On Tue, 20 Jun 2000, Andrea Arcangeli wrote:
> On Tue, 20 Jun 2000, Cyrille Chepelov wrote:
>
> >On a subjective side, ac22-class feels a little bit snappier on the long
>
> Well I think between classzone-31 and classzone-32 (classzone-32 is
> included into ac22-class) I fixed a bug that hurted performance. The right
> fix for such bug is tricky so for now I was forward porting what we
> added to 17pre4 (and it seems it's better to have the bug as in ac22-riel
> that to run with such a fix as in ac22-class and 17pre4). I'd suggest to
> try again ac22-class plus this below patch (the result will be interesting
> also for 17pre4). You can find explanation of the patch in recent email to
> hpa. In case it doesn't help please let me know!! Many thanks for the
> feedback.
<snip patch>
Hi Andrea,
I've been doing some comparisons between ac22.classzone and acc22.riel
with Juan's last two diffs. For make -j 30 bzImage test, ac22.classzone
was consistently beating ac22.riel by a full minute (8.5 vs 9.5). Adding
this patch to the mix took another solid 30 seconds off build time.
Machine is 128mb PIII UP box w. single IDE drive.
FarSideTesting:
While looking for a yardstick of which kernel can shove the most junk
through swap fastest if it really needs to, I constructed a test that
classzone handles _much_ better. I don't know if it means anything..
but it sure makes vmstat non-boring ;-)
#!/bin/sh
DD='dd ibs=1024k count=128 obs=65536k skip=32768'
pingpong()
{
for i in 1 2 3 4 5 6 7 8 9 0
do
($DD if=/dev/zero|$DD)&
($DD if=/dev/zero|$DD)&
wait
done > /dev/null 2>&1
}
time pingpong
ac22classzone: ac22riel: (+last2_juan_diffs)
-----------------------------------------------------------------
real| 1m54.170s 1m48.909s 1m50.367s 6m45.570s 10m2.700s 8m53.571s
user| 0m14.180s 0m14.450s 0m14.290s 0m12.730s 0m11.530s 0m11.260s
sys | 1m9.260s 1m8.860s 1m8.850s 1m15.490s 1m18.040s 1m17.880s
Well, SuSE _does_ say 'Have a lot of fun...'
-Mike
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Jun 23 2000 - 21:00:21 EST