| Age | Commit message (Collapse) | Author | Lines |
|---|
|
Improve performance of spsc_queue and stream.
This PR makes two main changes:
1. It switches the `spsc_queue` node caching strategy from keeping a shared
counter of the number of nodes in the cache to keeping a consumer only counter
of the number of node eligible to be cached.
2. It separates the consumer and producers fields of `spsc_queue` and `stream` into
a producer cache line and consumer cache line.
Overall, it speeds up `mpsc` in `spsc` mode by 2-10x.
Variance is higher than I'd like (that 2-10x speedup is on one benchmark), I believe this is due to the drop check in `send` (`fn stream::Queue::send:107`). I think this check can be combined with the sleep detection code into a version which only uses 1 shared variable, and only one atomic access per `send`, but I haven't looked through the select implementation enough to be sure.
The code currently assumes a cache line size of 64 bytes. I added a CacheAligned newtype in `mpsc` which I expect to reuse for `shared`. It doesn't really belong there, it would probably be best put in `core::sync::atomic`, but putting it in `core` would involve making it public, which I thought would require an RFC.
Benchmark runner is [here](https://github.com/JLockerman/queues/tree/3eca46279c53eb75833c5ecd416de2ac220bd022/shootout), benchmarks [here](https://github.com/JLockerman/queues/blob/3eca46279c53eb75833c5ecd416de2ac220bd022/queue_bench/src/lib.rs#L170-L293).
Fixes #44512.
|
|
Fixes #44771.
|
|
|
|
Queue::new is only used is tests atm, which causes warnings on emscripten which does not run queue tests.
|
|
This commit makes two main changes.
1. It switches the spsc_queue node caching strategy from keeping a shared
counter of the number of nodes in the cache to keeping a consumer only counter
of the number of node eligible to be cached.
2. It separate the consumer and producers fields of spsc_queue and stream into
a producer cache line and consumer cache line.
|
|
Like #43008 (f668999), but _much more aggressive_.
|
|
Implement Sync for SyncSender
r? @alexcrichton
|
|
|
|
|
|
|
|
Decompose Adjustment into smaller steps and remove the method map.
The method map held method callee information for:
* actual method calls (`x.f(...)`)
* overloaded unary, binary, indexing and call operators
* *every overloaded deref adjustment* (many can exist for each expression)
That last one was a historical ~~accident~~ hack, and part of the motivation for this PR, along with:
* a desire to compose adjustments more freely
* containing the autoderef logic better to avoid mutation within an inference snapshot
* not creating `TyFnDef` types which are incompatible with the original one
* i.e. we used to take a`TyFnDef`'s `for<'a> &'a T -> &'a U` signature and instantiate `'a` using a region inference variable, *then* package the resulting `&'b T -> &'b U` signature in another `TyFnDef`, while keeping *the same* `DefId` and `Substs`
* to fix #3548 by explicitly writing autorefs for the RHS of comparison operators
Individual commits tell their own story, of "atomic" changes avoiding breaking semantics.
Future work based on this PR could include:
* removing the signature from `TyFnDef`, now that it's always "canonical"
* some questions of variance remain, as subtyping *still* treats the signature differently
* moving part of the typeck logic for methods, autoderef and coercion into `rustc::traits`
* allowing LUB coercions (joining multiple expressions) to "stack up" many adjustments
* transitive coercions (e.g. reify or unsize after multiple steps of autoderef)
r? @nikomatsakis
|
|
|
|
|
|
These files are licensed under a different license
than the rest of the codebase. This causes potential
issues and inconveniences.
Relicense these files under the standard license.
I hold original copyright on that code.
Fixes #36556
|
|
Correct some stability versions
These were found by running tidy on stable versions of rust and finding
features stabilised with the wrong version numbers.
|
|
These were found by running tidy on stable versions of rust and finding
features stabilised with the wrong version numbers.
|
|
|
|
|
|
|
|
documentation to channel() and sync_channel(); adding more links #29377
|
|
|
|
|
|
|
|
This change adds links to to `Receiver`, `Iter`, `TryIter`, `IntoIter`,
`Sender`, `SyncSender`, `SendError`, `RecvError`, `TryRecvError`,
`RecvTimeoutError`, `TrySendError`, `Sender::send`, `SyncSender::send`,
`SyncSender::try_send`, `Receiver::recv`, `Receiver::recv_timeout`,
`Receiver::iter`, and `Receiver::try_iter`.
Examples added to `Receiver`, `Sender`, `Receiver::iter`.
|
|
|
|
This was never established as a convention we should follow in the 'More
API Documentation Conventions' RFC:
https://github.com/rust-lang/rfcs/blob/master/text/1574-more-api-documentation-conventions.md
|
|
Implement `fmt::Debug` for all structures in libstd.
Part of https://github.com/rust-lang/rust/issues/31869.
Also turn on the `missing_debug_implementations` lint at the crate
level.
|
|
Replace invalid use of `&mut` with `UnsafeCell` in `std::sync::mpsc`
Closes #36934
r? @alexcrichton
|
|
Part of https://github.com/rust-lang/rust/issues/31869.
Also turn on the `missing_debug_implementations` lint at the crate
level.
|
|
Closes #36934
|
|
|
|
Closes #37915
This commit enhances documentation with several links and
fixes an error in the `sync_channel` documentation as well:
`send` doesn't panic when the senders are all disconnected
|
|
The `bound` argument in `std::sync::mpsc::sync:channel(bound: usize)` was not defined in the documentation.
|
|
|
|
[tracking issue](https://github.com/rust-lang/rust/issues/27749)
|
|
Restore `DISCONNECTED` state in `oneshot::Packet::send`
Closes #32114
I'm not sure if this is the best approach, but the current action of swapping `DISCONNECTED` with `DATA` seems wrong. Additionally, it is strange that the `send` method (and others in the `oneshot` module) takes `&mut self` despite performing atomic operations, as this requires extra discipline to avoid data races and lets us use methods like `AtomicUsize::get_mut` instead of methods that require a memory ordering.
|
|
Closes #32114
|
|
std: Correct stability attributes for some implementations
These are displayed by rustdoc so should be correct.
|
|
This is intended to maintain existing standards of code organization
in hopes that the standard library will continue to be refactored to
isolate platform-specific bits, making porting easier; where "standard
library" roughly means "all the dependencies of the std and test
crates".
This generally means placing restrictions on where `cfg(unix)`,
`cfg(windows)`, `cfg(target_os)` and `cfg(target_env)` may appear,
the basic objective being to isolate platform-specific code to the
platform-specific `std::sys` modules, and to the allocation,
unwinding, and libc crates.
Following are the basic rules, though there are currently
exceptions:
- core may not have platform-specific code
- liballoc_system may have platform-specific code
- liballoc_jemalloc may have platform-specific code
- libpanic_abort may have platform-specific code
- libpanic_unwind may have platform-specific code
- other crates in the std facade may not
- std may have platform-specific code in the following places
- sys/unix/
- sys/windows/
- os/
There are plenty of exceptions today though, noted in the whitelist.
|
|
These are displayed by rustdoc so should be correct.
|
|
|
|
|
|
|
|
Add missing Eq implementations
Part of #36301.
|
|
|
|
|
|
|
|
|
|
Stabilized
* `Cell::as_ptr`
* `RefCell::as_ptr`
* `IpAddr::is_{unspecified,loopback,multicast}`
* `Ipv6Addr::octets`
* `LinkedList::contains`
* `VecDeque::contains`
* `ExitStatusExt::from_raw` - both on Unix and Windows
* `Receiver::recv_timeout`
* `RecvTimeoutError`
* `BinaryHeap::peek_mut`
* `PeekMut`
* `iter::Product`
* `iter::Sum`
* `OccupiedEntry::remove_entry`
* `VacantEntry::into_key`
Deprecated
* `Cell::as_unsafe_cell`
* `RefCell::as_unsafe_cell`
* `OccupiedEntry::remove_pair`
Closes #27708
cc #27709
Closes #32313
Closes #32630
Closes #32713
Closes #34029
Closes #34392
Closes #34285
Closes #34529
|
|
Add a method to the mpsc::Receiver for producing a non-blocking iterator
Currently, the `mpsc::Receiver` offers methods for receiving values in both blocking (`recv`) and non-blocking (`try_recv`) flavours. However only blocking iteration over values is supported. This PR adds a non-blocking iterator to complement the `try_recv` method, just as the blocking iterator complements the `recv` method.
Use-case
-------------
I predominantly use rust in my work on real-time systems and in particular real-time audio generation/processing. I use `mpsc::channel`s to communicate between threads in a purely non-blocking manner. I.e. I might send messages from the GUI thread to the audio thread to update the state of the dsp-graph, or from the audio thread to the GUI thread to display the RMS of each node. These are just a couple examples (I'm probably using 30+ channels across my various projects). I almost exclusively use the `mpsc::Receiver::try_recv` method to avoid blocking any of the real-time threads and causing unwanted glitching/stuttering. Now that I mention it, I can't think of a single time that I personally have used the `recv` method (though I can of course see why it would be useful, and perhaps the common case for many people).
As a result of this experience, I can't help but feel there is a large hole in the `Receiver` API.
| blocking | non-blocking |
|------------|--------------------|
| `recv` | `try_recv` |
| `iter` | 🙀 |
For the most part, I've been working around this using `while let Ok(v) = r.try_recv() { ... }`, however as nice as this is, it is clearly no match for the Iterator API.
As an example, in the majority of my channel use cases I only want to check for *n* number of messages before breaking from the loop so that I don't miss the audio IO callback or hog the GUI thread for too long when an unexpectedly large number of messages are sent. Currently, I have to write something like this:
```rust
let mut take = 100;
while let Ok(msg) = rx.try_recv() {
// Do stuff with msg
if take == 0 {
break;
}
take -= 1;
}
```
or wrap the `try_recv` call in a `Range<usize>`/`FilterMap` iterator combo.
On the other hand, this PR would allow for the following:
```rust
for msg in rx.try_iter().take(100) {
// Do stuff with msg
}
```
I imagine this might also be useful to game devs, embedded or anyone doing message passing across real-time threads.
|
