I've felt before that compilers often don't put much effort into optimizing the "trivial" cases.
Overly dramatic title for the content, though. I would have clicked "Async Rust Optimizations the Compiler Still Misses" too you know
The author seems to be obsessing about the overhead for trivial functions. He's bothered by overhead for states for "panicked" and "returned". That's not a big problem. Most useful async blocks are big enough that the overhead for the error cases disappears.
He may have a point about lack of inlining. But what tends to limit capacity for large numbers of activities is the state space required per activity.
Is it really though?
In my experience many Rust applications/libraries can be quite heavy on the indirection. One of the points from the article is that contrary to sync Rust, in async Rust each indirection has a runtime cost. Example from the article:
async fn bar(blah: SomeType) -> OtherType {
foo(blah).await
}
Depends somewhat on your expectations, I suppose. Compared to Python, Java, sure, but Rust off course strives to offer "zero-cost" high level concepts.
I think the critique is in the same realm of C++'s std::function. Convenience, sure, but far from zero-cost.
Most useful async blocks are deeply nested, so the overhead compounds rapidly. Check the size of futures in a decently large Tokio codebase sometime
not just too dramatic
given that all the things they list are
non essential optimizations,
and some fall under "micro optimizations I wouldn't be sure rust even wants",
and given how far the current async is away from it's old MVP state,
it's more like outright dishonest then overly dramatic
like the kind of click bait which is saying the author does cares neither about respecting the reader nor cares about honest communication, which for someone wanting to do open source contributions is kinda ... not so clever
through in general I agree rust should have more HIR/MIR optimizations, at least in release mode. E.g. its very common that a async function is not pub and in all places directly awaited (or other wise can be proven to only be called once), in that case neither `Returned` nor `Panicked` is needed, as it can't be called again after either. Similar `Unresumed` is not needed either as you can directly call the code up to the first await (and with such a transform their points about "inlining" and "asyncfns without await still having a state machine" would also "just go away"TM, at least in some places.). Similar the whole `.map_or(a,b)` family of functions is IMHO a anti-pattern, introducing more function with unclear operator ordering and removal of the signaling `unwrap_` and no benefits outside of minimal shortening a `.map(b).unwrap_or(a)` and some micro opt. is ... not productive on a already complicated language. Instead guaranteed optimizations for the kind of patterns a `.map(b).unwrap_or(a)` inline to would be much better.
So threads was the right programming model.
Now language runtimes prefer “green threads” for portability and performance but most languages don’t provide that properly. Instead we have awkward coloring of async/non-async and all these problems around scheduling, priority, and no-preemption. It’s a worse scheduling and process model than 1970.
There is one hill I'll die on, as far as programming languages go, which is that more people should study Céu's structured synchronous concurrency model. It specifically was designed to run on microcontrollers: it compiles down to a finite state machine with very little memory overhead (a few bytes per event).
It has some limitations in terms of how its "scheduler" scales when there are many trails activated by the same event, but breaking things up into multiple asynchronous modules would likely alleviate that problem.
I'm certain a language that would suppprt the "Globally Asynchronous, Locally Synchronous" (GALS) paradigm could have their cake and eat it too. Meaning something that combines support for a green threading model of choice for async events, with structured local reactivity a la Céu.
F'Santanna, the creator of Céu, actually has been chipping away at a new programming language called Atmos that does support the GALS paradigm. However, it's a research language that compiles to Lua 5.4. So it won't really compete with the low-level programming languages there.
Sure, but once you involve the kernel and OS scheduler things get 3 to 4 orders of magnitude slower than what they should be.
The last time I was working on our coroutine/scheduling code creating and joining a thread that exited instantly was ~200us, and creating one of our green threads, scheduling it and waiting for it was ~400ns.
You don't need to wait 10 years for someone else to design yet another absurdly complex async framework, you can roll your own green threads/stackful coroutines in any systems language with 20 lines of ASM.
2. Unchecked array operations are a lot faster. Manual memory management is a lot faster. Shared memory is a lot faster.
Usually when you see someone reach for sharp and less expressive tools it’s justified by a hot code path. But here we jump immediately to the perf hack?
3. How many simultaneous async operations does your program have?
For example, if you don't explicitly call the java.awt.Toolkit.sync() method after updating the UI state (which according to the docs "is useful for animation"), Swing will in my experience introduce seemingly random delays and UI lag because it just doesn't bother sending the UI updates to the window system.
I thought it was because they could copy chromium.
Which inputs are getting latency? The keyboard? The files?
> the non blocking nature
When it comes time to test your concurrent processing, to ensure you handle race conditions properly, that is much easier with callbacks because you can control their scheduling. Since each callback represents a discrete unit, you see which events can be reordered. This enables you to more easily consider all the different orderings.
Instead with threads it is easy to just ignore the orderings and not think about this complexity happening in a different thread and when it can influence the current thread. It isn't simpler, it is simplistic. Moreover, you cannot really change the scheduling and test the concurrent scenarios without introducing artificial barriers to stall the threads or stubbing the I/O so you can pass in a mock that you will then instrument with a callback to control the ordering...
The problem with callbacks is that the call stack when captured isn't the logical callstack unless you are in one of the few libraries/runtimes that put in the work to make the call stacks make sense. Otherwise you need good error definitions.
You can of course mix the paradigms and have the worst of both worlds.
Not really. I’ve observed async code often is written in such a way that it doesn’t maximize how much concurrency can be expressed (eg instead of writing “here’s N I/O operations to do them all concurrently” it’s “for operation X, await process(x)”). However, in a threaded world this concurrency problem gets worse because you have no way to optimize towards such concurrency - threads are inherently and inescapably too heavy weight to express concurrency in an efficient way.
This is is not a new lesson - work stealing executors have long been known to offer significantly lower latency with more consistent P99 than traditional threads. This has been known since forever - in the early 00s this is why Apple developed GCD. Threads simply don’t provide any richer information it needs in the scheduler to the kernel about the workload and kernel threads are an insanely heavy mechanism for achieving fine grained concurrency and even worse when this concurrency is I/O or a mixed workload instead of pure compute that’s embarrassingly easily to parallelize.
Do all programs need this level of performance? No, probably not. But it is significantly more trivial to achieve a higher performance bar and in practice achieve a latency and throughput level that traditional approaches can’t match with the same level of effort.
You can tell async is directionally kind of correct in that io_uring is the kernel’s approach to high performance I/O and it looks nothing like traditional threading and syscalls and completion looks a lot closer to async concurrency (although granted exploiting it fully is much harder in an async world because async/await is an insufficient number of colors to express how async tasks interrelate)
But as you observed, async/await fails to express concurrency any better. It’s also a thread, it’s just a worse implementation.
Every explanation of the feature starts with managing callback hell.
For problems that aren't overly concerned with performance/memory, yes. You should probably reach for threads as a default, unless you know a priori that your problem is not in this common bucket.
Unfortunately there is quite a lot of bookkeeping overhead in the kernel for threads, and context switches are fairly expensive, so in a number of high performance scenarios we may not be able to afford kernel threading
But what you said about kernel implementation is true. But are we really saying that the primary motivation for async/await is performance? How many programmers would give that answer? How many programs are actually hitting that bottleneck?
Doesn’t that buck the trend of every other language development in the past 20 years, emphasizing correctness and expressively over raw performance?
The original motivation for not using OS threads was indeed performance. Async/await is mostly syntax sugar to fix some of the ergonomic problems of writing continuation-based code (Rust more or less skipped the intermediate "callback hell" with futures that Javascript/Python et al suffered through).
It's all nuanced and what to choose requires careful evaluation.
Of course - what else would it be? The whole async trend started because moving away from each http request spawning (or being bound to) an OS thread gave quite extreme improvements in requests/second metrics, didn't it?
What I question is whether 1. Most programs resemble that, so that they make it an invasive feature of every general purpose language. 2. Whether programmers are making a conscious choice because they ruled out the perf overhead of the simpler model we have by default.
Examples in the blog seem too simple make any conclusions
So yes, it does really matter. Keep in mind that optimizations stack. We're preventing LLVM from doing it's thing. So if we make the futures themselves smaller, LLVM will be able to optimize more. So small changes really compound.
I never really liked the viral nature of async in rust when it was introduced.
I wish rust the best of luck and with more people like this rust could have a brighter future.
You could've deduced that from the fact that someone who puts this amount of energy in a detailed article about intricacies of an area of "foo", quite certainly does not "hate on foo".
The article is fine besides the bait title.
I don't know enough about the domain to be objectively helpful, so it's all wishy-washy feelings on my part. I keep reaching for orchestrating things with threads in Rust where most people would probably reach for async these days. The only language where I've felt fine embracing the blessed async system is Haskell and its green threads (which I understand come with their own host of problems).
The risk they took was very calculated. Unfortunately they’re bad at math and chose the wrong trade-offs.
Ah well. Shit happens.
They chose the exact same tradeoffs as C++'s async/await (and the same overall model as Python/NodeJS), so I'm not sure what that says about programming as a whole.
Not to mention Tokio (most popular runtime for Rust) is multi-threaded by default. So you have to deal with multithreading bugs as well as normal async ones. That is not the case with most async languages. For example both Python and NodeJS use a single thread to execute async code.
I somehow miss noticing that in C++ and I have no idea how it is working in other domains.
My only gripe is that a lot of it is feeling a bit kick-starter-y, with each of the goals needing specific funding. Is that the best model we've found so far?