CDO Segmentation fault
Added by Mao-Lin Shen almost 12 years ago
Hi,
I am running a climate model, COSMOS, on Hexagon (Cray XE6, http://docs.notur.no/uib), and have the segmentation fault when the data was processing by cdo.
The command is as follows.
cdo -s -f grb -g GR30s.nc selindexbox,2,121,1,101 -setgrid,r122x101 P31_mpiom.ext P31_mpiom.grb
Segmentation fault
The file can be processed correctly on another machine, DKRZ (http://www.dkrz.de/).
After some testing , I know the problem comes from “selindexbox,2,121,1,101 -setgrid,r122x101.”
Does anyone know how to fix it?
Following is the version of cdo on Hexagon.
Climate Data Operators version 1.5.4 (http://code.zmaw.de/projects/cdo)
Compiler: cc -g
version: /opt/cray/xt-asyncpe/5.08/bin/cc: INFO: Compiling with CRAYPE_COMPILE_TARGET=native.
with: PTHREADS OpenMP NC4 Z JASPER
Compiled: by aoltu on hexagon-5 (x86_64-unknown-linux-gnu) Apr 5 2012 14:21:01
CDI library version : 1.5.4 of Apr 5 2012 14:20:27
CGRIBEX library version : 1.5.1 of Aug 29 2011 20:30:27
netCDF library version : 4.1.3 of Nov 30 2011 13:50:42 $
HDF5 library version : 1.8.7
SERVICE library version : 1.3.0 of Apr 5 2012 14:20:23
EXTRA library version : 1.3.0 of Apr 5 2012 14:20:21
IEG library version : 1.3.0 of Apr 5 2012 14:20:22
FILE library version : 1.7.1 of Apr 5 2012 14:20:21
Thanks in advance.
Replies (6)
RE: CDO Segmentation fault - Added by Jaison-Thomas Ambadan almost 12 years ago
Hi,
Can you please post the output without the "-s"? It may give more info regarding the error. Also, can you please upload your input files?
Cheers,
J.
RE: CDO Segmentation fault - Added by Mao-Lin Shen almost 12 years ago
Hi,
Here is the result without "-s."
cdo -f grb -g GR30s.nc selindexbox,2,121,1,101 -setgrid,r122x101 P31_mpiom.ext P31_mpiom.grb
cdo selindexbox: Started child process "setgrid,r122x101 P31_mpiom.ext (pipe1.1)".
Segmentation fault
You can find the input data in the following links.
http://dl.dropbox.com/u/3311906/Web/Study/GR30s.nc
http://dl.dropbox.com/u/3311906/Web/Study/P31_mpiom.ext
Thanks in advance.
Best Regards,
Mao-Lin
RE: CDO Segmentation fault - Added by Jaison-Thomas Ambadan almost 12 years ago
Hi,
Sorry, I don't have any definite answer but [it's my wild guess] it is possible that the segmentation fault has to do with the "endian-ness" (Little/Big) of the data but I'm not sure (the data "P31_mpiom.ext" is Big-Endian) so may be try converting (say GR30s.nc to BigEndian ext) the files to same type with -b option and then do the selindexbox (in a second step 2). More info on "-b" option:
https://code.zmaw.de/embedded/cdo/1.5.4/cdo.html#x1-70001.2.1
By the way: using your original command sequence on DKRZ-Blizzard gives me a warning: "cgribexDefLevel : Changed zaxis type from generic to pressure" - I haven't seen such warnings in your report?
Cheers,
J.
RE: CDO Segmentation fault - Added by Jaison-Thomas Ambadan almost 12 years ago
Hi again,
just checking.. can you try again the command sequence with random numbers, i.e.,
cdo -f grb -g GR30s.nc selindexbox,2,121,1,101 -setgrid,r122x101 -random,r122x101 outfile.grb
If there is a segmentation fault (in CRAY) then there is a a bug, otherwise I suppose it has something to do with your data/file
Cheers,
J.
RE: CDO Segmentation fault - Added by Mao-Lin Shen almost 12 years ago
Hi,
Thanks for helping.
You are right. It is a bug.
There is a segmentation fault when we tested it with random data.
Does this imply that the bug is in cdo itself?
Or because the code was compiled incorrectly?
BTW, we do have the warning 'cgribexDefLevel'.
And I have to confess I pay no attention to that one. 
Best Regards,
Mao-Lin
RE: CDO Segmentation fault - Added by Jaison-Thomas Ambadan almost 12 years ago
Hi Mao-Lin,
I'm not sure if this is really a CDO bug or due to the compiler in CRAY. I suggest, you to report this issue (may be as a bug/support) in the "issues" section of this webpage (you may have to login first)
In the mean time, have a look at https://code.zmaw.de/projects/cdo/wiki/CRAY_Platform
In the above webpage there is a link to CRAY XE6 in Edinburgh, Scotland; you may contact those guys (just to check if they have similar issues)
Cheers,
J.