[CUDA] hello cuda

#include<iostream>
#include<stdio.h>
#include<stdlib.h>
#include<cutil_inline.h>

using namespace std;

#if __DEVICE_EMULATION__
bool InitCUDA(void){
 return true;
}
#else
bool InitCUDA(void){
 int count=0;
 int i=0;
 cudaError cue = cudaGetDeviceCount(&count);
 if(cue == cudaErrorNoDevice)
 {
  cerr<<"There is no device"<<endl;
  return false;
 }

 cout<<"i found "<<count<<" devices!"<<endl;
 for(i=0; i<count; i++){
  cudaDeviceProp prop;
  if(cudaGetDeviceProperties(&prop, i) == cudaSuccess)
  {
   cout<<"\tdevice name : "<<prop.name<<endl;
   if(prop.major >=1)
    break;
  }
 }
 if(i==count){
  cerr<<"There is no device supporting CUDA"<<endl;
  return false;
 }
 cudaSetDevice(i);
 cout<<"CUDA initialized"<<endl;
 return true;
}
#endif

__global__ static void HelloCUDA(char* result, int num){
 int i=0;
 char p_HelloCUDA[] = "Hello CUDA!";
 for(i=0; i<num; i++)
  result[i] = p_HelloCUDA[i];
}

int main(){
 if(! InitCUDA())
  return 0;

 char* device_result = 0;
 char host_result[12]={0};

 //1. device의 메모리 allocation
 cutilSafeCall(cudaMalloc((void**)&device_result, sizeof(char)*11));

 //2. CUDA함수 실행
 HelloCUDA<<<1,1,0>>>(device_result, 11);
 cutilCheckMsg("Kernel execusion failed\n");
 cudaThreadSynchronize();

 //3.device memory를 host로 할당
 cutilSafeCall(cudaMemcpy((void*)host_result, (const void*)device_result, (size_t)sizeof(char)*11, cudaMemcpyDeviceToHost));
 cout<<endl<<host_result<<endl;

 //4. device
 cutilSafeCall(cudaFree(device_result));
 cout<<"end the cuda hello.."<<endl;
 return 0;
}

어느 블로그에서 보고 따라해본 헬로쿠다