Julian,

There used to be a (IIRC) 5 second time limit for stuck / runaway real time 
threads where it would be terminated by the OS to save the rest of the system - 
in the old days of single core Macs they were capable of hanging the whole 
computer including the mouse pointer, so the OS put a limit on them otherwise 
it wasn’t possible to get control back. This limit could be changed using an 
NVRAM parameter if needed.

This implicitly conveys a point which is that it would be impossible for a 
single real time thread to freeze a multi core Mac, perhaps this is of help?

Tim.


On 1 Jun 2016, at 10:00, Julian Dessecker <[email protected]> wrote:

> Tim,
> 
> thanks for that information. I had a look at the xnu source - you are right. 
> The constraint won't prevent a real-time thread from starving non-realtime 
> threads.
> 
> Julian
> 
> On Tue, May 31, 2016 at 9:51 PM, Tim Hewett <[email protected]> wrote:
> Julian,
> 
> The constraints parameters were ignored in the kernel for many generations of 
> OS X in spite of the documentation defining their use. The real documentation 
> (the source code) said otherwise.
> 
> I am not sure if they are still ignored, IIRC the kernel-side code is not 
> hard to find to check.
> 
> Tim.
> 
> 
> On 31 May 2016, at 20:00, [email protected] wrote:
> 
> > Hi,
> >
> > I'm trying to create real-time threads via the
> > THREAD_TIME_CONSTRAINT_POLICY.
> > It works fine, an interrupt after each computation cycle can be seen inside
> > the profiler (Instruments / System Trace) but the constraint seems to be
> > ignored.
> >
> > I would expect a penalty after the constraint time is over but the thread
> > stays on the CPU starving other threads.
> >
> > That's the way I initialize the time constraint policy:
> >
> >    thread_time_constraint_policy_data_t time_constraints;
> >
> >    time_constraints.period = (int) NanosecondsToAbsolute (1000000); // 1
> > ms
> >
> >    time_constraints.computation = (int) NanosecondsToAbsolute (250000); //
> > 0.25 ms
> >
> >    time_constraints.constraint = (int) NanosecondsToAbsolute(500000); //
> > 0.5 ms
> >
> >    time_constraints.preemptible = 1;
> >
> >    thread_policy_set(mach_thread_self (), THREAD_TIME_CONSTRAINT_POLICY,
> > (thread_policy_t) &time_constraints, THREAD_TIME_CONSTRAINT_POLICY_COUNT);
> >
> > Any ideas how to get the constraint working?
> >
> > Regards,
> > Julian
> 
> 


 _______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list      ([email protected])
Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/coreaudio-api/archive%40mail-archive.com

This email sent to [email protected]

Reply via email to