x86/x86_64 intrinsics seem to be marked stable, but require feature flags

[afetisov]

I tried this code:

use std::arch::x86_64::*;

fn main() {
    unsafe { ud2(); }
}

I expected this code to successfully compile on stable, since all of std::arch, core::arch, core_arch::arch and core_arch::arch::x86/x86_64 are explicitly marked stable since 1.27:

#[doc = include_str!("core_arch_docs.md")]
#[stable(feature = "simd_arch", since = "1.27.0")]
pub mod arch {
    /// Platform-specific intrinsics for the `x86` platform.
    ///
    /// See the [module documentation](../index.html) for more details.
    #[cfg(any(target_arch = "x86", doc))]
    #[doc(cfg(target_arch = "x86"))]
    #[stable(feature = "simd_x86", since = "1.27.0")]
    pub mod x86 {
        #[stable(feature = "simd_x86", since = "1.27.0")]
        pub use crate::core_arch::x86::*;
    }

    /// Platform-specific intrinsics for the `x86_64` platform.
    ///
    /// See the [module documentation](../index.html) for more details.
    #[cfg(any(target_arch = "x86_64", doc))]
    #[doc(cfg(target_arch = "x86_64"))]
    #[stable(feature = "simd_x86", since = "1.27.0")]
    pub mod x86_64 {
        #[stable(feature = "simd_x86", since = "1.27.0")]
        pub use crate::core_arch::x86::*;
        #[stable(feature = "simd_x86", since = "1.27.0")]
        pub use crate::core_arch::x86_64::*;
    }

    /* snip */
}

Instead, I got the error

error[E0658]: use of unstable library feature 'stdsimd'
 --> src\main.rs:4:14
  |
4 |     unsafe { ud2(); }
  |              ^^^
  |
  = note: see issue #48556 <https://github.com/rust-lang/rust/issues/48556> for more information
  = help: add `#![feature(stdsimd)]` to the crate attributes to enable

However, this issue is closed, since x86/x86_64/aarch64 intrinsics were shipped long ago:

We discussed this in today's @rust-lang/lang meeting.

This seems to be the tracking issue for target-specific SIMD (not portable SIMD, which is being tracked and developed elsewhere). And we've shipped target-specific SIMD on x86-64 and aarch64. There will always be more CPUs to support, but that doesn't mean this issue needs to remain open indefinitely.

We've shipped this; closing.

The stdsimd feature flag was also removed from all x86/x86_64 intrinsics a long time ago. However, I can see that the core_arch crate is included as an unstable submodule guarded by #[feature(stdsimd)] inside of libcore:

// Pull in the `core_arch` crate directly into libcore. The contents of
// `core_arch` are in a different repository: rust-lang/stdarch.
//
// `core_arch` depends on libcore, but the contents of this module are
// set up in such a way that directly pulling it here works such that the
// crate uses the this crate as its libcore.
#[path = "../../stdarch/crates/core_arch/src/mod.rs"]
#[allow(
    missing_docs,
    missing_debug_implementations,
    dead_code,
    unused_imports,
    unsafe_op_in_unsafe_fn
)]
#[allow(rustdoc::bare_urls)]
// FIXME: This annotation should be moved into rust-lang/stdarch after clashing_extern_declarations is
// merged. It currently cannot because bootstrap fails as the lint hasn't been defined yet.
#[allow(clashing_extern_declarations)]
#[unstable(feature = "stdsimd", issue = "48556")]
mod core_arch;

Rust version

This issue reproduces on the current stable and nightly toolchains.

rustc 1.61.0 (fe5b13d68 2022-05-18)
binary: rustc
commit-hash: fe5b13d681f25ee6474be29d748c65adcd91f69e
commit-date: 2022-05-18
host: x86_64-pc-windows-msvc
release: 1.61.0
LLVM version: 14.0.0

rustc 1.63.0-nightly (4e725bad7 2022-06-04)
binary: rustc
commit-hash: 4e725bad73747a4c93a3ac53106e4b4006edc665
commit-date: 2022-06-04
host: x86_64-pc-windows-msvc
release: 1.63.0-nightly
LLVM version: 14.0.5

[C0D3-M4513R]

For core::arch::x86_64::{__cpuid, has_cpuid}; it is especially bad.
__cpuid is explicitly stable, and thus works.
has_cpuid doesn't work, since it is not explicitly stable.

That is worrysome for this code (checking for Long Mode).
https://play.rust-lang.org/?version=nightly&mode=debug&edition=2021&gist=11f2501972a1bc4f58cb93f86c4bf93e

[zertyz]

Still a thing in Rust 1.63
plus both core::arch::x86_64::cmpxchg16b & std::arch::x86_64::cmpxchg16b don't work either

[tomaka]

This happens as well for some std::arch::wasm functions, especially memory_atomic_notify

[C0D3-M4513R]

I think this is solved by: ea37e80 / #117372

So this issue can be closed, right?

[afetisov]

Seems so. The ud2 intrinsic was removed, so the original example is no longer valid, but I couldn't create any new counterexample, after a few attempts. Since nothing in x86_64 or x86 is marked as unstable now, and other arches don't seem to have the original inconsistent stable-unstable marking, I guess the issue should really be fixed. Closing.