https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84411
--- Comment #5 from Drea Pinski <pinskia at gcc dot gnu.org> --- Created attachment 64500 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=64500&action=edit Almost correct This patch is almost correct. The only issue I think is I should just do an atomic store still. So basically we end up with: if (!atomic_load_1 (&guard, __ATOMIC_ACQUIRE)) atomic_store_n (&guard, 1, __ATOMIC_ACQ_REL); So the overhead is still the same as before for the common case. But smaller for the first time. Maybe this is just a size optimization now.
