lkml.org 
[lkml]   [2002]   [Dec]   [8]   [last100]   RSS-feed
Views: [more markup]  [less markup]  [headers]  [forward] 
 
Messages in this thread
Complete thread (EXPERIMENTAL)
/
DateSun, 8 Dec 2002 08:25:59 -0500 (EST)
FromBill Davidsen <>
Subject[BENCHMARK] ctxbench 2.4.18 and 2.5.50
Digg This
The results of these tests indicate that there may be some odd SMP locking issues still in 2.5.xx which were either much less obvious or not present in 2.4.

The oddness I see

As these results show, best results were from using a uni kernel, but an SMP kernel with "nosmp" was faster than running SMP. The problem is that with 2.5 kernels the SMP results for multiple runs vary by up to 2:1, and for uni kernels it's more like 5-10% max. I have tried longer runs, more runs, and gotten the same results even in single user mode.
The ctxbench measures the number of times two processes can take turns running using various forms of IPC. I'll post a more complete sescription of the benchmark and a link later today. The IPC in question is signals, semiphores, and message queues. Pipes are disabled (didn't work in 2.5.50) and sockets are not currently tested.

I am going to enable the pipe testing code, and will rerun the tests and post the results as well as the link to source code by tomorrow at the latest.



================================================================
    Run information
================================================================
Run: 2.4.18-10-bl
CPU_MHz 498.014
CPUtype Celeron (Mendocino)
Ncpu 1
Run: 2.4.18-10smp-bl
CPU_MHz 498.007
CPUtype Celeron (Mendocino)
Ncpu 2
Run: 2.4.18-10uni-bl
CPU_MHz 498.016
CPUtype Celeron (Mendocino)
Ncpu 1


================================================================
    Results by IPC type
================================================================
loops/sec
SIGUSR1 low high average 2.4.18-10-bl 70703 71028 70914 2.4.18-10smp-bl 31772 32508 32181 2.4.18-10uni-bl 63017 63430 63165
loops/sec
message queue low high average 2.4.18-10-bl 129994 130446 130212 2.4.18-10smp-bl 45030 50691 48218 2.4.18-10uni-bl 105649 106529 106180
loops/sec
semiphore low high average 2.4.18-10-bl 141706 142895 142358 2.4.18-10smp-bl 44712 45401 44983 2.4.18-10uni-bl 113527 114013 113774
loops/sec
spin+yield low high average 2.4.18-10-bl 286882 300989 293864 2.4.18-10smp-bl 592585 648768 620266 2.4.18-10uni-bl 173367 185063 179160
loops/sec
spinlock low high average 2.4.18-10-bl 3 3 3 2.4.18-10smp-bl 1206672 1207471 1207192 2.4.18-10uni-bl 3 3 3



================================================================
    Run information
================================================================
Run: 2.5.50-bl
CPU_MHz 497.950
CPUtype Celeron (Mendocino)
Ncpu 1
Run: 2.5.50smp-bl
CPU_MHz 497.953
CPUtype Celeron (Mendocino)
Ncpu 2
Run: 2.5.50uni-bl
CPU_MHz 497.914
CPUtype Celeron (Mendocino)
Ncpu 1


================================================================
    Results by IPC type
================================================================
loops/sec
SIGUSR1 low high average 2.5.50-bl 66098 66638 66321 2.5.50smp-bl 6586 47026 31266 2.5.50uni-bl 55482 55647 55565
loops/sec
message queue low high average 2.5.50-bl 120720 122101 121574 2.5.50smp-bl 52365 74617 66059 2.5.50uni-bl 101765 101832 101795
loops/sec
semiphore low high average 2.5.50-bl 129564 130683 129992 2.5.50smp-bl 60871 87154 71212 2.5.50uni-bl 107290 107719 107523
loops/sec
spin+yield low high average 2.5.50-bl 327325 329358 328365 2.5.50smp-bl 391104 560007 492187 2.5.50uni-bl 249595 249734 249685
loops/sec
spinlock low high average 2.5.50-bl 3 3 3 2.5.50smp-bl 1195403 1196519 1196095 2.5.50uni-bl 3 3 3

-- 
bill davidsen, CTO TMR Associates, Inc <davidsen@tmr.com>
  Having the feature freeze for Linux 2.5 on Hallow'een is appropriate,
since using 2.5 kernels includes a lot of things jumping out of dark
corners to scare you.

\
ISP Services
Valid XHTML 1.0!\ /
Valid CSS! Last update: 2005-03-22 12:31    [W:0.431 / U:0.670 seconds]
©2003-2005 Jasper Spaans