Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 12 Apr 2007 11:05:14 -0700
From:      Julian Elischer <julian@elischer.org>
To:        Robert Watson <rwatson@FreeBSD.org>
Cc:        Ivan Voras <ivoras@fer.hr>, Randall Stewart <rrs@cisco.com>, freebsd-current@freebsd.org, Janardhan Iyengar <iyengar@conncoll.edu>, =?ISO-8859-1?Q?Dag-Erling_Sm=F8rgrav?= <des@des.no>, Brad Penoff <penoff@cs.ubc.ca>
Subject:   Re: CPU utilization
Message-ID:  <461E74DA.3020402@elischer.org>
In-Reply-To: <20070412153930.A99718@fledge.watson.org>
References:  <461E0078.3050001@cisco.com> <evl00c$89j$1@sea.gmane.org>	<461E092B.4080001@cisco.com> <86d529zxci.fsf@dwp.des.no> <20070412153930.A99718@fledge.watson.org>

next in thread | previous in thread | raw e-mail | index | archive | help
Robert Watson wrote:
> 
> On Thu, 12 Apr 2007, Dag-Erling Smørgrav wrote:
> 
>> Randall Stewart <rrs@cisco.com> writes:
>>> machdep.hyperthreading_allowed: 0
>>
>> Note that enabling hyperthreading is more likely to harm performance 
>> than to help it.  You should just disable it in the BIOS, and run a UP 
>> kernel.
> 
> Historically this has been true, but some more recent results I've seen 
> suggest that both hyperthreading hardware has improved, and the 
> efficiency of our SMP implementation and scheduler has lead to it being 
> more effective used. I would reevaluate this on more modern hardware and 
> using a more recent kernel before assuming this remains true for your 
> application.

In addition to this, to answer the original question, I remember a 
commit so that if you disable a cpu (or HT cpu) it doesn't get counted
in the CPU % so if you have 2 cpus and disable one hten prior to that
commit it was not possible to get > 50% busy but after that commit
you could get 100% "of the available CPUs". That fix is not (I believe)
in 6.2.

I have had applications where HT was useful. They had a mix of
integer and floating point work, and were long running, using all 
of their quanta. Usually with normal work it was a was or a loss.


> 
> Robert N M Watson
> Computer Laboratory
> University of Cambridge
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> freebsd-current@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org"




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?461E74DA.3020402>