[calebzulawski]

View all comments

Added algebraic intrinsics for SIMD. First half of rust-lang/portable-simd#539

[rustbot]

Some changes occurred to the CTFE / Miri interpreter

cc @rust-lang/miri

Some changes occurred to the CTFE machinery

cc @RalfJung, @oli-obk, @lcnr

Some changes occurred to the platform-builtins intrinsics. Make sure the
LLVM backend as well as portable-simd gets adapted for the changes.

cc @antoyo, @GuillaumeGomez, @bjorn3, @programmerjake

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

[rustbot]

r? @davidtwco

rustbot has assigned @davidtwco.
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: compiler
• compiler expanded to 77 candidates
• Random selection from 21 candidates

[rustbot]

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

[rustbot]

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

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

cc @bjorn3

[programmerjake]

maybe add a comment noting that this doesn't fully randomize the reduce order, since it only randomizes the order of the inputs then does an ordered reduce -- that will never give you an evaluation order like (a + b) + (c + d).

View changes since the review

[programmerjake]

an example implementation that does cover all possible evaluation orders: https://play.rust-lang.org/?version=stable&mode=debug&edition=2024&gist=c2f778ccd6b16d881b5afac8200ffaf7

pub fn random_reduce<T>(mut inputs: Vec<T>, mut reduce: impl FnMut(T, T) -> T) -> Option<T> {
    while inputs.len() >= 2 {
        let lhs_index = rand::random_range(0..inputs.len());
        let lhs = inputs.swap_remove(lhs_index);
        let rhs_index = rand::random_range(0..inputs.len());
        let rhs = inputs.swap_remove(rhs_index);
        inputs.push(reduce(lhs, rhs));
    }
    inputs.pop()
}

[calebzulawski]

I'm curious if this is actually necessary, miri's float nondeterminism doesn't generate every possible distribution of nondeterminism

[programmerjake]

maybe not, but in the case of catastrophic cancellation, you can get wildly different results for the different evaluation orders so imo it would be useful for miri to be able to give you those.

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[calebzulawski]

I think #164924 introduced a minor break to cg_gcc vector frem, I repaired that here

[RalfJung]

Miri and CTFE look good to me :)

View changes since this review