Lazy-page Symbol interner #136401

Mark-Simulacrum · 2025-02-01T21:09:59Z

This is a simple and only semi-correct impl that works well only on 64-bit platforms where we can lazily allocate zeroed pages (i.e., not actually use up memory for zeroed pages), meaning that we can grab a 4GB chunk of memory for all interned strings and then just offset allocate into that.

This already fixes unsoundness in the Symbol::as_str by leaking that 4gb memory chunk, but could also be faster (TBD).

A rewrite supporting ~any platform should be pretty straightforward by sharding the allocation and manually lazily allocating it in chunks.

r? @ghost -- this impl can't land due to being non-workable on Windows, opening for initial perf run

rustbot · 2025-02-01T21:10:04Z

These commits modify the Cargo.lock file. Unintentional changes to Cargo.lock can be introduced when switching branches and rebasing PRs.

If this was unintentional then you should revert the changes before this PR is merged.
Otherwise, you can ignore this comment.

Mark-Simulacrum · 2025-02-01T21:10:08Z

@bors try @rust-timer queue

…try> Lazy-page Symbol interner This is a simple and only semi-correct impl that works well only on 64-bit platforms where we can lazily allocate zeroed pages (i.e., not actually use up memory for zeroed pages), meaning that we can grab a 4GB chunk of memory for all interned strings and then just offset allocate into that. This already fixes unsoundness in the Symbol::as_str by leaking that 4gb memory chunk, but could also be faster (TBD). A rewrite supporting ~any platform should be pretty straightforward by sharding the allocation and manually lazily allocating it in chunks. r? `@ghost` -- this impl can't land due to being non-workable on Windows, opening for initial perf run

bors · 2025-02-01T21:11:19Z

⌛ Trying commit e6f870b with merge 085fad86adba2eafccd19c18a6c7ed5d3fbecbab...

bors · 2025-02-01T23:01:02Z

☀️ Try build successful - checks-actions
Build commit: 085fad8 (085fad86adba2eafccd19c18a6c7ed5d3fbecbab)

rust-timer · 2025-02-02T01:26:07Z

Finished benchmarking commit (085fad8): 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)	0.2%	[0.2%, 0.2%]	1
Regressions ❌ (secondary)	0.4%	[0.2%, 0.8%]	7
Improvements ✅ (primary)	-0.5%	[-0.8%, -0.2%]	4
Improvements ✅ (secondary)	-0.3%	[-0.6%, -0.1%]	20
All ❌✅ (primary)	-0.4%	[-0.8%, 0.2%]	5

Max RSS (memory usage)

Results (primary -0.2%, secondary -3.0%)

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)	3.5%	[3.2%, 3.8%]	3
Regressions ❌ (secondary)	2.4%	[1.7%, 3.4%]	3
Improvements ✅ (primary)	-2.1%	[-3.5%, -1.0%]	6
Improvements ✅ (secondary)	-3.7%	[-10.1%, -0.8%]	24
All ❌✅ (primary)	-0.2%	[-3.5%, 3.8%]	9

Cycles

Results (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)	2.9%	[2.3%, 3.3%]	6
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.2%	[-3.2%, -3.1%]	2
All ❌✅ (primary)	-	-	0

Binary size

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

Bootstrap: 777.851s -> 776.705s (-0.15%)
Artifact size: 328.84 MiB -> 328.88 MiB (0.01%)

This is a simple impl that works well only on 64-bit platforms where we can lazily allocate zeroed pages (i.e., not actually use up memory for zeroed pages), meaning that we can grab a 4GB chunk of memory for all interned strings and then just offset allocate into that. This already fixes unsoundness in the Symbol::as_str by leaking that 4gb memory chunk, but could also be faster (TBD). A rewrite supporting ~any platform should be pretty straightforward by sharding the allocation and manually lazily allocating it in chunks.

Mark-Simulacrum · 2025-02-02T03:48:46Z

@bors try @rust-timer queue

…try> Lazy-page Symbol interner This is a simple and only semi-correct impl that works well only on 64-bit platforms where we can lazily allocate zeroed pages (i.e., not actually use up memory for zeroed pages), meaning that we can grab a 4GB chunk of memory for all interned strings and then just offset allocate into that. This already fixes unsoundness in the Symbol::as_str by leaking that 4gb memory chunk, but could also be faster (TBD). A rewrite supporting ~any platform should be pretty straightforward by sharding the allocation and manually lazily allocating it in chunks. r? `@ghost` -- this impl can't land due to being non-workable on Windows, opening for initial perf run

bors · 2025-02-02T03:49:57Z

⌛ Trying commit b120ca9 with merge bad4b51...

bors · 2025-02-02T05:41:31Z

☀️ Try build successful - checks-actions
Build commit: bad4b51 (bad4b5105a2fc0c2d704f0d02aac6fd34ec7b4a9)

rust-timer · 2025-02-02T06:59:25Z

Finished benchmarking commit (bad4b51): 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%, 3.9%]	223
Regressions ❌ (secondary)	1.2%	[0.1%, 3.9%]	142
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-0.3%	[-0.4%, -0.2%]	14
All ❌✅ (primary)	0.7%	[0.2%, 3.9%]	223

Max RSS (memory usage)

Results (primary -0.2%, secondary -3.2%)

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)	3.2%	[2.8%, 3.5%]	2
Regressions ❌ (secondary)	3.0%	[2.1%, 3.7%]	4
Improvements ✅ (primary)	-1.6%	[-2.5%, -0.6%]	5
Improvements ✅ (secondary)	-4.4%	[-10.1%, -1.0%]	21
All ❌✅ (primary)	-0.2%	[-2.5%, 3.5%]	7

Cycles

Results (primary 2.0%, secondary 1.3%)

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)	2.0%	[0.9%, 3.1%]	25
Regressions ❌ (secondary)	2.9%	[1.8%, 4.3%]	34
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-2.9%	[-3.3%, -2.1%]	13
All ❌✅ (primary)	2.0%	[0.9%, 3.1%]	25

Binary size

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

Bootstrap: 777.275s -> 777.884s (0.08%)
Artifact size: 328.69 MiB -> 328.73 MiB (0.01%)

Mark-Simulacrum · 2025-02-02T17:00:30Z

@bors try @rust-timer queue

bors · 2025-02-02T17:01:40Z

⌛ Trying commit 79a6cff with merge ad08842...

…try> Lazy-page Symbol interner This is a simple and only semi-correct impl that works well only on 64-bit platforms where we can lazily allocate zeroed pages (i.e., not actually use up memory for zeroed pages), meaning that we can grab a 4GB chunk of memory for all interned strings and then just offset allocate into that. This already fixes unsoundness in the Symbol::as_str by leaking that 4gb memory chunk, but could also be faster (TBD). A rewrite supporting ~any platform should be pretty straightforward by sharding the allocation and manually lazily allocating it in chunks. r? `@ghost` -- this impl can't land due to being non-workable on Windows, opening for initial perf run

bors · 2025-02-02T18:55:44Z

☀️ Try build successful - checks-actions
Build commit: ad08842 (ad088427a5424ba3008dd911900b7aff41db7be2)

rust-timer · 2025-02-02T20:13:30Z

Finished benchmarking commit (ad08842): 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)	0.2%	[0.2%, 0.2%]	3
Regressions ❌ (secondary)	0.4%	[0.2%, 0.8%]	9
Improvements ✅ (primary)	-0.4%	[-0.7%, -0.2%]	4
Improvements ✅ (secondary)	-0.3%	[-0.6%, -0.1%]	20
All ❌✅ (primary)	-0.1%	[-0.7%, 0.2%]	7

Max RSS (memory usage)

Results (primary -0.3%, secondary -3.0%)

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)	2.8%	[2.6%, 3.1%]	2
Regressions ❌ (secondary)	3.2%	[0.9%, 8.0%]	5
Improvements ✅ (primary)	-1.6%	[-2.2%, -1.0%]	5
Improvements ✅ (secondary)	-4.7%	[-10.1%, -1.6%]	18
All ❌✅ (primary)	-0.3%	[-2.2%, 3.1%]	7

Cycles

Results (secondary -3.4%)

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)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.4%	[-4.2%, -2.8%]	5
All ❌✅ (primary)	-	-	0

Binary size

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

Bootstrap: 778.659s -> 777.495s (-0.15%)
Artifact size: 328.72 MiB -> 328.71 MiB (-0.00%)

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Feb 1, 2025