cuFFT memory allocation

[Christopher Wright]

Hi all,

Does numba/numbapro have a function which computes the memory allocation needed for a cuFFT?

Thanks!

[Stanley Seibert]

Unfortunately, I can't find a cuFFT function to estimate the memory required for the plan ahead of time.  Without that, we can't really expose the feature to Python.  NVIDIA seems to assume you'll just catch the memory error, but not tell you how close you were to meeting it.

--
You received this message because you are subscribed to the Google Groups "Numba Public Discussion - Public" group.
To unsubscribe from this group and stop receiving emails from it, send an email to numba-users...@continuum.io.
To post to this group, send email to numba...@continuum.io.
To view this discussion on the web visit https://groups.google.com/a/continuum.io/d/msgid/numba-users/7c1a19e9-4ea4-4bd5-9e77-f3dec934e500%40continuum.io.
For more options, visit https://groups.google.com/a/continuum.io/d/optout.

[Christopher Wright]

Do these functions do it?

http://docs.nvidia.com/cuda/cufft/#work-estimate-refined 

http://docs.nvidia.com/cuda/cufft/#function-cufftgetsize1d

To view this discussion on the web visit https://groups.google.com/a/continuum.io/d/msgid/numba-users/CADv3RKSVGPaen%2B499TAqotqBOygzoaNoE1pmZ%3DsHuqTTVVO%2BWA%40mail.gmail.com.