[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [rtl] Update rtl2.2 to 3.0p10
Well, the first thing i can pick out that might be making this not
work is that you need to set the scheduling for the thread where you do the
wait. See below where i inserted code as to how i would do it.
On Tue, Jan 09, 2001 at 07:58:38PM +0100, Peter Helbig wrote:
> Hello Everybody!
>
> I updated from rtlV2.2 to V3.0p10 and found a different behaviour of
> pthread_make_periodic_np(...,...,0). In V2.2 it works as suspected: The
> thread is called once and is sentenced to wait with a
> pthread_wait_np() afterwards. In V3.0p10 the call of
> pthread_wait_np() blocks the thread forever.
>
> Code example:
>
> void *thread_code(void *param)
> {
>
struct sched_param my_sparam;
> // blabla
> --------------- snip -----------------------------------------------
> do
> {
> min_diff = 2000000000;
> max_diff = -2000000000;
>
> for (i = 0; i < ntests; i++)
> {
> now = clock_gethrtime(rtl_getschedclock());
> if (advance && !mode)
> {
> if (now < expected)
> {
> rtl_delay (expected - now);
> }
> now = clock_gethrtime(CLOCK_UST);
> }
>
> /* start of next time element */
> do_rayonic_start();
> do_rayonic(loops);
> do_rayonic_fini();
>
> /* finished user part of measurement */
> fini = clock_gethrtime(rtl_getschedclock());
> samp.fini = fini;
>
> /* calculate time for next activation */
> /* 1) difference from expectation */
> diff = now - expected;
> samp.diff = diff;
> if (diff < min_diff)
> {
> min_diff = diff;
> }
>
> if (diff > max_diff)
> {
> max_diff = diff;
> }
>
> /* 2) theoretical sleeping duration */
> sleepfor = CYCLETIME-(fini-now);
> samp.sleepfor = sleepfor;
>
> /* 3) adaptive control towards expected time */
> sleepfor -= diff;
>
> /* 4) avoid processor overload */
> expected += CYCLETIME;
> while (sleepfor < MAXLOAD)
> {
> sleepfor += CYCLETIME;
> expected += CYCLETIME;
> samp.skipped ++;
> }
>
> /* 5) absolute time for reactivation */
> sleepto = (now+sleepfor);
> samp.sleepto = sleepto;
>
my_sparam.sched_priority = 1;
if (
(retval =
pthread_setschedparam(pthread_self(), SCHED_FIFO,
&my_sparam)) != 0) {
return (void *) retval;
}
> pthread_make_periodic_np
> (pthread_self(),sleepto,0);
> //++++++++++++++++++++++++++++++++++++++++++++++++
> // In V3.0p10 the next call blocks the thread
> forever,
> // in V2.2 it works properly....
> //++++++++++++++++++++++++++++++++++++++++++++++++
> pthread_wait_np();
> }
>
> samp.min = min_diff;
> samp.max = max_diff;
> write (fifo_monitor, &samp, sizeof(samp));
>
> } while (1); // endless...
>
> ----------------- snap
> -----------------------------------------------------------
> } // thread end
>
>
> Any hints will be welcome!
>
>
> Peter
>
>
> Peter Helbig
> Rayonic Sensor Systems GmbH
> Entwicklung Software
> Bayreuther Strasse 108
> D-91301 Forchheim
>
> Tel: ++49 (0) 9191 72 818-0
> Fax: ++49 (0) 9191 640 333
> Email: Peter.Helbig@rayonic.de
> http://www.rayonic.de
> -- [rtl] ---
> To unsubscribe:
> echo "unsubscribe rtl" | mail majordomo@rtlinux.org OR
> echo "unsubscribe rtl " | mail majordomo@rtlinux.org
> ---
> For more information on Real-Time Linux see:
> http://www.rtlinux.org/rtlinux/
>
>