Date: Sun, 28 Nov 2004 11:18:24 +0900 From: Rob <spamrefuse@yahoo.com> To: freebsd-stable@freebsd.org Subject: Re: make -j$n buildworld : use of -j investigated Message-ID: <41A93570.6050607@yahoo.com> In-Reply-To: <MDEHLPKNGKAHNMBLJOLKMEKPABAB.davids@webmaster.com> References: <41A91E7E.8050409@yahoo.com> <MDEHLPKNGKAHNMBLJOLKMEKPABAB.davids@webmaster.com>
next in thread | previous in thread | raw e-mail | index | archive | help
David Schwartz wrote: >>According to my formula: >> >> time(minutes) = 1e5 / ( speed(MHz) * nproc ) >> >>and taking nproc = 1, this results in >> >> time = 1e5 / 2798.66 = 36 minutes >> >>Quite accurate for your system as well. At least this formula gives a >>resonable estimate about the compile time. >> >>Apparently HT does not help much, since with SMP kernel, times do not >>get any close to 36/2 = 18 minutes. However, there is a slight improvement >>from -j1 to -j2 with SMP. > > > You're saying that HT does not help because it doesn't double the speed? > That makes no sense. HT helps if it brings the time down. If it brings it > down to 34 minutes, it helps a little. If it brings it down to 24 minutes, > it helps quite a bit. It's unrealistic to even imagine that HT will bring > anything close to the benefit of a second physical CPU. OK, sorry for me rushing into conclusions. All I tried to say was that in cases when 'nproc' plays are role, then nproc=2 with a second CPU, but it is somewhere between 1 and 2 for HT systems; possibly closer to 1 than to 2 (but indeed still larger than 1 !) Rob.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?41A93570.6050607>