Replace NLsolve.jl with NonlinearSolve.jl TrustRegion - #261
Draft
ChrisRackauckas-Claude wants to merge 1 commit into
Draft
Conversation
Copy link
Copy Markdown
Sorry, something went wrong.
…e solving
NLsolve.jl's trust region dogleg solver is replaced by NonlinearSolve.jl's
TrustRegion() in both KPS3 and KPS4 find_steady_state! functions. This
removes NLsolve and NLSolversBase as direct dependencies while NonlinearSolve
was already a transitive dependency via OrdinaryDiffEqNonlinearSolve.
The custom finite-difference Jacobian (make_jac) is retained because the
model's MVector{3, Float64} internals cannot hold ForwardDiff.Dual numbers.
NLsolve's autoscale=true is not needed — TrustRegion works without it and
actually produces slightly better results on the l=392 test case.
All 452 tests pass (441 pass + 11 pre-existing broken).
Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
Copy link
Copy Markdown
Member
|
This branch is not passing the tests: julia> include("test/test-steady-state-kps4.jl")
Activating project at `~/repos/KiteModels.jl/test`
┌ Warning: find_steady_state!: solver did not converge! retcode=Stalled
└ @ KiteModels ~/repos/KiteModels.jl/src/KPS4.jl:704
test_find_steady_state: Test Failed at /home/ufechner/repos/KiteModels.jl/test/test-steady-state-kps4.jl:42
Expression: pre_tension > 1.0001
Evaluated: 1.0 > 1.0001
Stacktrace:
[1] macro expansion
@ ~/.julia/juliaup/julia-1.11.9+0.x64.linux.gnu/share/julia/stdlib/v1.11/Test/src/Test.jl:680 [inlined]
[2] macro expansion
@ ~/repos/KiteModels.jl/test/test-steady-state-kps4.jl:42 [inlined]
[3] macro expansion
@ ~/.julia/juliaup/julia-1.11.9+0.x64.linux.gnu/share/julia/stdlib/v1.11/Test/src/Test.jl:1709 [inlined]
[4] top-level scope
@ ~/repos/KiteModels.jl/test/test-steady-state-kps4.jl:30
┌ Warning: find_steady_state!: solver did not converge! retcode=MaxIters
└ @ KiteModels ~/repos/KiteModels.jl/src/KPS4.jl:704
test_find_steady_state: Test Failed at /home/ufechner/repos/KiteModels.jl/test/test-steady-state-kps4.jl:42
Expression: pre_tension > 1.0001
Evaluated: 1.0 > 1.0001
Stacktrace:
[1] macro expansion
@ ~/.julia/juliaup/julia-1.11.9+0.x64.linux.gnu/share/julia/stdlib/v1.11/Test/src/Test.jl:680 [inlined]
[2] macro expansion
@ ~/repos/KiteModels.jl/test/test-steady-state-kps4.jl:42 [inlined]
[3] macro expansion
@ ~/.julia/juliaup/julia-1.11.9+0.x64.linux.gnu/share/julia/stdlib/v1.11/Test/src/Test.jl:1709 [inlined]
[4] top-level scope
@ ~/repos/KiteModels.jl/test/test-steady-state-kps4.jl:30
test_find_steady_state: Test Failed at /home/ufechner/repos/KiteModels.jl/test/test-steady-state-kps4.jl:45
Expression: isapprox(tether_length(kps4_local), 1.008954l, rtol = 0.01)
Evaluated: isapprox(197.42057783848486, 201.7908; rtol = 0.01)
Stacktrace:
[1] macro expansion
@ ~/.julia/juliaup/julia-1.11.9+0.x64.linux.gnu/share/julia/stdlib/v1.11/Test/src/Test.jl:680 [inlined]
[2] macro expansion
@ ~/repos/KiteModels.jl/test/test-steady-state-kps4.jl:45 [inlined]
[3] macro expansion
@ ~/.julia/juliaup/julia-1.11.9+0.x64.linux.gnu/share/julia/stdlib/v1.11/Test/src/Test.jl:1709 [inlined]
[4] top-level scope
@ ~/repos/KiteModels.jl/test/test-steady-state-kps4.jl:30
┌ Warning: find_steady_state!: solver did not converge! retcode=MaxIters
└ @ KiteModels ~/repos/KiteModels.jl/src/KPS4.jl:704
test_find_steady_state: Test Failed at /home/ufechner/repos/KiteModels.jl/test/test-steady-state-kps4.jl:42
Expression: pre_tension > 1.0001
Evaluated: 1.0 > 1.0001
Stacktrace:
[1] macro expansion
@ ~/.julia/juliaup/julia-1.11.9+0.x64.linux.gnu/share/julia/stdlib/v1.11/Test/src/Test.jl:680 [inlined]
[2] macro expansion
@ ~/repos/KiteModels.jl/test/test-steady-state-kps4.jl:42 [inlined]
[3] macro expansion
@ ~/.julia/juliaup/julia-1.11.9+0.x64.linux.gnu/share/julia/stdlib/v1.11/Test/src/Test.jl:1709 [inlined]
[4] top-level scope
@ ~/repos/KiteModels.jl/test/test-steady-state-kps4.jl:30
test_find_steady_state: Test Failed at /home/ufechner/repos/KiteModels.jl/test/test-steady-state-kps4.jl:45
Expression: isapprox(tether_length(kps4_local), 1.008954l, rtol = 0.01)
Evaluated: isapprox(389.9800737647081, 395.50996799999996; rtol = 0.01)
Stacktrace:
[1] macro expansion
@ ~/.julia/juliaup/julia-1.11.9+0.x64.linux.gnu/share/julia/stdlib/v1.11/Test/src/Test.jl:680 [inlined]
[2] macro expansion
@ ~/repos/KiteModels.jl/test/test-steady-state-kps4.jl:45 [inlined]
[3] macro expansion
@ ~/.julia/juliaup/julia-1.11.9+0.x64.linux.gnu/share/julia/stdlib/v1.11/Test/src/Test.jl:1709 [inlined]
[4] top-level scope
@ ~/repos/KiteModels.jl/test/test-steady-state-kps4.jl:30
Test Summary: | Pass Fail Total Time
test_find_steady_state | 10 5 15 1.6s
ERROR: LoadError: Some tests did not pass: 10 passed, 5 failed, 0 errored, 0 broken.
in expression starting at /home/ufechner/repos/KiteModels.jl/test/test-steady-state-kps4.jl:29
|
Sorry, something went wrong.
ufechner7
marked this pull request as draft
February 19, 2026 10:14
Copy link
Copy Markdown
|
v1.12 is fine but v1.11 is not. Peculiar. |
Sorry, something went wrong.
Copy link
Copy Markdown
|
now that the linear solve issue is fixed we should rerun tests |
Sorry, something went wrong.
Copy link
Copy Markdown
Author
|
Retested locally on Julia 1.11.9 with LinearSolve.jl v3.59.1 (which includes the fix from LinearSolve.jl#888). All tests pass: Julia 1.11.9 LinearSolve v3.59.1 NonlinearSolve v4.16.0 Test Summary: | Pass Broken Total Time Testing KiteModels... | 441 11 452 1m22.0s Aqua | 9 9 27.5s Helpers | 19 19 1.3s Inertia Calculation | 24 24 2.4s Interface | 14 14 0.7s Kps3 Fails | 4 4 1.0s Kps3 | 67 67 5.6s Kps4 | 200 11 211 17.6s Orientation | 81 81 1.0s Staticarrays | 1 1 3.2s Steady State Kps4 | 15 15 0.4s KPS3 benchmarking.... | 2 2 4.1s KPS4 benchmarking... | 5 5 16.7s The root cause on 1.11 was LinearSolve.jl#887: MKLLUFactorization does in-place getrf! which corrupts cache.A with L\U factors. When LU fails on singular/rank-deficient Jacobians, DefaultLinearSolver's QR safety fallback got the corrupted matrix instead of the original, producing garbage Newton steps. This only affected Julia 1.11 (where MKL_jll loads) — on 1.12 MKL_jll doesn't load so the regular LUFactorization was used, which allocates a new object and leaves cache.A intact. The fix in LinearSolve.jl#888 stores a backup of the original matrix and restores it before the QR fallback. With LinearSolve v3.59.1 this PR's KPS4 steady-state tests all converge correctly on Julia 1.11. |
Sorry, something went wrong.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Motivation
NLsolve.jl has known issues with subnormal Jacobian entries from NLSolversBase >= 7.10 (see PR #259). Rather than pinning NLSolversBase or maintaining workarounds, this PR switches to NonlinearSolve.jl which:
Changes
Test plan
Related
🤖 Generated with Claude Code