efriedma-quic wrote: > I think we'd get decent code by doing a sequence of LOAD <all outgoing byval > bytes to vregs> ; TOKENFACTOR ; STORE <all vregs to byval argument slots>.
Up to some threshold, sure. If the scheduler is sufficiently clever, we won't even get any spills: we can interleave the loads and stores. But we probably still need some fallback path that avoids emitting large numbers of inline load/store ops. https://github.com/llvm/llvm-project/pull/109943 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
