On Wed, 2003-10-29 at 17:39, Soeren Sonnenburg wrote: > On Wed, 2003-10-29 at 17:28, Michel Dänzer wrote: > > On Wed, 2003-10-29 at 09:51, Soeren Sonnenburg wrote: > > > On Wed, 2003-10-29 at 09:43, Benjamin Herrenschmidt wrote: > > > > > > > > You may be encountering a different issue ... (2.6 is usually overall > > > > faster than 2.4 here, > > > > Same here. Soeren, have you verified the difference actually is related > > to X, e.g. by redirecting the output to a file? Did you conduct both > > tests as soon after bootup as possible to minimize the influence of > > caching? ... > > Yes fresh reboot - both this morning. > A time find ./ >/dev/null says on 2.6. : > time find >/dev/null > > real 4m27.772s > user 0m0.855s > sys 0m11.292s > > while it seems slower than 2.4 (not yet tested) it does not justify the > 28minutes to display the files :-/
Weird. Maybe this (rather synthetic BTW - it doesn't feel slow doing 'real' work, does it?) test tickles very bad behaviour in something which has changed radically in the 2.6 kernel, the scheduler maybe? CONFIG_PREEMPT might make a difference, pity it causes crashes and all around weird behaviour about as soon as RAM is fully used. -- Earthling Michel Dänzer | Debian (powerpc), X and DRI developer Software libre enthusiast | http://svcs.affero.net/rm.php?r=daenzer
