Copy Computing Overlap in Python

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi all, I’m working on a problem involving the Longstaff Schwart algo for a gas storage. I need to work on many big matrices, and I was hoping to use copy computing overlap to speed-up the process. I was hoping to use python, but I have found no documentation about CC Overlap. Do you know if it is possible to use this technique using numba or pycuda -or any other python library? -

Thank you for your help.

-if this is not the most relevant section, please move it to the most appropriate one -

2

its possible. for pycuda, a number of the basic concepts (streams, async copy operations, etc.) are illustrated here. I’m sure there are other examples. For numba CUDA this seems to be a reasonable write-up. For general treatment of CUDA concurrency, which covers all the necessary elements (streams, pinned host data, async operations) see here, section 7.

3

Hi Robert, thank you very much for your reply. The references you posted seem to be very promising.
Thank you for your attention.