Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Mark TypeId::of as inline to avoid monomorphizing unnecessary code #74362

Closed
alecmocatta opened this issue Jul 15, 2020 · 5 comments · Fixed by #92135
Closed

Mark TypeId::of as inline to avoid monomorphizing unnecessary code #74362

alecmocatta opened this issue Jul 15, 2020 · 5 comments · Fixed by #92135
Labels
C-enhancement Category: An issue proposing an enhancement or a PR with one. I-slow Issue: Problems and improvements with respect to performance of generated code. T-libs Relevant to the library team, which will review and decide on the PR/issue.

Comments

@alecmocatta
Copy link
Contributor

alecmocatta commented Jul 15, 2020

TypeId can be used to specialize code by comparing types explicitly (TypeId::of::<u8>() == TypeId::of::<T>(), examples in the wild) or implicitly (<dyn Any>::downcast_mut).

In release mode this works well; in debug mode though the unused impl is unnecessarily monomorphized and codegened, slowing compile time as well as run time.

It isn't a major issue as it only affects debug mode, but I wondered if it was worth marking TypeId::of, <TypeId as PartialEq>::eq, <dyn Any>::is and the various downcast methods as inline(always) to potentially reduce compile times and provide a nice debug mode runtime boost?

use std::any::{Any, TypeId};

struct Container<T>(T);

impl<T: 'static> Container<T> {
    fn foo(&mut self) {
        if let Some(_) = <dyn Any>::downcast_mut::<Container<u8>>(self) {
            // specialized impl
            std::process::abort();
        }
        // general impl
    }
    fn bar(&mut self) {
        if TypeId::of::<u8>() == TypeId::of::<T>() {
            // specialized impl
            std::process::abort();
        }
        // general impl
    }
}

fn main() {
    let mut c = Container(String::new());
    c.foo();
    c.bar();
}

See e.g. abort is present in IR/assembly: Playground

@Alexendoo Alexendoo added C-enhancement Category: An issue proposing an enhancement or a PR with one. I-slow Issue: Problems and improvements with respect to performance of generated code. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. T-libs Relevant to the library team, which will review and decide on the PR/issue. and removed T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Jul 22, 2020
@AngelicosPhosphoros
Copy link
Contributor

It would be cool if this calls get inlined in -C opt-level=1. This would improve performance of debug builds of engines like bevy, compiled with weak optimizations.

@AngelicosPhosphoros
Copy link
Contributor

I found that this function always promoted to const on -C opt-level=1
godbolt

Another problem is that ParialEq::eq isn't inlined even if I use #[inline(always)]. Also, manual implementation of PartialEq would remove ability to use it in pattern matching like:
godbolt

#![feature(const_type_id)]

#[inline(always)]
pub const fn is_bool<T: 'static>()->bool{
    use std::any::TypeId;

    const BOOL_ID: TypeId = TypeId::of::<bool>();
    match TypeId::of::<T>(){
        BOOL_ID => true,
        _ => false,
    }
}

pub fn good()->bool{
    is_bool::<bool>()
}

pub fn bad()->bool{
    is_bool::<u8>()
}

@AngelicosPhosphoros
Copy link
Contributor

I think, we should close this issue in favor of #77125

Or in favor of #31844 which allows write such code:

#![feature(specialization)]

use std::marker::PhantomData;

struct IsSameType<T0: 'static, T1: 'static>(PhantomData<T0>, PhantomData<T1>);

trait IsSameTypeTrait{
    const IS_SAME: bool;
}

impl<T0: 'static, T1: 'static> IsSameTypeTrait for IsSameType<T0, T1>{
    default const IS_SAME: bool = false;
}

impl<T> IsSameTypeTrait for IsSameType<T, T>{
    const IS_SAME: bool = true;
}

@AngelicosPhosphoros
Copy link
Contributor

I decided to add #[inline] because we unlikely to get #77125 resolved soon, and #31844 still has soundness problems.

@AngelicosPhosphoros
Copy link
Contributor

AngelicosPhosphoros commented Dec 20, 2021

Adding #[inline] wouldn't have any effect without optimizations unfortunately because there must run InlinerPass which doesn't run on debug.

With New Pass Manager (which I hope would stabilized soon), users may set opt-level=1 for fast debug builds and get profits from this change.

bors added a commit to rust-lang-ci/rust that referenced this issue Dec 24, 2021
…62, r=dtolnay

Add `#[inline]` modifier to `TypeId::of`

It was already inlined but it happened only in 4th InlinerPass on my testcase.
With `#[inline]` modifier it happens on 2nd pass.

Closes rust-lang#74362
@bors bors closed this as completed in 756d163 Dec 24, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
C-enhancement Category: An issue proposing an enhancement or a PR with one. I-slow Issue: Problems and improvements with respect to performance of generated code. T-libs Relevant to the library team, which will review and decide on the PR/issue.
Projects
None yet
Development

Successfully merging a pull request may close this issue.

3 participants