spall wrote: I'd also update your description to mention fixing the vector > size 2 bug in FirstBitHigh as well, by making this code general purpose for both firstbitlow and firstbithigh. https://github.com/llvm/llvm-project/pull/116858