[llvm] [GlobalISel] Fold instructions with fully-known bits to a constant. (PR #224254)

[Vikash Gupta via llvm-commits]

vg0204 wrote:

> @vg0204 I am sorry #198808 got stuck in the review for a bit. I will push for another round of review this week.
> 
> One critical review I got is that we need to be faithful to what/how SelectionDAG does unless we have a concrete reason behind the intentional drift. I agree with this methodology.
> 
> So, having the fold in the 's walk is preferrable. Subsequent patches will be easier to land.

Understood, so for now keeping this PR, will close once your PR covers LIT tests covered here as @yxsamliu mentioned. In case, if you need any support from my end for your ongoing PR effort, let me know about it! 

https://github.com/llvm/llvm-project/pull/224254