Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 26 Aug 2003 14:14:57 -0400 (EDT)
From:      John Baldwin <jhb@FreeBSD.org>
To:        Yamada Ken Takeshi <ken@tydfam.jp>
Cc:        current@freebsd.org
Subject:   Re: HTT on current
Message-ID:  <XFMail.20030826141457.jhb@FreeBSD.org>
In-Reply-To: <20030826.233001.730554593.ken@tydfam.jp>

next in thread | previous in thread | raw e-mail | index | archive | help

On 26-Aug-2003 Yamada Ken Takeshi wrote:
>   JYI,
>   I tested machdep.hlt_logical_cpus=0/1, buildworld, 
> and found no particular reason to disable HTT 
> as below with Zeon 2.8Ghz x 2, 1GBmem, Slow IDE HDD.
> 
>  It may not be the common case, so just for your info.
> 
># sysctl machdep.hlt_logical_cpus=0
># /usr/bin/time make -j32 buildworld
>      1910.29 real      2520.53 user       777.30 sys
> 
># sysctl machdep.hlt_logical_cpus=1
># /usr/bin/time make -j32 buildworld
>      2289.33 real      2666.66 user       645.88 sys

One test is not sufficient.  -current is also not the best
place to test. :)  When I first implemented HTT in -current
and -stable, I did some worldstone benchmarks on -stable on
a machine with a single HTT CPU.  Note that I was comparing
a UP kernel with an SMP kernel as well, so some of the speed
decrease of the SMP kernel could be due to it being an SMP
kernel, not due to HTT.  I did 16 trials (first one was
throwaway) of back-to-back buildworlds of the same version
of -stable using make, make -j2, and make -j4 for the
following configurations: UP, HTT, HTT with smp_idle_hlt, and
HTT with pause instructions added to stable and smp_idle_hlt.
The fastest build time belonged to UP without any -j option.
However, the next fastest ended up being the HTT with pause
and smp_idle_hlt.  The averages were 19:03 and 19:42
respectively with a standard deviation in both cases of
around 3.5 seconds.  That is only with 15 real trials though,
so I'm not sure what kind of 90% confidence interval you would
get from that.  I did find that with the hlt and pause
additions, the HTT kernel did complete the -j2 and -j4 builds
faster than the UP kernel, but those builds still took longer
than a build without -j.  I can't remember if I committed the
addition of the pause instructions to stable or not.  If I
didn't then I really should do that prior to 4.9.

-- 

John Baldwin <jhb@FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve!"  -  http://www.FreeBSD.org/



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.20030826141457.jhb>