public inbox for kvm@vger.kernel.org
 help / color / mirror / Atom feed
* Floating point vs Integer performance
@ 2007-03-12 20:18 Benjamin Prosnitz
  2007-03-13  8:03 ` Avi Kivity
  2007-03-13  8:16 ` Michael Riepe
  0 siblings, 2 replies; 3+ messages in thread
From: Benjamin Prosnitz @ 2007-03-12 20:18 UTC (permalink / raw)
  To: kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

I ran some benchmarks inside of KVM-12 on my system and outside of KVM
(no virtualization on my system).  I found that floating point
performance when running both with and without KVM was approximately the
same, but integer performance in KVM was about 25% of what I get when
running the test outside of KVM. These were tests using Dhrystone 2 and
Double Precision Whetstone on a Core 2 Duo T7200 processor.

Any thoughts on why the integer and floating point performance appears
to differ so much?

Benjamin Prosnitz


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: Floating point vs Integer performance
  2007-03-12 20:18 Floating point vs Integer performance Benjamin Prosnitz
@ 2007-03-13  8:03 ` Avi Kivity
  2007-03-13  8:16 ` Michael Riepe
  1 sibling, 0 replies; 3+ messages in thread
From: Avi Kivity @ 2007-03-13  8:03 UTC (permalink / raw)
  To: Benjamin Prosnitz; +Cc: kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Benjamin Prosnitz wrote:
> I ran some benchmarks inside of KVM-12 on my system and outside of KVM
> (no virtualization on my system).  I found that floating point
> performance when running both with and without KVM was approximately the
> same, but integer performance in KVM was about 25% of what I get when
> running the test outside of KVM. These were tests using Dhrystone 2 and
> Double Precision Whetstone on a Core 2 Duo T7200 processor.
>
> Any thoughts on why the integer and floating point performance appears
> to differ so much?
>
>   

Perhaps the integer benchmarks thrash the mmu cache.  Try raising 
KVM_NUM_MMU_PAGES to 1024.

4MB per VM lowmem is a lot, but if it helps many workloads, it's worth it.


-- 
Do not meddle in the internals of kernels, for they are subtle and quick to panic.


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: Floating point vs Integer performance
  2007-03-12 20:18 Floating point vs Integer performance Benjamin Prosnitz
  2007-03-13  8:03 ` Avi Kivity
@ 2007-03-13  8:16 ` Michael Riepe
  1 sibling, 0 replies; 3+ messages in thread
From: Michael Riepe @ 2007-03-13  8:16 UTC (permalink / raw)
  To: Benjamin Prosnitz; +Cc: kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Hi!

Benjamin Prosnitz wrote:
> I ran some benchmarks inside of KVM-12 on my system and outside of KVM
> (no virtualization on my system).  I found that floating point
> performance when running both with and without KVM was approximately the
> same, but integer performance in KVM was about 25% of what I get when
> running the test outside of KVM. These were tests using Dhrystone 2 and
> Double Precision Whetstone on a Core 2 Duo T7200 processor.
> 
> Any thoughts on why the integer and floating point performance appears
> to differ so much?

I guess it's the benchmark. In contrast to its FP counterpart, Dhrystone
doesn't crunch numbers most of the time. It also performs lots of string
operations, for example (memory access == slow).

-- 
Michael "Tired" Riepe <michael-0QoEqw4nQxo@public.gmane.org>
X-Tired: Each morning I get up I die a little

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2007-03-13  8:16 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2007-03-12 20:18 Floating point vs Integer performance Benjamin Prosnitz
2007-03-13  8:03 ` Avi Kivity
2007-03-13  8:16 ` Michael Riepe

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox