Date: Tue, 03 Dec 1996 22:37:33 +0800 From: Peter Wemm <peter@spinner.dialix.com> To: "Eric J. Chet" <ejc@gargoyle.bazzle.com> Cc: smp@freebsd.org Subject: Re: performance Message-ID: <199612031437.WAA04281@spinner.DIALix.COM> In-Reply-To: Your message of "Tue, 03 Dec 1996 09:05:10 EST." <Pine.BSF.3.95.961203085847.4147A-100000@gargoyle.bazzle.com>
next in thread | previous in thread | raw e-mail | index | archive | help
"Eric J. Chet" wrote: > > On Mon, 2 Dec 1996, Eric J. Chet wrote: > > > Kernel compile times: > > > > smp-kernel compiled under -current (as of 12/01) > > >time make > > > 314.39s real 282.00s user 18.72s system > > > > smp-kernel compiled under -smp-current > > >time make -j8 > > > 213.57s real 338.08s user 77.12s system > > > > Hello > Just looking at the numbers again, food for thought. > > (u+s)/r == utilization. > > -current > (282.00+18.72)/314.39 = .96 > > -smp-current > (338.08+77.12)/213.57 = 1.94 !!! very good! > > I know there is a lot of development work to be done, but -smp is > definitly on the right path. Actually, those stats are biased and can't be trusted. The only thing you can really count on is the "elapsed" time improvement, especially since our u+s accounting is somewhat bogus at present... (ie: what one cpu is doing is accredited to all currently running processes, from memory). What's a much fairer indication is: 314.39/213.57 = 1.47 Not quite as good, but still a way to go. > Eric J. Chet > - ejc@bazzle.com Cheers, -Peter
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199612031437.WAA04281>