Segmentation fault occurs when generating large amounts of doubles with rand() in C

Question

According to gdb this line causes a segmentation fault

((double) rand()/(double) (RAND_MAX)) * (r*2)

but this is completely fine

((float) rand()/(float) (RAND_MAX)) * (r*2)

I can make do with floats, but this bugs me. Am I doing something wrong or is rand() incapable of dealing with doubles.

Notice that in the context of upper example all occurrences of float were changed to double.

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

struct cord {
    /* float or double */
    float x;
    float y;
};

int main() {
    float r = 3.0;
    int amount = 1000000;
    struct cord cords[amount];

    srand(time(NULL));
    for (int i = 0; i < amount; i++) {

        cords[i] = (struct cord) {
            /* Segfault occurs when (float) is replaced with (double) */
            ((float) rand()/(float) (RAND_MAX)) * (r*2),
            ((float) rand()/(float) (RAND_MAX)) * (r*2)
        };

    }
}

Answer 1

At the time the segfault occurs, there is no more stack space to allocate the new double value. The location of the next double would be at an illegal (out-of-bounds) memory address, hence the segfault.

The reason why your code does not segfault for float is because floats are single-precision and usually requires half the number of bytes (usually 4B) to store compared to a double (usually 8B). Your code still have sufficient space to store all the floats.