PyCuda: lets you access Nvidia's CUDA parallel computation API from Python(mathema.tician.de)
mathema.tician.de
PyCuda: lets you access Nvidia's CUDA parallel computation API from Python
http://mathema.tician.de/software/pycuda
1 comments
Greg and I talked about this last week, and it seems quite useful. I've been playing around with Cuda computations for dust temperatures and calling it from python would be useful. The biggest hurdle isn't the interface, though, it's rewriting your code to run on the GPU, so pycuda isn't such a big change since you still have to write the GPU code in C.
"you still have to write the GPU code in C"
You're almost better off writing python modules in C that take advantage of CUDA for specific applications instead of using a general one that really doesn't hide any of the more difficult pats of the CUDA API anyways.
You're almost better off writing python modules in C that take advantage of CUDA for specific applications instead of using a general one that really doesn't hide any of the more difficult pats of the CUDA API anyways.
I've been intending to play with CUDA compuation for a while, this seems like a great way to begin. It's especially nice that the documentation seems helpful. I'd be curious to hear peoples experiences with pycuda.