doubts regarding POSIX semaphores
sashti srinivasan
svasn_rtems at yahoo.com
Wed Mar 31 12:15:56 UTC 2004
Hello All,
I have a few doubts to ask:
(1) I configure with a maximum posix semaphores of
110. Create 100 semaphores, delete all of them, again
do sem_init 100 times, again delete them. i.e., I
created 100 posix semaphores and deleted them 100
times. I liked to repeat this for a large number of
times. But while doing on 3rd time, sem_init fails
and puts ENOSPC on errno. According to the posix_user
document, the reasons for this are:
(a) A resource required to initialize the semaphore
has been exhausted. I am not very clear with this.
May anyone give me some ideas regarding what might
have went wrong.
(b) The limit on semaphores(SEM_VALUE_MAX)reached.
I do not think this might be the reason. I gave 5 for
semvalue. Moreover, I gave same value all the times.
If this is wrong, sem_init could not have succeeded
first two(hundred) times.
Thanks in advance for ideas and suggestions.
(2) I have been doing some timing tests and have been
asking for ideas and suggestions based on them. Among
others, I have been asked to double check the timing
tests. I too have this suspicion on these results.
Particularly, I see a substantial overhead in
measurement. I suppose it takes 3 uSec to measure. I
ensured that effect of this overhead is minimum, by
measuring only before and after a large number of
operations and took average. Regarding errors, I am
not getting sufficient hints of bugs/errors, other
than the obvious ones which I have corrected even
before asking for ideas. Now, assuming that results
are OK, may I have some clarification on the following
aspect?
I run a test. sleep for a while. Run the tests
again. I get different results. But If I reboot the
system(pc686), I get repeatable results. May I have
some explanation of this?
With Regards
Srinivasan
__________________________________
Do you Yahoo!?
Yahoo! Finance Tax Center - File online. File on time.
http://taxes.yahoo.com/filing.html
More information about the users
mailing list