c++ - 无法读取文件并放置在 CUDA 中的二维相对矩阵地址中

Question

我正在使用 malloc 分配一个二维矩阵并尝试在相对地址中插入值。我不明白为什么它是核心转储错误。请在下面查看我的代码。

#include <stdio.h>
#include <stdlib.h>

int main()
{
    int width = 4;
    FILE *fp = fopen("matB.txt", "r");
    int *x;
    x = (int*)malloc(width*width*sizeof(int));
    int i, j;
    for(i=0; i<width; i++)
    {
        for(j=0; j<width; j++)
        {
            fscanf(fp, "%d", x[i*width+j]);
        }
    }

    for(i=0; i<width; i++)
    {
        for(j=0; j<width; j++)
        {
            printf("%d", x[i*width+j]);
        }
    }
    return 0;
}

matB.txt
1 2 3 4
1 2 3 4
1 2 3 4
1 2 3 4

我已经制作了上面的示例程序来检查相对地址，并将 &x[] 放在 fscanf 中清除了这个问题。

上面的示例C代码是因为在Cuda中有同样的读取问题而完成的。当使用相同的二维数组分配方式及其相对地址时，它正在读取文件并尝试打印相同的文件时..它打印 0 而不是 1,2,3,4..我正处于 CUDA 的学习阶段.我看到主机数组和放置在其相对地址中没有分配问题，但为什么读取的文件打印 0？？

Cuda程序如下

//Matrix multiplication using shared and non shared kernal
#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#define TILE_WIDTH 2

/*matrix multiplication kernels*/

//non shared
__global__ void MatrixMul( float *Md , float *Nd , float *Pd , const int WIDTH )
{
           // calculate thread id
        unsigned int col = TILE_WIDTH*blockIdx.x + threadIdx.x ;
        unsigned int row = TILE_WIDTH*blockIdx.y + threadIdx.y ;
        for (int k = 0 ; k<WIDTH ; k++ )
        {
            Pd[row*WIDTH + col]+= Md[row * WIDTH + k ] * Nd[ k * WIDTH + col] ;
        }
}

// shared
__global__ void MatrixMulSh( float *Md , float *Nd , float *Pd , const int WIDTH )
{
        //Taking shared array to break the MAtrix in Tile widht and fatch them in that array per ele
          __shared__ float Mds [TILE_WIDTH][TILE_WIDTH] ;
           __shared__ float Nds [TILE_WIDTH][TILE_WIDTH] ;

         // calculate thread id
          unsigned int col = TILE_WIDTH*blockIdx.x + threadIdx.x ;
          unsigned int row = TILE_WIDTH*blockIdx.y + threadIdx.y ;
        for (int m = 0 ; m<WIDTH/TILE_WIDTH ; m++ ) // m indicate number of phase
       {
            Mds[threadIdx.y][threadIdx.x] =  Md[row*WIDTH + (m*TILE_WIDTH + threadIdx.x)]  ;
            Nds[threadIdx.y][threadIdx.x] =  Nd[ ( m*TILE_WIDTH + threadIdx.y) * WIDTH + col] ;
         __syncthreads() ; // for syncronizeing the threads
         // Do for tile
           for ( int k = 0; k<TILE_WIDTH ; k++ )
                       Pd[row*WIDTH + col]+= Mds[threadIdx.x][k] * Nds[k][threadIdx.y] ;
         __syncthreads() ; // for syncronizeing the threads
     }
}

// main routine
int main (int argc, char* argv[])
{

    const int WIDTH = 4 ;
    printf("%d\n", WIDTH);
    //float array1_h[WIDTH][WIDTH] ,array2_h[WIDTH][WIDTH], M_result_array_h[WIDTH][WIDTH]  ;
    float *array1_h, *array2_h, *M_result_array_h;
    float *array1_d , *array2_d ,*result_array_d  ,*M_result_array_d; // device array
    int i , j ;
    cudaEvent_t start_full,stop_full;
    float time;
    cudaEventCreate(&start_full);
    cudaEventCreate(&stop_full);
    cudaEventRecord(start_full, 0);

    //char *file1 = argv[2];
    //char *file2 = argv[3];
    //char *file3 = argv[4];

    FILE *fp1 = fopen("matA.txt", "r");
    FILE *fp2 = fopen("matB.txt", "r");
    FILE *fp3 = fopen("matC.txt", "w");

    //create device array cudaMalloc ( (void **)&array_name, sizeofmatrixinbytes) ;
    cudaMallocHost((void **) &array1_h , WIDTH*WIDTH*sizeof (float) ) ;
    cudaMallocHost((void **) &array2_h , WIDTH*WIDTH*sizeof (float) ) ;
    cudaMallocHost((void **) &M_result_array_h , WIDTH*WIDTH*sizeof (float) ) ;
    //input in host array
    for ( i = 0 ; i<WIDTH ; i++ )
    {
        for (j = 0 ; j<WIDTH ; j++ )
        {
            fscanf(fp1, "%d", &array1_h[i*WIDTH + j]);
            printf("%d\t", array1_h[i*WIDTH + j]);
        }
      //  fscanf(fp1, "\n");
    }
    /*
    for ( i = 0 ; i<WIDTH ; i++ )
    {
        for (j = 0 ; j<WIDTH ; j++ )
        {
            printf("%d\t", array1_h[i*WIDTH+j]);
        }
        printf("\n");
    }*/
    for ( i = 0 ; i<WIDTH ; i++ )
    {
        for (j = 0 ; j<WIDTH ; j++ )
        {
            fscanf(fp2, "%d", &array2_h[i*WIDTH+j]);
        }
        fscanf(fp2, "\n");
    }

    //create device array cudaMalloc ( (void **)&array_name, sizeofmatrixinbytes) ;
    cudaMalloc((void **) &array1_d , WIDTH*WIDTH*sizeof (float) ) ;
    cudaMalloc((void **) &array2_d , WIDTH*WIDTH*sizeof (float) ) ;

    //copy host array to device array; cudaMemcpy ( dest , source , WIDTH , direction )
    cudaMemcpy ( array1_d , array1_h , WIDTH*WIDTH*sizeof (float) , cudaMemcpyHostToDevice ) ;
    cudaMemcpy ( array2_d , array2_h , WIDTH*WIDTH*sizeof (float) , cudaMemcpyHostToDevice ) ;

    //allocating memory for resultant device array
    cudaMalloc((void **) &result_array_d , WIDTH*WIDTH*sizeof (float) ) ;
    cudaMalloc((void **) &M_result_array_d , WIDTH*WIDTH*sizeof (float) ) ;

    //calling kernal
    dim3 dimGrid ( WIDTH/TILE_WIDTH , WIDTH/TILE_WIDTH ,1 ) ;
    dim3 dimBlock( TILE_WIDTH, TILE_WIDTH, 1 ) ;

    // Change if 0 to if 1 for running non shared code and make if 0 for shared memory code
    #if 0
            MatrixMul <<<dimGrid,dimBlock>>> ( array1_d , array2_d ,M_result_array_d , WIDTH) ;
    #endif

    #if 1
            MatrixMulSh<<<dimGrid,dimBlock>>> ( array1_d , array2_d ,M_result_array_d , WIDTH) ;
    #endif

    // all GPU function blocked till kernel is working
    //copy back result_array_d to result_array_h
    cudaMemcpy(M_result_array_h , M_result_array_d , WIDTH*WIDTH*sizeof(int), cudaMemcpyDeviceToHost) ;

    //printf the result array
    for ( i = 0 ; i<WIDTH ; i++ )
    {
        for ( j = 0 ; j < WIDTH ; j++ )
        {
            fprintf (fp3, "%d\t", M_result_array_h[i*WIDTH+j]) ;
        }
        fprintf (fp3, "\n") ;
    }
    //system("pause") ;
    cudaFree(array1_d);
    cudaFree(array2_d);
    cudaFree(M_result_array_d);

    cudaFreeHost(array1_h);
    cudaFreeHost(array2_h);
    cudaFreeHost(M_result_array_h);

    cudaEventRecord(stop_full, 0);
    cudaEventSynchronize(stop_full);

    cudaEventElapsedTime(&time, start_full, stop_full);
    printf ("Total execution Time is : %1.5f ms\n", time);

}

Answer 1

应该是 fscanf(fp, "%d", &x[i*width+j]);。 scanf 系列需要写入扫描值的位置地址。

此外，don't cast malloc .