Skip to content

Commit 9460311

Browse files
KobzolKobzol
authored
Merge pull request #1999 from pnkfelix/triage-2024-10-21
Triage 2024 10 21
2 parents 5d1e436 + a773b31 commit 9460311

File tree

1 file changed

+159
-0
lines changed

1 file changed

+159
-0
lines changed

‎triage/2024-10-21.md

Lines changed: 159 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,159 @@
1+
# 2024-10-21 Triage Log
2+
3+
Some tidy improvements from switching to next generation trait solver
4+
(solely for coherence checking) and from simplifying our dataflow
5+
analysis framework. There were some binary size regressions associated
6+
with PR 126557 (adding `#[track_caller]` to allocating methods of
7+
`Vec` and `VecDeque`), which I have handed off to T-libs to choose
8+
whether to investigate further.
9+
10+
Triage done by **@pnkfelix**.
11+
Revision range: [5ceb623a..3e33bda0](https://perf.rust-lang.org/?start=5ceb623a4abd66e91e7959d25caaf0523f1a7f7c&end=3e33bda0326586a6e1e34d0f5c060ca6d116e6a4&absolute=false&stat=instructions%3Au)
12+
13+
**Summary**:
14+
15+
| (instructions:u) | mean | range | count |
16+
|:----------------------------------:|:-----:|:--------------:|:-----:|
17+
| Regressions ❌ <br /> (primary) | 0.4% | [0.2%, 0.9%] | 43 |
18+
| Regressions ❌ <br /> (secondary) | 0.4% | [0.2%, 0.7%] | 36 |
19+
| Improvements ✅ <br /> (primary) | -0.8% | [-5.1%, -0.2%] | 92 |
20+
| Improvements ✅ <br /> (secondary) | -0.8% | [-2.0%, -0.1%] | 65 |
21+
| All ❌✅ (primary) | -0.4% | [-5.1%, 0.9%] | 135 |
22+
23+
24+
0 Regressions, 3 Improvements, 6 Mixed; 3 of them in rollups
25+
47 artifact comparisons made in total
26+
27+
#### Regressions
28+
29+
30+
31+
#### Improvements
32+
33+
stabilize `-Znext-solver=coherence` again [#130654](https://github.com/rust-lang/rust/pull/130654) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=f79fae3069c449993eda6b16934da3b144cb8a66&end=a0c2aba29aa9ea50a7c45c3391dd446f856bef7b&stat=instructions:u)
34+
35+
| (instructions:u) | mean | range | count |
36+
|:----------------------------------:|:-----:|:--------------:|:-----:|
37+
| Regressions ❌ <br /> (primary) | - | - | 0 |
38+
| Regressions ❌ <br /> (secondary) | - | - | 0 |
39+
| Improvements ✅ <br /> (primary) | -1.9% | [-4.5%, -0.2%] | 17 |
40+
| Improvements ✅ <br /> (secondary) | -0.3% | [-0.3%, -0.3%] | 3 |
41+
| All ❌✅ (primary) | -1.9% | [-4.5%, -0.2%] | 17 |
42+
43+
* improvements are to bitmaps and nalgebra
44+
* (presumably due to their heavy use of trait machinery; though, skimming bitmaps, I'm not sure if I see heavy use of trait machinery there)
45+
* what else can I say, except, amazing!
46+
47+
Remove `GenKillAnalysis` [#131481](https://github.com/rust-lang/rust/pull/131481) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=1f67a7aa8d5b30c43c28ed9b2621cf4b7b8bb963&end=d829780c4e4ef11f5e09c1c5ed9684c12aad7236&stat=instructions:u)
48+
49+
| (instructions:u) | mean | range | count |
50+
|:----------------------------------:|:-----:|:--------------:|:-----:|
51+
| Regressions ❌ <br /> (primary) | - | - | 0 |
52+
| Regressions ❌ <br /> (secondary) | 1.6% | [1.6%, 1.6%] | 1 |
53+
| Improvements ✅ <br /> (primary) | -0.4% | [-0.7%, -0.2%] | 24 |
54+
| Improvements ✅ <br /> (secondary) | -0.9% | [-1.9%, -0.1%] | 19 |
55+
| All ❌✅ (primary) | -0.4% | [-0.7%, -0.2%] | 24 |
56+
57+
* improvements are spread across html5ever, serde, cargo, libc, diesel, and bitmaps.
58+
* wow. I'm surprised `GenKillAnalysis` is a pessimization
59+
* (maybe this is a sign that our basic blocks tend to hold a small number of instructions...?)
60+
* anyway, overall amazing work.
61+
62+
Rollup of 8 pull requests [#131792](https://github.com/rust-lang/rust/pull/131792) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=bed75e7c21e8d18bd536a0f7c9e479d2f6707db3&end=7342830c05ec0996e9e4b7df550b1043dca7829c&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) | -0.2% | [-0.2%, -0.2%] | 5 |
69+
| Improvements ✅ <br /> (secondary) | - | - | 0 |
70+
| All ❌✅ (primary) | -0.2% | [-0.2%, -0.2%] | 5 |
71+
72+
* improvements are to incremental scenarios for stm32f4 (and one also for libc).
73+
74+
75+
#### Mixed
76+
77+
Add `#[track_caller]` to allocating methods of `Vec` & `VecDeque` [#126557](https://github.com/rust-lang/rust/pull/126557) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=5ceb623a4abd66e91e7959d25caaf0523f1a7f7c&end=f6648f252a05a0a46c865d7ec836b46290613bf9&stat=instructions:u)
78+
79+
| (instructions:u) | mean | range | count |
80+
|:----------------------------------:|:-----:|:--------------:|:-----:|
81+
| Regressions ❌ <br /> (primary) | 0.3% | [0.0%, 0.5%] | 26 |
82+
| Regressions ❌ <br /> (secondary) | - | - | 0 |
83+
| Improvements ✅ <br /> (primary) | - | - | 0 |
84+
| Improvements ✅ <br /> (secondary) | -0.2% | [-0.2%, -0.1%] | 4 |
85+
| All ❌✅ (primary) | 0.3% | [0.0%, 0.5%] | 26 |
86+
87+
* based on a prior perf run (which predicted 16 primary regressions of roughly the same magnitude as observed here), the T-libs team had [already approved this PR](https://github.com/rust-lang/rust/pull/126557#issuecomment-2329482185) under the assumption that there wouldn't be a *runtime impact* from this.
88+
* there was a note from nnethercote that it isn't totally clear if the binary size increases were anticipated.
89+
* marking as triaged
90+
91+
Rollup of 8 pull requests [#131690](https://github.com/rust-lang/rust/pull/131690) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=17a19e684cdf3ca088af8b4da6a6209d128913f4&end=9322d183f45e0fd5a509820874cc5ff27744a479&stat=instructions:u)
92+
93+
| (instructions:u) | mean | range | count |
94+
|:----------------------------------:|:-----:|:--------------:|:-----:|
95+
| Regressions ❌ <br /> (primary) | 0.4% | [0.4%, 0.4%] | 1 |
96+
| Regressions ❌ <br /> (secondary) | - | - | 0 |
97+
| Improvements ✅ <br /> (primary) | -0.6% | [-0.8%, -0.3%] | 2 |
98+
| Improvements ✅ <br /> (secondary) | - | - | 0 |
99+
| All ❌✅ (primary) | -0.2% | [-0.8%, 0.4%] | 3 |
100+
101+
* regression to cargo opt-full; improvements to image and html5ever opt-full
102+
* briefly skimmed detailed results. nothing stood out.
103+
* not worth digging into further; marking as triaged
104+
105+
Use `ThinVec` for PredicateObligation storage [#131422](https://github.com/rust-lang/rust/pull/131422) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=9ce3675b438aae22ef0c6147cde2003a418ab722&end=9618da7c9995a673af4841149ba2d1f53b69dd92&stat=instructions:u)
106+
107+
| (instructions:u) | mean | range | count |
108+
|:----------------------------------:|:-----:|:--------------:|:-----:|
109+
| Regressions ❌ <br /> (primary) | - | - | 0 |
110+
| Regressions ❌ <br /> (secondary) | 0.3% | [0.3%, 0.5%] | 8 |
111+
| Improvements ✅ <br /> (primary) | -0.3% | [-0.7%, -0.2%] | 15 |
112+
| Improvements ✅ <br /> (secondary) | -0.3% | [-0.7%, -0.2%] | 23 |
113+
| All ❌✅ (primary) | -0.3% | [-0.7%, -0.2%] | 15 |
114+
115+
* improvements outweigh regressions. (and reported performance matches what was anticipated via perf runs.)
116+
* marking as triaged
117+
118+
optimize str.replace [#130223](https://github.com/rust-lang/rust/pull/130223) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=3a85d3fa785d95a7b7bcf4f160b67bffba7afd4a&end=86bd45979a964678b40b79156744f0057759d840&stat=instructions:u)
119+
120+
| (instructions:u) | mean | range | count |
121+
|:----------------------------------:|:-----:|:--------------:|:-----:|
122+
| Regressions ❌ <br /> (primary) | 0.5% | [0.2%, 1.2%] | 5 |
123+
| Regressions ❌ <br /> (secondary) | - | - | 0 |
124+
| Improvements ✅ <br /> (primary) | -0.8% | [-0.8%, -0.8%] | 1 |
125+
| Improvements ✅ <br /> (secondary) | -0.4% | [-0.4%, -0.3%] | 6 |
126+
| All ❌✅ (primary) | 0.3% | [-0.8%, 1.2%] | 6 |
127+
128+
* primary regressions to miscellaneous cargo and clap scenarios. (single primary improvement was to cargo incr-patched::println).
129+
* overall seems like minor regressions, potentially "just" instrumentation bias, compared to the expected benefit for microbenchmark reported in PR description.
130+
* marking as triaged.
131+
132+
Rollup of 8 pull requests [#131934](https://github.com/rust-lang/rust/pull/131934) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=a2a1206811d864df2bb61b2fc27ddc45a3589424&end=8069f8d17a6c86a8fd881939fcce359a90c57ff2&stat=instructions:u)
133+
134+
| (instructions:u) | mean | range | count |
135+
|:----------------------------------:|:-----:|:--------------:|:-----:|
136+
| Regressions ❌ <br /> (primary) | 0.3% | [0.3%, 0.3%] | 2 |
137+
| Regressions ❌ <br /> (secondary) | 0.2% | [0.2%, 0.3%] | 11 |
138+
| Improvements ✅ <br /> (primary) | -0.2% | [-0.2%, -0.2%] | 1 |
139+
| Improvements ✅ <br /> (secondary) | - | - | 0 |
140+
| All ❌✅ (primary) | 0.1% | [-0.2%, 0.3%] | 3 |
141+
142+
* primary regressions to helloworld (doc) and hyper (opt); primary improvement to libc (doc).
143+
* all the secondary regressions are to doc benchmarks, which led lqd to hypothesize that this is due to PR 131908 which changed the hash used for filename generation to sha256 since it should be stable going forward.
144+
* hyper history seems really noisy.
145+
* marking as triaged
146+
147+
Update rustc-hash to version 2 but again [#131949](https://github.com/rust-lang/rust/pull/131949) [(Comparison Link)](https://perf.rust-lang.org/compare.html?start=de977a5acf210f7d71ff83f4b8bc42c274ce4ed9&end=662180b34d95f72d05b7c467b0baf4d23d36b1e1&stat=instructions:u)
148+
149+
| (instructions:u) | mean | range | count |
150+
|:----------------------------------:|:-----:|:--------------:|:-----:|
151+
| Regressions ❌ <br /> (primary) | 0.2% | [0.1%, 0.2%] | 8 |
152+
| Regressions ❌ <br /> (secondary) | 0.3% | [0.1%, 0.7%] | 9 |
153+
| Improvements ✅ <br /> (primary) | -0.3% | [-0.5%, -0.2%] | 10 |
154+
| Improvements ✅ <br /> (secondary) | -0.9% | [-1.4%, -0.6%] | 13 |
155+
| All ❌✅ (primary) | -0.1% | [-0.5%, 0.2%] | 18 |
156+
157+
* primary improvements to unicode-normalization and libc
158+
* primary regressions to typenum and serde
159+
* i don't think there's anything interesting to investigate here. marking as triaged.

0 commit comments

Comments
 (0)