MPAS-A Issue with RRTMG_DATA file

[rdtorn]

I am trying to run MPAS-A v6.2 on the Navy computer narwhal (see: Navy DSRC: Narwhal User Guide). This machine is a Cray EX with AMD cores that I am compiling with the Cray ftn and cc compilers using either the titan-cray or cray-nersc Makefile options (both behave the same). The model compiles fine, but once I run MPAS-A, I get the following error in log.atmosphere.0000.log:

ERROR:
ERROR: ------------------------------ FATAL CALLED ------------------------------
ERROR: module_ra_rrtmg_sw: error reading RRTMG_SW_DATA on unit 10
CRITICAL ERROR: MPAS core_physics abort

A quick google search suggests that this issue can arise from either not having the RRTMG files in your run directory, but that is not the case here:

rtorn@narwhal06:/p/work1/rtorn/MPAS-DART/ftest% ls -lL RRTMG_*
-rw-r--r-- 1 rtorn 5048G660 847552 Jul 24 2020 RRTMG_LW_DATA
-rw-r--r-- 1 rtorn 5048G660 1694976 Jul 24 2020 RRTMG_LW_DATA.DBL
-rw-r--r-- 1 rtorn 5048G660 680368 Jul 24 2020 RRTMG_SW_DATA
-rw-r--r-- 1 rtorn 5048G660 1360572 Jul 24 2020 RRTMG_SW_DATA.DBL

Moreover, I found another thread about the possibility of having the wrong endian, so I compiled with "-h byteswapio", but that has not made any difference (I get the same error).

Any suggestions for how to address this error?

 

[ldfowler58]

I would like to suggest doing the following test: If you compiled MPAS-Model with single precision, remove the files RRTMG_LW_DATA.DBL and RRTMG_SW_DATA.DBL; and vice versa, if you compiled MPAS-Model with double precisions, remove the files RRTMG_LW_DATA and RRTMG_SW_DATA from your run directory.

Unfortunately, I have not used the Cray ftn and cc compilers to help you test my suggestions above. Hope this will help,
Laura

 

[rdtorn]

Hi Laura,

Thank you for the suggestion. Removing these files did not make a difference, but I was able to successfully run MPAS by compiling the model with the ifort/Cray MPI wrapper (ftn) instead of using both the Cray compiler and MPI wrapper.

Cheers, Ryan

 

[MyAtmosphere]

Hi Laura,

Thank you for the suggestion. Removing these files did not make a difference, but I was able to successfully run MPAS by compiling the model with the ifort/Cray MPI wrapper (ftn) instead of using both the Cray compiler and MPI wrapper.

Cheers, Ryan

I'm using Cray compiler, could you specify what you mean by MPI wrapper ftn? I'm having the above issue too. Thanks!

 

[rdtorn]

ftn is the fortran compiler on a cray computer. The version that was on the Navy Narwhal machine does not seem to be compatible with the RRTM files, but when I used ifort instead, the model compiled and ran without issue.

 

[MyAtmosphere]

I have no issue compiling and running with other compilers. Only when I compile with Cray, does the run crash because of the above error.