Error compiling wrf 4.1. - module_ra_goddard.f90(127): error #5082

[Gippox]

Hi I have the following problem in compiling WRF 4.1 with Intel Fortran compiler (v.15.0)

module_ra_goddard.f90(127): error #5082: Syntax error, found INTEGER_CONSTANT '64' when expecting one of: %FILL <IDENTIFIER>
INTEGER, PARAMETER, PRIVATE:: 64 = 16
--------------------------------^
compilation aborted for module_ra_goddard.f90 (code 1)
make[3]: [../configure.wrf:355: module_ra_goddard.o] Error 1 (ignorato)

module_mp_gsfcgce_4ice_nuwrf.f90(33): error #5082: Syntax error, found INTEGER_CONSTANT '64' when expecting one of: %FILL <IDENTIFIER>
INTEGER, PARAMETER, PRIVATE:: 64 = 16
---------------------------------^
compilation aborted for module_mp_gsfcgce_4ice_nuwrf.f90 (code 1)
make[3]: [../configure.wrf:355: module_mp_gsfcgce_4ice_nuwrf.o] Error 1 (ignorato)

a plethora of errors follows

if you can give me some indication

Thanks.

 

[kwerner]

HI,
Can you do the following steps:
1) clean -a
2) request that the code only be compiled with 1 processor (csh e.g.)
setenv J "-j 1"
3) reconfigure and then recompile
4) send (attach) your configure.wrf file and your compile log

Thanks,
Kelly

 

[Gippox]

The problem happens the same (see the attachments)

If, however, in the files module_ra_goddard.F and module_mp_gsfcgce_4ice_nuwrf.f i uncomment this line, WRF 4.1 compile correctly :

! INTEGER, PARAMETER, PRIVATE:: CHUNK = 16 !size of vector (16 is optimized number for Haswell/Broadwell CPUs.

 

[kwerner]

Hi,
The problem is likely due to the older version of Intel you are using (V15). We tested with some of the newer versions and it seems to be okay. As long as you're not planning to use the Goddard radiation scheme, that shouldn't be a problem to comment it out to get it to build.

So that we can take a closer look at this problem, in an attempt to fix the code, I'm attaching a small script. Can you run this script using the same compiler you use to compile WRF and let me know what prints out?

Thanks,
Kelly

 

[Gippox]

I did a further test. The problem occurs with option 20 (dmpar) INTEL (ifort / icc): Xeon (SNB with AVX mods). In previous versions of WRF there were no compilation problems with this option.
If I compile with option 15. (dmpar) INTEL (ifort / icc), the executables are built (see the attachments).

 

[Gippox]

Hi,
The problem is likely due to the older version of Intel you are using (V15). We tested with some of the newer versions and it seems to be okay. As long as you're not planning to use the Goddard radiation scheme, that shouldn't be a problem to comment it out to get it to build.

So that we can take a closer look at this problem, in an attempt to fix the code, I'm attaching a small script. Can you run this script using the same compiler you use to compile WRF and let me know what prints out?

Thanks,
Kelly

Compiling with ifort and executing the script the result is :

andrea@localhost-localdomain-> /home/andrea/Scrivania/WRF/4.1/WRF-4.1/a.out
chunk = 0.0000000E+00


Note:
Compiling with gfortran and executing the script the result is :
andrea@localhost-localdomain-> /home/andrea/Scrivania/WRF/4.1/WRF-4.1/a.out
chunk = 16574421.0


Thanks

 

[RCarpenter]

I had the same problem. I changed "CHUNK =" to "chunk =" in phys/module_ra_goddard.F and phys/module_mp_gsfcgce_4ice_nuwrf.F to get it to compile. Perhaps it is a preprocessor issue ("-DCHUNK=64")? I have ifort v16.0.

 

[davegill]

Richard,
Would you please verify that this fix works for you?
https://github.com/wrf-model/WRF/pull/907

We would like to include your fix in the v4.1.1 release. Also, if you would include your affiliation, we like to send out props to folks who find and fix bugs during our release cycles.
Dave

 

[kwerner]

@RCarpenter,

We are in the process of putting together our release notes for the upcoming V4.2 release and we would like to include you in the acknowledgments for the work you provided. Do you mind sharing your affiliation so that I can note that?