Re: [gclist] Experience with conservative GC sought

[Douglas Thomas Crosher <dtc@scrooge.ee.swin.oz.au>]

> scm uses its own conservative collector which only scans the stack
> conservatively but does precise tracing for objects inside the heap.
> I have used scm to run programs with (comparatively) high rates of
> data allocation.  The experience has been similar as with PLT: The
> heap grows far beyond the size of the live data and doesn't shrink
> back.  Moreover, as the size of the live data within a huge heap was
> small, fragmentation caused substantial thrashing; the OS would
> typically try to keep the whole heap in memory.  Note that GNU Guile
> uses scm's collector.

The CMUCL generational GC similarly scavenges the stack for
conservative roots while having a precise heap object representation.
I use it on jobs with high allocation rates running over a week
without seeing significant problems due to leakage from conservative
roots. A little care is needed in the writing of the application so as
to limit the amount of objects retained due to conservative roots.  I
find that the retention of dead objects in old generations is the most
significant problem, and this can be limited by tuning the number of
generations.

Regards
Douglas Crosher