Repository navigation
Draft diagnostic: isolate Flash magma memory regression with five-flag ablation - #126
Draft
heathsanchez wants to merge 3 commits into
Draft
heathsanchez wants to merge 3 commits into
heathsanchez wants to merge 3 commits into
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Causal hypothesis only — do not merge
The published attributed Flash build (
78c7502bac8a5ba000057b3f083bc0595ac65750) has one known false REJECT: official Arena validperf/magma-list-pair-n21(published report ~14.6 GB RSS). Upstream sokonanoda accepts this exact source at ~7.6 GB.This isolated diagnostic disables five published Flash flags together (
DIRECT_VAR_EVAL,DIRECT_FRAMED_PRUNE,INFER_PI_DEMAND_BYPASS,RIGID_INDUCTIVE_NEUTRAL,APP_HC_PREALLOC) without modifying the algorithmic code. The official archive and n21 fixture have exact source hashes. Two binaries compile from exact source, with identical release settings and 2 threads; paired n7/n21 acceptance, wall, CPU and peak RSS are preserved, including failures.Interpretation: If the all-off version accepts n21 with lower memory while published fails, bisect these switches next. If both fail, the switches are not sufficient. If both accept, resource nonreproducibility remains. No combined correctness proof, no nonregression, and no valid grounds for Arena submission.
Follow-up only after a meaningful delta: isolate the responsible switch, keep full protected 179 cases + extra performance tests, then match full Mathlib. This draft does not include the six soundness repairs in PR #125.
Upstream source attribution preserved; no private internal research notes.