On Wed, 2004-10-06 at 16:38 +1000, Steffen Kluge wrote: > On Tue, 2004-10-05 at 15:15 -0400, Owen Taylor wrote: > > Well, the biggest problem with backing store is that it is per-window > > not per *toplevel* window. In X, toplevel windows can have subwindows > > Hmm, interesting. It seems, though, that the X11 forefathers have > thought of that, to a degree. There used to be (still are?) the > "WhenMapped" and "WhenUseful" backing store attributes. There still > seems to be the -wm server switch, turning on "WhenMapped" for all > windows. This doesn't really help - you want to backing store the *visible* portions of subwindows, what you don't want to do is to backing store portions of subwindows that are obscured by other subwindows. > > In certain circumstances, X can end up storing huge amounts of entirely > > useless pixel data for subwindows because it mistakes them for > > obscured toplevels. > > How much of an issue is this? I vaguely remember reading once that > backing store uses off-screen frame buffer memory (there should be > plenty of that on a 128MB graphics card running at 1280x1024). Does > backing store actually affect the X server memory footprint? Then again, > doesn't the reported memory footprint of the X server contain all the > mapped in frame buffer memory as well? The more I think about it the > more I realise my utter ignorance in all of this... Imagine a 10,000 item list that is 150,000 pixels by 1000 pixels - e.g., 450 megs of memory. Under some circumstances, the X backing store code can be fooled into backing storing the *entire list*. Regards, Owen
Attachment:
signature.asc
Description: This is a digitally signed message part
-- Fedora-desktop-list@xxxxxxxxxx http://www.redhat.com/mailman/listinfo/fedora-desktop-list
