Strange behavior of NEB-CI on Marconi-Cineca

Roberto Gaspari roby.g... at gmail.com
Tue Jul 26 13:02:34 UTC 2016


Dear Teo,

thank you for your advice. CP2K was actually compiled
 using intel compilers.
I will forward your suggestion to the CINECA staff,
and let you know when the problem is hopefully solved.

Best Regards,

Roberto

On Monday, July 25, 2016 at 5:48:44 PM UTC+2, Roberto Gaspari wrote:
>
> Dear CP2K users and developers,
>
> I am running NEB calculations on Marconi (as well as Fermi) machine at 
> Cineca center
> and I have a weird behavior:
> after the second NEB step is completed, i.e. when all the images have to 
> resync 
> to go on for the third step, the simulations stops. On Marconi sometimes 
> the 
> simulation simply goes on, without printing output anymore.
> The systems I am using for NEB (including the guesses along the path)
>  have all been optimized before NEB, so the geometries used should be ok.
>
> On Fermi I have had this error:
>
> *** ERROR in open_file (MODULE cp_files) *** 
>  ******************************************** 
>  
>  *** An error occurred opening the file <PROJ-r-0_localLog_p112.out> with 
> *** 
>  *** the unit number 1 (IOSTAT = 37)                                      
> *** 
>  
>  *** Program stopped at line number 439 of MODULE cp_files *** 
>  
>
> It  seems to me that the trouble is in the data access.
> Can somebody kindly clarify to me how the exchange of information
> between images is done? (i.e. is the exchange of information using 
> temporary files?)
> Did anybody step into this kind of behavior before?
> I am paralleling ask the help also of the CINECA support.
> Please tell me if more information on the calculation is needed.
>
> I thank you very much for your attention and help,
>
> Best Regards,
> Roberto 
>
> Istituto Italiano di Tecnologia
> CompuNet/Nanochemistry
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20160726/0a1d14ce/attachment.htm>


More information about the CP2K-user mailing list