Variadic templates?

Hello all,

We are working on a GPU-server without root-access, so I cannot check if this is due to the fact that our GCC is quite old (v4.3.4), but I was wondering if variadic templates are supported on CUDA 5. For some reason, I cannot find any information on this on the web.

If they are supported, are there any limitations to them? Can both device and global functions be declared as a variadic template? If so, I will ask the server-manager to update GCC.

Thanks in advance,
Joren

I’d be inclined to say that CUDA doesn’t support them, as I don’t think CUDA has any C++11 support, and variadic templates were just introduced to C++ in C++11. So you can update your gcc version and use variadic templates in separately compiled host code, but not with CUDA kernels (or really anything parsed by nvcc). I could be wrong though, so don’t take my word for it…

I would suggest that CUDA users interested in particular features of C++11 file an RFE (request for enhancement) through the bug reporting form linked from the registered developer website. This will help the compiler team prioritize features for future development. Thanks.

Thanks guys for your responses (disappointing as they are)… Any idea where I can submit such an RFE? Cheers!

C++11 added a huge number of individual features, so any information that helps prioritize the feature requests and determine the implementation sequence is helpful. To file bugs or RFEs:

[1] Log into the registered developer website at https://developer.nvidia.com/user/register
[2] Click on green link “CUDA/GPU Computing Registered Developer Program”
[3] Under the heading “Reporting an Issue”, click on green link “The Submit a Bug Form”