Please confirm: constant CUFFT plan size across different devices ... for a given FFT size and type

Hi everyone,

I would expect that, for a given FFT size and type, the plan created by, say, CUFFT’s cufftPlan1d() has a constant size (in memory). Thus it doesn’t vary with the actual device model (GPU and total/available global memory) being used.
Is this assumption correct?

Thanks!

[bump] Anyone…?