[ncl-talk] Is it possible optimizing the performance of the NCL program via parallel processing?

[Francesco Trotta]

Dear Dennis
   I'm calling some fortran codes for the interpolation
   ok, I will try the task parallelism you send me

many thanks

Francesco?



________________________________
Da: Dennis Shea <shea at ucar.edu>
Inviato: mercoledì 15 febbraio 2017 05:18
A: Francesco Trotta
Cc: ncl-talk at ucar.edu
Oggetto: Re: [ncl-talk] Is it possible optimizing the performance of the NCL program via parallel processing?

You did not indicate what you are using inside the procedure(s).

[1] What are you using to regrid?
[2] Are you regridding each day?
[3] If the grid structure is consistent, use ESMF regridding to generate a single weight file. The weight file  (ESMF_regrid_with_weights<http://www.ncl.ucar.edu/Document/Functions/ESMF/ESMF_regrid_with_weights.shtml>) can be used to to regrid an entire array (time,lev,lat,lon). It is *VERY* fast. No 'do loop'

    f         = addfiles("...","r")
    temp  = f[:]->TEMP                       ; (time,lev,lat,lon)
    sal     = f[:]->SALINITY
    Oce_temp_regrid = ESMF_regrid_with_weights(temp,wgt_filename,False)  ; (time,lev,LAT,LON)
    Oce_sal_regrid    = ESMF_regrid_with_weights(sal    ,wgt_filename,False)

----
**NOTE: ***   If the above does not address the problem ...

The following was recently posted to the Applications page:

    http://www.ncl.ucar.edu/Applications/task_parallelism.shtml

Perhaps your script would have to be rearranged.
One approach: for each day, the suite of procedures could be  could be sent to a different processor.





On Tue, Feb 14, 2017 at 4:16 PM, Francesco Trotta <francesco.trotta4 at unibo.it<mailto:francesco.trotta4 at unibo.it>> wrote:

Dear Ncl
   I'm using NCL - version 6.3.0.
   I'm using a computer with multiple cores and
   I would like to optimizing the performance of my NCL code via parallel processing.

   In particular the part of my script which is very time consuming and I would like to improve is this:

   ;======================================
   iday=-1
   SpatialInterpTemp(iday,date_less1,file_indataOce_temp_less1,file_extrapdataOce_temp_less1, \
                                     file_regriddataOce_temp_less1)
   SpatialInterpSal(iday,date_less1,file_indataOce_sal_less1,file_extrapdataOce_sal_less1, \
                                file_regriddataOce_sal_less1)
   SpatialInterpSsh(iday,date_less1,file_indataOce_ssh_less1,file_extrapdataOce_ssh_less1, \
                                 file_regriddataOce_ssh_less1)
   SpatialInterpVelU(iday,date_less1,file_indataOce_velU_less1,file_extrapdataOce_velU_less1, \
                                    file_regriddataOce_velU_less1)
   SpatialInterpVelV(iday,date_less1,file_indataOce_velV_less1,file_extrapdataOce_velV_less1, \
                                   file_regriddataOce_velV_less1)

   do iday=0,ndays-1

      SpatialInterpTemp(iday,date(iday),file_indataOce_temp(iday),file_extrapdataOce_temp(iday), \
                                        file_regriddataOce_temp(iday))
      SpatialInterpSal(iday,date(iday),file_indataOce_sal(iday),file_extrapdataOce_sal(iday), \
                                    file_regriddataOce_sal(iday))
      SpatialInterpSsh(iday,date(iday),file_indataOce_ssh(iday),file_extrapdataOce_ssh(iday), \
                                    file_regriddataOce_ssh(iday))
      SpatialInterpVelU(iday,date(iday),file_indataOce_velU(iday),file_extrapdataOce_velU(iday), \
                                       file_regriddataOce_velU(iday))
      SpatialInterpVelV(iday,date(iday),file_indataOce_velV(iday),file_extrapdataOce_velV(iday), \
                                      file_regriddataOce_velV(iday))

   end do
   ;======================================

   Is there a way to do this optimization with ncl?

Thanks
Francesco



_______________________________________________
ncl-talk mailing list
ncl-talk at ucar.edu<mailto:ncl-talk at ucar.edu>
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/ncl-talk/attachments/20170220/fe524a03/attachment.html