Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 16 Dec 2005 17:21:50 -0700
From:      Scott Long <scottl@samsco.org>
To:        Gea-Suan Lin <gslin@gslin.org>
Cc:        freebsd-performance@freebsd.org, leo.huang.gd@gmail.com, Rink Springer <rink@il.fontys.nl>
Subject:   Re: Benchmark MySQL Performance On FreeBSD And Linux
Message-ID:  <43A35A1E.3020408@samsco.org>
In-Reply-To: <20051216204322.GA66409@gslin.org>
References:  <20051216064615.GA41775@gslin.org>	<20051216070609.GA3216@il.fontys.nl> <20051216204322.GA66409@gslin.org>

next in thread | previous in thread | raw e-mail | index | archive | help
Gea-Suan Lin wrote:

> On Fri, Dec 16, 2005 at 08:06:09AM +0100, Rink Springer wrote:
> 
>>>And you should disable these options, it may increase ~10% again:
>>>
>>>-cpu             I486_CPU
>>>-cpu             I586_CPU
>>> cpu             I686_CPU
>>
>>A recent discussion on -STABLE warned against removing I586_CPU, refer
>>to
>>http://lists.freebsd.org/pipermail/freebsd-stable/2005-December/020696.html
>>for more information.
> 
> 
> Thanks, I'll try it later.
> 
> BTW, did anyone try using jemalloc to test mysql ?
> 

It was tried recently using supersmack and unix domain sockets.  The
results were slightly slower (around 6%, IIRC).  The theory is that
jemalloc doesn't pad out allocations as much, so there is high potential
for cache line sharing (and thus cache thrash) between CPUs.

Scott



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?43A35A1E.3020408>