cuda-gdb + python? When will cuda-gdb be updated to 7.x and will it have python support?

To the cuda-gdb team,
Is updating gdb to 7.x on the roadmap? If so, what’s the timeframe like and is Python support being considered?

If neither is planned for the near term, will it be possible to obtain a copy of the latest published cuda-gdb source? I’ve only seen executables, but I’d like to not be too presumptuous and hope that I’ve just missed it.


Never mind my question about the cuda-gdb source – I just found the public NVIDIA ftp site.

Still though, is there a better to communicate with the dev team directly? I have a registered developer account, but the only contact forms I see are for bug reports.

Beyond bug reports, this might be the only way to communicate with the dev team outside of conferences. (Although there is no guarantee that the part of the dev team who could answer your question is reading the forum.) If you have purchased Tesla cards, there might be another communication channel as part of the developer support.

Back to your original question: What do you mean by “Python support”? Have you seen PyCUDA? It allows you to perform some simple Numpy-style operations on the GPU in pure Python as well as compile your own CUDA C kernels on the fly and use those on numpy arrays.

Thank you, but PyCUDA is not what I’m looking for.

Since gdb 7.x releases, a Python API has been available to script the debugger. Documentation of this feature can be found here.

I just asked the debugger team–they’re actively working on this for a future release. And yes, they are very excited about the Python API as well (for QA purposes, so it should be well-supported).

That’s great news! Are they willing to say anything about a timeline? Will it have to wait for the next major SDK release?

Certainly won’t happen before then, and I can’t promise anything more than they’re working on it for a future release.

Whoa, neat!