[CPMD-list] problem with Wannier

[Yong Zhang]

Hi Axel,

I tried the two ways, it seems WANNIER SERIAL is a little faster. Here is
some results:

BG/L with 32 cpu: 2.88sec (serial)/3.00sec
BG/L/ with 128 cpu: 2.20sec(Serial)/3.94sec
p690 with 16 cpu: 2.98sec(Serial)/3.02sec

I guess the difference between 32 cpu and 128 cpu comes from that fact that
other calculation (like force...) was speeded up with more cpus.

Best wishes,
Yong

On 12/11/06, Axel Kohlmeyer <akohlmey at cmm.chem.upenn.edu> wrote:
>
> On Sun, 10 Dec 2006, Yong Zhang wrote:
>
> YZ> Hi Axel,
> YZ>
> YZ> I did not find Alessandro's poster in the archives. But just followed
> you
> YZ> hints, the problem seems gone. Thanks a lot.
>
> there was no post. the bug was _spotted_ in october, i reproduced it
> and notified developers that it was a real bug and alessandro
> wrote me about the workaround and how he plans to fix it properly
> (so that it scales well on ibm-BG/L and cray-xt3).
>
> btw, you should check which of the two workarounds is faster...
>
> cheers,
>    axel.
>
> YZ>
> YZ> Best wishes,
> YZ> Yong
> YZ>
>
> --
> =======================================================================
> Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
>    Center for Molecular Modeling   --   University of Pennsylvania
> Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
> tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
> =======================================================================
> If you make something idiot-proof, the universe creates a better idiot.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://cpmd.org/pipermail/cpmd-list/attachments/20061211/5db29956/attachment.html