Benjamin Goldberg <[EMAIL PROTECTED]> wrote: > There's no reason why the other scheme (whatever it is) can't be either > frequent DoD runs or my hybrid refcounting/dod scheme.
Frequent DOD runs are not a problem for an average program (for some definition of average). But when we come to ~1 million PMCs or so saving some of them does already count a lot. But doing the hybrid partial refcounting really depends on the var/value vtable split, which we don't have yet and on Dan, who seems to be strongly against it. leo
