T Compiler Meeting Agenda 2021 12 23

T-compiler Meeting Agenda 2021-12-23

Tracking Issue

Announcements

  • Reminder: if you see a PR/issue that seems like there might be legal implications due to copyright/IP/etc, please let the Core team know (or at least message @pnkfelix or @Wesley Wiser so we can pass it along).

MCPs/FCPs

  • New MCPs (take a look, see if you like them!)
  • Old MCPs (not seconded, take a look)
    • “CI should exercise (subset of) tests under –stage 1” compiler-team#439 (last review activity: 4 months 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: 5 months ago)
    • “Make -Z binary-dep-depinfo the default behavior” compiler-team#464 (last review activity: 2 months ago)
    • “Tier 3 target proposal: riscv64gc-linux-android (Android target for riscv64gc)” compiler-team#472 (last review activity: about 13 days ago)
    • -Dwarnings to cover all warnings” compiler-team#473 (last review activity: about 13 days ago)
  • Pending FCP requests (check your boxes!)
    • “Stabilize -Z instrument-coverage as -C instrument-coveragerust#90132
    • “Stabilize -Z print-link-args as -C print-link-argsrust#91606
  • Things in FCP (make sure you’re good with it)
    • “Tracking issue for RFC 2115: In-band lifetime bindings” rust#44524
  • Accepted MCPs
    • No new accepted proposals this time.
  • Finalized FCPs (disposition merge)
    • “Relax priv-in-pub lint on generic bounds and where clauses of trait impls.” rust#90586
    • “Make rustdoc --passes and rustdoc --no-defaults have no effect” rust#91714

WG checkins

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

Wesley Wiser: We’ve had a small amount of activity

Wesley Wiser: The main thing is that we now report sizes of binary artifacts generated on per.rlo

Backport nominations

T-compiler stable / T-compiler beta

  • :beta: “Re-introduce concept of projection cache ‘completion’” rust#89831
  • :beta: “Disable LLVM newPM by default” rust#91190
    • nominated by @nagisa, we want to beta-backport this or #92110
  • :beta: “Don’t perform any new queries while reading a query result on disk” rust#91919
  • :beta: “Backport LLVM changes to disable deferred inlining” rust#92110
    • nominated by @nagisa, we want to beta-backport this or #91190
  • :stable: “Re-introduce concept of projection cache ‘completion’” rust#89831
    • nominated by @Jack Huey, mentions that is as risky as rust#90423 (beta backported approved, stable backport declined)

T-rustdoc beta / T-rustdoc stable

  • No backport nominations for T-rustdoc this time.

:back: / :shrug: / :hand:

PRs S-waiting-on-team

T-compiler

  • None

Oldest PRs waiting for review

T-compiler

  • “Extract rustc_mir_pretty crate from rustc_middle” rust#89636
    • reviewed by @oli 3 months ago
    • PR author seems to need a little more help with squashing the perf. regression
  • “Store liveness in interval sets for region inference” rust#90637
    • last review about 20 days ago from @lqd
    • @simulacrum asks (IIUC) if this can move forward
  • “Fix ICE with inferred type in const or static item” rust#89161
    • last review from Niko about 3 months ago
    • There are some pending comments from author
  • “Extend check for UnsafeCell in consts to cover unions” rust#90383
    • last review from @RalfJ and activity from author about 2 months ago

Issues of Note

Short Summary

P-critical

T-compiler

  • “Huge compile-time regression in beta/nightly” rust#91128
    • @nagisa provided PR rust#92110 to beta-backport LLVM changes to fix this
  • “Compile-time regression between 1.56.1 and 1.57.0 for deeply nested decorator types” rust#91598
    • compilation time regression partially improved in latest nightlities

T-rustdoc

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

P-high regressions

P-high beta regressions

  • No P-high beta regressions this time.

Unassigned P-high nightly regressions

  • No unassigned P-high nightly regressions this time.

Performance logs

triage logs for 2021-12-21

Unfortunately a change introduced in rust-lang/rust#89836 has made performance across different compiler artifacts much more variable by embedding compiler version information (including a git commit hash) in demangled symbol names. This means that even if two compiler artifacts are built from the same exact source code (with only the git commit changed), the compiler will have slightly different performance characteristics. This makes comparisons across pull requests virtually impossible.

The compiler team is still deciding what to do to handle this, but in the mean time, performance testing is largely broken. This issue is currently being tracked in rust-lang/rustc-perf#1126.

Triage done by @rylev. Revision range: 404c8471aba60c2d837fa728e7c729a0f52d5830..3d57c61a9e04dcd3df633f41142009d6dcad4399

2 Regressions, 2 Improvements, 23 Mixed; 9 of them in rollups 23 Untriaged Pull Requests 38 comparisons made in total

Regressions

Update Clippy #92036

  • Very large regression in instruction counts (up to 133.5% on incr-unchanged builds of ctfe-stress-4)
  • Largely caused by a change in how we’re doing benchmarking that should not impact any other PRs other than this one.
  • Luckily we can generally rely on updating Clippy as having no effect whatsoever on compiler performance.

Rollup of 7 pull requests #92059

  • Very large regression in instruction counts (up to 20.0% on incr-unchanged builds of encoding)

Improvements

  • Rollup of 7 pull requests #91962
  • Rollup of 7 pull requests #92099

Mixed

Stabilize asm! and global_asm! #91728

  • Large improvement in instruction counts (up to -3.9% on incr-unchanged builds of deep-vector)
  • Moderate regression in instruction counts (up to 1.4% on incr-unchanged builds of html5ever)

Rollup of 7 pull requests #91959

  • Small improvement in instruction counts (up to -0.5% on incr-unchanged builds of helloworld)
  • Very large regression in instruction counts (up to 6.2% on incr-unchanged builds of inflate)

extend simplify_type #86986

  • Large improvement in instruction counts (up to -4.7% on full builds of hyper-2)
  • Very large regression in instruction counts (up to 8.4% on incr-patched: sparse set builds of regex)

Include rustc version in rustc_span::StableCrateId #89836

  • Very large improvement in instruction counts (up to -11.6% on incr-unchanged builds of tuple-stress)
  • Large regression in instruction counts (up to 4.7% on incr-unchanged builds of unicode_normalization)

Optimize vec::retain performance #91527

  • Large improvement in instruction counts (up to -4.4% on incr-unchanged builds of unicode_normalization)
  • Large regression in instruction counts (up to 4.1% on incr-patched: println builds of regression-31157)

Rollup of 6 pull requests #91996

  • Moderate improvement in instruction counts (up to -2.4% on incr-patched: println builds of regression-31157)
  • Very large regression in instruction counts (up to 46.9% on incr-unchanged builds of clap-rs)

rustdoc: don’t clone already owned Path and modify it inplace #91833

  • Very large improvement in instruction counts (up to -31.9% on incr-unchanged builds of clap-rs)
  • Large regression in instruction counts (up to 4.9% on incr-patched: println builds of regression-31157)

Fix #91306 by deriving all access from a single *mut T #91616

  • Moderate improvement in instruction counts (up to -1.3% on incr-unchanged builds of regression-31157)
  • Very large regression in instruction counts (up to 7.3% on incr-unchanged builds of ctfe-stress-4)

Rollup of 7 pull requests #92003

  • Very large improvement in instruction counts (up to -6.8% on incr-unchanged builds of ctfe-stress-4)
  • Very large regression in instruction counts (up to 13.0% on incr-unchanged builds of tuple-stress)

Implement normalize_erasing_regions queries in terms of ‘try’ version #91672

  • Very large improvement in instruction counts (up to -11.6% on incr-unchanged builds of tuple-stress)
  • Very large regression in instruction counts (up to 7.4% on incr-unchanged builds of ctfe-stress-4)

rustdoc: Pretty-print assoc const defaults on-demand #91812

  • Very large improvement in instruction counts (up to -6.8% on incr-unchanged builds of ctfe-stress-4)
  • Very large regression in instruction counts (up to 14.3% on incr-unchanged builds of tuple-stress)

Update cargo #92023

  • Very large improvement in instruction counts (up to -12.5% on incr-unchanged builds of tuple-stress)
  • Very large regression in instruction counts (up to 8.3% on incr-unchanged builds of wg-grammar)

Do array-slice equality via array equality, rather than always via slices #91838

  • Small improvement in instruction counts (up to -1.7% on full builds of piston-image)
  • Moderate regression in instruction counts (up to 1.7% on incr-patched: b9b3e592dd cherry picked builds of style-servo)

Rollup of 7 pull requests #92062

  • Large improvement in instruction counts (up to -4.7% on incr-patched: println builds of regression-31157)
  • Small regression in instruction counts (up to 1.0% on full builds of webrender)

Rollup of 7 pull requests #92064

  • Large improvement in instruction counts (up to -1.2% on full builds of keccak)
  • Moderate regression in instruction counts (up to 3.5% on incr-patched: println builds of regression-31157)

Avoid sorting in hash map stable hashing #91837

  • Large improvement in instruction counts (up to -4.9% on incr-full builds of clap-rs)
  • Moderate regression in instruction counts (up to 2.8% on incr-patched: println builds of regression-31157)

rustdoc: Add UrlPartsBuilder #91871

  • Moderate improvement in instruction counts (up to -3.6% on incr-patched: println builds of regression-31157)
  • Small regression in instruction counts (up to 0.4% on incr-unchanged builds of helloworld)

Remove SymbolStr #91957

  • Small improvement in instruction counts (up to -0.2% on full builds of externs)
  • Moderate regression in instruction counts (up to 3.1% on incr-patched: println builds of regression-31157)

Rollup of 4 pull requests #92106

  • Small improvement in instruction counts (up to -0.3% on incr-unchanged builds of helloworld)
  • Moderate regression in instruction counts (up to 0.8% on incr-unchanged builds of externs)

Eliminate ObligationCauseData #91844

rustdoc: make --passes and --no-defaults have no effect #91900

  • Moderate improvement in instruction counts (up to -0.8% on incr-unchanged builds of externs)
  • Small regression in instruction counts (up to 1.1% on incr-patched: println builds of regression-31157)

Remove ‘speculative evaluation’ of predicates #92041

  • Very large improvement in instruction counts (up to -5.8% on incr-unchanged builds of regression-31157)
  • Very large regression in instruction counts (up to 8.8% on incr-unchanged builds of deeply-nested)

Fully serialize AdtDef #91924

  • Very large improvement in instruction counts (up to -25.0% on incr-unchanged builds of issue-46449)
  • Very large regression in instruction counts (up to 15.8% on incr-full builds of webrender-wrench)

Nominated Issues

T-compiler

  • No I-compiler-nominated issues at this time

RFC

  • No nominated RFCs for T-compiler this time.

🎄 :)