c - 由于循环，RAM 正在爆炸

Question

所以我这里有这个循环:

/* loop */
omp_set_dynamic(0);
#pragma omp parallel num_threads(1)
#pragma omp for ordered private(iterationPoints, normalizedPoints, grid)
for(iterations = maxIter; iterations >= minIter; iterations -= stepSizeIter){

        /* calculate iterationPoints */
        iterationPoints = (double **) malloc(iterations *sizeof(double *));
        int i;
        for(i = 0; i < iterations; i++){
            iterationPoints[i] = (double *) malloc(DIMS *sizeof(double));
        }
        calcuTraj(iterationPoints,a,b,c,x,y,z,h,iterations, bounds, (iterations == maxIter));

        /* normalize Data */
        normalizedPoints = (double **) malloc(iterations *sizeof(double *));
        for(i = 0; i < iterations; i++){
            normalizedPoints[i] = (double *) malloc(DIMS * sizeof(double));
        }

        #pragma omp ordered
        printf("%d %f  %f  %f  %f  %f  %f\n",iterations,bounds[0][0] ,bounds[0][1], bounds[1][0], bounds[1][1], bounds[2][0], bounds[2][1]);

        normalize(iterationPoints, normalizedPoints, bounds, iterations); 

        /* creating 3D Array for the grid of boxes in space*/

        /* setting minimum for sidelength of the grid */
        double minGridSideLength = 1; 
        /* calculating array size */
        int boxesPerDim = ceil(minGridSideLength/epsion) +1;

        //printf("boxesPerDim: %d \n", boxesPerDim);

        /* create grid array */
        grid = (bool ***) malloc(boxesPerDim *sizeof(bool **));

        int j_X, j_Y; 
        for(j_X = 0; j_X < boxesPerDim; j_X++){

            grid[j_X] = (bool **) malloc(boxesPerDim *sizeof(bool *));

            for(j_Y = 0; j_Y < boxesPerDim; j_Y++){
               grid[j_X][j_Y] = (bool *) calloc(boxesPerDim,sizeof(bool *));
            }
        }   

        /* count hitted boxes */
        int boxesHit = boxCount(normalizedPoints, grid, iterations, epsion, boxesPerDim);

        #pragma omp ordered
        fprintf(file,"%d  %d\n", iterations, boxesHit);
        //printf("%d \n",boxesHit);


        /* free storage */
        free(iterationPoints);
        free(grid);
        free(normalizedPoints);

}

我将线程数设置为 1，因此不存在会导致高迭代时内存爆炸的多线程。因此，在此配置中，循环将一步一步运行。函数“calculatedTraj(..)”计算“迭代点”。如果我从 minIter = 20*10^6 到 maxIter = 50*10^6 迭代，步长 = 10*10^6，我的内存就会爆炸，进程就会被杀死。但是如果我运行 minIter = maxIter = 50*10^6 的程序，它就可以工作并且内存也很好。 那么为什么不能用 for 循环来做到这一点，因为在每个循环部分结束时我都会释放分配的内存，那么为什么 ram 会爆炸呢？

Answer 1

grid[j_X] = (bool **) malloc(boxesPerDim *sizeof(bool *));
grid[j_X][j_Y] = (bool *) calloc(boxesPerDim,sizeof(bool *));

您还没有释放这些。

类似于拥有一些结构:

struct whatever {
int something
char *stringly
};

struct whatever *whtvr = malloc(sizeof(struct whatever));
whtvr->stringly = malloc(strlen("whatever") + 1);

...

在放弃结构体的内存之前，您必须释放字符串的内存。

free(whtvr->stringly);
free(whtvr);