Repository navigation
Draft diagnostic: reproduce official Arena PGO build on magma false REJECT - #127
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Arena build-fidelity diagnostic (draft only)
The official MathGraph checker declaration specifies a three-stage profile-guided build that earlier local regression runs did not reproduce:
-Cprofile-generate, training on officialinit-prelude,llvm-profdata merge, and-Cprofile-use(all with-C target-cpu=native). Its runtime usesnum_threads:2.This experiment uses exact published source commit
78c7502bac8a5ba000057b3f083bc0595ac65750; nosrc/changes. It compiles an ordinary matched release and an Arena-style PGO binary on one runner, pins the official test tar SHA and the exactperf/magma-list-pair-n21source SHA, then compares n7 and n21 verdicts, wall, CPU and RSS while preserving profile and negative evidence.Why: official public Arena currently records an unexpected REJECT on valid n21 with ~14.6 GB RSS, while a plain release compiled on a GitHub runner accepted the identical input with ~12.7 GB RSS (run #41745436392). An all-fast-paths-off variant was slower and used more memory there, so the flags were not sufficient to explain the discrepancy.
Boundaries: exact build procedure tested but not necessarily the same Arena CPU, Nix toolchain or hardware. No change to Arena release, no soundness or speed claim beyond the measured gate. If the PGO binary reproduces the failure, follow with exact-source PGO candidate of six-guard PR #125 and a source-pinned memory causal tournament.