cuda kernel functions doesn't change value

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi there,

I’m new with cuda and for the first steps, I try to simply add two values and display the solution.
Now, I had write my code and compile it with

nvcc main.cu

, there are no errors, but the solution is wrong (I get a random number for c).

After that, I’m trying to search a solution and I have found many postings, that I have to add the arch flag to nvcc. Than I try to compile the files with

nvcc -gencode arch=compute_35,code=sm_35

, but the same issue happens. I don’t know the problem. It’s a Tesla K40m graphics card.

Can anyone help me? Here is the code:

#include <stdio.h>


__global__ void addKernel(int a, int b, int *c){
	
	*c = a +b;
}


int main(void) {
	int count = 0;
	cudaGetDeviceCount( &count );
	if(count == 0){
		printf("no cuda supported device found. EXIT.\n");
		return -1;
	}

	printf("number of nvidia devices = %d\n", count);

	cudaDeviceProp prop;
	cudaGetDeviceProperties(&prop, 0);
	printf(" ----- device information ----- \n");
	printf(" Name:   %s\n", prop.name );
	printf(" Max threads per block:   %d\n", prop.maxThreadsPerBlock);
	printf(" Max thread dimensions:   (%d, %d, %d)\n",
			prop.maxThreadsDim[0],
			prop.maxThreadsDim[1],
			prop.maxThreadsDim[2]);

	int c=19;
	printf(" c = %d\n", c);

	int *d_c;
	cudaMalloc( (void**) &d_c, sizeof(int) );

	addKernel<<<1,1>>>(7,17,d_c);

	cudaError_t err = cudaMemcpy(&c, d_c, sizeof(int), cudaMemcpyDeviceToHost); 
	if(err == cudaSuccess) printf("Cuda Success");

	printf(" c = %d\n", &c);
	return 0;

}

thank you in advance!

2

This is wrong:

printf(" c = %d\n", &c);

it should be:

printf(" c = %d\n", c);

identical to the previous statement like it in your code.

This programming error has nothing to do with CUDA.