Mario Lezcano Casado

My point is, when should `-1` be masked out when using this op in a decomp? What's an example where we want to do that.

Then that answers my question above -> We cannot remove the mask. Thanks!

@wconstab there's a (simple-looking) DTensor issue with this PR, who could be the best person to help with it?

ping @wanchaol

Once this PR passes CI, can you put one example PR on top of this where you use it in a decomposition? It'd be best to choose the simplest operation...

>The failure seems to come from special casing of bool Is it an easy fix? If so, can you prepend a PR with the fix?

`pivot=False` is often quite risky, as the class of matrices for which the decomposition has a not-so-nice structure (all its principal minors must be away from zero). For example, it...

The diagonal case you mentioned I don't think it's correct, but it's true that there is at least one family of matrices that I know of for which LU decompositions...

If you have a tall matrix, what you often want to do is to compute the PLU of the transposed matrix and use it to solve the given system. Even...

Also, thank you for digging this up https://github.com/data-apis/array-api/issues/627#issuecomment-1556881760. It looks to mee that we don't want to include the `permute_l` function, as it's just not useful. The matrix `P@L` does...