Skip to content

Allow LinearSolve v4#135

Open
ChrisRackauckas-Claude wants to merge 1 commit into
JuliaLinearAlgebra:masterfrom
ChrisRackauckas-Claude:linearsolve-v4-compat
Open

Allow LinearSolve v4#135
ChrisRackauckas-Claude wants to merge 1 commit into
JuliaLinearAlgebra:masterfrom
ChrisRackauckas-Claude:linearsolve-v4-compat

Conversation

@ChrisRackauckas-Claude

Copy link
Copy Markdown

LinearSolve.jl is preparing a breaking v4.0.0 (batched matrix right-hand sides, SciML/LinearSolve.jl#1072, at @ChrisRackauckas's direction). The breaking surface is only the solution shape for matrix right-hand sides (u becomes n×k instead of a flattened vector) plus an informative error for matrix b with iterative solvers — nothing AlgebraicMultigrid touches: SciMLLinearSolveAlgorithm, init/solve!, the LinearCache interface, build_linear_solution, and the precs API are all unchanged.

Verified locally against LinearSolve v4 (this branch dev'd together with the v4 branch):

  • AlgebraicMultigridJL() (Ruge–Stuben) and SmoothedAggregationAMG as LinearSolve algorithms solve correctly with ReturnCode.Success;
  • aspreconditioner(ruge_stuben(A)) with KrylovJL_GMRES yields bit-identical residuals under v3.87.0 and v4.0.0.

Includes a patch bump to 2.0.1 so this can be tagged directly. Widening compat before v4 is registered lets the ecosystem resolve as soon as v4 lands.

LinearSolve is releasing a breaking v4 (batched matrix right-hand sides,
SciML/LinearSolve.jl#1072). AlgebraicMultigrid uses only the stable
LinearSolve API (SciMLLinearSolveAlgorithm, build_linear_solution, the
LinearCache interface), none of which changes in v4, so widening compat
is sufficient.

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants