home *** CD-ROM | disk | FTP | other *** search
- x-gateway: rodan.UU.NET from help-lucid-emacs to alt.lucid-emacs.help; Mon, 9 Nov 1992 16:48:56 EST
- Message-ID: <9211091815.AA11299@thymus.synaptics>
- Subject: Re: GC idea
- Reply-To: daveg@synaptics.com
- Date: Mon, 09 Nov 1992 10:15:24 -0800
- From: daveg@thymus.synaptics.com (Dave Gillespie)
- Newsgroups: alt.lucid-emacs.help
- Path: sparky!uunet!wendy-fate.uu.net!help-lucid-emacs
- Sender: help-lucid-emacs-request@lucid.com
- Lines: 22
-
- Dirk Grunwald writes:
- > Actually, mayhaps it's time for a general discussion of the current GC
- > scheme. Switching to a generational scheme would reduce the relatively
- > large interactive ``hit'' you take in Emacs.
-
- It would be really great to have something like this. I find that
- after a while my Emacs process gets to be quite large, but mostly
- swapped out. As I understand it a generational GC would let most of
- that stuff stay swapped out, whereas the current GC has to bring it
- back into memory on a regular basis.
-
- I doubt you'd need something as complicated as a generational GC
- just to fix the interactive response time problem, though. Even
- the GC-during-idle-time scheme would fix 90% of the problem.
- Another idea I had was to arrange the current GC algorithm to be
- able to abort itself quickly if there is any user input. Right
- now the GC algorithm clears marks as it is sweeping; you'd need
- to change it to do an explicit unmarking phase at the beginning
- of GC, which would slow it down a bit. But since GC would be
- interruptable, it wouldn't matter nearly so much how long it took.
-
- -- Dave
-