[RFC patch 0/2] sched: dynamically adapt granularity with nr_running

[Mathieu Desnoyers]

Changing the minimum granularity is a double-edged sword: if we set it to a too
small value, then the scheduler will preempt tasks too often. If it is too
large, then the "latency" period can grow very large as the number of running
tasks increases.

The first patch leaves the same scheduling granularity when there are few tasks
on the system (3 or less), but dynamically adapts (shrinks) the sched
granularity when there are more. At a ceiling value of 8 running tasks (this
choice is arbitrary), it grows the latency rather than shrinking granularity
further to ensure we don't end up calling the scheduler too often.

The second patch ensures that awakened sleeping tasks don't get affected by
shrinked minimum granularity.

Comments are welcome,

Thanks,

Mathieu

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/