Cache more queries on disk #95418

cjgillot · 2022-03-28T19:36:03Z

One of the principles of incremental compilation is to allow saving results on disk to avoid recomputing them.
This PR investigates persisting a lot of queries whose result are to be saved into metadata.
Some of the queries are cheap reads from HIR, but we may also want to get rid of these reads for incremental lowering.

cjgillot · 2022-03-29T14:17:25Z

@bors try @rust-timer queue

rust-timer · 2022-03-29T14:17:26Z

Awaiting bors try build completion.

@rustbot label: +S-waiting-on-perf

bors · 2022-03-29T14:17:34Z

⌛ Trying commit ae9af5f539ba1eca44562ff9a400d8b46cdc66c6 with merge 36ace5e9edc63601569a7ef878ded180015f85d3...

bjorn3 · 2022-03-29T14:41:39Z

compiler/rustc_middle/src/query/mod.rs

        separate_provide_extern
    }

    query impl_constness(def_id: DefId) -> hir::Constness {
        desc { |tcx| "looking up whether `{}` is a const impl", tcx.def_path_str(def_id) }
+        cache_on_disk_if { def_id.is_local() }


Aren't those two just a cheap hir lookup?

Yes. For the moment, I'm trying some heuristic like "cache everything that will be stored in metadata (roughly if is has separate_provide_extern)". I plan to refine this from perf measurements.

bors · 2022-03-29T16:24:19Z

☀️ Try build successful - checks-actions
Build commit: 36ace5e9edc63601569a7ef878ded180015f85d3 (36ace5e9edc63601569a7ef878ded180015f85d3)

rust-timer · 2022-03-29T16:24:21Z

Queued 36ace5e9edc63601569a7ef878ded180015f85d3 with parent e2301ca, future comparison URL.

rust-timer · 2022-03-29T18:26:46Z

Finished benchmarking commit (36ace5e9edc63601569a7ef878ded180015f85d3): comparison url.

Summary: This benchmark run shows 83 relevant improvements 🎉 but 43 relevant regressions 😿 to instruction counts.

Arithmetic mean of relevant regressions: 1.3%
Arithmetic mean of relevant improvements: -2.1%
Arithmetic mean of all relevant changes: -0.9%
Largest improvement in instruction counts: -34.3% on incr-unchanged builds of externs debug
Largest regression in instruction counts: 4.5% on incr-unchanged builds of issue-58319 check

If you disagree with this performance assessment, please file an issue in rust-lang/rustc-perf.

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 led 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-review -S-waiting-on-perf +perf-regression

cjgillot · 2022-03-29T21:09:18Z

r? rust-lang/compiler

cjgillot · 2022-05-14T13:41:43Z

After investigation, I'm not sure how I can optimize the disk effect. One of the key effects is to avoid recomputing resolve_lifetimes (for instance here). This is done by skipping all "astconv" queries which access it indirectly. I'd rather avoid caching resolve_lifetimes, preferring queries whose results will be written into metadata.

davidtwco · 2022-05-16T04:14:48Z

If this is waiting on another review from me then all the changes in this PR still look good to me, so r=me if you don't want to do further changes.

After investigation, I'm not sure how I can optimize the disk effect. One of the key effects is to avoid recomputing resolve_lifetimes (for instance here). This is done by skipping all "astconv" queries which access it indirectly. I'd rather avoid caching resolve_lifetimes, preferring queries whose results will be written into metadata.

I'm not familiar enough with the relevant code to have any suggestions here unfortunately.

cjgillot · 2022-05-17T16:28:54Z

Let's confirm there are no unforeseen regressions, and I think we'll go ahead.

@bors try @rust-timer queue

rust-timer · 2022-05-17T16:28:55Z

Awaiting bors try build completion.

@rustbot label: +S-waiting-on-perf

bors · 2022-05-17T16:29:05Z

⌛ Trying commit 29f3b3f with merge 439d2588a96abfba733f7cef6e0f998d18042636...

bors · 2022-05-17T17:57:04Z

☀️ Try build successful - checks-actions
Build commit: 439d2588a96abfba733f7cef6e0f998d18042636 (439d2588a96abfba733f7cef6e0f998d18042636)

rust-timer · 2022-05-17T17:57:06Z

Queued 439d2588a96abfba733f7cef6e0f998d18042636 with parent 3655175, future comparison URL.

rust-timer · 2022-05-17T19:26:52Z

Finished benchmarking commit (439d2588a96abfba733f7cef6e0f998d18042636): comparison url.

Summary:

Primary benchmarks: mixed results
Secondary benchmarks: mixed results

	Regressions 😿 (primary)	Regressions 😿 (secondary)	Improvements 🎉 (primary)	Improvements 🎉 (secondary)	All 😿 🎉 (primary)
count¹	27	28	65	27	92
mean²	0.9%	1.7%	-1.0%	-4.8%	-0.5%
max	2.0%	4.1%	-3.7%	-36.4%	-3.7%

If you disagree with this performance assessment, please file an issue in rust-lang/rustc-perf.

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-review -S-waiting-on-perf +perf-regression

number of relevant changes ↩
the arithmetic mean of the percent change ↩

cjgillot · 2022-05-20T17:15:35Z

@bors r=davidtwco

bors · 2022-05-20T17:15:37Z

📌 Commit 29f3b3f has been approved by davidtwco

cjgillot · 2022-05-20T17:20:13Z

This PR caches many more queries' results on disk. The queries have been chosen to be: (1) outputs of typechecking, and (2) parts of metadata. This caching allows for wide and large compile time savings, up to 36%. The tradeoff is a compile time increase (4%) when these caches cannot be usefully used, and a increased on-disk size around 30%.
@rustbot label: +perf-regression-triaged

bors · 2022-05-20T20:49:58Z

⌛ Testing commit 29f3b3f with merge e6a4afc...

bors · 2022-05-20T23:11:05Z

☀️ Test successful - checks-actions
Approved by: davidtwco
Pushing e6a4afc to master...

rust-timer · 2022-05-21T03:46:52Z

Finished benchmarking commit (e6a4afc): comparison url.

Instruction count

Primary benchmarks: mixed results
Secondary benchmarks: mixed results

	Regressions 😿 (primary)	Regressions 😿 (secondary)	Improvements 🎉 (primary)	Improvements 🎉 (secondary)	All 😿 🎉 (primary)
count¹	33	32	65	28	98
mean²	0.8%	1.6%	-1.0%	-4.4%	-0.4%
max	2.1%	4.0%	-3.6%	-33.0%	-3.6%

Max RSS (memory usage)

Results

Primary benchmarks: 😿 relevant regressions found
Secondary benchmarks: mixed results

	Regressions 😿 (primary)	Regressions 😿 (secondary)	Improvements 🎉 (primary)	Improvements 🎉 (secondary)	All 😿 🎉 (primary)
count¹	17	5	2	8	19
mean²	1.3%	1.6%	-1.1%	-1.9%	1.1%
max	2.4%	1.9%	-1.2%	-2.5%	2.4%

Cycles

Results

Primary benchmarks: mixed results
Secondary benchmarks: mixed results

	Regressions 😿 (primary)	Regressions 😿 (secondary)	Improvements 🎉 (primary)	Improvements 🎉 (secondary)	All 😿 🎉 (primary)
count¹	11	16	8	9	19
mean²	4.0%	4.6%	-3.0%	-7.4%	1.1%
max	6.0%	8.9%	-4.4%	-25.0%	6.0%

If you disagree with this performance assessment, please file an issue in rust-lang/rustc-perf.

Next Steps: If you can justify the regressions found in this perf run, please indicate this with @rustbot label: +perf-regression-triaged along with sufficient written justification. If you cannot justify the regressions please open an issue or create a new PR that fixes the regressions, add a comment linking to the newly created issue or PR, and then add the perf-regression-triaged label to this PR.

@rustbot label: +perf-regression

number of relevant changes ↩ ↩² ↩³
the arithmetic mean of the percent change ↩ ↩² ↩³

rustbot added T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. T-rustdoc Relevant to the rustdoc team, which will review and decide on the PR/issue. labels Mar 28, 2022

This comment was marked as outdated.

Sign in to view

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Mar 28, 2022

This comment was marked as outdated.

Sign in to view

This comment has been minimized.

Sign in to view

This comment was marked as outdated.

Sign in to view

rustbot added perf-regression Performance regression. S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-perf Status: Waiting on a perf run to be completed. labels Mar 29, 2022

cjgillot force-pushed the more-disk branch from 84bcd94 to ae9af5f Compare March 29, 2022 14:16

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Mar 29, 2022

bjorn3 reviewed Mar 29, 2022

View reviewed changes

rustbot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Mar 29, 2022

rust-highfive assigned davidtwco Mar 29, 2022

This comment was marked as outdated.

Sign in to view

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Mar 29, 2022

cjgillot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Mar 29, 2022

cjgillot added 2 commits May 13, 2022 08:06

Cache more queries on disk.

9900ea3

Do not cache item_attrs.

29f3b3f

cjgillot force-pushed the more-disk branch from ae9af5f to 29f3b3f Compare May 13, 2022 16:39

davidtwco approved these changes May 16, 2022

View reviewed changes

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label May 17, 2022

rustbot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label May 17, 2022

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels May 20, 2022

rustbot added the perf-regression-triaged The performance regression has been triaged. label May 20, 2022

bors added the merged-by-bors This PR was explicitly merged by bors. label May 20, 2022

bors merged commit e6a4afc into rust-lang:master May 20, 2022

rustbot added this to the 1.63.0 milestone May 20, 2022

This was referenced May 20, 2022

Directly save a byte representation of the dep-graph and work-product index #83322

Closed

Implement pointee metadata unsizing via a TypedMetadata<T> container #97052

Closed

cjgillot deleted the more-disk branch May 21, 2022 07:37

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Cache more queries on disk #95418

Cache more queries on disk #95418

cjgillot commented Mar 28, 2022 •

edited

Loading

This comment was marked as outdated.

This comment was marked as outdated.

This comment was marked as outdated.

This comment has been minimized.

This comment was marked as outdated.

This comment was marked as outdated.

This comment was marked as outdated.

cjgillot commented Mar 29, 2022

rust-timer commented Mar 29, 2022

bors commented Mar 29, 2022

bjorn3 Mar 29, 2022 •

edited

Loading

cjgillot Mar 29, 2022

bors commented Mar 29, 2022

rust-timer commented Mar 29, 2022

rust-timer commented Mar 29, 2022

cjgillot commented Mar 29, 2022

This comment was marked as outdated.

cjgillot commented May 14, 2022

davidtwco commented May 16, 2022

cjgillot commented May 17, 2022

rust-timer commented May 17, 2022

bors commented May 17, 2022

bors commented May 17, 2022

rust-timer commented May 17, 2022

rust-timer commented May 17, 2022

cjgillot commented May 20, 2022

bors commented May 20, 2022

cjgillot commented May 20, 2022 •

edited

Loading

bors commented May 20, 2022

bors commented May 20, 2022

rust-timer commented May 21, 2022

Cache more queries on disk #95418

Cache more queries on disk #95418

Conversation

cjgillot commented Mar 28, 2022 • edited Loading

This comment was marked as outdated.

This comment was marked as outdated.

This comment was marked as outdated.

This comment has been minimized.

This comment was marked as outdated.

This comment was marked as outdated.

This comment was marked as outdated.

cjgillot commented Mar 29, 2022

rust-timer commented Mar 29, 2022

bors commented Mar 29, 2022

bjorn3 Mar 29, 2022 • edited Loading

Choose a reason for hiding this comment

cjgillot Mar 29, 2022

Choose a reason for hiding this comment

bors commented Mar 29, 2022

rust-timer commented Mar 29, 2022

rust-timer commented Mar 29, 2022

cjgillot commented Mar 29, 2022

This comment was marked as outdated.

cjgillot commented May 14, 2022

davidtwco commented May 16, 2022

cjgillot commented May 17, 2022

rust-timer commented May 17, 2022

bors commented May 17, 2022

bors commented May 17, 2022

rust-timer commented May 17, 2022

rust-timer commented May 17, 2022

Footnotes

cjgillot commented May 20, 2022

bors commented May 20, 2022

cjgillot commented May 20, 2022 • edited Loading

bors commented May 20, 2022

bors commented May 20, 2022

rust-timer commented May 21, 2022

Instruction count

Max RSS (memory usage)

Cycles

Footnotes

cjgillot commented Mar 28, 2022 •

edited

Loading

bjorn3 Mar 29, 2022 •

edited

Loading

cjgillot commented May 20, 2022 •

edited

Loading