Re: GC algorithms

Date view Thread view Subject view Author view

From: Archie Cobbs (archie@whistle.com)
Date: Fri Dec 04 1998 - 14:36:17 EST


peter@transvirtual.com writes:
> Just my two cents on the fly.. using page faults for GC barriers *heavily*
> depends on the OS exception speed. I remember we did that for OS/2 and ended
> just using it for "sbrk-ing" the bottom-most alloc areas (by means of guard
> pages). Besides its portability problems, its not a performance-panacea, too.

Interesting.. real-world experience is the best information.
Sounds like it wouldn't be worth the effort, although I think
FreeBSD at least should have acceptable exeception speed.

Perhaps the effort is better spent locating and coding in
write barriers manually... after all, manual write barriers
are surely faster (and more accurrate) than page faulting,
except in cases where there's very few changes (and hence
no page faults).

-Archie

___________________________________________________________________________
Archie Cobbs * Whistle Communications, Inc. * http://www.whistle.com


Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Sat Sep 23 2000 - 19:57:05 EDT