Add perf triage for this week

Author: Mark-Simulacrum

triage/2026/2026-05-17.md

@@ -0,0 +1,88 @@
+# 2026-05-17 Triage Log
+
+Fewer than usual PRs merged, mostly due to a shorter week than normal and some
+CI trouble. Overall a slightly positive week for performance.
+
+Triage done by **@simulacrum**.
+Revision range: [29b75901..281c97c3](https://perf.rust-lang.org/?start=29b7590130c83542a095cdf1323ed0f78eec2bb8&end=281c97c3240a9abd984ca0c6a2cd7389115e80d5&absolute=false&stat=instructions%3Au)
+
+**Summary**:
+
+| (instructions:u)                   | mean  | range          | count |
+|:----------------------------------:|:-----:|:--------------:|:-----:|
+| Regressions ❌ <br /> (primary)    | 0.5%  | [0.2%, 1.5%]   | 6     |
+| Regressions ❌ <br /> (secondary)  | 0.1%  | [0.0%, 0.3%]   | 9     |
+| Improvements ✅ <br /> (primary)   | -0.4% | [-1.3%, -0.2%] | 30    |
+| Improvements ✅ <br /> (secondary) | -0.5% | [-1.4%, -0.0%] | 27    |
+| All ❌✅ (primary)                 | -0.3% | [-1.3%, 1.5%]  | 36    |
+
+
+0 Regressions, 0 Improvements, 4 Mixed; 1 of them in rollups
+17 artifact comparisons made in total
+
+#### Regressions
+
+No strict regressions this (short) week.
+
+#### Improvements
+
+No strict improvements this (short) week.
+
+#### Mixed
+
+Rollup of 6 pull requests [#156506](https://github.com/rust-lang/rust/pull/156506) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=2aabf3ce0569dd7c8c2d2a27944850056dcf2566&end=8b03437a8ffc8f8b01e62ef5fce82a37ada09b12&stat=instructions:u)
+
+| (instructions:u)                   | mean  | range          | count |
+|:----------------------------------:|:-----:|:--------------:|:-----:|
+| Regressions ❌ <br /> (primary)    | 1.4%  | [1.4%, 1.4%]   | 1     |
+| Regressions ❌ <br /> (secondary)  | 0.1%  | [0.0%, 0.1%]   | 3     |
+| Improvements ✅ <br /> (primary)   | -     | -              | 0     |
+| Improvements ✅ <br /> (secondary) | -0.6% | [-1.0%, -0.1%] | 9     |
+| All ❌✅ (primary)                 | 1.4%  | [1.4%, 1.4%]   | 1     |
+
+Kicked off perf builds of two of the contained PRs. It seems possible the
+regression is due to `#[inline]` additions in the `core::io` movement PRs,
+though.
+
+Split the node_id_to_def_id table into a per-owner table [#138995](https://github.com/rust-lang/rust/pull/138995) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=800892799d7666fe1dc17abd862100a6cf273718&end=ff9a9ea07bdc74e9555126464b02be8ff277f521&stat=instructions:u)
+
+| (instructions:u)                   | mean  | range          | count |
+|:----------------------------------:|:-----:|:--------------:|:-----:|
+| Regressions ❌ <br /> (primary)    | 0.3%  | [0.2%, 0.3%]   | 2     |
+| Regressions ❌ <br /> (secondary)  | 0.4%  | [0.1%, 0.5%]   | 5     |
+| Improvements ✅ <br /> (primary)   | -0.3% | [-0.4%, -0.2%] | 8     |
+| Improvements ✅ <br /> (secondary) | -0.1% | [-0.3%, -0.0%] | 5     |
+| All ❌✅ (primary)                 | -0.1% | [-0.4%, 0.3%]  | 10    |
+
+Quite a bit of effort went into making this as neutral as possible, with the
+final status deemed warranted given future benefits. See also [this
+comment](https://github.com/rust-lang/rust/pull/138995#issuecomment-4406548798).
+
+DestinationPropagation: compute liveness as ranges instead of traveling bitsets [#156218](https://github.com/rust-lang/rust/pull/156218) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=88ba7fbe0a6eda36e0adbfd0482856f3784031b9&end=d7f14d3d89c39b2c12ed4b9864a0eb2c205d20bc&stat=instructions:u)
+
+| (instructions:u)                   | mean  | range          | count |
+|:----------------------------------:|:-----:|:--------------:|:-----:|
+| Regressions ❌ <br /> (primary)    | 0.1%  | [0.1%, 0.1%]   | 1     |
+| Regressions ❌ <br /> (secondary)  | 0.1%  | [0.1%, 0.1%]   | 3     |
+| Improvements ✅ <br /> (primary)   | -0.5% | [-1.3%, -0.2%] | 18    |
+| Improvements ✅ <br /> (secondary) | -0.4% | [-0.7%, -0.0%] | 16    |
+| All ❌✅ (primary)                 | -0.5% | [-1.3%, 0.1%]  | 19    |
+
+tt-muncher is the only regression, and that too in doc + incr-unchanged builds.
+Not actually sure this code runs in any of those scenarios... so marking
+as triaged, no follow up needed, especially given the improvement overall.
+
+rustdoc: properly support macros with multiple kinds [#152449](https://github.com/rust-lang/rust/pull/152449) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=be326a93323e9b929525b9298f43883ac66b3b2d&end=d3cd04068e406dd961c2fd666c049c3e38930e0a&stat=instructions:u)
+
+| (instructions:u)                   | mean  | range          | count |
+|:----------------------------------:|:-----:|:--------------:|:-----:|
+| Regressions ❌ <br /> (primary)    | 0.3%  | [0.3%, 0.3%]   | 1     |
+| Regressions ❌ <br /> (secondary)  | 0.3%  | [0.3%, 0.3%]   | 1     |
+| Improvements ✅ <br /> (primary)   | -     | -              | 0     |
+| Improvements ✅ <br /> (secondary) | -0.1% | [-0.1%, -0.1%] | 1     |
+| All ❌✅ (primary)                 | 0.3%  | [0.3%, 0.3%]   | 1     |
+
+> The impact is expected since we added some loops for iterating types for an
+> item (even though 99.9% of the time, there is only one type).
+
+(https://github.com/rust-lang/rust/pull/152449#issuecomment-4468468588)