Shim the LoongArch CRC intrinsics

[Gelbpunkt]

I'm adding support for using these in zlib-rs, and @folkertdev suggested implementing them in Miri (trifectatechfoundation/zlib-rs#511 (comment)). The LoongArch CRC intrinsics are extremely similar to those found on AArch64, with the main difference that the argument order is swapped.

Here's the test in Miri and on actual hardware:

$ ./miri run tests/pass/shims/loongarch/intrinsics-loongarch64-crc.rs
$ cargo +miri build -Zroot-dir=/home/aelin/miri --manifest-path /home/aelin/miri/./Cargo.toml --all-targets
    Finished `dev` profile [optimized + debuginfo] target(s) in 0.14s
$ cargo +miri build -Zroot-dir=/home/aelin/miri --manifest-path /home/aelin/miri/cargo-miri/Cargo.toml --all-targets
    Finished `dev` profile [optimized + debuginfo] target(s) in 0.07s
$ cargo miri setup
$ /home/aelin/miri/target/debug/miri --edition 2021 --sysroot /home/aelin/.cache/miri tests/pass/shims/loongarch/intrinsics-loongarch64-crc.rs

$ rustc tests/pass/shims/loongarch/intrinsics-loongarch64-crc.rs -o /tmp/meow && /tmp/meow

$ uname -a
Linux seraphine 7.1.0-rc4-0-stable #1-Alpine SMP PREEMPT_DYNAMIC 2026-05-23 17:58:11 loongarch64 Linux
$ cat /proc/cpuinfo | grep Model -m 1
Model Name		: Loongson-3B6000

i.e. it behaves the same on real hardware and in Miri.

The implementation is based on the ACLE code with only very slight adjustments; the test is also basically the same one.

Helpful links for reviewing:

• Architecture manual: https://loongson.github.io/LoongArch-Documentation/LoongArch-Vol1-EN.html#crc-check-instructions
• stdarch PR which aligns Rust's intrinsics with Clang's: Several fixes to the loongarch CRC intrinsics stdarch#2136
• Clang's intrinsics: https://github.com/llvm/llvm-project/blob/main/clang/lib/Headers/larchintrin.h#L58-L104
• LLVM's intrinsic definitions: https://github.com/llvm/llvm-project/blob/main/llvm/include/llvm/IR/IntrinsicsLoongArch.td#L78-L94 (note they all take i32 or i64 parameters)

[rustbot]

Thank you for contributing to Miri! A reviewer will take a look at your PR, typically within a week or two.
Please remember to not force-push to the PR branch except when you need to rebase due to a conflict or when the reviewer asks you for it.

[RalfJung]

Thanks a lot, just a few comments. :)

The code (and the test) are slightly ugly because the intrinsics take signed integers. I'd be very happy about suggestions to make it prettier.

What do you mean, you don't seem to be doing anything about the sign and just interpret the inputs as unsigned? (In the code, I mean. For the tests I guess it'd be up to stdarch to make the API less annoying...)

LLVM's intrinsic definitions: https://github.com/llvm/llvm-project/blob/main/llvm/include/llvm/IR/IntrinsicsLoongArch.td#L78-L94 (note they all take i32 or i64 parameters)

LLVM doesn't distinguish signed and unsigned integers so that does not mean much

@rustbot author

View changes since this review

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.

[Gelbpunkt]

What do you mean, you don't seem to be doing anything about the sign and just interpret the inputs as unsigned? (In the code, I mean. For the tests I guess it'd be up to stdarch to make the API less annoying...)

Yeah, I did change that in a follow-up commit since because CI was complaining about i32 -> u32 casts. I think it's fine now anyways.

LLVM doesn't distinguish signed and unsigned integers so that does not mean much

TIL!

@rustbot ready

[RalfJung]

Yeah, I did change that in a follow-up commit since because CI was complaining about i32 -> u32 casts. I think it's fine now anyways.

Okay, please update the PR description. :)

[Gelbpunkt]

Okay, please update the PR description. :)

Done!

[RalfJung]

Wait, we aren't even running these tests, are we? We'll have to add a loongarch target to CI. I'll take care of that.