https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118594
--- Comment #14 from Andrew Pinski <pinskia at gcc dot gnu.org> --- (In reply to Andrew Pinski from comment #11) > And yes both things I think are GCC 16 material at this point. I think we > should disallow subreg of mem completely and should do the adjust_address > change. I am not even sure adjust_address is needed for GCC 15 (that is this can wait until GCC 16 to deal with this report); this was only reported via a random code generator or by myself while testing out things after the reported bug. It is also not a regression and the duplicate has been ICEing since we allowed conversions from BF16 to float. Plus it requires volatile and vectors to be involved that is a cast from a volatile pointer to a non volatile pointer and then an element of that vector needs to show up.
