You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Follow-ups from review of #4793: Gauss4 commutator sign gives second-order convergence #4965
🚧 UNREVIEWED — AI-generated issue filed by an agent running as @ChrisRackauckas; not reviewed by Chris.
Follow-ups from the independent review (Codex gpt-6.1-sol, verdict MERGE-WITH-ISSUES) of the merged PR #4793. Review comment: #4793 (comment)
[P3 as labelled, non-blocking; numerical] lib/OrdinaryDiffEqLinear/src/linear_perform_step.jl:710 and :718: the Gauss4 commutator has the wrong sign, B*A - A*B where A is the later Gauss node and B the earlier one (defined near line 738); the Magnus double integral needs A*B - B*A. Reported (confirmed by running at review time, identical on base and head): on the exact rotating-frame problem the Float64 non-Krylov errors are 0.004298891155539751, 0.0010706786848178212, 0.0002674136399558906 at dt = 1/4, 1/8, 1/16, an observed order of 2.00 instead of 4; Krylov and ComplexF64 show the same loss. Change both expressions and add a noncommuting analytic convergence regression. The B * A - A * B expressions are still present on the default branch; the convergence numbers were not re-run (unverified).
Activity
Fix the sign of MagnusGauss4's commutator term (SciML#4968)