Hi Joonsoo,
On 08/05/2013 01:02 PM, Joonsoo Kim wrote:
>>> + if (busiest->group_imb) {
>>> + busiest->sum_weighted_load =
>>> + min(busiest->sum_weighted_load, sds->sd_avg_load);
>>
>> Right here we get confused as to why the total load is being compared
>> against load per task (although you are changing it to load per task above).
>
> Yes, you are right. I will add load_per_task to struct sg_lb_stats.
>
>
>>> @@ -4771,12 +4763,13 @@ static inline void calculate_imbalance(struct
>>> lb_env *env, struct sd_lb_stats *s
>>> * Be careful of negative numbers as they'll appear as very large values
>>> * with unsigned longs.
>>> */
>>> - max_pull = min(sds->max_load - sds->avg_load, load_above_capacity);
>>> + max_pull = min(busiest->avg_load - sds->sd_avg_load,
>>> + load_above_capacity);
>>
>> This is ok, but readability wise max_load is much better. max_load
>> signifies the maximum load per task on a group in this sd, and avg_load
>> signifies the total load per task across the sd. You are checking if
>> there is imbalance in the total load in the sd and try to even it out
>> across the sd. Here "busiest" does not convey this immediately.
>
> You may be confused. max_load doesn't means the maximum load per task.
> It means that the busiest group's load per cpu power. And here max doesn't
> mean maximum load in this sd, instead, load of busiest sg. IMO,
> busiest->avg_load convey proper meaning.
Yes sorry about that. It is load per cpu power. My point was that
max_load helped in readability. Look at the these two lines in your patch.
- max_pull = min(sds->max_load - sds->avg_load, load_above_capacity);
+ max_pull = min(busiest->avg_load - sds->sd_avg_load,
>>> + load_above_capacity);
The deleted line says that there is an imbalance in the load/cpu in some
part of the sd(max_load), which is above the average load of the sd, which
we are trying to even out. Something like the below diagram for the sd load.
max_load
_/\________ avg_load.
I believe the line you have added makes it hard to understand this.
Anyway this is a minor issue, you can ignore it.
Regards
Preeti U Murthy
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
