lkml.org 
[lkml]   [2009]   [Aug]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: Help Resource Counters Scale Better (v3)
* KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> [2009-08-10 15:22:05]:

> On Mon, 10 Aug 2009 14:45:59 +0900
> KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
>
> > > Do you agree?
> >
> > Ok. Config is enough at this stage.
> >
> > The last advice for merge is, it's better to show the numbers or
> > ask someone who have many cpus to measure benefits. Then, Andrew can
> > know how this is benefical.
> > (My box has 8 cpus. But maybe your IBM collaegue has some bigger one)
> >
> > In my experience (in my own old trial),
> > - lock contention itself is low. not high.
> > - but cacheline-miss, pingpong is very very frequent.
> >
> > Then, this patch has some benefit logically but, in general,
> > File-I/O, swapin-swapout, page-allocation/initalize etc..dominates
> > the performance of usual apps. You'll have to be careful to select apps
> > to measure the benfits of this patch by application performance.
> > (And this is why I don't feel so much emergency as you do)
> >
>
> Why I say "I want to see the numbers" again and again is that
> this is performance improvement with _bad side effect_.
> If this is an emergent trouble, and need fast-track, which requires us
> "fix small problems later", plz say so.
>

OK... I finally got a bigger machine (24 CPUs). I ran a simple
program called parallel_pagefault, which does pagefault's in parallel
(runs on every other CPU) and allocates 10K pages and touches the
data allocated, unmaps and repeats the process. I ran the program
for 300 seconds. With the patch, I was able to fault in twice
the number of pages as I was able to without the patch. I used
perf tool from tools/perf in the kernel

With patch

Performance counter stats for '/home/balbir/parallel_pagefault':

7188177.405648 task-clock-msecs # 23.926 CPUs
423130 context-switches # 0.000 M/sec
210 CPU-migrations # 0.000 M/sec
49851597 page-faults # 0.007 M/sec
5900210219604 cycles # 820.821 M/sec
424658049425 instructions # 0.072 IPC
7867744369 cache-references # 1.095 M/sec
2882370051 cache-misses # 0.401 M/sec

300.431591843 seconds time elapsed

Without Patch

Performance counter stats for '/home/balbir/parallel_pagefault':

7192804.124144 task-clock-msecs # 23.937 CPUs
424691 context-switches # 0.000 M/sec
267 CPU-migrations # 0.000 M/sec
28498113 page-faults # 0.004 M/sec
5826093739340 cycles # 809.989 M/sec
408883496292 instructions # 0.070 IPC
7057079452 cache-references # 0.981 M/sec
3036086243 cache-misses # 0.422 M/sec

300.485365680 seconds time elapsed


--
Balbir


\
 
 \ /
  Last update: 2009-08-10 10:39    [W:0.047 / U:0.752 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site