Use a more lightweight cache for `erase_regions` #139240

Zoxc · 2025-04-02T07:56:49Z

This changes erase_regions to use a global hashmap instead of a query to cache types with erased lifetimes. erase_regions_ty only depends on the input parameter so it doesn't need to be a query.

This is a rebase of #59505.

Benchmark	Before	After		Before	After		Before	After
Benchmark	Time	Time	%	Physical Memory	Physical Memory	%	Committed Memory	Committed Memory	%
🟣 clap:check	1.3960s	1.3883s	-0.55%	147.66 MiB	147.76 MiB	0.07%	199.83 MiB	199.86 MiB	0.01%
🟣 hyper:check	0.2346s	0.2337s	-0.38%	80.35 MiB	80.54 MiB	0.24%	129.99 MiB	130.12 MiB	0.10%
🟣 regex:check	0.7811s	0.7788s	-0.30%	107.75 MiB	108.07 MiB	0.29%	153.67 MiB	153.99 MiB	0.21%
🟣 syn:check	1.2999s	1.2978s	-0.17%	140.24 MiB	140.35 MiB	0.08%	188.59 MiB	188.58 MiB	-0.01%
Total	3.7117s	3.6986s	-0.35%	476.00 MiB	476.72 MiB	0.15%	672.08 MiB	672.55 MiB	0.07%
Summary	1.0000s	0.9965s	-0.35%	1 byte	1.00 bytes	0.17%	1 byte	1.00 bytes	0.08%

rustbot · 2025-04-02T07:56:54Z

r? @petrochenkov

rustbot has assigned @petrochenkov.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

lqd · 2025-04-02T08:29:15Z

Remember that you now have credentials for perf runs, which should be more accurate, and also more representative, with a LTOed/PGOed toolchain.

petrochenkov · 2025-04-02T13:08:09Z

compiler/rustc_query_system/src/dep_graph/graph.rs

@@ -268,6 +268,14 @@ impl<D: Deps> DepGraph<D> {
        D::with_deps(TaskDepsRef::Forbid, op)
    }

+    /// This checks that no dependencies are registered in `op` if debug assertions are enabled.
+    pub fn debug_assert_no_deps<OP, R>(op: OP) -> R


Suggested change

pub fn debug_assert_no_deps<OP, R>(op: OP) -> R

pub fn debug_assert_no_deps<R>(op: impl FnOnce() -> R) -> R

petrochenkov · 2025-04-02T13:11:29Z

compiler/rustc_middle/src/ty/context.rs

@@ -1376,6 +1376,9 @@ pub struct GlobalCtxt<'tcx> {
    /// Common consts, pre-interned for your convenience.
    pub consts: CommonConsts<'tcx>,

+    /// A cache for the `erase_regions` function.
+    pub(in crate::ty) erased_region_cache: ShardedHashMap<Ty<'tcx>, Ty<'tcx>>,


Could you add comments explaining why 1) the query cache is not used here and 2) a sharded hash map is used instead of a simple one.

petrochenkov · 2025-04-02T13:13:35Z

compiler/rustc_middle/src/ty/erase_regions.rs

-}
+impl<'tcx> TyCtxt<'tcx> {
+    /// Erases regions from `ty` to yield a new type.
+    pub fn erase_regions_ty(self, ty: Ty<'tcx>) -> Ty<'tcx> {


Could you add a comment telling that this function is expected to be a pure function of its input, and that's why it cannot call any other queries.

petrochenkov · 2025-04-02T13:14:45Z

@bors try @rust-timer queue

Use a more lightweight cache for `erase_regions` This changes `erase_regions` to use a global hashmap instead of a query to cache types with erased lifetimes. `erase_regions_ty` only depends on the input parameter so it doesn't need to be a query. This is a rebase of rust-lang#59505. <table><tr><td rowspan="2">Benchmark</td><td colspan="1">Before</th><td colspan="2">After</th><td colspan="1">Before</th><td colspan="2">After</th><td colspan="1">Before</th><td colspan="2">After</th></tr><tr><td align="right">Time</td><td align="right">Time</td><td align="right">%</th><td align="right">Physical Memory</td><td align="right">Physical Memory</td><td align="right">%</th><td align="right">Committed Memory</td><td align="right">Committed Memory</td><td align="right">%</th></tr><tr><td>🟣 clap:check</td><td align="right">1.3960s</td><td align="right">1.3883s</td><td align="right"> -0.55%</td><td align="right">147.66 MiB</td><td align="right">147.76 MiB</td><td align="right"> 0.07%</td><td align="right">199.83 MiB</td><td align="right">199.86 MiB</td><td align="right"> 0.01%</td></tr><tr><td>🟣 hyper:check</td><td align="right">0.2346s</td><td align="right">0.2337s</td><td align="right"> -0.38%</td><td align="right">80.35 MiB</td><td align="right">80.54 MiB</td><td align="right"> 0.24%</td><td align="right">129.99 MiB</td><td align="right">130.12 MiB</td><td align="right"> 0.10%</td></tr><tr><td>🟣 regex:check</td><td align="right">0.7811s</td><td align="right">0.7788s</td><td align="right"> -0.30%</td><td align="right">107.75 MiB</td><td align="right">108.07 MiB</td><td align="right"> 0.29%</td><td align="right">153.67 MiB</td><td align="right">153.99 MiB</td><td align="right"> 0.21%</td></tr><tr><td>🟣 syn:check</td><td align="right">1.2999s</td><td align="right">1.2978s</td><td align="right"> -0.17%</td><td align="right">140.24 MiB</td><td align="right">140.35 MiB</td><td align="right"> 0.08%</td><td align="right">188.59 MiB</td><td align="right">188.58 MiB</td><td align="right"> -0.01%</td></tr><tr><td>Total</td><td align="right">3.7117s</td><td align="right">3.6986s</td><td align="right"> -0.35%</td><td align="right">476.00 MiB</td><td align="right">476.72 MiB</td><td align="right"> 0.15%</td><td align="right">672.08 MiB</td><td align="right">672.55 MiB</td><td align="right"> 0.07%</td></tr><tr><td>Summary</td><td align="right">1.0000s</td><td align="right">0.9965s</td><td align="right"> -0.35%</td><td align="right">1 byte</td><td align="right">1.00 bytes</td><td align="right"> 0.17%</td><td align="right">1 byte</td><td align="right">1.00 bytes</td><td align="right"> 0.08%</td></tr></table>

bors · 2025-04-02T13:15:57Z

⌛ Trying commit bb40db8 with merge 35d98da...

bors · 2025-04-02T15:25:07Z

☀️ Try build successful - checks-actions
Build commit: 35d98da (35d98da4f0342063e91963b94fb43fa99e375349)

rust-timer · 2025-04-02T18:03:53Z

Finished benchmarking commit (35d98da): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. While you can manually mark this PR as fit for rollup, we strongly recommend not doing so since this PR may lead to changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please indicate this with @rustbot label: +perf-regression-triaged along with sufficient written justification. If you cannot justify the regressions please fix the regressions and do another perf run. If the next run shows neutral or positive results, the label will be automatically removed.

@bors rollup=never
@rustbot label: -S-waiting-on-perf +perf-regression

Instruction count

This is the most reliable metric that we have; it was used to determine the overall result at the top of this comment. However, even this metric can sometimes exhibit noise.

	mean	range	count
Regressions ❌ (primary)	0.7%	[0.2%, 1.1%]	4
Regressions ❌ (secondary)	0.5%	[0.3%, 0.9%]	7
Improvements ✅ (primary)	-0.9%	[-3.1%, -0.3%]	5
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-0.2%	[-3.1%, 1.1%]	9

Max RSS (memory usage)

Results (primary -0.7%, secondary -1.4%)

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	1.4%	[1.4%, 1.4%]	1
Improvements ✅ (primary)	-0.7%	[-0.7%, -0.7%]	1
Improvements ✅ (secondary)	-2.3%	[-3.1%, -1.4%]	3
All ❌✅ (primary)	-0.7%	[-0.7%, -0.7%]	1

Cycles

Results (primary -1.8%)

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.8%	[-1.8%, -1.8%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-1.8%	[-1.8%, -1.8%]	1

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 776.227s -> 774.214s (-0.26%)
Artifact size: 365.99 MiB -> 365.96 MiB (-0.01%)

petrochenkov · 2025-04-02T22:45:39Z

Not sure if it's worth it, the change is doing something unusual and non-obvious, but the results are mixed.
@rustbot author

rustbot · 2025-04-02T22:45:44Z

Reminder, once the PR becomes ready for a review, use @rustbot ready.

Use a more lightweight cache for erase_regions

bb40db8

rustbot assigned petrochenkov Apr 2, 2025

rustbot added A-query-system S-waiting-on-review T-compiler labels Apr 2, 2025

petrochenkov reviewed Apr 2, 2025

View reviewed changes

This comment has been minimized.

Sign in to view

rustbot added the S-waiting-on-perf label Apr 2, 2025

petrochenkov removed the S-waiting-on-review label Apr 2, 2025

This comment has been minimized.

Sign in to view

rustbot added perf-regression and removed S-waiting-on-perf labels Apr 2, 2025

rustbot added the S-waiting-on-author label Apr 2, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Use a more lightweight cache for `erase_regions` #139240

Use a more lightweight cache for `erase_regions` #139240

Zoxc commented Apr 2, 2025

rustbot commented Apr 2, 2025

lqd commented Apr 2, 2025 •

edited

Loading

petrochenkov Apr 2, 2025

petrochenkov Apr 2, 2025

petrochenkov Apr 2, 2025

petrochenkov commented Apr 2, 2025

This comment has been minimized.

bors commented Apr 2, 2025

bors commented Apr 2, 2025

This comment has been minimized.

rust-timer commented Apr 2, 2025

petrochenkov commented Apr 2, 2025

rustbot commented Apr 2, 2025

	pub fn debug_assert_no_deps<OP, R>(op: OP) -> R
	pub fn debug_assert_no_deps<R>(op: impl FnOnce() -> R) -> R

Use a more lightweight cache for erase_regions #139240

Are you sure you want to change the base?

Use a more lightweight cache for erase_regions #139240

Conversation

Zoxc commented Apr 2, 2025

rustbot commented Apr 2, 2025

lqd commented Apr 2, 2025 • edited Loading

petrochenkov Apr 2, 2025

Choose a reason for hiding this comment

petrochenkov Apr 2, 2025

Choose a reason for hiding this comment

petrochenkov Apr 2, 2025

Choose a reason for hiding this comment

petrochenkov commented Apr 2, 2025

This comment has been minimized.

bors commented Apr 2, 2025

bors commented Apr 2, 2025

This comment has been minimized.

rust-timer commented Apr 2, 2025

Overall result: ❌✅ regressions and improvements - please read the text below

Instruction count

Max RSS (memory usage)

Cycles

Binary size

petrochenkov commented Apr 2, 2025

rustbot commented Apr 2, 2025

Use a more lightweight cache for `erase_regions` #139240

Use a more lightweight cache for `erase_regions` #139240

lqd commented Apr 2, 2025 •

edited

Loading