Date: Fri, 18 Dec 1998 16:03:38 +0000 From: Karl Pielorz <kpielorz@tdx.co.uk> To: =?iso-8859-1?Q?Jos=E9=20M=AA?= Alcaide <jose@we.lc.ehu.es> Cc: freebsd-current@FreeBSD.ORG Subject: Re: SMP benchmarks Message-ID: <367A7CDA.EF67AF5E@tdx.co.uk> References: <367A78A5.3FBF4CBA@we.lc.ehu.es>
next in thread | previous in thread | raw e-mail | index | archive | help
"José Mª Alcaide" wrote: > After reading a recent message from Jordan (I lost its reference) > about the "makeworldbench" test results, I was a bit worried: is one > of the two CPUs of my new computer being wasted by the FreeBSD > kernel? Uh-oh! Then I did a little test running one of the BYTE > benchmarks (arithoh). I did three runs with one, two and three > simultaneous processes. These are the results: > > [snip] > > In summary, at least making pure arithmetic computing, SMP does > its work. Yes, this is what I was saying to Jordan - as the FreeBSD kernel currently has a 'big lock' around it, meaning on an SMP machine AFAIK only one CPU can be 'in the kernel' at a time, if the other tries - it will get blocked (and probably go carry on with something else?) SMP is still a winner on FreeBSD though - because of Unix's nature to be very 'multi-processed', i.e. when you run Apache you don't just run 1 instance of apache, you run 4 or 5 (or more as the load increases) - so you get to the advantages of more CPU's - up until the system gets kernel bound (or IO bound)... This is what your test results proved... :) So, if you have something like rc5des - your better off running 2 copies of it on an SMP system... ;-) -Kp To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?367A7CDA.EF67AF5E>