Oscar Smith

make preconditioners part of the solver rather than a random extra

4

This still needs tests, but I'm putting it up as a PR first to get feedback. The intended goal here is to make the `dolinsolve` functionality in OrdinaryDiffEq of being...

use LinearSolve precs interface

24

This is a post OrdinaryDiffEq solver split version of https://github.com/SciML/OrdinaryDiffEq.jl/pull/2247 (remaking was substantially easier than rebasing). It still needs a little bit of work (and for the LinearSolve side to...

`reshape` and `_vec` appropriately in more places

4

fixes https://github.com/SciML/OrdinaryDiffEq.jl/issues/2383 and related bugs.

reuse jacobians in NonlinearSolveAlg

1

This seems to just work. I think I only turned off reuse because it was making debugging the other bugs in this harder.

Draft: Make W_transform always true

2

Rosenbrock23 and Rosenbrock32 need updating to the W_transform versions, but other than that, this PR removes ~200 lines of complexity for free.

make `memorynew` intrinsic

14

This should speed up making new `Memory`s and allow the compiler to better understand what's going on.

This is a revival of https://github.com/JuliaLang/julia/pull/43738 with a lot of changes. The biggest by far is that with this version of the PR, I'm trying to get the BoundsError to...

remove precs

1

for comaptibility with https://github.com/SciML/OrdinaryDiffEq.jl/pull/2318. Unfortunately, I don't believe there is a way to correctly sequence these to prevent temporary breakage.

Don't setup a `JacVec` if `concrete_jac==true`, and use non `SciMLOperator` `W_prototype`

1

I'm not sure if this is a good idea yet, but I want to run it past CI.