On Thu, Jan 23, 2014 at 03:39:36PM +0100, Daniel Lezcano wrote: > On 01/23/2014 01:58 PM, Peter Zijlstra wrote: > >On Tue, Jan 21, 2014 at 12:17:57PM +0100, Peter Zijlstra wrote: > >>From: Daniel Lezcano <[email protected]> > >> > >>The idle_balance modifies the idle_stamp field of the rq, making this > >>information to be shared across core.c and fair.c. As we can know if the > >>cpu is going to idle or not with the previous patch, let's encapsulate the > >>idle_stamp information in core.c by moving it up to the caller. The > >>idle_balance function returns true in case a balancing occured and the cpu > >>won't be idle, false if no balance happened and the cpu is going idle. > >> > >>Cc: [email protected] > >>Cc: [email protected] > >>Cc: [email protected] > >>Signed-off-by: Daniel Lezcano <[email protected]> > >>Signed-off-by: Peter Zijlstra <[email protected]> > >>Link: > >>http://lkml.kernel.org/r/[email protected] > >>--- > >> kernel/sched/core.c | 2 +- > >> kernel/sched/fair.c | 14 ++++++-------- > >> kernel/sched/sched.h | 2 +- > >> 3 files changed, 8 insertions(+), 10 deletions(-) > >> > >>--- a/kernel/sched/core.c > >>+++ b/kernel/sched/core.c > >>@@ -2680,7 +2680,7 @@ static void __sched __schedule(void) > >> pre_schedule(rq, prev); > >> > >> if (unlikely(!rq->nr_running)) > >>- idle_balance(rq); > >>+ rq->idle_stamp = idle_balance(rq) ? 0 : rq_clock(rq); > > > >OK, spotted a problem here.. > > > >So previously idle_stamp was set _before_ actually doing idle_balance(), > >and that was RIGHT, because that way we include the cost of actually > >doing idle_balance() into the idle time. > > > >By not including the cost of idle_balance() you underestimate the 'idle' > >time in that if idle_balance() filled the entire idle time we account 0 > >idle, even though we had 'plenty' of time to run the entire thing. > > > >This leads to not running idle_balance() even though we have the time > >for it. > > Good catch. How did you notice that ?
Staring at that code for too long :-) > >So we very much want something like: > > > > > > if (!rq->nr_running) > > rq->idle_stamp = rq_clock(rq); > > > > p = pick_next_task(rq, prev); > > > > if (!is_idle_task(p)) > > rq->idle_stamp = 0; > > Is this code assuming idle_balance() is in pick_next_task ? Yeah, I'm trying to make that work. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
