Skip to contentSkip to conversation

sched_yield behavior

DISCUSSION Giovanni Tusa 3 participants 3 messages Feb 27, 2005

lore
Share & tools
lore
#0 lore raw

Hi all,
I have a question about the sched_yield behavior of Linux O(1) scheduler, for RT tasks.
By reading some documentation, I found that " ....real-time tasks are a special case, because
when they want to explicitly yield the processor to other waiting processes, they are merely
moved to the end of their priority list (and not inserted into the expired array, like conventional
processes)."
I have to implement an RT task with the highest priority in the system (it is also the only task within the
priority list for such priority level). Moreover, it has to be a SCHED_FIFO task, so that it can preempt
SCHED_RR ones, because of its strong real-time requirements. However, sometimes it should relinquish the
CPU, to give to other tasks a chance to run.
Now, what happen if it gives up the CPU by means of the sched_yield() system call?
If I am not wrong, the scheduler will choose it again (it will be still the higher priority task, and the only of its priority list).
I have to add an explicit sleep to effectively relinquish the CPU for some time, or the scheduler can deal with such a
situation in another way?

Any advice will be appreciated.
Giovanni

#1 lore raw
Quoted 16 lines · Giovanni Tusa

Hi all,
I have a question about the sched_yield behavior of Linux O(1) scheduler, for RT tasks.
By reading some documentation, I found that " ....real-time tasks are a special case, because
when they want to explicitly yield the processor to other waiting processes, they are merely
moved to the end of their priority list (and not inserted into the expired array, like conventional
processes)."
I have to implement an RT task with the highest priority in the system (it is also the only task within the
priority list for such priority level). Moreover, it has to be a SCHED_FIFO task, so that it can preempt
SCHED_RR ones, because of its strong real-time requirements. However, sometimes it should relinquish the

Quoted 8 lines

CPU, to give to other tasks a chance to run.
Now, what happen if it gives up the CPU by means of the sched_yield() system call?
If I am not wrong, the scheduler will choose it again (it will be still the higher priority task, and the only of its priority list).
I have to add an explicit sleep to effectively relinquish the CPU for some time, or the scheduler can deal with such a
situation in another way?

What exactly are you trying to do? I don't understand how the task could have "strong real-time requirements" if it's CPU bound. What is the exact nature of the real time constraint?

Lee

#2 lore raw
In reply to Giovanni Tusa
Quoted 5 lines · Giovanni Tusa

If I am not wrong, the scheduler will choose it again (it will be still the higher priority task, and the only of its priority list).
I have to add an explicit sleep to effectively relinquish the CPU for some time, or the scheduler can deal with such a
situation in another way?

Yes, the scheduler will choose it again. This behaviour is also specified in the relevant standards.

Your alternatives may be to use other methods of userspace synchronisation (eg. pipes, semaphores), or to use timers.