_mm_movemask_epi8 unexpectedly triggers Undefined Behavior

[Nugine]

Moved from rust-lang/miri#2617

---

playground

use core::arch::x86_64::*;

#[target_feature(enable = "sse2")]
pub unsafe fn is_ascii_sse2(a: &[u8; 16]) -> bool {
    let a = _mm_loadu_si128(a.as_ptr().cast());
    let m = _mm_movemask_epi8(a);
    m == 0
}

fn main() {
    assert!(cfg!(target_feature = "sse2"));
    let s = b"helloworld123456";
    assert!(unsafe { is_ascii_sse2(s) });
}

error: Undefined Behavior: each element of a SIMD mask must be all-0-bits or all-1-bits
    --> /playground/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/../../stdarch/crates/core_arch/src/x86/sse2.rs:1381:5
     |
1381 |     simd_bitmask::<_, u16>(a.as_i8x16()) as u32 as i32
     |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ each element of a SIMD mask must be all-0-bits or all-1-bits
     |
     = help: this indicates a bug in the program: it performed an invalid operation, and caused Undefined Behavior
     = help: see https://doc.rust-lang.org/nightly/reference/behavior-considered-undefined.html for further information
     = note: BACKTRACE:
     = note: inside `std::arch::x86_64::_mm_movemask_epi8` at /playground/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/../../stdarch/crates/core_arch/src/x86/sse2.rs:1381:5
note: inside `is_ascii_sse2` at src/main.rs:6:13

See also:

• https://doc.rust-lang.org/stable/core/arch/x86_64/fn._mm_movemask_epi8.html
• https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm_movemask_epi8

---

Related:

• Use simd_bitmask intrinsic in a couple of places #1331

---

_mm[256]_movemask_epi8 has defined behavior:

Returns a mask of the most significant bit of each element in a.

(regardless of the other bits)

Changing the requirement may introduce unsoundness into existing crates.

[Amanieu]

I commented on the miri issue, I think that one should be reopened.

[yanns]

Sorry for the naive question, but I still have this issue on nightly:

MIRIFLAGS="-Zmiri-disable-isolation -Zmiri-measureme=crox" cargo +nightly miri run
Preparing a sysroot for Miri (target: x86_64-apple-darwin)... done
    Finished dev [unoptimized + debuginfo] target(s) in 0.48s
     Running `/Users/yannsimon/.rustup/toolchains/nightly-x86_64-apple-darwin/bin/cargo-miri runner target/miri/x86_64-apple-darwin/debug/proto_refexp`
error: Undefined Behavior: each element of a SIMD mask must be all-0-bits or all-1-bits
    --> /Users/yannsimon/.rustup/toolchains/nightly-x86_64-apple-darwin/lib/rustlib/src/rust/library/core/src/../../stdarch/crates/core_arch/src/x86/avx2.rs:2004:5
     |
2004 |     simd_bitmask::<_, u32>(a.as_i8x32()) as i32
     |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ each element of a SIMD mask must be all-0-bits or all-1-bits

from:

pub unsafe fn _mm256_movemask_epi8(a: __m256i) -> i32 {
    simd_bitmask::<_, u32>(a.as_i8x32()) as i32
}

although this has been changed by #1354

Do we need some kind of release to use the new code?

[Amanieu]

The stdarch submodule in rust-lang/rust needs to be updated.

[yanns]

Is it something I can do? Locally? Or it has to be done by the rust-lang?

[Amanieu]

You can submit a pull request to rust-lang/rust that updates the submodule.

[yanns]

rust-lang/rust#105784

[yanns]

I don't know if you can help in the rust PR.
It seems that there is a new feature tag:

error: the feature named `gfni` is not valid for this target
  --> library/core/src/../../stdarch/crates/core_arch/src/x86/gfni.rs:68:18
   |
68 | #[target_feature(enable = "gfni,avx512bw,avx512f")]
   |                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ `gfni` is not valid for this target