Hi folks, I have a nagging feeling that this question has to have been asked and answered before but I've only found tantalizing hints that it is possible (and dare I say even easy). We use the NVIDIA Nsight tool to analyze our all-C/CUDA projects and I'd like to be able to do the same with the Python scripts I write that use the fabulous PyCUDA. How do I go about doing this?
In early 2011 [1], a question appeared on this list that implied that the poster was already doing this. I have tried attaching an already-running process (in this case, IPython running inside Sage) but get an error, "'Launching attach_launch' has encountered a problem', with the following details: Error in final launch sequence Failed to execute MI command: -target-attach 28235 Error message from debugger back end: ptrace: Operation not permitted. ptrace: Operation not permitted. I'll attempt to debug this problem, but I thought to ask about alternative ways to get this to happen, e.g., compiling the Python code into a C representation (Cython?), etc. Ubuntu 11.10 64-bit, Python 2.7.2, PyCUDA 2012.1. [1] http://lists.tiker.net/pipermail/pycuda/2011-March/002934.html Thanks, Ahmed -- Ahmed Fasih [email protected] [email protected] 614 547 3323 (Google Voice) _______________________________________________ PyCUDA mailing list [email protected] http://lists.tiker.net/listinfo/pycuda
