Re: Priority Inheritance Test (Real-Time Preemption)

[Ingo Molnar <mingo@elte.hu>]

* Esben Nielsen <simlo@phys.au.dk> wrote:

> > and i'm regularly testing this property with 'hackbench 50', which
> > creates over a 1000 wildly scheduling non-RT tasks. Latency is not
> > affected by such workloads.
> >
>  
> Probably not. Even while doing that you most likely wont build up wait
> lists of more than 10, maybe 100 tasks? Doing full traversals with irq
> disabled probably wont be meassureable!(?) compared to much other
> stuff increasing responsible for the meassured latency.

there is no full list traversal of SCHED_NORMAL tasks, ever.

but the best way is to test this yourself, download Florian's rtc_wakeup
from:

  http://www.affenbande.org/~tapas/wiki/index.php?rtc_wakeup

and run it with the highest possible resolution, 8192 Hz:

  chrt -f 98 -p `pidof 'IRQ 8'`
  chrt -f 99 -p `pidof 'IRQ 0'`

  ./rtc_wakeup -f 8192 -t 100000

in this mode rtc_wakeup will report the worst irq-delivery latency it
measures. It will thus measure the combined effect of any type of
scheduling or irqs-off latency to RT-tasks.

then download hackbench from:

  http://developer.osdl.org/craiger/hackbench/

and try e.g.:

  ./hackbench 50

this will start 2x20x50 == 2000 SCHED_NORMAL threads, all performing a
nice pattern of scheduling simulating a busy chat server workload with
tons of messages going back and forth.

	Ingo