__inline__ functions with template structs

Accelerated Computing CUDA CUDA Programming and Performance
1

The following compiles:

struct Foo() 

{

  __inline__  void Bar() {} 

};

__global__ void Test( Foo * p )

{

  p->Bar();

}

When I try and get fancy:

template< typename Type >

struct Foo2() 

{

  __inline__  void Bar() {} 

};

__global__ void Test2( Foo2<int> * p )

{

  p->Bar(); // generates an error

}

I get the following error:

Error: Calls are not supported (found non-inlined call to XXXX)

Is this a compiler bug or is this unsupported functionailty?

Thanks,

George

BTW - I’m still vague on what is supported by nvcc.

2

Okay, I refactored a bit to get it to compile.

template< typename Type >

struct Foo

{

  Type data;

};

template< typename Type >

__global__ void Bar( Foo<Type> * f )

{

}

I guess thats more “C-ish”-anyway. I would still like to know if class templates are/will be supported.

Thanks,

George

3

try

 __inline__ __device__ void Bar() {}
4

Awesome! That works. You just saved me a bunch of rework.

Thanks,
George

5

What’s the syntax for calling that templated global function?