stdarch-gen-hexagon-scalar: Do not emit inline(always) if target_feature is used

[taiki-e]

stdarch-gen-hexagon-scalar, which added in #2072, always emits #[inline(always)], but it cause compile errors on functions that have #[target_feature].

  error: cannot use `#[inline(always)]` with `#[target_feature]`
       --> /home/runner/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/../../stdarch/crates/core_arch/src/hexagon/scalar.rs:11428:1
        |
  11428 | #[inline(always)]
        | ^^^^^^^^^^^^^^^^^
        |
        = note: See this issue for full discussion: https://github.com/rust-lang/rust/issues/145574

hexagon-unknown-linux-musl build is broken due to this issue since nightly-2026-05-22: https://github.com/taiki-e/atomic-maybe-uninit/actions/runs/26266025199/job/77310173101#step:4:929

This PR fixes it by emitting #[inline] instead of #[inline(always)] if #[target_feature] is used.
(Confirmed with cargo +nightly-2026-05-21 build -Z build-std=core -p core_arch --target hexagon-unknown-linux-musl)

This also add missing stdarch-gen output check for stdarch-gen-hexagon-scalar.

cc @androm3da
cc rust-lang/rust#151523
Fixes #2134

[rustbot]

r? @adamgemmell

rustbot has assigned @adamgemmell.
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: @Amanieu, @adamgemmell, @davidtwco, @folkertdev, @sayantn
• @Amanieu, @adamgemmell, @davidtwco, @folkertdev, @sayantn expanded to Amanieu, adamgemmell, davidtwco, folkertdev, sayantn
• Random selection from Amanieu, adamgemmell, davidtwco, folkertdev, sayantn

[taiki-e]

CI failure seems to be unrelated

https://github.com/rust-lang/stdarch/actions/runs/26325152273/job/77501386688?pr=2133

+ cargo fmt --all -- --check
Diff in /home/runner/work/stdarch/stdarch/crates/core_arch/src/loongarch64/simd.rs:17:
 macro_rules! impl_simd_ext {
     ($v:ident, $e:ty) => {
         #[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
-        impl const SimdExt for crate::core_arch::simd::$v {
+        const impl SimdExt for crate::core_arch::simd::$v {
             type Elem = $e;
 
             #[inline(always)]
Diff in /home/runner/work/stdarch/stdarch/crates/core_arch/src/simd.rs:87:
 }
 
 #[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
-impl<T: SimdElement, const N: usize> const crate::cmp::PartialEq for Simd<T, N> {
+const impl<T: SimdElement, const N: usize> crate::cmp::PartialEq for Simd<T, N> {
     #[inline]
     fn eq(&self, other: &Self) -> bool {
         self.as_array() == other.as_array()
Diff in /home/runner/work/stdarch/stdarch/crates/core_arch/src/simd.rs:299:
 }
 
 #[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
-impl<T: SimdElement, const N: usize> const crate::cmp::PartialEq for SimdM<T, N> {
+const impl<T: SimdElement, const N: usize> crate::cmp::PartialEq for SimdM<T, N> {
     #[inline]
     fn eq(&self, other: &Self) -> bool {
         self.as_array() == other.as_array()

I don't know anything about the current status of const impl, but the code in rust-lang/rust is the same, and it seems that both const impl and impl const are valid in the first place? So why is rustfmt making changes here?

• https://github.com/rust-lang/rust/blob/54333ff079780f803f65dcee30c544050b35f544/library/stdarch/crates/core_arch/src/loongarch64/simd.rs#L20
• https://github.com/search?q=repo%3Arust-lang%2Frust+%2Fimpl.*+const%2F&type=code
• https://github.com/search?q=repo%3Arust-lang%2Frust+%2Fconst+impl%2F&type=code

[taiki-e]

CI failure seems to be unrelated

I don't know anything about the current status of const impl, but the code in rust-lang/rust is the same, and it seems that both const impl and impl const are valid in the first place? So why is rustfmt making changes here?

Added #[rustfmt::skip] for now: 44102ca