global_asm! macro causes non-fatal errors to be printed during compilation for some RISC-V extension instructions when targeting the GC extensions

[repnop]

I'm currently using the asm! macro on the riscv64gc-unknown-none-elf target, and have been getting some "errors" for a while about instructions requiring extensions -- except that GC includes the extensions of the instructions I'm using (IMAFDC). So far I've seen errors for F, D, and A instructions, everything else seems to be fine that I've used so far. Despite them being reported as errors, the build still succeeds however. @nbdd0121 suggested this may be an LLVM bug.

Example:

error: instruction requires the following: 'D' (Double-Precision Floating-Point)
        fsd f0, 248(sp)
        ^
error: instruction requires the following: 'A' (Atomic Instructions)
        sc.d zero, zero, 0(sp)
        ^

[Amanieu]

I can't seem to reproduce this: https://rust.godbolt.org/z/5hW8jn

[repnop]

Ah, finally was able to reproduce it, looks like its not asm! but global_asm! (I never made the connection between the output and global_asm!, I assumed it affected both since I don't use those instructions in non-global_asm! code), so I'll correct the title. It also requires the --emit=dep-info,link argument to be passed to rustc, at which point you can see all of the errors, here's a godbolt link: https://rust.godbolt.org/z/KreW3M

[usamoi]

I saw error: instruction requires the following: 'D' (Double-Precision Floating-Point) in my global_asm on the riscv64gc-unknown-none-elf target, too, but only in release mode.

[Amanieu]

I still don't know what the cause of this problem is (it seems to be in LLVM?). However I have found a workaround: you need to add .attribute arch, "rv64gc" to the assembly code to enable the f and d features in the assembler.

[deepaksirone]

I see the following in a release build with global_asm!:

error: instruction requires the following: 'A' (Atomic Instructions)
        amoswap.d.aq t0, t0, (t1)
        ^
error: instruction requires the following: 'M' (Integer Multiplication and Division) or 'Zmmul' (Integer Multiplication)
        mul t1, a3, t2

[Amanieu]

Upstream bug: llvm/llvm-project#61991

[nbdd0121]

This is now also affecting naked functions after the codegen switch. cc @folkertdev (example: https://github.com/nbdd0121/unwinding/actions/runs/12341696455/job/34440632980)

[folkertdev]

Well, that's unfortunate. The LLVM changes don't have any action since August, are there any strings we can pull to give that some extra attention?

[xobs]

Here's a minimal reproduction:

#![feature(naked_functions)]

#[naked]
pub extern "C-unwind" fn example() {
    #[cfg(target_feature = "d")]
    unsafe {
        core::arch::naked_asm!("fsd fs0, 0x140(sp)");
    }
}

Compile with cargo +nightly build --target riscv64gc-unknown-linux-gnu --release (the --release is important)