A problem with cudafe++

Accelerated Computing CUDA CUDA Programming and Performance
1

I’ve encounterd a very interesting problem. When i compile such code

template<typename Type>

Type* cudaHostNew(typename Type::InitStruct& data)

{avoided

   Type* result = NULL;

   cudaMalloc((void**)&devPtr,100);

}

cudafe always throws an error

(33): internal error: assertion failed: gen_expr: bad expr node kind (D:/Bld/rel/gpgpu/toolkit/r2.1/compiler/edg/EDG_3.9/src/cp_gen_be.c, line 9510)

Any idea about thisstrange behaviour? Could it be avoided? Is it a bug or only my incompetence? :unsure:

2

cudafe is known to be fragile especially when mixing C++ code with it. But CUDA 2.1 must be stable (coz templates are officially supported only in 2.1)

btw,

What is that “avoided” doing in your code? Is that a deliberate mistake that you have introduced?

3

I know, it seemed to me that i use cuda 2.1.

Simply nothing, just wrong copy+paste.

4 
template<typename Type>

Type* cudaHostNew(typename Type::InitStruct& data)

{   

   Type* result = NULL;

   cudaMalloc((void**)&devPtr,100);

}

You dont require “typename” in the arg list for “cudaHostNew” function.

Also I am not sure “type::InitStruct&” would work.

You better declare “Type” as “class Type” instead of “typename Type” – though am not very sure about it