Adi Sudewa wrote:
>
> On Fri, 9 Jun 2000, Stuart Hughes wrote:
> > > Suppose I have this procedure :
> > >
> > > void do_it()
> > > {
> > > complex_calculation();
> > > other_complex_calculation();
> > > }
> > >
> > > and this highest priority rtl thread(ignore syntax errors):
> > >
> > > pthread_make_periodic(PERIOD);
> > > while (1) {
> > > do_it();
> > > pthread_wait();
> > > }
> > >
> > > (2) How to measure the worst-case execution time of do_it() ?
> > > Can I just run it 100 times, do measurement, and pick the greatest
> > > value ?
> > >
> >
> > Set the period to be much longer than expected (so you won't lock-up).
> > Take the time before and after do_it() and then calculate the worst
> > case. This can then be output with printk (every once in a while). If
> > you then run it, you can load the system and observe the results.
>
> But if I port my application into another machine, then I must redo the
> measurement.
>
Yes & no. You can make an estimate of the time for the new machine by
comparing clock speeds. This is a very rough calculation, but on the
same architecture machine should give an idea of what to expect.
> > To know whether you got the absolute worst case is difficult, but you
> > should get a close idea of it from this test.
>
> Can I do statistic method to guess WCET (worst case execution time)
> from , say, 1000 measurement ?
>
You can run 1000 measurements and take the worst. Of course the results
will depend on test environment (I/O activity etc)
Regards, Stuart
-- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
---
For more information on Real-Time Linux see:
http://www.rtlinux.org/rtlinux/
