[llvm] [DAGCombine] Fold (icmp eq/ne (shift X, C), 0) -> (icmp eq/ne X, 0) (PR #88801)

[Björn Pettersson via llvm-commits]

bjope wrote:

Draft based on missing optimization found in https://github.com/llvm/llvm-project/pull/87646

Unfortunately it seems to cause other regressions. A bit surprised that for example RISCV is emitting more instructions when the DAG is simplified to get rid of the not-needed shift.

https://github.com/llvm/llvm-project/pull/88801