[llvm] [VectorCombine] Scalarize bin ops and cmps with two splatted operands (PR #137786)

[Min-Yih Hsu via llvm-commits]

https://github.com/mshockwave commented:

I'm a bit surprised the new pattern you proposed here wasn't created first, maybe it's because historically splat has been considered cheap?

https://github.com/llvm/llvm-project/pull/137786