|
| 1 | +# 2024-09-16 Triage Log |
| 2 | + |
| 3 | +A relatively quiet week, with overall neutral performance across our set of key |
| 4 | +metrics (instructions, cycles, memory). |
| 5 | + |
| 6 | +Triage done by **@simulacrum**. |
| 7 | +Revision range: [263a3aee..170d6cb8](https://perf.rust-lang.org/?start=263a3aeeb8f2d0e9cc85eee61774d1f5f23dc3f5&end=170d6cb845c8c3f0dcec5cdd4210df9ecf990244&absolute=false&stat=instructions%3Au) |
| 8 | + |
| 9 | +**Summary**: |
| 10 | + |
| 11 | +| (instructions:u) | mean | range | count | |
| 12 | +|:----------------------------------:|:-----:|:--------------:|:-----:| |
| 13 | +| Regressions ❌ <br /> (primary) | 0.7% | [0.2%, 4.0%] | 52 | |
| 14 | +| Regressions ❌ <br /> (secondary) | 0.7% | [0.2%, 1.4%] | 48 | |
| 15 | +| Improvements ✅ <br /> (primary) | -0.8% | [-2.6%, -0.2%] | 46 | |
| 16 | +| Improvements ✅ <br /> (secondary) | -0.5% | [-1.1%, -0.1%] | 25 | |
| 17 | +| All ❌✅ (primary) | 0.0% | [-2.6%, 4.0%] | 98 | |
| 18 | + |
| 19 | + |
| 20 | +2 Regressions, 3 Improvements, 4 Mixed; 2 of them in rollups |
| 21 | +54 artifact comparisons made in total |
| 22 | + |
| 23 | +#### Regressions |
| 24 | + |
| 25 | +interpret: make typed copies lossy wrt provenance and padding [#129778](https://github.com/rust-lang/rust/pull/129778) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=712463de61c65033a6f333f0a14fbb65e34efc50&end=304b7f801bab31233680879ca4fb6eb294706a59&stat=instructions:u) |
| 26 | + |
| 27 | +| (instructions:u) | mean | range | count | |
| 28 | +|:----------------------------------:|:----:|:------------:|:-----:| |
| 29 | +| Regressions ❌ <br /> (primary) | - | - | 0 | |
| 30 | +| Regressions ❌ <br /> (secondary) | 2.1% | [0.2%, 4.7%] | 13 | |
| 31 | +| Improvements ✅ <br /> (primary) | - | - | 0 | |
| 32 | +| Improvements ✅ <br /> (secondary) | - | - | 0 | |
| 33 | +| All ❌✅ (primary) | - | - | 0 | |
| 34 | + |
| 35 | +Most of the performance is recovered in https://github.com/rust-lang/rust/pull/130197. |
| 36 | + |
| 37 | +generalize: track relevant info in cache key [#130194](https://github.com/rust-lang/rust/pull/130194) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=4c5fc2c334f1cc305eea79ed0b8fdb276342d101&end=5a2dd7d4f3210629e65879aeecbe643ba3b86bb4&stat=instructions:u) |
| 38 | + |
| 39 | +| (instructions:u) | mean | range | count | |
| 40 | +|:----------------------------------:|:-----:|:--------------:|:-----:| |
| 41 | +| Regressions ❌ <br /> (primary) | - | - | 0 | |
| 42 | +| Regressions ❌ <br /> (secondary) | 1.0% | [0.4%, 1.3%] | 16 | |
| 43 | +| Improvements ✅ <br /> (primary) | - | - | 0 | |
| 44 | +| Improvements ✅ <br /> (secondary) | -0.2% | [-0.2%, -0.2%] | 1 | |
| 45 | +| All ❌✅ (primary) | - | - | 0 | |
| 46 | + |
| 47 | +Soundness fix and regressions only in secondary benchmarks. |
| 48 | + |
| 49 | +#### Improvements |
| 50 | + |
| 51 | +rustdoc: use strategic boxing to shrink `clean::Item` [#129789](https://github.com/rust-lang/rust/pull/129789) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=c2f74c3f928aeb503f15b4e9ef5778e77f3058b8&end=712463de61c65033a6f333f0a14fbb65e34efc50&stat=instructions:u) |
| 52 | + |
| 53 | +| (instructions:u) | mean | range | count | |
| 54 | +|:----------------------------------:|:-----:|:--------------:|:-----:| |
| 55 | +| Regressions ❌ <br /> (primary) | - | - | 0 | |
| 56 | +| Regressions ❌ <br /> (secondary) | - | - | 0 | |
| 57 | +| Improvements ✅ <br /> (primary) | -0.5% | [-1.0%, -0.2%] | 15 | |
| 58 | +| Improvements ✅ <br /> (secondary) | -0.5% | [-0.6%, -0.3%] | 19 | |
| 59 | +| All ❌✅ (primary) | -0.5% | [-1.0%, -0.2%] | 15 | |
| 60 | + |
| 61 | + |
| 62 | +rustdoc: unify the short-circuit on all lints [#129975](https://github.com/rust-lang/rust/pull/129975) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=6f7229c4da0471f1470bb1f86071848cba3a23d9&end=a9fb00bfa4b3038c855b2097b54e05e8c198c183&stat=instructions:u) |
| 63 | + |
| 64 | +| (instructions:u) | mean | range | count | |
| 65 | +|:----------------------------------:|:-----:|:--------------:|:-----:| |
| 66 | +| Regressions ❌ <br /> (primary) | - | - | 0 | |
| 67 | +| Regressions ❌ <br /> (secondary) | - | - | 0 | |
| 68 | +| Improvements ✅ <br /> (primary) | -0.7% | [-2.0%, -0.2%] | 10 | |
| 69 | +| Improvements ✅ <br /> (secondary) | -0.4% | [-0.4%, -0.4%] | 1 | |
| 70 | +| All ❌✅ (primary) | -0.7% | [-2.0%, -0.2%] | 10 | |
| 71 | + |
| 72 | + |
| 73 | +Rollup of 10 pull requests [#130253](https://github.com/rust-lang/rust/pull/130253) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=8d6b88b168e45ee1624699c19443c49665322a91&end=3afb2bb76ce56e860e45d8858ef5b5f00059e1b1&stat=instructions:u) |
| 74 | + |
| 75 | +| (instructions:u) | mean | range | count | |
| 76 | +|:----------------------------------:|:-----:|:--------------:|:-----:| |
| 77 | +| Regressions ❌ <br /> (primary) | 0.2% | [0.2%, 0.2%] | 1 | |
| 78 | +| Regressions ❌ <br /> (secondary) | - | - | 0 | |
| 79 | +| Improvements ✅ <br /> (primary) | -0.8% | [-1.5%, -0.2%] | 13 | |
| 80 | +| Improvements ✅ <br /> (secondary) | - | - | 0 | |
| 81 | +| All ❌✅ (primary) | -0.7% | [-1.5%, 0.2%] | 14 | |
| 82 | + |
| 83 | + |
| 84 | +#### Mixed |
| 85 | + |
| 86 | +Revert "Stabilize `-Znext-solver=coherence`" [#130249](https://github.com/rust-lang/rust/pull/130249) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=7c7372b6a1b9a8d6b787a3e5a10e4f8325b1fdfc&end=f753bc769b16ca9673f11a4cc06e5cc681efd84e&stat=instructions:u) |
| 87 | + |
| 88 | +| (instructions:u) | mean | range | count | |
| 89 | +|:----------------------------------:|:-----:|:--------------:|:-----:| |
| 90 | +| Regressions ❌ <br /> (primary) | 3.7% | [2.9%, 4.5%] | 7 | |
| 91 | +| Regressions ❌ <br /> (secondary) | - | - | 0 | |
| 92 | +| Improvements ✅ <br /> (primary) | -0.5% | [-0.9%, -0.2%] | 13 | |
| 93 | +| Improvements ✅ <br /> (secondary) | -0.5% | [-0.7%, -0.1%] | 9 | |
| 94 | +| All ❌✅ (primary) | 1.0% | [-0.9%, 4.5%] | 20 | |
| 95 | + |
| 96 | +Largely resets to the previous state on a couple benchmarks. Expected given the revert. |
| 97 | + |
| 98 | +Rollup of 5 pull requests [#130281](https://github.com/rust-lang/rust/pull/130281) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=8c0ec05f7dc9ef185eea9f36c5638e74e43935bf&end=2e8db5e9e39c2bf7729113b3041ef4011d90ac5a&stat=instructions:u) |
| 99 | + |
| 100 | +| (instructions:u) | mean | range | count | |
| 101 | +|:----------------------------------:|:-----:|:--------------:|:-----:| |
| 102 | +| Regressions ❌ <br /> (primary) | 0.6% | [0.6%, 0.6%] | 1 | |
| 103 | +| Regressions ❌ <br /> (secondary) | 0.2% | [0.2%, 0.2%] | 1 | |
| 104 | +| Improvements ✅ <br /> (primary) | -0.2% | [-0.2%, -0.2%] | 1 | |
| 105 | +| Improvements ✅ <br /> (secondary) | -1.8% | [-2.0%, -1.6%] | 6 | |
| 106 | +| All ❌✅ (primary) | 0.2% | [-0.2%, 0.6%] | 2 | |
| 107 | + |
| 108 | +Regressions are minor enough that given the rollup and lack of obviously |
| 109 | +relevant PRs I'm just marking as triaged. The primary regressed benchmark is |
| 110 | +webrender and only in full LLVM builds which doesn't seem like it maps well |
| 111 | +onto the contained PRs. |
| 112 | + |
| 113 | +Rescope temp lifetime in if-let into IfElse with migration lint [#107251](https://github.com/rust-lang/rust/pull/107251) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=d3a8524e80244d66968ab4522d31145ccba610f4&end=a5efa01895e82fa705146507d15311b29c715450&stat=instructions:u) |
| 114 | + |
| 115 | +| (instructions:u) | mean | range | count | |
| 116 | +|:----------------------------------:|:-----:|:--------------:|:-----:| |
| 117 | +| Regressions ❌ <br /> (primary) | - | - | 0 | |
| 118 | +| Regressions ❌ <br /> (secondary) | 0.6% | [0.2%, 1.3%] | 19 | |
| 119 | +| Improvements ✅ <br /> (primary) | -0.2% | [-0.2%, -0.2%] | 1 | |
| 120 | +| Improvements ✅ <br /> (secondary) | - | - | 0 | |
| 121 | +| All ❌✅ (primary) | -0.2% | [-0.2%, -0.2%] | 1 | |
| 122 | + |
| 123 | +Seems like the regression is primarily looking like a revert of #127313... |
| 124 | +@cjgillot do you have thoughts on what the delta here might be? I don't see a |
| 125 | +very obvious cause here. Regressions are minor enough that I'm not worried, but |
| 126 | +just checking if there's an easy explanation. |
| 127 | + |
| 128 | +Simplify the canonical clone method and the copy-like forms to copy [#128299](https://github.com/rust-lang/rust/pull/128299) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=02b1be16c65e5716ade771afde7116ebdfbf9b4a&end=e7386b361dda8c2222bee7a1ae79faba7e9a0351&stat=instructions:u) |
| 129 | + |
| 130 | +| (instructions:u) | mean | range | count | |
| 131 | +|:----------------------------------:|:-----:|:--------------:|:-----:| |
| 132 | +| Regressions ❌ <br /> (primary) | 0.4% | [0.2%, 0.8%] | 6 | |
| 133 | +| Regressions ❌ <br /> (secondary) | 0.4% | [0.2%, 2.1%] | 16 | |
| 134 | +| Improvements ✅ <br /> (primary) | -0.8% | [-1.7%, -0.3%] | 12 | |
| 135 | +| Improvements ✅ <br /> (secondary) | - | - | 0 | |
| 136 | +| All ❌✅ (primary) | -0.4% | [-1.7%, 0.8%] | 18 | |
| 137 | + |
| 138 | +Likely expected regressions from direct changes to codegen, possibly giving |
| 139 | +LLVM more freedom to optimize: |
| 140 | + |
| 141 | +> This pass improves performance by reducing the amount of MIR. The generated |
| 142 | +> copies will get merged/expanded/whatever by LLVM, with a completely different |
| 143 | +> heuristic. |
| 144 | +
|
| 145 | +(https://github.com/rust-lang/rust/pull/128299#pullrequestreview-2217057820) |
| 146 | + |
| 147 | + |
0 commit comments