Re: [Cake] cake review comments cake_drop

[Sebastian Moeller <moeller0@gmx.de>]

Hi Kevin,

On Oct 4, 2015, at 19:46 , Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk> wrote:

> On 04/10/15 11:56, Sebastian Moeller wrote:
>> Hi Jonathan, hi Kevin,
>> 
>> 
>> On Oct 4, 2015, at 11:03 , Jonathan Morton <chromatix99@gmail.com> wrote:
>> 
>>> I remember changing cake_drop so that it iterated over the list of active flows rather than the whole potential array. Hence an inactive class, which has no active queues, costs nothing to iterate over.  Usually the total number of active queues is tolerably small.
>> 	Ah, so there is the answer to my question ;) Jonathan, which of the public repositories is the closest to your internal version, so which is the semi-official one?
> It is exists in Dave's repo.  It iterates over the 'old_flows' and
> 'new_flows' lists for each class looking for the one in most 'distress',
> inherently if there's active data to be found it should be on one of
> those lists.

	Ah, that is what list_for_each_entry() does… Thanks for spelling it out for me…

Best Regards
	sebastian

> 
> In cpu cycle pedant mode, checking for a null list on a set of classes
> we know are going to be empty is unnecessary, in a single class config
> that's 7 (actually 14) null list searches we don't have to do.  I'm sure
> it's not much of a saving but every cycle helps.
> 
>