Raye Kimmerer

Tracking issue for integrating things like CHOLMOD: - [ ]

List of missing ArrayInterface functions:

No Julia level method to get `how` for a Vector/Array

1

Developing code that interacts with C it would be nice to know if an Array came from Julia or elsewhere (Libc etc). I believe this is determined by the `how`...

UMFPACK `serialization` and `copy` functions

5

We have a need in Julia to both serialize and copy UMFPACK objects (as well as perhaps other solver objects, but we'll leave those alone for now). I've added functions...

[UMFPACK] In-memory `save` and `load` for numeric and symbolic

1

We have a need to 1. copy `umfpack_numeric` and `umfpack_symbolic` types and 2. serialize them in memory. Both of these things can be done by an in-memory `save` and `load`,...

This is broken due to old glibc, we need at least 2.17 because of CUDA (at least because that's as far as I've gotten).

Enable retargeting SuiteSparse solvers to a different binary

1

I came across this while testing, but do we have the ability to choose a different SuiteSparse binary (or set of binaries really) at runtime? I know a number of...

`getrowval` and `getnzval` are defined on SparseMatrixCSC but not on SparseVector

4

`rowval` is defined on both, but curiously `getrowval` is not. I'm not sure exactly why we have two identical accessor functions.

Generalize matmul to arbitrary + and *

3

This request is a little out there, so it'd be totally reasonable to close it. TLDR and Tullio thoughts at the end. My JSOC is to wrap and provide ChainRules...

Hi there. I'm keen to use ArrayLayouts in some upcoming sparse ecosystem stuff. Not so much for the generic linalg but just for the traits. In light of that would...