[llvm] [AMDGPU][CodeGen] Implement SimplifyDemandedBitsForTargetNode for readfirstlane. (PR #190009)

[Jay Foad via llvm-commits]

jayfoad wrote:

> Propagate demanded bits through readfirstlane and readlane intrinsics in AMDGPUISelLowering with SimplifyDemandedBitsForTargetNode implementation.

Update the description: the patch does not handle readlane.

https://github.com/llvm/llvm-project/pull/190009