Juan Ignacio Polanco

> Maybe I'm answering my own question: it looks like there's a function `transpose!` provided by PencilArrays. Perhaps I can call `transpose!` to achieve steps 1, 7-8, and 10-11? Then...

But perhaps you can avoid the extra transpositions (which can be quite expensive), and work with the following order of transforms: ```julia transforms = ( PencilFFTs.Transforms.FFT!(), PencilFFTs.Transforms.FFT!(), PencilFFTs.Transforms.NoTransform!(), ) ```...

> Ah yes! We can use that algorithm if 1) we distribute _only_ in `y`, or 2) if PencilFFTs allows us to distribute data along the first dimension, so we...

> EDIT: a third possibility is that we support a configuration that explicitly disallows any component of the algorithm that requires z to be local (nitty gritty: vertical integrals for...

That's right, that sounds like a better alternative. Right now I'm not sure if this is possible due to [(artificial) limitations in PencilArrays](https://jipolanco.github.io/PencilArrays.jl/dev/Transpositions/#PencilArrays.Transpositions.Transposition): > when written as a sorted tuple,...

It just needed some very minor changes :) https://github.com/jipolanco/PencilArrays.jl/pull/57 I'll merge this and tag PencilArrays v0.17 once tests pass. Basically, as shown in the example in that PR, you will...

Hi, I'm glad you find this package useful! This definitely looks like a bug, thanks for reporting it. I'll try to get it fixed when I find some time within...

After thinking about this issue, I believe this is a limitation of the recombination procedure, which defines basis functions as linear combinations of B-splines near the boundaries. This can be...

I'll have to give it some more thought as well. I'll be happy to know if you come up with something.

Thank you for this! I think we can figure out later how to move things to a package extension. So I first tried to generate a file and open it...