andykaylor wrote: > > The inliner should either be inserting a call to > > llvm.launder.invariant.group() or removing the metadata, but it doesn't > > (though it does insert lifetime start/end markers). > > Is this intended or is it an LLVM bug? We might contact an LLVM guy to ensure > this.
I think it's probably a bug but, given that this is marked as an experimental feature in the LLVM Language Reference, I'm not sure how easy it would be to get someone to fix it. If you're familiar with the [history](https://llvm.org/devmtg/2021-11/slides/2021-ptr_provenanceAndLlvmNoaliasTheTaleOfFullRestrict.pdf) of `restrict/noalias` support in LLVM (which is conceptually similar to this), you'll know that this has been a very tricky type of problem to solve. https://github.com/llvm/llvm-project/pull/175037 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
