Linux and GC
Wed, 13 May 1998 06:34:39 +0000
P T Withington wrote:
> What looks difficult in your idea is that the list of dead objects to be
> communicated back to the mutator might be large. In a collector that is
> a co-routine, it is simple for the collector to pop those objects onto a
> free list. Where your collector is in a separate address space, some
> form of IPC is needed to communicate them back to the mutator.
Well, if the free list is maintained in some memory that is
accessable to both the collector thread and the process thread,
there will bo no problem about this no matter how large.