Messages in this thread | | | Date | Wed, 25 Mar 2009 10:30:52 +0100 | From | Ingo Molnar <> | Subject | Re: [RFC PATCH 00/11] sched: find_busiest_group() cleanup |
| |
* Gautham R Shenoy <ego@in.ibm.com> wrote:
> Hi, > > This patchset contains the cleanup of the humongous > find_busiest_group() function. > > Vaidy had tried a hand at this before. His approach can be > found here: > http://lkml.org/lkml/2008/9/24/201 and > http://lkml.org/lkml/2008/10/9/176 > > Though the code in this patchset has been written from scratch I > have reused some of the ideas that Vaidy had originally proposed. > Credit has been given whereever it is due :) > > The patches in this series are incremental. Each one is a > functional patch, which compiles fine. > > The steps followed in the cleanup are as follows: > - Fix indentations. > > - Group variables that serve a common high-level purpose into a single > structure. > > - Use helper functions to perform all the calculatations, like calculating > the sched_domain and sched_group statistics, calculating the imbalance, etc. > > - Move the power_savings_balance part, which depends on > (CONFIG_SCHED_MC || CONFIG_SCHED_SMT) into a different section, thereby > eliminating the #ifdef jungle in helper functions. > > - Add /** style comments for all the functions, including find_busiest_group() > > - Add additional comments whereever appropriate. > > After applying all the patches, the size of find_busiest_group() > goes down from 313 lines to 76 lines. Of course, there are the > helpers, but effort has been put to keep all the helper functions > within the 80 line limit.
Very nice series!
> Any feedback on the patches and how the functionality can be > tested is greatly appreciated.
I'll try my best to get this tested.
Ingo
| |