Make more float intrinsics generic by N1ark · Pull Request #162395 · rust-lang/rust

N1ark · 2026-09-07T00:48:19Z

Another (big) step of #153834 + #160989

Makemore remaining float intrinsics generic! Each commit is self-contained.

In order:

ceil, floor, trunc, round, round_ties_even, with const libm fallbacks (using #[rustc_do_not_const_check], see Intrinsics should allow a non-const fallback body if hooked for CTFE #150961, cc @RalfJung @tgross35)
sqrt, with libm fallbacks
powf, with libm fallbacks
fma and fmuladd, with libm fallbacks (+#[rustc_do_not_const_check]) and a shared fallback, respectively.
powi, with no fallback (it already didn't have any)

The first commit just removes the f16/f128 fallbacks for fabs, as mentioned here. I didn't try making the backends use more fallbacks, to make the PR less risky since it's quite large. Happy to do changes there though (cc @folkertdev)

copysign, minimum_number_nsz, maximum_number_nsz, minimum, maximum will be handled separately as they require rustc_allow_const_fn_unstable: #162414

rustbot · 2026-09-07T00:48:22Z

Some changes occurred to the CTFE / Miri interpreter

cc @rust-lang/miri

miri is developed in its own repository. If the Miri part of this change can be broken out, consider making this change to rust-lang/miri instead. However, if Miri needs adjusting for rustc changes, just ignore this message.

cc @rust-lang/miri

rustc_codegen_gcc is developed in its own repository. If possible, consider making this change to rust-lang/rustc_codegen_gcc instead.

cc @antoyo, @GuillaumeGomez

compiler-builtins is developed in its own repository. If possible, consider making this change to rust-lang/compiler-builtins instead.

cc @tgross35

Some changes occurred to the CTFE machinery

cc @RalfJung, @oli-obk, @lcnr

stdarch is developed in its own repository. If possible, consider making this change to rust-lang/stdarch instead.

cc @Amanieu, @folkertdev, @sayantn

Some changes occurred to the intrinsics. Make sure the CTFE / Miri interpreter
gets adapted for the changes, if necessary.

cc @rust-lang/miri, @RalfJung, @oli-obk, @lcnr

⚠️ #[miri::intrinsic_fallback_is_spec] must only be used if the function actively checks for all UB cases,
and explores the possible non-determinism of the intrinsic.

cc @rust-lang/miri

Any special-casing of Miri in the standard library requires review.

cc @rust-lang/miri

rustc_codegen_cranelift is developed in its own repository. If possible, consider making this change to rust-lang/rustc_codegen_cranelift instead.

cc @bjorn3

⚠️ #[rustc_intrinsic_const_stable_indirect] controls whether intrinsics can be exposed to stable const
code; adding it needs t-lang approval.

cc @rust-lang/wg-const-eval

rustbot · 2026-09-07T00:48:26Z

r? @JohnTitor

rustbot has assigned @JohnTitor.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

Owners of files modified in this PR: libs
libs expanded to 12 candidates
Random selection from Darksonn, JohnTitor, Mark-Simulacrum, clarfonthey

RalfJung · 2026-09-07T06:47:20Z

+#[rustc_const_unstable(feature = "core_intrinsics", issue = "none")]
 #[rustc_intrinsic_const_stable_indirect]


Noted this required adding #[rustc_const_unstable] to the whole intrinsic,

Why was this needed? I guess it is because the body now needs unstable const features, namely const traits? We should ensure const traits are stable enough before we do that.
(I now notice that you already did the same for fabs. We should have already checked this there then but forgot.)

@BoxyUwU @lcnr @oli-obk are const traits in a state where we can use them internally in const functions that can be called by public monomorphic stable const functions? I think this still allows us to change the syntax and logic for const traits pretty much arbitrarily as long as we keep some way of invoking trait functions in a const fn.

View changes since the review

yep, because it needs const_trait_impl, const_ops, const_cmp

these were moved to #162414

cc @fee1-dead

This is not the only case of such usage. Equivalent behaviour is here to stay and syntactical changes don't cause bootstrap churn anymore, so to me personally it is fine to add more, especially for intrinsic bodies which can always be replaced by an actual impl again

RalfJung · 2026-09-07T06:49:47Z

+#[rustc_const_unstable(feature = "core_intrinsics", issue = "none")]
 #[rustc_intrinsic_const_stable_indirect]


This combination of attributes is a strange one.

#[rustc_const_unstable(feature = "core_intrinsics", issue = "none")] #[rustc_intrinsic_const_stable_indirect]

I don't think I ever intended for this combination to be used. It looks quite unsafe, from a const-stability perspective. We should disallow it and require something like rustc_allow_const_fn_unstable instead.

View changes since the review

fabs seems to be the only intrinsic using this. I am quite uncomfortable with adding more such intrinsics. If possible, we should remove this from fabs as well. Unfortunately I won't have time to work on fixing this hole in our const stability checks any time soon. (Well I can easily add the check that rejects this but probably cannot explore adding alternatives.)

i admit this is me just trying to understand how to make things compile. very happy to do something else or exploring adding/fixing an attribute

without #[rustc_const_unstable(feature = "core_intrinsics", issue = "none")], we get "const function that might be (indirectly) exposed to stable cannot use #[feature(const_trait_impl)]"

without #[rustc_intrinsic_const_stable_indirect], we get "const function that might be (indirectly) exposed to stable cannot use #[feature(core_intrinsics)]"

fixed (folded into the relevant commits), there now is rustc_allow_const_fn_unstable; i guess this needs some approval

rustbot · 2026-09-07T08:16:39Z

clippy is developed in its own repository. If possible, consider making this change to rust-lang/rust-clippy instead.

cc @rust-lang/clippy

RalfJung · 2026-09-07T09:58:03Z

Please also update the PR description to reflect which intrinsics are actually being changed here now.

RalfJung · 2026-09-07T10:00:43Z

    where
        F: rustc_apfloat::Float + rustc_apfloat::FloatConvert<F> + Into<Scalar<M::Provenance>>,
    {
-        let x: F = x.to_float()?;


By removing this here, now every caller needs to call to_float first... is that really better?

View changes since the review

i found it nicer, because it means outside of SIMD we only do the x: F = x.to_float()? thing in one place, avoiding additional wrappers or calls elsewhere. it does make the SIMD intrinsics slightly bigger but i think it's nicer for the more general case

beetrees · 2026-09-07T10:19:06Z

 #[rustc_intrinsic]
 #[rustc_nounwind]
-pub fn powif128(a: f128, x: i32) -> f128;
+pub fn powi<T: bounds::FloatPrimitive>(a: T, x: i32) -> T;


Could this intrinsic_dispatch_on_type! calling __powi[sdtk]f2 in the fallback body?

View changes since the review

yes sure :) for f16, powi(x as f32, y) as f16 is correct right?

note that this requires declaring an extern block for these functions, since we dont have a compiler-builtins equivalent to libm. i'm happy adding a library/core/src/num/imp/builtins.rs for it, if you think it's not too much

powi has no precision guarantees, so yes. You'll also need to put #[cfg_attr(any(target_arch = "powerpc", target_arch = "powerpc64"), link_name = "__powikf2")] on the __powitf2 extern fn as PowerPC uses slightly different names for f128 intrinsics.

Adding library/core/src/num/imp/builtins.rs seems fine, e.g. complex number multiplication/division will also need to make direct calls to compiler-builtins.

Done in b6e3544

beetrees · 2026-09-07T11:07:18Z

@@ -0,0 +1,14 @@
+//! Bindings to functions provided by `compiler-builtins`.
+//!
+//! These are only ever provided by `compiler-builtins`, which is always linked in.


This isn't entirely true, compiler-builtins provides an implementation with weak linkage, meaning the implementation from e.g. libgcc_s or LLVM compiler-rt could also end up being used.

View changes since the review

how about this?

//! Bindings to functions provided by `compiler-builtins`. //! //! These are weakly provided by `compiler-builtins`, which is always weakly linked in, or //! possibly by e.g. libgcc_s or compiler-rt.

cranelift: use fallbacks for abs f16, f128

18d2205

rustbot assigned JohnTitor Sep 7, 2026

This comment has been minimized.

Sign in to view

RalfJung reviewed Sep 7, 2026

View reviewed changes

Comment thread library/core/src/intrinsics/mod.rs

RalfJung reviewed Sep 7, 2026

View reviewed changes

N1ark force-pushed the more-float-generics branch from d78924b to 06625b8 Compare September 7, 2026 08:16

rustbot added the T-clippy Relevant to the Clippy team. label Sep 7, 2026

N1ark force-pushed the more-float-generics branch from 06625b8 to da4599d Compare September 7, 2026 08:18

This comment has been minimized.

Sign in to view

N1ark force-pushed the more-float-generics branch 2 times, most recently from ca3f9a9 to b4275be Compare September 7, 2026 09:12

This comment was marked as resolved.

Sign in to view

RalfJung reviewed Sep 7, 2026

View reviewed changes

Comment thread library/core/src/intrinsics/mod.rs Outdated

This comment has been minimized.

Sign in to view

N1ark force-pushed the more-float-generics branch from b4275be to 49057d4 Compare September 7, 2026 09:53

N1ark changed the title ~~Make all float intrinsics generics~~ Make more float intrinsics generic Sep 7, 2026

This comment has been minimized.

Sign in to view

RalfJung reviewed Sep 7, 2026

View reviewed changes

Comment thread compiler/rustc_const_eval/src/interpret/intrinsics.rs

RalfJung reviewed Sep 7, 2026

View reviewed changes

N1ark force-pushed the more-float-generics branch from 49057d4 to 2bb83cf Compare September 7, 2026 10:10

N1ark added 6 commits September 7, 2026 11:12

Make rounding intrinsics generic

da81486

Make sqrt generic

c9d0c7a

Make powf generic

605acfb

Make fma and fmuladd generic

66614fa

Make powi generic

5f1bc1c

Add test for rustc_do_not_const_check intrinsics

33503c2

N1ark force-pushed the more-float-generics branch from 2bb83cf to 33503c2 Compare September 7, 2026 10:12

beetrees reviewed Sep 7, 2026

View reviewed changes

N1ark mentioned this pull request Sep 7, 2026

Make min/max and copysign intrinsics generic #162414

Open

Add a fallback to powi

b6e3544

beetrees reviewed Sep 7, 2026

View reviewed changes

		#[rustc_const_unstable(feature = "core_intrinsics", issue = "none")]
		#[rustc_intrinsic_const_stable_indirect]

Uh oh!

Uh oh!

Conversation

N1ark commented Sep 7, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Sep 7, 2026

Uh oh!

rustbot commented Sep 7, 2026

Uh oh!

This comment has been minimized.

Uh oh!

RalfJung Sep 7, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

N1ark Sep 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

RalfJung Sep 7, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

RalfJung Sep 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rustbot commented Sep 7, 2026

Uh oh!

This comment has been minimized.

This comment was marked as resolved.

Uh oh!

This comment has been minimized.

RalfJung commented Sep 7, 2026

Uh oh!

This comment has been minimized.

Uh oh!

RalfJung Sep 7, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

beetrees Sep 7, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

beetrees Sep 7, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

N1ark commented Sep 7, 2026 •

edited by rustbot

Loading

RalfJung Sep 7, 2026 •

edited by rustbot

Loading

N1ark Sep 7, 2026 •

edited

Loading

RalfJung Sep 7, 2026 •

edited by rustbot

Loading

RalfJung Sep 7, 2026 •

edited

Loading

RalfJung Sep 7, 2026 •

edited by rustbot

Loading

beetrees Sep 7, 2026 •

edited by rustbot

Loading

beetrees Sep 7, 2026 •

edited by rustbot

Loading