lkml.org 
[lkml]   [2013]   [Apr]   [30]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
SubjectRe: [PATCH 1/2] Make the batch size of the percpu_counter configurable
From
Date
On Tue, 2013-04-30 at 17:28 +0000, Christoph Lameter wrote:
> On Tue, 30 Apr 2013, Tim Chen wrote:
>
> > On Tue, 2013-04-30 at 13:32 +0000, Christoph Lameter wrote:
> > > On Mon, 29 Apr 2013, Tim Chen wrote:
> > >
> > > > diff --git a/include/linux/percpu_counter.h b/include/linux/percpu_counter.h
> > > > index d5dd465..5ca7df5 100644
> > > > --- a/include/linux/percpu_counter.h
> > > > +++ b/include/linux/percpu_counter.h
> > > > @@ -22,6 +22,7 @@ struct percpu_counter {
> > > > struct list_head list; /* All percpu_counters are on a list */
> > > > #endif
> > > > s32 __percpu *counters;
> > > > + int *batch ____cacheline_aligned_in_smp;
> > > > };
> > >
> > > What is this for and why does it have that alignmend?
> >
> > I was assuming that if batch is frequently referenced, it probably
> > should not share a cache line with the counters field.
>
> And why is it a pointer?

A pointer because the default percpu_counter_batch value could change
later when cpus come online after we initialize per cpu counter and
percpu_counter_batch will get computed again in percpu_counter_startup.
Making it a pointer will make it unnecessary to come back and change the
batch sizes if we use static batch value and default batch size.

>
> And the pointer is so frequently changed that it needs it own cache line?
>

On second thought, your're right. It is unnecessary for *batch to have
its own cache line as the counters pointer and head_list above it will
not change frequently. I'll remove the cache alignment.

Tim


\
 
 \ /
  Last update: 2013-04-30 20:21    [W:0.069 / U:0.044 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site