extern "C" void foo(double *a, double *b, double *c, int n);

__global__ void cudafoo(double *d_a, double *d_b, double *d_c, int n)
{
    int i = blockDim.x*blockIdx.x + threadIdx.x;

    if (i<n) {
        d_c[i] = d_a[i] + d_b[i];
    }
}

void foo(double *a, double *b, double *c, int n)
{
    double *d_a;
    double *d_b;
    double *d_c;
    cudaMalloc((void**)&d_a, n*sizeof(double)); 
    cudaMalloc((void**)&d_b, n*sizeof(double)); 
    cudaMalloc((void**)&d_c, n*sizeof(double)); 

    cudaMemcpy(d_a, a, n*sizeof(double), cudaMemcpyHostToDevice);
    cudaMemcpy(d_b, b, n*sizeof(double), cudaMemcpyHostToDevice);

    int threadsPerBlock = 256;
    int blocksPerGrid = 4;

    cudafoo<<<blocksPerGrid,threadsPerBlock>>>(d_a, d_b, d_c, n);

    cudaMemcpy(c, d_c, n*sizeof(double), cudaMemcpyDeviceToHost);
    cudaFree(d_a);
    cudaFree(d_b);
    cudaFree(d_c);
}