HankChang736 wrote: > I think this may not be the right way. > > 1. We should at least add vl/vtype to clobbered registers when V is > specified. > > 2. The asm may not depend on vl/type, but adding vl/type dependencies > unconditionally stop further scheduling.
Yes, you're right. But the purpose we add vl/vtype dependencies is to prevent the Post-RA scheduler moving vsetvl instruction across inline assembly. I'm not sure if there's better approach to solve this problem. https://github.com/llvm/llvm-project/pull/128636 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
