#include <stdio.h>

__global__ void calculate(float *a, float *b, float *c, float *d) {
  // Kernel declaration
  printf("test test\n");
  float temp = *a + *b; // Pointer variable operation
  *d = temp / *c;
}

int main(void) {
  // Host variables and input
  float a = 3.0, b = 7.0, c = 2.0;
  float d;
  // Device variables
  float *a_dev, *b_dev, *c_dev, *d_dev;
  int float_size = sizeof(float);

  // Allocate memory device variables
  cudaMalloc((void **)&a_dev, float_size);
  cudaMalloc((void **)&b_dev, float_size);
  cudaMalloc((void **)&c_dev, float_size);
  cudaMalloc((void **)&d_dev, float_size);

  // Copy input from host to device
  cudaMemcpy(a_dev, &a, float_size, cudaMemcpyHostToDevice);
  cudaMemcpy(b_dev, &b, float_size, cudaMemcpyHostToDevice);
  cudaMemcpy(c_dev, &c, float_size, cudaMemcpyHostToDevice);

  // Launch kernel
  calculate<<<1, 1>>>(a_dev, b_dev, c_dev, d_dev);

  // Copy result, and print
  cudaMemcpy(&d, d_dev, float_size, cudaMemcpyDeviceToHost);
  printf("d is: %f\n", d);

  // Clean up device memory
  cudaFree(a_dev);
  cudaFree(b_dev);
  cudaFree(c_dev);
  cudaFree(d_dev);
  return 0;
}