[llvm-branch-commits] [llvm] [InstCombine] Fold exp2(uitofp iN x) to ldexp when N equals the width of int (PR #227380)

[Harrison Hao via llvm-branch-commits]

harrisonGPU wrote:

@efriedma-quic @ZERICO2005  Thanks for the suggestions! Updated: without ninf we now emit ldexp(1.0, umin(x, Emax + 1)), and with ninf we fold directly to ldexp(1.0, x). And I also added an explicit isIntN check. It always holds today (Emax + 1 <= 16384), but it keeps the fold safe if a wider FP type is ever added. :-)

https://github.com/llvm/llvm-project/pull/227380