Hi, Raphael Hertzog wrote:
> Can you try with the dpkg version that is in the branch pu/async-sync > of git://git.hadrons.org/git/debian/dpkg.git ? Some rough numbers. master+sid+asyncsync: cold cache 16.86user 4.21system 0:50.52elapsed 41%CPU (119major)pagefaults warm cache 16.82user 4.05system 0:43.72elapsed 47%CPU (77major)pagefaults master+sid: cold cache 16.80user 4.36system 1:25.42elapsed 25%CPU (104major)pagefaults warm cache 16.75user 4.24system 1:19.49elapsed 26%CPU (64major)pagefaults master+sid-fsync [1]: cold cache 17.02user 4.21system 0:42.79elapsed 50%CPU (85major)pagefaults warm cache 16.76user 3.95system 0:32.36elapsed 64%CPU (33major)pagefaults lenny (1.14.29): cold cache 16.82user 4.32system 0:52.91elapsed 40%CPU (90major)pagefaults warm cache 16.72user 4.18system 0:45.77elapsed 46%CPU (58major)pagefaults kernel: Linux 2.6.33-2-686 (Debian version 2.6.33-1~experimental.5) filesystem: ext4 with data=ordered disk: an old 4200 rpm 50 GiB laptop drive. nothing interesting going on in the background. all times above are best of 3. I also tried inducing swapping in the background, but (because I am too impatient to run enough trials) that didn’t yield stable numbers. Roughly speaking, master+sid+asyncsync and master+sid-fsync were about the same in that case: one dpkg run takes three minutes or so with some huge margin of error. So: a little faster than lenny, 25% slower than it could be by deciding not to sync, 1½ times as fast as current sid. Not bad. Hope that helps, Jonathan [1] merge of current master with current sid with commits 8467206, 3927a9, 18b12083, 9cd41fd, 20fdb39, 8ccebf62, and 62668eb4 reverted. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
