Hello everyone,
For past few weeks I have been trying to compile out my LIGGGHTS executable "lmp_auto" into a HPC cluster.
I have been using the following modules from software stack of the cluster:
1. module load compiler/openmpi/4.1.4 (For MPI based parallelization)
I am not using any version of GCC compiler as I assume that OpenMPI already has a parallel version of GCC wrapped inside itself.
2. module load compiler/cmake/3.22.5 (For configuring the downloaded VTK).
before start of the compilation I have edited following files
1. Make.sh --> inside it I have set the "nlimits"=3240 to generate all possible combinations of contact models
After running "make -j8 auto" for first time
2. "src/MAKE/Makefile.user" ---> inside it I have set the 'AUTOINSTALL_VTK ="ON"'
Again ran "make -j8 auto"
Model list are Generated and compilation begins after the download of vtk libraries.
However, somewhere in the end following error is being encountered
-------------------------------------------------------------------------------------------------------------------------
make[1]: *** [lammps.o] Error 9
make[1]: Leaving directory '/home/19me92r10/app/LIGGGHTS-PUBLIC/src/Obj_auto'
make: *** [auto] Error 2
-------------------------------------------------------------------------------------------------------------------------
Things I have already tried:-
1. I have even tried compilation by turning of the USE_VTK ="OFF" and still I get the same above error.
2. I have also found a very unusual behavior, when I tried to compile the LIGGGHTS with default number of contact models (i.e. nLimit=1200) compilation had no issue and everything was smooth with "lmp_auto" generated at the end of compilation. But if I want to compile all 3240 possible contact model combinations (i.e. nLimit = 3240) I am constantly getting the same above error message. Also if try to compile at any nlimit_value < 3240, compilation takes place smoothly but there is no corresponding updates in the contact_model_list inside "style_contact_model.whitelist", it appears to remain same as in case of nlimit=1200.
3. Another thing to note here is that even if I try to update the "style_contact_model.whitelist" with some additional required contact model combinations apart from 1200 limited contact models provided and recompile every thing agian smoothly. During execution of the liggghts script with updated pair-style/contact_model, I got the following WARNING:
**********************************************************************
The contact model you specified is not located in any whitelist.
Because of this the model will be run in an unoptimized version (increasing runtime by up to 20%).
In order to optimize this model you have the following options:
(i) Run the genAutoExamplesWhitelist.sh script in your LIGGGHTS(R) source folder to automatically parse the input script
(ii) Add the model combination by hand to your style_contact_model_user.whitelist that can be found in your LIGGGHTS(R) source folder
If you perform one of the steps above LIGGGHTS(R) needs to be recompiled to generate the optimized code.
(../contact_models.h:364)
************************************************************************
!!!!Thanks for Reading the issue .