[haiku-gsoc] Re: some questions about haiku kernel

  • From: Yongcong Du <ycdu.vmcore@xxxxxxxxx>
  • To: haiku-gsoc@xxxxxxxxxxxxx
  • Date: Sat, 12 May 2012 21:17:52 +0800

Hi Axel,

On Sat, May 12, 2012 at 12:21 AM, Axel Dörfler <axeld@xxxxxxxxxxxxxxxx>wrote:

> On 11.05.2012 15:35, Yongcong Du wrote:
>> > I assume you mean if you remove the timer in case of the idle thread
>> > only?
>> all thread have the same behavior -- the oldThread and nextThread are
>> the same thread. But the phenomenon is obvious when the thread is the
>> idle thread.  And I found that there is no such phenomenon if I
>> commented out the quantumTimer related code
> The reason is simply scheduling: preemption is just one case of
> scheduling, but if you have longer running threads, they could run more or
> less forever without that timer, draining other long running threads (in
> this case with the same or a lower thread priority).
> The quantum is used to disrupt the thread, and look if another one should
> run next. You could disable the timer for the idle thread, as that one will
> always be preempted in order to save energy, but you cannot get rid of the
> timer completely.

Thanks very much. With your help, now I got the quantumTimer point. I can't
imagine haiku scheduler code is so tight(compared with linux thousands of
LOCs :D) and works well so far.

>  After digging into the source code for one fully day, I still can't
>> understand what's reschedule_even used for
> That's why you can always ask on this list ;-)
> If I didn't manage to write something understandable, please just ask
> again.
> HTH,

Appreciate your kind help,

Other related posts: