On 9 October 2014 13:36, Peter Zijlstra peterz@infradead.org wrote:
On Tue, Oct 07, 2014 at 02:13:35PM +0200, Vincent Guittot wrote:
Monitor the usage level of each group of each sched_domain level. The usage is the amount of cpu_capacity that is currently used on a CPU or group of CPUs. We use the utilization_load_avg to evaluate the usage level of each group.
The utilization_avg_contrib only takes into account the running time but not the uArch so the utilization_load_avg is in the range [0..SCHED_LOAD_SCALE] to reflect the running load on the CPU. We have to scale the utilization with the capacity of the CPU to get the usage of the latter. The usage can then be compared with the available capacity.
You say cpu_capacity, but in actual fact you use capacity_orig and fail to justify/clarify this.
you're right it's cpu_capacity_orig no cpu_capacity
cpu_capacity is the compute capacity available for CFS task once we have removed the capacity that is used by RT tasks.
We want to compare the utilization of the CPU (utilization_avg_contrib which is in the range [0..SCHED_LOAD_SCALE]) with available capacity (cpu_capacity which is in the range [0..cpu_capacity_orig]) An utilization_avg_contrib equals to SCHED_LOAD_SCALE means that the CPU is fully utilized so all cpu_capacity_orig are used. so we scale the utilization_avg_contrib from [0..SCHED_LOAD_SCALE] into cpu_usage in the range [0..cpu_capacity_orig]
The frequency scaling invariance is not taken into account in this patchset, it will be solved in another patchset
Maybe explain what the specific invariance issue is that is skipped over for now.
ok. I can add description on the fact that if the core run slower, the tasks will use more running time of the CPU for the same job so the usage of the cpu which is based on the amount of time, will increase
Signed-off-by: Vincent Guittot vincent.guittot@linaro.org
kernel/sched/fair.c | 13 +++++++++++++ 1 file changed, 13 insertions(+)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index d3e9067..7364ed4 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -4551,6 +4551,17 @@ static int select_idle_sibling(struct task_struct *p, int target) return target; }
+static int get_cpu_usage(int cpu) +{
unsigned long usage = cpu_rq(cpu)->cfs.utilization_load_avg;unsigned long capacity = capacity_orig_of(cpu);if (usage >= SCHED_LOAD_SCALE)return capacity + 1;Like Morten I'm confused by that +1 thing.
ok. the goal was to point out the erroneous case where usage is out of the range but if it generates confusion, it can remove it
return (usage * capacity) >> SCHED_LOAD_SHIFT;+}
A comment with that function that it returns capacity units might clarify shift confusion Morten raised the other day.
ok. i will add a comment