Main menu


SCAM netcdf pio crash

3 posts / 0 new
Last post
SCAM netcdf pio crash



It's a little soon I know but I am trying to build and run SCAM from a very recent CAM tag (cesm1_5_alpha02e)

And I am invoking the following modules on thorodin (CGD)

    module load compiler/pgi


Currently Loaded Modulefiles:
  1) mpi/pgi/mvapich2-1.8.1-qlc   2) tool/netcdf/4.3.2/pgi        3) compiler/pgi/default

SCAM compiles fine but crashes with the following

NetCDF: Index exceeds dimension bound  pio_support::pio_die:: myrank=           -1 : ERROR: pionfget_mod.F90:


Any ideas?



Richard Neale CESM AMWG Co-chair eval(unescape('%64%6f%63%75%6d%65%6e%74%2e%77%72%69%74%65%28%27%3c%61%20%68%72%65%66%3d%22%6d%61%69%6c%74%6f%3a%72%6e%65%61%6c%65%40%75%63%61%72%2e%65%64%75%22%20%63%6c%61%73%73%3d%22%62%62%2d%65%6d%61%69%6c%22%3e%72%6e%65%61%6c%65%40%75%63%61%72%2e%65%64%75%3c%2f%61%3e%27%29%3b')) Atmosphere Modeling and Predictability Section, AMP, NCAR Tel (303) 497-1380 Fax (303) 497-1324


Hi Rich:

Its hard to tell without seeing some more of the log file.  Its obviously having trouble reading one fo the netcdf files.  Usually if the model dates are beyond the IOP dates the model will die with this error.. Most of the time when this happens I try compiling with debug opitons and using pgdbg to stop at the error.  If its aborting you can usually break on mpi_abort or abort and go up the call stack once the model stops on abort.



Hi Rich:

Looked into this more and its a bug in the development branch.  The bug only shows up for initial files what have 3d dimensions that are ordered (time,lat,lev,lon).  For scam I think one of these datasets is a initial condition file.  The are other initial condition files (Jan I believe) that have their dimensions ordered (time,lev,lat,lon) and these do not trip the code that is halting your model.  Steve is aware of this and will commit a fix for the trunk.  As a work around in the mean time you can either find another initial condition file that has dimensions ordered (time,lev,lat,lon) or just use ncpdq (part of the nco utilites) and permute the dimensions of your current initial condition file.  This utility will only change the way the data is ordered on the netcdf file and should produce a bit for bit run with the original (that is if you could actually run the original).


ncpdq -a time,lev,lat,lon



Log in or register to post comments

Who's new

  • m.kliphuis@...
  • ddc3061993@...
  • hui.ding@...
  • zhouc@...
  • arianna.valmass...