https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119007
--- Comment #3 from Jeffrey A. Law <law at gcc dot gnu.org> --- I think that's reasonable as well and one of the options discussed weeks ago in the patchwork call. I was thinking that having the relevant intrinsics set the flag was slightly better solution because it automatically does the right thing with no additional need from the user. I guess one thing to look at is how LLVM handles this scenario and try to be compatible.
