__global__ 
void kernel_sum(float *x, float *y, float *z, int size) {
	
	/* global thread index (depends on grid organization) */ 
	int gtid =  ...;
	
	/* check if we are in the arrays x, y, z */
	if (gtid < size) {
		z[gtid] = x[gtid] + y[gtid]; // paralell part
	}
}