perf triage for 2024-08-29 -- better late than never.

Author: pnkfelix

triage/2024-08-29.md

@@ -1,8 +1,8 @@
 # 2024-08-29 Triage Log
 
-TODO: Summary
+Not a bad week. More improvements than regressions.
 
-Triage done by **@???**.
+Triage done by **@pnkfelix**.
 Revision range: [4fe1e2bd..acb4e8b6](https://perf.rust-lang.org/?start=4fe1e2bd5bf5a6f1cb245f161a5e9d315766f103&end=acb4e8b6251f1d8da36f08e7a70fa23fc581839e&absolute=false&stat=instructions%3Au)
 
 **Summary**:
@@ -31,6 +31,9 @@ Update stdarch submodule [#128866](https://github.com/rust-lang/rust/pull/128866
 | Improvements ✅ <br /> (secondary) | -    | -            | 0     |
 | All ❌✅ (primary)                 | 0.3% | [0.2%, 0.5%] | 2     |
 
+* primary regressions are to cratelift-codegen llvm-opt-full and helloworld llvm-doc-full
+* from the graphs, the cranelift regression looks like a blip, and the helloworld regression seems tiny compared to other changes in its history -- I infer the auto-computed significance threshold is quite low here, probably because historically the build time for helloworld is very stable?
+* In any case, doesn't seem worth investigating. Marking as triaged.
 
 Rollup of 9 pull requests [#129365](https://github.com/rust-lang/rust/pull/129365) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=a32d4a0e822a29a6682e08b75a8df4c29c7fa9f1&end=739b1fdb158a3216d1b592d0d79d77d256f59815&stat=instructions:u)
 
@@ -42,6 +45,10 @@ Rollup of 9 pull requests [#129365](https://github.com/rust-lang/rust/pull/12936
 | Improvements ✅ <br /> (secondary) | -    | -            | 0     |
 | All ❌✅ (primary)                 | 0.3% | [0.2%, 0.4%] | 15    |
 
+* primary regressions are solely to various incr-patched scenarios
+* confirmed that PR #127279 is root cause
+* note that PR #127279 is addressing an issue in incr-comp; it might be expected that it would cause some incr-comp benchmarks to be impacted.
+* left some comments and marked as triaged.
 
 Rollup of 9 pull requests [#129521](https://github.com/rust-lang/rust/pull/129521) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=697d9530d78aeba31458cadfc338adb2aea90bee&end=717aec0f8eebdc5eb51756e5129dde15e8b25710&stat=instructions:u)
 
@@ -53,6 +60,11 @@ Rollup of 9 pull requests [#129521](https://github.com/rust-lang/rust/pull/12952
 | Improvements ✅ <br /> (secondary) | -    | -            | 0     |
 | All ❌✅ (primary)                 | 0.7% | [0.4%, 0.9%] | 6     |
 
+* primary regressions are solely to various incremental scenarios for the bitmaps crate
+* PR #129513 and PR #129546 seem to have recovered much of the losses here... but after review, they are probably just masking the loss introduced by PR #129521
+* opened [discussion in zulip](https://rust-lang.zulipchat.com/#narrow/stream/247081-t-compiler.2Fperformance/topic/numeric.20precision.20in.20rustc-perf.20detailed.20results/near/466074054) about how to actually dig more deeply into regressions like this.
+* nothing here seems like it should have impacted incr-comp, and this regression seems minor enough that it does not warrant deeper investigation
+* marking as triaged
 
 Rollup of 9 pull requests [#129632](https://github.com/rust-lang/rust/pull/129632) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=515395af0efdbdd657ff08a1f6d28e553856654f&end=bf662eb80838008acabc307dd64d84935ce3a20d&stat=instructions:u)
 
@@ -64,6 +76,10 @@ Rollup of 9 pull requests [#129632](https://github.com/rust-lang/rust/pull/12963
 | Improvements ✅ <br /> (secondary) | -    | -            | 0     |
 | All ❌✅ (primary)                 | 1.4% | [0.6%, 2.3%] | 2     |
 
+* primary doc-full regressions to typenum and stm32f4
+* the regression for typenum's walltime is flagged as significant as well.
+* the self-profile results for typenum imply that we are spending more time in create_renderer
+* enqueued a rust-timer run against PR #192560 which is the most obvious candidate as a rustdoc change
 
 Implement a first version of RFC 3525: struct target features [#127537](https://github.com/rust-lang/rust/pull/127537) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=100fde5246bf56f22fb5cc85374dd841296fce0e&end=acb4e8b6251f1d8da36f08e7a70fa23fc581839e&stat=instructions:u)
 
@@ -75,6 +91,8 @@ Implement a first version of RFC 3525: struct target features [#127537](https://
 | Improvements ✅ <br /> (secondary) | -    | -            | 0     |
 | All ❌✅ (primary)                 | 0.4% | [0.2%, 1.0%] | 72    |
 
+* this is being actively investigated by the PR author after getting feedback from @nnethercote
+* not marking as triaged
 
 #### Improvements
 
@@ -123,6 +141,8 @@ Rollup of 6 pull requests [#129275](https://github.com/rust-lang/rust/pull/12927
 | Improvements ✅ <br /> (secondary) | -1.0% | [-1.0%, -1.0%] | 1     |
 | All ❌✅ (primary)                 | 0.2%  | [0.2%, 0.2%]   | 1     |
 
+* sole primary regression is to helloworld doc-full.
+* marked as triaged.
 
 Don't generate functions with the `rustc_intrinsic_must_be_overridden` attribute [#129239](https://github.com/rust-lang/rust/pull/129239) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=a971212545766fdfe0dd68e5d968133f79944a19&end=4d5b3b196284aded6ae99d12bcf149ffdc8ef379&stat=instructions:u)
 
@@ -134,6 +154,9 @@ Don't generate functions with the `rustc_intrinsic_must_be_overridden` attribute
 | Improvements ✅ <br /> (secondary) | -1.3% | [-2.2%, -0.6%] | 6     |
 | All ❌✅ (primary)                 | -0.1% | [-0.3%, 0.2%]  | 3     |
 
+* sole primary regression is to html5ever opt-full
+* from inspecting the PR, its hard to imagine doing this in a better way, and Isuspect the reported regression is noise.
+* marked as triaged
 
 Add a precondition check for Layout::from_size_align_unchecked [#126556](https://github.com/rust-lang/rust/pull/126556) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=59a74db37df8628f778faf75faa41958bb07a42a&end=982c6f8721416431ec62bb0b9105c0578a9fc603&stat=instructions:u)
 
@@ -146,40 +169,5 @@ Add a precondition check for Layout::from_size_align_unchecked [#126556](https:/
 | All ❌✅ (primary)                 | -0.3% | [-0.5%, -0.2%] | 9     |
 
 
-#### Untriaged Pull Requests
-
-- [#129632 Rollup of 9 pull requests](https://github.com/rust-lang/rust/pull/129632)
-- [#129521 Rollup of 9 pull requests](https://github.com/rust-lang/rust/pull/129521)
-- [#129365 Rollup of 9 pull requests](https://github.com/rust-lang/rust/pull/129365)
-- [#129275 Rollup of 6 pull requests](https://github.com/rust-lang/rust/pull/129275)
-- [#129239 Don't generate functions with the `rustc_intrinsic_must_be_overridden` attribute](https://github.com/rust-lang/rust/pull/129239)
-- [#128936 Support reading thin archives in ArArchiveBuilder](https://github.com/rust-lang/rust/pull/128936)
-- [#128866 Update stdarch submodule](https://github.com/rust-lang/rust/pull/128866)
-- [#128443 Properly mark loop as diverging if it has no breaks](https://github.com/rust-lang/rust/pull/128443)
-- [#128357 Detect non-lifetime binder params shadowing item params](https://github.com/rust-lang/rust/pull/128357)
-- [#128301 Rollup of 3 pull requests](https://github.com/rust-lang/rust/pull/128301)
-- [#128169 Rollup of 5 pull requests](https://github.com/rust-lang/rust/pull/128169)
-- [#128155 Rollup of 8 pull requests](https://github.com/rust-lang/rust/pull/128155)
-- [#128104 Not lint pub structs without pub constructors intentionally](https://github.com/rust-lang/rust/pull/128104)
-- [#127998 Rollup of 9 pull requests](https://github.com/rust-lang/rust/pull/127998)
-- [#127865 Rollup of 8 pull requests](https://github.com/rust-lang/rust/pull/127865)
-- [#127537 Implement a first version of RFC 3525: struct target features](https://github.com/rust-lang/rust/pull/127537)
-- [#127528 Replace ASCII control chars with Unicode Control Pictures](https://github.com/rust-lang/rust/pull/127528)
-- [#127486 Rollup of 5 pull requests](https://github.com/rust-lang/rust/pull/127486)
-- [#127439 Uplift elaboration into `rustc_type_ir`](https://github.com/rust-lang/rust/pull/127439)
-- [#127172 Make `can_eq` process obligations (almost) everywhere](https://github.com/rust-lang/rust/pull/127172)
-- [#127096 Rollup of 11 pull requests](https://github.com/rust-lang/rust/pull/127096)
-- [#127076 Rollup of 6 pull requests](https://github.com/rust-lang/rust/pull/127076)
-- [#126852 Also get `add nuw` from `uN::checked_add`](https://github.com/rust-lang/rust/pull/126852)
-- [#126591 [perf] More span update benchmarking](https://github.com/rust-lang/rust/pull/126591)
-- [#126578 Account for things that optimize out in inlining costs](https://github.com/rust-lang/rust/pull/126578)
-- [#126134 Rollup of 11 pull requests](https://github.com/rust-lang/rust/pull/126134)
-- [#126024 Do not use global caches if opaque types can be defined](https://github.com/rust-lang/rust/pull/126024)
-- [#125968 Store the types of `ty::Expr` arguments in the `ty::Expr`](https://github.com/rust-lang/rust/pull/125968)
-- [#125824 Rollup of 7 pull requests](https://github.com/rust-lang/rust/pull/125824)
-- [#125144 Rollup of 6 pull requests](https://github.com/rust-lang/rust/pull/125144)
-
-#### Nags requiring follow up
-
-TODO: Nags
-
+* this PR went through significant microoptimization during develoment and review
+* also, already marked as triaged