C++ 如何构建包含cuda函数和c++;模板函数
代码如下:C++ 如何构建包含cuda函数和c++;模板函数,c++,stl,cuda,C++,Stl,Cuda,代码如下: #include "cuda_runtime.h" #include "device_launch_parameters.h" #include <list> #include <iterator> #include <stdio.h> template<class T> void InsertValueNode(std::list<T> &ValueList,T NewValue) { std::list
#include "cuda_runtime.h"
#include "device_launch_parameters.h"
#include <list>
#include <iterator>
#include <stdio.h>
template<class T> void InsertValueNode(std::list<T> &ValueList,T NewValue)
{
std::list<T>::iterator itCurVal = ValueList.begin();
while(itCurVal != ValueList.end())
{
if(NewValue < (*itCurVal))
break;
itCurVal++;
}
std::insert_iterator< std::list<T> > itNewNode(ValueList, itCurVal);
*itNewNode = NewValue;
}
__global__ void addKernel(int *c, const int *a, const int *b)
{
int i = threadIdx.x;
c[i] = a[i] + b[i];
}
//to simplify the problem, just imply an empty main function
int main()
{
return 0;
}
如果我将该文件重命名为.cpp并重新生成,则可以编译代码,但会出现“error LNK2001:unresolved external symbol threadIdx”的链接错误。
那么如何构建代码? 这可能是CUDA C++前端/解析器对标准库模板的限制。我可以用CUDA 8和gcc 4.8.5复制它
作为一种解决办法,这似乎是可行的:
template<class T> void InsertValueNode(std::list<T> &ValueList,T NewValue)
{
auto itCurVal = ValueList.begin();
while(itCurVal != ValueList.end())
{
if(NewValue < (*itCurVal))
break;
itCurVal++;
}
std::insert_iterator< std::list<T> > itNewNode(ValueList, itCurVal);
*itNewNode = NewValue;
}
模板void InsertValueNode(std::list&ValueList,T NewValue)
{
auto itCurVal=ValueList.begin();
while(itCurVal!=ValueList.end())
{
如果(新值<(*itCurVal))
打破
itCurVal++;
}
std::insert_迭代器itNewNode(ValueList,itCurVal);
*itNewNode=NewValue;
}
使用nvcc引导编译时,只要使用-std=c++11
选项即可。您可能想考虑向NVIDIA提交一份bug报告
template<class T> void InsertValueNode(std::list<T> &ValueList,T NewValue)
{
auto itCurVal = ValueList.begin();
while(itCurVal != ValueList.end())
{
if(NewValue < (*itCurVal))
break;
itCurVal++;
}
std::insert_iterator< std::list<T> > itNewNode(ValueList, itCurVal);
*itNewNode = NewValue;
}