[nptl] Some statistics

[Sebastien Decugis <sebastien.decugis@ext.bull.net>]

The following stats were obtained with the attached script -- runing for
5 min ($ t0 300 ./script.sh) -- t0 can be found in the forum.

NOTE: the test internal timeout was set to 5 seconds (the machine is not
loaded at all).

SLES 9 with no modification:
Test passes                        : 1420 times
Test times out                     :   42 times
Test hangs during threads creation :   56 times

SLES 9 with vanilla kernel 2.6.8.1:
Test passes                        : 9132 times
Test times out                     :    0 times
Test hangs during threads creation :    8 times

* Note: with a greater number of threads (ex: ./test 500 98) we can
reproduce the timeout.

SLES 9 with vanilla kernel 2.6.8.1 & newer glibc (no patch):
Test passes                        :    0 times
Test times out                     :  110 times
Test hangs during threads creation :    0 times

* Note: with a smaller number of threads (ex: ./test 3 1) we can have
the test passes or hang (about 30% of the runs each).


Looks like this kernel problem is also dependent on the libc... Don't
know what to think... What's your opinion?


Cheers,
Seb.



-------------------------------
Sebastien DECUGIS
NPTL Test & Trace Project
http://nptl.bullopensource.org/

"You may fail if you try.
You -will- fail if you don't."


Liste de diffusion nptl
Pour se désinscrire : mailto:nptl_request@bullopensource.org?subject=unsubscribe