How to apply algorithms in parallel across blocks?

[ctessum]

It seems like it would make sense to allow the user to control how operations are applied (i.e. whether to use multithreading or not) when the same operation needs to be applied to all the blocks. The code that I added allows a MapAlgorithm argument when initializing the matrix, which specifies what algorithm to use for parallel operations. I didn't include the code for that, but it is here: https://github.com/EarthSciML/EarthSciMLBase.jl/blob/main/src/mtk_grid_func.jl#L7

There may be a better way to do it, though.

[oscardssmith]

I don't think making MapAlgorithm part of the type is a good idea. Depending on the circumstance, users may want to sometimes use the same matrix with different parallelism for different parts of an algorithm.

[ctessum]

Sure, I defer to your judgement.

[oscardssmith]

I think the best answer here might be hooking into the LinearSolve.jl interface in which case you can make algorithms BlockLu(threadmap=...) or something like that.

[ctessum]

Sounds good.