Re: Ottawa - kernel hackers ...



> [1] - it's quite amazing to quit X after a long session and see the X
> background pixmap re-render from swap: a huge area of flat memory - with
> uniform non-use for the last some hours - and the pixels are scattered
> all over the disk - hmm, at ~1 block per line it takes the ~10 seconds
> (1024*10ms) you'd expect to render. Just killall -9 nautilus to see.

Ah. I was wondering how many people were seeing this. So it isn't just me
after all.

I also see this when switching desktops sometimes. I reported it (along
with heavy disk thrashing/activity) to the Red Hat bugzilla. There was a
similar bug in which it was proclaimed that there was "a fundamental flaw
in the VM/cache design".

> 	Then again - perhaps it's only my machine where you can only tell if
> it's crashed or not when the crazy disk activity stops.

No, it's not just your machine. FWIW my SuSE 8 install never had
these kind of problems that I recall. Switching to a standard 2.4.20
kernel on Red Hat 9 improved performance noticeably, but unfortunately
killed Wine in the process, due to mucking about with NPTL it seems. Sigh.

If it's truly the case that the kernel cannot realistically prevent
fragmentation of large images in memory like this (and who knows what
else), perhaps it's worth altering GNOME/GTK to use mmapping more
extensively rather than loading the files into memory the traditional way.

thanks -mike




[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]