efriedma-quic wrote: Since nothing should be using preserve_none on aarch64 yet, it's probably fine to just fix forward.
> I'm not sure under what conditions issues present for this We only allocate a base pointer under restricted circumstances (primarily, functions with dynamic allocation), so maybe GHC never tripped over the issues. > It looks like X15 is clobbered so I don't think it will be usable on Windows. > Do you think it should be excluded from argument passing only on Windows or > always? You probably could make this work if you really wanted to: you can allocate small amounts of stack without __chkstk, which you could then use to spill X15. Not sure if that's the right tradeoff here. https://github.com/llvm/llvm-project/pull/91046 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
