[llvm] [AMDGPU] revert srl pattern for true16 mode (PR #208136)

[Jay Foad via llvm-commits]

https://github.com/jayfoad approved this pull request.

LGTM, but do you have any plans for how we might be able to fix the code quality regressions?

https://github.com/llvm/llvm-project/pull/208136