = 0; i < totalElements; i++) { if (fabs(C[i] - C_CPU[i]) > 0.01) { missmatchCount++; printf("mismatch at index %i: %f\t%f\n", i, C[i], C_CPU[i]); } } if (missmatchCount > 0) { printf("Computation is incorrect: outputs do not match in %d indexes\n", missmatchCount); } else { printf("Computation is correct: CPU and GPU outputs match\n"); } } // Main method int main(int argc, char** argv) { // The problem size N is the dimension of the arrays int N = (argc > 1) ? atoi(argv[1]) : 100; // Option whether or not to check GPU output against CPU implementation // Should not be included for large N, as CPU will be very slow int check_cpu = (argc > 2) ? atoi(argv[2]) : 0; // Options to specify the thread gridding int thread_x = (argc > 3) ? atoi(argv[3]) : 32; int thread_y = (argc > 4) ? atoi(argv[4]) : 32; // Option to specify the number of trials for run time tests of the GPU int n_trials = (argc > 5) ? atoi(argv[5]) : 10; // Compute block gridding from the thread gridding // This is the minimum size block gridding given the size of the array and the thread gridding that guarantees that all values // in C will be computed int grid_x = (int) ceil((double) N / thread_x); int grid_y = (int) ceil((double) N / thread_y); // Print run parameters printf("N = %d\nGrid: %d x %d\nThreads: %d x %d\nTrials: %d\n", N, grid_x, grid_y, thread_x, thread_y, n_trials); // Specify block and thread gridding dim3 block_per_grid(grid_x, grid_y , 1); dim3 thread_per_block(thread_x, thread_y, 1); // Create and allocate three arrays float* A = (float*) malloc(N * N * sizeof(float)); float* B = (float*) malloc(N * N * sizeof(float)); float* C = (float*) malloc(N * N * sizeof(float)); // Create pointers for GPU arrays, but do not allocate yet! float* A_GPU; float* B_GPU; float* C_GPU; // Perform random initialization of the arrays int i, j; float val; for (i = 0; i < N; i++) { for (j = 0; j < N; j++) { val = (rand() % 1000) * 0.001; A[i * N + j] = val; B[i * N + j] = val; C[i * N + j] = val; } } clock_t start, end; double elapsed; // If we want to check against CPU, perform CPU matrix multiplication and time it if (check_cpu) { start = clock(); matrix_mult_cpu(A, B, C, N); end = clock(); elapsed = (end - start) / (double) CLOCKS_PER_SEC; printf("CPU: %.10f seconds\n", elapsed); } double copy_avg_time = 0; double comp_avg_time = 0; // For each trial in run time analysis for (i = 0 ; i < n_trials; i++) { // Time the copy operation start = clock(); // Allocate arrays on GPU check_error(cudaMalloc((void **) &A_GPU, N * N * sizeof(float))); check_error(cudaMalloc((void **) &B_GPU, N * N * sizeof(float))); check_error(cudaMalloc((void **) &C_GPU, N * N * sizeof(float))); // Copy in values from A, B, and C check_error(cudaMemcpy(A_GPU, A, N * N * sizeof(float), cudaMemcpyHostToDevice)); check_error(cudaMemcpy(B_GPU, B, N * N * sizeof(float), cudaMemcpyHostToDevice)); check_error(cudaMemcpy(C_GPU, C, N * N * sizeof(float), cudaMemcpyHostToDevice)); end = clock(); elapsed = (end - start) / (double) CLOCKS_PER_SEC; copy_avg_time += elapsed; // Time the computation operation start = clock(); // Perform GPU matrix multiply matrix_mult_gpu<<<block_per_grid, thread_per_block>>>(A_GPU, B_GPU, C_GPU, N); cudaDeviceSynchronize(); end = clock(); elapsed = (end - start) / (double) CLOCKS_PER_SEC; comp_avg_time += elapsed; // Free arrays if this isn't our last trial if (i != n_trials - 1) { check_error(cudaFree(A_GPU)); check_error(cudaFree(B_GPU)); check_error(cudaFree(C_GPU)); } } // Print timing results printf("GPU_copy: %.10f seconds\n", copy_avg_time / n_trials); printf("GPU: %.10f seconds\n", comp_avg_time / n_trials); // If we wanted to check against CPU, do so if (check_cpu) { // Copy result from GPU float* C_GPU_Copy = (float*) malloc(N * N * sizeof(float)); check_error(cudaMemcpy(C_GPU_Copy, C_GPU, N * N * sizeof(float), cudaMemcpyDeviceToHost)); // Compare GPU and CPU output compareHostAndGpuOutput(C, C_GPU_Copy, N); } return 0; }
672
#include <iostream> #include <stdlib.h> #include "load_matrix.cuh" int load_matrix(double ***matrix, int *size, char* filename) { FILE* ip; int i, j; if ((ip = fopen(filename, "r")) == NULL) { return 1; } fscanf(ip, "%d\n\n", size); (*matrix) = alloc_mem(*size, (*size) + 1); for (i = 0; i < *size; ++i) { for (j = 0; j < *size; ++j) fscanf(ip, "%lf\t", &(*matrix)[i][j]); fscanf(ip, "\n"); } fscanf(ip, "\n"); for (i = 0; i < *size; ++i) fscanf(ip, "%lf\n", &(*matrix)[i][(*size - 1) + 1]); fclose(ip); return 0; } double **alloc_mem(int rows_count, int cols_count) { int i; double ** mem; mem = (double **)malloc(rows_count * sizeof(double*)); for (i = 0; i < rows_count; ++i) { mem[i] = (double *)malloc(cols_count * sizeof(double)); } return mem; } void dealloc_mem(double** matrix, int rows_count) { int i; for (i = 0; i < rows_count; ++i) { free(matrix[i]); } free(matrix); }
673
#include <stdio.h> #include <stdlib.h> #include <string.h> #include <math.h> #include <cuda.h> /** * Encrypt Program Cuda * * This program encrypts a file using a degree 2 formula using Cuda * Parallelization and then decrypts the file using another degree 2 * formula. * * @Author: Clayton Chase Glenn */ #define MAX 20 #define DEBUG 0 /** Kernel Function * First finds the Thread ID within the block of GPU Threads * and if the Thread is Correct, it Encrypts the corresponding * Character in the String. **/ __global__ void encrypt(char *p, char *c, int a, int b, int n) { int tid = blockIdx.x * blockDim.x + threadIdx.x; if(tid < n) c[tid] = (a*p[tid] + b) % 256; } /** Kernel Function * First finds the Thread ID within the block of GPU Threads * and if the Thread is Correct, it Encrypts the corresponding * Character in the String. **/ __global__ void decrypt(char *p, char *c, int a, int b, int n) { int tid = blockIdx.x * blockDim.x + threadIdx.x; if(tid < n) c[tid] = (a*p[tid] + b) % 256; } /** Kernel Function * First finds the Thread ID within the block of GPU Threads * and if the Thread is Correct, it checks if the corresponding * character in array a matches b. **/ __global__ void isMatch(char *p, char *c, int *a, int n) { int tid = blockIdx.x * blockDim.x + threadIdx.x; if(tid < n) { if (c[tid] != p[tid]) { *a = 1; } } } /** * Helper Function * Prints an string to standard error showing help * for valid arguments in the executable **/ void printerror(){ fprintf(stderr, "Invalid Arguments\n"); fprintf(stderr, "Correct Form: ./encrypt [File to Encrypt]\n"); fprintf(stderr, " or\n"); fprintf(stderr, " ./encrypt -n [2^(1:20)]\n"); exit(0); } /** * Main Program * This Program is for Homework 6 to encrypt some text or show * the encryption method of text that is 2 to the power of N * characters long all initialized to zero. **/ int main(int argc, char **argv) { // Declare a buffer of max size to start int N = MAX; char *buf; // Check for immediate errors in args if (argc < 2) printerror(); if (argc == 3 && strcmp(argv[1], "-n")) printerror(); // If args match for testing, Initiallize the program if(!strcmp(argv[1], "-n") && argc == 3){ // Get total characters from args N = strtol(argv[2], NULL, 10); // Show error if N isn't within constraints if(N < 1 || N > 20) printerror(); // N <- calc to 2^N as size and allocate space N = (int)pow(2, N); buf = (char*)malloc(N*sizeof(char)); //Initiallize the buffer to Zero int i = 0; while (i < N) buf[i++] = 48; } // If 2 args, this means file if(argc == 2) { // Declare a file pointer, character array, and single character for reading FILE *fp; char c; char chars[1048576]; int i = 0; // Open the file for reading fp = fopen(argv[1], "r"); // If file is null, file does not exist or error if (fp == NULL) { fprintf(stderr, "Not a Valid File\n"); return (-1); } // Read each character and keep within 2^20, add to array while((c = fgetc(fp)) != EOF) { if (i >= 1048576) { fprintf(stderr, "File Too Large\n"); return (-1); } chars[i++] = c; } // Increment i for space and allocate space for buffer N = i + 1; buf = (char*)malloc(N*sizeof(char)); // Copy read elements into buffer i = 0; while(i < N) buf[i] = chars[i++]; // Close File, not needed anymore fclose(fp); } // Initiallize Character Arrays for Encrypting and manual memset char h_p[N]; char h_c[N]; char h_r[N]; int i = 0; while (i < N) { h_p[i] = buf[i]; h_c[i] = 32; h_r[i++] = 32; } // Init all other variables char *dev_p, *dev_c, *dev_r; int *match; int h_match = 0; int h_a = 171, h_b = 55; int r_a = 3, r_b = 91; cudaEvent_t start1, start2, start3, startf, stop1, stop2, stop3, stopf; cudaEventCreate(&start1); cudaEventCreate(&stop1); cudaEventCreate(&start2); cudaEventCreate(&stop2); cudaEventCreate(&start3); cudaEventCreate(&stop3); cudaEventCreate(&startf); cudaEventCreate(&stopf); float final_time1 = 0.0, final_time2 = 0.0, final_time3 = 0.0, final_timef = 0.0; // Allocate Memory for match flag match = (int*)malloc(sizeof(int)); *match = 0; // Allocate memory in the GPU for the character arrays cudaMalloc(&dev_p, N*sizeof(char)); cudaMalloc(&dev_c, N*sizeof(char)); cudaMalloc(&dev_r, N*sizeof(char)); cudaMalloc(&match, sizeof(int)); // Print N for distinguish printf("N: %d\n", N); // If debug on, show plain text if(DEBUG) { printf("Plain Text: "); i = 0; while(i < N) printf("%c", h_p[i++]); printf("\n"); } // Copy the Memory from the arrays to the array pointers cudaMemcpy(dev_p, h_p, N*sizeof(char), cudaMemcpyHostToDevice); cudaMemcpy(dev_c, h_c, N*sizeof(char), cudaMemcpyHostToDevice); cudaMemcpy(dev_r, h_r, N*sizeof(char), cudaMemcpyHostToDevice); // Start Total Time Record cudaEventRecord(startf); // Encrypt the Plain Text and Record Start and Finish cudaEventRecord(start1); encrypt<<<128, 128>>>(dev_p, dev_c, h_a, h_b, N); cudaEventRecord(stop1); // Copy the results from GPU to the CPU cudaMemcpy(h_c, dev_c, N*sizeof(char), cudaMemcpyDeviceToHost); // If debug on, show encrypted text if(DEBUG) { printf("Encrypted Text: "); i = 0; while(i < N) printf("%c", h_c[i++]); printf("\n"); } // Syncronize all blocks and threads in GPU and get time cudaEventSynchronize(stop1); cudaEventElapsedTime(&final_time1, start1, stop1); // Decrypt the Encrypted Text cudaEventRecord(stop2); decrypt<<<128, 128>>>(dev_c, dev_r, r_a, r_b, N); cudaEventRecord(stop2); // Copy the results from GPU to CPU cudaMemcpy(h_r, dev_r, N*sizeof(char), cudaMemcpyDeviceToHost); // If debug on, show decrypted text if(DEBUG) { printf("Decrypted Text: ", h_r); i = 0; while(i < N) printf("%c", h_r[i++]); printf("\n"); } // Syncronize all blocks and threads in GPU and get time cudaEventSynchronize(stop2); cudaEventElapsedTime(&final_time2, start2, stop2); // Check if Plain Text and Encrypt<-->Decrypt Text is matching by GPU cudaEventRecord(start3); isMatch<<<128, 128>>>(dev_r, dev_p, match, N); cudaEventRecord(stop3); // Copy the Match Result from GPU to CPU cudaMemcpy(&h_match, match, sizeof(int), cudaMemcpyDeviceToHost); // If match is zero, success, else, no success if (h_match) fprintf(stdout, "Does not Match\n"); else fprintf(stdout, "Does Match\n"); // Syncronize all blocks and threads in GPU and get time cudaEventSynchronize(stop3); cudaEventElapsedTime(&final_time3, start3, stop3); // Syncronize all blocks and threads in GPU and get time cudaEventRecord(stopf); cudaEventSynchronize(stopf); cudaEventElapsedTime(&final_timef, startf, stopf); // Print Times printf("Encrypt Time: %4.10f seconds\n", final_time1/1000); printf("Decrypt Time: %4.10f seconds\n", final_time2/1000); printf("Match Time: %4.10f seconds\n", final_time3/1000); printf("Total Time: %4.10f seconds\n\n", final_timef/1000); // Free the GPU memory cudaFree(dev_p); cudaFree(dev_c); cudaFree(dev_r); }
674
/* Copyright (C) 2012 Fabrizio Gueli * * This file is part of Cuda-complex-sim * * Cuda-complex-sim is free software: you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation, either * version 3 of the License, or (at your option) any later version. * * Cuda-complex-sim is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public * License along with Cuda-complex-sim. If not, see <http://www.gnu.org/licenses/>. */ /* #include "device.cuh" #include "host.hpp" #include "graph_transf.hpp" #include "h_barabasi_game.hpp" #include "hygra.cuh" int main(int argc, char** argv) { bool* nodes_dev; Link* links_target_dev; bool* flagw_array; bool* flagr_array; uint32_t* counter; task_t* task_dev; task_arguments* task_args_dev; message_t* inbox_dev; n_attribute *prog; curandState *d_state; // if (argc!=3) // { // perror("\nErrore"); // exit(1); // } // uint32_t max_nodes=atoi(argv[1]); // uint8_t average_links=atoi(argv[2]); uint16_t supplementary_size= 10; uint32_t max_nodes = 1000000; uint8_t average_links= 2; uint16_t barabasi_initial_nodes=average_links+1; allocateDataStructures(&prog,&nodes_dev, &task_dev, &task_args_dev, &links_target_dev, &inbox_dev,max_nodes,average_links,supplementary_size,&d_state,&flagw_array,&flagr_array,&counter); h_allocateDataStructures(supplementary_size,max_nodes,average_links); Graph g = h_barabasi_game(barabasi_initial_nodes, 1, max_nodes); startSimulation(links_target_dev,nodes_dev,supplementary_size,g); srand(time(NULL)); // Link init; // init.target=-1; // init_data<<<BLOCKS,THREADS_PER_BLOCK>>>(); // h_initArray<bool>(false,h_nodes_array,h_max_nodes_number); // h_initArray<Link>(init, h_links_target_array, h_max_nodes_number*h_average_links_number); // adjlistToCompactList(g); // copyToDevice(nodes_dev,h_nodes_array , 0, h_max_nodes_number ); // copyToDevice(links_target_dev,h_links_target_array ,0, h_max_nodes_number*h_average_links_number ); size_t avail; size_t total; cudaMemGetInfo( &avail, &total ); size_t used = total - avail; printf("\nMemoria: totale %d, in uso %d, disponibile: %d", total, used, avail); cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); // Start record cudaEventRecord(start, 0); message_test<<<BLOCKS,THREADS_PER_BLOCK,h_average_links_number*THREADS_PER_BLOCK*sizeof(Link)>>>(); message_test2nd<<<BLOCKS,THREADS_PER_BLOCK,h_average_links_number*THREADS_PER_BLOCK*sizeof(Link)>>>(); message_test2nd<<<BLOCKS,THREADS_PER_BLOCK,h_average_links_number*THREADS_PER_BLOCK*sizeof(Link)>>>(); cudaEventRecord(stop, 0); cudaEventSynchronize(stop); float elapsedTime; cudaEventElapsedTime(&elapsedTime, start, stop); // that's our time! // Clean up: cudaEventDestroy(start); cudaEventDestroy(stop); FILE *file; file=fopen("times.txt","a"); fprintf(file, "%f\n",elapsedTime); fflush(file); fclose(file); cudaThreadExit(); return 0; } */
675
// dimensions layout: // 0-2 target offset_size (flat) // 3-5 source and inclusion size (incremental) // 6-8 filter size (not incremental) // 9 sample_count __global__ void cuda_filter_tips_fast( float *target_image, float const *const source_image, float const *const inclusion_image, float const *const filter_kernel, int const *const dimensions, int const *const filter_source_offsets, float const *const error_measure, unsigned int const *const real_grid_dimensions) { //indexing unsigned int block_index_x = blockIdx.x % real_grid_dimensions[0]; unsigned int block_index_y = blockIdx.x / real_grid_dimensions[0]; unsigned int block_index_z = blockIdx.y; unsigned int target_index_x = threadIdx.x + (block_index_x * blockDim.x); unsigned int target_index_y = threadIdx.y + (block_index_y * blockDim.y); unsigned int target_index_z = threadIdx.z + (block_index_z * blockDim.z); unsigned int target_index_frame_offset = target_index_x + dimensions[0]+ ((target_index_y + dimensions[1]) * dimensions[3])+ ((target_index_z + dimensions[2]) * dimensions[4]); float total_contribution = 0; for (int filter_index = 0; filter_index < dimensions[6]; filter_index++) { int source_index_frame_offset = target_index_frame_offset + filter_source_offsets[filter_index]; if (inclusion_image[source_index_frame_offset] == 0.0f) { continue; } // compute error, this is were the magic happens: // choose any function that has range {0, 1} to determine contribution float mse = 0; for (int time_index = 0; time_index < dimensions[9]; time_index++) { float time_error = source_image[target_index_frame_offset + (time_index * dimensions[5])] - source_image[source_index_frame_offset + (time_index * dimensions[5])]; mse += time_error * time_error; } // note magic number float k = (mse * mse) / (43245000.0f); if (k >= 4.61f) { continue; } float contribution = filter_kernel[filter_index] * inclusion_image[source_index_frame_offset] * expf(-k); // here the magic ends total_contribution += contribution; for (int time_index = 0; time_index < dimensions[9]; time_index++) { target_image[target_index_frame_offset + (time_index * dimensions[5])] += source_image[source_index_frame_offset + (time_index * dimensions[5])] * contribution; } } for (int time_index = 0; time_index < dimensions[9]; time_index++) { target_image[target_index_frame_offset + (time_index * dimensions[5])] /= total_contribution; } }
676
#include "test.cuh" __global__ void cuda_lanch_add(float *ptr, const int len) { int index = threadIdx.x + blockIdx.x * blockDim.x; int stride = blockDim.x * gridDim.x; for(int i = index; i < len; i = i + stride) { if(i < len){ ptr[i] = ptr[i] * 2; } } } void add(float *ptr, const int len) { dim3 blockSize(256); dim3 gridSize(10); cuda_lanch_add<<<gridSize, blockSize>>>(ptr, len); }
677
/*----------- * * square.cu * * This is the source file of an increment kernel. * * This kernel is from CUDA samples. simpleOccupancy.cu * * streamsOptBenchmark/square.cu * * By Hao Li * *------------ */ // #include "functions.h" //////////////////////////////////////////////////////////////////////////////// // Test kernel // // This kernel squares each array element. Each thread addresses // himself with threadIdx and blockIdx, so that it can handle any // execution configuration, including anything the launch configurator // API suggests. //////////////////////////////////////////////////////////////////////////////// __global__ void square(float *in_array, float *out_array, int arrayCount) { for(int l = 0; l < 1000000; l++) { // extern __shared__ int dynamicSmem[]; int idx = threadIdx.x + blockIdx.x * blockDim.x; out_array[idx] = in_array[idx]; if (idx < arrayCount) { out_array[idx] *= out_array[idx]; } } } // int main(int argc, char **argc){ // int matrixDataSize = MATRIX_SIZE * MATRIX_SIZE; // printf("%d\n", matrixDataSize); // Matrix h_A, d_A; // initMatrix(h_A, matrixDataSize, onHOST); // initMatrix(d_A, matrixDataSize, onDEVICE); // cudaMemcpy(d_A.elements, h_A.elements, matrixDataSize, cudaMemcpyHostToDevice); // // set kernel launch configuration // dim3 threads = dim3(512, 1); // dim3 blocks = dim3(matrixDataSize / threads.x, 1); // square<<<blocks, threads, 0, 0>>>(d_A.elements, 100000); // cudaMemcpy(h_A.elements, d_A.elements, matrixDataSize, cudaMemcpyDeviceToHost); // free(h_A.elements); // cudaFree(d_A.elements); // return 0; // }
678
#include "matrix.cuh" #define ROW_INDEX 0 #define COL_INDEX 1 #define NUM_INDEXES 2 __device__ matrix_t* device_roll_matrix_list(buffer_t* buffer, matrix_list_t* list) { unsigned int i; //assert(list != NULL); //for(i=0; i<list->num; i++) //{ // assert(list->matrix_list[i] != NULL); //} unsigned int vector_size=0; for(i=0; i<list->num; i++) { vector_size += list->matrix_list[i]->rows * list->matrix_list[i]->cols; } matrix_t* vector = device_matrix_constructor(buffer, 1, vector_size); float* current_index = vector->matrix; for(i=0; i<list->num; i++) { unsigned int matrix_size = list->matrix_list[i]->rows * list->matrix_list[i]->cols; memcpy(current_index, list->matrix_list[i]->matrix, matrix_size * sizeof(float)); current_index = current_index + matrix_size; } return vector; } __device__ matrix_list_t* device_unroll_matrix_list(buffer_t* buffer, matrix_t* vector, int num, unsigned int sizes[][NUM_INDEXES]) { //assert(vector != NULL); matrix_list_t* list = device_matrix_list_constructor(buffer, num); float* current_index = vector->matrix; unsigned int i; for(i=0; i<num; i++) { list->matrix_list[i] = device_matrix_constructor(buffer, sizes[i][ROW_INDEX], sizes[i][COL_INDEX]); unsigned int matrix_size = sizes[i][ROW_INDEX] * sizes[i][COL_INDEX]; memcpy(list->matrix_list[i]->matrix, current_index, matrix_size * sizeof(float)); current_index = current_index + matrix_size; } return list; }
679
#include <cstdio> int deviceQuery() { int deviceCount; cudaGetDeviceCount(&deviceCount); for (int dev = 0; dev < deviceCount; dev++) { cudaDeviceProp deviceProp; cudaGetDeviceProperties(&deviceProp, dev); if (dev == 0) { if (deviceProp.major == 9999 && deviceProp.minor == 9999) { printf( "No CUDA GPU has been detected\n"); return -1; } else if (deviceCount == 1) { printf( "There is 1 device supporting CUDA\n"); } else { printf("There are %d devices supporting CUDA\n", deviceCount); } } printf( "Device %d: %s", dev, deviceProp.name); printf( " Computational Capabilities: %d,%d\n", deviceProp.major, deviceProp.minor); printf( " Maximum global memory size: %lu\n", deviceProp.totalGlobalMem); printf( " Maximum constant memory size: %lu\n", deviceProp.totalConstMem); printf( " Maximum shared memory size per block: %lu\n", deviceProp.sharedMemPerBlock); printf( " Maximum block dimensions: %d x %d x %d\n", deviceProp.maxThreadsDim[0], deviceProp.maxThreadsDim[1], deviceProp.maxThreadsDim[2]); printf( " Maximum grid dimensions: %d x %d x %d\n", deviceProp.maxGridSize[0], deviceProp.maxGridSize[1], deviceProp.maxGridSize[2]); printf( " Warp size: %d\n", deviceProp.warpSize); } return 0; }
680
#include <iostream> #include <cstdio> #include <cstdlib> using namespace std; int main() { cudaDeviceProp prop; int count; cudaGetDeviceCount(&count); cout << count << endl; for(int i = 0; i < count;++i) { cudaGetDeviceProperties(&prop,i); printf("Name: %s\n",prop.name); printf("Cumpute capability: %d.%d\n",prop.major,prop.minor); printf("\n"); } }
681
#include "includes.h" __global__ void _bcnn_pow_kernel(int n, float *x, float a, float *y) { int i = (blockIdx.x + blockIdx.y * gridDim.x) * blockDim.x + threadIdx.x; if (i < n) y[i] = pow(x[i], a); }
682
#include <iostream> #include <cstdlib> using namespace std; __global__ void MatrixMultiplication(unsigned long* Md, unsigned long* Nd, unsigned long* Pd, int width) { int tx = threadIdx.x; // x-index of threads int ty = threadIdx.y; // y-index of threads // Pvalue stores the Pd element that is computed by thread unsigned long Pvalue = 0; for(int k = 0; k < width; k++) { unsigned long Mdelement = Md[ty * width + k]; unsigned long Ndelement = Nd[k * width + tx]; Pvalue += Mdelement * Ndelement; } // Write the matrix to device memory // Each thread write one element Pd[ty * width + tx] = Pvalue; } int main(int argc, char **argv) { int width = 1500; cout << width << " x " << width << endl; unsigned long size = width * width * sizeof(unsigned long); unsigned long* A; unsigned long* B; unsigned long* C; A = (unsigned long*) malloc(size); B = (unsigned long*) malloc(size); C = (unsigned long*) malloc(size); srand (time(NULL)); for (int i = 0; i < width; i++) { for (int j = 0; j < width; j++) { A[i * width + j] = (rand() % 100) + 1; B[i * width + j] = (rand() % 200) + 1; } } unsigned long* Ad; unsigned long* Bd; unsigned long* Cd; // transfer A and B to device memory cudaMalloc((void**) &Ad, size); cudaMemcpy(Ad, A, size, cudaMemcpyHostToDevice); cudaMalloc((void**) &Bd, size); cudaMemcpy(Bd, B, size, cudaMemcpyHostToDevice); // allocate c on the device // C is a result of matrix multiplication cudaMalloc((void**) &Cd, size); dim3 dimBlock(width, width); dim3 dimGrid(1,1); cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start); MatrixMultiplication<<<dimGrid,dimBlock>>>(Ad, Bd, Cd, width); cudaEventRecord(stop); cudaMemcpy(C, Cd, size, cudaMemcpyDeviceToHost); cudaEventSynchronize(stop); float milliseconds = 0; cudaEventElapsedTime(&milliseconds, start, stop); cout << endl; cout << milliseconds << endl; // for (int i = 0; i < width; i++) { // for (int j = 0; j < width; j++) { // cout << C[i * width + j] << "\t"; // } // cout << endl; // } return 0; }
683
#include <stdio.h> #include <string.h> #define tpb 32 __global__ void oddCheck(int* nums,int*len, int* out, int* last){ int index=threadIdx.x + blockIdx.x*tpb; if (index<*len) out[index]=nums[index]%2; if(index==((*len)-1)) *last=out[index]; } __global__ void exToIn(int* inp, int* out, int*len, int*last){ int index = threadIdx.x + blockIdx.x*tpb; if((index>0)&&(index<*len)){ out[index-1]=inp[index]; } if(index==((*len)-1)) { out[index]=inp[index]+*last; *last=out[index]; } } __global__ void upSweep(int* arr, int* len, int* tLen, int step){ int index=threadIdx.x + blockIdx.x*tpb; if(index>*tLen) return; if((((index+1)%(step*2))!=0) || index==0 || ((*len)<=index)) return; arr[index]=arr[index]+arr[index-step]; } __global__ void downSweep(int* arr, int* len, int* tLen, int step){ int index=threadIdx.x + blockIdx.x*tpb; if(2*step==*len) arr[(*len)-1]=0; if((((index+1)%(step*2))!=0) || (index==0) || ((*len)<=index)) return; int tmp=arr[index-step]; arr[index-step]=arr[index]; arr[index]+=tmp; } __global__ void printArr(int* arr,int*len){ for(int i=0;i<((*len)-1);i++) printf("%d, ",arr[i]); printf("%d",arr[(*len)-1]); } __global__ void copyOddsP(int*inp, int*prefix, int*inpLen,int*out){ if((blockIdx.x+threadIdx.x)==0){ out[0]=inp[0];} else if((blockIdx.x+threadIdx.x)<*inpLen){ int i=threadIdx.x + blockIdx.x*tpb; if(prefix[i]!=prefix[i-1]){ out[prefix[i-1]]=inp[i]; } } } int main(int argc,char **argv){ char buff[50000]; int inp[15000]; buff[0]=' '; char* token; FILE* fp = fopen("inp.txt", "r" ); fgets(buff+1, 50000, fp); token=strtok(buff,","); int numLen=0; while(token!=NULL){ inp[numLen]=atoi(token+1); numLen++; token=strtok(NULL,","); } int* nums = inp; int falseLen=1; while(falseLen<numLen) falseLen*=2; int Len=falseLen; int* cudLen; cudaMalloc(&cudLen,sizeof(int)); cudaMemcpy(cudLen,&Len,sizeof(int),cudaMemcpyHostToDevice); int* trueLen; cudaMalloc(&trueLen,sizeof(int)); cudaMemcpy(trueLen,&numLen,sizeof(int),cudaMemcpyHostToDevice); int* cudNum; cudaMalloc(&cudNum,(Len)*sizeof(int)); cudaMemcpy(cudNum,nums,(Len)*sizeof(int),cudaMemcpyHostToDevice); int* out; cudaMalloc(&out,(Len+1)*sizeof(int)); int* last; cudaMalloc(&last,sizeof(int)); oddCheck<<<(Len+tpb)/tpb,tpb>>>(cudNum,trueLen,out,last); for(int step=1; step<Len; step*=2){ upSweep<<<(Len+tpb)/tpb,tpb>>>(out,cudLen,trueLen,step); } for(int step=Len/2; step>0; step/=2){ downSweep<<<(Len+tpb)/tpb,tpb>>>(out,cudLen,trueLen,step); } Len=numLen; cudLen=trueLen; int* shifted; cudaMalloc(&shifted,Len*sizeof(int)); exToIn<<<(Len+tpb)/tpb,tpb>>>(out,shifted,cudLen,last); int* cudOut; cudaMalloc((void**) &cudOut, Len*sizeof(int)); copyOddsP<<<(Len+tpb)/tpb,tpb>>>(cudNum, shifted, cudLen,cudOut); int len; cudaMemcpy(&len,last,sizeof(int),cudaMemcpyDeviceToHost); cudaMemcpy(inp,cudOut,len*sizeof(int),cudaMemcpyDeviceToHost); fclose(fp); FILE* fp_end = fopen("q3.txt", "w"); for (int i = 0; i < len; i++) { fprintf(fp_end, "%d", inp[i]); if (i != len-1) { fprintf(fp_end, ", "); } } cudaFree(cudLen); cudaFree(cudNum); cudaFree(out); cudaFree(last); cudaFree(shifted); cudaFree(cudOut); return 0; }
684
#include<stdio.h> #include<cuda.h> #define row1 10 #define col1 10 #define row2 10 #define col2 10 typedef long long int LLI; __global__ void matproductsharedmemory(LLI *l,LLI *m, LLI *n) { LLI x=blockIdx.x; LLI y=blockIdx.y; __shared__ LLI p[col1]; LLI i; LLI k=threadIdx.x; n[col2*y+x]=0; p[k]=l[col1*y+k]*m[col2*k+x]; __syncthreads(); for(i=0;i<col1;i++) n[col2*y+x]=n[col2*y+x]+p[i]; } int main() { LLI a[row1][col1]; LLI b[row2][col2]; LLI c[row1][col2]; LLI *d,*e,*f; LLI i,j; for(i=0;i<row1;i++) { for(j=0;j<col1;j++) { a[i][j]= i*row1+j; } } for(i=0;i<row2;i++) { for(j=0;j<col2;j++) { b[i][j]=i*row2+j; } } cudaMalloc((void **)&d,row1*col1*sizeof(LLI)); cudaMalloc((void **)&e,row2*col2*sizeof(LLI)); cudaMalloc((void **)&f,row1*col2*sizeof(LLI)); cudaMemcpy(d,a,row1*col1*sizeof(LLI),cudaMemcpyHostToDevice); cudaMemcpy(e,b,row2*col2*sizeof(LLI),cudaMemcpyHostToDevice); dim3 grid(col2,row1); /* Here we are defining two dimensional Grid(collection of blocks) structure. Syntax is dim3 grid(no. of columns,no. of rows) */ matproductsharedmemory<<<grid,col1>>>(d,e,f); cudaMemcpy(c,f,row1*col2*sizeof(LLI),cudaMemcpyDeviceToHost); /* printf("\n Product of two matrices:\n "); for(i=0;i<row1;i++) { for(j=0;j<col2;j++) { printf("%Ld\t",c[i][j]); } printf("\n"); } */ cudaFree(d); cudaFree(e); cudaFree(f); return 0; } /* OUTPUT profile ==13287== NVPROF is profiling process 13287, command: ./a.out ==13287== Profiling application: ./a.out ==13287== Profiling result: Type Time(%) Time Calls Avg Min Max Name GPU activities: 94.72% 2.5322ms 1 2.5322ms 2.5322ms 2.5322ms matproductsharedmemory(__int64*, __int64*, __int64*) 3.68% 98.338us 2 49.169us 49.025us 49.313us [CUDA memcpy HtoD] 1.61% 42.913us 1 42.913us 42.913us 42.913us [CUDA memcpy DtoH] API calls: 98.22% 189.54ms 3 63.178ms 5.3290us 189.52ms cudaMalloc 1.43% 2.7661ms 3 922.02us 26.698us 2.6712ms cudaMemcpy 0.19% 361.76us 94 3.8480us 170ns 233.68us cuDeviceGetAttribute 0.08% 150.22us 3 50.073us 6.2080us 110.67us cudaFree 0.05% 89.941us 1 89.941us 89.941us 89.941us cuDeviceTotalMem 0.01% 27.216us 1 27.216us 27.216us 27.216us cuDeviceGetName 0.01% 24.939us 1 24.939us 24.939us 24.939us cudaLaunch 0.00% 2.2690us 3 756ns 186ns 1.7650us cuDeviceGetCount 0.00% 1.0820us 2 541ns 239ns 843ns cuDeviceGet 0.00% 955ns 3 318ns 172ns 542ns cudaSetupArgument 0.00% 724ns 1 724ns 724ns 724ns cudaConfigureCall */
685
#include <stdio.h> #include "cuda_profiler_api.h" #define SIZE 20480 __global__ void vecAdd(int *a,int *b, int *c, int len) { int i=threadIdx.x+blockDim.x*blockIdx.x; if(i<len) c[i] = a[i] + b[i]; } void vecAdd_CPU(int *a,int *b, int *c, int len) { int i=0; for(i=0;i<len;i++) c[i] =a[i]+b[i]; } void loadVal(int *a, int *b, int *c, int len) { int i=0; for(i=0;i<len;i++) { a[i] = i*5; b[i] = i*6; c[i] = 0; } } void dispRes(int *arr) { int i=0; printf("Results of the first 10 elements\n"); for(i=0;i<10;i++) { printf("%d, ",arr[i]); } } int main(void) { int *a,*b,*c; int *d_a,*d_b,*d_c; //device variables a = (int*)malloc(SIZE*sizeof(int)); b = (int*)malloc(SIZE*sizeof(int)); c = (int*)malloc(SIZE*sizeof(int)); //cuda memory allocation on the device cudaMalloc((void**)&d_a,SIZE*sizeof(int)); cudaMalloc((void**)&d_b,SIZE*sizeof(int)); cudaMalloc((void**)&d_c,SIZE*sizeof(int)); printf("Loading values to the array...\n"); loadVal(a,b,c,SIZE); //cuda memory copy from host to device cudaMemcpy(d_a,a,SIZE*sizeof(int),cudaMemcpyHostToDevice); cudaMemcpy(d_b,b,SIZE*sizeof(int),cudaMemcpyHostToDevice); printf("Calling vector adding function...\n"); dim3 DimGrid((SIZE-1)/256+1,1,1); dim3 DimBlock(256,1,1); cudaProfilerStart(); vecAdd<<<DimGrid,DimBlock>>>(d_a,d_b,d_c,SIZE); cudaProfilerStop(); //CPU equivalent vecAdd_CPU(a,b,c,SIZE); //cuda memory copy from device to host cudaMemcpy(c,d_c,SIZE*sizeof(int),cudaMemcpyDeviceToHost); dispRes(c); system("PAUSE"); free(a); free(b); free(c); cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); return 0; }
686
#include<stdlib.h> #include<stdio.h> #include<malloc.h> #include<time.h> #define arm 32 __device__ int globalArray[32]; __global__ void add(int *a,int *c) { int tid = threadIdx.x; int temp=a[tid]; int count=0; while(temp!=0) { count++; temp=temp/2; } atomicAdd(&globalArray[count], 1); c[count]=globalArray[count]; } int main(void) { int arr[20]; int count=20; int bitband[arm]={0}; for (int i = 0; i < 20; ++i) { arr[i]=i; } int *d_a,*d_c; int size = sizeof(int); cudaMalloc((void **)&d_a,size*count); cudaMalloc((void **)&d_c,size*arm); cudaMemcpy(d_a,arr,size*count,cudaMemcpyHostToDevice); add<<<1,count>>>(d_a,d_c); cudaMemcpy(bitband,d_c,size*arm,cudaMemcpyDeviceToHost); for (int i = 0; i < 20; ++i) { printf("%d\n",bitband[i]); } cudaFree(d_a); cudaFree(d_c); return 0; }
687
#include "includes.h" __global__ void calc_lut(int *lut, int * hist_in, int img_size, int nbr_bin){ __shared__ int shared_hist[256]; shared_hist[threadIdx.x] = hist_in[threadIdx.x]; __syncthreads(); __shared__ int cdf[256]; __syncthreads(); int i, min, d; //int cdf = 0; min = 0; i = 0; while(min == 0){ min = shared_hist[i++]; } d = img_size - min; for (unsigned int stride = 1; stride <= threadIdx.x; stride *= 2) { __syncthreads(); shared_hist[threadIdx.x] += shared_hist[threadIdx.x-stride]; } cdf[threadIdx.x] = shared_hist[threadIdx.x]; //printf("cdf = %d\n",cdf); __syncthreads(); //for(i = 0; i <= threadIdx.x; i ++){ //tha mporouse na ginei me prefix sum san veltistoipohsh FIXME // cdf += shared_hist[i]; // lut[i] = (cdf - min)*(nbr_bin - 1)/d; //} //printf("cdf = %d\n",cdf); lut[threadIdx.x] = (int)(((float)cdf[threadIdx.x] - min)*255/d + 0.5); if(lut[threadIdx.x] < 0){ lut[threadIdx.x] = 0; } }
688
#include <iostream> #include <cuda.h> #include <cstdlib> #include <stdlib.h> #include <stdio.h> #include <time.h> const int BLOCK = 256; __global__ void AddListK(float *I, float *O, int l) { int b = blockIdx.x; int t = threadIdx.x; __shared__ float pSum[BLOCK*2]; unsigned int start = 2*blockDim.x*b; (start+t < l) ? pSum[t] = I[start+t]: pSum[t] = 0.0;//First half (start+blockDim.x+t < l) ? pSum[t+blockDim.x] = I[start+blockDim.x+t] : pSum[t+blockDim.x] = 0.0;//Second half __syncthreads(); for(unsigned int s = blockDim.x; s > 0; s/=2){ __syncthreads(); (t < s) ? pSum[t] += pSum[t+s] : pSum[t]+= 0; } //printf("Sum =%f ", pSum[0]); O[b] = pSum[0]; } __host__ double addList(float *h_I, int h_l){ float *d_I, *d_O; int olen; olen = h_l / (BLOCK<<1); //The output length equals twice the total of the length divided by width if (olen % (BLOCK<<1)) { olen++; } float h_O[olen]; cudaMalloc((void **) &d_I, sizeof(float)*h_l); cudaMalloc((void **) &d_O, sizeof(float)*olen); cudaMemcpy(d_I, h_I, sizeof(float)*h_l, cudaMemcpyHostToDevice); dim3 dimGrid(olen, 1, 1); dim3 dimBlock(BLOCK, 1, 1); AddListK<<<dimGrid, dimBlock>>>(d_I, d_O, h_l); cudaMemcpy(h_O, d_O, sizeof(float)*olen, cudaMemcpyDeviceToHost); cudaFree(d_I);cudaFree(d_O); double total = 0.0; for(int i = 0; i < olen; i ++){ total+=h_O[i]; } return total; } void populateArray(float a[], int l){ srand48(time(NULL)); float prev = drand48()*100; float nxt; for(int i = 1; i < l; i++){ do{ nxt = drand48()*100; }while(nxt==prev); a[i] = nxt; prev = nxt; } } int main(){ srand(time(NULL)); //int ilen = (rand() % 6553) * BLOCK; int ilen = 2000000; float I[ilen]; populateArray(I, ilen); printf("Input length %d", ilen); time_t gstart = time(NULL); double gtotal = 0.0; for(int i = 0; i < 1000; i ++){ gtotal = addList(I,ilen); } time_t gstop = time(NULL); time_t start = time(NULL); double total = 0.0; for(int i = 0; i < 1000; i ++){ total = 0.0; for(int i = 0; i < ilen; i ++){ total+=I[i]; } } time_t stop = time(NULL); printf("Average times\n GPU: %f CPU: %f", difftime(gstop, gstart), difftime(stop, start)); printf("TOTAL: %f == %f \n DIF: %f", total, gtotal, total-gtotal); return 0; }
689
#include <stdio.h> #include <cuda_runtime.h> __global__ void test(const int *in, int *answer) { const int tid = threadIdx.x; if (tid == 0) printf("hello!\n"); int sum = in[tid]; if (tid == 0) printf("sum[0] is %d\n", sum); *answer = sum; } int main() { int *h_data, *d_data; int N = 64; int *h_answer, *d_answer; h_data = (int *)malloc(N*sizeof(int)); h_answer = (int *)malloc(sizeof(int)); cudaMalloc(&d_data, N*sizeof(int)); cudaMalloc(&d_answer, sizeof(int)); for (int i = 0; i < N; i += 1) { h_data[i] = N - i; } cudaMemcpy(d_data, h_data, N*sizeof(int), cudaMemcpyHostToDevice); test<<<1, N>>>(h_data, d_answer); cudaMemcpy(h_answer, d_answer, sizeof(int), cudaMemcpyDeviceToHost); printf("The answer is %d\n", *h_answer); }
690
#include <stdio.h> #define TILE 32 // Thread block dimension #define N 8192 // Side of the matrix #define MATSIZE N * N // Total size of the matrix #define MEMSIZE MATSIZE * sizeof(double) // Size of matrix in memory // Generic function to be called for bandwidth testing on GPUs. typedef void (*kernelFunc)(double *, double *, int); /** * @brief Performs an optimized version of a matrix transposition. * @param a The input matrix. * @param b The transposed matrix in output. * @param size The size of the matrix side. * * By exploiting GPU shared memory we may decompose the transposition in * multiple submatrices transpositions, minimizing global memory accesses * by doing them simultaneously for same-tile threads. */ __global__ void transposeOptimized(double *a, double *b, int size) { __shared__ double tile[TILE][TILE]; int col = blockIdx.x * blockDim.x + threadIdx.x; int row = blockIdx.y * blockDim.y + threadIdx.y; tile[threadIdx.x][threadIdx.y] = a[row * size + col]; __syncthreads(); b[col * size + row] = tile[threadIdx.x][threadIdx.y]; } /** * @brief Performs a naive version of a matrix transposition on GPU. * @param a The input matrix. * @param b The transposed matrix in output. * @param size The size of the matrix side. */ __global__ void transposeNaive(double *a, double *b, int size) { int row = blockIdx.x * blockDim.x + threadIdx.x; int col = blockIdx.y * blockDim.y + threadIdx.y; b[col * size + row] = a[row * size + col]; } /** * @brief Performs a serial version of a matrix transposition on CPU. * @param a The input matrix. * @param b The transposed matrix in output. * @param size The size of the matrix side. */ void transposeCpu(double *a, double *b, int size) { for (int i = 0; i < size; ++i) for (int j = 0; j < size; ++j) b[j * size + i] = a[i * size + j]; } /** * @brief Validates the equality of two matrices in input. * @param a Matrix a. * @param b Matrix b. * @param size The size of the matrix side. */ int isCorrect(double * a, double * b, int size) { for(int i = 0; i < size; ++i) for(int j = 0; j < size; ++j) if(b[i * size + j] != a[i * size + j]) return 0; return 1; } /** * @brief Tests execution time and bandwidth of a transposition kernel on a GPU. * @param kernel The kernel to be tested. * @param kernelName The name of the kernel to be tested. * @param block_x The x-dimension of the block used to perform blocking for cache. * @param block_y The y-dimension of the block used to perform blocking for cache. * * The transposition is performed as specified by the kernel function and then is * validated against a correctly-transposed matrix. GPU time and bandwidth are * provided as outputs of the function. */ int testCudaBandwidth(kernelFunc kernel, const char * kernelName, int block_x, int block_y) { double * h_in, * h_out; double * dev_in, * dev_out; double * cpu; dim3 block(block_x, block_y); dim3 grid(N / block.x, N / block.y); h_in = (double *)malloc(MEMSIZE); h_out = (double *)malloc(MEMSIZE); cpu = (double *)malloc(MEMSIZE); cudaMalloc((void **)&dev_in, MEMSIZE); cudaMalloc((void **)&dev_out, MEMSIZE); // Fill input matrix with some indices (for validating transposition). for(int i = 0; i < MATSIZE; ++i) h_in[i] = i; // Initial setup of input matrix and timing events. cudaMemcpy(dev_in, h_in, MEMSIZE, cudaMemcpyHostToDevice); cudaEvent_t start, stop; float exec_time = 0; cudaEventCreate(&start); cudaEventCreate(&stop); // Print some informations about the current task. printf("\nTransposing matrix on CPU for validation...\n"); transposeCpu(h_in, cpu, N); printf("\nMatrix size: %dx%d, tile: %dx%d, block:%dx%d \n", N, N, TILE, TILE, block_x, block_y); printf("\nKernel: %s\n\n", kernelName); // Time kernel execution. cudaEventRecord(start); kernel<<<grid, block>>>(dev_in, dev_out, N); cudaEventRecord(stop); cudaEventSynchronize(stop); cudaEventElapsedTime(&exec_time, start, stop); cudaMemcpy(h_out, dev_out, MEMSIZE, cudaMemcpyDeviceToHost); printf("%s: %s\n", kernelName, isCorrect(h_out, cpu, N) ? "CORRECT" : "FAIL"); printf("GPU Time: %f\n", exec_time); // Bandwidth given by reading and writing a matrix during exec_time, // converted to GB/s for readability. printf("Bandwidth (GB/s): %f\n", MEMSIZE * 2 / exec_time / 1000000); printf("-------------------------------\n"); // Cleanup free(h_in); free(h_out); free(cpu); cudaFree(dev_in); cudaFree(dev_out); cudaEventDestroy(start); cudaEventDestroy(stop); return 0; } int main(int argc, char *argv[]) { testCudaBandwidth(&transposeNaive, "Naive Transpose", 32, 32); int a[5] = {2, 4, 8, 16, 32}; for(int i = 0; i < 5; ++i) for(int j = 0; j < 5; ++j) testCudaBandwidth(&transposeOptimized, "Optimized Transpose", a[i], a[j]); return 0; }
691
__global__ void findCoordinate(float *A, int *keypoints, int *newKeypoints, int lenght){ int index = blockIdx.x * blockDim.x + threadIdx.x; if(index < lenght){ float a00 = A[0]; float a01 = A[1]; float a10 = A[3]; float a11 = A[4]; float t0 = A[2]; float t1 = A[5]; int x = keypoints[index]; int y = keypoints[index+lenght]; float xp = a00*y + a01*x + t0; float yp = a10*y +a11*x + t1; int x1 = (int)xp; int y1 = (int)yp; //printf("%d %d %f %f %d %d %f %f %f %f %f %f %f\n",x1, y1, xp, yp, x, y, a00,a01,a10,a11,t0,t1); int index2 = index*2+1; int index1 = index*2; newKeypoints[index1] = x1; newKeypoints[index2] = y1; } }
692
#include "Data.cuh" #include "Data_kernel.cuh" #include <fstream> #include <iostream> #include <limits> Data::Data(): rowSize(0), columnSize(0)//, flow(0) { } Data::~Data() { delete[] this->weightLeft; delete[] this->weightRight; delete[] this->weightUp; delete[] this->weightDown; delete[] this->weightS; delete[] this->weightT; delete[] this->height; delete[] this->capacity; // delete[] this->bfsTag; cudaFree(device_active); cudaFree(device_weightLeft); cudaFree(device_weightRight); cudaFree(device_weightUp); cudaFree(device_weightDown); cudaFree(device_weightS); cudaFree(device_weightT); cudaFree(device_height); cudaFree(device_capacity); } #define DEBUG #ifdef DEBUG void Data::Print() { std::cout << "Weight:" << std::endl << std::endl; std::cout << "Left:" << std::endl; for (int i = 0; i < this->rowSize; i++) { for (int j = 0; j < this->columnSize; j++) { size_t idx = (i * this->columnSize + j); std::cout << this->weightLeft[idx] << " "; } std::cout << std::endl; } std::cout << std::endl; std::cout << "Right:" << std::endl; for (int i = 0; i < this->rowSize; i++) { for (int j = 0; j < this->columnSize; j++) { size_t idx = (i * this->columnSize + j); std::cout << this->weightRight[idx] << " "; } std::cout << std::endl; } std::cout << std::endl; std::cout << "Up:" << std::endl; for (int i = 0; i < this->rowSize; i++) { for (int j = 0; j < this->columnSize; j++) { size_t idx = (i * this->columnSize + j); std::cout << this->weightUp[idx] << " "; } std::cout << std::endl; } std::cout << std::endl; std::cout << "Down:" << std::endl; for (int i = 0; i < this->rowSize; i++) { for (int j = 0; j < this->columnSize; j++) { size_t idx = (i * this->columnSize + j); std::cout << this->weightDown[idx] << " "; } std::cout << std::endl; } std::cout << std::endl; std::cout << "Capacity:" << std::endl; for (int i = 0; i < this->rowSize; i++) { for (int j = 0; j < this->columnSize; j++) { size_t idx = (i * this->columnSize + j); std::cout << this->capacity[idx] << " "; } std::cout << std::endl; } std::cout << std::endl; std::cout << "Height:" << std::endl; for (int i = 0; i < this->rowSize; i++) { for (int j = 0; j < this->columnSize; j++) { size_t idx = (i * this->columnSize + j); std::cout << this->height[idx] << " "; } std::cout << std::endl; } std::cout << std::endl; std::cout << std::endl; } #endif Data::Data(size_t rowSize, size_t columnSize) : rowSize(rowSize), columnSize(columnSize)//, flow(0) { //weight = new int[4 * rowSize * columnSize](); // (): set to zero weightLeft = new int[rowSize * columnSize](); weightRight = new int[rowSize * columnSize](); weightUp = new int[rowSize * columnSize](); weightDown = new int[rowSize * columnSize](); weightS = new int[rowSize * columnSize](); weightT = new int[rowSize * columnSize](); height = new int[rowSize * columnSize](); capacity = new int[rowSize * columnSize](); //int *bfsTag; cudaMalloc( &device_active, sizeof(bool) ); cudaMalloc( &device_weightLeft, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_weightRight, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_weightUp, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_weightDown, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_weightS, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_weightT, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_height, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_capacity, sizeof(int) * rowSize * columnSize ); cudaMemset( device_active, false, sizeof(bool) ); cudaMemset( device_height, 0, sizeof(int) * rowSize * columnSize ); cudaMemset( device_capacity, 0, sizeof(int) * rowSize * columnSize ); } void Data::Read(const char * filename) { using std::fstream; fstream fin; fin.open(filename, fstream::in); fin >> this->rowSize >> this->columnSize; this->weightLeft = new int[this->rowSize * this->columnSize](); this->weightRight = new int[this->rowSize * this->columnSize](); this->weightUp = new int[this->rowSize * this->columnSize](); this->weightDown = new int[this->rowSize * this->columnSize](); this->weightS = new int[this->rowSize * this->columnSize](); this->weightT = new int[this->rowSize * this->columnSize](); this->height = new int[this->rowSize * this->columnSize](); this->capacity = new int[this->rowSize * this->columnSize](); //this->bfsTag = new int[this->rowSize * this->columnSize](); for (int i = 0; i < this->rowSize * this->columnSize; i++) { // order: s, t, left, right, up, down fin >> this->weightS[i] >> this->weightT[i]; fin >> this->weightLeft[i] >> this->weightRight[i] >> this->weightUp[i] >> this->weightDown[i]; } fin.close(); cudaMalloc( &device_active, sizeof(bool) ); cudaMalloc( &device_weightLeft, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_weightRight, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_weightUp, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_weightDown, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_weightS, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_weightT, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_height, sizeof(int) * rowSize * columnSize ); cudaMalloc( &device_capacity, sizeof(int) * rowSize * columnSize ); cudaMemset( device_active, false, sizeof(bool) ); cudaMemset( device_height, 0, sizeof(int) * rowSize * columnSize ); cudaMemset( device_capacity, 0, sizeof(int) * rowSize * columnSize ); cudaMemcpy( device_weightS, weightS, sizeof(int) * rowSize * columnSize, cudaMemcpyHostToDevice); cudaMemcpy( device_weightT, weightT, sizeof(int) * rowSize * columnSize, cudaMemcpyHostToDevice); cudaMemcpy( device_weightLeft, weightLeft, sizeof(int) * rowSize * columnSize, cudaMemcpyHostToDevice); cudaMemcpy( device_weightRight, weightRight, sizeof(int) * rowSize * columnSize, cudaMemcpyHostToDevice); cudaMemcpy( device_weightUp, weightUp, sizeof(int) * rowSize * columnSize, cudaMemcpyHostToDevice); cudaMemcpy( device_weightDown, weightDown, sizeof(int) * rowSize * columnSize, cudaMemcpyHostToDevice); //cudaMemcpy(y, d_y, N*sizeof(float), cudaMemcpyDeviceToHost); } int Data::GetFlow() { dim3 grid( (this->rowSize + 31) / 32, (this->columnSize + 31) / 32, 1); dim3 block(32, 1, 1); //int count = 0;// debug bool *active = new bool(true); //this->active = true; while (*active) { //for (int _ = 0; _ < 100; _ ++) { //count ++; //this->Print(); //this->active = false; cudaMemset( device_active, false, sizeof(bool) ); Data_PushLeftForLine<<< grid, block >>>( this->device_active, this->device_weightLeft, this->device_weightRight, this->device_weightS, this->device_weightT, this->device_height, this->device_capacity, this->rowSize, this->columnSize ); //cudaMemcpy(active, device_active, sizeof(bool), cudaMemcpyDeviceToHost); Data_PushUpForLine<<< grid, block >>>( this->device_active, this->device_weightUp, this->device_weightDown, this->device_weightS, this->device_weightT, this->device_height, this->device_capacity, this->rowSize, this->columnSize ); //cudaMemcpy(active, device_active, sizeof(bool), cudaMemcpyDeviceToHost); Data_PushRightForLine<<< grid, block >>>( this->device_active, this->device_weightLeft, this->device_weightRight, this->device_weightS, this->device_weightT, this->device_height, this->device_capacity, this->rowSize, this->columnSize ); //cudaMemcpy(active, device_active, sizeof(bool), cudaMemcpyDeviceToHost); Data_PushDownForLine<<< grid, block >>>( this->device_active, this->device_weightUp, this->device_weightDown, this->device_weightS, this->device_weightT, this->device_height, this->device_capacity, this->rowSize, this->columnSize ); cudaMemcpy(active, device_active, sizeof(bool), cudaMemcpyDeviceToHost); } //std::cout << count << std::endl; //#ifdef DEBUG cudaMemcpy(weightS, device_weightS, sizeof(int) * rowSize * columnSize, cudaMemcpyDeviceToHost); cudaMemcpy(weightT, device_weightT, sizeof(int) * rowSize * columnSize, cudaMemcpyDeviceToHost); cudaMemcpy(weightLeft, device_weightLeft, sizeof(int) * rowSize * columnSize, cudaMemcpyDeviceToHost); cudaMemcpy(weightRight, device_weightRight, sizeof(int) * rowSize * columnSize, cudaMemcpyDeviceToHost); cudaMemcpy(weightUp, device_weightUp, sizeof(int) * rowSize * columnSize, cudaMemcpyDeviceToHost); cudaMemcpy(weightDown, device_weightDown, sizeof(int) * rowSize * columnSize, cudaMemcpyDeviceToHost); cudaMemcpy(height, device_height, sizeof(int) * rowSize * columnSize, cudaMemcpyDeviceToHost); cudaMemcpy(capacity, device_capacity, sizeof(int) * rowSize * columnSize, cudaMemcpyDeviceToHost); //#endif // this->Print(); // std::cout << "-----" << std::endl; return 0; //return this->flow; } void Data::BfsFromT() { dim3 grid((this->rowSize * this->columnSize + 32767) / 32768, 32, 1); dim3 block(32, 32, 1); Data_BfsFromT<<< grid, block >>>( this->device_weightT, this->device_height, this->rowSize, this->columnSize ); // debug //cudaMemcpy(height, device_height, sizeof(int) * rowSize * columnSize, cudaMemcpyDeviceToHost); this->Print(); // debug int maxHeight = 1; for (bool *active = new bool(true); *active; maxHeight ++) { cudaMemset( device_active, false, sizeof(bool) ); Data_BfsLevelK<<< grid, block >>>( this->device_active, this->device_weightUp, this->device_weightDown, this->device_weightLeft, this->device_weightRight, this->device_height, this->rowSize, this->columnSize, // parameter maxHeight ); cudaMemcpy(active, device_active, sizeof(bool), cudaMemcpyDeviceToHost); } cudaMemcpy(height, device_height, sizeof(int) * rowSize * columnSize, cudaMemcpyDeviceToHost); // debug //this->Print(); // debug // html debug int result[100][100] = { 0 }; for (int i = 0; i < this->rowSize; i++) { for (int j = 0; j < this->columnSize; j++) { const size_t index = i * this->columnSize + j; int v = this->height[index]; if (v != -1) { if (i != 0 && weightLeft[index] > 0 && this->height[(i - 1) * this->columnSize + j] == -1) { result[i][j] = 1; result[i - 1][j] = 2; } if (i != this->rowSize && weightRight[index] > 0 && this->height[(i + 1) * this->columnSize + j] == -1) { result[i][j] = 1; result[i + 1][j] = 2; } if (j != this->columnSize - 1 && weightUp[index] > 0 && this->height[i * this->columnSize + j + 1] == -1) { result[i][j] = 1; result[i][j + 1] = 2; } if (j != 0 && weightDown[index] > 0 && this->height[i * this->columnSize + j - 1] == -1) { result[i][j] = 1; result[i][j - 1] = 2; } } } } std::cout << this->rowSize << " " << this->columnSize << std::endl; for (int i = 0; i < this->rowSize; i++) { std::cout << "<tr>"; for (int j = 0; j < this->columnSize; j++) { int v = this->height[i * this->columnSize + j]; //if (v == 6) { // std::cout << "(" << i << " " << j << ")" << std::endl; //} //std::cout << (v > 10000 ? -1 : v) << " "; //if (v == max_k) { // std::cout << "<td class=\"yellow\">&nbsp;</td>"; if (result[i][j] == 1) { std::cout << "<td class=\"yellow\">&nbsp;</td>"; } else if (result[i][j] == 2) { std::cout << "<td class=\"purple\">&nbsp;</td>"; } else if (this->weightS[i * this->columnSize + j] > 0) { std::cout << "<td class=\"red\">&nbsp;</td>"; } else if (this->weightT[i * this->columnSize + j] > 0) { std::cout << "<td class=\"green\">&nbsp;</td>"; } else { if (v == -1) { std::cout << "<td>&nbsp;</td>"; } else { std::cout << "<td style=\"background-color: rgb(0, 0," << (5 * v) << ")\">&nbsp;</td>"; } } } std::cout << "</tr>"; //std::cout << std::endl; } // html debug } /* void Data::BfsFromT() { for (int i = 0; i < this->rowSize; i++) { for (int j = 0; j < this->columnSize; j++) { if (this->weightT[i * this->columnSize + j] > 0) { this->height[i * this->columnSize + j] = 1; } else { this->height[i * this->columnSize + j] = std::numeric_limits<int>::max(); } } } //int max_k = -1; int maxHeight = 0; bool check = true; for (int k = 1; check; k++) { check = false; for (int i = 0; i < this->rowSize; i++) { for (int j = 0; j < this->columnSize; j++) { if (this->height[i * this->columnSize + j] == k) { // if (i != 0 && weightRight[(i - 1) * this->columnSize + j] > 0 && this->height[(i - 1) * this->columnSize + j] > k) { this->height[(i - 1) * this->columnSize + j] = k + 1; check = true; } if (i != this->rowSize && weightLeft[(i + 1) * this->columnSize + j] > 0 && this->height[(i + 1) * this->columnSize + j] > k) { this->height[(i + 1) * this->columnSize + j] = k + 1; check = true; } if (j != this->columnSize - 1 && weightDown[i * this->columnSize + j + 1] > 0 && this->height[i * this->columnSize + j + 1] > k) { this->height[i * this->columnSize + j + 1] = k + 1; check = true; } if (j != 0 && weightUp[i * this->columnSize + j - 1] > 0 && this->height[i * this->columnSize + j - 1] > k) { this->height[i * this->columnSize + j - 1] = k + 1; check = true; } // } } } //std::cout << "k = " << k << std::endl; maxHeight = k; } //for (int i = 0; i < this->rowSize * this->columnSize; i++) { // std::cout << this->weight[i * 4] << " "; // std::cout << this->weight[i * 4 + 1] << " "; // std::cout << this->weight[i * 4 + 2] << " "; // std::cout << this->weight[i * 4 + 3] << std::endl; //} // return if there is a path from t to s // for (int i = 0; i < this->rowSize; i++) { // for (int j = 0; j < this->columnSize; j++) { // const size_t index = i * this->columnSize + j; // if (this->height[index] <= maxHeight && this->weightS[index] > 0) { // return true; // } // } // } // return false; #define HTML_DEBUG #ifdef HTML_DEBUG int result[100][100] = { 0 }; for (int i = 0; i < this->rowSize; i++) { for (int j = 0; j < this->columnSize; j++) { const size_t index = i * this->columnSize + j; int v = this->height[index]; if (v <= maxHeight) { if (i != 0 && weightLeft[index] > 0 && this->height[(i - 1) * this->columnSize + j] > maxHeight) { result[i][j] = 1; result[i - 1][j] = 2; } if (i != this->rowSize && weightRight[index] > 0 && this->height[(i + 1) * this->columnSize + j] > maxHeight) { result[i][j] = 1; result[i + 1][j] = 2; } if (j != this->columnSize - 1 && weightUp[index] > 0 && this->height[i * this->columnSize + j + 1] > maxHeight) { result[i][j] = 1; result[i][j + 1] = 2; } if (j != 0 && weightDown[index] > 0 && this->height[i * this->columnSize + j - 1] > maxHeight) { result[i][j] = 1; result[i][j - 1] = 2; } } } } std::cout << this->rowSize << " " << this->columnSize << std::endl; for (int i = 0; i < this->rowSize; i++) { std::cout << "<tr>"; for (int j = 0; j < this->columnSize; j++) { int v = this->height[i * this->columnSize + j]; //if (v == 6) { // std::cout << "(" << i << " " << j << ")" << std::endl; //} //std::cout << (v > 10000 ? -1 : v) << " "; //if (v == max_k) { // std::cout << "<td class=\"yellow\">&nbsp;</td>"; if (result[i][j] == 1) { std::cout << "<td class=\"yellow\">&nbsp;</td>"; } else if (result[i][j] == 2) { std::cout << "<td class=\"purple\">&nbsp;</td>"; } else if (this->weightS[i * this->columnSize + j] > 0) { std::cout << "<td class=\"red\">&nbsp;</td>"; } else if (this->weightT[i * this->columnSize + j] > 0) { std::cout << "<td class=\"green\">&nbsp;</td>"; } else { if (v > maxHeight) { std::cout << "<td>&nbsp;</td>"; } else { std::cout << "<td style=\"background-color: rgb(0, 0," << (5 * v) << ")\">&nbsp;</td>"; } } } std::cout << "</tr>"; //std::cout << std::endl; } #endif //std::cout << this->rowSize << " " << this->columnSize << std::endl; //for (int i = 0; i < this->rowSize; i++) { // for (int j = 0; j < this->columnSize; j++) { // int v = this->height[i * this->columnSize + j]; // if (v == 6) { // std::cout << "(" << i << ", " << j << ")" << std::endl; // } // } // std::cout << std::endl; //} } */
693
#include<stdio.h> #include<stdlib.h> #include<cuda.h> #include <sys/time.h> #include<thrust/sort.h> #include<math.h> #define BLOCKSIZE 1024 struct vehicle { float time ; int id; }; struct cmp { __host__ __device__ bool operator()(const vehicle& o1, const vehicle& o2) { if (o1.time == o2.time) return o1.id < o2.id ; else return o1.time < o2.time; } }; __global__ void dkernel(int n,int k,int size,float *matrix , int dis , int *speed) { unsigned id = blockIdx.x * blockDim.x + threadIdx.x ; if (id < size) { float time = (float)dis/(float)speed[id]; matrix[id] = time*(float)60; } } __global__ void dkernel1(float *matrix ,vehicle* AT, int i,int n) { unsigned idx = blockIdx.x * blockDim.x + threadIdx.x ; if(idx < n) { AT[idx].time=matrix[i*n + idx]; AT[idx].id = idx ; } } __global__ void dkernel2(float *matrix, vehicle* AT , float* ET,int i ,int n) { unsigned idx = blockIdx.x * blockDim.x + threadIdx.x ; if(idx < n) { AT[idx].time = ET[idx] + matrix[(i+1)*n + AT[idx].id]; AT[idx].id = AT[idx].id ; } } __global__ void dkernel3(int *total_time , vehicle* AT,int n) { unsigned idx = blockIdx.x * blockDim.x + threadIdx.x ; if(idx < n) { total_time[AT[idx].id]=(int)(AT[idx].time); } } //Complete the following function void operations ( int n, int k, int m, int x, int dis, int *speed, int **results ) { int size = n * (k+1) ; int nblocks = ceil(float(float(size)/float(BLOCKSIZE))) ; float *matrix , *travel_time ; cudaMalloc(&matrix , size*sizeof(float)); travel_time=(float*)malloc(size*sizeof(float)); int *gspeed ; cudaMalloc(&gspeed , size*sizeof(int)) ; cudaMemcpy(gspeed , speed, size*sizeof(int) , cudaMemcpyHostToDevice); dkernel<<<nblocks,1024>>>(n,k,size,matrix,dis,gspeed) ; cudaMemcpy(travel_time,matrix,size*(sizeof(float)), cudaMemcpyDeviceToHost); struct vehicle arrival_time[n] ; int no_blocks = ceil(float(float(n)/float(BLOCKSIZE))) ; vehicle *AT; cudaMalloc(&AT,n*sizeof(vehicle)); dkernel1<<<no_blocks,1024>>>(matrix,AT,0,n); cudaMemcpy(arrival_time,AT,n*sizeof(vehicle),cudaMemcpyDeviceToHost); thrust::sort(arrival_time, arrival_time+n, cmp()); float end_time[n]; float* ET; cudaMalloc(&ET, n*sizeof(float)) ; for(int i=0;i<k;i++) { results[0][i]=arrival_time[0].id+1; results[1][i]=arrival_time[n-1].id+1; for(int j=0;j<n;j++) { if(j<m) { end_time[j]=arrival_time[j].time + float(x) ; } else { if(end_time[j-m] > arrival_time[j].time) { float wait_time = end_time[j-m] - arrival_time[j].time ; end_time[j] = arrival_time[j].time + wait_time + float(x) ; } else { end_time[j] = arrival_time[j].time + float(x) ; } } } cudaMemcpy(ET , end_time, n*sizeof(float) , cudaMemcpyHostToDevice); cudaMemcpy(AT , arrival_time, n*sizeof(vehicle) , cudaMemcpyHostToDevice); dkernel2<<<no_blocks,1024>>>(matrix , AT , ET, i , n); cudaMemcpy(arrival_time,AT,n*sizeof(vehicle),cudaMemcpyDeviceToHost); thrust::sort(arrival_time, arrival_time+n, cmp()); } results[0][k]=arrival_time[0].id+1; results[1][k]=arrival_time[n-1].id+1; int *total_time ; cudaMalloc(&total_time,n*sizeof(int)); cudaMemcpy(AT , arrival_time, n*sizeof(vehicle) , cudaMemcpyHostToDevice); dkernel3<<<no_blocks , 1024>>>(total_time,AT,n); cudaMemcpy(results[2],total_time,n*sizeof(int),cudaMemcpyDeviceToHost); } int main(int argc,char **argv){ //variable declarations int n,k,m,x; int dis; //Input file pointer declaration FILE *inputfilepointer; //File Opening for read char *inputfilename = argv[1]; inputfilepointer = fopen( inputfilename , "r"); //Checking if file ptr is NULL if ( inputfilepointer == NULL ) { printf( "input.txt file failed to open." ); return 0; } fscanf( inputfilepointer, "%d", &n ); //scaning for number of vehicles fscanf( inputfilepointer, "%d", &k ); //scaning for number of toll tax zones fscanf( inputfilepointer, "%d", &m ); //scaning for number of toll tax points fscanf( inputfilepointer, "%d", &x ); //scaning for toll tax zone passing time fscanf( inputfilepointer, "%d", &dis ); //scaning for distance between two consecutive toll tax zones // scanning for speeds of each vehicles for every subsequent toll tax combinations int *speed = (int *) malloc ( n*( k+1 ) * sizeof (int) ); for ( int i=0; i<=k; i++ ) { for ( int j=0; j<n; j++ ) { fscanf( inputfilepointer, "%d", &speed[i*n+j] ); } } // results is in the format of first crossing vehicles list, last crossing vehicles list // and total time taken by each vehicles to pass the highway int **results = (int **) malloc ( 3 * sizeof (int *) ); results[0] = (int *) malloc ( (k+1) * sizeof (int) ); results[1] = (int *) malloc ( (k+1) * sizeof (int) ); results[2] = (int *) malloc ( (n) * sizeof (int) ); cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); float milliseconds = 0; cudaEventRecord(start,0); // Function given to implement operations ( n, k, m, x, dis, speed, results ); cudaDeviceSynchronize(); cudaEventRecord(stop,0); cudaEventSynchronize(stop); cudaEventElapsedTime(&milliseconds, start, stop); printf("Time taken by function to execute is: %.6f ms\n", milliseconds); // Output file pointer declaration char *outputfilename = argv[2]; FILE *outputfilepointer; outputfilepointer = fopen(outputfilename,"w"); // First crossing vehicles list for ( int i=0; i<=k; i++ ) { fprintf( outputfilepointer, "%d ", results[0][i]); } fprintf( outputfilepointer, "\n"); //Last crossing vehicles list for ( int i=0; i<=k; i++ ) { fprintf( outputfilepointer, "%d ", results[1][i]); } fprintf( outputfilepointer, "\n"); //Total time taken by each vehicles to pass the highway for ( int i=0; i<n; i++ ) { fprintf( outputfilepointer, "%d ", results[2][i]); } fprintf( outputfilepointer, "\n"); fclose( outputfilepointer ); fclose( inputfilepointer ); return 0; }
694
#include "includes.h" __device__ void out_of_bounds_function(void) { *(int*) 0x87654320 = 42; } __global__ void out_of_bounds_kernel(void) { out_of_bounds_function(); }
695
// Copyright (c) 2013 Craig Wright (kungfucraig@gmail.com) // // Permission is hereby granted, free of charge, to any person obtaining a copy of // this software and associated documentation files (the "Software"), to deal in // the Software without restriction, including without limitation the rights to use, // copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the // Software, and to permit persons to whom the Software is furnished to do so, // subject to the following conditions: // // The above copyright notice and this permission notice shall be included in all copies // or substantial portions of the Software. // // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, // INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A // PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT // HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION // OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE // OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include <cuda_runtime.h> //! Square the data. __global__ void square(float *data) { const unsigned int tid = threadIdx.x; float d = data[tid]; data[tid] = d*d; } //! Provide a wrapper for the kernel. extern "C" cudaError_t runKernel(int len, float *dataD) { dim3 grid(1, 1, 1); dim3 threads(len, 1, 1); square<<< grid, threads >>>(dataD); return cudaGetLastError(); }
696
#include <thrust/host_vector.h> #include <thrust/device_vector.h> #include <thrust/copy.h> #include <thrust/transform.h> #include <iostream> struct sample_functor { double alpha; double beta; sample_functor(double _alpha, double _beta) { alpha = _alpha; beta = _beta; } __device__ double operator() (const double& x, const double& y) const { return alpha * x + (1.0 - alpha) * (beta * y); } }; int main() { thrust::host_vector<double> host_x{1.1, 3.3, 2.2}; thrust::host_vector<double> host_y{6.6, 7.7, 8.8}; thrust::host_vector<double> host_output(3); thrust::device_vector<double> device_x(3); thrust::device_vector<double> device_y(3); thrust::device_vector<double> device_output(3); double alpha = 0.005; double beta = 0.1; thrust::copy(host_x.begin(), host_x.end(), device_x.begin()); thrust::copy(host_y.begin(), host_y.end(), device_y.begin()); thrust::transform(device_x.begin(), device_x.end(), device_y.begin(), device_output.begin(), sample_functor(alpha, beta)); thrust::copy(device_output.begin(), device_output.end(), host_output.begin()); std::cout << host_output[0] << ", " << host_output[1] << ", " << host_output[2] << std::endl; // 0.6622, 0.78265, 0.8866000000000002 return 0; }
697
#include <stdio.h> #include <stdlib.h> #include <cstdio> #include <cstdlib> #include <fstream> #include <iostream> #include <cuda_runtime.h> #include <chrono> using namespace std; const int INF = ((1 << 30) - 1); // const int V = 50010; void input(char* inFileName); void output(char* outFileName); void block_FW(int B); int ceil(int a, int b); // void cal(int B, int Round, int block_start_x, int block_start_y, int block_width, int block_height); int n, m; // static int Dist[V][V]; static __align__(32) int *dist, *Result; __global__ void PHASE_ONE(int Round, int B, int* distance, int V){ // calculate target pivot extern __shared__ __align__(32) int shared_distance[]; int x = threadIdx.x; int y = threadIdx.y; int p_x = Round * B + x; int p_y = Round * B + y; if(p_y < V && p_x < V){ shared_distance[y * B + x] = distance[p_y * V + p_x]; }else{ shared_distance[y * B + x] = INF; } __syncthreads(); #pragma unroll for(int k = 0; k < B; k++){ if(shared_distance[y * B + x] > shared_distance[y * B + k] + shared_distance[k * B + x]){ shared_distance[y * B + x] = shared_distance[y * B + k] + shared_distance[k * B + x]; } __syncthreads(); } if(p_y < V && p_x < V){ distance[p_y * V + p_x] = shared_distance[y * B + x]; } } __global__ void PHASE_TWO(int Round, int B, int* distance, int V){ // grid size (Round * 2) if(blockIdx.x == Round){ return; } extern __shared__ __align__(32) int shared[]; // first block stores pivot block // rest of spaces store distance col/row's block int* shared_p_distacne = shared; int* shared_distance = shared + B*B; int x = threadIdx.x; int y = threadIdx.y; int p_x = Round * B + x; int p_y = Round * B + y; int idx_y, idx_x; // fill present distance if(p_y < V && p_x < V){ shared_p_distacne[y * B + x] = distance[p_y * V + p_x]; }else{ shared_p_distacne[y * B + x] = INF; } __syncthreads(); if(blockIdx.x == Round) { return; } if(blockIdx.y == 0){ // pivot row idx_x = blockIdx.x * B + x; idx_y = p_y; }else{ // pivot col idx_x = p_x; idx_y = blockIdx.x * B + y; } if(idx_y >= V || idx_x >= V) { return; } if(idx_y < V && idx_x < V){ shared_distance[y * B + x] = distance[idx_y * V + idx_x]; }else{ shared_distance[y * B + x] = INF; } __syncthreads(); // calculate for each row/col if(blockIdx.y == 0){ // #pragma unroll for(int k = 0; k < B; k++){ int temp = shared_p_distacne[y * B + k] + shared_distance[k * B + x]; if(shared_distance[y * B + x] > temp){ shared_distance[y * B + x] = temp; } __syncthreads(); } }else{ // #pragma unroll for(int k = 0; k < B; k++){ int temp = shared_distance[y * B + k] + shared_p_distacne[k * B + x]; if(shared_distance[y * B + x] > temp){ shared_distance[y * B + x] = temp; } __syncthreads(); } } if(idx_y < V && idx_x < V){ distance[idx_y * V + idx_x] = shared_distance[y * B + x]; } } __global__ void PHASE_THREE(int Round, int B, int* distance, int V){ // grid size (Round*Round) if(blockIdx.x == Round || blockIdx.y == Round){ return; } extern __shared__ __align__(32) int shared[]; int* shared_row = shared; int* shared_col = shared + B*B; int x = threadIdx.x; int y = threadIdx.y; int idx_x = blockIdx.x * B + x; int idx_y = blockIdx.y * B + y; int row_i = Round * B + y; int row_j = idx_x; int col_i = idx_y; int col_j = Round * B + x; if(row_i < V && row_j < V){ shared_row[y * B + x] = distance[row_i * V + row_j]; }else{ shared_row[y * B + x] = INF; } if(col_i < V && col_j < V){ shared_col[y * B + x] = distance[col_i * V + col_j]; }else{ shared_col[y * B + x] = INF; } __syncthreads(); if(idx_y >= V || idx_x >= V) { return; } int temp = distance[idx_y * V + idx_x]; // #pragma unroll // #pragma GCC ivdep for(int k = 0; k < B; k++){ if(temp > shared_col[y * B + k] + shared_row[k * B + x]){ temp = shared_col[y * B + k] + shared_row[k * B + x]; } } distance[idx_y * V + idx_x] = temp; } int main(int argc, char* argv[]) { input(argv[1]); int B = 32; block_FW(B); output(argv[2]); return 0; } void input(char* infile) { FILE* file = fopen(infile, "rb"); fread(&n, sizeof(int), 1, file); fread(&m, sizeof(int), 1, file); dist = (int*)malloc(n * n * sizeof(int)); Result = (int*)malloc(n * n * sizeof(int)); for (int i = 0; i < n; ++i) { for (int j = 0; j < n; ++j) { if (i == j) { dist[i * n + j] = 0; } else { dist[i * n + j] = INF; } } } int pair[3]; for (int i = 0; i < m; ++i) { fread(pair, sizeof(int), 3, file); dist[pair[0] * n + pair[1]] = pair[2]; } fclose(file); } void output(char* outFileName) { FILE* outfile = fopen(outFileName, "w"); for (int i = 0; i < n; ++i) { for (int j = 0; j < n; ++j) { if(Result[i * n + j] >= INF){ Result[i * n + j] = INF; } } fwrite(&Result[i * n], sizeof(int), n, outfile); } fclose(outfile); } int ceil(int a, int b) { return (a + b - 1) / b; } void block_FW(int B) { int round = ceil(n, B); int *device_distance; int deviceID; cudaDeviceProp prop; dim3 grid_p1(1, 1); dim3 grid_p2(round, 2); dim3 grid_p3(round, round); dim3 blk(B, B); cudaGetDevice(&deviceID); cudaGetDeviceProperties(&prop, deviceID); if (!prop.deviceOverlap){ printf("!prop.deviceOverlap\n"); } cudaSetDevice(0); cudaMalloc(&device_distance, n * n * sizeof(int)); cudaMemcpy(device_distance, dist, n * n * sizeof(int), cudaMemcpyHostToDevice); for (int i = 0; i < round; i++) { PHASE_ONE<<<grid_p1, blk, B * B * sizeof(int)>>>(i, B, device_distance, n); PHASE_TWO<<<grid_p2, blk, B * B * sizeof(int) * 2>>>(i, B, device_distance, n); PHASE_THREE<<<grid_p3, blk, B * B * sizeof(int) * 2>>>(i, B, device_distance, n); } cudaMemcpy(Result, device_distance, n * n * sizeof(int), cudaMemcpyDeviceToHost); cudaFree(device_distance); } // void cal( // int B, int Round, int block_start_x, int block_start_y, int block_width, int block_height) { // int block_end_x = block_start_x + block_height; // int block_end_y = block_start_y + block_width; // for (int b_i = block_start_x; b_i < block_end_x; ++b_i) { // for (int b_j = block_start_y; b_j < block_end_y; ++b_j) { // // To calculate B*B elements in the block (b_i, b_j) // // For each block, it need to compute B times // for (int k = Round * B; k < (Round + 1) * B && k < n; ++k) { // // To calculate original index of elements in the block (b_i, b_j) // // For instance, original index of (0,0) in block (1,2) is (2,5) for V=6,B=2 // int block_internal_start_x = b_i * B; // int block_internal_end_x = (b_i + 1) * B; // int block_internal_start_y = b_j * B; // int block_internal_end_y = (b_j + 1) * B; // if (block_internal_end_x > n) block_internal_end_x = n; // if (block_internal_end_y > n) block_internal_end_y = n; // for (int i = block_internal_start_x; i < block_internal_end_x; ++i) { // for (int j = block_internal_start_y; j < block_internal_end_y; ++j) { // if (Dist[i][k] + Dist[k][j] < Dist[i][j]) { // Dist[i][j] = Dist[i][k] + Dist[k][j]; // } // } // } // } // } // } // }
698
#include <stdio.h> #include <cuda_runtime.h> __global__ void kernel(void) { printf("GPU: Hello world\n"); } int main(int argc, char* argv[]) { kernel<<<2, 4>>>(); cudaDeviceSynchronize(); return 0; }
699
#include "user.cuh" using namespace std; int main(int argc, char *argv[]) { // Preapre file I/O string fileName; if (argc == 1) fileName = "maxFloatNum.csv"; else fileName = argv[1]; CSV_Data cd(fileName, true); // Generate Vector printf("[INFO] Generating vector...\n"); float minVal = INFINITY, maxVal = -INFINITY; float *p_host_vector = (float *)malloc(VECTOR_SIZE * sizeof(float)); generateVector(p_host_vector, VECTOR_SIZE, &minVal, &maxVal); printf("[INFO] Max value:%.4f, Min Value:%.4f\n", maxVal, minVal); // Copy vector from host to device float *p_host_max_val = (float *)malloc(sizeof(float)), *p_device_vector; float *p_host_device_max_val = (float *)malloc(sizeof(float)), *p_device_max_val; cudaMalloc((void **)&p_device_vector, VECTOR_SIZE * sizeof(float)); cudaMalloc((void **)&p_device_max_val, sizeof(float)); cudaMemcpy(p_device_vector, p_host_vector, VECTOR_SIZE * sizeof(float), cudaMemcpyHostToDevice); printf("Starting Benchmark (ITERATION: %d)\n", ITERATION); // Host Code struct timeval startTime, endTime; float host_exec_time = 0.0; for (int iter = 0; iter < ITERATION; iter++) { gettimeofday(&startTime, NULL); host_maxValueVector(p_host_vector, VECTOR_SIZE, p_host_max_val); gettimeofday(&endTime, NULL); float tmp_exec_time = (endTime.tv_sec - startTime.tv_sec) * 1000. + (endTime.tv_usec - startTime.tv_usec) / 1000.; host_exec_time += tmp_exec_time; } host_exec_time /= ITERATION; cd.AddHostData(host_exec_time, p_host_max_val, sizeof(float)); // Device Code for (int numThreadsPerBlk = NUM_THREADS_BASE; numThreadsPerBlk <= NUM_THREADS_MAX; numThreadsPerBlk*=2) { for (int numBlks = NUM_THREAD_BLKS_FROM; numBlks <= NUM_THREAD_BLKS_TO; numBlks*=2) { int numElements = VECTOR_SIZE; int numOps = numElements > (numBlks * numThreadsPerBlk) ? numElements / (numBlks * numThreadsPerBlk) + (numElements % (numBlks * numThreadsPerBlk) ? 1 : 0) : 1; int *p_device_block_cnt, host_block_cnt; cudaMalloc((void **)&p_device_block_cnt, sizeof(int)); dim3 gridSize(numBlks); dim3 blockSize(numThreadsPerBlk); float avg_exec_time = 0.0; for (int iter = 0; iter < ITERATION; iter++) { cudaEvent_t cuda_start, cuda_end; float tmp_exec_time; cudaEventCreate(&cuda_start); cudaEventCreate(&cuda_end); // Prepare data cudaMemset(p_device_max_val, 0, sizeof(float)); cudaMemset(p_device_block_cnt, 0, sizeof(int)); // Execute Kernel cudaEventRecord(cuda_start, 0); device_maxValueVector<<<gridSize, blockSize, numThreadsPerBlk*sizeof(float)>>> (p_device_vector, p_device_max_val, VECTOR_SIZE, p_device_block_cnt, numOps); cudaDeviceSynchronize(); cudaEventRecord(cuda_end, 0); cudaEventSynchronize(cuda_end); cudaEventElapsedTime(&tmp_exec_time, cuda_start, cuda_end); cudaEventDestroy(cuda_start); cudaEventDestroy(cuda_end); avg_exec_time += tmp_exec_time; } avg_exec_time /= ITERATION; // Get average of execution time. // Print Results cudaMemcpy(&host_block_cnt, p_device_block_cnt, sizeof(int), cudaMemcpyDeviceToHost); cudaMemcpy(p_host_device_max_val, p_device_max_val, sizeof(float), cudaMemcpyDeviceToHost); cd.AddDeviceData(numThreadsPerBlk, numBlks, avg_exec_time, p_host_device_max_val, sizeof(float)); cudaFree(p_device_block_cnt); } } // Deallocation cudaFree(p_device_vector); cudaFree(p_device_max_val); free(p_host_vector); free(p_host_max_val); free(p_host_device_max_val); }
700
#include <iostream> __global__ void vecAddKernel(float *A, float *B, float *C, int n) { int i = blockDim.x * blockIdx.x + threadIdx.x; if(i < n) { C[i] = A[i] + B[i]; } } void vecAdd(float* h_A, float* h_B, float* h_C, int n) { int size = n * sizeof(float); float *d_A, *d_B, *d_C; cudaMalloc((void **) &d_A, size); cudaMalloc((void **) &d_B, size); cudaMalloc((void **) &d_C, size); cudaError_t err = cudaMemcpy(d_A, h_A, size, cudaMemcpyHostToDevice); if (err != cudaSuccess) { printf("%s in %s at line %d\n", cudaGetErrorString(err), __FILE__,__LINE__); exit(EXIT_FAILURE); } cudaMemcpy(d_B, h_B, size, cudaMemcpyHostToDevice); vecAddKernel<<<ceil(n/256.0), 256>>>(d_A, d_B, d_C, n); cudaMemcpy(h_C, d_C, size, cudaMemcpyDeviceToHost); cudaFree(d_A); cudaFree(d_B); cudaFree(d_C); } int main() { float A[] = {1, 2, 3, 4}; float B[] = {1, 2, 3, 4}; float C[4] = {}; vecAdd(A, B, C, 4); for(float i : A) { std::cout << i << " "; } std::cout << "\n"; for(float i : B) { std::cout << i << " "; } std::cout << "\n"; for(float i : C) { std::cout << i << " "; } std::cout << "\n"; }