[llvm] [AMDGPU][True16][MC] true16 for v_minmax/maxmin_f16 (PR #119586)

Brox Chen via llvm-commits llvm-commits at lists.llvm.org
Thu Dec 19 07:48:19 PST 2024


broxigarchen wrote:

> Hi @broxigarchen this is breaking compilation on GFX12 for us. Can you please fix or revert? Test case:
> 
> ```
> ; RUN: llc -mtriple=amdgcn -mcpu=gfx1200 < %s
> define half @test(half %arg) {
>   %i = call half @llvm.minnum.f16(half %arg, half 0xH0000)
>   %i1 = call half @llvm.maxnum.f16(half %i, half 0xH0000)
>   ret half %i1
> }
> ```
> 
> I get:
> 
> ```
> llc: /home/jayfoad2/git/llvm-project/llvm/include/llvm/MC/MCInstrInfo.h:64: const MCInstrDesc &llvm::MCInstrInfo::get(unsigned int) const: Assertion `Opcode < NumOpcodes && "Invalid opcode!"' failed.
> PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace.
> ```

Revert this commit here https://github.com/llvm/llvm-project/pull/120594

https://github.com/llvm/llvm-project/pull/119586


More information about the llvm-commits mailing list