serial_no
int64
1
24.2k
cuda_source
stringlengths
11
9.01M
7,701
struct ProgramTestClass { __device__ ProgramTestClass() { } int A; int B; int C; }; // CudaTest.Program extern "C" __global__ void doTheThing( ProgramTestClass* tests, int testsLen0); // CudaTest.Program extern "C" __global__ void addArrays( int* a, int aLen0, int* b, int bLen0, int* c, int cLen0); // CudaTest.Program extern "C" __global__ void findPrimes( int* toCheck, int toCheckLen0, int* results, int resultsLen0); // CudaTest.Program __device__ int isPrime(int a); // CudaTest.Program extern "C" __global__ void doTheThing( ProgramTestClass* tests, int testsLen0) { int x = blockIdx.x; if (x < 1000000) { ProgramTestClass testClass = tests[(x)]; testClass.C = testClass.A + testClass.B; } } // CudaTest.Program extern "C" __global__ void addArrays( int* a, int aLen0, int* b, int bLen0, int* c, int cLen0) { int x = blockIdx.x; if (x < 1000000) { c[(x)] = a[(x)] + b[(x)]; } } // CudaTest.Program extern "C" __global__ void findPrimes( int* toCheck, int toCheckLen0, int* results, int resultsLen0) { int x = blockIdx.x; if (x < 1000000) { results[(x)] = isPrime(toCheck[(x)]); } } // CudaTest.Program __device__ int isPrime(int a) { int result; if (a == 1 || a == 2) { result = 1; } else { int num = a % 2; if (num == 0) { result = 0; } else { int num2 = a / 2; for (int num3 = 3; num3 <= num2; num3++) { if (a % num3 == 0) { result = 0; return result; } } result = 1; } } return result; }
7,702
#include <iostream> #include <cstdlib> //matrix A[i][j] template<int N> __device__ int l(int a, int b){ return a*N + b; } template<int N> __global__ void multiply(double *A, double *B, double *C){ // C[i][k] = \sum_{j} A[i][j] * B[j][k] int i = blockIdx.y*blockDim.y+threadIdx.y; int k = blockIdx.x*blockDim.x+threadIdx.x; if(i<N && k<N){ double result = 0; for(int j=0; j<N; j++){ result += A[l<N>(i, j)] * B[l<N>(j, k)]; } C[l<N>(i, k)] = result; } } void random(double* x, int size){ for(int i=0; i<size; i++){ x[i] = (double)(rand()%100)/100; } } template<int N> void dispMat(const double *A){ for(int i=0; i<N; i++){ for(int j=0; j<N; j++){ std::cout << A[i*N+j] << ","; } std::cout << std::endl; } } void checkError(){ std::cout << cudaGetErrorString(cudaGetLastError()) << std::endl; } int main(void){ const int N = 1000; const int THREADS_PER_BLOCK = 16; double *A; double *B; double *C; double *d_A; double *d_B; double *d_C; int memsize = N*N*sizeof(double); A = (double *)malloc(memsize); B = (double *)malloc(memsize); C = (double *)malloc(memsize); random(A, N*N); random(B, N*N); std::cout << "A=" << std::endl; dispMat<N>(A); std::cout << "B=" << std::endl; dispMat<N>(B); cudaMalloc((void**)&d_A, memsize); checkError(); cudaMalloc((void**)&d_B, memsize); checkError(); cudaMalloc((void**)&d_C, memsize); checkError(); cudaMemcpy(d_A, A, memsize, cudaMemcpyHostToDevice); checkError(); cudaMemcpy(d_B, B, memsize, cudaMemcpyHostToDevice); checkError(); dim3 blocks((N/THREADS_PER_BLOCK)+1, (N/THREADS_PER_BLOCK)+1, 1); dim3 thPerBlock(THREADS_PER_BLOCK, THREADS_PER_BLOCK, 1); multiply<N><<<blocks,thPerBlock>>>(d_A, d_B, d_C); checkError(); cudaMemcpy(C, d_C, memsize, cudaMemcpyDeviceToHost); checkError(); std::cout << "C=" << C[10*N+10] << std::endl; dispMat<N>(C); cudaFree(d_A); cudaFree(d_B); cudaFree(d_C); free(A); free(B); free(C); }
7,703
#include <stdio.h> #define cudaCheckErrors(msg) \ do { \ cudaError_t __err = cudaGetLastError(); \ if (__err != cudaSuccess) { \ fprintf(stderr, "Fatal error: %s (%s at %s:%d)\n", \ msg, cudaGetErrorString(__err), \ __FILE__, __LINE__); \ fprintf(stderr, "*** FAILED - ABORTING\n"); \ exit(1); \ } \ } while (0) const int N=10; __global__ void add(int *a, int *b, int *c) { int tid = threadIdx.x; c[tid] = a[tid] + b[tid]; } int main(){ int a[N], b[N], c[N]; int *dev_a, *dev_b, *dev_c; cudaMalloc( (void**)&dev_a, N * sizeof(int) ); cudaMalloc( (void**)&dev_b, N * sizeof(int) ); cudaMalloc( (void**)&dev_c, N * sizeof(int) ); cudaCheckErrors("cudamalloc fail"); for (int i=0; i<N; i++) { a[i] = -i; b[i] = i * i; } cudaMemcpy ( dev_a, a, N * sizeof(int), cudaMemcpyHostToDevice ); cudaMemcpy ( dev_b, b, N * sizeof(int), cudaMemcpyHostToDevice ); cudaCheckErrors("cuda memcpy fail"); add<<<1,N>>>(dev_a, dev_b, dev_c); cudaMemcpy(c, dev_c, N * sizeof(int),cudaMemcpyDeviceToHost ); cudaCheckErrors("cudamemcpy or cuda kernel fail"); for (int i=0; i<N; i++) { printf("%d + %d = %d\n", a[i],b[i],c[i]); } cudaFree (dev_a); cudaFree (dev_b); cudaFree (dev_c); return 0; }
7,704
#include "includes.h" __global__ void imagePaddingKernel(float3 *ptr, float3 *dst, int width, int height, int top, int bottom, int left, int right) { int x = threadIdx.x + blockIdx.x * blockDim.x; int y = threadIdx.y + blockIdx.y * blockDim.y; if(x < left || x >= (width - right) || y < top || y > (height - bottom)) { return; } float3 color = ptr[(y - top) * (width - top - right) + (x - left)]; dst[y * width + x] = color; }
7,705
#include <thrust/host_vector.h> #include <thrust/device_vector.h> #include <thrust/reduce.h> #include <iostream> int main(void) { int N = 100000; thrust::host_vector<int> h_data(N); thrust::device_vector<int> d_data(N); //Method 1 (one cudaMemcpy per element, slow) // for (int i = 0; i < N; i++) // d_data[i] = i; //Method 2 (one cudaMemcpy per entire array, faster) for (int i = 0; i < N; i++) h_data[i] = i; thrust::copy(h_data.begin(), h_data.end(), d_data.begin()); return 0; }
7,706
#include <stdio.h> #include <ctype.h> #include <stdlib.h> #include <unistd.h> #include <cstdlib> //Data parameters #define SHOW_ELAPSED_TIME 1 //DSP parameters #define SAMP_FREQ 1e6 #define FREQ_SHIFT -350000 //Coefficients for FIR #define DECIMATION_RATE 2 #define FIR_SIZE 64 const int fir_size_in_bytes = FIR_SIZE * sizeof(float); __constant__ float fir_coef [FIR_SIZE]; float cpu_fir_coef[FIR_SIZE] = {0.0}; // Declare arrays dynamically float *cpu_I_in_buffer; float *cpu_I_result_buffer; float *cpu_Q_in_buffer; float *cpu_Q_result_buffer; //Default values, can be overwritten by providing command line arguments unsigned int array_size; unsigned int array_size_in_bytes; unsigned int num_threads = FIR_SIZE; unsigned int num_blocks = 4096; //Function to copy data into shared memory. Includes thread sync __device__ void copy_data_to_shared(float * src, float * dst, const unsigned int tid) { // Copy data dst[tid] = src[tid]; // Sync threads before accessing shared memory __syncthreads(); } //Function to copy data out of shared memory. Includes thread sync __device__ void copy_data_from_shared(float * src, float * dst, const unsigned int tid) { // Sync threads before accessing shared memory __syncthreads(); // Copy data dst[tid] = src[tid]; } // Custom complex multiplication kernel __device__ void cMult(const float Ai, const float Aq, const float Bi, const float Bq, float* Ri, float* Rq) { *Ri = Ai*Bi - Aq*Bq; *Rq = Ai*Bq + Aq*Bi; } //Method to quickly sum an array and put the result at the begining of the array //http://developer.download.nvidia.com/compute/cuda/1.1-Beta/x86_website/projects/reduction/doc/reduction.pdf __device__ void sum_array(float * sdata, const unsigned int blockSize, const unsigned int tid) { for (unsigned int s=blockSize/2; s>0; s>>=1) { if(tid < s) { sdata[tid] += sdata[tid + s]; } __syncthreads(); } } // Perform a frequency shift via complex multiply // Parameters: // I_in, Q_in, I_out, Q_out: 4 data buffers, all must be the same length // n0: Used to calculate the phase of the first point of the mixing signal // freq_shift: Frequency to shift in Hz // Fs: sample frequency in Hz __global__ void freq_shift(float * I_in, float * Q_in, float * I_out, float * Q_out, const unsigned int n0, const float freq_shift, const float Fs) { //Who am I? //const unsigned int thread_id = threadIdx.x; //const unsigned int block_id = blockIdx.x; const unsigned int global_index = (blockIdx.x * blockDim.x) + threadIdx.x; float I_shift; float Q_shift; float theta_nopi = 2.0*freq_shift*(n0 + global_index)/Fs; sincospif(theta_nopi, &Q_shift, &I_shift); cMult(I_in[global_index], Q_in[global_index], I_shift, Q_shift, &I_out[global_index], &Q_out[global_index]); } // FIR based decimation __global__ void decimate(float * input_buffer, float * output_buffer, const unsigned int decimation_factor) { __shared__ float conv[FIR_SIZE]; //Who am I? const unsigned int thread_id = threadIdx.x; const unsigned int block_id = blockIdx.x; //Perform the convolution as a copy from global (num samples) to shared (FIR width) float sample = 0.0; int sample_index = block_id - thread_id; if(sample_index >= 0) sample = input_buffer[sample_index]; conv[thread_id] = sample*fir_coef[thread_id]; __syncthreads(); //Sum results vector using loop unrolling and shared memory sum_array(conv, blockDim.x, thread_id); //Decimate if(thread_id == 0) { if((block_id % decimation_factor) == 0) output_buffer[block_id / decimation_factor] = conv[0]; } } // main_sub0 : Method to copy an input buffer into cuda and copy the results out void main_sub0() { // Declare pointers for GPU based params float *gpu_I_in_buffer; float *gpu_I_mixed_buffer; float *gpu_I_result_buffer; float *gpu_Q_in_buffer; float *gpu_Q_mixed_buffer; float *gpu_Q_result_buffer; // Allocate memory in the GPU cudaMalloc((void **)&gpu_I_in_buffer, array_size_in_bytes); cudaMalloc((void **)&gpu_I_mixed_buffer, array_size_in_bytes); cudaMalloc((void **)&gpu_I_result_buffer, array_size_in_bytes); cudaMalloc((void **)&gpu_Q_in_buffer, array_size_in_bytes); cudaMalloc((void **)&gpu_Q_mixed_buffer, array_size_in_bytes); cudaMalloc((void **)&gpu_Q_result_buffer, array_size_in_bytes); //Copy Constant data cudaMemcpyToSymbol(fir_coef, &cpu_fir_coef, fir_size_in_bytes); #if SHOW_ELAPSED_TIME float ms; // Setup Start and Stop event cudaEvent_t startEvent, stopEvent; cudaEventCreate(&startEvent); cudaEventCreate(&stopEvent); // Start timer cudaEventRecord(startEvent, 0); #endif // Copy data from CPU to GPU cudaMemcpy(gpu_I_in_buffer, cpu_I_in_buffer, array_size_in_bytes, cudaMemcpyHostToDevice); cudaMemcpy(gpu_Q_in_buffer, cpu_Q_in_buffer, array_size_in_bytes, cudaMemcpyHostToDevice); //Run kernels freq_shift<<<num_blocks/32, 32>>>(gpu_I_in_buffer, gpu_Q_in_buffer, gpu_I_mixed_buffer, gpu_Q_mixed_buffer, 0, FREQ_SHIFT, SAMP_FREQ); decimate<<<num_blocks, num_threads>>>(gpu_I_mixed_buffer, gpu_I_result_buffer, DECIMATION_RATE); decimate<<<num_blocks, num_threads>>>(gpu_Q_mixed_buffer, gpu_Q_result_buffer, DECIMATION_RATE); // Copy results from GPU to CPU cudaMemcpy(cpu_I_result_buffer, gpu_I_result_buffer, array_size_in_bytes/DECIMATION_RATE, cudaMemcpyDeviceToHost); cudaMemcpy(cpu_Q_result_buffer, gpu_Q_result_buffer, array_size_in_bytes/DECIMATION_RATE, cudaMemcpyDeviceToHost); #if SHOW_ELAPSED_TIME // Stop timer cudaEventRecord(stopEvent, 0); cudaEventSynchronize(stopEvent); cudaEventElapsedTime(&ms, startEvent, stopEvent); printf("Elapsed Time: %f ms\n", ms); // Destroy timer cudaEventDestroy(startEvent); cudaEventDestroy(stopEvent); #endif //Destory streams //cudaStreamDestroy(stream1); //cudaStreamDestroy(stream2); // Free the arrays on the GPU cudaFree(gpu_I_in_buffer); cudaFree(gpu_I_mixed_buffer); cudaFree(gpu_I_result_buffer); cudaFree(gpu_Q_in_buffer); cudaFree(gpu_Q_mixed_buffer); cudaFree(gpu_Q_result_buffer); } //main : parse command line arguments and run GPU code int main(int argc, char *argv[]) { // Argument parsing using getopt // http://www.gnu.org/software/libc/manual/html_node/Example-of-Getopt.html#Example-of-Getopt // -b <int> sets the number of GPU blocks // -t <int> sets the nubmer of GPU threads // -v Sets verbose flag - shows math results int c; bool showMathResults = false; while ((c = getopt (argc, argv, "b:t:v")) != -1) switch (c) { case 'b': num_blocks = atoi(optarg); break; case 't': num_threads = atoi(optarg); break; case 'v': showMathResults = true; break; default: printf("USAGE:\n-b <int> GPU blocks\n-t <int> GPU threads (each block)\n-v Verbose\n"); return EXIT_SUCCESS; } printf("Blocks: %d\nThreads: %d\n", num_blocks, num_threads); // Calculate buffer size array_size = num_blocks; array_size_in_bytes = sizeof(float) * (array_size); // Allocate memory on the CPU cpu_I_in_buffer = new float[array_size]; cpu_I_result_buffer = new float[array_size]; cpu_Q_in_buffer = new float[array_size]; cpu_Q_result_buffer = new float[array_size]; //Load fir FILE * iFile; char fileName[100]; sprintf(fileName, "fir_dec_%d_taps_%d.txt", DECIMATION_RATE, FIR_SIZE); iFile = fopen(fileName, "r"); for(unsigned int i=0; i<FIR_SIZE; i++) fscanf(iFile, "%f\r\n", &cpu_fir_coef[i]); fclose(iFile); // Generate data to be processed float I, Q; iFile = fopen("inputIQ.txt", "r+"); for(unsigned int i=0; i<array_size; i++) { fscanf(iFile, "%f,%f\r\n", &I, &Q); cpu_I_in_buffer[i] = I; cpu_Q_in_buffer[i] = Q; } fclose(iFile); // Run main_sub0(); // Output results if(showMathResults) { for(unsigned int i = 0; i < array_size/DECIMATION_RATE; i++) { printf("%.5f, %.5f\n", cpu_I_result_buffer[i], cpu_Q_result_buffer[i]); } } printf("\n"); FILE * oFile; oFile = fopen("outputIQ.txt", "w+"); for(unsigned int i=0; i<array_size/DECIMATION_RATE; i++) { fprintf(oFile, "%f,%f\r\n", cpu_I_result_buffer[i], cpu_Q_result_buffer[i]); } fclose(oFile); // Done return EXIT_SUCCESS; }
7,707
#include <stdlib.h> #include <iostream> #include "test_hashmap.cuh" #include "test_polyaurn.cuh" #include "test_spalias.cuh" #include "test_topics.cuh" namespace gpulda_test { void run_tests() { std::cout << "running tests" << std::endl; std::cout << "testing hashmap" << std::endl; test_hashmap(); std::cout << "testing polya_urn_init" << std::endl; test_polya_urn_init(); std::cout << "testing polya_urn_sample" << std::endl; test_polya_urn_sample(); std::cout << "testing polya_urn_transpose" << std::endl; test_polya_urn_transpose(); std::cout << "testing polya_urn_reset" << std::endl; test_polya_urn_reset(); std::cout << "testing polya_urn_colsums" << std::endl; test_polya_urn_colsums(); std::cout << "testing build_alias" << std::endl; test_build_alias(); std::cout << "testing compute_d_idx" << std::endl; test_compute_d_idx(); std::cout << "testing sample_topics_functions" << std::endl; test_sample_topics_functions(); std::cout << "testing sample_topics" << std::endl; test_sample_topics(); std::cout << "tests completed" << std::endl; } }
7,708
#include <cuda.h> #include <cuda_runtime.h> #include <iostream> using namespace std; //DEVICE __global__ void kernelVector_x_constant( float* arr, int n, int k ) { //Obtengo el indice del hilo fisico int idx = blockIdx.x * blockDim.x + threadIdx.x; //Mientras el hilo sea valido para la operaci�n if( idx<n ) { //Multiplico el elemento por la constante arr[ idx ] = arr[ idx ] * k; } } //HOST int main() { int size = 1000000; //Separo memoria en la RAM del HOST float* arr = new float[size]; float* arr_DEVICE = NULL; //Inicializo el arreglo en el HOST for( int index = 0; index<size ; index++ ) { arr[index] = index; } //Separo memoria en la RAM del DEVICE ( la misma cantidad de bytes que en el HOST ) cudaMalloc((void**)&arr_DEVICE, size * sizeof(float)); //Copio el bloque de memoria del HOST al DEVICE cudaMemcpy( arr_DEVICE, arr, size * sizeof(float), cudaMemcpyHostToDevice); ///////////////////////// EJECUTO EL KERNEL DE CUDA //////////////////////////// //////// 512 Hilos //////// ceil(1000000/512) Bloques kernelVector_x_constant<<< ceil(size/512.0), 512 >>>( arr_DEVICE, size, 65 ); //Fuerzo una llamada Sincrona cudaThreadSynchronize(); //Copio mis datos ya procesados a la RAM del HOST cudaMemcpy( arr, arr_DEVICE, size * sizeof(float), cudaMemcpyDeviceToHost); //Con una impresi�n de los primeros 100 visualizo el resultado for( int index = 0; index<100 ; index++ ) { cout<<arr[index]<<endl; } //Libero memoria en la RAM del DEVICE cudaFree( arr_DEVICE ); //Libero memoria en la RAM del HOST delete[] arr; cin.get(); }
7,709
#include <stdio.h> #include <string.h> #include <stdlib.h> #include <math.h> #include <unistd.h> #include <time.h> #include <cuda_runtime.h> #define PI 3.14159265 const int val = 180.0 / PI; const int thread = 256; const float fix = 1 / 60 * PI / 180; const int bins = 720; using namespace std; __global__ void angles(volatile float *a0, volatile float *b0, volatile float *a1, volatile float *b1, int xind, int yind, int max_x, int max_y, volatile int *histi) { int idx = blockIdx.x * blockDim.x + threadIdx.x; float angle; idx += xind; __shared__ unsigned int shared[bins]; // Always clear the first iteration if (threadIdx.x == 0) { for (int i = 0; i < bins; i++) { shared[i] = 0; } } __syncthreads(); // Check for (int i = yind; i < yind + max_y; i++) { angle = acosf((sin(b0[idx]) * sin(b1[i]) + cos(b0[idx]) * cos(b1[i]) * cos(a0[idx] * -a1[i])) * 1 / val); shared[int(angle)]++; // TODO: Make atomicAddwork :D // atomicAdd(&shared[int(angle)], 1); } __syncthreads(); if (threadIdx.x == 0) { for (int i = 0; i < bins; i++) { histi[i] = shared[i]; } } } void read_the_files() { // Reading files 1 and 2 FILE *real_g; FILE *synthetic_g; int galaxies_r, galaxies_s; float *a0, *a1, *b0, *b1; real_g = fopen("data_100k_arcmin.txt", "r"); synthetic_g = fopen("flat_100k_arcmin.txt", "r"); fscanf(real_g, "%d", &galaxies_r); fscanf(synthetic_g, "%d", &galaxies_s); a0 = (float*) malloc(galaxies_r * sizeof(float)); b0 = (float*) malloc(galaxies_r * sizeof(float)); a1 = (float*) malloc(galaxies_s * sizeof(float)); b1 = (float*) malloc(galaxies_s * sizeof(float)); for (int i = 0; i < galaxies_r; i++) { fscanf(real_g, "%e %e", &a0[i], &b0[i]); fscanf(synthetic_g, "%e %e", &a1[i], &b1[i]); a0[i] *= fix; b0[i] *= fix; a1[i] *= fix; b1[i] *= fix; } fclose(real_g); fclose(synthetic_g); dim3 grid, block; grid.x = 1024; grid.y = 1024; block.x = 1; float *aa1, *bb1, *aa0, *bb0; cudaMalloc((void **) &aa0, galaxies_r * sizeof(float)); cudaMalloc((void **) &bb0, galaxies_r * sizeof(float)); cudaMalloc((void **) &aa1, galaxies_s * sizeof(float)); cudaMalloc((void **) &bb1, galaxies_s * sizeof(float)); // Do we have enough memory? // Initialize array to all 0's cudaMemset(aa0, 0, galaxies_r * sizeof(float)); cudaMemset(bb0, 0, galaxies_r * sizeof(float)); cudaMemset(aa1, 0, galaxies_s * sizeof(float)); cudaMemset(bb1, 0, galaxies_s * sizeof(float)); cudaMemcpy(aa0, a0, galaxies_r * sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(bb0, b0, galaxies_r * sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(aa1, a1, galaxies_s * sizeof(float), cudaMemcpyHostToDevice); cudaMemcpy(bb1, b1, galaxies_s * sizeof(float), cudaMemcpyHostToDevice); int x, y; // Preparing the histogram array int *hist, *histi, *tmp; int size_h_bytes = 720 * sizeof(int); hist = (int*)malloc(size_h_bytes); memset(hist, 0, size_h_bytes); cudaMalloc((void **) &tmp, (size_h_bytes)); cudaMemset(tmp, 0, size_h_bytes); unsigned long *hist_array; int hist_array_size = 720 * sizeof(unsigned long); hist_array = (unsigned long*)malloc(hist_array_size); memset(hist_array, 0, hist_array_size); cudaMemset(tmp, 0, size_h_bytes); for (int i = 0; i < 9; i++) { y = i * 512 * 512; for (int j = 0; j < 9; j++) { x = 512 * 512 * i; angles<<<grid,block>>>(aa0, bb0, aa1, bb1, x, y, 512, 512, tmp); cudaMemcpy(hist, tmp, size_h_bytes, cudaMemcpyDeviceToHost); for(int i = 0; i < 720; i++) { hist_array[i] += hist[i]; } } } for(int i = 0; i < 720; i++) { printf("%d ", hist_array[i]); } free(a1); free(b1); free(a0); free(b0); cudaFree(aa1); cudaFree(aa0); cudaFree(bb0); cudaFree(bb1); cudaFree(tmp); } // Preparation for the kernel int main() { float alpha1 = 4646.98; float b1 = 3749.51; float a2 = 4644.35; float b2 = 3749.52; float theta1 = acos(sin(b1) * sin(b2) + cos(b1) * cos(b2) * cos(alpha1 - a2)); printf("%f\n", b1); printf("%f\n", theta1); clock_t start, end; double cpu_time_used; start = clock(); read_the_files(); // Do some calculations end = clock(); cpu_time_used = ((double) (end - start)) / CLOCKS_PER_SEC; printf("%f", cpu_time_used); return EXIT_SUCCESS; }
7,710
#include <iostream> #include <algorithm> using namespace std; __global__ void VectorsPairMaximums(size_t size, double *first, double *second, double *res) { size_t begin = (size_t) (blockDim.x * blockIdx.x + threadIdx.x); size_t offset = gridDim.x * blockDim.x; for (size_t i = begin; i < size; i += offset) { res[i] = max(first[i], second[i]); } } __host__ int main(void) { size_t size; cin >> size; double *first = new double[size]; double *second = new double[size]; double *res = new double[size]; for (size_t i = 0; i < size; i++) { cin >> first[i]; //first[i] = i; } for (size_t i = 0; i < size; i++) { cin >> second[i]; //second[i] = i; } double *cudaFirst; double *cudaSecond; double *cudaRes; cudaMalloc((void**) &cudaFirst, sizeof(double) * size); cudaMalloc((void**) &cudaSecond, sizeof(double) * size); cudaMalloc((void**) &cudaRes, sizeof(double) * size); cudaMemcpy(cudaFirst, first, sizeof(double) * size, cudaMemcpyHostToDevice); cudaMemcpy(cudaSecond, second, sizeof(double) * size, cudaMemcpyHostToDevice); VectorsPairMaximums<<<256, 256>>>(size, cudaFirst, cudaSecond, cudaRes); cudaEvent_t syncEvent; cudaEventCreate(&syncEvent); cudaEventRecord(syncEvent, 0); cudaEventSynchronize(syncEvent); cudaMemcpy(res, cudaRes, sizeof(double) * size, cudaMemcpyDeviceToHost); //double *testArr = new double[size]; //cudaMemcpy(testArr, cudaFirst, sizeof(double) * size, cudaMemcpyDeviceToHost); cudaEventDestroy(syncEvent); cudaFree(cudaFirst); cudaFree(cudaSecond); cudaFree(cudaRes); for (size_t i = 0; i < size; i++) { if (i > 0) { cout << " "; } cout << scientific << res[i]; } cout << endl; delete [] first; delete [] second; delete [] res; return 0; }
7,711
# include <math.h> __constant__ const float MAX_FLOAT32 = 3.4028e+038; __constant__ const float MIN_FLOAT32 = -3.4028e+038; __constant__ const float EPS_FLOAT32 = 2.22045e-016; __constant__ float MU_WATER = 0.037; __constant__ float MU_AIR = 0.00046; __global__ void backprojectPixel( const int h, const int w, float * dsts, const float * minv, const float * kinv, const int z_sign, const bool down, const float sid) { int blockId = blockIdx.x + blockIdx.y*gridDim.x; int localId = (threadIdx.y*blockDim.x) + threadIdx.x; int threadId = blockId * (blockDim.x*blockDim.y) + localId; __shared__ float sKinv[9]; __shared__ float sMinv[12]; if (localId < 9){ sKinv[localId] = kinv[localId]; } else if (localId < 21){ sMinv[localId - 9] = minv[localId - 9]; } __syncthreads(); int i = threadId / w; int j = threadId % w; if (down == true) { int temp = i; i = j; j = temp; } if (threadId < h*w) { float dotx = sid*z_sign*(sKinv[0]*i + sKinv[1]*j + sKinv[2]*1); float doty = sid*z_sign*(sKinv[3]*i + sKinv[4]*j + sKinv[5]*1); float dotz = sid*z_sign*(sKinv[6]*i + sKinv[7]*j + sKinv[8]*1); dsts[3*threadId + 0] = sMinv[0]*dotx + sMinv[1]*doty + sMinv[2]*dotz + sMinv[3]*1; dsts[3*threadId + 1] = sMinv[4]*dotx + sMinv[5]*doty + sMinv[6]*dotz + sMinv[7]*1; dsts[3*threadId + 2] = sMinv[8]*dotx + sMinv[9]*doty + sMinv[10]*dotz + sMinv[11]*1; } } __device__ int getIJK( const float & src, const float & dst, const float & minAxyz, const float & amin, const float & b, const float & s) { return floor((src + 0.5*(minAxyz + amin)*(dst-src) - b)/s); } __device__ void getAlphas( const float & b, const float & s, const float & p1, const float & p2, const int & n, float & amin, float & amax) { if (fabsf(p1 - p2) < EPS_FLOAT32) { amin = MIN_FLOAT32; amax = MAX_FLOAT32; } else{ amin = (b - p1)/(p2 - p1); amax = (b+(n-1)*s - p1)/(p2 - p1); if (amin > amax){ float temp = amin; amin = amax; amax = temp; } } } __device__ float getAx( const float & p1, const float & p2, const int & n, const float & b, const float & s, const float & axmin, const float & axmax, const float & amin, const float & amax) { float ax = 0; int imin = 1; int imax = n-2; if(fabsf(p2 - p1) < EPS_FLOAT32){ ax = MAX_FLOAT32; } else if (p1 < p2){ if(fabsf(amin - axmin) > EPS_FLOAT32){ imin = floor((p1 + amin*(p2-p1) - b)/s + 1); } ax = ((b + imin*s) - p1)/(p2 - p1); } else { if(fabsf(amin - axmin) > EPS_FLOAT32){ imax = ceil((p1 + amin*(p2 - p1) - b)/s - 1); } ax = ((b + imax*s) - p1)/(p2 - p1); } return ax; } __global__ void traceRay( const float * src, const float * dsts, float * raysums, const float * rho, const float * b, const float * sp, const int * n, const int h, const int w, const float threshold) { int blockId = blockIdx.x + blockIdx.y*gridDim.x; int localId = (threadIdx.y*blockDim.x) + threadIdx.x; int threadId = blockId * (blockDim.x*blockDim.y) + localId; __shared__ float sB[3]; __shared__ float sSp[3]; __shared__ float sSrc[3]; __shared__ int sN[3]; if(localId < 3){ sB[localId] = b[localId]; } else if(localId < 6){ sSp[localId - 3] = sp[localId - 3]; } else if(localId < 9){ sN[localId - 6] = n[localId - 6]; } else if(localId < 12){ sSrc[localId - 9] = src[localId - 9]; } __syncthreads(); if (threadId < h*w){ float3 dst = make_float3(dsts[3*threadId], dsts[3*threadId + 1], dsts[3*threadId + 2]); float axmin, axmax, aymin, aymax, azmin, azmax; getAlphas(sB[0], sSp[0], sSrc[0], dst.x, sN[0], axmin, axmax); getAlphas(sB[1], sSp[1], sSrc[1], dst.y, sN[1], aymin, aymax); getAlphas(sB[2], sSp[2], sSrc[2], dst.z, sN[2], azmin, azmax); float amin = fmaxf(axmin, fmaxf(aymin, azmin)); float amax = fminf(axmax, fminf(aymax, azmax)); if(amin > amax || (amin < 0)){ raysums[threadId] = 1; return; } else { float ax = getAx(sSrc[0], dst.x, sN[0], sB[0], sSp[0], axmin, axmax, amin, amax); float ay = getAx(sSrc[1], dst.y, sN[1], sB[1], sSp[1], aymin, aymax, amin, amax); float az = getAx(sSrc[2], dst.z, sN[2], sB[2], sSp[2], azmin, azmax, amin, amax); float dconv = sqrtf( (dst.x-sSrc[0])*(dst.x-sSrc[0]) + (dst.y-sSrc[1])*(dst.y-sSrc[1]) + (dst.z-sSrc[2])*(dst.z-sSrc[2])); float d12 = 0; float ac = amin; float minAxyz = fminf(ax, fminf(ay, az)); int i = getIJK(sSrc[0], dst.x, minAxyz, amin, sB[0], sSp[0]); int j = getIJK(sSrc[1], dst.y, minAxyz, amin, sB[1], sSp[1]); int k = getIJK(sSrc[2], dst.z, minAxyz, amin, sB[2], sSp[2]); while((-1 < i && i < (sN[0]-1)) && (-1 < j && j < (sN[1]-1)) && (-1 < k && k < (sN[2]-1))){ float hu = rho[k + j*(sN[2]-1) + i*(sN[2]-1)*(sN[1]-1)]; float mu = (hu*(MU_WATER-MU_AIR)/1000 + MU_WATER); if (hu < threshold){ mu = 0; } if(ax == minAxyz){ d12 = d12 + (ax - ac)*dconv*mu; i = (sSrc[0] < dst.x)?(i+1): (i-1); ac = ax; ax = ax + sSp[0]/fabsf(dst.x - sSrc[0]); } else if(ay == minAxyz){ d12 = d12 + (ay - ac)*dconv*mu; j = (sSrc[1] < dst.y)?(j+1): (j-1); ac = ay; ay = ay + sSp[1]/fabsf(dst.y - sSrc[1]); } else { d12 = d12 + (az - ac)*dconv*mu; k = (sSrc[2] < dst.z)?(k+1): (k-1); ac = az; az = az + sSp[2]/fabsf(dst.z - sSrc[2]); } minAxyz = fminf(ax, fminf(ay, az)); } raysums[threadId] = expf(-d12); } } }
7,712
#include <stdio.h> #include <stdlib.h> #include <math.h> #include <string> #include <fstream> #include <cmath> #include <string> #include <iostream> using namespace std; typedef struct { double x,y,z; } XYZ; typedef struct { XYZ p[8]; double val[8]; } GRIDCELL; typedef struct { XYZ p[3]; /* Vertices */ // XYZ c; /* Centroid */ // XYZ n[3]; /* Normal */ } TRIANGLE; #define ABS(x) (x < 0 ? -(x) : (x)) // Prototypes //__global__ //int PolygoniseCube(GRIDCELL,double,TRIANGLE *); //XYZ VertexInterp(double,XYZ,XYZ,double,double); /* #define NX 200 #define NY 160 #define NZ 160 */ #define NX 200//200 #define NY 160//160 #define NZ 160//160 //68 x 256 x 256 void fillMatrix(XYZ* a, int n) { int i; for (i = 0; i < n; ++i) { a[i].x = 3; a[i].y = 2; a[i].z = 5;//rand()%5; } } __global__ void matrixAdition(XYZ * b, XYZ *a,int n) { int ij = threadIdx.x + blockDim.x * blockIdx.x; if(ij<n) { b[ij].x = a[ij].x+2; b[ij].y = a[ij].y+3; b[ij].z = a[ij].z+0; //printf("da %d \n" , b[ij].x); } } void printMatrix(string s, XYZ *a , int tam){ cout<<s; for(int i=0;i<tam;i++) { if(a[i].x!=0 && a[i].y!=0 && a[i].z!=0) { cout<<a[i].x<<" "<<a[i].y<<" "<<a[i].z<<" "; cout<<endl; } } } void assingMem(int *** data) { int i,j; data = (int ***)malloc(NX*sizeof(short int **)); for (i=0;i<NX;i++) data[i] = (int **)malloc(NY*sizeof(short int *)); for (i=0;i<NX;i++) for (j=0;j<NY;j++) data[i][j] = (int *)malloc(NZ*sizeof(short int)); } void readFile(FILE *fptr, const char * namefile , int themin , int themax, int *** data) { int i,j,k,c; fprintf(stderr,"Load data ...\n"); if ((fptr = fopen(namefile,"rb")) == NULL) { fprintf(stderr,"Error al leer archivo\n"); exit(-1); } for (k=0;k<NZ;k++) { for (j=0;j<NY;j++) { for (i=0;i<NX;i++) { if ((c = fgetc(fptr)) == EOF) { fprintf(stderr,"Error en tamaño\n"); exit(-1); } data[i][j][k] = c; cout<<"leyendo :"<<c<<endl; if (c > themax) themax = c; if (c < themin) themin = c; } } } fclose(fptr); fprintf(stderr,"Rango del volumen: %d -> %d\n",themin,themax); } int constructCubes(GRIDCELL * vectGrids, int *** data) { int i,j,k; //fprintf(stderr,"Construyendo Cubos ...\n"); int cont=0; for (i=0;i<NX-1;i++) { //cout<<i<<endl; //if (i % (NX/10) == 0) //fprintf(stderr," Slice %d de %d\n",i,NX); for (j=0;j<NY-1;j++) { for (k=0;k<NZ-1;k++) { GRIDCELL grid; grid.p[0].x = i; grid.p[0].y = j; grid.p[0].z = k; grid.val[0] = data[i][j][k]; grid.p[1].x = i+1; grid.p[1].y = j; grid.p[1].z = k; grid.val[1] = data[i+1][j][k]; grid.p[2].x = i+1; grid.p[2].y = j+1; grid.p[2].z = k; grid.val[2] = data[i+1][j+1][k]; grid.p[3].x = i; grid.p[3].y = j+1; grid.p[3].z = k; grid.val[3] = data[i][j+1][k]; grid.p[4].x = i; grid.p[4].y = j; grid.p[4].z = k+1; grid.val[4] = data[i][j][k+1]; grid.p[5].x = i+1; grid.p[5].y = j; grid.p[5].z = k+1; grid.val[5] = data[i+1][j][k+1]; grid.p[6].x = i+1; grid.p[6].y = j+1; grid.p[6].z = k+1; grid.val[6] = data[i+1][j+1][k+1]; grid.p[7].x = i; grid.p[7].y = j+1; grid.p[7].z = k+1; grid.val[7] = data[i][j+1][k+1]; vectGrids[cont]=grid; cont++; } } } cout<<"numero de datos ingresados "<<cont<<endl; return cont; } __device__ XYZ VertexInterp(double isolevel,XYZ p1,XYZ p2,double valp1,double valp2) { double mu; XYZ p; if (ABS(isolevel-valp1) < 0.00001) return(p1); if (ABS(isolevel-valp2) < 0.00001) return(p2); if (ABS(valp1-valp2) < 0.00001) return(p1); mu = (isolevel - valp1) / (valp2 - valp1); p.x = p1.x + mu * (p2.x - p1.x); p.y = p1.y + mu * (p2.y - p1.y); p.z = p1.z + mu * (p2.z - p1.z); return p; } __device__ void copyXYZ(XYZ &a, XYZ &b) { a.x=b.x ; a.y=b.y ; a.z = b.z; } __device__ XYZ defect() { XYZ a; a.x=300 ; a.y=300 ; a.z = 300; return a; } /* __global__ void coyGRID(GRIDCELL * a, GRIDCELL * b, int x, int y, int z) { int i = threadIdx.x + blockDim.x * blockIdx.x; int j = threadIdx.y + blockDim.y * blockIdx.y; int k = threadIdx.z + blockDim.z * blockIdx.z; /*if(i<x && j<y && k<z) { a[ij].p = b[ij].p; a[ij].val = b[ij].val; } } */ /* __global__ void copyGRID1(GRIDCELL * a, GRIDCELL * b, int x, int y, int z) { int i = threadIdx.x + blockDim.x * blockIdx.x; int j = threadIdx.y + blockDim.y * blockIdx.y; int k = threadIdx.z + blockDim.z * blockIdx.z; if(i<x && j<y && k<z) { for(int w=0;w<8;w++) { a[i].p[w] = b[i].p[w]; a[i].val[w] = b[i].val[w]; } } }*/ /* __global__ void PolygoniseCube(XYZ * vertlist ,GRIDCELL * g ,double iso, int x ,int y , int z) */ __device__ int ntri=0; __global__ void PolygoniseCube(TRIANGLE * d_vectTriangles,GRIDCELL * g , int n) { int iso=80; int ind = threadIdx.x + blockDim.x * blockIdx.x; //printf("este si %d \n",ind); //return ; if(ind<n) { //printf("thread %d \n", g[i].p[7].x); int cubeindex; //int tamVert=12; XYZ vertlist[12]; int edgeTable[256]={ 0x0 , 0x109, 0x203, 0x30a, 0x406, 0x50f, 0x605, 0x70c, 0x80c, 0x905, 0xa0f, 0xb06, 0xc0a, 0xd03, 0xe09, 0xf00, 0x190, 0x99 , 0x393, 0x29a, 0x596, 0x49f, 0x795, 0x69c, 0x99c, 0x895, 0xb9f, 0xa96, 0xd9a, 0xc93, 0xf99, 0xe90, 0x230, 0x339, 0x33 , 0x13a, 0x636, 0x73f, 0x435, 0x53c, 0xa3c, 0xb35, 0x83f, 0x936, 0xe3a, 0xf33, 0xc39, 0xd30, 0x3a0, 0x2a9, 0x1a3, 0xaa , 0x7a6, 0x6af, 0x5a5, 0x4ac, 0xbac, 0xaa5, 0x9af, 0x8a6, 0xfaa, 0xea3, 0xda9, 0xca0, 0x460, 0x569, 0x663, 0x76a, 0x66 , 0x16f, 0x265, 0x36c, 0xc6c, 0xd65, 0xe6f, 0xf66, 0x86a, 0x963, 0xa69, 0xb60, 0x5f0, 0x4f9, 0x7f3, 0x6fa, 0x1f6, 0xff , 0x3f5, 0x2fc, 0xdfc, 0xcf5, 0xfff, 0xef6, 0x9fa, 0x8f3, 0xbf9, 0xaf0, 0x650, 0x759, 0x453, 0x55a, 0x256, 0x35f, 0x55 , 0x15c, 0xe5c, 0xf55, 0xc5f, 0xd56, 0xa5a, 0xb53, 0x859, 0x950, 0x7c0, 0x6c9, 0x5c3, 0x4ca, 0x3c6, 0x2cf, 0x1c5, 0xcc , 0xfcc, 0xec5, 0xdcf, 0xcc6, 0xbca, 0xac3, 0x9c9, 0x8c0, 0x8c0, 0x9c9, 0xac3, 0xbca, 0xcc6, 0xdcf, 0xec5, 0xfcc, 0xcc , 0x1c5, 0x2cf, 0x3c6, 0x4ca, 0x5c3, 0x6c9, 0x7c0, 0x950, 0x859, 0xb53, 0xa5a, 0xd56, 0xc5f, 0xf55, 0xe5c, 0x15c, 0x55 , 0x35f, 0x256, 0x55a, 0x453, 0x759, 0x650, 0xaf0, 0xbf9, 0x8f3, 0x9fa, 0xef6, 0xfff, 0xcf5, 0xdfc, 0x2fc, 0x3f5, 0xff , 0x1f6, 0x6fa, 0x7f3, 0x4f9, 0x5f0, 0xb60, 0xa69, 0x963, 0x86a, 0xf66, 0xe6f, 0xd65, 0xc6c, 0x36c, 0x265, 0x16f, 0x66 , 0x76a, 0x663, 0x569, 0x460, 0xca0, 0xda9, 0xea3, 0xfaa, 0x8a6, 0x9af, 0xaa5, 0xbac, 0x4ac, 0x5a5, 0x6af, 0x7a6, 0xaa , 0x1a3, 0x2a9, 0x3a0, 0xd30, 0xc39, 0xf33, 0xe3a, 0x936, 0x83f, 0xb35, 0xa3c, 0x53c, 0x435, 0x73f, 0x636, 0x13a, 0x33 , 0x339, 0x230, 0xe90, 0xf99, 0xc93, 0xd9a, 0xa96, 0xb9f, 0x895, 0x99c, 0x69c, 0x795, 0x49f, 0x596, 0x29a, 0x393, 0x99 , 0x190, 0xf00, 0xe09, 0xd03, 0xc0a, 0xb06, 0xa0f, 0x905, 0x80c, 0x70c, 0x605, 0x50f, 0x406, 0x30a, 0x203, 0x109, 0x0 }; cubeindex = 0; if (g[ind].val[0] < iso) cubeindex |= 1; if (g[ind].val[1] < iso) cubeindex |= 2; if (g[ind].val[2] < iso) cubeindex |= 4; if (g[ind].val[3] < iso) cubeindex |= 8; if (g[ind].val[4] < iso) cubeindex |= 16; if (g[ind].val[5] < iso) cubeindex |= 32; if (g[ind].val[6] < iso) cubeindex |= 64; if (g[ind].val[7] < iso) cubeindex |= 128; //printf("Est parte normal %d \n",cubeindex ); int triTable[256][16] = {{-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 8, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 1, 9, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {1, 8, 3, 9, 8, 1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {1, 2, 10, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 8, 3, 1, 2, 10, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {9, 2, 10, 0, 2, 9, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {2, 8, 3, 2, 10, 8, 10, 9, 8, -1, -1, -1, -1, -1, -1, -1}, {3, 11, 2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 11, 2, 8, 11, 0, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {1, 9, 0, 2, 3, 11, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {1, 11, 2, 1, 9, 11, 9, 8, 11, -1, -1, -1, -1, -1, -1, -1}, {3, 10, 1, 11, 10, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 10, 1, 0, 8, 10, 8, 11, 10, -1, -1, -1, -1, -1, -1, -1}, {3, 9, 0, 3, 11, 9, 11, 10, 9, -1, -1, -1, -1, -1, -1, -1}, {9, 8, 10, 10, 8, 11, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {4, 7, 8, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {4, 3, 0, 7, 3, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 1, 9, 8, 4, 7, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {4, 1, 9, 4, 7, 1, 7, 3, 1, -1, -1, -1, -1, -1, -1, -1}, {1, 2, 10, 8, 4, 7, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {3, 4, 7, 3, 0, 4, 1, 2, 10, -1, -1, -1, -1, -1, -1, -1}, {9, 2, 10, 9, 0, 2, 8, 4, 7, -1, -1, -1, -1, -1, -1, -1}, {2, 10, 9, 2, 9, 7, 2, 7, 3, 7, 9, 4, -1, -1, -1, -1}, {8, 4, 7, 3, 11, 2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {11, 4, 7, 11, 2, 4, 2, 0, 4, -1, -1, -1, -1, -1, -1, -1}, {9, 0, 1, 8, 4, 7, 2, 3, 11, -1, -1, -1, -1, -1, -1, -1}, {4, 7, 11, 9, 4, 11, 9, 11, 2, 9, 2, 1, -1, -1, -1, -1}, {3, 10, 1, 3, 11, 10, 7, 8, 4, -1, -1, -1, -1, -1, -1, -1}, {1, 11, 10, 1, 4, 11, 1, 0, 4, 7, 11, 4, -1, -1, -1, -1}, {4, 7, 8, 9, 0, 11, 9, 11, 10, 11, 0, 3, -1, -1, -1, -1}, {4, 7, 11, 4, 11, 9, 9, 11, 10, -1, -1, -1, -1, -1, -1, -1}, {9, 5, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {9, 5, 4, 0, 8, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 5, 4, 1, 5, 0, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {8, 5, 4, 8, 3, 5, 3, 1, 5, -1, -1, -1, -1, -1, -1, -1}, {1, 2, 10, 9, 5, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {3, 0, 8, 1, 2, 10, 4, 9, 5, -1, -1, -1, -1, -1, -1, -1}, {5, 2, 10, 5, 4, 2, 4, 0, 2, -1, -1, -1, -1, -1, -1, -1}, {2, 10, 5, 3, 2, 5, 3, 5, 4, 3, 4, 8, -1, -1, -1, -1}, {9, 5, 4, 2, 3, 11, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 11, 2, 0, 8, 11, 4, 9, 5, -1, -1, -1, -1, -1, -1, -1}, {0, 5, 4, 0, 1, 5, 2, 3, 11, -1, -1, -1, -1, -1, -1, -1}, {2, 1, 5, 2, 5, 8, 2, 8, 11, 4, 8, 5, -1, -1, -1, -1}, {10, 3, 11, 10, 1, 3, 9, 5, 4, -1, -1, -1, -1, -1, -1, -1}, {4, 9, 5, 0, 8, 1, 8, 10, 1, 8, 11, 10, -1, -1, -1, -1}, {5, 4, 0, 5, 0, 11, 5, 11, 10, 11, 0, 3, -1, -1, -1, -1}, {5, 4, 8, 5, 8, 10, 10, 8, 11, -1, -1, -1, -1, -1, -1, -1}, {9, 7, 8, 5, 7, 9, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {9, 3, 0, 9, 5, 3, 5, 7, 3, -1, -1, -1, -1, -1, -1, -1}, {0, 7, 8, 0, 1, 7, 1, 5, 7, -1, -1, -1, -1, -1, -1, -1}, {1, 5, 3, 3, 5, 7, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {9, 7, 8, 9, 5, 7, 10, 1, 2, -1, -1, -1, -1, -1, -1, -1}, {10, 1, 2, 9, 5, 0, 5, 3, 0, 5, 7, 3, -1, -1, -1, -1}, {8, 0, 2, 8, 2, 5, 8, 5, 7, 10, 5, 2, -1, -1, -1, -1}, {2, 10, 5, 2, 5, 3, 3, 5, 7, -1, -1, -1, -1, -1, -1, -1}, {7, 9, 5, 7, 8, 9, 3, 11, 2, -1, -1, -1, -1, -1, -1, -1}, {9, 5, 7, 9, 7, 2, 9, 2, 0, 2, 7, 11, -1, -1, -1, -1}, {2, 3, 11, 0, 1, 8, 1, 7, 8, 1, 5, 7, -1, -1, -1, -1}, {11, 2, 1, 11, 1, 7, 7, 1, 5, -1, -1, -1, -1, -1, -1, -1}, {9, 5, 8, 8, 5, 7, 10, 1, 3, 10, 3, 11, -1, -1, -1, -1}, {5, 7, 0, 5, 0, 9, 7, 11, 0, 1, 0, 10, 11, 10, 0, -1}, {11, 10, 0, 11, 0, 3, 10, 5, 0, 8, 0, 7, 5, 7, 0, -1}, {11, 10, 5, 7, 11, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {10, 6, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 8, 3, 5, 10, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {9, 0, 1, 5, 10, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {1, 8, 3, 1, 9, 8, 5, 10, 6, -1, -1, -1, -1, -1, -1, -1}, {1, 6, 5, 2, 6, 1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {1, 6, 5, 1, 2, 6, 3, 0, 8, -1, -1, -1, -1, -1, -1, -1}, {9, 6, 5, 9, 0, 6, 0, 2, 6, -1, -1, -1, -1, -1, -1, -1}, {5, 9, 8, 5, 8, 2, 5, 2, 6, 3, 2, 8, -1, -1, -1, -1}, {2, 3, 11, 10, 6, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {11, 0, 8, 11, 2, 0, 10, 6, 5, -1, -1, -1, -1, -1, -1, -1}, {0, 1, 9, 2, 3, 11, 5, 10, 6, -1, -1, -1, -1, -1, -1, -1}, {5, 10, 6, 1, 9, 2, 9, 11, 2, 9, 8, 11, -1, -1, -1, -1}, {6, 3, 11, 6, 5, 3, 5, 1, 3, -1, -1, -1, -1, -1, -1, -1}, {0, 8, 11, 0, 11, 5, 0, 5, 1, 5, 11, 6, -1, -1, -1, -1}, {3, 11, 6, 0, 3, 6, 0, 6, 5, 0, 5, 9, -1, -1, -1, -1}, {6, 5, 9, 6, 9, 11, 11, 9, 8, -1, -1, -1, -1, -1, -1, -1}, {5, 10, 6, 4, 7, 8, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {4, 3, 0, 4, 7, 3, 6, 5, 10, -1, -1, -1, -1, -1, -1, -1}, {1, 9, 0, 5, 10, 6, 8, 4, 7, -1, -1, -1, -1, -1, -1, -1}, {10, 6, 5, 1, 9, 7, 1, 7, 3, 7, 9, 4, -1, -1, -1, -1}, {6, 1, 2, 6, 5, 1, 4, 7, 8, -1, -1, -1, -1, -1, -1, -1}, {1, 2, 5, 5, 2, 6, 3, 0, 4, 3, 4, 7, -1, -1, -1, -1}, {8, 4, 7, 9, 0, 5, 0, 6, 5, 0, 2, 6, -1, -1, -1, -1}, {7, 3, 9, 7, 9, 4, 3, 2, 9, 5, 9, 6, 2, 6, 9, -1}, {3, 11, 2, 7, 8, 4, 10, 6, 5, -1, -1, -1, -1, -1, -1, -1}, {5, 10, 6, 4, 7, 2, 4, 2, 0, 2, 7, 11, -1, -1, -1, -1}, {0, 1, 9, 4, 7, 8, 2, 3, 11, 5, 10, 6, -1, -1, -1, -1}, {9, 2, 1, 9, 11, 2, 9, 4, 11, 7, 11, 4, 5, 10, 6, -1}, {8, 4, 7, 3, 11, 5, 3, 5, 1, 5, 11, 6, -1, -1, -1, -1}, {5, 1, 11, 5, 11, 6, 1, 0, 11, 7, 11, 4, 0, 4, 11, -1}, {0, 5, 9, 0, 6, 5, 0, 3, 6, 11, 6, 3, 8, 4, 7, -1}, {6, 5, 9, 6, 9, 11, 4, 7, 9, 7, 11, 9, -1, -1, -1, -1}, {10, 4, 9, 6, 4, 10, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {4, 10, 6, 4, 9, 10, 0, 8, 3, -1, -1, -1, -1, -1, -1, -1}, {10, 0, 1, 10, 6, 0, 6, 4, 0, -1, -1, -1, -1, -1, -1, -1}, {8, 3, 1, 8, 1, 6, 8, 6, 4, 6, 1, 10, -1, -1, -1, -1}, {1, 4, 9, 1, 2, 4, 2, 6, 4, -1, -1, -1, -1, -1, -1, -1}, {3, 0, 8, 1, 2, 9, 2, 4, 9, 2, 6, 4, -1, -1, -1, -1}, {0, 2, 4, 4, 2, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {8, 3, 2, 8, 2, 4, 4, 2, 6, -1, -1, -1, -1, -1, -1, -1}, {10, 4, 9, 10, 6, 4, 11, 2, 3, -1, -1, -1, -1, -1, -1, -1}, {0, 8, 2, 2, 8, 11, 4, 9, 10, 4, 10, 6, -1, -1, -1, -1}, {3, 11, 2, 0, 1, 6, 0, 6, 4, 6, 1, 10, -1, -1, -1, -1}, {6, 4, 1, 6, 1, 10, 4, 8, 1, 2, 1, 11, 8, 11, 1, -1}, {9, 6, 4, 9, 3, 6, 9, 1, 3, 11, 6, 3, -1, -1, -1, -1}, {8, 11, 1, 8, 1, 0, 11, 6, 1, 9, 1, 4, 6, 4, 1, -1}, {3, 11, 6, 3, 6, 0, 0, 6, 4, -1, -1, -1, -1, -1, -1, -1}, {6, 4, 8, 11, 6, 8, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {7, 10, 6, 7, 8, 10, 8, 9, 10, -1, -1, -1, -1, -1, -1, -1}, {0, 7, 3, 0, 10, 7, 0, 9, 10, 6, 7, 10, -1, -1, -1, -1}, {10, 6, 7, 1, 10, 7, 1, 7, 8, 1, 8, 0, -1, -1, -1, -1}, {10, 6, 7, 10, 7, 1, 1, 7, 3, -1, -1, -1, -1, -1, -1, -1}, {1, 2, 6, 1, 6, 8, 1, 8, 9, 8, 6, 7, -1, -1, -1, -1}, {2, 6, 9, 2, 9, 1, 6, 7, 9, 0, 9, 3, 7, 3, 9, -1}, {7, 8, 0, 7, 0, 6, 6, 0, 2, -1, -1, -1, -1, -1, -1, -1}, {7, 3, 2, 6, 7, 2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {2, 3, 11, 10, 6, 8, 10, 8, 9, 8, 6, 7, -1, -1, -1, -1}, {2, 0, 7, 2, 7, 11, 0, 9, 7, 6, 7, 10, 9, 10, 7, -1}, {1, 8, 0, 1, 7, 8, 1, 10, 7, 6, 7, 10, 2, 3, 11, -1}, {11, 2, 1, 11, 1, 7, 10, 6, 1, 6, 7, 1, -1, -1, -1, -1}, {8, 9, 6, 8, 6, 7, 9, 1, 6, 11, 6, 3, 1, 3, 6, -1}, {0, 9, 1, 11, 6, 7, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {7, 8, 0, 7, 0, 6, 3, 11, 0, 11, 6, 0, -1, -1, -1, -1}, {7, 11, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {7, 6, 11, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {3, 0, 8, 11, 7, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 1, 9, 11, 7, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {8, 1, 9, 8, 3, 1, 11, 7, 6, -1, -1, -1, -1, -1, -1, -1}, {10, 1, 2, 6, 11, 7, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {1, 2, 10, 3, 0, 8, 6, 11, 7, -1, -1, -1, -1, -1, -1, -1}, {2, 9, 0, 2, 10, 9, 6, 11, 7, -1, -1, -1, -1, -1, -1, -1}, {6, 11, 7, 2, 10, 3, 10, 8, 3, 10, 9, 8, -1, -1, -1, -1}, {7, 2, 3, 6, 2, 7, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {7, 0, 8, 7, 6, 0, 6, 2, 0, -1, -1, -1, -1, -1, -1, -1}, {2, 7, 6, 2, 3, 7, 0, 1, 9, -1, -1, -1, -1, -1, -1, -1}, {1, 6, 2, 1, 8, 6, 1, 9, 8, 8, 7, 6, -1, -1, -1, -1}, {10, 7, 6, 10, 1, 7, 1, 3, 7, -1, -1, -1, -1, -1, -1, -1}, {10, 7, 6, 1, 7, 10, 1, 8, 7, 1, 0, 8, -1, -1, -1, -1}, {0, 3, 7, 0, 7, 10, 0, 10, 9, 6, 10, 7, -1, -1, -1, -1}, {7, 6, 10, 7, 10, 8, 8, 10, 9, -1, -1, -1, -1, -1, -1, -1}, {6, 8, 4, 11, 8, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {3, 6, 11, 3, 0, 6, 0, 4, 6, -1, -1, -1, -1, -1, -1, -1}, {8, 6, 11, 8, 4, 6, 9, 0, 1, -1, -1, -1, -1, -1, -1, -1}, {9, 4, 6, 9, 6, 3, 9, 3, 1, 11, 3, 6, -1, -1, -1, -1}, {6, 8, 4, 6, 11, 8, 2, 10, 1, -1, -1, -1, -1, -1, -1, -1}, {1, 2, 10, 3, 0, 11, 0, 6, 11, 0, 4, 6, -1, -1, -1, -1}, {4, 11, 8, 4, 6, 11, 0, 2, 9, 2, 10, 9, -1, -1, -1, -1}, {10, 9, 3, 10, 3, 2, 9, 4, 3, 11, 3, 6, 4, 6, 3, -1}, {8, 2, 3, 8, 4, 2, 4, 6, 2, -1, -1, -1, -1, -1, -1, -1}, {0, 4, 2, 4, 6, 2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {1, 9, 0, 2, 3, 4, 2, 4, 6, 4, 3, 8, -1, -1, -1, -1}, {1, 9, 4, 1, 4, 2, 2, 4, 6, -1, -1, -1, -1, -1, -1, -1}, {8, 1, 3, 8, 6, 1, 8, 4, 6, 6, 10, 1, -1, -1, -1, -1}, {10, 1, 0, 10, 0, 6, 6, 0, 4, -1, -1, -1, -1, -1, -1, -1}, {4, 6, 3, 4, 3, 8, 6, 10, 3, 0, 3, 9, 10, 9, 3, -1}, {10, 9, 4, 6, 10, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {4, 9, 5, 7, 6, 11, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 8, 3, 4, 9, 5, 11, 7, 6, -1, -1, -1, -1, -1, -1, -1}, {5, 0, 1, 5, 4, 0, 7, 6, 11, -1, -1, -1, -1, -1, -1, -1}, {11, 7, 6, 8, 3, 4, 3, 5, 4, 3, 1, 5, -1, -1, -1, -1}, {9, 5, 4, 10, 1, 2, 7, 6, 11, -1, -1, -1, -1, -1, -1, -1}, {6, 11, 7, 1, 2, 10, 0, 8, 3, 4, 9, 5, -1, -1, -1, -1}, {7, 6, 11, 5, 4, 10, 4, 2, 10, 4, 0, 2, -1, -1, -1, -1}, {3, 4, 8, 3, 5, 4, 3, 2, 5, 10, 5, 2, 11, 7, 6, -1}, {7, 2, 3, 7, 6, 2, 5, 4, 9, -1, -1, -1, -1, -1, -1, -1}, {9, 5, 4, 0, 8, 6, 0, 6, 2, 6, 8, 7, -1, -1, -1, -1}, {3, 6, 2, 3, 7, 6, 1, 5, 0, 5, 4, 0, -1, -1, -1, -1}, {6, 2, 8, 6, 8, 7, 2, 1, 8, 4, 8, 5, 1, 5, 8, -1}, {9, 5, 4, 10, 1, 6, 1, 7, 6, 1, 3, 7, -1, -1, -1, -1}, {1, 6, 10, 1, 7, 6, 1, 0, 7, 8, 7, 0, 9, 5, 4, -1}, {4, 0, 10, 4, 10, 5, 0, 3, 10, 6, 10, 7, 3, 7, 10, -1}, {7, 6, 10, 7, 10, 8, 5, 4, 10, 4, 8, 10, -1, -1, -1, -1}, {6, 9, 5, 6, 11, 9, 11, 8, 9, -1, -1, -1, -1, -1, -1, -1}, {3, 6, 11, 0, 6, 3, 0, 5, 6, 0, 9, 5, -1, -1, -1, -1}, {0, 11, 8, 0, 5, 11, 0, 1, 5, 5, 6, 11, -1, -1, -1, -1}, {6, 11, 3, 6, 3, 5, 5, 3, 1, -1, -1, -1, -1, -1, -1, -1}, {1, 2, 10, 9, 5, 11, 9, 11, 8, 11, 5, 6, -1, -1, -1, -1}, {0, 11, 3, 0, 6, 11, 0, 9, 6, 5, 6, 9, 1, 2, 10, -1}, {11, 8, 5, 11, 5, 6, 8, 0, 5, 10, 5, 2, 0, 2, 5, -1}, {6, 11, 3, 6, 3, 5, 2, 10, 3, 10, 5, 3, -1, -1, -1, -1}, {5, 8, 9, 5, 2, 8, 5, 6, 2, 3, 8, 2, -1, -1, -1, -1}, {9, 5, 6, 9, 6, 0, 0, 6, 2, -1, -1, -1, -1, -1, -1, -1}, {1, 5, 8, 1, 8, 0, 5, 6, 8, 3, 8, 2, 6, 2, 8, -1}, {1, 5, 6, 2, 1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {1, 3, 6, 1, 6, 10, 3, 8, 6, 5, 6, 9, 8, 9, 6, -1}, {10, 1, 0, 10, 0, 6, 9, 5, 0, 5, 6, 0, -1, -1, -1, -1}, {0, 3, 8, 5, 6, 10, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {10, 5, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {11, 5, 10, 7, 5, 11, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {11, 5, 10, 11, 7, 5, 8, 3, 0, -1, -1, -1, -1, -1, -1, -1}, {5, 11, 7, 5, 10, 11, 1, 9, 0, -1, -1, -1, -1, -1, -1, -1}, {10, 7, 5, 10, 11, 7, 9, 8, 1, 8, 3, 1, -1, -1, -1, -1}, {11, 1, 2, 11, 7, 1, 7, 5, 1, -1, -1, -1, -1, -1, -1, -1}, {0, 8, 3, 1, 2, 7, 1, 7, 5, 7, 2, 11, -1, -1, -1, -1}, {9, 7, 5, 9, 2, 7, 9, 0, 2, 2, 11, 7, -1, -1, -1, -1}, {7, 5, 2, 7, 2, 11, 5, 9, 2, 3, 2, 8, 9, 8, 2, -1}, {2, 5, 10, 2, 3, 5, 3, 7, 5, -1, -1, -1, -1, -1, -1, -1}, {8, 2, 0, 8, 5, 2, 8, 7, 5, 10, 2, 5, -1, -1, -1, -1}, {9, 0, 1, 5, 10, 3, 5, 3, 7, 3, 10, 2, -1, -1, -1, -1}, {9, 8, 2, 9, 2, 1, 8, 7, 2, 10, 2, 5, 7, 5, 2, -1}, {1, 3, 5, 3, 7, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 8, 7, 0, 7, 1, 1, 7, 5, -1, -1, -1, -1, -1, -1, -1}, {9, 0, 3, 9, 3, 5, 5, 3, 7, -1, -1, -1, -1, -1, -1, -1}, {9, 8, 7, 5, 9, 7, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {5, 8, 4, 5, 10, 8, 10, 11, 8, -1, -1, -1, -1, -1, -1, -1}, {5, 0, 4, 5, 11, 0, 5, 10, 11, 11, 3, 0, -1, -1, -1, -1}, {0, 1, 9, 8, 4, 10, 8, 10, 11, 10, 4, 5, -1, -1, -1, -1}, {10, 11, 4, 10, 4, 5, 11, 3, 4, 9, 4, 1, 3, 1, 4, -1}, {2, 5, 1, 2, 8, 5, 2, 11, 8, 4, 5, 8, -1, -1, -1, -1}, {0, 4, 11, 0, 11, 3, 4, 5, 11, 2, 11, 1, 5, 1, 11, -1}, {0, 2, 5, 0, 5, 9, 2, 11, 5, 4, 5, 8, 11, 8, 5, -1}, {9, 4, 5, 2, 11, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {2, 5, 10, 3, 5, 2, 3, 4, 5, 3, 8, 4, -1, -1, -1, -1}, {5, 10, 2, 5, 2, 4, 4, 2, 0, -1, -1, -1, -1, -1, -1, -1}, {3, 10, 2, 3, 5, 10, 3, 8, 5, 4, 5, 8, 0, 1, 9, -1}, {5, 10, 2, 5, 2, 4, 1, 9, 2, 9, 4, 2, -1, -1, -1, -1}, {8, 4, 5, 8, 5, 3, 3, 5, 1, -1, -1, -1, -1, -1, -1, -1}, {0, 4, 5, 1, 0, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {8, 4, 5, 8, 5, 3, 9, 0, 5, 0, 3, 5, -1, -1, -1, -1}, {9, 4, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {4, 11, 7, 4, 9, 11, 9, 10, 11, -1, -1, -1, -1, -1, -1, -1}, {0, 8, 3, 4, 9, 7, 9, 11, 7, 9, 10, 11, -1, -1, -1, -1}, {1, 10, 11, 1, 11, 4, 1, 4, 0, 7, 4, 11, -1, -1, -1, -1}, {3, 1, 4, 3, 4, 8, 1, 10, 4, 7, 4, 11, 10, 11, 4, -1}, {4, 11, 7, 9, 11, 4, 9, 2, 11, 9, 1, 2, -1, -1, -1, -1}, {9, 7, 4, 9, 11, 7, 9, 1, 11, 2, 11, 1, 0, 8, 3, -1}, {11, 7, 4, 11, 4, 2, 2, 4, 0, -1, -1, -1, -1, -1, -1, -1}, {11, 7, 4, 11, 4, 2, 8, 3, 4, 3, 2, 4, -1, -1, -1, -1}, {2, 9, 10, 2, 7, 9, 2, 3, 7, 7, 4, 9, -1, -1, -1, -1}, {9, 10, 7, 9, 7, 4, 10, 2, 7, 8, 7, 0, 2, 0, 7, -1}, {3, 7, 10, 3, 10, 2, 7, 4, 10, 1, 10, 0, 4, 0, 10, -1}, {1, 10, 2, 8, 7, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {4, 9, 1, 4, 1, 7, 7, 1, 3, -1, -1, -1, -1, -1, -1, -1}, {4, 9, 1, 4, 1, 7, 0, 8, 1, 8, 7, 1, -1, -1, -1, -1}, {4, 0, 3, 7, 4, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {4, 8, 7, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {9, 10, 8, 10, 11, 8, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {3, 0, 9, 3, 9, 11, 11, 9, 10, -1, -1, -1, -1, -1, -1, -1}, {0, 1, 10, 0, 10, 8, 8, 10, 11, -1, -1, -1, -1, -1, -1, -1}, {3, 1, 10, 11, 3, 10, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {1, 2, 11, 1, 11, 9, 9, 11, 8, -1, -1, -1, -1, -1, -1, -1}, {3, 0, 9, 3, 9, 11, 1, 2, 9, 2, 11, 9, -1, -1, -1, -1}, {0, 2, 11, 8, 0, 11, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {3, 2, 11, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {2, 3, 8, 2, 8, 10, 10, 8, 9, -1, -1, -1, -1, -1, -1, -1}, {9, 10, 2, 0, 9, 2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {2, 3, 8, 2, 8, 10, 0, 1, 8, 1, 10, 8, -1, -1, -1, -1}, {1, 10, 2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {1, 3, 8, 9, 1, 8, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 9, 1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {0, 3, 8, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, {-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}}; if (edgeTable[cubeindex] == 0) return; /* Find the vertices where the surface intersects the cube */ if (edgeTable[cubeindex] & 1) { vertlist[0] = VertexInterp(iso,g[ind].p[0],g[ind].p[1],g[ind].val[0],g[ind].val[1]); } if (edgeTable[cubeindex] & 2) { vertlist[1] = VertexInterp(iso,g[ind].p[1],g[ind].p[2],g[ind].val[1],g[ind].val[2]); } if (edgeTable[cubeindex] & 4) { vertlist[2] = VertexInterp(iso,g[ind].p[2],g[ind].p[3],g[ind].val[2],g[ind].val[3]); } if (edgeTable[cubeindex] & 8) { vertlist[3] = VertexInterp(iso,g[ind].p[3],g[ind].p[0],g[ind].val[3],g[ind].val[0]); } if (edgeTable[cubeindex] & 16) { vertlist[4] = VertexInterp(iso,g[ind].p[4],g[ind].p[5],g[ind].val[4],g[ind].val[5]); } if (edgeTable[cubeindex] & 32) { vertlist[5] = VertexInterp(iso,g[ind].p[5],g[ind].p[6],g[ind].val[5],g[ind].val[6]); } if (edgeTable[cubeindex] & 64) { vertlist[6] = VertexInterp(iso,g[ind].p[6],g[ind].p[7],g[ind].val[6],g[ind].val[7]); } if (edgeTable[cubeindex] & 128) { vertlist[7] = VertexInterp(iso,g[ind].p[7],g[ind].p[4],g[ind].val[7],g[ind].val[4]); } if (edgeTable[cubeindex] & 256) { vertlist[8] = VertexInterp(iso,g[ind].p[0],g[ind].p[4],g[ind].val[0],g[ind].val[4]); } if (edgeTable[cubeindex] & 512) { vertlist[9] = VertexInterp(iso,g[ind].p[1],g[ind].p[5],g[ind].val[1],g[ind].val[5]); } if (edgeTable[cubeindex] & 1024) { vertlist[10] = VertexInterp(iso,g[ind].p[2],g[ind].p[6],g[ind].val[2],g[ind].val[6]); } if (edgeTable[cubeindex] & 2048) { vertlist[11] = VertexInterp(iso,g[ind].p[3],g[ind].p[7],g[ind].val[3],g[ind].val[7]); } int t=0; //printf("indice %d \n", ind); while(triTable[cubeindex][t] != -1) { d_vectTriangles[ind].p[0] = vertlist[triTable[cubeindex][t]]; d_vectTriangles[ind].p[1] = vertlist[triTable[cubeindex][t+1]]; d_vectTriangles[ind].p[2] = vertlist[triTable[cubeindex][t+2]]; ntri++; t+=3; } //printf("ntri %d \n", ntri); } } //return ntri; void printGrid(string a, GRIDCELL * g, int tam) { cout<<a; for(int i =0; i<tam ;i++) for(int j=0;j<8;j++) //printf("%f %f %f \n", g[i].p[j].x ,g[i].p[j].y,g[i].p[j].z); printf("%f \n", g[i].val[j]); } void printTriangles(string a , TRIANGLE * t, int tam) { //numtri=ntri; //printTriangles("Printing Tringles \n",cpy_vectTriangles,N); cout<<a; int cont=0; ofstream myfile; myfile.open ("Triangulos.txt"); //myfilen.open("Normales.txt"); myfile<<"# .PCD v.5 - Point Cloud Data file format"<<endl; myfile<<"VERSION .5"<<endl; myfile<<"FIELDS x y z"<<endl; myfile<<"SIZE 4 4 4"<<endl; myfile<<"TYPE F F F"<<endl; myfile<<"COUNT 1 1 1"<<endl; myfile<<"WIDTH 742716"<<endl; myfile<<"HEIGHT 1"<<endl; myfile<<"POINTS 742716"<<endl; myfile<<"DATA ascii"<<endl; for(int i =0; i<tam ;i++) { for (int k=0;k<3;k++) { cont++; if(t[i].p[k].x != 0 && t[i].p[k].y != 0 && t[i].p[k].z !=0 ) { //fprintf(fptr,"%g %g %g ",t[i].p[k].x,t[i].p[k].y,t[i].p[k].z); myfile<<t[i].p[k].x<<" "<<t[i].p[k].y<<" "<<t[i].p[k].z<<endl; } } //t++; } myfile.close(); cout<<"numero de triangulos"<<cont<<endl; } /* XYZ * CalcNormals(string a , XYZ * data, int tam) { XYZ normales[tam]; for (int i=0;i<NX-1;i++) { for (int j=0;j<NY-1;j++) { for (int k=0;k<NZ-1;k++) { Normales[n].x=(data[i+1][j][k]+data[i-1,j,k])/0.0001; Normales[n].y=(data[i][j+1][k]-data[i][j-1][k])/0.0001; Normales[n].z=(data[i][j][k+1]-data[i][j][k-1])/0.0001; n++; } } return normales; } */ int main(int argc, char *argv[]) { int i,j,k,c; int numtri=0; int ***data; FILE *fptr; int N= ((NX-1)*(NY-1)*(NZ-1)); cout<<N<<endl; //return 1; int THREADS_PER_BLOCK =1024; int themin=255; int themax=0; int isolevel=80; //const char* FILENAME = "mri.raw"; //assingMem(data); //readFile(fptr,FILENAME,themin, themax,data); // Malloc the volumetric data, hardwired size! data = (int***)malloc(NX*sizeof(int **)); for (i=0;i<NX;i++) data[i] = (int**)malloc(NY*sizeof(int *)); for (i=0;i<NX;i++) for (j=0;j<NY;j++) data[i][j] = (int*)malloc(NZ*sizeof(int)); //cout<<data[199][60][0]<<endl; // Open and read the raw data fprintf(stderr,"Reading data ...\n"); if ((fptr = fopen(argv[argc-1],"rb")) == NULL) { fprintf(stderr,"File open failed\n"); exit(-1); } cout<<"llega"<<endl; for (k=0;k<NZ;k++) { for (j=0;j<NY;j++) { for (i=0;i<NX;i++) { if ((c = fgetc(fptr)) == EOF) { fprintf(stderr,"Unexpected end of file\n"); exit(-1); } data[i][j][k] = c; //cout<<i<<" "<<j <<" "<<k <<" data : "<<data[i][j][k]<<endl; if (c > themax) themax = c; if (c < themin) themin = c; } } } fclose(fptr); fprintf(stderr,"Volumetric data range: %d -> %d\n",themin,themax); int sizeGRID= N*sizeof(GRIDCELL); int sizeTRI = N*sizeof(TRIANGLE); cout<<"sizeGRID "<<sizeGRID<<endl; cout<<"sizeTRI "<<sizeTRI<<endl; GRIDCELL * vectGrids; GRIDCELL * d_vectGrids; TRIANGLE * vectTriangles; TRIANGLE * d_vectTriangles; vectGrids = (GRIDCELL *)malloc(sizeGRID); vectTriangles= (TRIANGLE *)malloc(sizeTRI); cout<<"mem of grid "<<vectGrids<<endl; cout<<"mem of triangulos "<<vectTriangles<<endl; cout<<"asigna memoria sin problemas"<<endl; int numCubos = constructCubes(vectGrids,data); cout<<"pasa"<<endl; /* typedef struct { XYZ p[3]; // XYZ c; // XYZ n[3]; } TRIANGLE; typedef struct { XYZ p[8]; double val[8]; } GRIDCELL; */ size_t available, total; cudaMemGetInfo(&available, &total); cout<<"available: " << available<<" total: "<<total <<endl; cudaMalloc((void **)&d_vectGrids, sizeGRID); cudaMemcpy(d_vectGrids,vectGrids, sizeGRID, cudaMemcpyHostToDevice); cout<<"asignacion en RAM"<<endl; cudaMalloc((void **)&d_vectTriangles, sizeTRI); cudaMemcpy(d_vectTriangles,vectTriangles,sizeTRI,cudaMemcpyHostToDevice); cout<<"mem of grid "<<d_vectGrids<<endl; cout<<"mem of triangulos "<<d_vectTriangles<<endl; cout<<"separa memoria en cuda sin problemas"<<endl; //printGrid("imprimiendo Grid inicial en Host \n ",vectGrids,N); cudaEvent_t start, stop; float elapsedTime; cudaEventCreate(&start); //int x = NX-1; int y = NY-1 ; int z = NZ-1; /*int blockX= (x + THREADS_PER_BLOCK -1)/THREADS_PER_BLOCK; int blockY= (y + THREADS_PER_BLOCK -1)/THREADS_PER_BLOCK; int blockZ= (z + THREADS_PER_BLOCK -1)/THREADS_PER_BLOCK; cout<<"blocks : "<<blockX<<" threds: "<<THREADS_PER_BLOCK<<endl; cout<<"blocks : "<<blockY<<" threds: "<<THREADS_PER_BLOCK<<endl; cout<<"blocks : "<<blockZ<<" threds: "<<THREADS_PER_BLOCK<<endl;*/ //int ntri=0;int d_ntri=0; //cudaMalloc((void **)ntri,int); //cudaMemcpy(d_ntri,ntri,cudaMemcpyHostToDevice); int blocks= (N + THREADS_PER_BLOCK -1)/THREADS_PER_BLOCK; cout<<"blocks : \n"<<blocks<<"\n threds: \n "<<THREADS_PER_BLOCK<<endl; //dim3 dimGrid(blockX, blockY, blockZ); //dim3 dimBlock(THREADS_PER_BLOCK,THREADS_PER_BLOCK, THREADS_PER_BLOCK); cudaEventRecord(start,0); //copyGRID1<<<dimGrid,dimBlock>>>(d_res,d_vectGrid,x,y,z); //PolygoniseCube<<<dimGrid,dimBlock>>>(d_vectTriangles,d_vectGrids,x,y,z); PolygoniseCube<<<blocks,THREADS_PER_BLOCK>>>(d_vectTriangles,d_vectGrids,N); //matrixAdition<<<blocks,THREADS_PER_BLOCK>>>(d_a, d_points,10); //matrixAditionCol<<<blocks2,THREADS_PER_BLOCK>>>( d_c, d_a, d_b,N); cudaEventCreate(&stop); cudaEventRecord(stop,0); cudaEventSynchronize(stop); cudaEventElapsedTime(&elapsedTime, start,stop); printf("Elapsed time : %f ms\n" ,elapsedTime); TRIANGLE * cpy_vectTriangles; cpy_vectTriangles= (TRIANGLE *)malloc(sizeTRI); cout<<"crea sin problemas en host"<<endl; cudaMemcpy(cpy_vectTriangles,d_vectTriangles, sizeTRI, cudaMemcpyDeviceToHost); printTriangles("Printing Tringles \n",cpy_vectTriangles,N); //CalcNormals("Calculate Normals \n",data,N); int n=0; free(vectTriangles); free(vectGrids); free(cpy_vectTriangles); cudaFree(d_vectTriangles); cudaFree(d_vectGrids); return 0; } /* for(int i=0;i<N;i++) { cudaMalloc((void**)&d_p,8*sizeof(XYZ)); //cudaMemGetInfo(&available, &total); //cout<<"available: " << available<<" total: "<<total <<endl; cudaMalloc((void**)&d_val,8*sizeof(double)); //cudaMemGetInfo(&available, &total); //cout<<"available: " << available<<" total: "<<total <<endl; cudaMemcpy(d_p,vectGrids[i].p,8*sizeof(XYZ),cudaMemcpyHostToDevice); //for(int w=0;w<8;w++) //{ cout<<vectGrids[i].p[w].y<<endl; //} //cudaMemGetInfo(&available, &total); //cout<<"available: " << available<<" total: "<<total <<endl; cudaMemcpy(d_val,vectGrids[i].val,8*sizeof(double),cudaMemcpyHostToDevice); //cudaMemGetInfo(&available, &total); //cout<<"available: " << available<<" total: "<<total <<endl cudaMemGetInfo(&available, &total); //cout<<"available: " << available<<" total: "<<total <<endl; cudaMemcpy(d_vectGrid[i].val, d_val, 8*sizeof(double),cudaMemcpyHostToDevice); cudaMemcpy(d_vectGrid[i].p, d_p, 8*sizeof(XYZ),cudaMemcpyHostToDevice); }*/
7,713
#define MAXBLOCK 512 __device__ int sgn(float x){ if(x == 0.0){ return 0; }else if(x < 0.0){ return -1; } return 1; } __device__ int square(int x){ return x*x; } __global__ void binarize(int* output, int2 size, float* implicit){ const int x = blockIdx.x * blockDim.x + threadIdx.x; const int y = blockIdx.y * blockDim.y + threadIdx.y; const int i = x + y * size.x; const int replace = square(max(size.x,size.y)); if(size.x <= x || size.y <= y) return; output[i] = sgn(implicit[i]) * replace; } __global__ void edt_pass(int* samples, const int width, const int height, const int dim){ __shared__ int coeffs[MAXBLOCK]; __shared__ int verts[MAXBLOCK]; __shared__ int signs[MAXBLOCK]; // this requires that the thread size is (length of data,1) const int x = blockIdx.x * blockDim.x + threadIdx.x; const int y = blockIdx.y * blockDim.y + threadIdx.y; int* sample = &(samples[x + y * width]); int frame; int pos; int size; if(dim){ frame = y & ~1; pos = y; size = height; }else{ frame = x & ~1; pos = x; size = width; } // Perform the first set of reductions int out = *sample; int sign = sgn(out); out = abs(out); signs[pos] = sign * pos; coeffs[pos] = out; __syncthreads(); int otherindex = pos ^ 1; int otherdata = coeffs[otherindex]; int othersign = signs[otherindex]; if(othersign * sign < 0){ if(sign == -1){ out = 0; coeffs[pos] = 0; verts[pos] = pos; signs[pos] = 0; }else{ out = 1; coeffs[pos] = 1; verts[pos] = otherindex; } }else if(out > otherdata){ coeffs[pos] = otherdata + 1; verts[pos] = otherindex; }else{ verts[pos] = pos; } __syncthreads(); int mask = 3; while(size > 0){ size >>= 1; int base = frame & ~3; int dest = base >> 1; int offset = base ^ frame; int half = offset >> 1; offset = offset | half; int par = pos & mask; int lowvertex = verts[base + 1]; int highvertex = verts[base + 2]; int lowcoeff = coeffs[base + 1]; int highcoeff = coeffs[base + 2]; int lowsgn = signs[base+1]; int highsgn = signs[base+2]; if((0 > lowsgn * highsgn) && (size > 1)){ lowvertex = abs(min(lowsgn,highsgn)); lowcoeff = 0; } int low = square(pos - lowvertex) + lowcoeff; int high = square(pos - highvertex) + highcoeff; int extreme = square(pos - verts[base + offset]) + coeffs[base + offset]; out = min(out,min(high,min(low,extreme))); if(par == 0 || par == mask){ int vertex; int coefficient; if(high < extreme || low < extreme){ if(high < low){ vertex = highvertex; coefficient = highcoeff; }else{ vertex = lowvertex; coefficient = lowcoeff; } } else { vertex = verts[base + offset]; coefficient = coeffs[base + offset]; } int s = signs[base + 3 * half]; __syncthreads(); signs[dest + half] = s; coeffs[dest + half] = coefficient; verts[dest + half] = vertex; __syncthreads(); }else{ __syncthreads(); __syncthreads(); } frame = dest; mask = (mask << 1) + 1; } *sample = sign * out; } __global__ void signed_sqrt(int* values, int2 size, float* output){ // int width, int height){ const int x = blockIdx.x * blockDim.x + threadIdx.x; const int y = blockIdx.y * blockDim.y + threadIdx.y; if(size.x <= x || size.y <= y) return; const int i = x + size.x * y; float out = 1.0; int value = values[i]; if(value < 0){ out = -1.0; value = -1 * value; } output[i] = out * sqrtf((float) value) / ((float) min(size.y,size.x)); }
7,714
#include <stdio.h> #include <stdlib.h> #include <math.h> #include <time.h> #define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); } inline void gpuAssert(cudaError_t code, char *file, int line, bool abort=true) { if (code != cudaSuccess) { fprintf(stderr,"GPUassert: %s %s %d\n", cudaGetErrorString(code), file, line); if (abort) exit(code); } } // CUDA kernel to pause for at least num_cycle cycles __global__ void sleep(int64_t *completed_cycles, int64_t requested_cycles) { completed_cycles[0] = 0; int64_t start = clock64(); while(completed_cycles[0] < requested_cycles) { completed_cycles[0] = clock64() - start; } } extern "C" void allocate_mem(int64_t **device_value) { gpuErrchk( cudaMalloc((void**)device_value, sizeof(int64_t)) ); } extern "C" void copy_mem(int64_t *host_value, int64_t *device_value) { gpuErrchk( cudaMemcpy(host_value, device_value, sizeof(int64_t), cudaMemcpyDeviceToHost) ); } // Returns number of cycles required for requested seconds extern "C" int64_t get_cycles(float seconds) { // Get device frequency in KHz int64_t Hz; cudaDeviceProp prop; cudaGetDeviceProperties(&prop, 0); Hz = int64_t(prop.clockRate) * 1000; // Calculate number of cycles to wait int64_t num_cycles; num_cycles = (int64_t)(seconds * Hz); return num_cycles; } // Launches a kernel that sleeps for num_cycles extern "C" void sleep_kernel(int64_t *completed_cycles, int64_t requested_cycles) { // Our kernel will launch a single thread to sleep the kernel int blockSize, gridSize; blockSize = 1; gridSize = 1; // Execute the kernel in default stream sleep<<< gridSize, blockSize >>>(completed_cycles, requested_cycles); } // Wait for all work to complete extern "C" void wait_for_gpu() { cudaDeviceSynchronize(); }
7,715
//pass //--blockDim=[64,1] --gridDim=[64,1] #include <cuda.h> #define BIN_COUNT 64 //////////////////////////////////////////////////////////////////////////////// // GPU-specific definitions //////////////////////////////////////////////////////////////////////////////// //Fast mul on G8x / G9x / G100 #define IMUL(a, b) a * b //////////////////////////////////////////////////////////////////////////////// // Merge blockN histograms into gridDim.x histograms // blockDim.x == BIN_COUNT // gridDim.x == BLOCK_N2 //////////////////////////////////////////////////////////////////////////////// #define MERGE_THREADS 64 __global__ void mergeHistogram64Kernel( unsigned int *d_Histogram, unsigned int *d_PartialHistograms, unsigned int blockN ){ __shared__ unsigned int data[MERGE_THREADS]; unsigned int sum = 0; for(unsigned int i = threadIdx.x; i < blockN; i += MERGE_THREADS) { sum += d_PartialHistograms[blockIdx.x + i * BIN_COUNT]; } data[threadIdx.x] = sum; for(unsigned int stride = MERGE_THREADS / 2; stride > 0; stride >>= 1){ __syncthreads(); /* BUGINJECT: ADD_BARRIER, DOWN */ if(threadIdx.x < stride) { #ifdef MUTATION __syncthreads(); #endif data[threadIdx.x] += data[threadIdx.x + stride]; } } if(threadIdx.x == 0) d_Histogram[blockIdx.x] = data[0]; }
7,716
#include <stdio.h> #include <stdlib.h> __global__ void add(int *da, int *db, int *dc) { dc[threadIdx.x] = da[threadIdx.x] + db[threadIdx.x]; } int main(int argc, char **argv) { int a_in = atoi(argv[1]); // Read the addends from the command line int b_in = atoi(argv[2]); int N = atoi(argv[3]); // Read the length of the vectors int gpuThreads = atoi(argv[4]); // Read the number of CUDA threads to use int *a, *b, *c; int *da, *db, *dc; int bytes = N *sizeof(int); a = (int *)malloc(bytes); b = (int *)malloc(bytes); c = (int *)malloc(bytes); for (int i=0; i<N; i++) { a[i] = a_in; b[i] = b_in; c[i] = 0; } cudaMalloc((void **)&da, bytes); cudaMalloc((void **)&db, bytes); cudaMalloc((void **)&dc, bytes); cudaMemcpy(da, a, bytes, cudaMemcpyHostToDevice); cudaMemcpy(db, b, bytes, cudaMemcpyHostToDevice); add<<<1,gpuThreads>>>(da, db, dc); cudaMemcpy(c, dc, bytes, cudaMemcpyDeviceToHost); cudaDeviceSynchronize(); printf("%d + %d -> %d\n", a[0], b[0], c[0]); printf(" ...\n"); printf("%d + %d -> %d\n", a[N-1], b[N-1], c[N-1]); cudaFree(da); cudaFree(db); cudaFree(dc); free(a); free(b); free(c); }
7,717
#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__ unsigned long long shiftCount[LIST_SIZE]; extern "C" __device__ unsigned long long shiftVal[LIST_SIZE]; extern "C" __device__ unsigned long long record_flag; void bambooLogRecordOff(){ long long local_record = 0; cudaMemcpyToSymbol(record_flag, &local_record, sizeof(long long), 0, cudaMemcpyHostToDevice); } void bambooLogKernelBegin(long long i) { i = 1; cudaMemcpyToSymbol(record_flag, &i, sizeof(long long), 0, cudaMemcpyHostToDevice); } void bambooLogKernelEnd() { #ifdef KERNELTRACE cudaDeviceSynchronize(); #endif unsigned long long shift_count[LIST_SIZE] = {0}; unsigned long long shift_val[LIST_SIZE] = {0}; cudaMemcpyFromSymbol(shift_count, shiftCount, LIST_SIZE * sizeof(long long), 0, cudaMemcpyDeviceToHost); cudaMemcpyFromSymbol(shift_val, shiftVal, LIST_SIZE * sizeof(long long), 0, cudaMemcpyDeviceToHost); FILE *profileFile = fopen("profile_shift_value_result.txt", "w"); for(long long i=0; i < LIST_SIZE; i++){ if(shift_count[i] != 0){ fprintf(profileFile, "%lld: %lld %lld\n", i, shift_val[i]/shift_count[i], shift_count[i]); } } fclose(profileFile); }
7,718
#include <stdio.h> __device__ static inline unsigned int argmin(unsigned int a, unsigned int b, unsigned int c) { if (a < b) { if (a < c) return 0; else return 2; } if (b < c) return 1; else return 2; } __global__ void matching(int *D, int *phi, int m, int n) { // matrix m x n int pnt = 2; const int tid = threadIdx.x;// + blockDim.x * blockIdx.x; if (tid == 0) { printf("%d", tid); int j = 0; for (j = 0; j < n; j++) { int tmp[3] = { D[(tid * n) + j], D[(tid+1)*n+j] + pnt, D[(tid * n) + j + 1] + pnt }; int arg = argmin(tmp[0], tmp[1], tmp[2]); int dmin = tmp[arg]; D[((tid+1) * n) + j + 1] = D[((tid+1) * n) + j + 1] + dmin; phi[(tid * n) + j] = arg + 1; } } else if (tid < m) { int j = 0; for (j = 0; j < n; j++) { while(1) { if (phi[(tid-1) * n + j]) { int tmp[3] = {D[(tid * n) + j], D[(tid+1)*n+j] + pnt, D[(tid * n) + j + 1] + pnt}; int arg = argmin(tmp[0], tmp[1], tmp[2]); int dmin = tmp[arg]; D[((tid+1) * n) + j + 1] = D[((tid+1) * n) + j + 1] + dmin; phi[(tid * n) + j] = arg + 1; break; } } } } }
7,719
#include <iostream> #include <chrono> #include <functional> #include <cmath> #include <stdio.h> using namespace std::chrono; typedef std::function<void*() > func; class Matrix; __global__ void cudaMatMulClass(const int* A, int an, int am, const int* B, int bn, int bm, int* out); class Matrix{ private: int m_dimX; int m_dimY; int* m_content; void deleteContent(){ if(m_content != NULL){ cudaFree(m_content); m_content = NULL; } } class helper{ public: int& operator[](int j){ return m_M->m_content[m_i*(m_M->dimY()) +j]; } helper(const Matrix* m, int i): m_M(m), m_i(i){} private: const Matrix* m_M; int m_i; }; // class helperC{ // public: // int operator[](int j){ // return m_M->m_content[m_i*j]; // } // helper(const Matrix* m, int i): m_M(m), m_i(i){} // private: // const Matrix* m_M; // int m_i; // }; public: Matrix():m_dimX(-1), m_dimY(-1), m_content(NULL){} Matrix(int dy, int dx, int val): m_dimX(dx), m_dimY(dy){ cudaMallocManaged(&m_content, m_dimX*m_dimY*sizeof(int)); for(int i=0; i<(m_dimY*m_dimX); i++){ m_content[i] = val; } } Matrix(int dy, int dx): m_dimX(dx), m_dimY(dy){ cudaMallocManaged(&m_content, m_dimX*m_dimY*sizeof(int)); } int nElem(){return m_dimX*m_dimY;} int dimX() const{return m_dimX;} int dimY() const{return m_dimY;} Matrix& operator=(const Matrix& other){ if(this != &other){ this->deleteContent(); this->m_content = other.m_content; this->m_dimX = other.m_dimX; this->m_dimY = other.m_dimY; } return *this; } void print(){ //TODO for(int i=0; i<m_dimY; i++){ for(int j=0; j<m_dimX; j++){ std::cout << (*this)[i][j] << ","; } std::cout << std::endl; } std::cout << "("<<m_dimY<<","<<m_dimX<<")"<<std::endl; } helper operator[] (const int i) const{ return helper(this, i); } Matrix matMul(const Matrix& other) const{ if( this->m_dimX != other.m_dimY ){ std::cout << "dimenstion don't fit " << std::endl; throw 0; } int blockSize = 256; int numBlocks = ((m_dimY * other.m_dimX) + blockSize - 1) / blockSize; Matrix ret(m_dimY, other.m_dimX); cudaMatMulClass<<<numBlocks,blockSize>>>(m_content, m_dimY, m_dimX, other.m_content, other.dimY(), other.dimX(), ret.m_content); //cudaMatMul<<<(2,2),(3,3)>>>(A, an, am, B, bn, bm, out); cudaDeviceSynchronize(); return ret; } }; __global__ void cudaMatMulClass(const int* A, int an, int am, const int* B, int bn, int bm, int* out){ int i,j,ij, k; int ijS = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x; //how many threads ber bock * how many blocks in the grid; //int jS = blockIdx.y * blockDim.y + threadIdx.y; //int jStride = blockDim.y * gridDim.y; //how many threads ber bock * how many blocks in the grid; if(blockIdx.x==0 && threadIdx.x==0){ printf("blockInd.x= %d ,blockDim.x= %d, threadIdx.x= %d, GridDim.x= %d\n", blockIdx.x ,blockDim.x , threadIdx.x ,gridDim.x); printf("blockInd.y= %d ,blockDim.y= %d, threadIdx.y= %d, GridDim.y= %d\n", blockIdx.y ,blockDim.y , threadIdx.y ,gridDim.y); printf("ijS= %d ,stride= %d \n", ijS, stride); } for(ij=ijS; ij < an*bm; ij+=stride){ i = ij/bm; j = ij%bm; out[ij]=0; for(k=0; k<am; k++){ out[ij] += A[i*an + k]*B[k*bn +j]; } //printf("(i:%d, j:%d)=%d \n", i,j, out[i][j]); } } __global__ void cudaMatMul(int** A, int an, int am, int** B, int bn, int bm, int** out){ int i,j,ij, k; int ijS = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x; //how many threads ber bock * how many blocks in the grid; //int jS = blockIdx.y * blockDim.y + threadIdx.y; //int jStride = blockDim.y * gridDim.y; //how many threads ber bock * how many blocks in the grid; if(blockIdx.x==0 && threadIdx.x==0){ printf("blockInd.x= %d ,blockDim.x= %d, threadIdx.x= %d, GridDim.x= %d\n", blockIdx.x ,blockDim.x , threadIdx.x ,gridDim.x); printf("blockInd.y= %d ,blockDim.y= %d, threadIdx.y= %d, GridDim.y= %d\n", blockIdx.y ,blockDim.y , threadIdx.y ,gridDim.y); printf("ijS= %d ,stride= %d \n", ijS, stride); } //std::cout << "blockInd.x="<< blockIdx.x << " blockDim.x=" << blockDim.x << " threadIdx.x=" << threadIdx.x << " GridDim.x" << gridDim.x << std::endl; //std::cout << "blockInd.y="<< blockIdx.y << " blockDim.y=" << blockDim.y << " threadIdx.y=" << threadIdx.y << " GridDim.y" << gridDim.y << std::endl; //for(i=iS; i < an; i+=iStride){ // for(j=0; j<bm; j++){ // // go threw all fields in output // // initialize output field as 0 // out[i][j]=0; // // for(k=0; k < am; k++){ // out[i][j] += A[i][k] * B[k][j]; // } // } //} //version with single loop for(ij=ijS; ij < an*bm; ij+=stride){ i = ij/bm; j = ij%bm; out[i][j]=0; for(k=0; k<am; k++){ out[i][j] += A[i][k]*B[k][j]; } //printf("(i:%d, j:%d)=%d \n", i,j, out[i][j]); } } int** matMul(int** A, int an, int am, int** B, int bn, int bm){ if( am != bn ){ std::cout << "dimenstion don't fit" << std::endl; throw 0; } int N = std::max(an, bm); int blockSize = 256; int numBlocks = ((an * bm) + blockSize - 1) / blockSize; int i; int **out; //out = new int*[an]; cudaMallocManaged(&out, an*sizeof(int*)); for(i=0; i < an ; i++){ //out[i] = new int[bm]; cudaMallocManaged(&out[i], bm*sizeof(int)); } cudaMatMul<<<numBlocks,blockSize>>>(A, an, am, B, bn, bm, out); //cudaMatMul<<<(2,2),(3,3)>>>(A, an, am, B, bn, bm, out); cudaDeviceSynchronize(); return out; } /// initialize matrix int** init(int an, int am, int value){ //int** out = new int*[an]; int** out; cudaMallocManaged(&out, an*sizeof(int*)); for(int i=0; i<an; i++){ //out[i] = new int[am]; cudaMallocManaged(&out[i], am*sizeof(int)); for(int j=0; j<am; j++){ out[i][j] = value; } } return out; } void print(int** A, int an, int am){ std::cout << "A = " << std::endl; for(int i=0; i<an; i++){ for(int j=0; j<am; j++){ std::cout << A[i][j] << ","; } std::cout << std::endl; } std::cout << "("<<an<<","<<am<<")"<<std::endl; } template <class retType> retType measureTime(func& f){ std::chrono::high_resolution_clock::time_point t1 = std::chrono::high_resolution_clock::now(); void* result = f(); std::chrono::high_resolution_clock::time_point t2 = std::chrono::high_resolution_clock::now(); auto duration = std::chrono::duration_cast<milliseconds>( t2 - t1 ).count(); std::cout << "execution took " << duration << " milliseconds" << std::endl; return (retType)result; } int main(){ std::cout << "gpu version" << std::endl; int A[2][2]={{1,2},{3,4}}; int B[2][3]={{1,1,1},{1,1,1}}; int **out; int **C = init(300, 500, 1); int **D = init(500, 900, 1); func f = [C,D](){return (void*)matMul(C, 300, 500, D, 500, 900);}; out=measureTime<int**>(f); std::cout << "the new one " << std:: endl; //print(out, 300, 900); Matrix MA(300,500,1); Matrix MB(500,900,1); //func f = [A,B](){return (void*)A.matMul(B);} //Matrix C = measureTime<Matrix>(f); Matrix MC = MA.matMul(MB); //MC.print(); return 0; }
7,720
#include <cuda.h> #include <stdio.h> #include <stdlib.h> __device__ int mandel(float c_re, float c_im, int maxIteration) { float z_re = c_re, z_im = c_im; int i; for (i = 0; i < maxIteration; ++i) { if (z_re * z_re + z_im * z_im > 4.f) break; float new_re = z_re * z_re - z_im * z_im; float new_im = 2.f * z_re * z_im; z_re = c_re + new_re; z_im = c_im + new_im; } return i; } __global__ void mandelKernel(float x1, float y1, float x0, float y0, int* output, int width, int height, int maxIterations, int pitch, int pixels) { // To avoid error caused by the floating number, use the following pseudo code // float x = lowerX + thisX * stepX; // float y = lowerY + thisY * stepY; int i = (blockIdx.x * blockDim.x + threadIdx.x) * pixels; int j = blockIdx.y * blockDim.y + threadIdx.y; // if (i >= width || j >= height) return; if (i >= width || j >= height) return; // i -> 1600, j -> 1200 float dx = (x1 - x0) / width; float dy = (y1 - y0) / height; float y = y0 + j * dy; for(int pixel = 0; pixel < pixels; pixel++) { float x = x0 + (i + pixel) * dx; int index = (j * pitch + i) + pixel; output[index] = mandel(x, y, maxIterations); } } // Host front-end function that allocates the memory and launches the GPU kernel #define N 1600 #define BLOCK_SIZE 64 void hostFE (float x1, float y1, float x0, float y0, int* output, int width, int height, int maxIterations) { int *h_img = NULL, *d_img = NULL; size_t pitch; // Locate CPU memory and GPU memory cudaHostAlloc((void**)&h_img, width * height * sizeof(int), cudaHostAllocDefault); cudaMallocPitch((void**)&d_img, &pitch, (size_t)width * sizeof(int), (size_t)height); // Copy memory from CPU to GPU cudaMemcpy2D(d_img, pitch, h_img, width * sizeof(int), width * sizeof(int), height, cudaMemcpyHostToDevice); // dim3 blockSize(BLOCK_SIZE, BLOCK_SIZE); dim3 numBlock(N / BLOCK_SIZE, N / BLOCK_SIZE); mandelKernel<<<blockSize, numBlock>>>(x1, y1, x0, y0, d_img, width, height, maxIterations, pitch / sizeof(int), 2); // Sync cudaDeviceSynchronize(); // // Copy memory from GPU to CPU cudaMemcpy2D(h_img, width * sizeof(int), d_img, pitch, width * sizeof(int), height, cudaMemcpyDeviceToHost); // // Copy memory from CPU to CPU(answers) memcpy(output, h_img, width * height * sizeof(int)); cudaFree(h_img); cudaFree(d_img); }
7,721
#include <iostream> #include <stdlib.h> #include <stdio.h> using namespace std; int main( int argc, char** argv ){ int args_needed = 1; if (argc < args_needed + 1 ){ printf(" Arg number error, needed: %d \n", args_needed); return 0; } printf(" CUDA - Template \n"); return 0; }
7,722
#define BLOCK_SIZE 32 __global__ void opticalFlowKernel(int* o_gx, int* o_gy, int* arr1dI, int* arr1dJ, int rows, int cols, int wx, int wy, int K) { __shared__ int I_tile[BLOCK_SIZE][BLOCK_SIZE]; __shared__ int J_tile[BLOCK_SIZE][BLOCK_SIZE]; int tx = threadIdx.x; int ty = threadIdx.y; int xidx = blockIdx.x * blockDim.x + threadIdx.x; int yidx = blockIdx.y * blockDim.y + threadIdx.y; I_tile[ty][tx] = arr1dI[yidx * cols + xidx]; J_tile[ty][tx] = arr1dJ[yidx * cols + xidx]; __syncthreads(); // Initialization of pyramidal guess float gx = 0, gy = 0; // Spatial gradient matrix float Gxx = 0, Gxy = 0, Gyy = 0; for(int dy = -wy; dy <= wy; dy++) { for(int dx = -wx; dx <= wx; dx++) { if(ty + dy - 1 >= 0 && ty + dy + 1 < BLOCK_SIZE && tx + dx - 1 >= 0 && tx + dx + 1 < BLOCK_SIZE) { int Ix = (I_tile[ty+dy][tx+dx+1] - I_tile[ty+dy][tx+dx-1]) / 2; int Iy = (I_tile[ty+dy+1][tx+dx] - I_tile[ty+dy-1][tx+dx]) / 2; Gxx += Ix * Ix; Gxy += Ix * Iy; Gyy += Iy * Iy; } } } float det = Gxx * Gyy - Gxy * Gxy; if(det > 0.00001f) { // Initialization of iterative L-K float vx = 0, vy = 0; for(int k = 0; k < K; k++) { int Jx = tx + gx + vx; int Jy = ty + gy + vy; if(Jx - 1 >= 0 && Jx + 1 < BLOCK_SIZE && Jy - 1 >= 0 && Jy + 1 < BLOCK_SIZE) { // Image mismatch vector float bkx = 0, bky = 0; for(int dy = -wy; dy <= wy; dy++) { for(int dx = -wx; dx <= wx; dx++) { if(ty + dy - 1 >= 0 && ty + dy + 1 < BLOCK_SIZE && tx + dx - 1 >= 0 && tx + dx + 1 < BLOCK_SIZE) { int Ik = I_tile[ty+dy][tx+dx] - J_tile[Jy][Jx]; int Ix = (I_tile[ty+dy][tx+dx+1] - I_tile[ty+dy][tx+dx-1]) / 2; int Iy = (I_tile[ty+dy+1][tx+dx] - I_tile[ty+dy-1][tx+dx]) / 2; bkx += Ik * Ix; bky += Ik * Iy; } } } // Optical Flow float eta_x = (-Gyy * bkx + Gxy * bky) / det; float eta_y = (Gxy * bkx - Gxx * bky) / det; // Guess for next iteration vx += eta_x; vy += eta_y; } } // Guess for next level gx += vx; gy += vy; } o_gx[yidx * cols + xidx] = gx; o_gy[yidx * cols + xidx] = gy; }
7,723
#include "includes.h" #define N 128*128 __global__ void kernelMontecarlo(float *x, float *y,int *contador) { //int i = threadIdx.x + blockIdx.x*blockDim.x; //int j = threadIdx.y + blockIdx.y*blockDim.y; int indice = threadIdx.x + blockIdx.x*blockDim.x; //int indice=i; //printf("Indice: %f\n",(x[indice]*x[indice] + y[indice]*y[indice])); if((x[indice]*x[indice] + y[indice]*y[indice]) <=1.0) { atomicAdd(contador,1);//contador++; //printf("Contador: %d\n",*contador); } }
7,724
#include <stdio.h> int DIM_LIM = 10; int MAT_COUNT = 20; int SEED = 15; //seed for rand class matrix { public: int row; //number of rows, y int col; //number of columns, x double* data; __host__ __device__ matrix(int columns, int rows) : col(columns), row(rows), data(new double[col * row]) {} __host__ __device__ double& getdata(int x, int y){ return data[y * col + x]; //vertical position * row length + pos in row }; }; __global__ void d_printMat(matrix *mat) { int dimxn = mat->col; int dimyn = mat->row; printf("Dim x %d, Dim y %d\n", dimxn, dimyn); for(int y = 0; y<dimyn; y++){ for(int x = 0; x<dimxn; x++){ printf("%lf ", mat->getdata(x,y)); } printf("\n"); } printf("\n"); } __host__ void printMat(matrix *mat) { int dimxn = mat->col; int dimyn = mat->row; printf("Dim x %d, Dim y %d\n", dimxn, dimyn); for(int y = 0; y<dimyn; y++){ for(int x = 0; x<dimxn; x++){ printf("%lf ", mat->getdata(x,y)); } printf("\n"); } printf("\n"); } __global__ void mat_mult(matrix *a, matrix *b, matrix *ans){ if(a->row == b->col){ int iter = a->row; //number of mults needed printf("a(%d, %d) b(%d,%d)\n",a->col,a->row,b->col,b->row); //ans = new matrix(a->col, b->row); printf("result %d rows %d cols\n", ans->row, ans->col); for(int x = 0; x < ans->col; x++){ for(int y = 0; y < ans->row; y++){ ans->getdata(x,y) = 0; //initialize for(int z = 0; z < iter; z++){ ans->getdata(x,y) += (a->getdata(x,y) * b->getdata(y,x)); } //printf("value at %d %d is %f\n", x,y, ans->getdata(x,y)); } } } else{ printf("matrix size mismatch"); } }; matrix** initialize(){ srand(SEED); //init random gen int dim[MAT_COUNT + 1]; //stores matrix sizes for(int z = 0; z <= MAT_COUNT; z++){ dim[z] = rand()%DIM_LIM + 1;//random between 1 and limit } //declare matrix array as pointer matrix **mat = (matrix **)malloc(MAT_COUNT * sizeof(matrix*)); for(int z = 0; z < MAT_COUNT; z++){ //each matrix shares a dimension with the previous int dimx = dim[z]; int dimy = dim[z+1]; mat[z] = new matrix(dimx,dimy); //dimx columns, dimy rows for(int x = 0; x<dimx; x++){ for(int y = 0; y<dimy; y++){ //TODO change to random double mat[z]->getdata(x,y) = 5; //initialize each element } } } return mat; } matrix* copyMatrixDev(matrix *host){ matrix *d_mat; double *tmp_data; cudaMalloc(&d_mat, sizeof(matrix)); cudaMemcpy(d_mat, host, sizeof(matrix), cudaMemcpyHostToDevice); cudaMalloc(&tmp_data, sizeof(double) * host->col * host->row); cudaMemcpy(tmp_data, host->data, sizeof(double) * host->col * host->row, cudaMemcpyHostToDevice); cudaMemcpy(&(d_mat->data),&tmp_data, sizeof(double *), cudaMemcpyHostToDevice); return d_mat; } matrix* hostMultMat(matrix *a, matrix *d_a, matrix *b, matrix *d_b){ matrix *result = new matrix(a->col,b->row); matrix *d_result = copyMatrixDev(result); cudaDeviceSynchronize(); mat_mult<<<1,1>>>(d_a,d_b,d_result); cudaDeviceSynchronize(); return d_result; } int main(){ matrix **mat_arr = initialize(); matrix *d_mat[MAT_COUNT]; double *mat_data[MAT_COUNT]; for(int i = 0; i < MAT_COUNT; i++){ cudaMalloc(&d_mat[i], sizeof(matrix)); cudaMemcpy(d_mat[i], mat_arr[i], sizeof(matrix), cudaMemcpyHostToDevice); cudaMalloc(&mat_data[i], sizeof(double) * mat_arr[i]->col * mat_arr[i]->row); cudaMemcpy(mat_data[i], mat_arr[i]->data, sizeof(double) * mat_arr[i]->col * mat_arr[i]->row, cudaMemcpyHostToDevice); cudaMemcpy(&(d_mat[i]->data),&mat_data[i], sizeof(double *), cudaMemcpyHostToDevice); // printMat(mat_arr[i]); // d_printMat<<<1,1>>>(d_mat[i]); } //matrix *d_result = hostMultMat(mat_arr[0], d_mat[0], mat_arr[1], d_mat[1]); //d_printMat<<<1,1>>>(d_result); //cudaDeviceSynchronize(); for(int i = 0; i < MAT_COUNT-1; i++){ matrix *d_result = hostMultMat(mat_arr[i], d_mat[i], mat_arr[i+1], d_mat[i+1]); d_printMat<<<1,1>>>(d_result); cudaFree(d_result); } }
7,725
#include "includes.h" // Taken from the NVIDIA "2_Graphics\simpleGL" sample: // A kernel that modifies the z-coordinates of a rectangular // grid of vertices, based on a time value, so that they // form an animated sine wave extern "C" __global__ void simple_vbo_kernel( float4 *pos, unsigned int width, unsigned int height, float time) { unsigned int x = blockIdx.x*blockDim.x + threadIdx.x; unsigned int y = blockIdx.y*blockDim.y + threadIdx.y; // calculate uv coordinates float u = x / (float) width; float v = y / (float) height; u = u*2.0f - 1.0f; v = v*2.0f - 1.0f; // calculate simple sine wave pattern float freq = 4.0f; float w = sinf(u*freq + time) * cosf(v*freq + time) * 0.5f; // write output vertex pos[y*width+x] = make_float4(u, w, v, 1.0f); }
7,726
#ifdef __INC_CUDA_VECTOR_UTIL_H__ #include "cuda_runtime.h" __device__ void Mat4x4_Mul(float *A, float *B, float *C) { C[0] = A[0]*B[0]+A[4]*B[1]+A[8]*B[2]+A[12]*B[3]; C[1] = A[0]*B[4]+A[4]*B[5]+A[8]*B[6]+A[12]*B[7]; C[2] = A[0]*B[8]+A[4]*B[9]+A[8]*B[10]+A[12]*B[11]; C[3] = A[0]*B[12]+A[4]*B[13]+A[8]*B[14]+A[12]*B[15]; C[4] = A[1]*B[0]+A[5]*B[1]+A[9]*B[2]+A[13]*B[3]; C[5] = A[1]*B[4]+A[5]*B[5]+A[9]*B[6]+A[13]*B[7]; C[6] = A[1]*B[8]+A[5]*B[9]+A[9]*B[10]+A[13]*B[11]; C[7] = A[1]*B[12]+A[5]*B[13]+A[9]*B[14]+A[13]*B[15]; C[8] = A[2]*B[0]+A[6]*B[1]+A[10]*B[2]+A[14]*B[3]; C[9] = A[2]*B[4]+A[6]*B[5]+A[10]*B[6]+A[14]*B[7]; C[10] = A[2]*B[8]+A[6]*B[9]+A[10]*B[10]+A[14]*B[11]; C[11] = A[2]*B[12]+A[6]*B[13]+A[10]*B[14]+A[14]*B[15]; C[12] = A[3]*B[0]+A[7]*B[1]+A[11]*B[2]+A[15]*B[3]; C[13] = A[3]*B[4]+A[7]*B[5]+A[11]*B[6]+A[15]*B[7]; C[14] = A[3]*B[8]+A[7]*B[9]+A[11]*B[10]+A[15]*B[11]; C[15] = A[3]*B[12]+A[7]*B[13]+A[11]*B[14]+A[15]*B[15]; } __device__ void Mat4x4_Mul_Vec4(float *A, float *B, float *C) { C[0] = A[0]*B[0]+A[4]*B[1]+A[8]*B[2]+A[12]*B[3]; C[1] = A[1]*B[0]+A[5]*B[1]+A[9]*B[2]+A[13]*B[3]; C[2] = A[2]*B[0]+A[6]*B[1]+A[10]*B[2]+A[14]*B[3]; C[3] = A[3]*B[0]+A[7]*B[1]+A[11]*B[2]+A[15]*B[3]; } #endif
7,727
/* * @author Connie Shi * Lab 3: Write a reduction program in CUDA that finds the maximum * of an array of M integers. * Part 3 (Improved): * Write a CUDA version that makes use of shared memory, * prefetching, and different granularities. Performs better * than original cudashared.cu version, because it does not * divide the data into subsets to sequential search. * * Should be run on cuda1 machine with 1024 max threads per block. */ #include <stdio.h> #include <stdlib.h> #include <time.h> #include <cuda.h> #define THREADS_PER_BLOCK 1024 #define WARP 32 #define MAX(a, b) ((a) > (b) ? (a) : (b)) /* Function Declarations */ void generate_random(int random[], int num_elements); __global__ void max_in_blocks(int random[], int num_elements); __device__ void sequential(int random[], int num_elements); /* Generates M random numbers from 1 to 100000*/ void generate_random(int random[], int num_elements) { int i; time_t t; srand((unsigned)time(&t)); //randomizes seed for (i = 0; i < num_elements; i++) { random[i] = (int)(((double)rand()/RAND_MAX)*100000); } } /* global function called from host and executed on kernel * Uses a tree-like structure to do parallel max reduction. * Avoids branch diversion, uses prefetching and shared memory. */ __global__ void max_in_blocks(int random[], int num_elements) { __shared__ int sdata[THREADS_PER_BLOCK]; unsigned int tid = threadIdx.x; unsigned int index = blockIdx.x * blockDim.x + threadIdx.x; int stride; // Take data from global memory to shared memory for faster access sdata[tid] = random[index]; __syncthreads(); for (stride = blockDim.x/2; stride >= 32; stride >>= 1) { if (tid < stride && tid + stride < num_elements) { int current = sdata[tid + stride]; if (sdata[tid] < current) { sdata[tid] = current; } } __syncthreads(); } // Prevents branch divergence because it will stop running // Through while loop when it reaches the size of the warp // At which point, the max is in the first 32 positions. // Sequential search 32 elements is very fast. if (tid < 32) { sequential(sdata, num_elements); random[blockIdx.x] = sdata[0]; } } /* Sequential searches through the first 32 positions of the block * to prevent further divvying up of the warp into different tasks. */ __device__ void sequential(int sdata[], int num_elements) { int i; int max = 0; int tid = threadIdx.x; for (i = tid; i < tid + WARP && i < num_elements; i++) { if (max < sdata[i]) { max = sdata[i]; } } // Put in index position, first element of the block sdata[0] = max; } /**************************************************************/ int main(int argc, char*argv[]) { int* h_random; int* d_random; int i; int largest = 0; clock_t start, end; if (argc != 2) { printf("Invalid number of commands: usage ./cudadivshared M\n"); exit(1); } // Generate array of random elements int num_elements = atoi(argv[1]); h_random = (int*)malloc(sizeof(int) * num_elements); generate_random(h_random, num_elements); start = clock(); // Calculation for grid dimensions int leftover = num_elements % WARP; int d_elements = num_elements - leftover; int n_blocks = (int)ceil((double)d_elements/THREADS_PER_BLOCK); int n_threads = (d_elements > THREADS_PER_BLOCK) ? THREADS_PER_BLOCK : d_elements; // Allocate space on device and copy over elements cudaError_t err = cudaMalloc((void**)&d_random, sizeof(int) * d_elements); if (err != cudaSuccess) { printf("cudaMalloc failure\n"); } err = cudaMemcpy(d_random, h_random, sizeof(int) * d_elements, cudaMemcpyHostToDevice); if (err != cudaSuccess) { printf("cudaMemcpy failure\n"); } // Execute kernel max_in_blocks<<<n_blocks, n_threads>>>(d_random, d_elements); // While kernel is executing, find the max in leftover elements for (i = d_elements; i < num_elements; i++) { if (largest < h_random[i]) { largest = h_random[i]; } } // Retrieve reduction results, only the first n_blocks element cudaMemcpy(h_random, d_random, sizeof(int) * n_blocks, cudaMemcpyDeviceToHost); // Check through n_blocks elements for the max for (i = 0; i < n_blocks; i ++) { if (largest < h_random[i]) { largest = h_random[i]; } } end = clock(); printf("Time to find max %f\n", (double)(end-start)/CLOCKS_PER_SEC); printf("Largest: %d\n", largest); // Clean up resources cudaFree(d_random); free(h_random); }
7,728
#include "includes.h" //#define NDEBUG const static float eps = 1e-6; const static size_t blocSize = 8; const static size_t size = 1024; __global__ void matMultiply2D(float* matA, float* matB, float* Dest, int dimensions) { int ix = threadIdx.x + blockIdx.x*blockDim.x; int iy = threadIdx.y + blockIdx.y*blockDim.y; if (ix < dimensions&&iy < dimensions) { float res = 0.0f; for (unsigned k = 0; k != dimensions; ++k) { res += matA[ix*dimensions + k] * matB[k*dimensions + iy]; } Dest[ix*dimensions + iy] = res; } }
7,729
#include <stdint.h> #include <stdlib.h> #include <stdio.h> #include <string.h> #include <stdbool.h> #include <time.h> #define ALPHABET_LEN 4 #define NOT_FOUND patlen #define max(a, b) ((a < b) ? b : a) void make_delta1(int *delta1, int8_t *pat, int32_t patlen) { int i; for (i=0; i < ALPHABET_LEN; i++) { delta1[i] = NOT_FOUND; } for (i=0; i < patlen-1; i++) { delta1[pat[i]] = patlen-1 - i; } } int is_prefix(int8_t *word, int wordlen, int pos) { int i; int suffixlen = wordlen - pos; for (i=0; i < suffixlen; i++) { if (word[i] != word[pos+i]) { return 0; } } return 1; } int suffix_length(int8_t *word, int wordlen, int pos) { int i; for (i = 0; (word[pos-i] == word[wordlen-1-i]) && (i < pos); i++); return i; } void make_delta2(int *delta2, int8_t *pat, int32_t patlen) { int p; int last_prefix_index = 1; for (p=patlen-1; p>=0; p--) { if (is_prefix(pat, patlen, p+1)) { last_prefix_index = p+1; } delta2[p] = (patlen-1 - p) + last_prefix_index; } for (p=0; p < patlen-1; p++) { int slen = suffix_length(pat, patlen, p); if (pat[p - slen] != pat[patlen-1 - slen]) { delta2[patlen-1 - slen] = patlen-1 - p + slen; } } } __device__ int d_retval; __global__ void boyer_moore (int8_t *string, int32_t stringlen, int8_t *pat, int32_t patlen, int *delta1, int *delta2, int n) { int i; d_retval = -1; int tid = blockIdx.x*blockDim.x+threadIdx.x; __syncthreads(); if (tid<n) { int beg = tid*patlen; int end = min (beg+(2*patlen), stringlen); i = beg+patlen-1; while (i < end) { int j = patlen-1; while (j >= 0 && (string[i] == pat[j])) { //printf("here in loop\n"); --i; --j; } if (j < 0) { d_retval = i+1; break; } i += max(delta1[string[i]], delta2[j]); } } } int8_t h_string[1000000]; int8_t h_pat[100]; int main(int argc, char const *argv[]) { int8_t *d_s, *d_p; int *d_d1, *d_d2; int32_t strlen = 1000000; int32_t patlen = 100; srand(time(NULL)); int i; char con [] = "ACGT"; for(i=0;i<strlen;i++) h_string[i] = rand ()%4; int patid = rand ()%10000; for(i=0;i<patlen;i++) h_pat[i] = h_string[patid++]; printf("The String is: "); for (i=0;i<strlen;i++) printf("%c", con[h_string[i]]); printf("\nThe search keyword is: "); for (i=0;i<patlen;i++) printf("%c", con[h_pat[i]]); int delta1[ALPHABET_LEN]; int delta2[patlen]; make_delta1(delta1, h_pat, patlen); make_delta2(delta2, h_pat, patlen); cudaMalloc(&d_s, strlen*sizeof(int8_t)); cudaMemcpy(d_s, h_string,strlen*sizeof(int8_t),cudaMemcpyHostToDevice); cudaMalloc(&d_p, patlen*sizeof(int8_t)); cudaMemcpy(d_p, h_pat,patlen*sizeof(int8_t),cudaMemcpyHostToDevice); cudaMalloc(&d_d1, ALPHABET_LEN*sizeof(int)); cudaMemcpy(d_d1, delta1,ALPHABET_LEN*sizeof(int),cudaMemcpyHostToDevice); cudaMalloc(&d_d2, patlen*sizeof(int)); cudaMemcpy(d_d2, delta2,patlen*sizeof(int),cudaMemcpyHostToDevice); int n = strlen/patlen; int block_size = 1024; int n_blocks = n/block_size + (n%block_size==0?0:1); boyer_moore<<<n_blocks,block_size>>>(d_s, strlen, d_p, patlen, d_d1, d_d2, n); cudaDeviceSynchronize(); int answer; cudaMemcpyFromSymbol(&answer, d_retval, sizeof(int), 0, cudaMemcpyDeviceToHost); printf("\nString found at %d\n", answer); return 0; }
7,730
#include <stdio.h> #include <stdlib.h> #include <math.h> #include <iostream> #include <time.h> #include <stdio.h> #include <stdlib.h> #define N 5 #define BLOCK_DIM 10 using namespace std; __global__ void adicion_Matrices (int *a, int *b, int *c) { int col = blockIdx.x * blockDim.x + threadIdx.x; int fila = blockIdx.y * blockDim.y + threadIdx.y; int index = col + fila * N; if (col < N && fila < N) { c[index] = a[index] + b[index]; } } __global__ void matrix_vector (int *a, int *b, int *c) { int col = blockIdx.x * blockDim.x + threadIdx.x; int sum = 0; if (col < N) { for(int i=0;i<N;i++){ sum += b[i]*a[(i*N)+col]; } c[col] = sum; } } __global__ void adicion_matrices_Filas (int *a, int *b, int *c) { int col = blockIdx.x * blockDim.x + threadIdx.x; int fila = blockIdx.y * blockDim.y + threadIdx.y; for(int i=col; i<N; i++){ int index = i + fila * N; c[index] = a[index] + b[index]; } } __global__ void adicion_matrices_Columnas (int *a, int *b, int *c) { int col = blockIdx.x * blockDim.x + threadIdx.x; int fila = blockIdx.y * blockDim.y + threadIdx.y; for(int i=fila; i<N; i++){ int index = col + i * N; c[index] = a[index] + b[index]; } } void imprimir(int matrix[N][N]){ for(int i=0;i<N;i++){ for(int j=0; j<N; j++){ std::cout << matrix[i][j] << '\t'; } std::cout << std::endl; } } void print_vector(int vector[N]){ for(int j=0; j<N; j++){ std::cout << vector[j] << '\t'; } } void adicion_matrix_to_Kernel(){ int a[N][N], b[N][N], c[N][N]; int *dev_a, *dev_b, *dev_c; int size = N * N * sizeof(int); srand(time(NULL)); for(int i=0; i<N; i++) for (int j=0; j<N; j++){ a[i][j] = rand() % 3; b[i][j] = rand() % 3; } imprimir(a); std::cout << std::endl; imprimir(b); cudaMalloc((void**)&dev_a, size); cudaMalloc((void**)&dev_b, size); cudaMalloc((void**)&dev_c, size); cudaMemcpy(dev_a, a, size, cudaMemcpyHostToDevice); cudaMemcpy(dev_b, b, size, cudaMemcpyHostToDevice); dim3 dimBlock(BLOCK_DIM, BLOCK_DIM); dim3 dimGrid((N+dimBlock.x-1)/dimBlock.x, (N+dimBlock.y-1)/dimBlock.y); adicion_Matrices<<<dimGrid,dimBlock>>>(dev_a,dev_b,dev_c); cudaDeviceSynchronize(); cudaMemcpy(c, dev_c, size, cudaMemcpyDeviceToHost); std::cout << std::endl; imprimir (c); /* for(int i=0; i<N; i++){ for (int j=0; j<N; j++){ printf("%d\t", c[i][j] ); } printf("\n"); } */ } int main() { int a[N][N], b[N], c[N]; int *dev_a, *dev_b, *dev_c; int size = N * N * sizeof(int); srand(time(NULL)); for(int i=0; i<N; i++){ for (int j=0; j<N; j++){ a[i][j] = rand() % 9; } b[i] = rand() % 9; } imprimir(a); std::cout << std::endl; print_vector(b); cudaMalloc((void**)&dev_a, size); cudaMalloc((void**)&dev_b, N * sizeof(int)); cudaMalloc((void**)&dev_c, N* sizeof(int)); cudaMemcpy(dev_a, a, size, cudaMemcpyHostToDevice); cudaMemcpy(dev_b, b, N*sizeof(int), cudaMemcpyHostToDevice); matrix_vector<<<N/256+1,256>>>(dev_a,dev_b,dev_c); cudaMemcpy(c, dev_c, size, cudaMemcpyDeviceToHost); std::cout << std::endl; print_vector(c); return 0; }
7,731
// Matrix multiplication using CUDA // Based on code from: https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#shared-memory #include <stdio.h> #include <string.h> #include <cuda.h> #include "cuda_runtime.h" #include "device_launch_parameters.h" #include "math.h" #include "time.h" #include <iostream> #include <fstream> #include <iomanip> #define BLOCK_SIZE 64 //CPU matrix multiplication __host__ void cpu_matrix_mult(float *h_a, float *h_b, float *h_result, int m) { for (int i = 0; i < m; ++i) { for (int j = 0; j < m; ++j) { float tmp = 0.0; for (int h = 0; h < m; ++h) { tmp += h_a[i * m + h] * h_b[h * m + j]; } h_result[i * m + j] = tmp; } } } // GPU matrix multiplication // Uses square matrices // Based on code from: https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#shared-memory __global__ void gpu_multiply(float *left, float *right, float *res, int dim) { float temp = 0; int row = blockIdx.y * blockDim.y + threadIdx.y; int col = blockIdx.x * blockDim.x + threadIdx.x; for (int e = 0; e < dim; ++e) temp += left[row * dim + e] * right[e * dim + col]; res[row * dim + col] = temp; } // Function to fill a non-square matrix with zeros to make it a square matrix __host__ int fill(float **Lmatrix, float **Rmatrix, int LdimX, int LdimY, int RdimX, int RdimY) { int sqr_dim_X, sqr_dim_Y, size; sqr_dim_X = RdimX; if (LdimX > RdimX) { sqr_dim_X = LdimX; } sqr_dim_Y = RdimY; if (LdimY > RdimY) { sqr_dim_Y = LdimY; } size = sqr_dim_Y; if (sqr_dim_X > sqr_dim_Y) { size = sqr_dim_X; } int temp = size / BLOCK_SIZE + (size % BLOCK_SIZE == 0 ? 0 : 1); size = temp * BLOCK_SIZE; size_t pt_size = size * size * sizeof(float); *Lmatrix = (float *)malloc(pt_size); *Rmatrix = (float *)malloc(pt_size); memset(*Lmatrix, 0, pt_size); memset(*Rmatrix, 0, pt_size); for (int i = 0; i < LdimX; i++) { for (int j = 0; j < LdimY; j++) { int dummy = size * i + j; (*Lmatrix)[dummy] = sinf(dummy); } } for (int i = 0; i < RdimX; i++) { for (int j = 0; j < RdimY; j++) { int dummy = size * i + j; (*Rmatrix)[dummy] = cosf(dummy); } } return size; } // GPU matrix multiplication - optimized // uses shared memory // based on code from: https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#shared-memory __global__ void multiply(float *left, float *right, float *res, int dim) { int i, j; float temp = 0; __shared__ float Left_shared_t[BLOCK_SIZE][BLOCK_SIZE]; __shared__ float Right_shared_t[BLOCK_SIZE][BLOCK_SIZE]; // Row i of matrix left int row = blockIdx.y * blockDim.y + threadIdx.y; int col = blockIdx.x * blockDim.x + threadIdx.x; for (int tileNUM = 0; tileNUM < gridDim.x; tileNUM++) { // Column j of matrix left j = tileNUM * BLOCK_SIZE + threadIdx.x; i = tileNUM * BLOCK_SIZE + threadIdx.y; // Load left[i][j] to shared mem Left_shared_t[threadIdx.y][threadIdx.x] = left[row * dim + j];// Coalesced access // Load right[i][j] to shared mem Right_shared_t[threadIdx.y][threadIdx.x] = right[i * dim + col]; // Coalesced access // Synchronize before computation __syncthreads(); // Accumulate one tile of res from tiles of left and right in shared mem for (int k = 0; k < BLOCK_SIZE; k++) { temp += Left_shared_t[threadIdx.y][k] * Right_shared_t[k][threadIdx.x]; //no shared memory bank conflict } // Synchronize __syncthreads(); } // Store accumulated value to res res[row * dim + col] = temp; } int main(void) { // Matrix sizes int Left_matrix_x, Left_matrix_y, Right_matrix_x, Right_matrix_y, Left_vector_size, Right_vector_size; // Declaring pointers for arrays float *Left_Vector_h, *Right_Vector_h, *Left_Vector_d, *Right_Vector_d, *Res_h, *Res_d, *CPU; printf("Enter m n n k :\n"); scanf("%d %d %d %d", &Left_matrix_x, &Left_matrix_y, &Right_matrix_x, &Right_matrix_y); int dim = fill(&Left_Vector_h, &Right_Vector_h, Left_matrix_x, Left_matrix_y, Right_matrix_x, Right_matrix_y); size_t vector_size; vector_size = dim * dim * sizeof(float); //Allocate memory on host Res_h = (float *)malloc(vector_size); CPU = (float *)malloc(vector_size); //Allocate memory on device cudaMalloc((void **)&Left_Vector_d, vector_size); cudaMalloc((void **)&Right_Vector_d, vector_size); cudaMalloc((void **)&Res_d, vector_size); //Copy values to device cudaMemcpy(Left_Vector_d, Left_Vector_h, vector_size, cudaMemcpyHostToDevice); //Copy values to device cudaMemcpy(Right_Vector_d, Right_Vector_h, vector_size, cudaMemcpyHostToDevice); //Block dimension is directly from block_size dim3 Block_dim(BLOCK_SIZE, BLOCK_SIZE); //Grid dimension is found by dividing matrix dimension to block_size dim3 Grid_dim(dim / BLOCK_SIZE, dim / BLOCK_SIZE); cudaEvent_t start, stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start, 0); //GPU Kernel call - optimized matrix multiplication multiply <<< Grid_dim, Block_dim >>> (Left_Vector_d, Right_Vector_d, Res_d, dim); cudaEventRecord(stop, 0); cudaEventSynchronize(stop); float et; cudaEventElapsedTime(&et, start, stop); cudaEventDestroy(start); cudaEventDestroy(stop); //Retrieve result from device and store it in host array cudaMemcpy(Res_h, Res_d, vector_size, cudaMemcpyDeviceToHost); clock_t begin = clock(); //CPU Matrix multiplication cpu_matrix_mult(Left_Vector_h, Right_Vector_h, CPU, dim); clock_t end = clock(); double time_spent = (double)1000 * (end - begin) / CLOCKS_PER_SEC; //Block dimension is directly from block_size dim3 Block_dim_1(BLOCK_SIZE, BLOCK_SIZE); //Grid dimension is found by dividing matrix dimension to block_size dim3 Grid_dim_1(dim / BLOCK_SIZE, dim / BLOCK_SIZE); //commented out the functions which helps to calculate time cudaEvent_t start1, stop1; cudaEventCreate(&start1); cudaEventCreate(&stop1); cudaEventRecord(start1, 0); //GPU Kernel call - non optimized matrix multiplication gpu_multiply <<< Grid_dim_1, Block_dim_1 >>> (Left_Vector_d, Right_Vector_d, Res_d, dim); cudaEventRecord(stop1, 0); cudaEventSynchronize(stop1); float mm; cudaEventElapsedTime(&mm, start1, stop1); cudaEventDestroy(start1); cudaEventDestroy(stop1); printf("GPU time= %f ms\n", et); printf("CPU time= %lf ms\n", time_spent); printf("GPU (not optimized) time = %f ms\n", mm); //Cleanup free(Left_Vector_h); free(Right_Vector_h); free(Res_h); free(CPU); cudaFree(Left_Vector_d); cudaFree(Right_Vector_d); cudaFree(Res_d); }
7,732
#include <stdio.h> #include <cuda_runtime.h> #include <unistd.h> #define N 20*1024*1024 #define THREADS_PER_BLOCK 512 #ifdef __cplusplus #define restrict __restrict__ #endif __global__ void gpu_vector_add(int* a, int* b, int* c) { int index = threadIdx.x + blockIdx.x*blockDim.x; c[index] = a[index] + b[index]; } void cpu_vector_add(int* restrict a, int* restrict b, int* restrict c) { int i; for (i = 0; i < N; ++i) c[i] = a[i] + b[i]; } int main() { //static int a[N], b[N], c[N]; /*Variáveis na memória principal*/ int *a, *b, *c; int *d_a, *d_b, *d_c; /*Variáveis que alocaremos na memória da GPU*/ int i; size_t size = N*sizeof(int); int err; a = (int*) malloc(size); b = (int*) malloc(size); c = (int*) malloc(size); err = cudaMalloc((void **)&d_a, size); /*Aloque um inteiro na memória de vídeo e faça d_a apontar para ele.*/ if (err != cudaSuccess) { return 1; } err = cudaMalloc((void **)&d_b, size); if (err != cudaSuccess) { return 1; } err = cudaMalloc((void **)&d_c, size); if (err != cudaSuccess) { return 1; } for (i = 0; i < N; ++i) a[i] = 42; for (i = 0; i < N; ++i) b[i] = 1337; /*Copie a e b para os seus respectivos espaços alocados na GPU*/ cudaMemcpy(d_a, a, size, cudaMemcpyHostToDevice); /* O que é isso? * a flag cudaMemcpyHostToDevice * é parte de uma enum que especifica * o fluxo de dados. HostToDevice * especifica que copiaremos os * dados da RAM para a GRAM */ cudaMemcpy(d_b, b, size, cudaMemcpyHostToDevice); gpu_vector_add<<<N/THREADS_PER_BLOCK, THREADS_PER_BLOCK>>>(d_a, d_b, d_c); /*N = número de blocos. 1 bloco = um conjunto de threads*/ cudaDeviceSynchronize(); cudaMemcpy(c, d_c, size, cudaMemcpyDeviceToHost); printf("Resultado: %d\n", c[0]); /*Libera a memória na placa. E se eu não liberar?*/ cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); free(a); free(b); free(c); return 0; }
7,733
#include <stdio.h> #include <cuda.h> #include <cuda_runtime.h> #include <sys/time.h> #define size 4096 #define row 16 #define loop 256 //row is size/(32*sqrt(brock_num)) and loop % 2^2n == 0 loop = row^2 #define N 100 #define r 0.1 #define blocksize 1024 //size is a grid size and N is how many times to calculate #define block_num 64 __global__ void calc(float *u1, float *u2, int n){ int i2 = threadIdx.x; int i3 = blockIdx.x; int i4 = blockDim.x; for(int j = 0; j < loop; j++){ //int i = i2 + blockrow*(j%row) + (j/row) * blocksize * row; int i = i2 + i4 * j + i4*loop*i3; i = i + ((i/size)*2) + 1 + (size + 2); if(n%2==0){ float x1 = u1[i]; float x2 = u1[i+1]; float x3 = u1[i-1]; float x4 = u1[i+(size+2)]; float x5 = u1[i-(size+2)]; //__syncthreads(); u2[i] = (1 - 4*r) * x1 + r * (x2 + x3 + x4 + x5); __syncthreads(); } else { float x1 = u2[i]; float x2 = u2[i+1]; float x3 = u2[i-1]; float x4 = u2[i+(size+2)]; float x5 = u2[i-(size+2)]; //__syncthreads(); u1[i] = (1 - 4*r) * x1 + r * (x2 + x3 + x4 + x5); __syncthreads(); } } return ; } int main(){ float u_odd [(size+2)*(size+2)]; float u_even [(size+2)*(size+2)]; float u_result [(size+2)*(size+2)]; float* u1; float* u2; for(int i = 0; i < (size+2)*(size+2); i++){ u_even[i] = 0; u_odd[i] = 0; u_result[i] = 0; } //initialize u_even for(int i = 0; i < (size+2)*(size+2); i++){ if(i % (size+2) != 0 && i % (size+2) != (size+1) && i/(size+2) != 0 && i/(size+2) != (size+1)){ u_even[i] = 1; } } struct timeval t0,t1; const int csize = (size+2)*(size+2)*sizeof(float); cudaMalloc((void**)&u1,csize); cudaMalloc((void**)&u2,csize); if (N%2 == 0){ cudaMemcpy(u1,u_even,csize,cudaMemcpyHostToDevice); cudaMemcpy(u2,u_odd,csize,cudaMemcpyHostToDevice); }else{ cudaMemcpy(u2,u_even,csize,cudaMemcpyHostToDevice); cudaMemcpy(u1,u_odd,csize,cudaMemcpyHostToDevice); } dim3 dimBlock(blocksize,1); dim3 dimGrid(block_num,1); //時間ぶんloopする gettimeofday(&t0, NULL); for (int i = N; i > 0; i--){ calc<<<dimGrid,dimBlock>>>(u1,u2,i); cudaThreadSynchronize(); } gettimeofday(&t1, NULL); cudaMemcpy(u_result,u1,csize,cudaMemcpyDeviceToHost); cudaFree(u1); cudaFree(u2); printf("Elapsed time = %lf\n",((double)(t1.tv_sec - t0.tv_sec) + (double)(t1.tv_usec - t0.tv_usec)*1.0e-6)); /* for(int i = 0; i < (size+2)*(size+2); i++){ if (i % (size+2) != 0 && i % (size+2) != (size+1) && i/(size+2) != 0 && i/(size+2) != (size+1)){ printf("%.3f ",u_result[i]); } if(i % (size+2) == (size+1)){ printf("\n"); } } */ return EXIT_SUCCESS; }
7,734
// 使用GPU输出Hello world #include <stdio.h> #include "cuda_runtime.h" #include "device_launch_parameters.h" __global__ void helloFromGPU (void) { printf("Hello World from GPU!\n"); } int main(void) { // hello from cpu printf("Hello World from CPU!\n"); for(int i=0;i<1000;i++) { helloFromGPU <<<1, 100>>>(); cudaDeviceReset(); } return 0; }
7,735
//A1.cu #include <stdio.h> #include <stdlib.h> #include <cuda.h> #include <cuda_runtime.h> __global__ void A1_kernel(double* r, double* v, double dt) { size_t id = blockIdx.x * blockDim.x + threadIdx.x; r[id] += v[id] * dt; } extern "C" { void A1(double* r_h, double* v_h, double dt, int numParticles) { size_t N = 3 * numParticles; size_t N_bytes = N * sizeof(double); double *r_d, *v_d; cudaMalloc(&r_d, N_bytes); cudaMalloc(&v_d, N_bytes); cudaMemcpy(r_d, r_h, N_bytes, cudaMemcpyHostToDevice); cudaMemcpy(v_d, v_h, N_bytes, cudaMemcpyHostToDevice); A1_kernel<<<N, 3>>>(r_d, v_d, dt); cudaMemcpy(r_h, r_d, N_bytes, cudaMemcpyDeviceToHost); cudaFree(r_d); cudaFree(v_d); } }
7,736
#include "includes.h" // Save wave state as binary file. __global__ void fillSpaceTSteps(int N, int T, float c, float dt, float dd, float *waveSpace, float *waveSpaceTMin1, float *waveSpaceTMin2) { int i = blockIdx.y * blockDim.y + threadIdx.y; int j = blockIdx.x * blockDim.x + threadIdx.x; waveSpace[N * i + j] = 2 * waveSpaceTMin1[N * i + j] - waveSpaceTMin2[N * i + j] + (c * c) * (dt/dd * dt/dd) * (waveSpaceTMin1[N * (i + 1) + j] + waveSpaceTMin1[N * (i - 1) + j] + waveSpaceTMin1[N * i + (j - 1)] + waveSpaceTMin1[N * i + (j + 1)] - 4 * waveSpaceTMin1[N * i + j]); __syncthreads(); }
7,737
#include <stdio.h> #include <stdlib.h> #include <cuda.h> __global__ void revArray(int N, float *a) { int n = threadIdx.x + blockIdx.x*blockDim.x; if(n<N/2) { float a1 = a[n]; a[n] = a[N-1-n]; a[N-1-n] = a1; } } int main(int argc, char **argv) { int N = 100; //Host memory allocation float *h_a = (float*) malloc(N*sizeof(float)); float *h_b = (float*) malloc(N*sizeof(float)); int n; for(n=0;n<N;n++) { h_a[n] = 1+n; } // Device memory allocation float *d_a; cudaMalloc(&d_a, N*sizeof(float)); // Copy data from host to device cudaMemcpy(d_a, h_a, N*sizeof(float), cudaMemcpyHostToDevice); //save this for later int NthreadsPerBlock = 10; int NthreadBlocks = ((N/2)+NthreadsPerBlock-1)/NthreadsPerBlock ; revArray<<<NthreadBlocks, NthreadsPerBlock>>>(N,d_a); //copy result from device to host cudaMemcpy(h_a, d_a, N*sizeof(float), cudaMemcpyDeviceToHost); for(n=0;n<N;++n) { printf("h_a[%d] = %g\n",n,h_a[n]); } free(h_a); cudaFree(d_a); return 0; }
7,738
#include "includes.h" __global__ void sum_of_mults_kernel(float *a1, float *a2, float *b1, float *b2, size_t size, float *dst) { const int index = blockIdx.x*blockDim.x + threadIdx.x; if (index < size) { dst[index] = a1[index] * a2[index] + b1[index] * b2[index]; } }
7,739
#include <cuda_runtime.h> #include <device_launch_parameters.h> #include <cufft.h> #include <stdlib.h> #include <stdio.h> #include <time.h> #define TPBx 32 // TPBx * TPBy = number of threads per block #define TPBy 32 __global__ void real2complex(cufftComplex *c, float *a, int n); __global__ void complex2real_scaled(float *a, cufftComplex *c, float scale, int n); __global__ void solve_poisson(cufftComplex *c, float *kx, float *ky, int n); void exportData(const char *file, const float *X, const float *Y, const float *Z, const int n); void gaussian(float *bin, const float *X, const float *Y, const int n); int main(){ ///////////////////////////// INITIZALIZATION //////////////////////////// int N, R; printf("Phase 1: Set Up The Environment for Testing\n"); printf("Input the range of x and y: "); // the range of x and y will be from -R to R scanf("%d", &R); printf("Input the number of samples: "); // the number of samples will be N * N scanf("%d", &N); printf("Allocating memory...\n"); clock_t startTime11 = clock(); char *uFile = (char *)"u_data.dat"; char *rFile = (char *)"r_data.dat"; float *X = (float *)malloc(sizeof(float) * N); float *Y = (float *)malloc(sizeof(float) * N); float *kx = (float *)malloc(sizeof(float) * N); float *ky = (float *)malloc(sizeof(float) * N); float *r = (float *)malloc(sizeof(float) * N * N); float *u = (float *)malloc(sizeof(float) * N * N); const float EPSILON = 8.85418782 * pow(10, -12); // Permitivity of free space const float PI = 4 * atan(1); float *kx_d, *ky_d, *r_d; cufftComplex *r_complex_d; cudaMalloc((void **)&kx_d, sizeof(float) * N); cudaMalloc((void **)&ky_d, sizeof(float) * N); cudaMalloc((void **)&r_d, sizeof(float) * N * N); cudaMalloc((void **)&r_complex_d, sizeof(cufftComplex) * N * N); int m = 0; for(int i = N/-2; i < N/2; i++){ if(m < N){ X[m] = i * (float)R / (N / 2); Y[m] = -1 * i * (float)R / (N / 2); kx[m] = i * PI * N / (float)R; // Centers kx values to be at the origin ky[m] = - i * PI * N / (float)R; // Centers ky values to be at the origin } m += 1; } clock_t endTime11 = clock(); clock_t startTime12 = clock(); gaussian(r, X, Y, N); // Generate a Gaussian Distribution for r clock_t endTime12 = clock(); for (int i = 0; i < N * N; i++){ u[i] = 0.f; } double totalTime11 = (double)(endTime11 - startTime11) / CLOCKS_PER_SEC; double totalTime12 = (double)(endTime12 - startTime12) / CLOCKS_PER_SEC; printf("Phase 1 ended\n"); printf("Time spent on allocating memory: %f sec\n", totalTime11); printf("Time spent on generating function: %f sec\n\n", totalTime12); ////////////////////////////////////////////////////////////////////////// printf("Phase 2: Evaluation\n"); printf("Copying data from the host to the device...\n"); clock_t startTime21 = clock(); cudaMemcpy(kx_d, kx, sizeof(float) * N, cudaMemcpyHostToDevice); cudaMemcpy(ky_d, ky, sizeof(float) * N, cudaMemcpyHostToDevice); cudaMemcpy(r_d, r, sizeof(float) * N * N, cudaMemcpyHostToDevice); cufftHandle plan; cufftPlan2d(&plan, N, N, CUFFT_C2C); // Compute the execution configuration dim3 dimBlock(TPBx, TPBy); dim3 dimGrid(N / dimBlock.x, N / dimBlock.y); // Handle N not multiple of TPBx or TPBy if(N % TPBx != 0){ dimGrid.x += 1; } if(N % TPBy != 0){ dimGrid.y += 1; } clock_t endTime21 = clock(); printf("Start to solve the Poisson equation...\n"); clock_t startTime22 = clock(); real2complex<<<dimGrid, dimBlock>>>(r_complex_d, r_d, N); cufftExecC2C(plan, r_complex_d, r_complex_d, CUFFT_FORWARD); solve_poisson<<<dimGrid, dimBlock>>>(r_complex_d, kx_d, ky_d, N); cufftExecC2C(plan, r_complex_d, r_complex_d, CUFFT_INVERSE); float scale = 1.f / (EPSILON * N * N); complex2real_scaled<<<dimGrid, dimBlock>>>(r_d, r_complex_d, scale, N); clock_t endTime22 = clock(); clock_t startTime23 = clock(); cudaMemcpy(u, r_d, sizeof(float) * N * N, cudaMemcpyDeviceToHost); clock_t endTime23 = clock(); printf("Phase 2 ended\n"); double totalTime21 = (double)(endTime22 - startTime22) / CLOCKS_PER_SEC; double totalTime22 = (double)(endTime21 + endTime23 - startTime21 - endTime23) / CLOCKS_PER_SEC; printf("Time spent on calculation: %f sec\n", totalTime21); printf("Data spent on data transfer: %f sec\n\n", totalTime22); printf("Phase 3: Data Exportation"); printf("Exporting data...\n"); clock_t startTime31 = clock(); exportData(rFile, X, Y, r, N); exportData(uFile, X, Y, u, N); clock_t endTime31 = clock(); printf("Finish!\n"); printf("Phase 3 ended\n"); double totalTime31 = (double)(endTime31 - startTime31) / CLOCKS_PER_SEC; printf("Time spent on exporting files: %f sec\n", totalTime31); // Destroy plan and clean up memory on device free(kx); free(ky); free(X); free(Y); free(r); free(u); cufftDestroy(plan); cudaFree(r_complex_d); cudaFree(kx_d); cudaFree(ky_d); return 0; } __global__ void real2complex(cufftComplex *c, float *a, int n){ /* compute idx and idy, the location of the element in the original NxN array */ int idxX = blockIdx.x * blockDim.x + threadIdx.x; int idxY = blockIdx.y * blockDim.y + threadIdx.y; if(idxX < n && idxY < n){ int idx = idxX + idxY * n; c[idx].x = a[idx]; c[idx].y = 0.0f; } } __global__ void complex2real_scaled(float *a, cufftComplex *c, float scale, int n){ /* Compute index X and index Y, the location of the element in the original NxN array */ int idxX = blockIdx.x * blockDim.x + threadIdx.x; int idxY = blockIdx.y * blockDim.y + threadIdx.y; if(idxX < n && idxY < n){ int idx = idxX + idxY * n; a[idx] = scale * c[idx].x; } } __global__ void solve_poisson(cufftComplex *c, float *kx, float *ky, int n){ /* compute idx and idy, the location of the element in the original NxN array */ int idxX = blockIdx.x * blockDim.x + threadIdx.x; int idxY = blockIdx.y * blockDim.y + threadIdx.y; if (idxX < n && idxY < n){ int idx = idxX + idxY * n; float scale = -(kx[idxX] * kx[idxX] + ky[idxY] * ky[idxY]); if(idxX == n/2 && idxY == n/2){ scale = 1.0f; } scale = 1.0f / scale; c[idx].x *= scale; c[idx].y *= scale; } } void exportData(const char *file, const float *X, const float *Y, const float *Z, const int n){ FILE *dataFile = fopen(file, "w"); if(dataFile != NULL){ for(int j = 0; j < n ; j++){ for(int i = 0; i < n; i++){ fprintf(dataFile, "%f\t%f\t%f\n", X[i], Y[j], Z[i+j*n]); } } printf("All data have been stored in \"%s\".\n", file); fclose(dataFile); }else{ printf("File not found!"); } } void gaussian(float *bin, const float *X, const float *Y, const int n){ int sNum; // Number of signal int dim = 2; const float PI = 4 * atan(1); float x, y; // Ask for essential parameters printf("Number of signal: "); scanf("%d", &sNum); float *sPos = (float *)malloc(sizeof(float) * dim * sNum); // Position of signal float *scale = (float *)malloc(sizeof(float) * sNum); // Normalization factor float *var = (float *)malloc(sizeof(float) * sNum); // Variances for(int s = 0; s < sNum; s++){ printf("Position of signal %d(e.g. 1.2 -3): ", s+1); scanf("%f %f", &sPos[0+s*dim], &sPos[1+s*dim]); printf("Value of variance %d: ", s+1); scanf("%f", &var[s]); } // Generate required function printf("Generating density distribution..."); for(int s = 0; s < sNum; s++){ scale[s] = 1.0f / sqrt(2 * PI * var[s]); } for(int j = 0; j < n-1; j++){ for(int i = 0; i < n-1; i++){ bin[i+j*n] = 0; for(int s = 0; s < sNum; s++){ x = X[i] - sPos[0+s*dim]; y = Y[j] - sPos[1+s*dim]; bin[i+j*n] += scale[s] * exp(-(x * x + y * y)/(2 * var[s])); } } } // Fix boundary for(int i = 0; i < n; i++){ bin[i+(n-1)*n] = bin[i]; bin[(n-1)+i*n] = bin[i*n]; } // Clean up free(sPos); free(scale); free(var); }
7,740
#include "includes.h" __global__ void call_kALACSearch(int16_t * mCoefsU, int16_t * mCoefsV, int32_t kALACMaxCoefs) { int x = blockIdx.x; int y = threadIdx.x; int index = x * 16 * 16 + y * 16; int32_t k; int32_t den = 1 << DENSHIFT_DEFAULT; mCoefsU[index + 0] = (AINIT * den) >> 4; mCoefsU[index + 1] = (BINIT * den) >> 4; mCoefsU[index + 2] = (CINIT * den) >> 4; mCoefsV[index + 0] = (AINIT * den) >> 4; mCoefsV[index + 1] = (BINIT * den) >> 4; mCoefsV[index + 2] = (CINIT * den) >> 4; for (k = 3; k < kALACMaxCoefs; k++) { mCoefsU[index + k] = 0; mCoefsV[index + k] = 0; } }
7,741
#include <stdio.h> #include <cuda.h> __global__ void kernel() { printf("Hello World! My threadId is %d\n", blockDim.x * blockIdx.x + threadIdx.x); } int main() { kernel<<<1, 256>>>(); cudaDeviceSynchronize(); }
7,742
__global__ void add_vectors( const int* a, const int* b, int *c, const int n) { const int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx >= n) return; c[idx] = a[idx] + b[idx]; }
7,743
/* * Copyright 2015 Netherlands eScience Center, VU University Amsterdam, and Netherlands Forensic Institute * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ /** * This file contains CUDA kernels for applying a Wiener filter to a * PRNU pattern, as proposed by: * M. Chen et al. "Determining image origin and integrity using sensor * noise", IEEE Trans. Inf. Forensics Secur. 3 (2008) 74-90. * * The Wiener filter is used to remove JPEG artifacts from a PRNU pattern. * * To apply the complete filter: * apply Fourier transform to the input image * call computeSquaredMagnitudes() on the frequencies * call computeVarianceEstimates() on the squared magnitudes * call computeVarianceZeroMean() on the squared magnitudes * call scaleWithVariances() scaling the frequencies using the local and global variance * apply inverse Fourier transform * normalize result by calling normalizeToReal() * * @author Ben van Werkhoven <b.vanwerkhoven@esciencecenter.nl> * @version 0.1 */ #ifndef block_size_x #define block_size_x 32 #endif #ifndef block_size_y #define block_size_y 16 #endif #ifndef reuse_computation #define reuse_computation 1 #endif //set the number and size of filters, also adjust max_border #define FILTERS 4 #define FILTER_SIZES {3, 5, 7, 9} #define MAX_BORDER 4 //the largest (filter size/2) #define FLT_MAX 3.40282347e+38f //function interfaces to prevent C++ garbling the kernel keys extern "C" { __global__ void computeSquaredMagnitudes(int h, int w, float* output, float* frequencies); __global__ void scaleWithVariances(int h, int w, float* output, float* input, float* varianceEstimates, float* variance); __global__ void toComplex(int h, int w, float* complex, float* input); __global__ void toReal(int h, int w, float* output, float* complex); __global__ void computeVarianceZeroMean(int n, float* output, float *input); __global__ void computeVarianceEstimates(int h, int w, float* varest, float* input); __global__ void computeVarianceEstimates_naive(int h, int w, float* varest, float* input); __global__ void normalizeToReal(int h, int w, float* output, float* complex); __global__ void normalize(int h, int w, float* output, float* complex); __global__ void sumFloats(float *output, float *input, int n); } /** * Computes the square of each frequency and stores the result as a real. */ __global__ void computeSquaredMagnitudes(int h, int w, float* output, float* frequencies) { int i = threadIdx.y + blockIdx.y * block_size_y; int j = threadIdx.x + blockIdx.x * block_size_x; if (j < w && i < h) { float re = frequencies[i*2*w+(2 * j)]; float im = frequencies[i*2*w+(2 * j + 1)]; output[i*w+j] = (re * re) + (im * im); } } /** * This kernel scales the frequencies in input with a combination of the global variance and an estimate * for the local variance at that position. Effectively this cleans the input pattern from low frequency * noise. */ __global__ void scaleWithVariances(int h, int w, float* output, float* input, float* varianceEstimates, float* variance) { int i = threadIdx.y + blockIdx.y * block_size_y; int j = threadIdx.x + blockIdx.x * block_size_x; float var = variance[0]; if (j < w && i < h) { float scale = var / max(var, varianceEstimates[i*w+j]); output[i*2*w+(j * 2)] = input[i*2*w+(j*2)] * scale; output[i*2*w+(j * 2 + 1)] = input[i*2*w+(j * 2 + 1)] * scale; } } /** * Simple helper kernel to convert an array of real values to an array of complex values */ __global__ void toComplex(int h, int w, float* complex, float* input) { int i = threadIdx.y + blockIdx.y * block_size_y; int j = threadIdx.x + blockIdx.x * block_size_x; if (i < h && j < w) { complex[i * w * 2 + 2 * j] = input[i * w + j]; complex[i * w * 2 + (2 * j + 1)] = 0.0f; } } /** * Simple helper kernel to convert a complex array to an array of real values */ __global__ void toReal(int h, int w, float* output, float* complex) { int i = threadIdx.y + blockIdx.y * block_size_y; int j = threadIdx.x + blockIdx.x * block_size_x; if (i < h && j < w) { output[i*w+j] = complex[i * w * 2 + 2 * j]; } } /** * This kernel normalizes the input by dividing it by the number of pixels in the image. * It takes an array of complex numbers as input, but only stores the real values. */ __global__ void normalizeToReal(int h, int w, float* output, float* complex) { int i = threadIdx.y + blockIdx.y * block_size_y; int j = threadIdx.x + blockIdx.x * block_size_x; if (i < h && j < w) { output[i*w+j] = (complex[i * w * 2 + 2 * j] / (float)(w * h)); } } /** * This kernel normalizes the complex input by dividing it by the number of pixels in the image. */ __global__ void normalize(int h, int w, float* complex_out, float* complex_in) { int i = threadIdx.y + blockIdx.y * block_size_y; int j = threadIdx.x + blockIdx.x * block_size_x; if (i < h && j < w) { complex_out[i*w*2+2*j ] = (complex_in[i*w*2+2*j ] / (float)(w*h)); complex_out[i*w*2+2*j+1] = (complex_in[i*w*2+2*j+1] / (float)(w*h)); } } /** * computeVarianceEstimates uses a number of simple filters to compute a minimum local variance * * Instead of using multiple arrays with zeroed borders around them, the loading phase of this * kernel writes a zero to shared memory instead of loading a border value from global memory. * The filters can then be performed as normal on the data in shared memory. Because of this * MAX_BORDER needs to be set accordingly. * */ __global__ void computeVarianceEstimates(int h, int w, float* varest, float* input) { int ty = threadIdx.y; int tx = threadIdx.x; int i = blockIdx.y * block_size_y; int j = blockIdx.x * block_size_x; __shared__ float shinput[block_size_y+2*MAX_BORDER][block_size_x+2*MAX_BORDER]; //the loading phase of the kernel, which writes 0.0f to shared memory if the index //is outside the input int yEnd = block_size_y+2*MAX_BORDER; int xEnd = block_size_x+2*MAX_BORDER; for (int y=ty; y < yEnd; y+= block_size_y) { for (int x=tx; x < xEnd; x+= block_size_x) { float in = 0.0f; int indexy = i+y-MAX_BORDER; int indexx = j+x-MAX_BORDER; if (indexy >= 0 && indexy < h) { if (indexx >= 0 && indexx < w) { in = input[indexy*w+indexx]; } } shinput[y][x] = in; } } __syncthreads(); const int filter[FILTERS] = FILTER_SIZES; float res = FLT_MAX; #if reuse_computation == 0 //perform filtering without reusing the sum from smaller filters for (int f = 0; f < FILTERS; f++) { int filterSize = filter[f]; int offset = MAX_BORDER-(filterSize/2); //do a convolution float sum = 0.0f; for (int fi = 0; fi < filterSize; fi++) { for (int fj = 0; fj < filterSize; fj++) { sum += shinput[ty+fi+offset][tx+fj+offset]; } } sum /= (float)(filterSize * filterSize); //store minimum res = sum < res ? sum : res; } #elif reuse_computation == 1 //perform filtering while reusing the sum from smaller filters //start from center pixel float sum = shinput[ty+MAX_BORDER][tx+MAX_BORDER]; //add sides of the square filter to sum and store minimum average for (int f = 0; f < FILTERS; f++) { int filterSize = filter[f]; int offset = MAX_BORDER-(filterSize/2); //top and bottom row for (int fj=0; fj<filterSize; fj++) { sum += shinput[ty+0+offset][tx+fj+offset]; sum += shinput[ty+filterSize-1+offset][tx+fj+offset]; } //two sides (between top and bottom rows) for (int fi=1; fi<filterSize-1; fi++) { sum += shinput[ty+fi+offset][tx+0+offset]; sum += shinput[ty+fi+offset][tx+filterSize-1+offset]; } //store minimum float avg = sum / (filterSize*filterSize); res = avg < res ? avg : res; } #endif //write output if (i + ty < h) { if (j + tx < w) { varest[(i+ty)*w + (j+tx)] = res; } } } /** * This method is a naive implementation of computeVarianceEstimates used for correctness checks */ __global__ void computeVarianceEstimates_naive(int h, int w, float* varest, float* input) { int i = threadIdx.y + blockIdx.y * block_size_y; int j = threadIdx.x + blockIdx.x * block_size_x; float res = FLT_MAX; if (i < h && j < w) { const int filter[FILTERS] = FILTER_SIZES; for (int f = 0; f < FILTERS; f++) { int filterSize = filter[f]; int border = filterSize/2; //do a convolution float sum = 0.0f; for (int fi = 0; fi < filterSize; fi++) { for (int fj = 0; fj < filterSize; fj++) { //original //sum += input[(i + fi)*(w+border*2)+(j + fj)]; int row = i+fi-border; int col = j+fj-border; //the following ifs are a hack to save redundant copying if (row >= 0 && row < h) { if (col >= 0 && col < w) { sum += input[row*w + col]; } } } } sum /= (float)(filterSize * filterSize); if (sum < res) { res = sum; } } //write output varest[i*w+j] = res; } } /* * This method computes the variance of an input array, assuming the mean is equal to zero * * Thread block size should be power of two because of the reduction. * The implementation currently assumes only one thread block is used for the entire input array * * In case of multiple thread blocks initialize output to zero and use atomic add or another kernel * * block_size_x power of 2 */ #ifndef grid_size_x //hack to see if the Kernel Tuner is being used #undef block_size_x #define block_size_x 128 #endif __global__ void computeVarianceZeroMean(int n, float *output, float *input) { int x = blockIdx.x * block_size_x + threadIdx.x; int ti = threadIdx.x; int step_size = block_size_x * gridDim.x; float sum = 0.0f; if (x < n) { //compute thread-local sums of squares for (int i=x; i < n; i+=step_size) { sum += input[i]*input[i]; } } //store local sums in shared memory __shared__ float shmem[block_size_x]; shmem[ti] = sum; __syncthreads(); //reduce local sums for (unsigned int s=block_size_x/2; s>0; s>>=1) { if (ti < s) { shmem[ti] += shmem[ti + s]; } __syncthreads(); } //write result if (ti == 0) { output[blockIdx.x] = ( shmem[0] * n ) / ( n - 1 ); //in case of multiple threadblocks write back using atomicAdd } } /* * Simple CUDA Helper function to reduce the output of a * reduction kernel with multiple thread blocks to a single value * * This function performs a sum of an array of floats * * This function is to be called with only a single thread block */ __global__ void sumFloats(float *output, float *input, int n) { int ti = threadIdx.x; __shared__ float shmem[block_size_x]; //compute thread-local sums float sum = 0.0f; for (int i=ti; i < n; i+=block_size_x) { sum += input[i]; } //store local sums in shared memory shmem[ti] = sum; __syncthreads(); //reduce local sums for (unsigned int s=block_size_x/2; s>0; s>>=1) { if (ti < s) { shmem[ti] += shmem[ti + s]; } __syncthreads(); } //write result if (ti == 0) { output[0] = shmem[0]; } }
7,744
#include <stdio.h> #include <stdlib.h> #include <sys/time.h> #define STAR 10001 char *RULES_FILE = (char *)"rule_2M.csv"; const int RULES_COUNT = 2000000; char *TR_FILE = (char *)"transactions_tiny.csv"; const int TR_COUNT = 20000; const int RULE_SIZE = 11; const int TR_SIZE = RULE_SIZE - 1; int rules_f[RULES_COUNT*RULE_SIZE]; int data_f[TR_COUNT*TR_SIZE]; int** alloc_two_d(int rows, int cols) { int **array = (int **)calloc(rows, sizeof(int*)); for (int row = 0; row < rows; row++) { array[row] = (int *)calloc(cols, sizeof(int)); } return array; } int** load_csv(char *csv_file, int rows, int cols){ int **data = alloc_two_d(rows, cols); FILE* file = fopen(csv_file, "r"); for (int row = 0; row < rows; row++) { for (int col = 0; col < cols; col++) { if(!fscanf(file, "%d;", &data[row][col])) { fscanf(file, "%*c;"); data[row][col] = STAR; } } } fclose(file); return data; } int cmpfunc (const void * a, const void * b) { const int **r1 = (const int**)a; const int **r2 = (const int**)b; int i = 0; int cmp = 0; while(i<10){ cmp = (*r1)[i]-(*r2)[i]; if ( cmp != 0){ return cmp; }else{ i++; } } return 0; } __global__ void gpu_kernel(int* data, int tr_count, int tr_size, int* rules, int rules_count,int rule_size, int* result, int result_size){ int tr = blockIdx.x*blockDim.x + threadIdx.x; if(tr >= tr_count)return; int start_col = 0; for (int row = 0; row < rules_count; row++) { int ok = 1; while(rules[row*rule_size + start_col] == STAR){ start_col++; } for (int col = start_col; ok && col < tr_size; col++) { if (data[tr*tr_size + col] != rules[row*rule_size+col] && rules[row*rule_size+col] != STAR) { ok = 0; } } if (ok) { result[tr*result_size+ rules[row*rule_size+rule_size-1]]+=1; } } } void gpu_search(int* data_f, int tr_count,int tr_size, int* rules_f, int rules_count,int rule_size){ int* data_g; int* rules_g; int* result_g; int result_f [tr_count*100]; for(int i=0;i<tr_count;i++){ for(int j=0;j<100;j++){ result_f[i*100+j]=0; } } cudaError_t err; err = cudaMalloc((void **)&data_g, tr_count*tr_size*sizeof(int)); if(err!=cudaSuccess){printf("%s in %s at line %d\n",cudaGetErrorString(err),__FILE__,__LINE__);} err = cudaMalloc((void **)&rules_g, rules_count*rule_size*sizeof(int)); if(err!=cudaSuccess){printf("%s in %s at line %d\n",cudaGetErrorString(err),__FILE__,__LINE__);} err = cudaMalloc((void **)&result_g, tr_count*100*sizeof(int)); if(err!=cudaSuccess){printf("%s in %s at line %d\n",cudaGetErrorString(err),__FILE__,__LINE__);} err = cudaMemcpy(data_g, data_f, tr_count*tr_size*sizeof(int), cudaMemcpyHostToDevice); if(err!=cudaSuccess){printf("%s in %s at line %d\n",cudaGetErrorString(err),__FILE__,__LINE__);} err = cudaMemcpy(rules_g, rules_f, rules_count*rule_size*sizeof(int), cudaMemcpyHostToDevice); if(err!=cudaSuccess){printf("%s in %s at line %d\n",cudaGetErrorString(err),__FILE__,__LINE__);} err = cudaMemcpy(result_g, result_f, tr_count*100*sizeof(int), cudaMemcpyHostToDevice); if(err!=cudaSuccess){printf("%s in %s at line %d\n",cudaGetErrorString(err),__FILE__,__LINE__);} int BLOCK_SIZE =1024; int BLOCK_DIM = 5; //for(int i=0;i<){ gpu_kernel<<<BLOCK_DIM,BLOCK_SIZE>>>(data_g, tr_count, tr_size, rules_g, rules_count, rule_size, result_g, 100); //} err = cudaMemcpy(result_f,result_g , tr_count*100*sizeof(int), cudaMemcpyDeviceToHost); if(err!=cudaSuccess){printf("%s in %s at line %d\n",cudaGetErrorString(err),__FILE__,__LINE__);} // for(int i=0;i<BLOCK_DIM*BLOCK_SIZE;i++){ // for(int j=0;j<100;j++){ // printf("%d,",result_f[i*100+j]); // } // printf("\n"); // } err = cudaFree(data_g); if(err!=cudaSuccess){printf("%s in %s at line %d\n",cudaGetErrorString(err),__FILE__,__LINE__);} err = cudaFree(rules_g); if(err!=cudaSuccess){printf("%s in %s at line %d\n",cudaGetErrorString(err),__FILE__,__LINE__);} err = cudaFree(result_g); if(err!=cudaSuccess){printf("%s in %s at line %d\n",cudaGetErrorString(err),__FILE__,__LINE__);} } int main(){ struct timeval start, end; gettimeofday(&start, NULL); printf("Loading rules\n"); int **rules = load_csv(RULES_FILE, RULES_COUNT, RULE_SIZE); printf("Loading transactions\n"); int **data = load_csv(TR_FILE, TR_COUNT, TR_SIZE); printf("Sorting rules\n"); qsort(rules, RULES_COUNT, sizeof(rules[0]), cmpfunc); for(int i=0;i<RULES_COUNT;i++){ for(int j=0;j<RULE_SIZE;j++){ rules_f[i*RULE_SIZE + j]=rules[i][j]; } } for(int i=0;i<TR_COUNT;i++){ for(int j=0;j<TR_SIZE;j++){ data_f[i*TR_SIZE + j]=data[i][j]; } } printf("GPU: start\n"); gettimeofday(&start, NULL); gpu_search(data_f,TR_COUNT,TR_SIZE,rules_f,RULES_COUNT,RULE_SIZE); gettimeofday(&end, NULL); printf("GPU: %f\n",(end.tv_sec - start.tv_sec)+ (end.tv_usec - start.tv_usec) / 1.e6); return 0; }
7,745
#include "includes.h" __global__ void clearLabel(bool *label, unsigned int size) { unsigned int id = blockDim.x * blockIdx.x + threadIdx.x; if(id < size) label[id] = false; }
7,746
#include "includes.h" __global__ void reduceUnrollWarps8(int *g_idata, int *g_odata, unsigned int n){ unsigned int tid = threadIdx.x; unsigned int idx = (8 * blockIdx.x) * blockDim.x + threadIdx.x; int *idata = g_idata + (8 * blockIdx.x) * blockDim.x; if(idx + 7 * blockDim.x < n){ g_idata[idx] += g_idata[idx + blockDim.x]; g_idata[idx] += g_idata[idx + 2 * blockDim.x]; g_idata[idx] += g_idata[idx + 3 * blockDim.x]; g_idata[idx] += g_idata[idx + 4 * blockDim.x]; g_idata[idx] += g_idata[idx + 5 * blockDim.x]; g_idata[idx] += g_idata[idx + 6 * blockDim.x]; g_idata[idx] += g_idata[idx + 7 * blockDim.x]; } __syncthreads(); for(int stride = blockDim.x / 2; stride > 32; stride >>= 1){ if(tid < stride) idata[tid] += idata[tid + stride]; __syncthreads(); } if(tid < 32){ volatile int *vmem = idata; vmem[tid] += vmem[tid + 32]; vmem[tid] += vmem[tid + 16]; vmem[tid] += vmem[tid + 8]; vmem[tid] += vmem[tid + 4]; vmem[tid] += vmem[tid + 2]; vmem[tid] += vmem[tid + 1]; } if(tid == 0) g_odata[blockIdx.x] = idata[0]; }
7,747
#include <iostream> #include <math.h> #include <chrono> // Kernel function to add the elements of two arrays __global__ void add(int n, float *x, float *y){ int index = blockIdx.x * blockDim.x + threadIdx.x; int stride = blockDim.x * gridDim.x; for (int i = index; i < n; i += stride) y[i] = x[i] + y[i]; } int main(int argc ,char *argv[]){ std::chrono::high_resolution_clock::time_point start_beig,stop_end; int N = 1<<20; float *x, *y; float *x_device, *y_device; //cpu data x = (float *)malloc(N*sizeof(float)); y = (float *)malloc(N*sizeof(float)); start_beig = std::chrono::high_resolution_clock::now(); // initialize x and y arrays on the host for (int i = 0; i < N; i++) { x[i] = 1.0f; y[i] = 2.0f; } stop_end = std::chrono::high_resolution_clock::now(); auto duration = std::chrono::duration_cast<std::chrono::microseconds>(stop_end - start_beig).count(); std::cout << "cudaMalloc set data time : " << duration / 1000.0f << " ms\n"; //device data start_beig = std::chrono::high_resolution_clock::now(); cudaMalloc(&x_device,N*sizeof(float)); cudaMalloc(&y_device,N*sizeof(float)); cudaMemcpy(x_device,x,N*sizeof(float),cudaMemcpyHostToDevice); cudaMemcpy(y_device,y,N*sizeof(float),cudaMemcpyHostToDevice); stop_end = std::chrono::high_resolution_clock::now(); duration = std::chrono::duration_cast<std::chrono::microseconds>(stop_end - start_beig).count(); std::cout << "cudaMalloc time : " << duration / 1000.0f << " ms\n"; int blockSize = 256; int numBlocks = (N + blockSize - 1) / blockSize; start_beig = std::chrono::high_resolution_clock::now(); // Run kernel on 1M elements on the GPU add<<<numBlocks, blockSize>>>(N, x_device, y_device); // Wait for GPU to finish before accessing on host cudaDeviceSynchronize(); cudaMemcpy(y,y_device,N*sizeof(float),cudaMemcpyDeviceToHost); stop_end = std::chrono::high_resolution_clock::now(); duration = std::chrono::duration_cast<std::chrono::microseconds>(stop_end - start_beig).count(); std::cout << "cudaMalloc exe : " << duration / 1000.0f << " ms\n"; // Check for errors (all values should be 3.0f) float maxError = 0.0f; for (int i = 0; i < N; i++){ maxError = fmax(maxError, fabs(y[i]-3.0f)); if(maxError > 0){ std::cout << "i: " << i << " y[i]: " << y[i] << std::endl; } } std::cout << "Max error: " << maxError << std::endl; // Free memory cudaFree(x); cudaFree(y); return 0; }
7,748
#include <iostream> #include <math.h> __global__ void add(int n, float *x, float *y) { int index = blockIdx.x * blockDim.x + threadIdx.x; //int stride = blockDim.x * gridDim.x; if (index >= n) { return; } y[index] = x[index] + y[index]; } int main(void) { int N = 1<<20; float *x, *y; cudaMallocManaged(&x, N*sizeof(float)); cudaMallocManaged(&y, N*sizeof(float)); for (int i=0; i<N; i++) { x[i] = 1.0f; y[i] = 2.0f; } int blockSize = 256; int numBlocks = (N + blockSize - 1) / blockSize; add<<<numBlocks, blockSize>>>(N, x, y); cudaDeviceSynchronize(); float maxError = 0.0f; for (int i=0; i<N; i++) { maxError = fmax(maxError, fabs(y[i]-3.0f)); } std::cout << "Max error: " << maxError << std::endl; cudaFree(x); cudaFree(y); return 0; }
7,749
#include<thrust/host_vector.h> #include<thrust/device_vector.h> #include<thrust/reduce.h> #include<thrust/scan.h> #include<device_launch_parameters.h> #include<stdio.h> #include<stdlib.h> #define DATA_SIZE 66292994 int main(void) { float t_reduce; cudaEvent_t start1,stop1, start2, stop2; cudaEventCreate(&start1); cudaEventCreate(&stop1); cudaEventCreate(&start2); cudaEventCreate(&stop2); //initialze random values on host thrust::host_vector<int> data(DATA_SIZE); thrust::generate(data.begin(), data.end(), rand); //compute sum on host(CPU) int h_sreduce = thrust::reduce(data.begin(), data.end()); //for inclusive time cudaEventRecord(start1, NULL); //copy values on device thrust::device_vector<int> gpudata = data; cudaEventRecord(start2, NULL); //compute sum on device(GPU) int d_sreduce = thrust::reduce(gpudata.begin(), gpudata.end()); //copy back to host thrust::copy(gpudata.begin(), gpudata.end(), data.begin()); cudaEventRecord(stop1, NULL); cudaEventSynchronize(stop1); cudaEventElapsedTime(&t_reduce, start1, stop1); printf("\n Reduce time is %f ms", t_reduce); // thrust::copy(d_vec.begin(), d_vec.end(), h_vec.begin()); printf("\n host sum = %d, gpu sum = %d",h_sreduce,d_sreduce); }
7,750
#include<stdio.h> const int ARRAY_SIZE =500000; // size greater than 32M could not be achieved const int ARRAY_BYTES = ARRAY_SIZE * sizeof(float); const int MAX_NO_THREADS = 512; __global__ void vector_reduce(float *d_in1, float *d_in2, float *d_out){ int index = threadIdx.x + blockIdx.x*blockDim.x ; *(d_out+index) = *(d_in1+index) + *(d_in2+index); } int check( float *h_in1, float *h_in2, float *h_out){ int i,flag = 1; for(i=0;i<ARRAY_SIZE;i++){ if(h_in1[i]+h_in2[i]!=h_out[i]){ flag=0; break; } } return flag; } int main(){ //allocating size for host arrays float h_in1[ARRAY_SIZE], h_in2[ARRAY_SIZE], h_out[ARRAY_SIZE]; //generating the input arrays int i; for(i=0;i<ARRAY_SIZE;i++){ h_in1[i]=(float)i; h_in2[i]=(float)(ARRAY_SIZE-i); } //declaring device memory pointers float *d_in1, *d_in2, *d_out; //allocating device memory cudaMalloc(&d_in1, ARRAY_BYTES); cudaMalloc(&d_in2, ARRAY_BYTES); cudaMalloc(&d_out, ARRAY_BYTES); //transferring memory from host to device cudaMemcpy(d_in1, h_in1, ARRAY_BYTES, cudaMemcpyHostToDevice); cudaMemcpy(d_in2, h_in2, ARRAY_BYTES, cudaMemcpyHostToDevice); //starting kernel vector_reduce<<<(int)(ARRAY_SIZE/MAX_NO_THREADS)+1, MAX_NO_THREADS>>>(d_in1, d_in2, d_out); //transferring memory from device to host cudaMemcpy(h_out, d_out, ARRAY_BYTES, cudaMemcpyDeviceToHost); //checking correctness if(check(h_in1, h_in2, h_out)) printf("the result is correct\n"); else printf("the result is incorrect\n"); //freeing memory cudaFree(d_in1); cudaFree(d_in2); cudaFree(d_out); return 0; }
7,751
// // Created by Peter Rigole on 2019-04-26. // #ifndef AXONBITS_PHASE_H #define AXONBITS_PHASE_H enum Phase { ExpectationPhase, OutcomePhase }; #endif //AXONBITS_PHASE_H
7,752
#include "includes.h" __global__ void initGuessBuffers( const uchar4* srcImg, float3* guess1, float3* guess2, const uint nRows, const uint nCols ) { const uint nSamps = nRows*nCols; const uint samp = threadIdx.x + blockDim.x * blockIdx.x; if( samp < nSamps ) { guess1[samp].x = srcImg[samp].x; guess2[samp].x = srcImg[samp].x; guess1[samp].y = srcImg[samp].y; guess2[samp].y = srcImg[samp].y; guess1[samp].z = srcImg[samp].z; guess2[samp].z = srcImg[samp].z; } }
7,753
#include "includes.h" __global__ void init(int order, const int matrices, double * C) { int i = blockIdx.x * blockDim.x + threadIdx.x; int j = blockIdx.y * blockDim.y + threadIdx.y; for (int b=0; b<matrices; ++b) { if ((i<order) && (j<order)) { C[b*order*order+i*order+j] = 0; } } }
7,754
#include "includes.h" __global__ void cu_padding(const float* src, float* dst, const int rows1, const int cols1, const int cols2, const int n){ int tid = threadIdx.x + blockIdx.x * blockDim.x; int stride = blockDim.x * gridDim.x; while(tid < n){ int pad = (cols2 - cols1) / 2; int c1 = tid % cols1; int r1 = tid / cols1; int r2 = r1 + pad; int c2 = c1 + pad; dst[r2 * cols2 + c2] = src[tid]; tid += stride; } }
7,755
/*************************************************** * Module that multiply a matrix by the transpose of other * Author: Alonso Vidales <alonso.vidales@tras2.es> * * To be compiled with nvcc -ptx matrix_mult_trans.cu * Debug: nvcc -arch=sm_20 -ptx matrix_mult_trans.cu * **************************************************/ //#include <stdio.h> #ifdef __cplusplus extern "C" { #endif // CUDA Kernel __global__ void matrixMulTrans(double* C, double* A, double* B, int wA, int resW, int resH, int resultWidth, int resultSize) { int x = threadIdx.x + (blockIdx.x * resW); int y = threadIdx.y + (blockIdx.y * resH); int resultPos = y * resultWidth + x; //printf("Thread %d - %d: %d. Final: x: %d y: %d Size: %d\n", threadIdx.x, threadIdx.y, resultPos, x, y, resultSize); if (resultPos < resultSize && x < resultWidth) { // value stores the element that is // computed by the thread double value = 0; for (int i = 0; i < wA; ++i) { value += A[y * wA + i] * B[x * wA + i]; //printf("Pos %d - %d, thread %d - %d : pos: %d %d H: %d Pos: %d Val: %f\n", blockIdx.x, blockIdx.y, threadIdx.x, threadIdx.y, x, y, resultWidth, resultPos, value); } // Write the matrix to device memory each // thread writes one element C[resultPos] = value; } } #ifdef __cplusplus } #endif
7,756
#include <stdio.h> #include <stdlib.h> #include "cuda_runtime.h" #include "device_launch_parameters.h" #define ErrorCheck(ans) { CheckFun((ans), __FILE__, __LINE__); } inline void CheckFun(cudaError_t code, const char *file, int line){ if (code != cudaSuccess) { fprintf(stderr,"ERROR: %s %s %d\n", cudaGetErrorString(code), file, line); exit(0); } } __global__ void multiply(double* dev_A, double* dev_B, size_t arrLen) { size_t index = threadIdx.x + blockIdx.x * blockDim.x; while (index < arrLen) { dev_A[index] = dev_A[index] * dev_B[index]; index += blockDim.x * gridDim.x; } } int main() { size_t arrLen; scanf("%zd", &arrLen); size_t size = sizeof(double) * arrLen; double *arrA = (double*)malloc(size); double *arrB = (double*)malloc(size); for (size_t i = 0; i < arrLen; ++i) { scanf("%lf", &arrA[i]); } for (size_t i = 0; i < arrLen; ++i) { scanf("%lf", &arrB[i]); } double *dev_A, *dev_B; ErrorCheck(cudaMalloc((void**)&dev_A, size)); ErrorCheck(cudaMalloc((void**)&dev_B, size)); ErrorCheck(cudaMemcpy(dev_A, arrA, size, cudaMemcpyHostToDevice)); ErrorCheck(cudaMemcpy(dev_B, arrB, size, cudaMemcpyHostToDevice)); dim3 blockSize = dim3(512,1,1); dim3 gridSize = dim3((unsigned int)arrLen / 512 + 1, 1, 1); multiply <<<gridSize, blockSize >>> (dev_A, dev_B, arrLen); ErrorCheck(cudaGetLastError()); ErrorCheck(cudaMemcpy(arrA, dev_A, size, cudaMemcpyDeviceToHost)); for (size_t i = 0; i < arrLen; ++i) { printf("%.10lf ", arrA[i]); } printf("\n"); free(arrA); free(arrB); ErrorCheck(cudaFree(dev_A)); ErrorCheck(cudaFree(dev_B)); return 0; }
7,757
#include <iostream> #include <algorithm> #include <cstdlib> #include <ctime> #include <cuda.h> #include <stdio.h> #include <cassert> //define the chunk sizes that each threadblock will work on #define BLKXSIZE 32 #define BLKYSIZE 4 #define BLKZSIZE 4 #define Q 19 #define lx 10 #define ly 10 #define lz 5 // for cuda error checking #define cudaCheckErrors(msg) \ do { \ cudaError_t __err = cudaGetLastError(); \ if (__err != cudaSuccess) { \ fprintf(stderr, "Fatal error: %s (%s at %s:%d)\n", \ msg, cudaGetErrorString(__err), \ __FILE__, __LINE__); \ fprintf(stderr, "*** FAILED - ABORTING\n"); \ return 1; \ } \ } while (0) template <typename T> __device__ void swap ( T& a, T& b ) { T c(a); a=b; b=c; } __global__ void gpu_array_swap(int ptr_gpu[][ly][lz][Q]) { // int thread_id = thread_idx(grid_dim, block_dim); unsigned idx = blockIdx.x * blockDim.x + threadIdx.x; unsigned idy = blockIdx.y * blockDim.y + threadIdx.y; unsigned idz = blockIdx.z * blockDim.z + threadIdx.z; if ((idx < lx) && (idy < ly) && (idz < lz)) { for (size_t i = 1; i <= 9; i++) swap(ptr_gpu[idx][idy][idz][i], ptr_gpu[idx][idy][idz][i + 9]); } } void set_array(int array[][ly][lz][Q]) { int m = 0; for (int l = 0; l < Q; ++l) { for (int i = 0; i < lz; ++i) { for (int j = 0; j < ly; ++j) { for (int k = 0; k < lx; ++k) { array[i][j][k][l] = ++m; } } } } } void print_array(int array[][ly][lz][Q]) { for (int i = 0; i < lx; ++i) { for (int j = 0; j < ly; ++j) { for (int k = 0; k < lz; ++k) { for (int l = 0; l < Q; ++l) { std::cout << array[i][j][k][l] << " "; if (l == (Q - 1)) std::cout << std::endl; } } } } } int main() { typedef int array_3d[ly][lz]; typedef int array_4d[ly][lz][Q]; const dim3 blockSize(BLKXSIZE, BLKYSIZE, BLKZSIZE); const dim3 gridSize(((lx + BLKXSIZE - 1) / BLKXSIZE), ((ly + BLKYSIZE - 1) / BLKYSIZE), ((lz + BLKZSIZE - 1) / BLKZSIZE)); // pointers for data set storage via malloc array_4d* c; // storage for result stored on host array_4d* d_c; // storage for result computed on device // allocate storage for data set if ((c = (array_4d*)malloc((lx * ly * lz * Q) * sizeof(int))) == 0) { fprintf(stderr, "malloc1 Fail \n"); return 1; } set_array(c); print_array(c); // allocate GPU device buffers cudaMalloc((void**)&d_c, (lx * ly * lz * Q) * sizeof(int)); cudaCheckErrors("Failed to allocate device buffer"); cudaMemcpy(d_c, c, ((lx * ly * lz * Q) * sizeof(int)), cudaMemcpyHostToDevice); // compute result gpu_array_swap<<<gridSize, blockSize>>>(d_c); cudaCheckErrors("Kernel launch failure"); // copy output data back to host cudaMemcpy(c, d_c, ((lx * ly * lz * Q) * sizeof(int)), cudaMemcpyDeviceToHost); cudaCheckErrors("CUDA memcpy failure"); free(c); cudaFree(d_c); cudaCheckErrors("cudaFree fail"); return 0; }
7,758
#include "includes.h" __global__ void __pairmult2(int nrows, int bncols, int brows1, int brows2, float *A, int lda, float *A2, int lda2, float *Bdata, int *Bir, int *Bjc, int broff, int bcoff, float *C, int ldc, int transpose) {}
7,759
#include <stdio.h> #include <stdlib.h> #include <cuda.h> #include <time.h> #include <iostream> #define TIMER_CREATE(t) \ cudaEvent_t t##_start, t##_end; \ cudaEventCreate(&t##_start); \ cudaEventCreate(&t##_end); #define TIMER_START(t) \ cudaEventRecord(t##_start); \ cudaEventSynchronize(t##_start); \ #define TIMER_END(t) \ cudaEventRecord(t##_end); \ cudaEventSynchronize(t##_end); \ cudaEventElapsedTime(&t, t##_start, t##_end); \ cudaEventDestroy(t##_start); \ cudaEventDestroy(t##_end); #define TILE_SIZE 16 #define BLOCK_SIZE_1D 256 #define NUM_BINS 256 #define CUDA_TIMING #define DEBUG #define WARP_SIZE 32 #define R 9 unsigned char *input_gpu; unsigned char *output_gpu; double CLOCK() { struct timespec t; clock_gettime(CLOCK_MONOTONIC, &t); return (t.tv_sec * 1000)+(t.tv_nsec*1e-6); } /*******************************************************/ /* Cuda Error Function */ /*******************************************************/ inline cudaError_t checkCuda(cudaError_t result) { #if defined(DEBUG) || defined(_DEBUG) if (result != cudaSuccess) { fprintf(stderr, "CUDA Runtime Error: %s\n", cudaGetErrorString(result)); exit(-1); } #endif return result; } // Add GPU kernel and functions // HERE!!! __global__ void kernel(unsigned char *input, unsigned int *output_cdf, //unsigned char *output, unsigned int im_size, unsigned int *cdf_min){ /* int x = blockIdx.x*TILE_SIZE+threadIdx.x; int y = blockIdx.y*TILE_SIZE+threadIdx.y; int location = y*TILE_SIZE*gridDim.x+x; input[location] = round(float(output_cdf[input[location]] - *cdf_min)/float(im_size/4 - *cdf_min) * (NUM_BINS - 1)); */ int location = blockIdx.x * blockDim.x+threadIdx.x; input[location] = float(output_cdf[input[location]] - *cdf_min)/float(im_size/64 - *cdf_min) * (NUM_BINS - 1); //printf("the final: %d .", int(output[location])); } __global__ void get_histogram(unsigned char *input, unsigned int *output_histogram //int offset ){ /* int x = blockIdx.x*TILE_SIZE+threadIdx.x; int y = blockIdx.y*TILE_SIZE+threadIdx.y; //if(x % 4 == 0 && y % 4 == 0){ if(x & 1 && y & 1){ int location = offset + y*TILE_SIZE*gridDim.x+x; atomicAdd(&(output_histogram[input[location]]), 1); }*/ if( !(threadIdx.x & 63)){ int location = blockIdx.x * blockDim.x+threadIdx.x; atomicAdd(&(output_histogram[input[location]]), 1); } //__syncthreads(); } __global__ void get_cdf_prefixSum(unsigned int *histogram) { int tid = threadIdx.x; //USE SHARED MEMORY - COMON WE ARE EXPERIENCED PROGRAMMERS __shared__ int Cache[256]; Cache[tid] = histogram[tid]; __syncthreads(); int space = 1; //BEGIN for (int i = 0; i < 8; i++) { int temp = Cache[tid]; int neighbor = 0; if ((tid - space) >= 0) { neighbor = Cache[tid - space]; } __syncthreads(); //AFTER LOADING if (tid < space) { //DO NOTHING } else { Cache[tid] = temp + neighbor; } space = space * 2; __syncthreads(); } //REWRITE RESULTS TO MAIN MEMORY histogram[tid] = Cache[tid]; } /* __global__ void get_cdf_naive(unsigned int *output_histogram, unsigned int *output_cdf, int n) { unsigned int d_hist_idx = blockDim.x * blockIdx.x + threadIdx.x; if (d_hist_idx >= n) { return; } unsigned int cdf_val = 0; for (int i = 0; i <= d_hist_idx; ++i) { cdf_val = cdf_val + output_histogram[i]; } output_cdf[d_hist_idx] = cdf_val; }*/ __global__ void reductionMin(unsigned int *sdata, unsigned int *results, int n) { // extern __shared__ int sdata[]; unsigned int tx = threadIdx.x; // block-wide reduction for(unsigned int offset = blockDim.x>>1; offset > 0; offset >>= 1) { __syncthreads(); if(tx < offset) { if(sdata[tx + offset] < sdata[tx] || sdata[tx] == 0) sdata[tx] = sdata[tx + offset]; } } // finally, thread 0 writes the result if(threadIdx.x == 0) { // the result is per-block *results = sdata[0]; } } __global__ void kernel_warmup(unsigned char *input, unsigned char *output){ int x = blockIdx.x*TILE_SIZE+threadIdx.x; int y = blockIdx.y*TILE_SIZE+threadIdx.y; int location = y*TILE_SIZE*gridDim.x+x; output[location] = x%255; } void histogram_gpu(unsigned char *data, unsigned int height, unsigned int width){ int gridXSize = 1 + (( width - 1) / TILE_SIZE); int gridYSize = 1 + ((height - 1) / TILE_SIZE); int gridSize_1D = 1 + (NUM_BINS - 1)/ BLOCK_SIZE_1D; int gridSize1D_2D = 1 + (( width*height - 1) / BLOCK_SIZE_1D); int XSize = gridXSize*TILE_SIZE; int YSize = gridYSize*TILE_SIZE; // Both are the same size (CPU/GPU). int size = XSize*YSize; // CPU unsigned int *cdf_gpu = new unsigned int [NUM_BINS]; // Pinned //unsigned char *data_pinned; // GPU unsigned int *output_histogram; //unsigned int *output_cdf; unsigned int *cdf_min; // Pageable to Pinned memory //cudaMallocHost((void**)&data_pinned, size*sizeof(unsigned char)); //memcpy(data_pinned, data, size*sizeof(unsigned char)); // Allocate arrays in GPU memory checkCuda(cudaMalloc((void**)&input_gpu , size*sizeof(unsigned char))); checkCuda(cudaMalloc((void**)&output_histogram , NUM_BINS*sizeof(unsigned int))); //checkCuda(cudaMalloc((void**)&output_cdf , NUM_BINS*sizeof(unsigned int))); checkCuda(cudaMalloc((void**)&cdf_min , sizeof(unsigned int))); checkCuda(cudaMemset(output_histogram , 0 , NUM_BINS*sizeof(unsigned int))); //checkCuda(cudaMemset(output_cdf , 0 , NUM_BINS*sizeof(unsigned int))); checkCuda(cudaMemset(cdf_min, 0, sizeof(unsigned int))); // Grid & Block Size //dim3 dimGrid2D(gridXSize, gridYSize); //dim3 dimBlock2D(TILE_SIZE, TILE_SIZE); // create streams /* const int nStreams = 2; const int streamSize = width*height/nStreams; std:: cout << "stream size: " << streamSize<<std::endl; const int streamBytes = streamSize * sizeof(unsigned char); cudaStream_t stream[nStreams]; for (int i = 0; i < nStreams; ++i){ checkCuda(cudaStreamCreate(&stream[i])); } */ // Copy data to GPU checkCuda(cudaMemcpy(input_gpu, data, size*sizeof(unsigned char), cudaMemcpyHostToDevice)); //checkCuda(cudaDeviceSynchronize()); // Execute algorithm dim3 dimGrid1D(gridSize_1D); dim3 dimBlock1D(BLOCK_SIZE_1D); dim3 dimGrid1D_2D(gridSize1D_2D); dim3 dimBlock1D_2D(BLOCK_SIZE_1D); // Kernel Call #if defined(CUDA_TIMING) float Ktime; TIMER_CREATE(Ktime); TIMER_START(Ktime); #endif get_histogram<<<dimGrid1D_2D, dimBlock1D_2D>>>(input_gpu, output_histogram); //get_cdf_naive<<<dimGrid1D, dimBlock1D>>>(output_histogram, output_cdf, NUM_BINS); get_cdf_prefixSum<<<1, 256>>>(output_histogram); reductionMin<<<1, 256>>>(output_histogram, cdf_min, 256); //checkCuda(cudaPeekAtLastError()); //checkCuda(cudaDeviceSynchronize()); // Retrieve results from the GPU /* checkCuda(cudaMemcpy(cdf_gpu, output_cdf, NUM_BINS*sizeof(unsigned int), cudaMemcpyDeviceToHost)); // Free resources and end the program */ /* unsigned int cdf_min = INT_MAX; for (int i = 0; i < NUM_BINS; i++){ if(cdf_gpu[i] != 0 && cdf_gpu[i] < cdf_min){ cdf_min = cdf_gpu[i]; } }*/ // std::cout << "cdf min : " << cdf_min << std::endl; kernel<<<dimGrid1D_2D, dimBlock1D_2D>>>(input_gpu, output_histogram, width*height, cdf_min); checkCuda(cudaPeekAtLastError()); checkCuda(cudaDeviceSynchronize()); #if defined(CUDA_TIMING) TIMER_END(Ktime); printf("Kernel Execution Time: %f ms\n", Ktime); #endif checkCuda(cudaMemcpy(data, input_gpu, size*sizeof(unsigned char), cudaMemcpyDeviceToHost)); //memcpy(data, data_pinned, size*sizeof(unsigned char)); //checkCuda(cudaFreeHost(data_pinned)); checkCuda(cudaFree(output_histogram)); checkCuda(cudaFree(cdf_min)); //checkCuda(cudaFree(output_cdf)); //checkCuda(cudaFree(output_gpu)); checkCuda(cudaFree(input_gpu)); /* for(int i = 0; i < NUM_BINS; i++){ std::cout << "Value " << i << " : " << probability_gpu[i] << " " << cdf_gpu[i] << std::endl; }*/ /* for(int i = 0; i < NUM_BINS*NUM_PARTS; i++){ std::cout << "Value " << i << " : " << hist_local_gpu[i] << " " << std::endl; } */ /* for (long int i = 0; i < 4990464; i++){ std::cout << data[i] << " "; }*/ } void histogram_gpu_warmup(unsigned char *data, unsigned int height, unsigned int width){ int gridXSize = 1 + (( width - 1) / TILE_SIZE); int gridYSize = 1 + ((height - 1) / TILE_SIZE); int XSize = gridXSize*TILE_SIZE; int YSize = gridYSize*TILE_SIZE; // Both are the same size (CPU/GPU). int size = XSize*YSize; // Allocate arrays in GPU memory checkCuda(cudaMalloc((void**)&input_gpu , size*sizeof(unsigned char))); checkCuda(cudaMalloc((void**)&output_gpu , size*sizeof(unsigned char))); checkCuda(cudaMemset(output_gpu , 0 , size*sizeof(unsigned char))); // Copy data to GPU checkCuda(cudaMemcpy(input_gpu, data, size*sizeof(char), cudaMemcpyHostToDevice)); checkCuda(cudaDeviceSynchronize()); // Execute algorithm dim3 dimGrid(gridXSize, gridYSize); dim3 dimBlock(TILE_SIZE, TILE_SIZE); kernel_warmup <<<dimGrid, dimBlock>>>(input_gpu, output_gpu); checkCuda(cudaDeviceSynchronize()); // Retrieve results from the GPU checkCuda(cudaMemcpy(data, output_gpu, size*sizeof(unsigned char), cudaMemcpyDeviceToHost)); // Free resources and end the program checkCuda(cudaFree(output_gpu)); checkCuda(cudaFree(input_gpu)); }
7,760
#include <iostream> #include <array> #include <cuda.h> #include <numeric> #include <vector> template<typename T> T* copy_to_host(T* dest, T* src, size_t size){ auto rc = cudaMemcpy(dest, src, sizeof(T) * size, cudaMemcpyDeviceToHost); if(rc != cudaSuccess) throw std::bad_alloc(); return dest; } template<class T> T* cuda_malloc(size_t size){ T* temp; auto rc = cudaMalloc((void**)&temp, sizeof(T) * size); if (rc != cudaSuccess){ printf("device side\n"); throw std::bad_alloc(); } return temp; } template<typename T> cudaError_t copy_to_device(T* dest, T* src, size_t size){ cudaError_t rc = cudaMemcpy(dest, src, sizeof(T) * size, cudaMemcpyHostToDevice); if(rc != cudaSuccess) throw std::bad_alloc(); return rc; } template<typename T> T* alloc_and_copy_to_device(T* src, size_t size){ T* tmp = cuda_malloc<T>(size); cudaError_t rc = copy_to_device<T>(tmp, src, size); if(rc != cudaSuccess) throw std::runtime_error("Bad copy to device"); return tmp; } __global__ void atomic_addition(double* src, double* out, size_t size){ size_t tid = blockIdx.x * blockDim.x + threadIdx.x; size_t total_num_threads = gridDim.x * blockDim.x; for(size_t i = tid; i < size; i += total_num_threads){ for(int i=0; i < 8;++i) atomicAdd(&out[threadIdx.x], src[i]); //out[threadIdx.x] += src[i]; } } int main(){ const size_t num_threads = 64; std::vector<double> src; src.resize(32768 * 1025 * 2); std::iota(src.begin(), src.end(), 1.); std::array<double, num_threads> output = {0.}; std::cout<<"size:"<<src.size()<<std::endl; std::cout<<"Memory:"<<src.size()*8/1e9<<"GB\n"; double* d_src = alloc_and_copy_to_device<double>(src.data(), src.size()); double* d_output = alloc_and_copy_to_device<double>(output.data(), output.size()); size_t num_blocks = src.size()/num_threads; atomic_addition<<<num_blocks, num_threads>>>(d_src, d_output, src.size()); cudaDeviceSynchronize(); copy_to_host(output.data(), d_output, output.size()); for(int i = 0; i < output.size(); ++i) printf("output %i, %e\n", i, output[i]); cudaFree(d_src); cudaFree(d_output); return 0; }
7,761
/* Furthest point sampling GPU implementation * Author Zhaoyu SU * All Rights Reserved. Sep., 2019. * Happy Mid-Autumn Festival! :) */ #include <stdio.h> #include <time.h> #include <sys/time.h> #include <iostream> #define USECPSEC 1000000ULL __global__ void get_bev_gt_bbox_gpu_kernel(int batch_size, int npoint, int nbbox, int bbox_attr, int num_anchor, int anchor_attr, int diff_thres, int cls_thres, float expand_ratio, const float* input_coors, const float* label_bbox, const int* input_num_list, const float* anchor_param_list, int* input_accu_list, float* gt_bbox, int* gt_conf, int* label_idx) { if (batch_size * nbbox * bbox_attr <=0 || npoint <=0) { // printf("Get RoI Logits Op exited unexpectedly.\n"); return; } if (threadIdx.x == 0 && blockIdx.x == 0) { input_accu_list[0] = 0; for (int b=1; b<batch_size; b++) { input_accu_list[b] = input_accu_list[b-1] + input_num_list[b-1]; } } __syncthreads(); // printf("%d\n", input_accu_list[5]); for (int b=blockIdx.x; b<batch_size; b+=gridDim.x) { for (int i=threadIdx.x; i<input_num_list[b]; i+=blockDim.x) { for (int k=0; k<num_anchor; k++) { float point_x = input_coors[input_accu_list[b]*2 + i*2 + 0]; float point_y = input_coors[input_accu_list[b]*2 + i*2 + 1]; float point_z = anchor_param_list[k*anchor_attr + 3]; gt_bbox[input_accu_list[b]*num_anchor*7 + i*num_anchor*7 + k*7 + 0] = 0.1; gt_bbox[input_accu_list[b]*num_anchor*7 + i*num_anchor*7 + k*7 + 1] = 0.1; gt_bbox[input_accu_list[b]*num_anchor*7 + i*num_anchor*7 + k*7 + 2] = 0.1; gt_conf[input_accu_list[b]*num_anchor + i*num_anchor + k] = 0; label_idx[input_accu_list[b]*num_anchor + i*num_anchor + k] = -1; for (int j=0; j<nbbox; j++) { // [w, l, h, x, y, z, r, cls, diff_idx] // 0 1 2 3 4 5 6 7 8 float bbox_w = label_bbox[b*nbbox*bbox_attr + j*bbox_attr + 0]; float bbox_l = label_bbox[b*nbbox*bbox_attr + j*bbox_attr + 1]; float bbox_h = label_bbox[b*nbbox*bbox_attr + j*bbox_attr + 2]; float bbox_x = label_bbox[b*nbbox*bbox_attr + j*bbox_attr + 3]; float bbox_y = label_bbox[b*nbbox*bbox_attr + j*bbox_attr + 4]; float bbox_z = label_bbox[b*nbbox*bbox_attr + j*bbox_attr + 5]; float bbox_r = label_bbox[b*nbbox*bbox_attr + j*bbox_attr + 6]; float bbox_cls = label_bbox[b*nbbox*bbox_attr + j*bbox_attr + 7]; float bbox_diff = label_bbox[b*nbbox*bbox_attr + j*bbox_attr + 8]; // printf("bbox:[%.2f,%.2f,%.2f], point:[%.2f,%.2f,%.2f]\n",bbox_w,bbox_l,bbox_h,point_x,point_y,point_z); if (bbox_l*bbox_h*bbox_w > 0) { float rel_point_x = point_x - bbox_x; float rel_point_y = point_y - bbox_y; float rel_point_z = point_z - bbox_z; float rot_rel_point_x = rel_point_x*cosf(bbox_r) + rel_point_y*sinf(bbox_r); float rot_rel_point_y = -rel_point_x*sinf(bbox_r) + rel_point_y*cosf(bbox_r); if (abs(rot_rel_point_x)<=bbox_w * (1 + expand_ratio) / 2 && abs(rot_rel_point_y)<=bbox_l * (1 + expand_ratio) / 2 && abs(rel_point_z)<=bbox_h * (1 + expand_ratio) / 2) { gt_bbox[input_accu_list[b]*num_anchor*7 + i*num_anchor*7 + k*7 + 0] = bbox_w; gt_bbox[input_accu_list[b]*num_anchor*7 + i*num_anchor*7 + k*7 + 1] = bbox_l; gt_bbox[input_accu_list[b]*num_anchor*7 + i*num_anchor*7 + k*7 + 2] = bbox_h; gt_bbox[input_accu_list[b]*num_anchor*7 + i*num_anchor*7 + k*7 + 3] = bbox_x; gt_bbox[input_accu_list[b]*num_anchor*7 + i*num_anchor*7 + k*7 + 4] = bbox_y; gt_bbox[input_accu_list[b]*num_anchor*7 + i*num_anchor*7 + k*7 + 5] = bbox_z; gt_bbox[input_accu_list[b]*num_anchor*7 + i*num_anchor*7 + k*7 + 6] = bbox_r; if (bbox_diff <= diff_thres && bbox_cls <= cls_thres) { // Here we only take cars into consideration, while vans are excluded and give the foreground labels as -1 (ignored). // TODO: need to change the category class accordingly to the expected detection target. gt_conf[input_accu_list[b]*num_anchor + i*num_anchor + k] = 1; label_idx[input_accu_list[b]*num_anchor + i*num_anchor + k] = b * nbbox + j; }else{ gt_conf[input_accu_list[b]*num_anchor + i*num_anchor + k] = -1; } } } } } } } } long long dtime_usec(unsigned long long start){ timeval tv; gettimeofday(&tv, 0); return ((tv.tv_sec*USECPSEC)+tv.tv_usec)-start; } void get_bev_gt_bbox_gpu_launcher(int batch_size, int npoint, int nbbox, int bbox_attr, int num_anchor, int anchor_attr, int diff_thres, int cls_thres, float expand_ratio, const float* input_coors, const float* label_bbox, const int* input_num_list, const float* anchor_param_list, int* input_accu_list, float* gt_bbox, int* gt_conf, int* label_idx) { long long dt = dtime_usec(0); get_bev_gt_bbox_gpu_kernel<<<32,512>>>(batch_size, npoint, nbbox, bbox_attr, num_anchor, anchor_attr, diff_thres, cls_thres, expand_ratio, input_coors, label_bbox, input_num_list, anchor_param_list, input_accu_list, gt_bbox, gt_conf, label_idx); dt = dtime_usec(dt); // std::cout << "Voxel Sample (forward) CUDA time: " << dt/(float)USECPSEC << "s" << std::endl; }
7,762
#include "cuda.h" #include <stdio.h> #include <stdlib.h> #include <sys/time.h> inline double gettime_ms() { struct timeval t; gettimeofday(&t,NULL); return (t.tv_sec+t.tv_usec*1e-6)*1000; } __global__ void mem_kernel(int *arr, int N, int *rdom_arr, int *result) { int tem=0; int random; int tid=blockIdx.x*blockDim.x+threadIdx.x; int totalthreads=blockDim.x*gridDim.x; for(int i=0;i<10;++i){ for(int j=tid ;j<N; j+=totalthreads){ random = rdom_arr[j]; int idx=(j+random)%N; tem += arr[idx]; result[random%N]=tem; } } //write result } int main(int argc, char **argv){ //input: Size N, Utilization U, Affinity A, BlockNum B, ThreadsNum T, Randomizer X int N, U, A, B, T, X; if (argc!=7) { printf("\nInput arguments wrong!\n input: Size N, Utilization U, Affinity A, BlockNum B, ThreadsNum T, Random X \n "); return 0; } N=atoi(argv[1]); U=atof(argv[2]); A=atoi(argv[3]); B=atoi(argv[4]); T=atoi(argv[5]); X=atoi(argv[6]); printf("\ninput: Size N:%d, Utilization U:%d, Affinity A:%d, BlockNum B:%d, ThreadsNum T:%d, Random X:%d \n ",N,U,A,B,T,X); if(N%32!=0) { printf("\nArray size N has to be multiple of 32\n"); return 0; } cudaSetDevice(0); srand(0); int *array_h=(int *)malloc(N*sizeof(int)); for (int i=0;i<N;++i) array_h[i]=1; int *array_d; cudaMalloc(&array_d,N*sizeof(int)); cudaMemcpy(array_d,array_h,N*sizeof(int),cudaMemcpyHostToDevice); int *rdom_arr_h=(int *)malloc(N*sizeof(int)); for (int i=0;i<N;++i) rdom_arr_h[i]= rand() % X; //generate random number in range [0, X) int *rdom_arr_d; cudaMalloc(&rdom_arr_d,N*sizeof(int)); cudaMemcpy(rdom_arr_d,rdom_arr_h,N*sizeof(int),cudaMemcpyHostToDevice); int *result_h=(int *)malloc(N*sizeof(int)); int *result_d; cudaMalloc(&result_d,N*sizeof(int)); double ktime=gettime_ms(); mem_kernel<<<B,T>>>(array_d,N,rdom_arr_d,result_d); cudaDeviceSynchronize(); ktime=gettime_ms()-ktime; FILE* fp=fopen("gpu_result.txt","a+"); fprintf(fp,"%f ",ktime); printf("Kernel time:%f \n",ktime); fclose(fp); cudaMemcpy(result_h,result_d,N*sizeof(int),cudaMemcpyDeviceToHost); cudaDeviceSynchronize(); printf("results:\n"); for(int i=0;i<10;++i){ printf("%d ",result_h[i]); } printf("\n"); free(array_h); free(rdom_arr_h); free(result_h); cudaFree(array_d); cudaFree(rdom_arr_d); cudaFree(result_d); return 0; }
7,763
__global__ void fillOneIntegerArrayKernel( int numberEntries, int numberIterations, int* array, int constant) { int start = blockIdx.x * blockDim.x * numberIterations + threadIdx.x * numberIterations; for(int index = start; index < min(start + numberIterations, numberEntries); index++) { array[index] = constant; } }
7,764
//#ifndef _MATRIXMUL_KERNEL_H_ //#define _MATRIXMUL_KERNEL_H_ /* (define (gpu-info) (let* ([info (cuGPUinfo)]) (values (gridDim-x info) ......))) (: cpyTestDrv_kernel ((Listof Float) Integer -> (Listof Float) (Listof Integer) Integer) (define (cpyTestDrv_kernel d_array_in d_single_in) (let*-values ([(d_array_out) (take d_array_in 0)] [(memstruct) (gpu-info)] [(d_single_out) d_single_in]) (values d_array_out memstruct d_single_out))) */ extern "C" /* Signature: float* d_array_in, int count, uint single_in -> float* d_array_out, int* memstruct, uint* single_out */ __global__ void cpyTestDrv_kernel(float* d_array_out, float* d_array_in, int count, unsigned int* memstruct, unsigned int* d_single_out, unsigned int d_single_in) { // copy - single value *d_single_out = d_single_in; // copy of default variables unsigned int gdm_x = gridDim.x; unsigned int gdm_y = gridDim.y; unsigned int gdm_z = gridDim.z; unsigned int bdm_x = blockDim.x; unsigned int bdm_y = blockDim.y; unsigned int bdm_z = blockDim.z; unsigned int bid_x = blockIdx.x; unsigned int bid_y = blockIdx.y; unsigned int bid_z = blockIdx.z; unsigned int tid_x = threadIdx.x; unsigned int tid_y = threadIdx.y; unsigned int tid_z = threadIdx.z; *memstruct = gdm_x; *(memstruct+1) = gdm_y; *(memstruct+2) = gdm_z; *(memstruct+3) = bdm_x; *(memstruct+4) = bdm_y; *(memstruct+5) = bdm_z; *(memstruct+6) = bid_x; *(memstruct+7) = bid_y; *(memstruct+8) = bid_z; *(memstruct+9) = tid_x; *(memstruct+10) = tid_y; *(memstruct+11) = tid_z; // copy of array variables for(int j = 0 ; j < count ; j++) { *(d_array_out+j) = d_array_in[j]; *(d_array_out+j) = j; } }
7,765
#include "includes.h" __global__ void VanLeerThetaKernel (double *Rsup, double *Rinf, double *Surf, double dt, int nrad, int nsec, int UniformTransport, int *NoSplitAdvection, double *QRStar, double *DensStar, double *Vazimutal_d, double *Qbase) { int j = threadIdx.x + blockDim.x*blockIdx.x; int i = threadIdx.y + blockDim.y*blockIdx.y; double dxrad, invsurf, varq; if (i<nrad && j<nsec){ if ((UniformTransport == NO) || (NoSplitAdvection[i] == NO)){ dxrad = (Rsup[i]-Rinf[i])*dt; invsurf = 1.0/Surf[i]; varq = dxrad*QRStar[i*nsec + j]*DensStar[i*nsec + j]*Vazimutal_d[i*nsec + j]; varq -= dxrad*QRStar[i*nsec + (j+1)%nsec]*DensStar[i*nsec + (j+1)%nsec]*Vazimutal_d[i*nsec + (j+1)%nsec]; Qbase[i*nsec + j] += varq*invsurf; } } }
7,766
#include "includes.h" __global__ void loadOP() { /*output[0] = Load<DF>(input); output[1] = Load<CA>(input + 1); output[2] = Load<CG>(input + 2); output[3] = Load<CS>(input + 3); output[4] = Load<CV>(input + 4); output[5] = Load<NC>(input + 5); output[6] = Load<NC_CA>(input + 6); output[7] = Load<NC_CG>(input + 7); output[8] = Load<NC_CS>(input + 8);*/ }
7,767
template <typename T> __global__ void add1(T * n){ T i = threadIdx.x; n[i] += 1; }
7,768
#include "includes.h" __global__ void checkAggregationFillAggregates(int size, int *adjIndices, int *adjacency, int* aggregation, int* valuesIn, int* valuesOut, int* incomplete) { int idx = blockIdx.x * blockDim.x + threadIdx.x; if (idx < size) { // Find the currently marked distance int currentVal = valuesIn[idx]; int currentAgg = aggregation[idx]; // Checking if any neighbors have a better value int start = adjIndices[idx]; int end = adjIndices[idx + 1]; for (int i = start; i < end; i++) { int neighborAgg = aggregation[adjacency[i]]; int neighborVal = valuesIn[adjacency[i]]; if (neighborAgg == currentAgg && neighborVal > currentVal) { currentVal = neighborVal; incomplete[0] = 1; } } // Write out the distance to the output vector: valuesOut[idx] = currentVal; } }
7,769
extern "C" { __device__ float add(float a, float b) { return a+b; } }
7,770
//******************************************************************************************************** #include <cstdio>// a simple matrix matrix multiplication in CUDA #include <iostream> #include <fstream> #include <vector> #include <cmath> using namespace std; //#ifdef DOLOG //#define LOG(msg) std::cerr<<msg<<std::endl #define LOG(msg) fprintf(stderr, msg "\n"); //#else //#define LOG(msg) //#endif // host code for validating last cuda operation (not kernel launch) //using namespace std; #define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); } inline void gpuAssert(cudaError_t code, const char *file, int line, bool abort=true) { if (code != cudaSuccess) { fprintf(stderr,"GPUassert: %s %s %d\n", cudaGetErrorString(code), file, line); if (abort) exit(code); } } // CPU matrix multiplication: code(the normal way to mulitiply matrices) __global__ void multiply (int width,int height ,int firstCol ,int** A,int** B,int** C,int** A_O,int** B_O, int** C_O) { int x_idx = threadIdx.x + blockIdx.x * blockDim.x; int y_idx = threadIdx.y + blockIdx.y * blockDim.y; if (x_idx < width && y_idx <height ) { //int temp = 0; for (int i = 0; i <3; i++) { //printf ("i= %d", i); //printf ("\n"); for (int j = 0; j <3; j++) { //printf ("j= %d", j); //printf ("\n"); int temp = 0; //printf ("temp= %d", temp); //printf ("\n"); for (int k = 0; k < firstCol; k++) { //printf ("k= %d", k); // printf ("\n"); temp += A[i][k] * B[k][j]; //C is fixed while A is rowWise and B is columnWise C_O[i][j] =temp ; //printf ("Cij= %d", temp); //printf ("\n"); } } } } } //************************************************************************************** __global__ void display (int width, int height,int** A,int** B,int** C ){ for (int i = 0; i <3; i++) { for (int j = 0; j < 3; j++) { //printf ("%d\t", C[i][j]); } printf ("\n"); } } //*********************************************************************************** int main () { int width, height; //vector<vector<int> > A { {1, 2, 3}, {4, 5, 6}, {7, 8, 9} }; //vector<vector<int> > B { {1, 2, 3}, {4, 5, 6}, {7, 8, 9} }; //vector<vector<int> > C { {0, 0, 0}, {0, 0, 0}, {0, 0, 0} }; int A[][3]= { {1, 2, 3}, {4, 5, 6}, {7, 8, 9} }; int B[][3]= { {1, 2, 3}, {4, 5, 6}, {7, 8, 9} }; int C[][3]= { {0, 0, 0}, {0, 0, 0}, {0, 0, 0} }; //********************************************************************************** //Memory allocaction width=3; height=3; int firstCol= 3; int **d_A, **d_B, **d_C; //allocate memory on device //copy matrix to GPU gpuErrchk(cudaMalloc((void**)&d_A, (width)*(height)*sizeof(int))); gpuErrchk(cudaMalloc((void**)&d_B, (width)*(height)*sizeof(int))); gpuErrchk(cudaMalloc((void**)&d_C, (width)*(height)*sizeof(int))); //same int **A_O,**B_O,**C_O; gpuErrchk(cudaMalloc((void**)&A_O,(width)*(height)*sizeof(int))); gpuErrchk(cudaMalloc((void**)&B_O,(width)*(height)*sizeof(int))); gpuErrchk(cudaMalloc((void**)&C_O,(width)*(height)*sizeof(int))); gpuErrchk(cudaMemcpy(d_A, &A[0][0],(width)*(height)*sizeof(int) , cudaMemcpyHostToDevice)); gpuErrchk(cudaMemcpy(d_B, &B[0][0],(width)*(height)*sizeof(int) , cudaMemcpyHostToDevice)); gpuErrchk(cudaMemcpy(d_C, &C[0][0],(width)*(height)*sizeof(int) , cudaMemcpyHostToDevice)); //printf ("matrix A= %d", A); //printf ("matrix B= %d", B); //call kernel dim3 tpb(16,16); dim3 bpg((width+tpb.x-1)/tpb.x, (height+tpb.y-1)/tpb.y); multiply<<<bpg,tpb>>>(width, height , firstCol ,d_A,d_B,d_C, A_O, B_O, C_O); display<<<bpg,tpb>>>(width,height,d_A,d_B,d_C); //copy matrix back to CPU //gpuErrchk(cudaMemcpy(&A[0][0], d_A, (width)*(height)*sizeof(int), cudaMemcpyDeviceToHost)); //gpuErrchk(cudaMemcpy(&B[0][0], d_B, (width)*(height)*sizeof(int), cudaMemcpyDeviceToHost)); gpuErrchk(cudaMemcpy(&C[0][0], (void**)C_O, (width)*(height)*sizeof(int), cudaMemcpyDeviceToHost)); cudaFree(d_A);cudaFree(d_B);cudaFree(d_C);cudaFree(A_O);cudaFree(B_O);cudaFree(C_O); return 0; }
7,771
#include "assignmentHPC1.cuh" #include <iostream> #include <cstdlib> #include <chrono> #include <cmath> using namespace std; using namespace std::chrono; double find_sd_cpu(double *arr_host, unsigned int N) { double mean = 0.0f; for(unsigned int i = 0; i < N; i++) { mean += arr_host[i]; } mean /= N; double result = 0.0f; for(unsigned int i = 0; i < N; i++) { result += pow(arr_host[i] - mean, 2); } result = sqrt(result/N); return result; } void find_sd(double *arr_host, unsigned int N) { // ----------------------------------------- CPU Code ------------------------------------------------- auto startCPU = high_resolution_clock::now(); double result = find_sd_cpu(arr_host, N); auto stopCPU = high_resolution_clock::now(); cout<<"\n\n--------------- CPU ---------------\n"<<endl; cout<<"Answer CPU : "<<result<<endl; cout<<"\nTime on CPU : "<<duration_cast<microseconds>(stopCPU - startCPU).count()/1000<<" milli seconds\n\n"<<endl; }
7,772
#include<iostream> #include<fstream> #include<math.h> #include<string> #include<stdio.h> #include<cuda_runtime.h> #include<device_launch_parameters.h> using namespace std; //#define length 8 #define PI 3.14159265 #define length 8 #define block_len 16 cudaError_t dctWithCuda_1(const double *d, double *D); cudaError_t dctWithCuda_2(const double *f, double *F); /*__global__ void dct1(float *f, float *F){ int tidy = blockIdx.x*blockDim.x + threadIdx.x; int tidx = blockIdx.y*blockDim.y + threadIdx.y; int index = tidx*len + tidy; float tmp; float beta,alfa; if(tidx == 0) beta = sqrt(1.0/length); else beta = sqrt(2.0/length); if(tidy == 0) alfa = sqrt(1.0/length); else alfa = sqrt(2.0/length); if(tidx<length && tidy<length){ for(i=0; i<length; i++){ int x = i/length; int y = i%length; tmp+=((int)data[i])*cos((2*x+1)*tidx*PI/(2.0*length))* cos((2*y+1)*tidy*PI/(2.0*length)); } F[index]=(float)alfa*beta*tmp; } }*/ __global__ void dct_1(const double *f,double *F){ int bid = blockIdx.x; //int tid = threadIdx.x; int i,j; //double data[length]={0.0}; double tmp; if(bid<length){ __shared__ double data[length]; for (i=0; i<length; i++) data[i] = f[bid*length+i];//load row data from f. __syncthreads(); for(i=0; i<length; i++){ if(i==0){ tmp = (double)(1.0/sqrt(1.0*length)); F[bid] = 0;//why use F[bid]? Do transpose at the same time. for(j=0; j<length; j++) F[bid] +=data[j] ; F[bid] *= tmp; } else{ tmp = (double)(sqrt(2.0/(1.0*length))); for(i=1; i<length; i++){ F[i*length+bid] = 0; for(j=0; j<length; j++) F[i*length+bid] += (double)(data[j]*cos((2*j+1)*i*PI/(2*length))); F[i*length+bid] *= tmp; } } } __syncthreads(); for(i=0; i<length; i++) data[i] = F[bid*length+1]; __syncthreads(); for(i=0; i<length; i++){ if(i==0){ tmp=(double)(1.0/sqrt(1.0*length)); F[bid]=0; for(j=0; j<length; j++) F[bid] += data[i]; F[bid] *= tmp; } else{ tmp = (double)(sqrt(2.0/(1.0*length))); for(i=1; i<length; i++){ F[i*length+bid] = 0; for(j=0; j<length; j++) F[i*length+bid] += (double)(data[j]*cos((2*j+1)*i*PI/(2*length))); F[i*length+bid] *= tmp; } } } __syncthreads(); } } __global__ void dct_2(const double *f, double *F){ int tidy = blockIdx.x*blockDim.x + threadIdx.x; int tidx = blockIdx.y*blockDim.y + threadIdx.y; int index = tidx*length + tidy; int i; double tmp; double beta ,alfa; if(tidx == 0) beta = sqrt(1.0/length); else beta = sqrt(2.0/length); if(tidy == 0) alfa = sqrt(1.0/length); else alfa = sqrt(2.0/length); if(tidx<length && tidy<length){ for(i=0; i<length*length; i++){ int x = i/length; int y = i%length; tmp += ((double)f[i])*cos((2*x+1)*tidx*PI/(2.0*length))* cos((2*y+1)*tidy*PI/(2.0*length)); } F[index]=(double)alfa * beta * tmp; } } int main(){ ifstream infile("/home/zhujian/cuda-workspace/dct_10.16/gradient.txt"); int i=0; string line; double f[length*length] = {0.0}; double F[length*length] = {0.0}; while(i<length*length){ if(getline(infile, line)){ f[i] = atof(line.c_str()); cout<<"f[i]: "<<f[i]<<endl; } i++; } cout<<"before"<<endl; for(i=0; i<length*length; i++){ cout<<f[i]<<" "; if ((i+1)%length==0) cout<<endl; } cout<<endl; for(i=0; i<length*length; i++){ cout<<F[i]<<" "; if ((i+1)%length==0) cout<<endl; } cudaError_t cudaStatus = dctWithCuda_1(f,F); if (cudaStatus != cudaSuccess) { fprintf(stderr, "dctWithCuda_1 failed!"); return 1; } cout<<"after"<<endl; for(i=0; i<length*length; i++){ cout<<f[i]<<" "; if ((i+1)%length==0) cout<<endl; } cout<<endl; for(i=0; i<length*length; i++){ cout<<F[i]<<" "; if ((i+1)%length==0) cout<<endl; } return 0; } cudaError_t dctWithCuda_1(const double *d, double *D){ double *dev_d = 0; double *dev_D = 0; cudaError_t cudaStatus; cudaStatus = cudaSetDevice(0); if(cudaStatus != cudaSuccess){ fprintf(stderr, "cudaSetDevice failed! Do you have a CUDA-capable GPU installed?"); goto Error; } cudaStatus = cudaMalloc((void**)&dev_d,length *length* sizeof(double)); if(cudaStatus != cudaSuccess){ fprintf(stderr, "cudaMalloc failed!"); goto Error; } cudaStatus = cudaMalloc((void**)&dev_D,length *length* sizeof(double)); if(cudaStatus != cudaSuccess){ fprintf(stderr, "cudaMalloc failed!"); goto Error; } //copy input vectors from host memory to GPU buffers. cudaStatus = cudaMemcpy(dev_d, d,length *length*sizeof(double),cudaMemcpyHostToDevice); if(cudaStatus != cudaSuccess){ fprintf(stderr, "cudaMemcpy-- failed"); goto Error; } //launch a kernel on the GPU dct_1<<<length,1>>>(dev_d, dev_D); cudaStatus = cudaThreadSynchronize(); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaThreadSynchronize returned error code %d after launching addKernel!\n", cudaStatus); goto Error; } cudaStatus = cudaMemcpy(D, dev_D, length*length* sizeof(double), cudaMemcpyDeviceToHost); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaMemcpy failed!"); goto Error; } Error: cudaFree(dev_d); cudaFree(dev_D); return cudaStatus; } cudaError_t dctWithCuda_2(const double *d, double *D){ double *dev_d = 0; double *dev_D = 0; cudaError_t cudaStatus; cudaStatus = cudaSetDevice(0); if(cudaStatus != cudaSuccess){ fprintf(stderr, "cudaSetDevice failed! Do you have a CUDA-capable GPU installed?"); goto Error; } cudaStatus = cudaMalloc((void**)&dev_d,length * sizeof(double)); if(cudaStatus != cudaSuccess){ fprintf(stderr, "cudaMalloc failed!"); goto Error; } cudaStatus = cudaMalloc((void**)&dev_D,length * sizeof(double)); if(cudaStatus != cudaSuccess){ fprintf(stderr, "cudaMalloc failed!"); goto Error; } //copy input vectors from host memory to GPU buffers. cudaStatus = cudaMemcpy(dev_d, d,length *sizeof(double),cudaMemcpyHostToDevice); if(cudaStatus != cudaSuccess){ fprintf(stderr, "cudaMalloc failed"); goto Error; } //launch a kernel on the GPU dct_2<<<1, (length/block_len)*(length/block_len), block_len*block_len>>>(dev_d, dev_D); cudaStatus = cudaThreadSynchronize(); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaThreadSynchronize returned error code %d after launching addKernel!\n", cudaStatus); goto Error; } cudaStatus = cudaMemcpy(D, dev_D, length*length * sizeof(double), cudaMemcpyDeviceToHost); if (cudaStatus != cudaSuccess) { fprintf(stderr, "cudaMemcpy failed!"); goto Error; } Error: cudaFree(dev_d); cudaFree(dev_D); return cudaStatus; }
7,773
#include <stdio.h> /* -------- KERNEL -------- */ __global__ void reduce_kernel(int * d_out, int * d_in, int size) { // position and threadId int pos = blockIdx.x * blockDim.x + threadIdx.x; int tid = threadIdx.x; // do reduction in global memory for (unsigned int s = blockDim.x / 2; s>0; s>>=1) { if (tid < s) { if (pos+s < size) // Handling out of bounds { d_in[pos] = d_in[pos] + d_in[pos+s]; } } __syncthreads(); } // only thread 0 writes result, as thread if ((tid==0) && (pos < size)) { d_out[blockIdx.x] = d_in[pos]; } } /* -------- KERNEL WRAPPER -------- */ void reduce(int * d_out, int * d_in, int size, int num_threads) { // setting up blocks and intermediate result holder int num_blocks; if(((size) % num_threads)) { num_blocks = ((size) / num_threads) + 1; } else { num_blocks = (size) / num_threads; } int * d_intermediate; cudaMalloc(&d_intermediate, sizeof(int)*num_blocks); cudaMemset(d_intermediate, 0, sizeof(int)*num_blocks); int prev_num_blocks; int i = 1; int size_rest = 0; // recursively solving, will run approximately log base num_threads times. do { printf("Round:%.d\n", i); printf("NumBlocks:%.d\n", num_blocks); printf("NumThreads:%.d\n", num_threads); printf("size of array:%.d\n", size); i++; reduce_kernel<<<num_blocks, num_threads>>>(d_intermediate, d_in, size); size_rest = size % num_threads; size = size / num_threads + size_rest; // updating input to intermediate cudaMemcpy(d_in, d_intermediate, sizeof(int)*num_blocks, cudaMemcpyDeviceToDevice); // Updating num_blocks to reflect how many blocks we now want to compute on prev_num_blocks = num_blocks; if(size % num_threads) { num_blocks = size / num_threads + 1; } else { num_blocks = size / num_threads; } // updating intermediate cudaFree(d_intermediate); cudaMalloc(&d_intermediate, sizeof(int)*num_blocks); } while(size > num_threads); // if it is too small, compute rest. // computing rest reduce_kernel<<<1, size>>>(d_out, d_in, prev_num_blocks); } /* -------- MAIN -------- */ int main(int argc, char **argv) { printf("@@STARTING@@ \n"); // Setting num_threads int num_threads = 512; // Making non-bogus data and setting it on the GPU const int size = 1<<19; const int size_out = 1; int * d_in; int * d_out; cudaMalloc(&d_in, sizeof(int)*size); cudaMalloc(&d_out, sizeof(int)*size_out); int * h_in = (int *)malloc(size*sizeof(int)); for (int i = 0; i < size; i++) h_in[i] = 1; cudaMemcpy(d_in, h_in, sizeof(int)*size, cudaMemcpyHostToDevice); // Running kernel wrapper reduce(d_out, d_in, size, num_threads); int result; cudaMemcpy(&result, d_out, sizeof(int), cudaMemcpyDeviceToHost); printf("\nFINAL SUM IS: %d\n", result); }
7,774
// Author: Brian Nguyen #include <stdio.h> #include <stdlib.h> #include <sys/time.h> #include <math.h> #define N 5000 // number of array elements #define B 64 // number of elements in a block __global__ void scan(float *g_odata, float *g_idata, int n); __global__ void prescan(float *g_odata, float *g_idata, int n, float *g_sums); __global__ void uniform_add(float *o_array, float *sum_array); void scanCPU(float *f_out, float *f_in, int i_n); double myDiffTime(struct timeval &start, struct timeval &end) { /* Calculate the time difference. */ double d_start, d_end; d_start = (double)(start.tv_sec + start.tv_usec/1000000.0); d_end = (double)(end.tv_sec + end.tv_usec/1000000.0); return (d_end - d_start); } int main() { timeval start, end; int new_N; // temporary pointer arrays for computation float *dev_a, *dev_g, *dev_sums; double d_gpuTime, d_cpuTime; // handle padding of the array for non-powers of 2 if ((N % B) != 0) { new_N = ((N + B - 1) / B) * B; } else{ new_N = N; } printf("%d", new_N); //float a[new_N]; float *a; a = (float *) malloc(new_N * sizeof(float)); // initialize matrix a with random floats between 0 and 1000 for (int i = 1; i <= N; i++) { a[i-1] = i; } for (int i = N; i < new_N; i++) { a[i] = 0; } int size = new_N * sizeof(float); int grid_size = ceil(new_N / B); // size of grids for first prefix-scan int grid_size2 = ceil(grid_size / B); // size of grids for second prefix-scan int thread_size = B / 2; // thread size for each block int size_sums = grid_size * sizeof(float); int size_sums2 = grid_size2 * sizeof(float); float c[N], g[new_N], sums[grid_size]; // CPU version (serial) of prefix-sum gettimeofday(&start, NULL); scanCPU(c, a, N); gettimeofday(&end, NULL); d_cpuTime = myDiffTime(start, end); // START OF FIRST PRE-SCAN RUN // initialize a and b matrices here for CUDA cudaMalloc((void **) &dev_a, size); cudaMalloc((void **) &dev_g, size); cudaMalloc((void **) &dev_sums, size_sums); // GPU version (CUDA) of prefix-sum gettimeofday(&start, NULL); cudaMemcpy(dev_a, a, size, cudaMemcpyHostToDevice); // work-efficient scan for SUMS array prescan<<<grid_size, thread_size, B*sizeof(float)>>>(dev_g, dev_a, new_N, dev_sums); cudaDeviceSynchronize(); cudaMemcpy(g, dev_g, size, cudaMemcpyDeviceToHost); cudaMemcpy(sums, dev_sums, size_sums, cudaMemcpyDeviceToHost); cudaFree(dev_a); cudaFree(dev_g); cudaFree(dev_sums); // START OF SECOND PRE-SCAN RUN float inc[grid_size], sums_inc[grid_size2], inc_final[grid_size2]; float *dev_inc, *dev_sums_inc, *dev_inc_final, *dev_sums_input; cudaMalloc((void **) &dev_sums_input, size_sums); cudaMalloc((void **) &dev_inc, size_sums); cudaMalloc((void **) &dev_sums_inc, size_sums2); cudaMemcpy(dev_sums_input, sums, size_sums, cudaMemcpyHostToDevice); prescan<<<grid_size2, thread_size, B*sizeof(float)>>>(dev_inc, dev_sums_input, grid_size2, dev_sums_inc); cudaDeviceSynchronize(); cudaMemcpy(inc, dev_inc, size_sums, cudaMemcpyDeviceToHost); cudaMemcpy(sums_inc, dev_sums_inc, size_sums2, cudaMemcpyDeviceToHost); cudaFree(dev_inc); cudaFree(dev_sums_inc); cudaFree(dev_sums_input); scanCPU(inc_final, sums_inc, size_sums2); // START OF UPDATING SUMS float g2[grid_size]; float *dev_g2; cudaMalloc((void **) &dev_g2, size); cudaMalloc((void **) &dev_inc_final, size_sums); cudaMemcpy(dev_inc_final, inc_final, size_sums, cudaMemcpyHostToDevice); cudaMemcpy(dev_g2, inc, size_sums, cudaMemcpyHostToDevice); uniform_add<<<grid_size2, thread_size, B*sizeof(float)>>>(dev_g2, dev_inc_final); cudaDeviceSynchronize(); cudaMemcpy(g2, dev_g2, size, cudaMemcpyDeviceToHost); cudaFree(dev_g2); cudaFree(dev_inc_final); // START OF FINAL UPDATE TO FIRST PREFIX SCAN float g3[new_N]; float *dev_g3, *dev_first_add; cudaMalloc((void **) &dev_g3, size); cudaMalloc((void **) &dev_first_add, size_sums); cudaMemcpy(dev_first_add, g2, size_sums, cudaMemcpyHostToDevice); cudaMemcpy(dev_g3, g, size, cudaMemcpyHostToDevice); uniform_add<<<grid_size, thread_size, B*sizeof(float)>>>(dev_g3, dev_first_add); cudaDeviceSynchronize(); cudaMemcpy(g3, dev_g3, size, cudaMemcpyDeviceToHost); gettimeofday(&end, NULL); d_gpuTime = myDiffTime(start, end); cudaFree(dev_g3); cudaFree(dev_first_add); free(a); // display results of the prefix-sum for (int i = 0; i < N; i++) { printf("c[%i] = %0.3f, g3[%i] = %0.3f\n", i, c[i], i, g3[i]); //if (c[i] != g[i]) //{ // printf("Results do not match! c[%i]=%f, g[%i]=%f\n", i, c[i], i, g[i]); // break; //} } printf("GPU Time for scan size %i: %f\n", N, d_gpuTime); printf("CPU Time for scan size %i: %f\n", N, d_cpuTime); return 0; } __global__ void scan(float *g_odata, float *g_idata, int n) { /* CUDA Naive Scan Algorithm (double buffered). */ extern __shared__ float temp[]; // allocated on invocation int thid = threadIdx.x; int pout = 0, pin = 1; // Load input into shared memory. // This is exclusive scan, so shift right by one // and set first element to 0 temp[thid] = (thid > 0) ? g_idata[thid-1] : 0; __syncthreads(); for (int offset = 1; offset < n; offset *= 2) { pout = 1 - pout; // swap double buffer indices pin = 1 - pout; if (thid >= offset) temp[pout*n+thid] += temp[pin*n+thid - offset]; else temp[pout*n+thid] = temp[pin*n+thid]; __syncthreads(); } g_odata[thid] = temp[pout*n+thid]; // write output } __global__ void prescan(float *g_odata, float *g_idata, int n, float *g_sums) { /* CUDA Work-Efficient Scan Algorithm. */ extern __shared__ float temp[]; // allocated on invocation int thid = threadIdx.x; // thread id of a thread in a block int gthid = (blockIdx.x * blockDim.x) + thid; // global thread id of grid int offset = 1; /* // for each thread in a block, put data into shared memory if (gthid > n) { // handle non-power of two arrays by padding elements in last block temp[2*thid] = 0; temp[2*thid+1] = 0; } else { // grab data from input array temp[2*thid] = g_idata[2*gthid]; temp[2*thid+1] = g_idata[2*gthid+1]; } */ temp[2*thid] = g_idata[2*gthid]; temp[2*thid+1] = g_idata[2*gthid+1]; // build sum in place up the tree for (int d = B>>1; d > 0; d >>= 1) { __syncthreads(); if (thid < d) { int ai = offset*(2*thid+1)-1; int bi = offset*(2*thid+2)-1; temp[bi] += temp[ai]; } offset *= 2; } if (thid == 0) { g_sums[blockIdx.x] = temp[B - 1]; temp[B - 1] = 0; } // clear the last element // traverse down tree & build scan for (int d = 1; d < B; d *= 2) { offset >>= 1; __syncthreads(); if (thid < d) { int ai = offset*(2*thid+1)-1; int bi = offset*(2*thid+2)-1; float t = temp[ai]; temp[ai] = temp[bi]; temp[bi] += t; } } __syncthreads(); // write results to device memory g_odata[2*gthid] = temp[2*thid]; g_odata[2*gthid+1] = temp[2*thid+1]; } __global__ void uniform_add(float *o_array, float *sum_array) { int bid = blockIdx.x; int gthid = (bid * blockDim.x) + threadIdx.x; // global thread id of grid o_array[2*gthid] = o_array[2*gthid] + sum_array[bid]; o_array[2*gthid+1] = o_array[2*gthid+1] + sum_array[bid]; } void scanCPU(float *f_out, float *f_in, int i_n) { /* Apply all-prefix sums to an array on the CPu without parallelization. */ f_out[0] = 0; /* for each array element, the value is the previous sum plus the current array value */ for (int i = 1; i < i_n; i++) f_out[i] = f_out[i-1] + f_in[i-1]; }
7,775
#include <stdio.h> #include <assert.h> #include "cuda.h" #define N 2 __global__ void foo(int* p) { p[threadIdx.x] = 2; __syncthreads(); }
7,776
#include <stdio.h> #include <stdlib.h> #include <sys/time.h> #include <math.h> double get_walltime() { struct timeval tp; gettimeofday(&tp, NULL); return (double) (tp.tv_sec + tp.tv_usec*1e-6); } void file_results(double time_elapsed, int k) { FILE *fptr; fptr = fopen("cuda_times.txt", "a"); if(fptr == NULL){ printf("\nError creating or opening file\n"); exit(1); } fprintf(fptr, "cuda\npower of 2: %d\ntime elapsed: %f\n\n", k, time_elapsed); fclose(fptr); } void compare_and_swap(int *arr,int a, int b, int dir){ if(dir == (arr[a] > arr[b])){ int temp = arr[a]; arr[a] = arr[b]; arr[b] = temp; } } void generate_vector(int *D, int n){ for (size_t i = 0; i < n; i++) D[i] = rand()%100; } void print_array(int *D, int n) { for (size_t i = 0; i < n; i++) { printf("%d ", D[i]); } } __global__ void bitonic_sort_step(int *D, int n, int a, int b){ int i, ixa, temp; i = threadIdx.x + (blockDim.x * blockIdx.x); ixa = i^a; if (ixa > i){ if ((i & b) == 0){ if (D[i] > D[ixa]){ temp = D[i]; D[i] = D[ixa]; D[ixa] = temp; } } if ((i & b) != 0){ if (D[i] < D[ixa]){ temp = D[i]; D[i] = D[ixa]; D[ixa] = temp; } } } //print_array(D, n); //printf("\t\ti: %d\tixa: %d\ta: %d\tb: %d\n", i, ixa, a, b); } void bitonic_sort(int *D, int n, int k){ int a, b; int thread_num=512; int block_num=1; int *dev_D; if (k > 9){ block_num = pow(2, k-8); } while (thread_num > n){ thread_num >>= 1; } cudaMalloc((void**) &dev_D, sizeof(int) * n); cudaMemcpy(dev_D, D, sizeof(int) * n, cudaMemcpyHostToDevice); dim3 blocks(block_num); dim3 threads(thread_num); for (b = 2; b <= n; b <<= 1){ for (a = b>>1; a > 0; a >>= 1){ bitonic_sort_step<<<blocks, threads>>>(dev_D, n, a, b); } } cudaMemcpy(D, dev_D, sizeof(int) * n, cudaMemcpyDeviceToHost); cudaFree(dev_D); } int main(int argc, char const *argv[]) { /* code */ int *D; // Array of values int n, k; // 2^k = n, number of elements in the array double time0, time1; if (argc > 1){ k = atoi(argv[1]); } else { printf("number of nodes = 2^k\nenter the value of k: \n"); scanf("%d", &k); } if (argc > 2){ srand(atoi(argv[2])); } else { srand(get_walltime()); } printf("\npopulating array\n"); n = pow(2, k); D = (int*)malloc(sizeof(int)*n); generate_vector(D, n); if (n < 33){ printf("printing array\n\n"); for (size_t i = 0; i < n; i++) { printf("%d ", D[i]); } } printf("\n\nbegin bitonic sort\n"); time0 = get_walltime(); bitonic_sort(D, n, k); time1 = get_walltime(); if (n < 33){ printf("finalize bitonic sort\n\n"); for (size_t i = 0; i < n; i++) { printf("%d ", D[i]); } } printf("\n\ntime elapsed: %f s\n\n ", time1 - time0); file_results(time1-time0, k); free(D); return 0; }
7,777
#include "includes.h" // Copyright 2019 Adam Campbell, Seth Hall, Andrew Ensor // Copyright 2019 High Performance Computing Research Laboratory, Auckland University of Technology (AUT) // Redistribution and use in source and binary forms, with or without // modification, are permitted provided that the following conditions are met: // 1. Redistributions of source code must retain the above copyright notice, // this list of conditions and the following disclaimer. // 2. Redistributions in binary form must reproduce the above copyright // notice, this list of conditions and the following disclaimer in the // documentation and/or other materials provided with the distribution. // 3. Neither the name of the copyright holder nor the names of its // contributors may be used to endorse or promote products derived from this // software without specific prior written permission. // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" // AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE // ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE // LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR // CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF // SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS // INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN // CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) // ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE // POSSIBILITY OF SUCH DAMAGE. /** * Check the return value of the CUDA runtime API call and exit * the application if the call has failed. */ __device__ double2 complex_multiply(double2 z1, double2 z2) { double real = z1.x*z2.x - z1.y*z2.y; double imag = z1.y*z2.x + z1.x*z2.y; return make_double2(real, imag); } __global__ void inverse_dft_with_w_correction(double2 *grid, size_t grid_pitch, const double3 *visibilities, const double2 *vis_intensity, int vis_count, int batch_count, int x_offset, int y_offset, int render_size, double cell_size) { // look up id of thread int idx = blockIdx.x*blockDim.x + threadIdx.x; int idy = blockIdx.y*blockDim.y + threadIdx.y; if(idx >= render_size || idy >= render_size) return; double real_sum = 0; double imag_sum = 0; // convert to x and y image coordinates double x = (idx+x_offset) * cell_size; double y = (idy+y_offset) * cell_size; double2 vis; double2 theta_complex = make_double2(0.0, 0.0); // precalculate image correction and wCorrection double image_correction = sqrt(1.0 - (x * x) - (y * y)); double w_correction = image_correction - 1.0; // NOTE: below is an approximation... Uncomment if needed // double wCorrection = -((x*x)+(y*y))/2.0; // loop through all visibilities and create sum using iDFT formula for(int i = 0; i < batch_count; ++i) { double theta = 2.0 * M_PI * (x * visibilities[i].x + y * visibilities[i].y + (w_correction * visibilities[i].z)); sincos(theta, &(theta_complex.y), &(theta_complex.x)); vis = complex_multiply(vis_intensity[i], theta_complex); real_sum += vis.x; imag_sum += vis.y; } // adjust sum by image correction real_sum *= image_correction; imag_sum *= image_correction; // look up destination in image (grid) and divide by amount of visibilities (N) double2 *row = (double2*)((char*)grid + idy * grid_pitch); row[idx].x += (real_sum / vis_count); row[idx].y += (imag_sum / vis_count); }
7,778
#include "includes.h" __global__ void add(float *a, float *b, int nx, int ny) { int x = blockIdx.x * blockDim.x + threadIdx.x; int y = blockIdx.y * blockDim.y + threadIdx.y; int idx = x + y*nx; if (x<nx && y<ny) b[idx] += a[idx] * .125; }
7,779
#include<iostream> using namespace std; #define THREADS_PER_BLOCK 256 __global__ void max_per_block(int *a_d,int *b_d,int n){ int block = blockDim.x*blockIdx.x; int max=0; for(int i=block;i<min(block+blockDim.x,n);i++){ if(max<a_d[i]){ max=a_d[i]; } } b_d[blockIdx.x]=max; } int main() { int n; cout<<"Enter the no of elements"; cin>>n; int *arr = new int[n]; for(int i=0;i<n;i++){ arr[i]=i+1; } int no_of_blocks = (n+THREADS_PER_BLOCK-1)/THREADS_PER_BLOCK; int size = n*sizeof(int); int *arr_d,*b_d; cudaMalloc(&arr_d,size); cudaMalloc(&b_d,no_of_blocks*sizeof(int)); cudaMemcpy(arr_d,arr,size,cudaMemcpyHostToDevice); while(n>1){ max_per_block<<<no_of_blocks,THREADS_PER_BLOCK>>>(arr_d,b_d,n); n=(n+THREADS_PER_BLOCK-1)/THREADS_PER_BLOCK; cudaMemcpy(arr_d,b_d,no_of_blocks*sizeof(int),cudaMemcpyDeviceToDevice); } int ans; cudaMemcpy(&ans,arr_d,sizeof(int),cudaMemcpyDeviceToHost); cout<<ans; // clock_t cpu_start = clock(); // clock_t cpu_stop = clock(); // clock_t cpu_elapsed_time = 1000*(cpu_stop - cpu_start)/CLOCKS_PER_SEC; //int( pow(double(input[i]- *mean),2.0)); // cudaEvent_t gpu_start,gpu_stop; // cudaEventCreate(&gpu_start); // cudaEventCreate(&gpu_stop); // cudaEventRecord(gpu_start,0); // cudaEventRecord(gpu_stop, 0); // cudaEventSynchronize(gpu_stop); // cudaEventElapsedTime(&gpu_elapsed_time, gpu_start, gpu_stop); }
7,780
#include <iostream> #include <stdio.h> #include <time.h> #include <stdlib.h> #include <cuda_runtime.h> #include <cmath> #define VECTOR_SIZE 4096 #define VECTOR_NUM 4096 #define blocksize 8 #define TILE_WIDTH 16 using namespace std; __global__ void rbf_cuda(const float *a, const float *b, float *c, int vec_size, int vec_num) // use cuda to calculate rbf with only global memory { // get the thread parameters int index_a = blockIdx.y * blockDim.y + threadIdx.y; int index_b = blockIdx.x * blockDim.x + threadIdx.x; float tmp = 0; //if (index_a >= VECTOR_NUM || index_b >= VECTOR_NUM) return; for (int i = 0; i < vec_size; ++i) { tmp += (a[index_a * vec_size + i] - b[index_b * vec_size + i]) * \ (a[index_a * vec_size + i] - b[index_b * vec_size + i]); } tmp = sqrt(tmp); c[index_a * vec_num + index_b] = tmp; } __global__ void rbf_cuda_shareMem(const float *a, const float *b, float *c, int vec_size, int vec_num) // use cuda to calculate rbf with global memory and shared memory { // shared memory allocation __shared__ float a_s[TILE_WIDTH][TILE_WIDTH]; __shared__ float b_s[TILE_WIDTH][TILE_WIDTH]; // thread parameters int index_a = blockIdx.y * blockDim.y + threadIdx.y; int index_b = blockIdx.x * blockDim.x + threadIdx.x; int ty = threadIdx.y; int tx = threadIdx.x; float tmp = 0; for (int i = 0; i < vec_size / TILE_WIDTH; i++) { // in each tiling, a thread load two data into shared memory // which will be used for the whole thread block. a_s[ty][tx] = \ a[index_a * vec_size + i * TILE_WIDTH + tx]; b_s[ty][tx] = \ b[index_b * vec_size + i * TILE_WIDTH + tx]; // synchronize threads to avoid data race __syncthreads(); for (int j = 0; j < TILE_WIDTH; j++) { tmp += (a_s[ty][j] - b_s[ty][j]) * \ (a_s[ty][j] - b_s[ty][j]); } __syncthreads(); } c[index_a * vec_num + index_b] = sqrt(tmp); } void rbf_cpu(float *a, float *b, float *c) // naive CPU function { float tmp = 0; for (int i = 0; i < VECTOR_NUM; i++) for (int j = 0; j < VECTOR_NUM; j++) { tmp = 0; for (int k = 0; k < VECTOR_SIZE; k++) { tmp += (a[i * VECTOR_SIZE + k] - b[j * VECTOR_SIZE + k]) * \ (a[i * VECTOR_SIZE + k] - b[j * VECTOR_SIZE + k]); } c[i * VECTOR_NUM + j] = sqrt(tmp); } } void genNum(float *m, int num, int size, float value) { for (int i = 0; i < num; ++i) for (int j = 0; j < size; ++j) m[i * size + j] = value; } int main(int argc, char *argv[]) { // parse command line int mode = 0; if (argc == 1) { cout << "Too few arguments, please add \'-c\', \'-n\' or \'-s\'.\n"; return 0; } if (!strcmp(argv[1], "-c")) mode = 0; else if (!strcmp(argv[1], "-n")) mode = 1; else if (!strcmp(argv[1], "-s")) mode = 2; else { cout << "Invalid run mode!\n"; return 0; } clock_t start, end; // memory on host(CPU) float *a, *b, *c; int calMatrixSize = sizeof(float) * VECTOR_SIZE * VECTOR_NUM; int ansMatrixSize = sizeof(float) * VECTOR_NUM * VECTOR_NUM; a = (float*)malloc(calMatrixSize); b = (float*)malloc(calMatrixSize); c = (float*)malloc(ansMatrixSize); // initialize the memory of matrix on CPU genNum(a, VECTOR_NUM, VECTOR_SIZE, 1.0); genNum(b, VECTOR_NUM, VECTOR_SIZE, 2.0); genNum(c, VECTOR_NUM, VECTOR_NUM, 3.0); // only for debug, can be deleted later // memory on device(GPU) float *d_a, *d_b, *d_c; cudaMalloc((void**)&d_a, calMatrixSize); cudaMalloc((void**)&d_b, calMatrixSize); cudaMalloc((void**)&d_c, ansMatrixSize); // cpu mode if (mode == 0) { cout << "Using CPU to calculate...\n"; // start timing start = clock(); rbf_cpu(a, b, c); end = clock(); } // normal mode else if (mode == 1) { cout << "Using GPU with global memory to calculate...\n"; // prepare the size of block and grid dim3 blockSize(blocksize, blocksize); dim3 gridSize((VECTOR_NUM + blocksize - 1) / blocksize, (VECTOR_NUM + blocksize - 1) / blocksize); // start timing start = clock(); // copy the matrix from host to device cudaMemcpy((void*)d_a, (void*)a, calMatrixSize, cudaMemcpyHostToDevice); cudaMemcpy((void*)d_b, (void*)b, calMatrixSize, cudaMemcpyHostToDevice); // calculate RBF kernel function rbf_cuda<<<gridSize, blockSize>>>(d_a, d_b, d_c, VECTOR_SIZE, VECTOR_NUM); // copy the ans matrix from device to host cudaMemcpy((void*)c, (void*)d_c, ansMatrixSize, cudaMemcpyDeviceToHost); end = clock(); } // shared memory mode else { cout << "Using GPU with global and shared memory to calculate...\n"; // prepare the size of block and grid dim3 blockSize(TILE_WIDTH, TILE_WIDTH); dim3 gridSize((VECTOR_NUM + TILE_WIDTH - 1) / TILE_WIDTH, (VECTOR_NUM + TILE_WIDTH - 1) / TILE_WIDTH); // start timing start = clock(); // copy the matrix from host to device cudaMemcpy((void*)d_a, (void*)a, calMatrixSize, cudaMemcpyHostToDevice); cudaMemcpy((void*)d_b, (void*)b, calMatrixSize, cudaMemcpyHostToDevice); // calculate RBF kernel function rbf_cuda_shareMem<<<gridSize, blockSize>>>(d_a, d_b, d_c, VECTOR_SIZE, VECTOR_NUM); // copy the ans matrix from device to host cudaMemcpy((void*)c, (void*)d_c, ansMatrixSize, cudaMemcpyDeviceToHost); end = clock(); } cout << "Time consuming: " << double(end - start) / CLOCKS_PER_SEC << "s\n"; // check the error float max_error = 0.0; float correct_ans = sqrt(VECTOR_SIZE); int error_cnt = 0; for (int i = 0; i < VECTOR_NUM; ++i) for (int j = 0; j < VECTOR_NUM; ++j) { float error = c[i * VECTOR_NUM + j] - correct_ans; error = error >= 0 ? error : -1 * error; if (error != 0.0) error_cnt++; if (error > max_error) max_error = error; } cout << "Max error: " << max_error << endl; cout << "Total error counts: " << error_cnt << endl; // free all memory cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); free(a); free(b); free(c); return 0; }
7,781
/** * simple hello world program, demonstrated that nvcc is fully downward-compatible with c */ #include <stdio.h> int main(void) { printf("Hello World! 123\n"); return 0; }
7,782
#include "includes.h" /** * Given a input tensor x with shape (N, C, D), compute x.mean(2).mean(0) * This function is useful in batch normalization. * Refer to https://people.maths.ox.ac.uk/gilesm/cuda/prac4/reduction.pdf. * But the unrolling warps seems to be not working correctly for now. */ const int N = 256; const int C = 1024; const int D = 28*28; __global__ void reduce0(const float* in, float* out) { __shared__ float buffer[CUDA_NUM_THREADS]; const unsigned int tid = threadIdx.x; const unsigned int c = blockIdx.x; // load and accumulate data to buffer buffer[tid] = 0; for (int i = tid; i < N * D; i += blockDim.x) { const unsigned int n = i / D; const unsigned int d = i % D; const unsigned int index = n * C * D + c * D + d; buffer[tid] += in[index]; } __syncthreads(); // do tree reduction in buffer for (int s = 1; s < blockDim.x; s *= 2) { if (tid % (2*s) == 0) { // <-- bad: divergent branching buffer[tid] += buffer[tid + s]; } __syncthreads(); } if (tid == 0) out[c] = buffer[0] / (N * D); }
7,783
#include "includes.h" __global__ void skip_add(size_t sz, float_t* f1, float* skip_out_sum) { size_t index = blockDim.x * blockIdx.x + threadIdx.x; if(index < sz) { skip_out_sum[index] += f1[index]; } }
7,784
#include <thrust/host_vector.h> #include <thrust/device_vector.h> #include <thrust/extrema.h> #include <iostream> #include <cstdlib> int main(int argc, char *argv[]) { if (argc < 2 || argc > 2) { std::cout << "Usage: " << argv[0] << " size\n"; return -1; } const size_t len = 1L << (atoi(argv[1])); thrust::host_vector<int> h_a(len); srand(0); clock_t begin = clock(); for (size_t i = 0; i < len; ++i) { h_a[i] = rand() / (float)RAND_MAX; } std::cout << "Elapsed time: " << double(clock() - begin) / CLOCKS_PER_SEC * 1000 << " ms\n"; // Copy host_vector to device_vector thrust::device_vector<int> d_a = h_a; thrust::min_element(d_a.begin(), d_a.end()); return 0; }
7,785
#define TILE_SIZE 4 #define KBLOCK 8 #define TILE_SIZEB 8 #define TILE_M 8 #define TILE_N 4 #define BLOCK_M 128 #define BLOCK_N 64 #define BLOCK_SIZEB 128 #define BLOCK_SIZE 64 #define KBLOCKB 8 #define STRIDE BLOCK_SIZEB/TILE_SIZEB #define M 2560 #define N 2048 #define K 4096 #include <cuda_runtime.h> #include <stdlib.h> #include <stdio.h> /** Algorithm 0: The most naive CUDA matmul kernel. The result serves as the baseline. */ __global__ void matmul_0(float *A, float *B, float *C) { int m_i = blockIdx.y * blockDim.y + threadIdx.y; int n_i = blockIdx.x * blockDim.x + threadIdx.x; float c = 0; for(int i=0; i<K; i++){ c += A[m_i*K + i]*B[i*N + n_i]; } if(m_i < M && n_i < N){ C[m_i*N + n_i] = c; } } __global__ void matmul_s(float *A, float *B, float *C) { int m_b = blockIdx.y * blockDim.y; int n_b = blockIdx.x * blockDim.x; int tx = threadIdx.x; int ty = threadIdx.y; int m_i = m_b + ty; int n_i = n_b + tx; float c = 0; __shared__ float sa[16][16]; __shared__ float sb[16][16]; for(int k_i=0; k_i<K; k_i += 16){ sa[ty][tx] = A[(m_i)*K + k_i + tx]; sb[ty][tx] = B[(k_i + ty)*N + n_i]; __syncthreads(); for(int k=0; k<16; k++){ c += sa[ty][k]*sb[k][tx]; } __syncthreads(); } if(m_i < M && n_i < N){ C[m_i*N + n_i] = c; } } __global__ void matmul_1(float *A, float *B, float *C) { int m_i = blockIdx.y * BLOCK_SIZE + threadIdx.y * TILE_SIZE; int n_i = blockIdx.x * BLOCK_SIZE + threadIdx.x * TILE_SIZE; float a[TILE_SIZE][KBLOCK]; float b[KBLOCK][TILE_SIZE]; float c[TILE_SIZE][TILE_SIZE]; for(int i=0; i<TILE_SIZE; i++){ for(int j=0; j<TILE_SIZE; j++){ c[i][j]=0; } } for(int k_i=0; k_i<K; k_i += KBLOCK){ for(int i=0; i<TILE_SIZE; i++){ for(int k=0; k<KBLOCK; k++){ a[i][k] = A[(m_i+i)*K + k_i + k]; b[k][i] = B[(k_i + k)*N + n_i + i]; } } for(int i=0; i<TILE_SIZE; i++){ for(int j=0; j<TILE_SIZE; j++){ #pragma unroll for(int k=0; k<KBLOCK; k++) c[i][j] += a[i][k] * b[k][j]; } } } for(int i=0; i<TILE_SIZE; i++){ for(int j=0; j<TILE_SIZE; j++){ if(m_i+i < M && n_i+j < N){ C[(m_i+i)*N + n_i + j] = c[i][j]; } } } } __global__ void matmul_2(float *A, float *B, float *C) { int m_b = blockIdx.y * BLOCK_SIZE; int n_b = blockIdx.x * BLOCK_SIZE; int m_t = threadIdx.y * TILE_SIZE; int n_t = threadIdx.x * TILE_SIZE; int m_i = m_b + m_t; int n_i = n_b + n_t; float a[TILE_SIZE][KBLOCK]; float b[KBLOCK][TILE_SIZE]; float c[TILE_SIZE][TILE_SIZE]; __shared__ float sa[BLOCK_SIZE*KBLOCK]; __shared__ float sb[KBLOCK*BLOCK_SIZE]; for(int i=0; i<TILE_SIZE; i++){ for(int j=0; j<TILE_SIZE; j++){ c[i][j]=0; } } for(int k_i=0; k_i<K; k_i += KBLOCK){ for (int t=threadIdx.y*blockDim.x + threadIdx.x; t<BLOCK_SIZE*KBLOCK; t += blockDim.x*blockDim.y){ sa[t] = A[(m_b + t/KBLOCK)*K + k_i + t%KBLOCK]; sb[t] = B[(k_i + t/BLOCK_SIZE)*N + n_b + t%BLOCK_SIZE]; } __syncthreads(); for(int i=0; i<TILE_SIZE; i++){ for(int k=0; k<KBLOCK; k++){ a[i][k] = sa[(m_t + i)*KBLOCK + k]; b[k][i] = sb[k*BLOCK_SIZE + n_t + i]; } } for(int i=0; i<TILE_SIZE; i++){ for(int j=0; j<TILE_SIZE; j++){ #pragma unroll for(int k=0; k<KBLOCK; k++) c[i][j] += a[i][k] * b[k][j]; } } __syncthreads(); } for(int i=0; i<TILE_SIZE; i++){ for(int j=0; j<TILE_SIZE; j++){ if(m_i+i < M && n_i+j < N){ C[(m_i+i)*N + n_i + j] = c[i][j]; } } } } __global__ void matmul_3(float *A, float *B, float *C) { int m_b = blockIdx.y * BLOCK_SIZEB; int n_b = blockIdx.x * BLOCK_SIZEB; int m_t = threadIdx.y * TILE_SIZEB; int n_t = threadIdx.x * TILE_SIZEB; int m_i = m_b + m_t; int n_i = n_b + n_t; float a[TILE_SIZEB]; float b[TILE_SIZEB]; float c[TILE_SIZEB][TILE_SIZEB]; __shared__ float sa[BLOCK_SIZEB*KBLOCKB]; __shared__ float sb[KBLOCKB*BLOCK_SIZEB]; for(int i=0; i<TILE_SIZEB; i++){ for(int j=0; j<TILE_SIZEB; j++){ c[i][j]=0; } } for(int k_i=0; k_i<K; k_i += KBLOCKB){ for (int t=threadIdx.y*blockDim.x + threadIdx.x; t<BLOCK_SIZEB*KBLOCKB; t += blockDim.x*blockDim.y){ sa[t] = A[(m_b + t/KBLOCKB)*K + k_i + t%KBLOCKB]; sb[t] = B[(k_i + t/BLOCK_SIZEB)*N + n_b + t%BLOCK_SIZEB]; } __syncthreads(); for(int k=0; k<KBLOCKB; k++){ #pragma unroll for(int i=0; i<TILE_SIZEB; i++){ a[i] = sa[(m_t + i)*KBLOCKB + k]; b[i] = sb[k*BLOCK_SIZEB + n_t + i]; } for(int i=0; i<TILE_SIZEB; i++){ #pragma unroll for(int j=0; j<TILE_SIZEB; j++){ c[i][j] += a[i] * b[j]; } } } __syncthreads(); } for(int i=0; i<TILE_SIZEB; i++){ for(int j=0; j<TILE_SIZEB; j++){ if(m_i+i < M && n_i+j < N){ C[(m_i+i)*N + n_i + j] = c[i][j]; } } } } __global__ void matmul_4(float *A, float *B, float *C) { int m_b = blockIdx.y * BLOCK_SIZEB; int n_b = blockIdx.x * BLOCK_SIZEB; int m_i = m_b + threadIdx.y; int n_i = n_b + threadIdx.x; float a[TILE_SIZEB]; float b[TILE_SIZEB]; float c[TILE_SIZEB][TILE_SIZEB]; __shared__ float sa[BLOCK_SIZEB*KBLOCKB]; __shared__ float sb[KBLOCKB*BLOCK_SIZEB]; for(int i=0; i<TILE_SIZEB; i++){ for(int j=0; j<TILE_SIZEB; j++){ c[i][j]=0; } } for(int k_i=0; k_i<K; k_i += KBLOCKB){ for (int t=threadIdx.y*STRIDE + threadIdx.x; t<BLOCK_SIZEB*KBLOCKB; t += STRIDE*STRIDE){ sa[t] = A[(m_b + t/KBLOCKB)*K + k_i + t%KBLOCKB]; sb[t] = B[(k_i + t/BLOCK_SIZEB)*N + n_b + t%BLOCK_SIZEB]; } __syncthreads(); for(int k=0; k<KBLOCKB; k++){ #pragma unroll for(int i=0; i<TILE_SIZEB; i++){ a[i] = sa[(i*STRIDE + threadIdx.y)*KBLOCKB + k]; b[i] = sb[k*BLOCK_SIZEB + i*STRIDE + threadIdx.x]; } for(int i=0; i<TILE_SIZEB; i++){ #pragma unroll for(int j=0; j<TILE_SIZEB; j++){ c[i][j] += a[i] * b[j]; } } } __syncthreads(); } for(int i=0; i<TILE_SIZEB; i++){ for(int j=0; j<TILE_SIZEB; j++){ if(m_i+i*STRIDE < M && n_i+j*STRIDE < N){ C[(m_i+i*STRIDE)*N + n_i + j*STRIDE] = c[i][j]; } } } } __global__ void matmul_5(float *A, float *B, float *C) { int m_b = blockIdx.y * BLOCK_SIZEB; int n_b = blockIdx.x * BLOCK_SIZEB; int m_i = m_b + threadIdx.y; int n_i = n_b + threadIdx.x; float a[TILE_SIZEB]; float b[TILE_SIZEB]; float a1[TILE_SIZEB]; float b1[TILE_SIZEB]; float c[TILE_SIZEB][TILE_SIZEB]; __shared__ float sa[BLOCK_SIZEB*KBLOCKB]; __shared__ float sb[KBLOCKB*BLOCK_SIZEB]; for(int i=0; i<TILE_SIZEB; i++){ for(int j=0; j<TILE_SIZEB; j++){ c[i][j]=0; } } for(int k_i=0; k_i<K; k_i += KBLOCKB){ for (int t=threadIdx.y*STRIDE + threadIdx.x; t<BLOCK_SIZEB*KBLOCKB; t += STRIDE*STRIDE){ sa[t] = A[(m_b + t/KBLOCKB)*K + k_i + t%KBLOCKB]; sb[t] = B[(k_i + t/BLOCK_SIZEB)*N + n_b + t%BLOCK_SIZEB]; } __syncthreads(); for(int i=0; i<TILE_SIZEB; i++){ a[i] = sa[(i*STRIDE + threadIdx.y)*KBLOCKB]; b[i] = sb[i*STRIDE + threadIdx.x]; } int k = 1; while(k<KBLOCKB-1){ for(int i=0; i<TILE_SIZEB; i++){ a1[i] = sa[(i*STRIDE + threadIdx.y)*KBLOCKB + k]; b1[i] = sb[k*BLOCK_SIZEB + i*STRIDE + threadIdx.x]; #pragma unroll for(int j=0; j<TILE_SIZEB; j++){ c[i][j] += a[i] * b[j]; } } k++; for(int i=0; i<TILE_SIZEB; i++){ a[i] = sa[(i*STRIDE + threadIdx.y)*KBLOCKB + k]; b[i] = sb[k*BLOCK_SIZEB + i*STRIDE + threadIdx.x]; #pragma unroll for(int j=0; j<TILE_SIZEB; j++){ c[i][j] += a1[i] * b1[j]; } } k++; } for(int i=0; i<TILE_SIZEB; i++){ a1[i] = sa[(i*STRIDE + threadIdx.y)*KBLOCKB + k]; b1[i] = sb[k*BLOCK_SIZEB + i*STRIDE + threadIdx.x]; #pragma unroll for(int j=0; j<TILE_SIZEB; j++){ c[i][j] += a[i] * b[j]; } } for(int i=0; i<TILE_SIZEB; i++){ #pragma unroll for(int j=0; j<TILE_SIZEB; j++){ c[i][j] += a1[i] * b1[j]; } } __syncthreads(); } for(int i=0; i<TILE_SIZEB; i++){ for(int j=0; j<TILE_SIZEB; j++){ if(m_i+i*STRIDE < M && n_i+j*STRIDE < N){ C[(m_i+i*STRIDE)*N + n_i + j*STRIDE] = c[i][j]; } } } } int main(int argc, char *argv[]) { srand(0); float *cpu_a = (float*)malloc(sizeof(float)*M*K); float *cpu_b = (float*)malloc(sizeof(float)*K*N); float *cpu_at = (float*)malloc(sizeof(float)*K*M); float *cpu_c = (float*)malloc(sizeof(float)*M*N); float *cpu_c2 = (float*)malloc(sizeof(float)*M*N); for (int i=0; i<M; i++){ for (int k=0; k<K; k++){ cpu_a[i*K+k] = rand()/65536; cpu_at[k*M+i] = cpu_a[i*K+k]; } } for (int i=0; i<N*K; i++) cpu_b[i] = rand()/65536; //cudaEvent_t start, end; //cudaEventCreate(&start); //cudeEventCreate(&end); float *gpu_a, *gpu_at, *gpu_b, *gpu_c; cudaMalloc((void**)&gpu_a, sizeof(float)*M*K); cudaMalloc((void**)&gpu_b, sizeof(float)*N*K); cudaMalloc((void**)&gpu_at, sizeof(float)*M*K); cudaMalloc((void**)&gpu_c, sizeof(float)*M*N); cudaMemcpy(gpu_a, cpu_a, sizeof(float)*M*K, cudaMemcpyHostToDevice); cudaMemcpy(gpu_at, cpu_at, sizeof(float)*M*K, cudaMemcpyHostToDevice); cudaMemcpy(gpu_b, cpu_b, sizeof(float)*N*K, cudaMemcpyHostToDevice); dim3 grid0(M/16,N/16); dim3 block0(16, 16); matmul_s<<<grid0, block0>>>(gpu_a, gpu_b, gpu_c); cudaDeviceSynchronize(); printf("%s\n",cudaGetErrorString(cudaPeekAtLastError())); cudaMemcpy(cpu_c2, gpu_c, sizeof(float)*M*N, cudaMemcpyDeviceToHost); cudaDeviceSynchronize(); //cudaEventRecord(start, 0); dim3 grid(M/BLOCK_SIZE,N/BLOCK_SIZE); dim3 block(BLOCK_SIZE/TILE_SIZE, BLOCK_SIZE/TILE_SIZE); matmul_2<<<grid, block>>>(gpu_a, gpu_b, gpu_c); cudaDeviceSynchronize(); printf("%s\n",cudaGetErrorString(cudaPeekAtLastError())); cudaMemcpy(cpu_c2, gpu_c, sizeof(float)*M*N, cudaMemcpyDeviceToHost); cudaDeviceSynchronize(); matmul_1<<<grid, block>>>(gpu_a, gpu_b, gpu_c); cudaDeviceSynchronize(); printf("%s\n",cudaGetErrorString(cudaPeekAtLastError())); cudaMemcpy(cpu_c, gpu_c, sizeof(float)*M*N, cudaMemcpyDeviceToHost); cudaDeviceSynchronize(); dim3 gridb(M/BLOCK_SIZEB,N/BLOCK_SIZEB); dim3 blockb(BLOCK_SIZEB/TILE_SIZEB, BLOCK_SIZEB/TILE_SIZEB); matmul_4<<<gridb, blockb>>>(gpu_a, gpu_b, gpu_c); cudaDeviceSynchronize(); printf("%s\n",cudaGetErrorString(cudaPeekAtLastError())); cudaMemcpy(cpu_c, gpu_c, sizeof(float)*M*N, cudaMemcpyDeviceToHost); cudaDeviceSynchronize(); matmul_5<<<gridb, blockb>>>(gpu_a, gpu_b, gpu_c); cudaDeviceSynchronize(); printf("%s\n",cudaGetErrorString(cudaPeekAtLastError())); cudaMemcpy(cpu_c, gpu_c, sizeof(float)*M*N, cudaMemcpyDeviceToHost); cudaDeviceSynchronize(); float sum = 0.0; for (int i=0; i<M*N; i++) sum += cpu_c2[i]; printf("%f\n",sum); sum = 0.0; for (int i=0; i<M*N; i++) sum += cpu_c[i]; printf("%f\n",sum); sum = 0.0; for (int i=0; i<M*N; i++) sum += cpu_c[i] - cpu_c2[i]; printf("%f\n",sum); //cudaEventRecord(end,0); cudaFree(gpu_a); cudaFree(gpu_at); cudaFree(gpu_b); cudaFree(gpu_c); free(cpu_a); free(cpu_at); free(cpu_b); free(cpu_c); free(cpu_c2); }
7,786
#define BLOCK_SIZE_KERNEL 16 #define BLOCK_SIZE 16 /** * This macro checks return value of the CUDA runtime call and exits * the application if the call failed. */ #define CUDA_CHECK_RETURN(value) { \ cudaError_t _m_cudaStat = value; \ if (_m_cudaStat != cudaSuccess) { \ fprintf(stderr, "Error %s at line %d in file %s\n", \ cudaGetErrorString(_m_cudaStat), __LINE__, __FILE__); \ exit(1); \ } } __device__ int kerrors; __global__ void GoldChkKernel(double *gk, double *ck, int n) //, int *kerrors) { //================== HW Accelerated output validation int tx = blockIdx.x * BLOCK_SIZE + threadIdx.x; int ty = blockIdx.y * BLOCK_SIZE + threadIdx.y; if ((fabs((gk[ty * n + tx] - ck[ty * n + tx]) / gk[ty * n + tx]) > 0.0000000001) || (fabs((gk[ty * n + tx] - ck[ty * n + tx]) / ck[ty * n + tx]) > 0.0000000001)) atomicAdd(&kerrors, 1); } /* ======================================================= */ /* CUDA implementation of dGEMM using shared memory /* ======================================================= */ __global__ void cuda_dgemm_shmem(int n, double alpha, const double *B, const double *A, double beta, double *C) { // Block index int block_col = blockIdx.x; int block_row = blockIdx.y; // Thread index int thread_col = threadIdx.x; int thread_row = threadIdx.y; //printf("row = %d col = %d n= %d\n", block_col, block_row, n); //int row = blockDim.y * blockIdx.y + threadIdx.y; //int col = blockDim.x * blockIdx.x + threadIdx.x; int aBegin = n * blockDim.x * block_row; int aEnd = aBegin + n-1; int bBegin = blockDim.x * block_col; int bStep = n * blockDim.x; double Csub = 0; for (int a=aBegin, b=bBegin, istep=0; a <= aEnd; a+= blockDim.x, b+=bStep, ++istep){ __shared__ double As[BLOCK_SIZE_KERNEL][BLOCK_SIZE_KERNEL]; __shared__ double Bs[BLOCK_SIZE_KERNEL][BLOCK_SIZE_KERNEL]; if ((istep*blockDim.x+thread_col < n) && (block_row*blockDim.x+ thread_row < n)) As[thread_row][thread_col] = A[a + n * thread_row + thread_col]; else As[thread_row][thread_col] = 0; if ((block_col*blockDim.x+thread_col < n) && (istep*blockDim.x + thread_row < n)) Bs[thread_row][thread_col] = B[b + n * thread_row + thread_col]; else Bs[thread_row][thread_col] = 0; __syncthreads(); // calculate the cell for (int k = 0; k < blockDim.x; ++k) Csub += As[thread_row][k] * Bs[k][thread_col]; __syncthreads(); } // Write the block sub-matrix to global memory; // each thread writes one element int c = n * blockDim.x * block_row + blockDim.x * block_col; if ((block_col*blockDim.x+thread_col < n) && (block_row*blockDim.x+ thread_row < n)) C[c + n * thread_row + thread_col] = alpha * Csub + beta * C[c +n * thread_row + thread_col]; }
7,787
//Sum of two vectors in a parallel fashion #include <stdio.h> #include <iostream> #define N 512 __global__ void add(int *aa, int *bb, int *cc) { //int index = threadIdx.x + blockIdx.x* blockDim.x; // int index = threadIdx.x; int index = blockIdx.x; cc[index] = aa[index] + bb[index]; } int main(void) { int *a, *b, *c;// host copies of a, b, c int *d_a, *d_b, *d_c;// device copies of a, b, c int size = N * sizeof(int); a = (int *)malloc(size); b = (int *)malloc(size); c = (int *)malloc(size); for (int i = 0; i < N; i++) { a[i] = 1; b[i] = 2; c[i] = -1; } cudaMalloc((void**)&d_a, size); cudaMalloc((void**)&d_b, size); cudaMalloc((void**)&d_c, size); cudaMemcpy(d_a, a, size, cudaMemcpyHostToDevice); cudaMemcpy(d_b, b, size, cudaMemcpyHostToDevice); // Perform SAXPY on 1M elements add<<<N, 1>>>(d_a, d_b, d_c); cudaMemcpy(c, d_c, size, cudaMemcpyDeviceToHost); //free(a); //free(b); //free(c); cudaFree(d_a); cudaFree(d_b); cudaFree(d_c); //return 0; //std::cout << "Result c\n"; std::cout << "a = " << a[1] << "\n"; std::cout << "b = " << b[1] << "\n"; std::cout << "c = " << c[1] << "\n"; }
7,788
#include "rand.cuh" __device__ void init_fast_random(fast_random_t &fr, int uniq_id, int dest_id, rand_init_t init) { int init2 = init.a * 17; int init3 = init.b * 59; int init4 = init.c * 23; fr.x = uniq_id * init.d + dest_id * init3; fr.y = uniq_id * init2 + dest_id * init2; fr.z = uniq_id * init3 + dest_id * init3; fr.w = uniq_id * init4 + dest_id * init.d; } __device__ uint32_t rand_full(fast_random_t &fr) { uint32_t t = fr.x ^ (fr.x << 11); fr.x = fr.y; fr.y = fr.z; fr.z = fr.w; return fr.w = fr.w ^ (fr.w >> 19) ^ t ^ (t >> 8); } __device__ float rand_f(fast_random_t &fr) { uint32_t r = rand_full(fr); return 100.0f / ((r & 0xFFFF) + 1); } __device__ uint32_t rand256(fast_random_t &fr) { uint32_t t = fr.x ^ (fr.x << 11); fr.x = fr.y; fr.y = fr.z; fr.z = fr.w; return (fr.w = fr.w ^ (fr.w >> 19) ^ t ^ (t >> 8)) & 255; } void init_rand(rand_init_t &rand_init) { rand_init.a = rand(); rand_init.b = rand(); rand_init.c = rand(); rand_init.d = rand(); }
7,789
#include "includes.h" __global__ void Kernel10(int N, int M, int P, float *A, float *B, float *C) { __shared__ float sA[SIZE][SIZE]; __shared__ float sB[SIZE][SIZE]; int bx = blockIdx.x; int by = blockIdx.y; int tx = threadIdx.x; int ty = threadIdx.y; int row = by * SIZE + ty; int col = bx * SIZE + tx; float tmp = 0.0; for (int m=0; m < P; m=m+SIZE) { sA[ty][tx] = A[row*P + m + tx]; sB[ty][tx] = B[col + (m + ty)*M]; __syncthreads(); for (int k=0; k<SIZE; k++) tmp += sA[ty][k] * sB[k][tx]; __syncthreads(); } C[row*M+col] = tmp; }
7,790
__global__ void BNAvg(const float lr, float *w, const float *dzdw, const float ts, const float len) { int index = blockIdx.x * blockDim.x + threadIdx.x; if (index >= len) return; w[index] = (1-lr)*w[index]+lr*dzdw[index]/ts; }
7,791
//This file contains code for performing convolution on a GPU //Author:-Ajay Singh #include<stdio.h> #include<stdlib.h> #include<time.h> __global__ void convolution_1D_basic_kernel(float *N, float *M, float *P, int mask_width, int size){ int i=threadIdx.x+blockIdx.x*blockDim.x; int start_pos=(int)(i-(mask_width/2)); float pvalue; for(int j=start_pos; j<start_pos+mask_width; j++){ if(j>=0 && j<size) pvalue+=M[j-start_pos]*N[j]; } P[i]=pvalue; } #define size 10000 #define mask_width 100 int main(){ float mask[mask_width]; float arr[size]; float Output[size]; float *Pd, *arrd, *Md; double elapsed; for(int i=0; i<size; i++){ if(i<mask_width){ mask[i]=i/size; } arr[i]=(i*i)/size; } if(cudaMalloc((void **)&Pd, sizeof(float)*size)!=cudaSuccess){ printf("error while allocating Pd\n"); } if(cudaMalloc((void **)&Md, sizeof(float)*mask_width)!=cudaSuccess){ printf("error while allocating Md\n"); } if(cudaMalloc((void **)&arrd, sizeof(float)*size)!=cudaSuccess){ printf("error while allocating arrd\n"); } if(cudaMemcpy(arrd, arr, sizeof(float)*size, cudaMemcpyHostToDevice)!=cudaSuccess){ printf("error while copyting arr from host to device\n"); } if(cudaMemcpy(Md, mask, sizeof(float)*mask_width, cudaMemcpyHostToDevice)!=cudaSuccess){ printf("error while copyting mask from host to device\n"); } elapsed= -clock(); convolution_1D_basic_kernel<<<size/20, 20>>>(arrd, Md, Pd, mask_width, size); elapsed+=clock(); if(cudaMemcpy(Output, Pd, sizeof(float)*size, cudaMemcpyDeviceToHost)!=cudaSuccess){ printf("error while copyting arr from host to device\n"); } printf("Time Taken = %lf\n",elapsed); cudaFree(arrd); cudaFree(Pd); cudaFree(Md); return 0; }
7,792
#include <stdio.h> #include <stdlib.h> #include <cuda.h> #include <time.h> #include <curand_kernel.h> #include <iostream> #include <string> #include <fstream> using namespace std; #define CURAND_CALL(x) do { if((x)!=CURAND_STATUS_SUCCESS) { \ printf("Error at %s:%d\n",__FILE__,__LINE__);\ return EXIT_FAILURE;}} while(0) __global__ void setup_kernel(curandState *state,unsigned long long seed) { int id = threadIdx.x + blockIdx.x * blockDim.x; curand_init(/*seed*/seed,/*sequence*/id, /*offset*/0, &state[id]); } __global__ void generate_kernel(curandState *state, unsigned int *result) { int id = threadIdx.x + blockIdx.x * blockDim.x; /* Copy state to local memory for efficiency */ curandState localState = state[id]; /* Generate pseudo-random unsigned ints */ result[id] = curand(&localState); /* Copy state back to global memory */ state[id] = localState; } __global__ void generate_kernel2(unsigned int *result,unsigned long long seed) { int id = threadIdx.x + blockIdx.x * blockDim.x; curandState state; curand_init(/*seed*/seed,/*sequence*/id, /*offset*/0, &state); result[id] = curand(&state); } int main(int argc, char *argv[]) { int *h_data,*d_data,L,tam,print; size_t size; curandState *devStates; L=10; if(argc > 1) L = atoi(argv[1]); if(argc > 2) print = atoi(argv[2]); tam = L*L; size = tam*sizeof(int); // Allocate memory for the vectors on host memory. h_data = (int*) malloc(size); for (int i = 0; i < tam; i++) h_data[i] = 0; cudaMalloc((void **)&d_data, size); cudaMalloc((void **)&devStates, tam * sizeof(curandState)); //setup_kernel<<<L,L>>>(devStates,time(NULL)); //generate_kernel<<<L,L>>>(devStates, (unsigned int *) d_data); generate_kernel2<<<L,L>>>((unsigned int *) d_data,time(NULL)); cudaMemcpy(h_data, d_data, size, cudaMemcpyDeviceToHost); ofstream out("data.txt"); if(print) printf("\n\n"); for (int i = 0; i < tam; i++) { if(print) if(i%L==0) printf("\n"); out << h_data[i] << " "; if(print) printf(" %u",h_data[i]); } if(print) printf("\n\n"); out.close(); /* Free host memory */ cudaFree(devStates); cudaFree(d_data); free(h_data); return 0; }
7,793
#include <cmath> __global__ void myexp(double* value) { value[threadIdx.x] = std::exp(value[threadIdx.x]); }
7,794
#include "includes.h" /* * Naive sort * used if the quicksort uses too many levels */ __global__ void kernel_quicksort(int* values, int n) { #define MAX_LEVELS 1000 int pivot, L, R; int idx = threadIdx.x + blockIdx.x * blockDim.x; int start[MAX_LEVELS]; int end[MAX_LEVELS]; start[idx] = idx; end[idx] = n - 1; while (idx >= 0) { L = start[idx]; R = end[idx]; if (L < R) { pivot = values[L]; while (L < R) { while (values[R] >= pivot && L < R) R--; if(L < R) values[L++] = values[R]; while (values[L] < pivot && L < R) L++; if (L < R) values[R--] = values[L]; } values[L] = pivot; start[idx + 1] = L + 1; end[idx + 1] = end[idx]; end[idx++] = L; if (end[idx] - start[idx] > end[idx - 1] - start[idx - 1]) { int tmp = start[idx]; start[idx] = start[idx - 1]; start[idx - 1] = tmp; tmp = end[idx]; end[idx] = end[idx - 1]; end[idx - 1] = tmp; } } else idx--; } }
7,795
#include "includes.h" __global__ void PolynomialFunctionKernel_Double(float a3, float a2, float a1, float a0, double* input, double* output, int size) { int id = blockDim.x * blockIdx.y * gridDim.x + blockDim.x * blockIdx.x + threadIdx.x; if (id < size) { double x = input[id]; output[id] = a3 * x * x * x + a2 * x * x + a1 * x + a0; } }
7,796
#include "includes.h" __global__ void jacobi_copy_u( const int x_inner, const int y_inner, const double* src, double* dest) { const int gid = threadIdx.x+blockIdx.x*blockDim.x; if(gid < x_inner*y_inner) { dest[gid] = src[gid]; } }
7,797
#include "includes.h" __global__ void ComputeVelocitiesKernel (double *Vrad, double *Vtheta, double *Dens, double *Rmed, double *ThetaMomP, double *ThetaMomM, double *RadMomP, double *RadMomM, int nrad, int nsec, double OmegaFrame) { int j = threadIdx.x + blockDim.x*blockIdx.x; int i = threadIdx.y + blockDim.y*blockIdx.y; if (i<nrad && j<nsec){ if (i == 0) Vrad[i*nsec + j] = 0.0; else { Vrad[i*nsec + j] = (RadMomP[(i-1)*nsec + j] + RadMomM[i*nsec + j])/(Dens[i*nsec + j] + Dens[(i-1)*nsec + j] + 1e-20); } Vtheta[i*nsec + j] = (ThetaMomP[i*nsec + ((j-1)+nsec)%nsec] + ThetaMomM[i*nsec + j])/(Dens[i*nsec + j] + Dens[i*nsec + ((j-1)+nsec)%nsec] + 1e-15)/Rmed[i] - Rmed[i]*OmegaFrame; /* It was the angular momentum */ } }
7,798
#include <stdio.h> // Функция сложения двух векторов __global__ void addVector(float* left, float* right, float* result) { //Получаем id текущей нити. int idx = threadIdx.x; //Расчитываем результат. result[idx] = left[idx] + right[idx]; } #define SIZE 512 __host__ int main() { //Выделяем память под вектора float* vec1 = new float[SIZE]; float* vec2 = new float[SIZE]; float* vec3 = new float[SIZE]; //Инициализируем значения векторов for (int i = 0; i < SIZE; i++) { vec1[i] = i; vec2[i] = i; } //Указатели на память видеокарте float* devVec1; float* devVec2; float* devVec3; //Выделяем память для векторов на видеокарте cudaMalloc((void**)&devVec1, sizeof(float) * SIZE); cudaMalloc((void**)&devVec2, sizeof(float) * SIZE); cudaMalloc((void**)&devVec3, sizeof(float) * SIZE); //Копируем данные в память видеокарты cudaMemcpy(devVec1, vec1, sizeof(float) * SIZE, cudaMemcpyHostToDevice); cudaMemcpy(devVec2, vec2, sizeof(float) * SIZE, cudaMemcpyHostToDevice); dim3 gridSize = dim3(1, 1, 1); //Размер используемого грида dim3 blockSize = dim3(SIZE, 1, 1); //Размер используемого блока //Выполняем вызов функции ядра addVector<<<gridSize, blockSize>>>(devVec1, devVec2, devVec3); //Выполняем вызов функции ядра //addVector<<<blocks, threads>>>(devVec1, devVec2, devVec3); //Хендл event'а cudaEvent_t syncEvent; cudaEventCreate(&syncEvent); //Создаем event cudaEventRecord(syncEvent, 0); //Записываем event cudaEventSynchronize(syncEvent); //Синхронизируем event //Только теперь получаем результат расчета cudaMemcpy(vec3, devVec3, sizeof(float) * SIZE, cudaMemcpyDeviceToHost); //Результаты расчета for (int i = 0; i < SIZE; i++) { printf("Element #%i: %.1f\n", i , vec3[i]); } // // Высвобождаем ресурсы // cudaEventDestroy(syncEvent); cudaFree(devVec1); cudaFree(devVec2); cudaFree(devVec3); delete[] vec1; vec1 = 0; delete[] vec2; vec2 = 0; delete[] vec3; vec3 = 0; }
7,799
#include "includes.h" __global__ void sync_conv_groups() { }
7,800
#include <iostream> #define NUM_BANKS 32 #define LOG_NUM_BANKS 5 #define GET_OFFSET(idx) (idx >> LOG_NUM_BANKS) // Must be divisible by 32 and a divisor of 1024 const int block_size = 256; // Akhtyamov Pavel's realization of Scan // https://github.com/akhtyamovpavel/ParallelComputationExamples/blob/master/CUDA/05-scan/05-scan_bank_conflicts.cu __global__ void Scan(int *in_data, int *out_data) { // in_data -> [1 2 3 4 5 6 7 8], block_size 4 // block_idx -> [0 0 0 0 1 1 1 1 ] extern __shared__ int shared_data[]; // block_idx = 0 unsigned int tid = threadIdx.x; unsigned int index = blockIdx.x * blockDim.x + threadIdx.x; shared_data[tid + GET_OFFSET(tid)] = in_data[index]; // shared_data[tid + (tid >> LOG_NUM_BANKS)] = in_data[index]; // shared_data -> [1, 2, 3, 4] __syncthreads(); // shift = 2^(d - 1) for (unsigned int shift = 1; shift < blockDim.x; shift <<= 1) { int ai = shift * (2 * tid + 1) - 1; // tid = 0, shift = 1, ai = 0; // tid = 16, shift = 1, ai = 32 = 0 int bi = shift * (2 * tid + 2) - 1; if (bi < blockDim.x) { shared_data[bi + GET_OFFSET(bi)] += shared_data[ai + GET_OFFSET(ai)]; } __syncthreads(); } if (tid == 0) { shared_data[blockDim.x - 1 + GET_OFFSET(blockDim.x - 1)] = 0; } __syncthreads(); int temp; for (unsigned int shift = blockDim.x / 2; shift > 0; shift >>= 1) { int bi = shift * (2 * tid + 2) - 1; int ai = shift * (2 * tid + 1) - 1; int ai_offset = ai + GET_OFFSET(ai); int bi_offset = bi + GET_OFFSET(bi); if (bi < blockDim.x) { temp = shared_data[ai_offset]; // blue in temp // temp = 4 shared_data[ai_offset] = shared_data[bi_offset]; // orange // 1 2 1 0 1 2 1 0 // temp = 4 shared_data[bi_offset] = temp + shared_data[bi_offset]; } __syncthreads(); } // if (blockIdx.x == 16383) { // printf("%d %d %d %d\n", tid, tid + GET_OFFSET(tid), shared_data[tid + GET_OFFSET(tid)], index); // // std::cout << shared_data[tid] << std::endl; // } // block_idx = 0 -> [a0, a1, a2, a3] // block_idx = 1 -> [a4, a5, a6, a7] out_data[index] = shared_data[tid + GET_OFFSET(tid)]; __syncthreads(); // out_data[block_idx == 0] = [1, 3, 6, 10] // out_data[block_idx == 1] = [5, 11, 18, 26] } /* * Calculates grid size for array * size - array size */ int GridSize(int size) { // Calc grid size, the whole array must be covered with blocks int grid_size = size / block_size; if (size % block_size) { grid_size += 1; } return grid_size; } /* * Put prefix sums of each block in one array (d_blocks) * * d_array - source array * d_prefix_sum - calculated prefix sum on blocks * d_blocks - result */ __global__ void FindBlocks(int *d_array, int *d_prefix_sum, int *d_blocks) { int index = threadIdx.x + blockIdx.x * blockDim.x; int last_elem_in_block_index = (index + 1) * block_size - 1; d_blocks[index] = d_array[last_elem_in_block_index] + d_prefix_sum[last_elem_in_block_index]; __syncthreads(); } /* * Fills masks of elements greater than pivot and less than or equal to pivot * using GPU * * d_array - array of elements under consideration * d_greater_mask - mask of elements greater than pivot * d_less_or_equal_mask - mask of elements less (or equal) than pivot */ __global__ void FillMasks(int *d_array, int size, int pivot, int *d_greater_mask, int *d_less_or_equal_mask) { int index = threadIdx.x + blockIdx.x * blockDim.x; if (index < size) { if (d_array[index] > pivot) { d_greater_mask[index] = 1; d_less_or_equal_mask[index] = 0; } else { d_greater_mask[index] = 0; d_less_or_equal_mask[index] = 1; } } __syncthreads(); } /* * Add prefix sum of previous blocks to prefix sum, calculated on one block * * d_prefix_sum - prefix sum calculated on each block * d_block_prefix_sum - prefix sum calculated on previous blocks */ __global__ void AddBlocksPrefixSum(int *d_prefix_sum, int *d_block_prefix_sum) { int index = threadIdx.x + blockIdx.x * blockDim.x; d_prefix_sum[index] += d_block_prefix_sum[blockIdx.x]; __syncthreads(); } /* * Calculates scan (cumsum) for input array * using GPU * * d_array - input array * size - input array size * d_prefix_sum - result array */ void ParallelPrefixSum(int *d_array, int size, int *d_prefix_sum) { // Calc prefix sum on each block int grid_size = GridSize(size); Scan<<<grid_size, block_size, sizeof(int) * (block_size + GET_OFFSET(block_size))>>>(d_array, d_prefix_sum); // If size of array greater than size of one block, // we have to add to each block prefix sum of previous blocks if (grid_size > 1) { int block_grid_size = GridSize(grid_size); // Array of last elements of blocks int *d_block_array; cudaMalloc(&d_block_array, sizeof(int) * grid_size); // Array for prefix sum on blocks int *d_block_prefix_sum; cudaMalloc(&d_block_prefix_sum, sizeof(int) * grid_size); // Put prefix sum of each block in d_block_array FindBlocks<<<block_grid_size, block_size>>>(d_array, d_prefix_sum, d_block_array); // Find prefix sum on blocks ParallelPrefixSum(d_block_array, grid_size, d_block_prefix_sum); // Add prefix sum of all previous blocks to prefix sum AddBlocksPrefixSum<<<grid_size, block_size>>>(d_prefix_sum, d_block_prefix_sum); cudaFree(d_block_array); cudaFree(d_block_prefix_sum); } } /* * Divide elements into two arrays * d_array - source array * size - size of array * * d_less_or_equal_mask, d_greater_mask - masks of elements in d_array * d_less_or_equal_prefix_sum, d_greater_prefix_sum - prefix sums of masks * * d_less_or_equal_elems, d_greater_elems - result */ __global__ void ParallelDivide(int *d_array, int *d_less_or_equal_prefix_sum, int *d_greater_prefix_sum, int *d_less_or_equal_mask, int *d_greater_mask, int *d_less_or_equal_elems, int *d_greater_elems, int size) { int index = threadIdx.x + blockIdx.x * blockDim.x; if (index < size) { if (d_less_or_equal_mask[index]) { d_less_or_equal_elems[d_less_or_equal_prefix_sum[index]] = d_array[index]; } else { d_greater_elems[d_greater_prefix_sum[index]] = d_array[index]; } } __syncthreads(); } /* * Copy elements from d_source to d_array */ __global__ void ParallelCopy(int *d_array, int *d_source, int size) { int index = threadIdx.x + blockIdx.x * blockDim.x; if (index < size) { d_array[index] = d_source[index]; } __syncthreads(); } /* * Returns last value of array in global memory * d_array - array in global memory * size - d_array size */ int GetLast(int *d_array, int size) { int result = 0; cudaMemcpy(&result, d_array + size - 1, sizeof(int), cudaMemcpyDeviceToHost); return result; } /* * Place elements in correct order * Firstly, elements less (or equal) than pivot * Then, pivot, * Then, elements greater than pivot */ void ArrangeElements(int *d_array, int size, int pivot, int *d_less_or_equal_elems, int *d_greater_elems, int less_or_equal_size, int greater_size) { int grid_size = GridSize(size); ParallelCopy<<<grid_size, block_size>>>(d_array, d_less_or_equal_elems, less_or_equal_size); cudaMemcpy(d_array + less_or_equal_size, &pivot, sizeof(int), cudaMemcpyHostToDevice); ParallelCopy<<<grid_size, block_size>>>(d_array + less_or_equal_size + 1, d_greater_elems, greater_size); } /* * Fills arrays of elements less than pivot and greater than pivot */ void ProcessMasks(int *d_array, int size, int pivot, int **d_less_or_equal_elems, int **d_greater_elems, int &less_or_equal_size, int &greater_size) { // Storages of masks for elements less (or equal) than pivot // and greater than pivot int *d_less_or_equal_mask = nullptr; int *d_greater_mask = nullptr; // Storages of its prefix sums int *d_less_or_equal_prefix_sum = nullptr; int *d_greater_prefix_sum = nullptr; cudaMalloc(&d_less_or_equal_mask, sizeof(int) * size); cudaMalloc(&d_greater_mask, sizeof(int) * size); cudaMalloc(&d_less_or_equal_prefix_sum, sizeof(int) * size); cudaMalloc(&d_greater_prefix_sum, sizeof(int) * size); int grid_size = GridSize(size); FillMasks<<<grid_size, block_size>>>(d_array, size, pivot, d_greater_mask, d_less_or_equal_mask); ParallelPrefixSum(d_less_or_equal_mask, size, d_less_or_equal_prefix_sum); ParallelPrefixSum(d_greater_mask, size, d_greater_prefix_sum); // Get count of elements less or equal than pivot, held in last elem of prefix sum less_or_equal_size = GetLast(d_less_or_equal_prefix_sum, size); // Get count of elements greater than pivot, held in last elem of prefix sum greater_size = GetLast(d_greater_prefix_sum, size); cudaMalloc(d_less_or_equal_elems, sizeof(int) * less_or_equal_size); cudaMalloc(d_greater_elems, sizeof(pivot) * greater_size); // Divide elements into two arrays ParallelDivide<<<grid_size, block_size>>>(d_array, d_less_or_equal_prefix_sum, d_greater_prefix_sum, d_less_or_equal_mask, d_greater_mask, *d_less_or_equal_elems, *d_greater_elems, size - 1); cudaFree(d_less_or_equal_mask); cudaFree(d_greater_mask); cudaFree(d_less_or_equal_prefix_sum); cudaFree(d_greater_prefix_sum); } /* * Sorts d_array using GPU * d_array - array in global memory * size - d_array size */ void QuickSort(int *d_array, int size) { // Already sorted if (size <= 1) return; /// Partition // Take last element of array as a pivot int pivot = GetLast(d_array, size); // Storages of elements less (or equal) than pivot // and greater than pivot int *d_less_or_equal_elems = nullptr; int *d_greater_elems = nullptr; // Its sizes int less_or_equal_size = 0; int greater_size = 0; ProcessMasks(d_array, size, pivot, &d_less_or_equal_elems, &d_greater_elems, less_or_equal_size, greater_size); ArrangeElements(d_array, size, pivot, d_less_or_equal_elems, d_greater_elems, less_or_equal_size, greater_size); cudaFree(d_less_or_equal_elems); cudaFree(d_greater_elems); // Recursively process parts QuickSort(d_array, less_or_equal_size); QuickSort(d_array + less_or_equal_size + 1, greater_size); } void Shuffle(int *array, int size) { for (int i = size - 1; i > 0; --i) { std::swap(array[i], array[std::rand() % (i + 1)]); } } int main() { size_t size = 0; std::cout << "Insert the size of array" << std::endl; std::cin >> size; int *h_array_to_be_sorted = new int[size]; for (int i = 0; i < size; ++i) { h_array_to_be_sorted[i] = i; } // Create shuffled example to be sorted later Shuffle(h_array_to_be_sorted, size); // Create array to be processed to GPU int *d_array_to_be_sorted; cudaMalloc(&d_array_to_be_sorted, size * sizeof(float)); cudaMemcpy(d_array_to_be_sorted, h_array_to_be_sorted, sizeof(int) * size, cudaMemcpyHostToDevice); cudaEvent_t start; cudaEvent_t stop; cudaEventCreate(&start); cudaEventCreate(&stop); cudaEventRecord(start, 0 /* Stream ID */); QuickSort(d_array_to_be_sorted, size); cudaEventRecord(stop, 0); cudaMemcpy(h_array_to_be_sorted, d_array_to_be_sorted, sizeof(int) * size, cudaMemcpyDeviceToHost); cudaEventSynchronize(stop); float elapsedTime; cudaEventElapsedTime(&elapsedTime, start, stop); std::cout << "Elapsed time is :" << elapsedTime << std::endl; cudaEventDestroy(start); cudaEventDestroy(stop); std::cout << "Sorted array:" << std::endl; for (int i = 0; i < size; ++i) { std::cout << h_array_to_be_sorted[i] << " "; } std::cout << "\n"; cudaFree(d_array_to_be_sorted); delete[] h_array_to_be_sorted; return 0; }