dd

test.cu

#include<cuda.h>

#define THREAD_NUM 256
#define BLOCK_NUM 32
#define R_SIZE THREAD_NUM * BLOCK_NUM
#define M_SIZE R_SIZE * R_SIZE


__global__ void mat_mul(int* mat1, int* mat2, int* result){

    int bid = blockIdx.x;
    int tid = threadIdx.x;
    
    int index = bid * THRREAD_NUM  + tid;
    
    for(int c=0; i< R_SIZE; i++) {
        for(int n= 0; n < R_SIZE;n++) {
            result[index * R_SIZE + c] +=mat1[index*R_SIZE+n]* mat2[n*R_SIZE+c];
        }
       
    {

}

int main(){
    int* mat1;
    int* mat2;
    int* result;
    
    int* g_mat1;
    int* g_mat2;
    int* g_mat_result;
    
    
    mat1 = (int*) malloc(M_SIZE*sizeof(int);
    // same to mat2 result
    
    cudaMalloc((void**)&g_mat1, sizeof(int)* M_SIZE);
    // some to g_mat2  g_mat_result;
    
    
    cudaMemcpy(g_mat1, mat1, sizeof(int)* M_SIZE, cudaMemcpyHosttoDevice);
    // some from mat2 to g_mat2;
    
    mat_mul<<BLOCK_NUM, THREAD_NUM>>>(g_mat1, g_mat2, g_mat_result);
    
    
    cudaMemcpy(result, g_mat_result, sizeof(int) * M_SIZE, cudaMemcpyDevicetohost);
    
    
    

}