Subject: pthread concurrency & time(1) results
To: NetBSD current <current-users@netbsd.org>
From: Nicolas Joly <njoly@pasteur.fr>
List: current-users
Date: 04/06/2004 18:32:43
Hi,

I made some pthread concurrency tests on a DELL PowerEdge 6300 i386
machine (running -current 20040406), with 4 Pentium III 550MHz CPUs.

It seems to work fine, except that time(1) report some strange values
: 739% cpu for 3 threads running on a 4CPU machine seems a little too
much ;-)

The tests were run with PTHREAD_CONCURRENCY=4 :

hmmcalibrate --cpu 1 globin.hmm
        49.51s user 0.01s system 107% cpu 45.965 total
        50.20s user 0.05s system 107% cpu 46.894 total
        49.81s user 0.03s system 104% cpu 47.745 total
hmmcalibrate --cpu 2 globin.hmm
        57.50s user 0.10s system 196% cpu 29.274 total
        51.30s user 0.07s system 182% cpu 28.196 total
        100.28s user 0.14s system 353% cpu 28.394 total
hmmcalibrate --cpu 3 globin.hmm
        75.23s user 0.19s system 353% cpu 21.309 total
        63.15s user 0.09s system 297% cpu 21.237 total
        161.37s user 0.29s system 739% cpu 21.854 total
hmmcalibrate --cpu 4 globin.hmm
        74.25s user 0.19s system 423% cpu 17.562 total
        78.77s user 0.16s system 451% cpu 17.470 total
        74.74s user 0.21s system 421% cpu 17.790 total

Did i missed something, or do we have some counters problems ?

Thanks in advance,
Regards.

-- 
Nicolas Joly

Biological Software and Databanks.
Institut Pasteur, Paris.