lkml.org 
[lkml]   [2003]   [Dec]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: 2.6.0 performance problems
From
Date
On Tue, 2003-12-30 at 00:58, Thomas Molina wrote:
> On Mon, 29 Dec 2003, Linus Torvalds wrote:
>
> >
> >
> > On Mon, 29 Dec 2003, Thomas Molina wrote:
> > >
> > > I just finished a couple of comparisons between 2.4 and 2.6 which seem to
> > > confirm my impressions. I understand that the comparison may not be
> > > apples to apples and my methods of testing may not be rigorous, but here
> > > it is. In contrast to some recent discussions on this list, this test is
> > > a "real world" test at which 2.6 comes off much worse than 2.4.
> >
> > Are you sure you have DMA enabled on your laptop disk? Your 2.6.x system
> > times are very high - much bigger than the user times. That sounds like
> > PIO to me.
>
> It certainly looks like DMA is enabled. Under 2.4 I get:
>
> [root@lap root]# hdparm /dev/hda
>
> /dev/hda:
> multcount = 16 (on)
> IO_support = 1 (32-bit)
> unmaskirq = 1 (on)
> using_dma = 1 (on)
> keepsettings = 0 (off)
> readonly = 0 (off)
> readahead = 8 (on)
> geometry = 2584/240/63, sectors = 39070080, start = 0
>
>
> Under 2.6 I get:
>
> [root@lap root]# hdparm /dev/hda
>
> /dev/hda:
> multcount = 16 (on)
> IO_support = 1 (32-bit)
> unmaskirq = 1 (on)
> using_dma = 1 (on)
> keepsettings = 0 (off)
> readonly = 0 (off)
> readahead = 256 (on)
> geometry = 38760/16/63, sectors = 39070080, start = 0
>

Increase your readahead:

# hdparm -a 8192 /dev/hda


BTW: As we really do get this question a _lot_ of times, why
don't the ide layer automatically set a higher readahead
if there is enough cache on the drive or something?


Thanks,

--
Martin Schlemmer
[unhandled content-type:application/pgp-signature]
\
 
 \ /
  Last update: 2005-03-22 13:59    [from the cache]
©2003-2014 Jasper Spaans. Advertise on this site