[llvm] [AMDGPU] Add a trap lowering workaround for gfx11 (PR #85854)

[Scott Linder via llvm-commits]

slinder1 wrote:

What is the usual criteria for whether a workaround should be based on a target feature or a GFXIP# check?

Is there a chance some gfx11 revision fixes the underlying bug and we will need to switch to a feature? Can we just do that when/if it happens?

https://github.com/llvm/llvm-project/pull/85854