goldsteinn wrote: > I can see the argument for dereferenceable_or_null, but I don't see any way > in which range will be useful.
Well I'd say if the callsite in the new inlined function gets inlined itself, we now have `range` info on that parameter we can optimize around. Further, if we ever get the attributor online IIRC it can add function parameters from the union of callsite arguments. But the more salient point to me at least, is the general principal that we should try to reduce information loss in our transforms when possible. We don't know all the potential goals of either downstream users or other LLVM contributors working on far away parts of the code base, and shouldn't essentially preclude use cases we can keep open. https://github.com/llvm/llvm-project/pull/91101 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
