#include <stdio.h>

__global__ 
void device_hello(){

  //uncomment this line to print only one time (unless you have multiple blocks)
  //if(threadIdx.x==0)
  printf("Hello world! from the device! thread:%d,%d\n",blockIdx.x,threadIdx.x);
  return;
}

int main(void){

  // rather than calling fflush    
  setbuf(stdout, NULL);

  // greet from the host
  printf("Hello world! from the host!\n");

  // launch a kernel with a block of threads to greet from the device
  dim3 blockSize(32,1,1);
  dim3 gridSize(1,1,1);
  // run several variations by playing with the block and grid sizes
  // above -- if you change the y or z dimensions change the print 
  // statement to reflect that.
  device_hello<<<gridSize,blockSize>>>();

  // comment this line out and see what happens
  cudaDeviceSynchronize();

  printf("Goodbye world! from the host!\n");

  return 0;
}