steakhal added a comment. In D138037#3941777 <https://reviews.llvm.org/D138037#3941777>, @xazax.hun wrote:
> I guess there are some more options. We could try keeping representatives > alive no matter what. It could be a good exercise to see if doing that makes > any difference in the analysis results. Yes, I should have mentioned it. I was thinking about exactly this, prior convincing myself for removing this assertion. However, my feeling is that it would keep too many symbols. Basically leaking them. Think about that all symbol starts as a trivial eqclass. Whenever we would merge, we would leak the other eqclass. Idk but this also seems scarry. On the other side, I dont have the infra for conducting a detailed performance analysis yet - unlike with making report diffs. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D138037/new/ https://reviews.llvm.org/D138037 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
