miri: optimize zeroed alloc #136035

SpecificProtagonist · 2025-01-25T01:27:11Z

When allocating zero-initialized memory in MIR interpretation, rustc allocates zeroed memory, marks it as initialized and then re-zeroes it. Remove the last step.

I don't expect this to have much of an effect on performance normally, but in my case in which I'm creating a large allocation via mmap it gets in the way.

rustbot · 2025-01-25T01:27:20Z

Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @wesleywiser (or someone else) some time within the next two weeks.

Please see the contribution instructions for more information. Namely, in order to ensure the minimum review times lag, PR authors and assigned reviewers should ensure that the review label (S-waiting-on-review and S-waiting-on-author) stays updated, invoking these commands when appropriate:

@rustbot author: the review is finished, PR author should check the comments and take action accordingly
@rustbot review: the author is ready for a review, this PR will be queued again in the reviewer's queue

rustbot · 2025-01-25T01:27:22Z

Some changes occurred to the CTFE machinery

cc @rust-lang/wg-const-eval

The Miri subtree was changed

cc @rust-lang/miri

Some changes occurred to the CTFE / Miri interpreter

cc @rust-lang/miri, @rust-lang/wg-const-eval

Some changes occurred to the CTFE / Miri interpreter

cc @rust-lang/miri

jieyouxu · 2025-01-25T07:17:49Z

r? miri

SpecificProtagonist · 2025-01-25T09:51:34Z

Sorry, I'm not sure how I closed this – misclick?

Kobzol · 2025-01-25T11:36:51Z

@bors try @rust-timer queue

bors · 2025-01-25T11:38:02Z

⌛ Trying commit bd28faf with merge 837b710...

…c, r=<try> miri: optimize zeroed alloc When allocating zero-initialized memory in MIR interpretation, rustc allocates zeroed memory, marks it as initialized and then re-zeroes it. Remove the last step. I don't expect this to have much of an effect on performance normally, but in my case in which I'm creating a large allocation via mmap miri is unusable without this. There's probably a better way – with less code duplication – to implement this. Maybe adding a zero_init flag to the relevant methods, but then `Allocation::uninit` & co need a new name :)

bors · 2025-01-25T13:29:27Z

☀️ Try build successful - checks-actions
Build commit: 837b710 (837b710e5dd54b53b888f1ab109a3b93efc9a144)

oli-obk · 2025-01-25T14:11:32Z

This is not gonna show up in perf. No code path outside miri is changed

RalfJung · 2025-01-25T14:21:48Z

We should definitely explore ways to do this with less code duplication. :)
Adding a flag sounds like a good idea. The name of the method could just be Allocation::new etc?

rust-timer · 2025-01-25T15:26:43Z

Finished benchmarking commit (837b710): comparison URL.

Overall result: ✅ improvements - no action needed

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.

@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
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-0.2%	[-0.2%, -0.2%]	2
All ❌✅ (primary)	-	-	0

Max RSS (memory usage)

Results (primary -2.2%, secondary 2.2%)

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)	-	-	0
Regressions ❌ (secondary)	2.2%	[2.2%, 2.2%]	2
Improvements ✅ (primary)	-2.2%	[-2.2%, -2.2%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-2.2%	[-2.2%, -2.2%]	1

Cycles

Results (primary -1.6%, secondary -0.2%)

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)	-	-	0
Regressions ❌ (secondary)	1.3%	[1.3%, 1.3%]	1
Improvements ✅ (primary)	-1.6%	[-1.6%, -1.6%]	1
Improvements ✅ (secondary)	-1.6%	[-1.6%, -1.6%]	1
All ❌✅ (primary)	-1.6%	[-1.6%, -1.6%]	1

Binary size

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

Bootstrap: 771.261s -> 771.165s (-0.01%)
Artifact size: 325.82 MiB -> 325.82 MiB (0.00%)

SpecificProtagonist · 2025-01-25T18:34:42Z

Adding a flag sounds like a good idea. The name of the method could just be Allocation::new etc?

Changed 👍

oli-obk · 2025-01-25T19:32:14Z

compiler/rustc_const_eval/src/interpret/memory.rs

@@ -289,7 +291,7 @@ impl<'tcx, M: Machine<'tcx>> InterpCx<'tcx, M> {

        // For simplicities' sake, we implement reallocate as "alloc, copy, dealloc".
        // This happens so rarely, the perf advantage is outweighed by the maintenance cost.
-        let new_ptr = self.allocate_ptr(new_size, new_align, kind)?;
+        let new_ptr = self.allocate_ptr(new_size, new_align, kind, zero_init)?;


Since we'll memcopy right over the alloc, zeroing is not needed for realloc, just use false here

The memcopy only initializes part of the allocation, but mremap needs everything to be initialized.

oh I see, thanks!

Please add a comment explaining this.

Also seems worth saying that we assume here that zeroing the entire allocation is more efficient than zeroing just the parts not copied from the old allocation.

oli-obk · 2025-01-25T19:37:08Z

compiler/rustc_middle/src/mir/interpret/allocation.rs

@@ -315,8 +320,8 @@ impl<Prov: Provenance, Bytes: AllocBytes> Allocation<Prov, (), Bytes> {

    /// Try to create an Allocation of `size` bytes, failing if there is not enough memory
    /// available to the compiler to do so.
-    pub fn try_uninit<'tcx>(size: Size, align: Align) -> InterpResult<'tcx, Self> {
-        Self::uninit_inner(size, align, || {
+    pub fn try_new<'tcx>(size: Size, align: Align, zero_init: bool) -> InterpResult<'tcx, Self> {


Bools are a bit icky, as the call sites don't make it clear what the bool means. Maybe an enum with Uninit and Zeroed variants would be better? What do you think @RalfJung

Yeah, I'm not happy with the unlabeled bools. An additional (very simple) possibility could be to use comments – this is used by existing code: mem_copy(ptr, new_ptr.into(), old_size.min(new_size), /*nonoverlapping*/ true).

Yeah a dedicated enum makes sense.

miri: optimize zeroed alloc

185e3ef

rustbot assigned wesleywiser Jan 25, 2025

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 Jan 25, 2025

This comment has been minimized.

Sign in to view

rustbot assigned oli-obk and unassigned wesleywiser Jan 25, 2025

leftover import

bd28faf

SpecificProtagonist closed this Jan 25, 2025

SpecificProtagonist reopened this Jan 25, 2025

This comment has been minimized.

Sign in to view

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Jan 25, 2025

This comment has been minimized.

Sign in to view

rustbot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Jan 25, 2025

implement via flag

e685a36

oli-obk reviewed Jan 25, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

miri: optimize zeroed alloc #136035

miri: optimize zeroed alloc #136035

SpecificProtagonist commented Jan 25, 2025 •

edited

Loading

rustbot commented Jan 25, 2025

rustbot commented Jan 25, 2025

This comment has been minimized.

jieyouxu commented Jan 25, 2025

SpecificProtagonist commented Jan 25, 2025

Kobzol commented Jan 25, 2025

This comment has been minimized.

bors commented Jan 25, 2025

bors commented Jan 25, 2025

This comment has been minimized.

oli-obk commented Jan 25, 2025

RalfJung commented Jan 25, 2025

rust-timer commented Jan 25, 2025

SpecificProtagonist commented Jan 25, 2025

oli-obk Jan 25, 2025

SpecificProtagonist Jan 25, 2025

oli-obk Jan 25, 2025

RalfJung Jan 25, 2025

RalfJung Jan 26, 2025

oli-obk Jan 25, 2025

SpecificProtagonist Jan 25, 2025 •

edited

Loading

RalfJung Jan 25, 2025

miri: optimize zeroed alloc #136035

Are you sure you want to change the base?

miri: optimize zeroed alloc #136035

Conversation

SpecificProtagonist commented Jan 25, 2025 • edited Loading

rustbot commented Jan 25, 2025

rustbot commented Jan 25, 2025

This comment has been minimized.

jieyouxu commented Jan 25, 2025

SpecificProtagonist commented Jan 25, 2025

Kobzol commented Jan 25, 2025

This comment has been minimized.

bors commented Jan 25, 2025

bors commented Jan 25, 2025

This comment has been minimized.

oli-obk commented Jan 25, 2025

RalfJung commented Jan 25, 2025

rust-timer commented Jan 25, 2025

Overall result: ✅ improvements - no action needed

SpecificProtagonist commented Jan 25, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

SpecificProtagonist Jan 25, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

SpecificProtagonist commented Jan 25, 2025 •

edited

Loading

SpecificProtagonist Jan 25, 2025 •

edited

Loading