[RFC] [PATCH 0/1] hrtimers: Remove unnecessary clock events deviceprogramming

From: Ashwin Chaugule
Date: Thu Aug 27 2009 - 17:49:00 EST

hrtimer_force_reprogram() was being called quite often unnecessarily. The following patch caches the hrtimer that causes the cpu_base->expires_next value to change. This way, we force the reprogram only when remove_hrtimer() finds that its going to remove an hrtimer which was about expire next. IOW, we forgo reprogramming the clock events device, if the current timer to be removed isn't at the root of the expires tree.

Also removed the hrtimer_is_queued check from remove_hrtimer and made appropriate changes to all callers. That check seemed redundant.


Shows ~60% reduction in the following calls to program clock events device.
- tick_nohz_restart_sched_tick (tick_sched_timer)
- hrtimer_start (tick_sched_timer)

I haven't traced all the callers yet, this was from a hack to timer_stats. If you guys have a better way to test this, I'll give that a go.

Tested on v2.6.29 on an ARM based embedded device.

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/