owned this note   changed 4 years ago  
Published Linked with GitHub

T-compiler Meeting Agenda 2021-09-30

Tracking Issue

Announcements

  • New MCPs (take a look, see if you like them!)
  • Old MCPs (not seconded, take a look)
    • "rustdoc is using rustc_ast_pretty, would it be possible to make it somewhat "stable"?" compiler-team#403 (last review activity: GH none, Zulip: +2months ago)
    • "CI should exercise (subset of) tests under –stage 1" compiler-team#439 (last review activity: GH none, Zulip: about 1 month ago)
    • "Accept pc in place of unknown and unknown in place of pc for x86_64 and i?86 targets" compiler-team#441 (last review activity: GH none, Zulip: about 1 week ago)
    • "Tier 3 target proposal: x86_64-unknown-none (freestanding/bare-metal x86-64)" compiler-team#462 (last review activity: GH: 10 days ago, Zulip: about 4 days ago)
  • Pending FCP requests (check your boxes!)
    • "Write text output files to stdout if options like -o - or --emit asm=- are provided" compiler-team#431
  • Things in FCP (make sure you're good with it)
    • No FCP requests this time.
  • Accepted MCPs
  • Finalized FCPs (disposition merge)
    • "Make *const (), *mut () okay for FFI" rust#84267
    • "Make #[derive(A, B, ...)] cfg-eval its input only for A, B, ... and stabilize feature(macro_attributes_in_derive_output)" rust#87220

WG checkins

@WG-self-profile by @mw and @Wesley Wiser (previous checkin):

@rylev and @mw have been working towards recording non-timestamp data like the sizes of files that the compiler emits – e.g. object files and incr. comp. on-disk data as part of the self-profile infrastructure. This hasn't quite landed yet, but significant progress has been made.

We've also introduced new apis that allow reading older versions of the on-disk data which will help perf.rlo work more smoothly as new versions are released.

Backport nominations

T-compiler stable / T-compiler beta

  • :beta: "Disable the evaluation cache when in intercrate mode" rust#88994
    • beta nominated by @simulacrum to fix #88869, a beta regresion
    • @Mara comments this causes the regression rust#89119 now fixed by rust#89125
    • seems to be perf neutral
    • should also #89125 be beta-backport nominated?
  • :beta: "Don't use projection cache or candidate cache in intercrate mode" rust#89125
    • opened by @Aaron1011
    • fixes rust#88969 a regression from stable to beta
  • :beta: "2229: Mark insignificant dtor in stdlib" rust#89144
  • :beta: "[rfc 2229] Drop fully captured upvars in the same order as the regular drop code" rust#89208
    • Opened by @Wesley Wiser
    • We need this for Rust 2021
  • :beta: "Don't normalize opaque types with escaping late-bound regions" rust#89285
    • Opened by @Jack Huey
    • Fixes a P-medium rust#88862 causing a cargo build hang
  • No stable nominations for T-compiler this time.

T-rustdoc stable / T-rustdoc beta

  • :beta: "Use the correct edition for syntax highlighting doctests" rust#89277
    • nominated by @simulacrum as needed for 2021 edition
    • T-rustdoc (cc @GuillaumeGomez) is fine with the backport
  • No stable nominations for T-rustdoc this time.

Image Not Showing Possible Reasons
  • The image file may be corrupted
  • The server hosting the image is unavailable
  • The image path is incorrect
  • The image format is not supported
Learn More →
/
Image Not Showing Possible Reasons
  • The image file may be corrupted
  • The server hosting the image is unavailable
  • The image path is incorrect
  • The image format is not supported
Learn More →
/
Image Not Showing Possible Reasons
  • The image file may be corrupted
  • The server hosting the image is unavailable
  • The image path is incorrect
  • The image format is not supported
Learn More →

PRs S-waiting-on-team

T-compiler

  • "On macOS, make strip="symbols" not pass any options to strip" rust#88137
    • @Wesley Wiser has a note about asking if Alex has any insight into macOS

Oldest PRs waiting for review

T-compiler

  • "Replace dominators algorithm with simple Lengauer-Tarjan" rust#85013 (last review activity: 4 months ago)
  • "Account for incorrect impl Foo<const N: ty> {} syntax" rust#85346 (last review activity: 3 months ago)
  • "Diagnostic tweaks" rust#85102 (last review activity: 3 months ago)
  • "might_permit_raw_init: also check arrays (take two)" rust#87041 (last review activity: 2 months ago)
  • "When recovering from a : in a pattern, use adequate AST pattern" rust#87160 (last review activity: 2 months ago)

Issues of Note

Short Summary

P-critical

T-compiler

  • No P-critical issues for T-compiler this time.

T-rustdoc

  • No P-critical issues for T-rustdoc this time.

P-high regressions

P-high beta regressions

  • "no errors encountered even though delay_span_bug issued" rust#87757
  • "regression: cycle in MIR opts" rust#88972
  • "Trait upcasting shadows (trait object) deref coercion" rust#89190
    • Currently unassigned
    • There a Zulip topic, involves also T-lang

Unassigned P-high nightly regressions

  • No unassigned P-high nightly regressions this time.

Performance logs

triage logs for 2021-09-28

The largest story for the week are the massive improvements that come from enabling the new pass manager in LLVM which leads to consistent 5% to 30% improvements across almost all test cases. The regressions were mostly minor with clear paths for addressing the ones that were not made with some specific trade off in mind.

Triage done by @rylev. Revision range: 7743c9fadd64886d537966ba224b9c20e6014a59..83f147b3baf21acfc367a6da1045d212cd3957e4

4 Regressions, 4 Improvements, 3 Mixed; 0 of them in rollups
43 comparisons made in total

Regressions

Revise never type fallback algorithm #88804

Introduce Rvalue::ShallowInitBox #89030

  • Moderate regression in instruction counts (up to 1.9% on incr-patched: println builds of syn)
  • Perf regression is happening in real-world optimized builds which we would expect if we're making LLVM do more work.
  • The author has an idea for how to reduce the pressure on LLVM and perhaps win back some of the perf, but was unsure if the regression was large enough to warrant that investigation.
  • Left a comment asking the author to think about prioritizing that investigation.

Fix spacing of links in inline code. #88343

  • Large regression in instruction counts (up to 2.1% on incr-unchanged builds of webrender-wrench)
  • This change is only in doc comments for the standard library so large regressions are quite surprising.
  • This has impacted the incr-unchanged scenario of only one benchmark. Perhaps the docs led to a change in how incremental cache was being stored which could have an impact?

Suggest both of immutable and mutable trait implementations #89263

  • Moderate regression in instruction counts (up to 2.0% on full builds of diesel)
  • The regression occurs in evaluate_obligation which seems like it would be effected by this change, but the only impacted benchmark is diesel doc which doesn't trigger this diagnostic
  • The change is large enough though that it being noise would be quite surprising.
  • Left a comment

Improvements

  • Migrate in-tree crates to 2021 #89103
  • Disable visible path calculation for PrettyPrinter in Ok path of compiler #89120
  • Make Duration respect width when formatting using Debug #88999
  • Enable new pass manager with LLVM 13 #88243

Mixed

Use ZST for fmt unsafety #89139

  • Moderate improvement in instruction counts (up to -1.6% on full builds of cranelift-codegen)
  • Moderate regression in instruction counts (up to 0.5% on full builds of deeply-nested-async)
  • The regressions were already triaged here.

Support #[track_caller] on closures and generators #87064

  • Large improvement in instruction counts (up to -1.8% on incr-unchanged builds of webrender-wrench)
  • Small regression in instruction counts (up to 0.5% on incr-unchanged builds of helloworld)
  • The large majority of performance regressions were quite small and many were historically noisy benchmarks.
  • Looking at the non-noisy benchmarks nothing seemed to stand out as a clear culprit. -
  • There were some test cases that indicated large jumps in incr_comp_persist_dep_graph but this did not hold true across all incremental test cases, and nothing in the changes seemed to jump out at me as potentially causing this issue.
  • Added a comment in the PR asking for additional opinions.

Don't normalize opaque types with escaping late-bound regions #89285

  • Very large improvement in instruction counts (up to -86.7% on full builds of issue-88862)
  • Moderate regression in instruction counts (up to 1.3% on full builds of deeply-nested-async)
  • This has already been triaged.
  • Essentially it's a targeted fix for a severe perf regression that can happen in limited circumstances. The fix will be backported to the latest beta.

Untriaged Pull Requests

Nominated Issues

T-compiler

  • "On macOS, make strip="symbols" not pass any options to strip" rust#88137
    • discussed in PR waiting-on-team
  • "Rust can't infer appropriate generics for function when it should" rust#89242

RFC

  • No nominated RFCs for T-compiler this time.

Misc

Select a repo