[llvm] [AMDGPU][GlobalISel] Look through readfirstlane/readlane in computeKnownBitsForTargetInstr. (PR #224275)

[Vikash Gupta via llvm-commits]

vg0204 wrote:

> We don't do this on the DAG path yet? I'd rather do both to gather, or the DAG path first (I believe I opened an issue to implement this, and there may be existing open PRs from a long time ago)

If you plan to land your patch that contains both paths, let me know. Else, I could include SDAG path as well in this PR as you say!

https://github.com/llvm/llvm-project/pull/224275