Windows: Support sub-millisecond sleep #116461

ChrisDenton · 2023-10-05T23:20:12Z

Use CreateWaitableTimerExW with CREATE_WAITABLE_TIMER_HIGH_RESOLUTION. Does not work before Windows 10, version 1803 so in that case we fallback to using Sleep.

I've created a WaitableTimer type so it can one day be adapted to also support waiting to an absolute time (which has been talked about). Note though that it currently returns Err(()) because we can't do anything with the errors other than fallback to the old Sleep. Feel free to tell me to do errors properly. It just didn't seem worth constructing an io::Error if we're never going to surface it to the user. And it should all be infallible anyway unless the OS is too old to support it.

Closes #43376

rustbot · 2023-10-05T23:20:19Z

r? @thomcc

(rustbot has picked a reviewer for you, use r? to override)

ChrisDenton · 2023-10-05T23:46:02Z

That error looks like it's just miri again. I'll try to fix it.

rustbot · 2023-10-06T00:16:47Z

The Miri subtree was changed

cc @rust-lang/miri

Use `CreateWaitableTimerExW` with `CREATE_WAITABLE_TIMER_HIGH_RESOLUTION`. Does not work before Windows 10, version 1803 so in that case we fallback to using `Sleep`.

RalfJung · 2023-10-06T05:36:55Z

src/tools/miri/src/shims/windows/foreign_items.rs

@@ -267,6 +267,14 @@ pub trait EvalContextExt<'mir, 'tcx: 'mir>: crate::MiriInterpCxExt<'mir, 'tcx> {

                this.Sleep(timeout)?;
            }
+            "CreateWaitableTimerExW" => {
+                let [_attributes, _name, _flags, _access] =
+                    this.check_shim(abi, Abi::System { unwind: false }, link_name, args)?;


Generally we do at least read all the arguments of a shim to make sure they are initialized -- or we mark the shim as "only run in std".

Ok, I've added the reads.

Thanks! LGTM.

joboet · 2023-10-06T08:53:31Z

library/std/src/sys/windows/thread.rs

@@ -87,7 +88,17 @@ impl Thread {
    }

    pub fn sleep(dur: Duration) {
-        unsafe { c::Sleep(super::dur2timeout(dur)) }
+        fn high_precision_sleep(dur: Duration) -> Result<(), ()> {
+            let timer = WaitableTimer::high_resolution()?;


How good is the performance of CreateWaitableTimerExW? If it is slow, it might be better to cache the timer in a thread local.

It's around 2,000 nanoseconds on my machine (minus the time it takes to call QueryPerformanceCounter). That is well within noise for waiting on my machine. Granted, other hardware may be able to perform better but I'd rather avoid adding a thread local unless or until someone can show it's worth it.

This function will always fail, allowing std's `Sleep` fallback path to be taken instead.

thomcc · 2023-10-23T23:44:30Z

Looks fine. My recollection is there's sometimes some other steps to do to get the most accurate possible timing (such as adjusting thread priority before/after sleep), but that probably has negative consequences for power usage.

@bors r+

bors · 2023-10-23T23:44:33Z

📌 Commit fbf2567 has been approved by thomcc

It is now in the queue for this repository.

bors · 2023-10-24T11:14:18Z

⌛ Testing commit fbf2567 with merge cee6db1...

bors · 2023-10-24T13:08:12Z

☀️ Test successful - checks-actions
Approved by: thomcc
Pushing cee6db1 to master...

rust-timer · 2023-10-24T14:43:39Z

Finished benchmarking commit (cee6db1): comparison URL.

Overall result: no relevant changes - no action needed

@rustbot label: -perf-regression

Instruction count

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

Max RSS (memory usage)

Results

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)	1.2%	[1.2%, 1.2%]	1
Improvements ✅ (primary)	-3.6%	[-3.6%, -3.6%]	1
Improvements ✅ (secondary)	-2.0%	[-2.0%, -2.0%]	1
All ❌✅ (primary)	-3.6%	[-3.6%, -3.6%]	1

Cycles

Results

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)	-7.1%	[-7.1%, -7.1%]	1
All ❌✅ (primary)	-	-	0

Binary size

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

Bootstrap: missing data
Artifact size: 304.37 MiB -> 304.32 MiB (-0.02%)

klensy · 2023-10-25T08:17:22Z

Should there be documentation update, describing updated behavior?

ChrisDenton · 2023-10-25T09:51:35Z

Hm, this should probably be in the release notes. Not sure if anything changed documentation wise? I intentionally kept the behaviour described there (i.e. zero duration calls Sleep(0)).

klensy · 2023-10-25T13:33:23Z

https://doc.rust-lang.org/nightly/std/thread/fn.sleep.html#platform-specific-behavior
This part.

Currently, specifying a zero duration on Unix platforms returns immediately without invoking the underlying [nanosleep](https://linux.die.net/man/2/nanosleep) syscall, whereas on Windows platforms the underlying [Sleep](https://docs.microsoft.com/en-us/windows/win32/api/synchapi/nf-synchapi-sleep) syscall is always invoked. If the intention is to yield the current time-slice you may want to use [yield_now](https://doc.rust-lang.org/nightly/std/thread/fn.yield_now.html) instead.

ChrisDenton · 2023-10-25T15:30:16Z

Currently, specifying a zero duration on... Windows platforms the underlying Sleep syscall is always invoked.

This should still be true even after this PR

Pkgsrc changes: * Adjust patches and cargo checksums to new versions. * For an external LLVM, set dependency of llvm >= 16, in accordance with the upstream changes. * Mark that on NetBSD we now need >= 9.0, so 8.x is no longer supported. * On NetBSD/sparc64 10.x, we now need GCC 12 to build the embedded LLVM, which is version 17; apparently GCC 10.4 or 10.5 mis-compiles it, resulting in an illegal instruction fault during the build. Ref. rust-lang/rust#117231 Upstream changes: Version 1.75.0 (2023-12-28) ========================== - [Stabilize `async fn` and return-position `impl Trait` in traits.] (rust-lang/rust#115822) - [Allow function pointer signatures containing `&mut T` in `const` contexts.] (rust-lang/rust#116015) - [Match `usize`/`isize` exhaustively with half-open ranges.] (rust-lang/rust#116692) - [Guarantee that `char` has the same size and alignment as `u32`.] (rust-lang/rust#116894) - [Document that the null pointer has the 0 address.] (rust-lang/rust#116988) - [Allow partially moved values in `match`.] (rust-lang/rust#103208) - [Add notes about non-compliant FP behavior on 32bit x86 targets.] (rust-lang/rust#113053) - [Stabilize ratified RISC-V target features.] (rust-lang/rust#116485) Compiler -------- - [Rework negative coherence to properly consider impls that only partly overlap.] (rust-lang/rust#112875) - [Bump `COINDUCTIVE_OVERLAP_IN_COHERENCE` to deny, and warn in dependencies.] (rust-lang/rust#116493) - [Consider alias bounds when computing liveness in NLL.] (rust-lang/rust#116733) - [Add the V (vector) extension to the `riscv64-linux-android` target spec.] (rust-lang/rust#116618) - [Automatically enable cross-crate inlining for small functions] (rust-lang/rust#116505) - Add several new tier 3 targets: - [`csky-unknown-linux-gnuabiv2hf`] (rust-lang/rust#117049) - [`i586-unknown-netbsd`] (rust-lang/rust#117170) - [`mipsel-unknown-netbsd`] (rust-lang/rust#117356) Refer to Rust's [platform support page][platform-support-doc] for more information on Rust's tiered platform support. Libraries --------- - [Override `Waker::clone_from` to avoid cloning `Waker`s unnecessarily.] (rust-lang/rust#96979) - [Implement `BufRead` for `VecDeque<u8>`.] (rust-lang/rust#110604) - [Implement `FusedIterator` for `DecodeUtf16` when the inner iterator does.] (rust-lang/rust#110729) - [Implement `Not, Bit{And,Or}{,Assign}` for IP addresses.] (rust-lang/rust#113747) - [Implement `Default` for `ExitCode`.] (rust-lang/rust#114589) - [Guarantee representation of None in NPO] (rust-lang/rust#115333) - [Document when atomic loads are guaranteed read-only.] (rust-lang/rust#115577) - [Broaden the consequences of recursive TLS initialization.] (rust-lang/rust#116172) - [Windows: Support sub-millisecond sleep.] (rust-lang/rust#116461) - [Fix generic bound of `str::SplitInclusive`'s `DoubleEndedIterator` impl] (rust-lang/rust#100806) - [Fix exit status / wait status on non-Unix `cfg(unix)` platforms.] (rust-lang/rust#115108) Stabilized APIs --------------- - [`Atomic*::from_ptr`] (https://doc.rust-lang.org/stable/core/sync/atomic/struct.AtomicUsize.html#method.from_ptr) - [`FileTimes`] (https://doc.rust-lang.org/stable/std/fs/struct.FileTimes.html) - [`FileTimesExt`] (https://doc.rust-lang.org/stable/std/os/windows/fs/trait.FileTimesExt.html) - [`File::set_modified`] (https://doc.rust-lang.org/stable/std/fs/struct.File.html#method.set_modified) - [`File::set_times`] (https://doc.rust-lang.org/stable/std/fs/struct.File.html#method.set_times) - [`IpAddr::to_canonical`] (https://doc.rust-lang.org/stable/core/net/enum.IpAddr.html#method.to_canonical) - [`Ipv6Addr::to_canonical`] (https://doc.rust-lang.org/stable/core/net/struct.Ipv6Addr.html#method.to_canonical) - [`Option::as_slice`] (https://doc.rust-lang.org/stable/core/option/enum.Option.html#method.as_slice) - [`Option::as_mut_slice`] (https://doc.rust-lang.org/stable/core/option/enum.Option.html#method.as_mut_slice) - [`pointer::byte_add`] (https://doc.rust-lang.org/stable/core/primitive.pointer.html#method.byte_add) - [`pointer::byte_offset`] (https://doc.rust-lang.org/stable/core/primitive.pointer.html#method.byte_offset) - [`pointer::byte_offset_from`] (https://doc.rust-lang.org/stable/core/primitive.pointer.html#method.byte_offset_from) - [`pointer::byte_sub`] (https://doc.rust-lang.org/stable/core/primitive.pointer.html#method.byte_sub) - [`pointer::wrapping_byte_add`] (https://doc.rust-lang.org/stable/core/primitive.pointer.html#method.wrapping_byte_add) - [`pointer::wrapping_byte_offset`] (https://doc.rust-lang.org/stable/core/primitive.pointer.html#method.wrapping_byte_offset) - [`pointer::wrapping_byte_sub`] (https://doc.rust-lang.org/stable/core/primitive.pointer.html#method.wrapping_byte_sub) These APIs are now stable in const contexts: - [`Ipv6Addr::to_ipv4_mapped`] (https://doc.rust-lang.org/stable/core/net/struct.Ipv6Addr.html#method.to_ipv4_mapped) - [`MaybeUninit::assume_init_read`] (https://doc.rust-lang.org/stable/core/mem/union.MaybeUninit.html#method.assume_init_read) - [`MaybeUninit::zeroed`] (https://doc.rust-lang.org/stable/core/mem/union.MaybeUninit.html#method.zeroed) - [`mem::discriminant`] (https://doc.rust-lang.org/stable/core/mem/fn.discriminant.html) - [`mem::zeroed`] (https://doc.rust-lang.org/stable/core/mem/fn.zeroed.html) Cargo ----- - [Add new packages to `[workspace.members]` automatically.] (rust-lang/cargo#12779) - [Allow version-less `Cargo.toml` manifests.] (rust-lang/cargo#12786) - [Make browser links out of HTML file paths.] (rust-lang/cargo#12889) Rustdoc ------- - [Accept less invalid Rust in rustdoc.] (rust-lang/rust#117450) - [Document lack of object safety on affected traits.] (rust-lang/rust#113241) - [Hide `#[repr(transparent)]` if it isn't part of the public ABI.] (rust-lang/rust#115439) - [Show enum discriminant if it is a C-like variant.] (rust-lang/rust#116142) Compatibility Notes ------------------- - [FreeBSD targets now require at least version 12.] (rust-lang/rust#114521) - [Formally demote tier 2 MIPS targets to tier 3.] (rust-lang/rust#115238) - [Make misalignment a hard error in `const` contexts.] (rust-lang/rust#115524) - [Fix detecting references to packed unsized fields.] (rust-lang/rust#115583) - [Remove support for compiler plugins.] (rust-lang/rust#116412)

Rust 1.75 changed thread::sleep to use high-resolution timers on Windows instead of Sleep() see rust-lang/rust#116461

rustbot assigned thomcc Oct 5, 2023

rustbot added O-windows Operating system: Windows S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-libs Relevant to the library team, which will review and decide on the PR/issue. labels Oct 5, 2023

This comment has been minimized.

Sign in to view

ChrisDenton force-pushed the sleep branch 2 times, most recently from 571b4cb to b9255cb Compare October 6, 2023 03:53

Windows: Support sub-millisecond sleep

fca182b

Use `CreateWaitableTimerExW` with `CREATE_WAITABLE_TIMER_HIGH_RESOLUTION`. Does not work before Windows 10, version 1803 so in that case we fallback to using `Sleep`.

ChrisDenton force-pushed the sleep branch from b9255cb to de31b40 Compare October 6, 2023 05:07

RalfJung reviewed Oct 6, 2023

View reviewed changes

joboet reviewed Oct 6, 2023

View reviewed changes

Miri: Add CreateWaitableTimerEx stub

fbf2567

This function will always fail, allowing std's `Sleep` fallback path to be taken instead.

ChrisDenton force-pushed the sleep branch from de31b40 to fbf2567 Compare October 6, 2023 12:10

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 Oct 23, 2023

bors added the merged-by-bors This PR was explicitly merged by bors. label Oct 24, 2023

bors merged commit cee6db1 into rust-lang:master Oct 24, 2023
11 checks passed

rustbot added this to the 1.75.0 milestone Oct 24, 2023

ChrisDenton deleted the sleep branch October 24, 2023 14:52

ChrisDenton added the relnotes Marks issues that should be documented in the release notes of the next release. label Oct 25, 2023

github-actions bot mentioned this pull request Oct 31, 2023

Automated pull from upstream master ferrocene/ferrocene#74

Merged

alexheretic mentioned this pull request Jan 2, 2024

Use Windows WaitableTimer with high resolution if available alexheretic/spin-sleep#16

Closed

poszu mentioned this pull request Mar 18, 2024

Bump rust to 1.76 and disable sleeping between kernels on Win spacemeshos/post-rs#210

Merged

jsgroth added a commit to jsgroth/jgenesis that referenced this pull request Dec 20, 2024

remove Windows-specific sleep code; no longer necessary as of Rust 1.75

3f46e65

Rust 1.75 changed thread::sleep to use high-resolution timers on Windows instead of Sleep() see rust-lang/rust#116461

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Windows: Support sub-millisecond sleep #116461

Windows: Support sub-millisecond sleep #116461

ChrisDenton commented Oct 5, 2023

rustbot commented Oct 5, 2023

This comment has been minimized.

ChrisDenton commented Oct 5, 2023 •

edited

Loading

rustbot commented Oct 6, 2023

This comment has been minimized.

RalfJung Oct 6, 2023

ChrisDenton Oct 6, 2023

RalfJung Oct 6, 2023

joboet Oct 6, 2023

ChrisDenton Oct 6, 2023

thomcc commented Oct 23, 2023

bors commented Oct 23, 2023

bors commented Oct 24, 2023

bors commented Oct 24, 2023

rust-timer commented Oct 24, 2023

klensy commented Oct 25, 2023

ChrisDenton commented Oct 25, 2023

klensy commented Oct 25, 2023 •

edited

Loading

ChrisDenton commented Oct 25, 2023

Windows: Support sub-millisecond sleep #116461

Windows: Support sub-millisecond sleep #116461

Conversation

ChrisDenton commented Oct 5, 2023

rustbot commented Oct 5, 2023

This comment has been minimized.

ChrisDenton commented Oct 5, 2023 • edited Loading

rustbot commented Oct 6, 2023

This comment has been minimized.

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

thomcc commented Oct 23, 2023

bors commented Oct 23, 2023

bors commented Oct 24, 2023

bors commented Oct 24, 2023

rust-timer commented Oct 24, 2023

Overall result: no relevant changes - no action needed

klensy commented Oct 25, 2023

ChrisDenton commented Oct 25, 2023

klensy commented Oct 25, 2023 • edited Loading

ChrisDenton commented Oct 25, 2023

ChrisDenton commented Oct 5, 2023 •

edited

Loading

klensy commented Oct 25, 2023 •

edited

Loading