[CP2K-user] [CP2K:11141] van der Waals regtests fail on Intel KNL, and build glitches

Anton Kudelin archm... at gmail.com
Tue Jan 8 20:08:25 CET 2019


Why is the toolchain taking so much time? Did you set -j option? Usually it 
runs several hours on a multicore workstation.

On Tuesday, January 8, 2019 at 10:14:40 PM UTC+3, Ronald Cohen wrote:
>
> So I understand ( and have from the beginning) that the PSMP does not give 
> correct answers with the Intel compilers for vdW.
> I run Quantum Espresso and ABINIT and other codes using the same compilers 
> and have no problems. I am not sure
> I understand what is special about cp2k that makes it sensitive to the 
> compiler.
> Anyway, I am still doing the toolchain pure gnu build . It takes days and 
> is still building. I will let you know how that goes.
>
> BTW, the water benchmark (and many hundreds of regtests) work OK with my 
> intel PSMP build. It seems to be mainly a problem with vdW.
> But the first cycle works so I don’t understand that either.
>
> Thanks again!
>
> ---
> Ronald Cohen
> Extreme Materials Initiative
> Geophysical Laboratory
> Carnegie Institution
> 5251 Broad Branch Rd., N.W.
> Washington, D.C. 20015
> rco... at carnegiescience.edu <javascript:>
> office: 202-478-8937
> skype: ronaldcohen
> twitter: @recohen3
>
> On Jan 8, 2019, at 7:58 PM, Ronald Cohen <rec... at gmail.com <javascript:>> 
> wrote:
>
> The Intel PSMP OMP=1 was with 64 MPI processes, and MPI1 was with one 
> process.
>
> The times I quoted before were under:
>
>
> ------------------------------------------------------------------------------
>  -                                                                         
>     -
>  -                                T I M I N G                             
>      -
>  -                                                                         
>     -
>
>  -------------------------------------------------------------------------------
>  SUBROUTINE                       CALLS  ASD         SELF TIME       
>  TOTAL TIME
>                                 MAXIMUM       AVERAGE  MAXIMUM  AVERAGE 
>  MAXIMUM
>  CP2K                                 1  1.0    0.040    0.040    6.885   
>  6.885
>
> Thank you!
>
> Ron
>
>
>
> ---
> Ron Cohen
> rec... at gmail.com <javascript:>
> skypename: ronaldcohen
> twitter: @recohen3
>
>
>
>
> On Jan 8, 2019, at 7:41 PM, Alfio Lazzaro <alfi... at gmail.com 
> <javascript:>> wrote:
>
> Sorry, I'm confused... In your previous table, I see that you have
>
> Intel PSMP OMP=1 gives
> 85.1906685165 time 434.5
>
> In your last table, MPI1OMP1 takes 
>
> Used time                  =                1.739
>
> Could you compare these two cases? I don't see any reason for such a large 
> difference besides a numerical problem...
>
> OK, I see that CP2K people will soon have a test based on 18.0.5 and PSMP 
> (see 
> https://github.com/cp2k/cp2k/pull/158/commits/11ad8588ee2cf6bbdbb9dcd78e62a30da96d8c1c 
> ). Let's see if there is something wrong for the compiler, at least on the 
> Xeon. Sorry to repeat myself, the only strategy I see is to have a common 
> baseline... 
>
> -- 
> You received this message because you are subscribed to a topic in the 
> Google Groups "cp2k" group.
> To unsubscribe from this topic, visit 
> https://groups.google.com/d/topic/cp2k/gzmRqKNt62U/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to 
> cp2k+... at googlegroups.com <javascript:>.
> To post to this group, send email to cp... at googlegroups.com <javascript:>.
> Visit this group at https://groups.google.com/group/cp2k.
> For more options, visit https://groups.google.com/d/optout.
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190108/f7ba9e61/attachment-0001.html>


More information about the CP2K-user mailing list