r? @LukasKalbertodt

(rust_highfive has picked a reviewer for you, use r? to override)

⚠️ Warning ⚠️

• These commits modify submodules.

r? @Amanieu

@bors r+

📌 Commit 5b01cd5e1b51c89be8e13a7924e70645363dab09 has been approved by Amanieu

@Amanieu failed on Linux, might be hwcap issue?

It's because the stage0 compiler doesn't support the tme extension. Rustc stage0 uses the current beta compiler to compile master.

You need to add #[cfg(not(bootstrap))] to the tme module in stdarch to exclude it from the stage0 build.

@bors r-

(Continuing from the other thread)

We generally want x.py doc src/libstd --stage 0 to "just work" because it's significantly better for docs contributors as opposed to forcing them to build a full std/compiler toolchain just to document std. However, that said, I am surprised to see this working with the rustc bootstrap compiler which ran x.py check on that crate already? Is there some bug in bootstrap perhaps that is only sometimes passing --cfg bootstrap?

It seems that CI is running a stage0 rustdoc which doesn't understand the "tme" feature as part of the mingw-check builder.

I'm not exactly sure what the right course of action is here. cc @pietroalbini

So it seems that this is caused by #56627:

So far this starts to enable the std::arch in stage0 builds of rustc.
This means that we may need stage0/not(stage0) in stdsimd itself, but
more and more code is starting to use std::arch so I think it's time
to start shifting the balance of work here.

I guess we'll just need to add cfg(bootstrap) to stdarch.

In the cases where we do need it I'm hoping we can make sure that the current nightly compiler will still work -- i.e., that you just get a less featureful stdsimd but not actually breakage when we upgrade the bootstrap compiler. If that's not possible then please ping me when landing such a change, as it means we need to bump stdsimd when bumping the bootstrap compiler in rust-lang/rust (and in the long run probably means we should subtree/merge in stdsimd to this repo).

The TME instructions are not available on any physical CPU yet (and I don't think even QEMU supports them), so I doubt this will cause any breakage in practice.

@bors r+

📌 Commit 2915d8e4224c80d08f0f1d75ad6e745964f652c7 has been approved by Amanieu

The TME instructions are not available on any physical CPU yet (and I don't think even QEMU supports them), so I doubt this will cause any breakage in practice.

Yes, yes, and yes. So far only llvm and gcc incorporated this support.

Can someone add this to rollup if it is ok? @rust-lang/release

@bors r+

📌 Commit 15201af has been approved by Amanieu