AIM9 slowdowns between 2.6.11 and 2.6.12-rc1

AIM9 slowdowns between 2.6.11 and 2.6.12-rc1

[Mel Gorman]

While testing the page placement policy patches on 2.6.12-rc1, I noticed
that aim9 is showing significant slowdowns on page allocation-related
tests. An excerpt of the results is at the end of this mail but it shows
that page_test is allocating 18000 less pages.

I did not check who has been recently changing the buddy allocator but
they might want to run a benchmark or two to make sure this is not
something specific to my setup.

root@monocle:~# grep _test vmregressbench-2.6.11-standard/aim9/log.txt
     7 page_test           60.01       4420   73.65439       125212.46 System Allocations & Pages/second
     8 brk_test            60.00       1732   28.86667       490733.33 System Memory Allocations/second
     9 jmp_test            60.01     252898 4214.26429      4214264.29 Non-local gotos/second
    10 signal_test         60.00       5983   99.71667        99716.67 Signal Traps/second
    11 exec_test           60.01        788   13.13114           65.66 Program Loads/second
    12 fork_test           60.06        986   16.41692         1641.69 Task Creations/second
    13 link_test           60.00       6302  105.03333         6617.10 Link/Unlink Pairs/second
root@monocle:~# grep _test vmregressbench-2.6.12-rc1-standard/aim9/log.txt
     7 page_test           60.01       3784   63.05616       107195.47 System Allocations & Pages/second
     8 brk_test            60.02       1194   19.89337       338187.27 System Memory Allocations/second
     9 jmp_test            60.00     252312 4205.20000      4205200.00 Non-local gotos/second
    10 signal_test         60.00       3731   62.18333        62183.33 Signal Traps/second
    11 exec_test           60.08        762   12.68309           63.42 Program Loads/second
    12 fork_test           60.04        864   14.39041         1439.04 Task Creations/second
    13 link_test           60.01       4723   78.70355         4958.32 Link/Unlink Pairs/second

-- 
Mel Gorman
Part-time Phd Student                          Java Applications Developer
University of Limerick                         IBM Dublin Software Lab
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>

Re: AIM9 slowdowns between 2.6.11 and 2.6.12-rc1

[Dave Hansen]

On Sun, 2005-04-03 at 15:37 +0100, Mel Gorman wrote:
> While testing the page placement policy patches on 2.6.12-rc1, I noticed
> that aim9 is showing significant slowdowns on page allocation-related
> tests. An excerpt of the results is at the end of this mail but it shows
> that page_test is allocating 18000 less pages.
> 
> I did not check who has been recently changing the buddy allocator but
> they might want to run a benchmark or two to make sure this is not
> something specific to my setup.

Can you get some kernel profiles to see what, exactly, is causing the
decreased performance?  Also, what kind of system do you have?  Does
backing this out help?  If not, can you test some BK snapshots to see
when this started occurring?  

http://linus.bkbits.net:8080/linux-2.5/cset@422de02c1628MP_noKSum9sGlTaC-Q

-- Dave

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>

Re: AIM9 slowdowns between 2.6.11 and 2.6.12-rc1 (probably false positive)

[Mel Gorman]

On Sun, 3 Apr 2005, Dave Hansen wrote:

> On Sun, 2005-04-03 at 15:37 +0100, Mel Gorman wrote:
> > While testing the page placement policy patches on 2.6.12-rc1, I noticed
> > that aim9 is showing significant slowdowns on page allocation-related
> > tests. An excerpt of the results is at the end of this mail but it shows
> > that page_test is allocating 18000 less pages.
> >
> > I did not check who has been recently changing the buddy allocator but
> > they might want to run a benchmark or two to make sure this is not
> > something specific to my setup.
>
> Can you get some kernel profiles to see what, exactly, is causing the
> decreased performance?  Also, what kind of system do you have?  Does
> backing this out help?  If not, can you test some BK snapshots to see
> when this started occurring?
>
> http://linus.bkbits.net:8080/linux-2.5/cset@422de02c1628MP_noKSum9sGlTaC-Q
>

The machine is a quad xeon with P III 733 processors. I don't have profile
information available as I wasn't collecting as I went along. However,
backing out the patch made little difference

However, I reran the test on 2.6.11 and this time the performance
difference was a lot less. Something else must have been happening when I
collected the first set of poor results. I'll revisit this when I'm next
working on kernel stuff and see can I reproduce it again reliably.

-- 
Mel Gorman
Part-time Phd Student                          Java Applications Developer
University of Limerick                         IBM Dublin Software Lab
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>