[llvm] [InstCombine] Fold zext(i1) == lshr(A, BW - 1) => i1 == A s< 0 (PR #68244)

[Nikita Popov via llvm-commits]

nikic wrote:

And I think the tests need an update too? It doesn't look like the most basic case is tested at all (where the zext i1 and lshr are not correlated).

https://github.com/llvm/llvm-project/pull/68244