At 8:57 PM +0000 12/23/02, Nicholas Clark wrote:
If the interpreters are reasonably separate, then one won't affect the other, except in cases where the system has exhausted the memory available to the process, so there's not a huge reason to do this.On Mon, Dec 23, 2002 at 01:37:18PM -0500, Dan Sugalski wrote:At 1:50 PM +0000 12/22/02, Leopold Toetsch (via RT) wrote: >Here is another approach, to finally unify the marking routines: >- next_for_GC remains what it is >- instead of current_end_of_list and a return value, mark_used() uuses > a file static mark_ptr. As we don't have recursive DOD runs, this > ought to be safe.Nope. Threads. Two interpreters can be DODing at the same time. Static variables aren't tenable for much of anything with threads... :(Unless protected by a mutex. I have now read the programming POSIX threads book. I take it that makes me a fully qualified l33t threads d00d who knows everything. How much of a speed impact could it be to serialise DOD runs between threads? Presumably 1 thread doing a DOD run could actually manage to free up enough memory to make both happy?
Also, it is a choke point. Since the interpreters are generally separate, the better place for this is in the interpreter structure.
--
Dan
--------------------------------------"it's like this"-------------------
Dan Sugalski even samurai
[EMAIL PROTECTED] have teddy bears and even
teddy bears get drunk
