Implement matrix-matrix and matrix-matrix-matrix product

[fverdugo]

This would be needed for multigrid preconditioners.

[stevengj]

Wouldn't you generally want to leave such products implicit, ala https://github.com/JuliaSmoothOptimizers/LinearOperators.jl, i.e. A * B returns an operator that computes (A * B) * x via A * (B * x)?

That being said, matrix–matrix products are nice to have in general.

[carlodev]

I would try to implement this feature. I was looking at the PSparseMatrix and PVector product. Do you have any suggestions on where I can start from? or reference to better understand how the code/package internally works (I get lost in the indexing notation)?

[fverdugo]

Hi Carlos, I am refactoring the code with the main objective of improving the documentation so that others can contribute easier.

See the work going on in branch https://github.com/fverdugo/PartitionedArrays.jl/tree/remove_abstractpdata

(the name is a bit confusing, sorry...)

If you are not in a hurry, it would better to wait.

In the meanwhile, you can build and read the documentation in the branch and start providing feedback for the parts that are relatively mature (parallel primitives and partitioned ranges).

[stevengj]

@fverdugo, can you link the commit where this was completed?

[fverdugo]

Yes, I should have mentioned this, while closing the issue.

It was in PR #134 as part of the implementation of smoothed aggregation AMG solvers.