tschuett wrote: Regarding optimizations that exist in the middle-end, if that becomes a big issue, we could move it one floor up. And discuss on discourse the relationship between optimizations in the-middle-end and the two backends. https://github.com/llvm/llvm-project/pull/105991