Timing of events in an ISR
Joel Sherrill <joel@OARcorp.com>
joel.sherrill at OARcorp.com
Mon Oct 18 20:24:12 UTC 2004
Till Straumann wrote:
> Mark VanderVoord wrote:
>
>> I am sending an event from an ISR. The task waiting
>> on the event is the highest priority task in the
>> system (priority 5... where two others are priority
>> 100+).
>> It appears to take 150 usec on my AU1100 between the
>> actual interrupt firing, and the task being allowed to
>> run. This processor runs at 400MHz, so this delay
>> sounds excessive to me. If I poll the flag instead of
>> using the interrupt, I can act on it within 4 usec
>> (but I am obviously using up a lot more CPU time).
>
>
> I have done similar, excessive tests on powerpcs
> (~300Mhz) and the average latency observed is in the
> 'a few microseconds' ballpark which is what you'd expect.
>
> Could you post your code (including the timing)?
>
> Did you make sure:
> - your ISR doesn't do excessive work
> - your task really has the correct priority
> - task preemption is not disabled (globally or
> for one of your lower priority tasks)
I am beginning to think this is the case. The Init taks is no-preempt
so if he just let that become the application task, it would
exhibit this behavior.
> - your code which determines the latency is correct
>
>>
>> What factors contribute to this delay? Is there
>> anything that I can do about it?
>
>
> There *must* be something wrong...
>
> -- Till
>
>>
>> As a side note, I noticed that there are a lot of
>> timing tests which come with RTEMS. Would any of
>> these be useful in nailing down just how much time
>> these calls should take?
>>
>> Thanks!
>>
>> Mark S VanderVoord
>> Self-Guided Systems, LLC
>>
>> __________________________________________________
>> Do You Yahoo!?
>> Tired of spam? Yahoo! Mail has the best spam protection around
>> http://mail.yahoo.com
>
>
>
--
Joel Sherrill, Ph.D. Director of Research & Development
joel at OARcorp.com On-Line Applications Research
Ask me about RTEMS: a free RTOS Huntsville AL 35805
Support Available (256) 722-9985
More information about the users
mailing list