CUDA Programming - Running same code as other person, getting wrong output

Accelerated Computing CUDA CUDA Programming and Performance
1

I’m following a course for CUDA Programming and I wanted to try the square program myself. I re-wrote the code from the course but I’m not getting the same output. Here is my code

#include <stdio.h>

__global__ void square (int *d_out, int *d_in){
	int idx = threadIdx.x;
	int f = d_in[idx];
	d_out[idx] = f * f;
}

int main(int argc, char ** argv){
	const int ARRAY_SIZE = 64;
	const int ARRAY_BYTES = ARRAY_SIZE * sizeof(float);
	
	int h_in[ARRAY_SIZE];
	for(int i = 0; i < ARRAY_SIZE; i ++){
		h_in[i] = i;
	//	printf("%d ", h_in[i]);
	}
	int h_out[ARRAY_SIZE];
	
	int *d_in;
	int *d_out;
	
	cudaMalloc(&d_in, ARRAY_BYTES);
	cudaMalloc(&d_out, ARRAY_BYTES);
	
	cudaMemcpy(d_in, h_in, ARRAY_BYTES, cudaMemcpyHostToDevice);
	
	square<<<1, ARRAY_SIZE>>>(d_out, d_in);
	
	cudaMemcpy(h_out, d_out, ARRAY_BYTES, cudaMemcpyDeviceToHost);

	printf("\n\n");	
	
	for(int i = 0; i < ARRAY_SIZE; i ++){
		printf("%d", h_out[i]);
		printf(((i % 4) != 3) ? "\t" : "\n");
	}
	
	cudaFree(d_in);
	cudaFree(d_out);
	
	return 0;
}

The course shows a square of floats but I changed it and do it with integers. Here’s my output and every time I compile it it’s difference. Considering the fact of different outputs, I think that the mistake is in the h_out array. Here’s what I got as an output my last time:

-1042321192	22092	627381103	32615
0	0	0	0
0	0	-808220160	-700231102
-1042321200	22092	635949264	32615
-339029120	32764	635950901	32615
644673632	32615	0	0
-1064290840	22092	635949431	32615
-1064280056	22092	-1066442772	22092
644678000	32615	-1066442772	22092
-1066488425	22092	630040104	32615
-1066489136	22092	626185353	32615
21	0	4	0
-1066743629	22092	-1064290840	22092
1	0	-1066464195	22092
642480544	32615	0	0
-1066464272	22092	-1066819952	22092
2

there is nothing wrong with your code.
The problem may be in your machine setup.

Add proper cuda error checking (google for that, and apply it to your code) and also run your code with cuda-memcheck, to get an indication of what may be the problem