nikic wrote: I think the assumption may be that for uitofp nsz applies to the return value (contrary to our general definition) so that a 0.0 result can non-deterministically flip to -0.0?
I agree that having it on fptoui would make more sense for this transform, and match normal nsz semantics. https://github.com/llvm/llvm-project/pull/198470 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
