Uh oh!
There was an error while loading. Please reload this page.
This repository was archived by the owner on Aug 20, 2026. It is now read-only.
fix(reg_flow): route bending==0 Lamé calls to matvec_lame/diag_lame - #39
Merged
balbasty merged 1 commit intoAug 1, 2026
Conversation
_flow_matvec/_flow_diag unconditionally selected the full combined matvec_all/diag_all stencil whenever shears/div were non-zero, even when bending==0, leaving the cheaper Lamé-only matvec_lame/diag_lame stencils (already exercised via relax_lame_) unreachable from the public dispatch. Mirror _flow_kernel's/_flow_relax's existing correct "bending != 0" branch so an elastic-only (non-bending) registration configuration uses the cheaper stencil. Also add regression tests for the corner cross-term sign bug fixed upstream in fastfields-kernels (diag_bending / diag_all boundary correction): on a square domain with the same boundary condition on every axis, diag(0,j,c) must be symmetric under axis (and, for diag_all with shears==div, channel) swap. Confirmed these tests fail against the pre-fix kernels and pass against the fix. Adds the analogous test_diag_boundary_symmetry_2d() to test_reg_field.cpp for the field-side diag_bending fix.
4 tasks
balbasty
commented
Jul 30, 2026
ContributorAuthor
Agent: Generated by Claude Code |
balbasty pushed a commit
that referenced
this pull request
Jul 31, 2026
…flict Both main (via cpu-impl#38's merge, 9e12e7d) and this branch (via cpu-impl#39's merge, b6eae48) independently re-pinned impl since their common ancestor. b6eae48 is a descendant of 9e12e7d (#39 was built on top of #38's merged state), so it is the correct resolution -- kept. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016AjQcY78NgbagPSbPJRr6Z
balbasty
merged commit Aug 1, 2026
8afcab4
into
claude/fastfields-teeny-refactor-js42id
1 check failed
Uh oh!
There was an error while loading. Please reload this page.
This was referenced Aug 1, 2026
Closed
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 freeto subscribe to this conversation on GitHub.
Already have an account?
Sign in.
2 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
_flow_matvec/_flow_diaginreg_flow.cppunconditionally routed any non-zeroshears/divto the full combinedmatvec_all/diag_allstencil, even whenbending == 0, leaving the cheaper Lamé-onlymatvec_lame/diag_lamestencils (already exercised viarelax_lame_) unreachable from the public dispatch._flow_kerneland_flow_relaxalready branch correctly onbending != 0; this mirrors that pattern in_flow_matvec/_flow_diagso an elastic-only (non-bending) registration configuration uses the cheaper ~9-tap stencil instead of the ~25-tap combined one.tests/test_reg_flow.cpp,tests/test_reg_field.cpp) for the corner cross-term sign bug fixed upstream in fastfields-kernels#32: on a square domain with the same boundary condition on every axis,diag(0,j,c)must be symmetric under axis swap (and, for flow'sdiag_allwithshears == div, channel swap too). Confirmed these fail against the pre-fix kernels and pass against the fix.Depends on
fastfields-kernels#32 — the
impl/kernelssubmodule pins here are unchanged (dev-tree symlinks were used to validate locally against the fix); a real (non-symlinked) checkout needs the kernels PR merged and the pin bumped for the new diag-symmetry tests to pass.Test plan
make test CXX=clang++) green: 0 failures across all 11 test binaries, reg_flow now 854 checks (830 + 24 new).matvec_lame/diag_lameinstead ofmatvec_all/diag_alldoes not change any existing test's expected values (mathematically the two stencils must agree whenbending == 0) — this is a performance fix, not a behavior change.Generated by Claude Code