I am trying to configure & compile WRF 4.0 on my new server (I just left an institution with a lot of resources and an HPC and I’m now sort of on my own here). I’ve searched the forum and reached out to my IS dept multiple times, but nothing I’ve found has worked yet. Basically, I have all of the executables created, ungrib, metgrid, and real.exe run perfectly, wrf.exe segmentation faults every time. I’m not sure where the error lies.
The server is comprised of two CPUs that are Intel Xeon-Gold wit 20 cores each.
I’ve gone through this Configure/Compile wrf walkthrough (http://www2.mmm.ucar.edu/wrf/OnLineTutorial/compilation_tutorial.php#STEP1) and all of the tests produced the correct output - although for test 2 under Library Compatibility Tests – each of the lines printed 40 times, which I assumed was for each core – perhaps this is in error and one of the libraries is installed incorrectly.
Then I go on to configure and compile wrf. It only creates the executables when I chose option 20 under configure (which is dmpar). I actually think I should be configuring it with 19/smpar, because as I understand it, compiling with serial – one core, SMpar – shared memory parallel (all the cores in a multi-core computer?) and DMpar (distributed memory) – all the cores in every computer of a cluster. But, the compilation does not succeed when I try that one or when I try 21. I have attached the compilation logs for when it works when I choose options 19 and 20.
I went forward with it configured/compiled using option 20 and the executables built and tried the test case study. The first three executables/steps (geogrid.exe, ungrib.exe, real.exe) work totally fine. And then, when I try and run the climate model itself - it segmentation faults every time. Both when I just try and run it just using ‘./wrf.exe’ and when I try and run it on multiple nodes (I’ve used both ‘mpirun –np 2 ./wrf.exe’ and ‘export OMP_NUM_THREADS = 2 / mpirun –np 2 ./wrf.exe’ – both of these might be the incorrect syntax, but regardless it *should* work using use ./wrf.exe). The error and out files are attached. Is there just something wrong in the namelist (last file attached) – this would be the ideal outcome, haha.
The server is comprised of two CPUs that are Intel Xeon-Gold wit 20 cores each.
I’ve gone through this Configure/Compile wrf walkthrough (http://www2.mmm.ucar.edu/wrf/OnLineTutorial/compilation_tutorial.php#STEP1) and all of the tests produced the correct output - although for test 2 under Library Compatibility Tests – each of the lines printed 40 times, which I assumed was for each core – perhaps this is in error and one of the libraries is installed incorrectly.
Then I go on to configure and compile wrf. It only creates the executables when I chose option 20 under configure (which is dmpar). I actually think I should be configuring it with 19/smpar, because as I understand it, compiling with serial – one core, SMpar – shared memory parallel (all the cores in a multi-core computer?) and DMpar (distributed memory) – all the cores in every computer of a cluster. But, the compilation does not succeed when I try that one or when I try 21. I have attached the compilation logs for when it works when I choose options 19 and 20.
I went forward with it configured/compiled using option 20 and the executables built and tried the test case study. The first three executables/steps (geogrid.exe, ungrib.exe, real.exe) work totally fine. And then, when I try and run the climate model itself - it segmentation faults every time. Both when I just try and run it just using ‘./wrf.exe’ and when I try and run it on multiple nodes (I’ve used both ‘mpirun –np 2 ./wrf.exe’ and ‘export OMP_NUM_THREADS = 2 / mpirun –np 2 ./wrf.exe’ – both of these might be the incorrect syntax, but regardless it *should* work using use ./wrf.exe). The error and out files are attached. Is there just something wrong in the namelist (last file attached) – this would be the ideal outcome, haha.