Re: 2.6.15-rc5-mm3

[Con Kolivas <kernel@kolivas.org>]

Jesper Juhl writes:

> On 12/15/05, Andrew Morton <akpm@osdl.org> wrote:
> >
> <!-- snip -->
> > -mm-implement-swap-prefetching.patch
> > -mm-implement-swap-prefetching-default-y.patch
> > -mm-implement-swap-prefetching-tweaks.patch
> > -mm-implement-swap-prefetching-tweaks-2.patch
> > -mm-swap-prefetch-magnify.patch
> >
> >  Dropped swap prefetching, sorry.  I wasn't able to notice much benefit from
> >  it in my testing, and the number of mm/ patches in getting crazy, so we don't
> >  have capacity for speculative things at present.
> <!-- snip -->
>
> This is a bit sad.
> On my system (1.4GHz Athlon w/512MB RAM, 768MB swap) this did have an effect.
> One situation in particular where it helped (and which is a common
> case for me) was when I had OpenOffice2 + Eclipse (with CDT) + xchat +
> nedit + Firefox + a few konsole windows open (running KDE 3.5 btw and
> all apps usually have a lot of content loaded), minimized all the apps
> and then started an allyesconfig build in one window - the
> allyesconfig build would drag the machine to its knees and eat up more
> or less all RAM + swap, so I usually left it alone for a while to
> finish and when I then came back later and reactivated the apps I had
> minimized they came back pretty fast. Without the swap prefetch
> patches things come back somewhat slower - it's not an earth
> shattering difference, but it's definately noticable.

Apart from the feedback I've gotten directly from users of the -ck patchset 
incorporating this, Andrew and lkml has had very few reports about it until 
he decided to drop it. I'm still maintaining the code in -ck indefinitely 
for the moment. Here's hoping he picks it up again in the future.

Cheers,
Con

Attachment: pgpieuotOhym8.pgp
Description: PGP signature