Skip to content

Commit e34d759

Browse files
JonathanBrouwerJonathanBrouwer
committed
Add perf triage for 2026-08-09
1 parent 8501598 commit e34d759

1 file changed

Lines changed: 219 additions & 0 deletions

File tree

‎triage/2026/2026-08-09.md‎

Lines changed: 219 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,219 @@
1+
# 2026-08-09 Triage Log
2+
3+
This week so many new performance improvements landed that we needed to roll 10 of them up together to keep the bors queue manageable, great work!
4+
Also new is the LLVM 23 update which caused massive compile-time, run-time, bootstrap time and artifact size improvements.
5+
These improvements are reduced by the merge of Polonius Alpha on nightly causing a 3.0% regression. There still seems to be some potential to mitigate this regression.
6+
7+
Triage done by **@JonathanBrouwer**.
8+
Revision range: [65dd30fb..771916f9](https://perf.rust-lang.org/?start=65dd30fb9e882a7e8f0be10caca62936db2a98b8&end=771916f9028e7fe56d2685f2c4f698de5d7d6a45&absolute=false&stat=instructions%3Au)
9+
10+
**Summary**:
11+
12+
| (instructions:u) | mean | range | count |
13+
|:----------------------------------:|:-----:|:---------------:|:-----:|
14+
| Regressions ❌ <br /> (primary) | 4.6% | [0.2%, 12.8%] | 24 |
15+
| Regressions ❌ <br /> (secondary) | 4.3% | [0.2%, 14.3%] | 30 |
16+
| Improvements ✅ <br /> (primary) | -3.3% | [-16.4%, -0.2%] | 251 |
17+
| Improvements ✅ <br /> (secondary) | -5.2% | [-34.8%, -0.2%] | 308 |
18+
| All ❌✅ (primary) | -2.6% | [-16.4%, 12.8%] | 275 |
19+
20+
21+
1 Regression, 4 Improvements, 7 Mixed; 5 of them in rollups
22+
25 artifact comparisons made in total
23+
24+
#### Regressions
25+
26+
Rollup of 22 pull requests [#160517](https://github.com/rust-lang/rust/pull/160517) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=1ed2df61a19042f231709eb05d032ae9e2cb2084&end=7218ebe93668f51a94a572b690c433dfdbdc2c3d&stat=instructions:u)
27+
28+
| (instructions:u) | mean | range | count |
29+
|:----------------------------------:|:-----:|:--------------:|:-----:|
30+
| Regressions ❌ <br /> (primary) | 0.6% | [0.2%, 1.9%] | 11 |
31+
| Regressions ❌ <br /> (secondary) | 0.5% | [0.2%, 1.8%] | 13 |
32+
| Improvements ✅ <br /> (primary) | -0.8% | [-0.8%, -0.8%] | 1 |
33+
| Improvements ✅ <br /> (secondary) | - | - | 0 |
34+
| All ❌✅ (primary) | 0.5% | [-0.8%, 1.9%] | 12 |
35+
36+
Most of this regression is noise, showing up on all PRs that were part of this rollup identically.
37+
The one exception is [#160480](https://github.com/rust-lang/rust/pull/160480), which changes the implementation of `to_ascii_uppercase` and `to_ascii_lowercase` to be faster.
38+
While the run-time performance of these functions improve, they are now slower to compile because they use more generics. Why this effect is so significant is unclear.
39+
40+
#### Improvements
41+
42+
perf: Lock-free root fast paths for hygiene queries [#160494](https://github.com/rust-lang/rust/pull/160494) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=26bfbf28f5f79dff333ce5a232f131b496ec9cfe&end=f73951df0a5566d94d13b7954acd9f4ab1fa3734&stat=instructions:u)
43+
44+
| (instructions:u) | mean | range | count |
45+
|:----------------------------------:|:-----:|:--------------:|:-----:|
46+
| Regressions ❌ <br /> (primary) | - | - | 0 |
47+
| Regressions ❌ <br /> (secondary) | - | - | 0 |
48+
| Improvements ✅ <br /> (primary) | -0.3% | [-0.7%, -0.2%] | 61 |
49+
| Improvements ✅ <br /> (secondary) | -0.8% | [-2.3%, -0.1%] | 42 |
50+
| All ❌✅ (primary) | -0.3% | [-0.7%, -0.2%] | 61 |
51+
52+
53+
Optimize handling of solver errors [#160160](https://github.com/rust-lang/rust/pull/160160) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=88f7399cb4912680976e1c342ea7661b9dc84940&end=ae45457594a670c59cd4d5591eaa243d9a3d44d5&stat=instructions:u)
54+
55+
| (instructions:u) | mean | range | count |
56+
|:----------------------------------:|:-----:|:--------------:|:-----:|
57+
| Regressions ❌ <br /> (primary) | - | - | 0 |
58+
| Regressions ❌ <br /> (secondary) | 0.6% | [0.5%, 0.6%] | 3 |
59+
| Improvements ✅ <br /> (primary) | -0.3% | [-0.6%, -0.2%] | 27 |
60+
| Improvements ✅ <br /> (secondary) | -0.6% | [-2.7%, -0.3%] | 36 |
61+
| All ❌✅ (primary) | -0.3% | [-0.6%, -0.2%] | 27 |
62+
63+
64+
Optimize `try_evaluate_obligations` [#160479](https://github.com/rust-lang/rust/pull/160479) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=ae45457594a670c59cd4d5591eaa243d9a3d44d5&end=65bcac45b3d8a8b2126e5cc844cf6fff5795d32a&stat=instructions:u)
65+
66+
| (instructions:u) | mean | range | count |
67+
|:----------------------------------:|:-----:|:--------------:|:-----:|
68+
| Regressions ❌ <br /> (primary) | - | - | 0 |
69+
| Regressions ❌ <br /> (secondary) | - | - | 0 |
70+
| Improvements ✅ <br /> (primary) | - | - | 0 |
71+
| Improvements ✅ <br /> (secondary) | -1.6% | [-5.1%, -0.2%] | 20 |
72+
| All ❌✅ (primary) | - | - | 0 |
73+
74+
75+
[perf] Reuse existing trait reference instead of recreating it. [#160539](https://github.com/rust-lang/rust/pull/160539) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=7dfb9d0c40d3338e1a27f0f787182c6b90e6791e&end=49c80b5a0db8abd63315c04ff34f58629e388281&stat=instructions:u)
76+
77+
| (instructions:u) | mean | range | count |
78+
|:----------------------------------:|:-----:|:--------------:|:-----:|
79+
| Regressions ❌ <br /> (primary) | - | - | 0 |
80+
| Regressions ❌ <br /> (secondary) | - | - | 0 |
81+
| Improvements ✅ <br /> (primary) | - | - | 0 |
82+
| Improvements ✅ <br /> (secondary) | -4.4% | [-7.5%, -0.3%] | 7 |
83+
| All ❌✅ (primary) | - | - | 0 |
84+
85+
86+
#### Mixed
87+
88+
Rollup of 10 perf-sensitive pull requests [#160506](https://github.com/rust-lang/rust/pull/160506) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=0b63defc8fd68957c635fe61f097b84df8fd9611&end=1ed2df61a19042f231709eb05d032ae9e2cb2084&stat=instructions:u)
89+
90+
| (instructions:u) | mean | range | count |
91+
|:----------------------------------:|:-----:|:---------------:|:-----:|
92+
| Regressions ❌ <br /> (primary) | - | - | 0 |
93+
| Regressions ❌ <br /> (secondary) | 0.5% | [0.5%, 0.5%] | 1 |
94+
| Improvements ✅ <br /> (primary) | -1.7% | [-6.2%, -0.2%] | 237 |
95+
| Improvements ✅ <br /> (secondary) | -4.7% | [-34.7%, -0.1%] | 230 |
96+
| All ❌✅ (primary) | -1.7% | [-6.2%, -0.2%] | 237 |
97+
98+
This is a rollup of 10 perf-sensitive pull requestions, this was made for the first time because the queue was too large.
99+
After the rollup was merged, every PR in the rollup was individually benchmarked.
100+
The following 10 PRs were rolled up: (below each PR the effect on benchmarks is listed)
101+
102+
* [perf: skip irrelevant foreign impls when building the specialization graph - #157281](https://github.com/rust-lang/rust/pull/157281#issuecomment-5188979580)
103+
104+
-1.5% [-6.1%, -0.2%] 209
105+
* [Next steps for FnDef binder changes](https://github.com/rust-lang/rust/pull/159403#issuecomment-5189362799)
106+
107+
0.3% [-0.2%, 0.4%] 19
108+
109+
This change is a part of the ongoing effort to clean up binder/turbofishing behavior in the compiler.
110+
While the perf regression was known before merging, it wasn't explained by the author. Author has been pinged.
111+
* [Optimize crate resolution for large workspace](https://github.com/rust-lang/rust/pull/159763#issuecomment-5189793061)
112+
113+
-7.5% [-15.3%, -1.7%] 11 (secondary, no effect on primaries)
114+
* [Speed up EverInitializedPlaces](https://github.com/rust-lang/rust/pull/160033#issuecomment-5190192416)
115+
116+
-0.4% [-0.6%, -0.2%] 10
117+
* [store the fulfillment engine inline in ObligationCtxt](https://github.com/rust-lang/rust/pull/160268#issuecomment-5190941859)
118+
119+
-0.5% [-1.4%, -0.2%] 71
120+
* [Cache already-checked types in the privacy visitor](https://github.com/rust-lang/rust/pull/160317#issuecomment-5191318930)
121+
122+
-0.5% [-1.1%, -0.2%] 57
123+
* [interpret: skip deref-projection validity checks when they are not needed](https://github.com/rust-lang/rust/pull/160399#issuecomment-5191715368)
124+
125+
-12.8% [-13.9%, -11.9%] 6 (secondary, no effect on primaries)
126+
* [Deduplicate target and host filesearch](https://github.com/rust-lang/rust/pull/160451#issuecomment-5192147316)
127+
128+
-0.7% [-1.4%, -0.3%] 11 (secondary, no effect on primaries)
129+
* [Add fast path to escape_string_symbol](https://github.com/rust-lang/rust/pull/160453#issuecomment-5192623577)
130+
131+
-19.2% [-34.8%, -4.4%] 14 (secondary, no effect on primaries)
132+
* [Add offload guard flags to typeck to prevent perf regressions](https://github.com/rust-lang/rust/pull/160454#issuecomment-5193112704)
133+
134+
-0.4% [-0.5%, -0.2%] 10 (secondary, no effect on primaries)
135+
136+
137+
Update to LLVM 23 [#158734](https://github.com/rust-lang/rust/pull/158734) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=7218ebe93668f51a94a572b690c433dfdbdc2c3d&end=1b29cfdbee8ee37333f4b24afb936350eb57e705&stat=instructions:u)
138+
139+
| (instructions:u) | mean | range | count |
140+
|:----------------------------------:|:-----:|:---------------:|:-----:|
141+
| Regressions ❌ <br /> (primary) | 0.5% | [0.2%, 0.7%] | 14 |
142+
| Regressions ❌ <br /> (secondary) | 1.7% | [0.0%, 9.1%] | 27 |
143+
| Improvements ✅ <br /> (primary) | -2.6% | [-16.8%, -0.2%] | 208 |
144+
| Improvements ✅ <br /> (secondary) | -2.0% | [-11.9%, -0.0%] | 246 |
145+
| All ❌✅ (primary) | -2.4% | [-16.8%, 0.7%] | 222 |
146+
147+
Not only does the LLVM 23 update have an overwhelmingly positive effect on compile-time performance,
148+
it also seriously improves run-time performance (-6.9% on the raytracer benchmark), it reduces bootstrap time by 7.7%, and reduces binary size by 0.2%.
149+
Thanks to everyone who worked on landing this update!
150+
151+
Type fallback refactorings [#159026](https://github.com/rust-lang/rust/pull/159026) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=1b29cfdbee8ee37333f4b24afb936350eb57e705&end=22950ab9e29e11620a10905f027f141ea126966e&stat=instructions:u)
152+
153+
| (instructions:u) | mean | range | count |
154+
|:----------------------------------:|:-----:|:--------------:|:-----:|
155+
| Regressions ❌ <br /> (primary) | - | - | 0 |
156+
| Regressions ❌ <br /> (secondary) | 0.3% | [0.1%, 0.6%] | 12 |
157+
| Improvements ✅ <br /> (primary) | - | - | 0 |
158+
| Improvements ✅ <br /> (secondary) | -2.2% | [-3.1%, -0.1%] | 13 |
159+
| All ❌✅ (primary) | - | - | 0 |
160+
161+
The perf regression is noise, while the improvements are real.
162+
163+
Rollup of 25 pull requests [#160559](https://github.com/rust-lang/rust/pull/160559) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=22950ab9e29e11620a10905f027f141ea126966e&end=87212cef77e7bfa92ba0c4850be1d089745ba6fc&stat=instructions:u)
164+
165+
| (instructions:u) | mean | range | count |
166+
|:----------------------------------:|:-----:|:--------------:|:-----:|
167+
| Regressions ❌ <br /> (primary) | 0.2% | [0.1%, 0.2%] | 2 |
168+
| Regressions ❌ <br /> (secondary) | 0.3% | [0.1%, 0.5%] | 16 |
169+
| Improvements ✅ <br /> (primary) | - | - | 0 |
170+
| Improvements ✅ <br /> (secondary) | -0.4% | [-0.8%, -0.2%] | 13 |
171+
| All ❌✅ (primary) | 0.2% | [0.1%, 0.2%] | 2 |
172+
173+
The performance improvements and some of the regressions are noise.
174+
[The rest of the regressions are caused by changes in rustc-perf.](https://github.com/rust-lang/rust/pull/160510#issuecomment-5216265249)
175+
176+
Enable polonius alpha on nightly [#159343](https://github.com/rust-lang/rust/pull/159343) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=87212cef77e7bfa92ba0c4850be1d089745ba6fc&end=7608eb7b07eaf93f16d7cf5bcb2098eca87503df&stat=instructions:u)
177+
178+
| (instructions:u) | mean | range | count |
179+
|:----------------------------------:|:-----:|:--------------:|:-----:|
180+
| Regressions ❌ <br /> (primary) | 3.0% | [0.2%, 15.7%] | 89 |
181+
| Regressions ❌ <br /> (secondary) | 2.3% | [0.2%, 17.3%] | 108 |
182+
| Improvements ✅ <br /> (primary) | - | - | 0 |
183+
| Improvements ✅ <br /> (secondary) | -0.9% | [-1.3%, -0.3%] | 9 |
184+
| All ❌✅ (primary) | 3.0% | [0.2%, 15.7%] | 89 |
185+
186+
This enables polonius alpha on nightly, which represents years of work on making the borrow checker stronger.
187+
This is not enabled on stable yet, thus the performance regression won't be present there.
188+
Polonius does strictly more work so getting this to zero will be impossible.
189+
Serde is the biggest regression with a 15.6% regression.
190+
191+
While some performance work has already been done on polonius alpha, there seems to be quite some opportunities left.
192+
One of those that @panstromek mentioned is that the `IntervalSet` datatype could be optimized further.
193+
194+
Rollup of 17 pull requests [#160627](https://github.com/rust-lang/rust/pull/160627) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=887804d5686c3bf9a090b5bbc23e3dcdfd9e316a&end=84b36a78a28a63f134171c670be1932ffa2485f8&stat=instructions:u)
195+
196+
| (instructions:u) | mean | range | count |
197+
|:----------------------------------:|:-----:|:--------------:|:-----:|
198+
| Regressions ❌ <br /> (primary) | - | - | 0 |
199+
| Regressions ❌ <br /> (secondary) | 0.4% | [0.2%, 0.5%] | 7 |
200+
| Improvements ✅ <br /> (primary) | -0.3% | [-0.3%, -0.2%] | 5 |
201+
| Improvements ✅ <br /> (secondary) | -0.3% | [-0.3%, -0.2%] | 4 |
202+
| All ❌✅ (primary) | -0.3% | [-0.3%, -0.2%] | 5 |
203+
204+
The new-solver benchmark improvements seem to be noise.
205+
Some improvements caused by https://github.com/rust-lang/rust/pull/160555.
206+
Perf regression caused by https://github.com/rust-lang/rust/pull/160464. Pinged the authors.
207+
208+
Rollup of 28 pull requests [#160725](https://github.com/rust-lang/rust/pull/160725) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=1a98b1e135b254f209c67d447b6d8bcd56a859e0&end=7dfb9d0c40d3338e1a27f0f787182c6b90e6791e&stat=instructions:u)
209+
210+
| (instructions:u) | mean | range | count |
211+
|:----------------------------------:|:-----:|:--------------:|:-----:|
212+
| Regressions ❌ <br /> (primary) | - | - | 0 |
213+
| Regressions ❌ <br /> (secondary) | 0.7% | [0.5%, 0.9%] | 2 |
214+
| Improvements ✅ <br /> (primary) | -0.4% | [-0.5%, -0.3%] | 2 |
215+
| Improvements ✅ <br /> (secondary) | -0.3% | [-0.3%, -0.2%] | 3 |
216+
| All ❌✅ (primary) | -0.4% | [-0.5%, -0.3%] | 2 |
217+
218+
Improvements outweigh regressions.
219+
The regressions may or may not be caused by a separate PR, but are so small that they are not worth hunting down.

0 commit comments

Comments
 (0)