nikic wrote: For the purposes of this PR, I think you should only change canCreateUndefOrPoison to return true and just eat the regressions. We can follow up with using isValidAddrSpaceCast() in a followup, because it will be less straightforward. I'm not willing to accept direct use of TTI in ValueTracking without laundering it through an abstraction first (see https://discourse.llvm.org/t/constant-propagation-for-target-specific-intrinsics/85881/5).
https://github.com/llvm/llvm-project/pull/144686 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
