A quiet week overall. A few smaller crate (e.g., helloworld) benchmarks saw significant improvements in #118568, but this merely restores performance regressed earlier.
Triage done by @simulacrum. Revision range: 9358642..5701093
Summary:
| (instructions:u) | mean | range | count |
|---|---|---|---|
| Regressions ❌ (primary) | 0.5% | [0.2%, 1.3%] | 111 |
| Regressions ❌ (secondary) | 0.7% | [0.1%, 1.5%] | 34 |
| Improvements ✅ (primary) | -5.2% | [-11.1%, -0.5%] | 18 |
| Improvements ✅ (secondary) | -4.6% | [-10.5%, -0.4%] | 72 |
| All ❌✅ (primary) | -0.3% | [-11.1%, 1.3%] | 129 |
5 Regressions, 2 Improvements, 3 Mixed; 2 of them in rollups 69 artifact comparisons made in total
compile-time evaluation: detect writes through immutable pointers #118324 (Comparison Link)
| (instructions:u) | mean | range | count |
|---|---|---|---|
| Regressions ❌ (primary) | - | - | 0 |
| Regressions ❌ (secondary) | 0.9% | [0.3%, 1.4%] | 12 |
| Improvements ✅ (primary) | - | - | 0 |
| Improvements ✅ (secondary) | - | - | 0 |
| All ❌✅ (primary) | - | - | 0 |
I feel like the remaining regression is small enough to be acceptable. We're doing more checks in const-eval, and there are no regressions in the builds of real crates.
(https://github.com/rust-lang/rust/pull/118324#issuecomment-1832836626)
Introduce support for async gen blocks #118420 (Comparison Link)
| (instructions:u) | mean | range | count |
|---|---|---|---|
| Regressions ❌ (primary) | 0.2% | [0.1%, 0.3%] | 23 |
| Regressions ❌ (secondary) | 0.4% | [0.2%, 0.7%] | 3 |
| Improvements ✅ (primary) | - | - | 0 |
| Improvements ✅ (secondary) | - | - | 0 |
| All ❌✅ (primary) | 0.2% | [0.1%, 0.3%] | 23 |
New functionality costs some additional time across some benchmarks.
Rollup of 6 pull requests #118763 (Comparison Link)
| (instructions:u) | mean | range | count |
|---|---|---|---|
| Regressions ❌ (primary) | 0.4% | [0.2%, 0.5%] | 18 |
| Regressions ❌ (secondary) | - | - | 0 |
| Improvements ✅ (primary) | - | - | 0 |
| Improvements ✅ (secondary) | - | - | 0 |
| All ❌✅ (primary) | 0.4% | [0.2%, 0.5%] | 18 |
We do a little bit more work after that PR because the reachable set is slightly bigger and that in turn means more MIR gets encoded and things like that.
(https://github.com/rust-lang/rust/pull/118763#issuecomment-1850073772)
Don‘t warn an empty pattern unreachable if we’re not sure the data is valid #118308 (Comparison Link)
| (instructions:u) | mean | range | count |
|---|---|---|---|
| Regressions ❌ (primary) | 0.6% | [0.2%, 0.9%] | 4 |
| Regressions ❌ (secondary) | 1.0% | [1.0%, 1.0%] | 6 |
| Improvements ✅ (primary) | - | - | 0 |
| Improvements ✅ (secondary) | - | - | 0 |
| All ❌✅ (primary) | 0.6% | [0.2%, 0.9%] | 4 |
Yeah I expected a perf regression because of the validity tracking. It adds an extra step for every constructor we try, and match-stress has an enum with 8000 of these. I don‘t think there’s a way around it, I need to know the validity of every inspected place.
(https://github.com/rust-lang/rust/pull/118308#issuecomment-1831103831)
This adds costs particularly to crates with lots or large matches, seen across a few primary workloads and within rustc bootstrap. This is a diagnostic correctness fix, so some cost is acceptable.
Restore const PartialEq #118661 (Comparison Link)
| (instructions:u) | mean | range | count |
|---|---|---|---|
| Regressions ❌ (primary) | 0.3% | [0.2%, 0.4%] | 9 |
| Regressions ❌ (secondary) | 0.3% | [0.2%, 0.6%] | 11 |
| Improvements ✅ (primary) | - | - | 0 |
| Improvements ✅ (secondary) | -0.8% | [-0.8%, -0.8%] | 1 |
| All ❌✅ (primary) | 0.3% | [0.2%, 0.4%] | 9 |
Feature addition carrying likely unavoidable cost in some benchmarks.
Avoid adding builtin functions to symbols.o #118568 (Comparison Link)
| (instructions:u) | mean | range | count |
|---|---|---|---|
| Regressions ❌ (primary) | - | - | 0 |
| Regressions ❌ (secondary) | - | - | 0 |
| Improvements ✅ (primary) | -6.6% | [-11.0%, -0.7%] | 14 |
| Improvements ✅ (secondary) | -5.0% | [-10.7%, -0.3%] | 66 |
| All ❌✅ (primary) | -6.6% | [-11.0%, -0.7%] | 14 |
Smaller benchmarks show significant improvements, recovering a past performance regression. See PR description for details.
Rollup of 8 pull requests #118838 (Comparison Link)
| (instructions:u) | mean | range | count |
|---|---|---|---|
| Regressions ❌ (primary) | - | - | 0 |
| Regressions ❌ (secondary) | - | - | 0 |
| Improvements ✅ (primary) | - | - | 0 |
| Improvements ✅ (secondary) | -0.9% | [-1.0%, -0.9%] | 6 |
| All ❌✅ (primary) | - | - | 0 |
https://github.com/rust-lang/rust/pull/118822 is the likely cause, but not worth digging deeper on this improvement.
split NormalizesTo out of Projection 3 #118725 (Comparison Link)
| (instructions:u) | mean | range | count |
|---|---|---|---|
| Regressions ❌ (primary) | 0.4% | [0.2%, 0.9%] | 17 |
| Regressions ❌ (secondary) | 0.3% | [0.2%, 0.3%] | 3 |
| Improvements ✅ (primary) | -2.6% | [-2.6%, -2.6%] | 1 |
| Improvements ✅ (secondary) | - | - | 0 |
| All ❌✅ (primary) | 0.3% | [-2.6%, 0.9%] | 18 |
Moving code around in hot area, with slightly different work being done. Seems to add some cost to a few primary benchmarks.
See details in https://github.com/rust-lang/rust/pull/118725#issuecomment-1847121861
never_patterns: Parse match arms with no body #118527 (Comparison Link)
| (instructions:u) | mean | range | count |
|---|---|---|---|
| Regressions ❌ (primary) | 0.1% | [0.1%, 0.1%] | 1 |
| Regressions ❌ (secondary) | - | - | 0 |
| Improvements ✅ (primary) | -0.9% | [-1.0%, -0.8%] | 2 |
| Improvements ✅ (secondary) | - | - | 0 |
| All ❌✅ (primary) | -0.5% | [-1.0%, 0.1%] | 3 |
Regression is justified by the extra work for this new feature.
Improve print_tts #114571 (Comparison Link)
| (instructions:u) | mean | range | count |
|---|---|---|---|
| Regressions ❌ (primary) | 0.2% | [0.2%, 0.2%] | 1 |
| Regressions ❌ (secondary) | 0.8% | [0.5%, 1.0%] | 8 |
| Improvements ✅ (primary) | - | - | 0 |
| Improvements ✅ (secondary) | -0.2% | [-0.3%, -0.2%] | 2 |
| All ❌✅ (primary) | 0.2% | [0.2%, 0.2%] | 1 |
Regressions mostly limited to secondary benchmarks. Single primary regression is in incr-unchanged and is relatively small.