Skip to content

Commit 0aee941

Browse files
KobzolKobzol
authored
Merge pull request #2100 from Mark-Simulacrum/triage
Perf triage for this week
2 parents 17f2551 + 11a8ba7 commit 0aee941

File tree

1 file changed

+153
-0
lines changed

1 file changed

+153
-0
lines changed

‎triage/2025-05-04.md

+153
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,153 @@
1+
# 2025-05-04 Triage Log
2+
3+
A relatively noisy week due to addition of new benchmarks as part of our [2025
4+
benchmark update], and a number of large regressions in a rollup landing late
5+
in the week, with investigation [here](https://github.com/rust-lang/rust/pull/140652).
6+
7+
[2025 benchmark update]: https://github.com/rust-lang/rustc-perf/issues/2024
8+
9+
Triage done by **@simulacrum**.
10+
Revision range: [25cdf1f6..62c5f58f](https://perf.rust-lang.org/?start=25cdf1f67463c9365d8d83778c933ec7480e940b&end=62c5f58f57670ce65e7fec34f8c4ba00c27da2d9&absolute=false&stat=instructions%3Au)
11+
12+
**Summary**:
13+
14+
| (instructions:u) | mean | range | count |
15+
|:----------------------------------:|:-----:|:--------------:|:-----:|
16+
| Regressions ❌ <br /> (primary) | 1.9% | [0.2%, 14.4%] | 47 |
17+
| Regressions ❌ <br /> (secondary) | 2.0% | [0.1%, 11.8%] | 156 |
18+
| Improvements ✅ <br /> (primary) | -0.6% | [-3.4%, -0.2%] | 34 |
19+
| Improvements ✅ <br /> (secondary) | -1.6% | [-3.7%, -0.3%] | 20 |
20+
| All ❌✅ (primary) | 0.9% | [-3.4%, 14.4%] | 81 |
21+
22+
2 Regressions, 2 Improvements, 6 Mixed; 3 of them in rollups
23+
31 artifact comparisons made in total
24+
25+
#### Regressions
26+
27+
Introduce Arena::try_alloc_from_iter. [#140023](https://github.com/rust-lang/rust/pull/140023) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=74509131e85a97353c67c503ea32e148a56cf4bd&end=0fbb922e530399599aab8296ea975cb9e7ed67bf&stat=instructions:u)
28+
29+
| (instructions:u) | mean | range | count |
30+
|:----------------------------------:|:----:|:------------:|:-----:|
31+
| Regressions ❌ <br /> (primary) | 0.3% | [0.3%, 0.4%] | 3 |
32+
| Regressions ❌ <br /> (secondary) | 0.3% | [0.1%, 0.4%] | 15 |
33+
| Improvements ✅ <br /> (primary) | - | - | 0 |
34+
| Improvements ✅ <br /> (secondary) | - | - | 0 |
35+
| All ❌✅ (primary) | 0.3% | [0.3%, 0.4%] | 3 |
36+
37+
Regressions look plausibly real (they've held up for a bit) but given they're
38+
small and limited to just doc builds I'm inclined to accept. It seems plausible
39+
this is just some slight differences in the ABI of the functions here or
40+
similar. Marking as triaged.
41+
42+
Set groundwork for proper const normalization [#140549](https://github.com/rust-lang/rust/pull/140549) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=3559e0ab0e370b98bc3f9e6b688152c47eb9e6c8&end=1bea580f364c65bd5f7380a1056e150df7b8a1a6&stat=instructions:u)
43+
44+
| (instructions:u) | mean | range | count |
45+
|:----------------------------------:|:----:|:------------:|:-----:|
46+
| Regressions ❌ <br /> (primary) | 0.7% | [0.3%, 1.4%] | 3 |
47+
| Regressions ❌ <br /> (secondary) | 0.5% | [0.1%, 0.6%] | 7 |
48+
| Improvements ✅ <br /> (primary) | - | - | 0 |
49+
| Improvements ✅ <br /> (secondary) | - | - | 0 |
50+
| All ❌✅ (primary) | 0.7% | [0.3%, 1.4%] | 3 |
51+
52+
Mostly what looks like inlining noise. See [comment](https://github.com/rust-lang/rust/pull/140549#issuecomment-2849664075)
53+
for more details.
54+
55+
#### Improvements
56+
57+
perf: delay checking of `#[rustc_no_implicit_autorefs]` in autoref lint [#140406](https://github.com/rust-lang/rust/pull/140406) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=7c96085b64580af5c5f619384f0e3f082138ff13&end=4824c2bb7445cb2478aab0190c268c939d77a0f6&stat=instructions:u)
58+
59+
| (instructions:u) | mean | range | count |
60+
|:----------------------------------:|:-----:|:--------------:|:-----:|
61+
| Regressions ❌ <br /> (primary) | - | - | 0 |
62+
| Regressions ❌ <br /> (secondary) | - | - | 0 |
63+
| Improvements ✅ <br /> (primary) | -1.2% | [-3.4%, -0.2%] | 13 |
64+
| Improvements ✅ <br /> (secondary) | -0.4% | [-0.4%, -0.4%] | 1 |
65+
| All ❌✅ (primary) | -1.2% | [-3.4%, -0.2%] | 13 |
66+
67+
Use a closure instead of three chained iterators [#140464](https://github.com/rust-lang/rust/pull/140464) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=5fe04cbebd593c6530a21cd7bd0e8e9503b1ffe4&end=d7df5bdf2986e596aeaeec38e732711c69ebbce1&stat=instructions:u)
68+
69+
| (instructions:u) | mean | range | count |
70+
|:----------------------------------:|:-----:|:--------------:|:-----:|
71+
| Regressions ❌ <br /> (primary) | - | - | 0 |
72+
| Regressions ❌ <br /> (secondary) | - | - | 0 |
73+
| Improvements ✅ <br /> (primary) | -0.3% | [-0.6%, -0.2%] | 13 |
74+
| Improvements ✅ <br /> (secondary) | -0.1% | [-0.1%, -0.1%] | 2 |
75+
| All ❌✅ (primary) | -0.3% | [-0.6%, -0.2%] | 13 |
76+
77+
#### Mixed
78+
79+
Rollup of 7 pull requests [#140474](https://github.com/rust-lang/rust/pull/140474) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=efcbb94dccbe0fcc1076f031f88332db4a6a2a52&end=74509131e85a97353c67c503ea32e148a56cf4bd&stat=instructions:u)
80+
81+
| (instructions:u) | mean | range | count |
82+
|:----------------------------------:|:-----:|:--------------:|:-----:|
83+
| Regressions ❌ <br /> (primary) | 0.3% | [0.3%, 0.3%] | 1 |
84+
| Regressions ❌ <br /> (secondary) | - | - | 0 |
85+
| Improvements ✅ <br /> (primary) | -2.9% | [-2.9%, -2.9%] | 1 |
86+
| Improvements ✅ <br /> (secondary) | -0.2% | [-0.2%, -0.2%] | 2 |
87+
| All ❌✅ (primary) | -1.3% | [-2.9%, 0.3%] | 2 |
88+
89+
Regression is limited to unicode-normalization-0.1.19 and appears within noise
90+
limit for that benchmark. Marking as triaged.
91+
92+
Simplify `LazyAttrTokenStream` [#127516](https://github.com/rust-lang/rust/pull/127516) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=0fbb922e530399599aab8296ea975cb9e7ed67bf&end=f242d6c26cc6fc187257bd1be9590b4b39632425&stat=instructions:u)
93+
94+
| (instructions:u) | mean | range | count |
95+
|:----------------------------------:|:-----:|:--------------:|:-----:|
96+
| Regressions ❌ <br /> (primary) | - | - | 0 |
97+
| Regressions ❌ <br /> (secondary) | 0.6% | [0.3%, 0.8%] | 7 |
98+
| Improvements ✅ <br /> (primary) | - | - | 0 |
99+
| Improvements ✅ <br /> (secondary) | -2.1% | [-4.0%, -1.4%] | 16 |
100+
| All ❌✅ (primary) | - | - | 0 |
101+
102+
> Improvements outweigh regressions, and all the movement is in secondary benchmarks.
103+
104+
Streamline the `format` macro. [#140188](https://github.com/rust-lang/rust/pull/140188) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=f242d6c26cc6fc187257bd1be9590b4b39632425&end=427288b3ce2d574847fdb41cc3184c893750e09a&stat=instructions:u)
105+
106+
| (instructions:u) | mean | range | count |
107+
|:----------------------------------:|:-----:|:--------------:|:-----:|
108+
| Regressions ❌ <br /> (primary) | 0.2% | [0.2%, 0.2%] | 1 |
109+
| Regressions ❌ <br /> (secondary) | 1.8% | [1.7%, 1.9%] | 6 |
110+
| Improvements ✅ <br /> (primary) | -0.3% | [-0.4%, -0.1%] | 10 |
111+
| Improvements ✅ <br /> (secondary) | - | - | 0 |
112+
| All ❌✅ (primary) | -0.2% | [-0.4%, 0.2%] | 11 |
113+
114+
> The improvements on primary benchmarks were deemed to outweigh the regressions on secondary benchmarks.
115+
116+
shared-generics: Do not share instantiations that contain local-only types [#138522](https://github.com/rust-lang/rust/pull/138522) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=6e23095adf9209614a45f7f75fea36dad7b92afb&end=0e517d38ad0e72f93c734b14fabd4bb9b7441de6&stat=instructions:u)
117+
118+
| (instructions:u) | mean | range | count |
119+
|:----------------------------------:|:-----:|:--------------:|:-----:|
120+
| Regressions ❌ <br /> (primary) | 0.9% | [0.4%, 1.5%] | 3 |
121+
| Regressions ❌ <br /> (secondary) | 0.1% | [0.1%, 0.1%] | 1 |
122+
| Improvements ✅ <br /> (primary) | -0.5% | [-0.7%, -0.3%] | 6 |
123+
| Improvements ✅ <br /> (secondary) | -1.0% | [-1.0%, -1.0%] | 1 |
124+
| All ❌✅ (primary) | -0.0% | [-0.7%, 1.5%] | 9 |
125+
126+
> Marking perf as triaged. There's some improvements, some regressions, but the
127+
> regressions look like potentially bimodality -- and where they're possibly
128+
> real hard to justify further digging in. The change seems overall a slight
129+
> improvement on our benchmarks.
130+
131+
Rollup of 9 pull requests [#140596](https://github.com/rust-lang/rust/pull/140596) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=4824c2bb7445cb2478aab0190c268c939d77a0f6&end=2d5ffc513f1c56b7bc95bacb2519705096e8cc2b&stat=instructions:u)
132+
133+
| (instructions:u) | mean | range | count |
134+
|:----------------------------------:|:-----:|:--------------:|:-----:|
135+
| Regressions ❌ <br /> (primary) | - | - | 0 |
136+
| Regressions ❌ <br /> (secondary) | 0.4% | [0.2%, 0.5%] | 2 |
137+
| Improvements ✅ <br /> (primary) | -1.2% | [-1.2%, -1.2%] | 1 |
138+
| Improvements ✅ <br /> (secondary) | - | - | 0 |
139+
| All ❌✅ (primary) | -1.2% | [-1.2%, -1.2%] | 1 |
140+
141+
Regression triaged. Regressions look like bimodality/noise, not worth further follow up.
142+
143+
Rollup of 7 pull requests [#140633](https://github.com/rust-lang/rust/pull/140633) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=1bea580f364c65bd5f7380a1056e150df7b8a1a6&end=622ac043764d5d4ffff8de8cf86a1cc938a8a71b&stat=instructions:u)
144+
145+
| (instructions:u) | mean | range | count |
146+
|:----------------------------------:|:-----:|:--------------:|:-----:|
147+
| Regressions ❌ <br /> (primary) | 1.2% | [0.2%, 13.4%] | 77 |
148+
| Regressions ❌ <br /> (secondary) | 1.9% | [0.2%, 11.0%] | 155 |
149+
| Improvements ✅ <br /> (primary) | -1.5% | [-1.5%, -1.5%] | 1 |
150+
| Improvements ✅ <br /> (secondary) | - | - | 0 |
151+
| All ❌✅ (primary) | 1.2% | [-1.5%, 13.4%] | 78 |
152+
153+
Per investigation on the PR, the root cause is [#139675](https://github.com/rust-lang/rust/pull/139675). Per discussion [here](https://github.com/rust-lang/rust/pull/140652) we expect that this will be mostly resolved in our next LLVM bump, but is likely an acceptable regression in the meantime (primarily to very short compiles) for the feature in question.

0 commit comments

Comments
 (0)