C中二维数组的分割错误

Question

I'm making what we call an integral image, but I have a problem. When I reach a certain number I have an Segmentation fault. The image is 273*273 it works, 274*274 not working. So i tried to Malloc my double array. Same. I've read that arrays in a function are in the Stack, and the stack has small space. How can I make my array declared in the memory?

I am using C99 and SDL:

void integral_img(SDL_Surface * img)
{
    int **M;

    M = malloc(img->w * sizeof(int *));
    for (int i = 0; i < img->w; i++)
    {
        M[i] = malloc(img->h * sizeof(int));
    }
    int sum_int = 0;
    Uint8 sum = 0;

    for (int i = 0; i < img->h; i++)
    {
        for (int j = 0; j < img->w; j++)
        {
            SDL_GetRGB(getpixel(img, i, j), img->format, &sum, &sum, &sum);

            if (i == 0 && j == 0)
                sum_int = (int)sum;
            if (i == 0 && j > 0)
                sum_int = (int)sum + M[i][j - 1];
            if (j == 0 && i > 0)
                sum_int = (int)sum + M[i - 1][j];
            if (i > 0 && j > 0)
                sum_int = (int)sum + M[i - 1][j] + M[i][j - 1] - M[i - 1][j - 1];
            M[i][j] = sum_int;
        }
    }
}

Answer 1

• You are allocating a pointer-based lookup table, not a 2D array. It will work, but it is wide-spread incorrect practice: it makes the program needlessly slow for no gain. Consider using a 2D array instead:

   int (*M)[w] = malloc( sizeof(int[h][w]) ); ... free(M); 

• One bug is here:

   M = malloc(img->w * sizeof(int *)); for (int i = 0; i < img->w; i++) { M[i] = malloc(img->h * sizeof(int)); } 

  Here you declare the inner-most dimension of your lookup table to be w , width.

  But later on

   for (int i = 0; i < img->h; i++) { for (int j = 0; j < img->w; j++) ... M[i][j] = sum_int; 

  you say that the inner-most dimension is h , height.

• Furthermore, your program leaks a lot of memory each time you call the function, since you don't free memory anywhere. In fact, you don't seem to even use the lookup table after filling it up.

Answer 2

A simplified version of your code, modified to illustrate 2D array initialization and cleanup (you must free what you malloc):

In short, your array initialization has to follow the way you create memory for row , column :

typedef struct  {
    int h;
    int w;
}SDL_Surface;
//if you change your prototype to pass the initializer then you can
//free it in the calling function.  Otherwise, as written in your OP
//you will have a memory leak.
void integral_img(SDL_Surface * img, int **m);

int main(int argc, char* argv[])
{
    SDL_Surface img;
    int **M; 

    img.h = 100;
    img.w = 50;
    //create memory for array in calling function
    M = malloc(img.w * sizeof(int *));
    if(!M){ /*handle error and leave*/ }//test results of malloc
    for (int i = 0; i < img.w; i++)//... If you create space for w rows and h columns...
    {
        M[i] = malloc(img.h * sizeof(int));// note modified prototype
        if(!M[i]) {/*handle error and leave*/} //test results of malloc
    }

    integral_img(&img, M);

    //use img...

    //free M :(no memory leaks)

    for (int i = 0; i < img.w; i++)
    {
        if(M[i]) free(M[i]);
    }
    if(M) free(M);


    return 0;
}

void integral_img(SDL_Surface * img, int **m)
{


    for (int i = 0; i < img->w; i++)//... Then you must access w rows and h columns
    {                               // (w & h were reversed in your code)
        for (int j = 0; j < img->h; j++)
        {
            m[i][j] = (i+1)*(j+1);
        }
    }
}

Answer 3

i don't think that your code is bogus, but i am suspecting this line:

SDL_GetRGB(getpixel(img, i, j), img->format, &sum, &sum, &sum);

any way, why you have to allocate a 2D array such way? why not just one continues block?

take a look to my code:

void integral_img(SDL_Surface * img)
{
    #define _IDX(x,y) ((x)+((y)*cx))
   int cx=img->w,
        cy=img->h;
    int *M;

    //allocating only one continues block of memory
    M = malloc(cx*cy * sizeof(int ));
    if(!M){
        perror("not enough memory");
        exit(1);
    }

    int sum_int = 0;
    Uint8 t sum = 0;

    for (int i = 0; i < cy; i++)
    {
        for (int j = 0; j < cx; j++)
        {
            SDL_GetRGB(getpixel(img, i, j), img->format, &sum, &sum, &sum);

            if (i == 0 && j == 0)
                sum_int = (int)sum;
            if (i == 0 && j > 0)
                sum_int = (int)sum + M[_IDX(i,j - 1)];
            if (j == 0 && i > 0)
                sum_int = (int)sum + M[_IDX(i- 1,j )];
            if (i > 0 && j > 0)
                sum_int = (int)sum + M[_IDX(i- 1,j )] + M[_IDX(i,j - 1)] - M[_IDX(i- 1,j - 1)];
            M[_IDX(i,j )] = sum_int;
        }
    }

    // easy to free
    free(M); 
    return;
}

it is cleaner and uses less iterations on allocating and freeing memory.