VV: Compute Schmids Fully V&V'ed - #1723
Open
imikejackson wants to merge 3 commits into
Open
Conversation
4 tasks
imikejackson
force-pushed
the
vv/ComputeSchmids
branch
from
August 26, 2026 13:44
1b865fd to
3f9ae91
Compare
Summary: - Found and fixed 4 bugs (two in EbsdLib on topic/3_1_1_staging: CubicOps normalized the Schmid factor with the float literals 1.732f/1.414f instead of sqrt(3)/sqrt(2), inflating every value by +0.0180353% and letting it exceed the physical maximum of 0.5; and HexagonalLowOps read schmidfactor uninitialized while the seven no-slip-system Laue classes left angleComps untouched, leaking the previous Feature's angles. Two in simplnx: all five output arrays were created with no fill value where legacy passed initValue 0, so the filter stated no requirement about the rows it never writes -- a latent port regression, since the in-core store factory supplies a zero initValue of its own, established by experiment and recorded as such; and crystalStructures[featurePhases[i]] was an unbounded index); - documented 5 deviations from DREAM3D 6.5.171 (D1 +0.018% normalizer bias, D2 -301 sentinel in the legacy angle arrays, D3 undefined EbsdLib Schmid outputs plus twelve known-open SC-2-class hexagonal divisors and the new slipsys=0 sentinel, D4 the Phis/Lambdas cosines-vs-radians unit flip and renamed arrays, D5 the table-relative override-path symmetry-operator index); - retired 0 tests (both pre-existing tests kept; the archive-consuming test's three float comparisons were converted from equality to the exact per-component EbsdLib bias relationship, which is strictly stronger and keeps all 214 regression checks); - augmented existing tests with 12 value-asserting *Class 1 (Analytical) + Class 4 (Invariant)* fixtures plus 5 option/guard fixtures across 4 new test cases, 601 assertions total (measured); - added 3 V&V source-tree deliverables (report, deviations, provenance); - documented the loading-direction normalization, the tie-break rule, both Slip Systems numbering schemes and their two "no slip system found" zero cases, the Phis/Lambdas unit flip, and corrected the claim that Poles is a Miller index plus the (05) -> (04) example-pipeline citation in the user-facing doc. MERGE DEPENDENCY: this PR is blocked until EbsdLib 3.1.1 is released. Two of the four fixes live in EbsdLib and the Class 1 oracle asserts the post-fix values to 1e-6, so the branch carries a vcpkg.json bump of ebsdlib from "version>=" 3.1.0 to 3.1.1. Until that version is published, vcpkg rejects the manifest at configure time in EVERY preset -- including the local SIMPLNX_USE_LOCAL_EBSD_LIB=ON preset, because manifest version constraints are resolved regardless of which features are enabled. All verification was performed with the bump not applied, against the local EbsdLib source through the NX-Com-Qt69-Vtk96-Rel-EbsdLib preset; the bump was applied last as a manifest-only change. Do not reconfigure a build directory on this branch until 3.1.1 exists. This consequence was accepted by the requester on 2026-08-20. Signed-off-by: Michael Jackson <mike.jackson@bluequartz.net>
imikejackson
force-pushed
the
vv/ComputeSchmids
branch
from
August 26, 2026 14:48
3f9ae91 to
26160af
Compare
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
1 participant
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
Full V&V of
ComputeSchmidsFilter, including fixes to EbsdLib itself on itstopic/3_1_1_stagingbranch. Exact-arithmetic Class 1 oracles (textbook Schmid factors: 1/√6 for ⟨001⟩ loading, etc.) replace the circular exemplar comparison against6_6_stats_test_v2(archive lines untouched — other consumers remain; the retained archive test now asserts the exact pre-fix bias relationship instead of raw equality).vcpkg.jsonto EbsdLib>= 3.1.1and cannot configure or pass CI until EbsdLib 3.1.1 is released (staging commits exist locally: exact Schmid normalizer constants replacing1.732f/1.414f— which allowed Schmid factors above the physical 0.5 bound — plus HexagonalLow/stub-class output initialization and a stub-locking test).-13501/-13502); explicit zero fills on all created outputs — investigated experimentally and reclassified as a latent port regression (the in-core store factory hard-codes zero init; the fill is load-bearing only for external large-data store factories, logged open); per-iteration reinit hardening against EbsdLib stub leaks.SlipSystemsnumbering divergence vs legacy traced to reordered symmetry tables (same six tying group elements, different first-hit — a labeling convention, not a value difference); legacy initializesPhis/Lambdasto −301; theangleCompscosine-vs-radian unit flip documented in the user doc; twelve hexagonal sqrt-divisor literals recorded as known-open follow-up (one carries 3.4× the cubic bias).Test Plan
ctest -R "OrientationAnalysis::ComputeSchmids"— 6/6 pass (601 assertions), built against local EbsdLib staging via theNX-Com-Qt69-Vtk96-Rel-EbsdLibpresetctest -R "OrientationAnalysis::"— 269/269 unit tests in that buildNote for merge sequencing: this PR and the ComputeShapes PR both add
vv/provenance/6_6_stats_test_v2.md— whichever merges second needs a trivial add/add content reconciliation.