[PATCH v2] Improve the performance of --num-threads -d 31
Minfei Huang
mhuang at redhat.com
Wed Mar 2 02:25:43 PST 2016
On 03/01/16 at 04:16pm, Minfei Huang wrote:
> On 03/01/16 at 02:59pm, "Zhou, Wenjian/周文剑" wrote:
> > Hi Minfei,
> >
> > Could you help me test the patch in the attachment?
>
> Ok. I will verify it with this patch. Feedback will be given in next few
> days. Thanks for your effort.
>
> Thanks
> Minfei
>
> >
> > I have run it a thousand times and haven't got a failure.
> > Previously, I can always get a failure after running hundreds of times.
> >
> > But I'm not sure if it is the same problem you have met.
Hi.
I have verified this patch based on 1.5.9. makedumpfile works well both
with and without option --num-threads after applying this patch.
There are 128 active CPUs and 4T memory in kdump kernel.
with option -d 31
real 3m45.916s
with option --num-threads 1 -d 31
real 6m30.286s
with option --num-threads 2 -d 31
real 7m45.597s
with option --num-threads 8 -d 31
real 13m7.914s
with option --num-threads 32 -d 31
real 10m10.248s
original makedumpfile with option -d 31
real 4m27.617s
Seem more time will be taken by increasing thread number.
Thanks
Minfei
> >
> > --
> > Thanks
> > Zhou
> >
> > On 02/24/2016 10:24 AM, Minfei Huang wrote:
> > >
> > >>On Feb 24, 2016, at 10:20, Zhou, Wenjian/周文剑 <zhouwj-fnst at cn.fujitsu.com> wrote:
> > >>
> > >>Hi,
> > >>
> > >>On 02/24/2016 09:43 AM, Minfei Huang wrote:
> > >>>On 02/23/16 at 01:47pm, "Zhou, Wenjian/周文剑" wrote:
> > >>>>Hello, Minfei,
> > >>>>
> > >>>>Does it occur every time?
> > >>>>If not, I think I have known the reason.
> > >>>
> > >>>Hi, Wenjian.
> > >>>
> > >>>This patch is applied directly on version 1.5.9. And makedumpfile hangs
> > >>>if option num-thread is appended.
> > >>>
> > >>
> > >>I see.
> > >>I'm working on it.
> > >>
> > >>BTW, did you only test it with --num-threads 128?
> > >>How is it with --num-threads 32(or smaller value)?
> > >
> > >Yes. I have tested with num-thread 1, 2, 8 and 32. Except for —num-threads 1,
> > >all of the test fail.
> > >
> > >Thanks
> > >Minfei
More information about the kexec
mailing list