Simon Jakobi pushed to branch wip/sjakobi/T27602 at Glasgow Haskell Compiler / GHC Commits: a4ddab82 by Simon Jakobi at 2026-08-15T11:54:49+02:00 DEMO: Tighten T9675 residency tolerance (do not merge) Revert the T4830 regression from the previous demo commit; T4830's 'bytes allocated' turns out to be fully deterministic across CI samples, so it cannot show a spread. Instead tighten T9675's residency tolerance from 15% to 2%: its peak_megabytes_allocated jitters by up to ~12% between runs of the same commit, so with a multi-sample baseline the natural jitter fails the window and the new sample-spread output shows a baseline built from genuinely disagreeing samples. Context: #27602 Assisted-by: Claude Fable 5 - - - - - 1 changed file: - testsuite/tests/perf/compiler/all.T Changes: ===================================== testsuite/tests/perf/compiler/all.T ===================================== @@ -133,7 +133,8 @@ test('T9020', test('T9675', [ only_ways(['optasm']), - collect_compiler_residency(15), + # Demo-only: tightened from 15 so natural residency jitter fails. + collect_compiler_residency(2), collect_compiler_stats('bytes allocated',2), ], compile, View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/a4ddab82385619a28b7fd6de1b925e13... -- View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/a4ddab82385619a28b7fd6de1b925e13... You're receiving this email because of your account on gitlab.haskell.org. Manage all notifications: https://gitlab.haskell.org/-/profile/notifications | Help: https://gitlab.haskell.org/help