[llvm] [AMDGPU] Add new llvm.amdgcn.wave.shuffle intrinsic (PR #167372)

via llvm-commits llvm-commits at lists.llvm.org
Tue Dec 30 09:20:42 PST 2025


saxlungs wrote:

> Throw in some GFX6 & GFX9 tests, but apart from that and the inline comment it LGTM.

@nhaehnle Thanks for the information and suggestions! In regards to this comment, from what I can tell support for the `ds_bpermute_b32` instruction wasn't added until GFX8, so currently this implementation will not work for GFX6 or GFX7. Is there a way that I can enforce that this new intrinsic that I'm adding is only legal for GFX8+? I think that would be the easiest way to solve this if so.

https://github.com/llvm/llvm-project/pull/167372


More information about the llvm-commits mailing list