Alexey Radul

Another option is to allow the user to specify an integer dump frequency in time steps rather than time duration. It would be nice for the user-facing interface to allow...

Same labeling rationale as https://github.com/edoddridge/aronnax/issues/99#issuecomment-295060593, except with a higher degree of belief that an actual bug is involved.

I presume this is a rather low-priority issue, as MIM's target users are unlikely to want to run MIM with very many layers -- if the number of layers becomes...

@edoddridge Please edit the description with the list you actually intend to do :) We can also use this ticket (or maybe individual tickets per paper) to keep track of...

Somehow it didn't connect before that NetCDF is a binary format with embedded strings, rather than a text format. That means it shouldn't be materially larger than the native stuff,...

Any sort of Python wrapping should have a negligible effect on performance, or on parallelism, provided the chunks of work done by the wrappee are large enough. For instance, I...

Yes. That overlaps with #36, which calls for a suite of realistic examples to measure; one of the things to measure on it is performance.

As of PR #68, I am somewhat surprised by how much time the toplevel program takes -- on the beta plane bump example, 50% of the runtime in the reduced...

Interesting. I assume you mean nesting `do i = 1, nx` inside `do j = 1, ny`, in contrast to status quo? For some reason, compiling with `gfortran -g -Ofast`...

I am very confused by my attempts to find performance tweaks that work on my machine. There should be two improvements to SOR_solve that should just work: - Replacing the...