Date: Fri, 11 May 2007 23:40:05 +0200
From: Benjamin Lutz <mail@maxlor.com>
To: ports@freebsd.org
Subject: Benchmark: -j{$NCPU} vs -j{$NCPU+1}
Message-ID: <200705112340.09640.mail@maxlor.com>
next in thread | raw e-mail | index | archive | help
--nextPart19990703.rZQdD007cc Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Hello, Now that I'm back from having to serve in the armed services, I've=20 picked up work on parallel building for the ports again. There is some=20 conventional wisdom floating around these parts that says you should=20 run make -j with a number one higher than the number of CPUs you have.=20 As I wondered what difference that makes, I ran some benchmarks. Since=20 maybe someone else is curious about this too, I'll publish my results=20 here. Here's the summary: on my core 2 duo machine (=3D 2 CPUs), this is the=20 time improvement to build the port (just configuring and building, no=20 installation) with -j3 over -j2: x11-toolkits/gtk20: 2% x11/kdebase3: 1% graphics/gimp-app: 24% lang/gcc41: 12% So basically, -j3 really is faster than -j2, although the difference is=20 substantial only for some ports. Here's the script I use to produce the numbers, as well as my raw=20 results which contain much more detail: http://www.maxlor.com/freebsd/files/parallel_benchmark.sh http://www.maxlor.com/freebsd/files/parallel_benchmark.ods Cheers Benjamin --nextPart19990703.rZQdD007cc Content-Type: application/pgp-signature -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux) iD8DBQBGROK5zZEjpyKHuQwRAjlaAJ4lGWxnwHSOkGEfKd+Ba14yfWamLACeK+Nx jMHFByCWHjOsTxXnQqIpsn8= =di5W -----END PGP SIGNATURE----- --nextPart19990703.rZQdD007cc--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200705112340.09640.mail>