Hello Thomas, I am curious about your test.
I am assuming that you built kernel-2.4.4 under 2.2.19 and 2.4.4?? Why did you use -j4 instead of -j2 on a dual processor box?? Best Regards, --George On Monday 14 May 2001 08:32, T. Weyergraf wrote: > Hi all, > > triggered by a discussion in irc.openprojects.net, #alphalinux, I've spend > some time on benchmarking various linux kernels and came across some very > strange results. > > Platform: Dual 667 UP2000, 1.5gig ram, 3 adaptecs with 5 disks total, > serial console. > > Kernel 2.4.4 seems to suffer a *tremendous* performance impact, compared to > 2.2.19. A 'time make -j4 boot' of the kernel-2.4.4 sources gives the > following compile-times: > > 2.4.4 2.2.19 > real 3m58 2m38 > user 4m23 2m23 > sys 0m14 0m11 > > Several other tests seem to indicate a similar decrease in performance wrt > real-time, some of them being CPU and/or memory-bound only. > > In addition to that, i've noticed, that kernel 2.2.19 frequently crashes, > after giving the well-known "stuck spinlock" messages in rather random > processes. No panics are given, the system just freezes completely ( no > ping and the like ). > > So, obviously, my two questions are: > > 1. Has anybody ever observed the performance impact, i'm seeing with kernel > 2.4.x ? 2. Since i moved to kernel 2.4.x very early for various reasons, I > would like to know, which of the recent 2.2.x kernels can be considered > stable on a alpha/SMP platform ( ideally being a UP2000 as well ). > > I'm a bit confused here ;-) > > Regards, > Thomas Weyergraf
