https://gcc.gnu.org/bugzilla/show_bug.cgi?id=29231
--- Comment #8 from Iain Sandoe <iains at gcc dot gnu.org> --- A secondary comment - the wiring up of the built-ins that allocate/deallocate trampoline entries makes the underlying mechanism opaque to the middle end consumer. So, although the current example implementations use mmap / heap, I do not believe that there is anything preventing an implementation from choosing some alternate scheme for allocation of tables (e.g. a non-hosted impl could use a fixed allocation - presumably with a hard fail if that allocation is exceeded.)
