|
From: | Chen,Jinzhen |
Subject: | Re: [EXT] Re: getfem installation |
Date: | Tue, 7 Dec 2021 17:31:48 +0000 |
User-agent: | Microsoft-MacOutlook/16.43.20110804 |
Dear Dr. Poulios, I was able to follow your steps and built all dependencies. In addition, I also built metis/5.1.0, as parmetis I built has no metis.h and metis library but they are required for getfem. I passed the getfem compiling with this configuration.
I could not include some libraries on LIBS somehow. However, I defined LD_LIBRARY_PATH environment variable which contains all dependency libraries. So no problem on building getfem. ./configure --prefix=/risapps/rhel7/getfem-mpi/5.4.1-new \ DFLAGS="-L/risapps/rhel7/gcc/9.3.0/lib64 -L/risapps/rhel7/openmpi/4.1.1/lib -L/risapps/rhel7mumps/5.4.1/lib -L/risapps/rhel7/parmetis/4.0.3/lib -L/risapps/rhel7/metis/5.1.0/lib -L/risapps/rhel7/scalapack/2.1.0/lib -L/risapps/rhel7/scotch/6.1.1/lib
-L/risapps/rhel7/qhull/2020.2/lib" \ CPPFLAGS="-I/risapps/rhel7/gcc/9.3.0/include -I/risapps/rhel7/openmpi/4.1.1/include -I/risapps/rhel7/mumps/5.4.1/include -I/risapps/rhel7/parmetis/4.0.3/include -I/risapps/rhel7/metis/5.1.0/include -I/risapps/rhel7/qhull/2020.2/include"
\ LIBS="-lmetis -lzmumps -ldmumps -lcmumps -lsmumps -lmumps_common -lpord -lgfortran" \ --with-blas="/risapps/rhel7/openblas/0.3.18/lib/libopenblas.a" \ --with-pic --with-optimization=-O3 --disable-metlab --enable-python --enable-paralevel=2 --enable-shared However, I got below error when I run a basic testing. I noticed that most built dependency libraries are static libraries. I.e. *.a files. How can I build shared libraries that getfem python needs at run time? Thank you very much ! [jchen24@tdragon4 testmpi]$ mpirun -n 4 ./parallel.py Traceback (most recent call last): File "/risapps/rhel7/getfem-mpi/5.4.1-new/lib/python3.9/site-packages/getfem/getfem.py", line 46, in <module> Traceback (most recent call last): File "/risapps/rhel7/getfem-mpi/5.4.1-new/lib/python3.9/site-packages/getfem/getfem.py", line 46, in <module> Traceback (most recent call last): File "/risapps/rhel7/getfem-mpi/5.4.1-new/lib/python3.9/site-packages/getfem/getfem.py", line 46, in <module> Traceback (most recent call last): File "/risapps/rhel7/getfem-mpi/5.4.1-new/lib/python3.9/site-packages/getfem/getfem.py", line 46, in <module> from ._getfem import * ImportError: libsmumps-5.3.so: cannot open shared object file: No such file or directory During handling of the above exception, another exception occurred: Traceback (most recent call last): from ._getfem import * File "/rsrch3/home/itops/jchen24/testmpi/./parallel.py", line 39, in <module> ImportError: libsmumps-5.3.so: cannot open shared object file: No such file or directory During handling of the above exception, another exception occurred: Traceback (most recent call last): File "/rsrch3/home/itops/jchen24/testmpi/./parallel.py", line 39, in <module> from ._getfem import * ImportError: libsmumps-5.3.so: cannot open shared object file: No such file or directory …. Actually, I was able to compile getfem using system level mumps and metis with some code changes before I received your previous email. But I got below errors in testing. I could be that the system level mumps was built using different
mpi as we are using or it missed other dependencies. This is just FYI.. as I forgot to mention it in my previous email.
[jchen24@tdragon4 testmpi]$ mpirun -n 4 ./parallel.py initializing ... numthread = 1 initializing ... numthread = 1 initializing ... numthread = 1 initializing ... numthread = 1 *** The MPI_Comm_rank() function was called before MPI_INIT was invoked. *** This is disallowed by the MPI standard. *** Your MPI job will now abort. [cdragon003:240842] Local abort before MPI_INIT completed completed successfully, but am not able to aggregate error messages, and not able to guarantee that all other processes were killed! *** The MPI_Comm_rank() function was called before MPI_INIT was invoked. *** This is disallowed by the MPI standard. *** Your MPI job will now abort. [cdragon003:240843] Local abort before MPI_INIT completed completed successfully, but am not able to aggregate error messages, and not able to guarantee that all other processes were killed! *** The MPI_Comm_rank() function was called before MPI_INIT was invoked. *** This is disallowed by the MPI standard. *** Your MPI job will now abort. [cdragon003:240844] Local abort before MPI_INIT completed completed successfully, but am not able to aggregate error messages, and not able to guarantee that all other processes were killed! *** The MPI_Comm_rank() function was called before MPI_INIT was invoked. *** This is disallowed by the MPI standard. *** Your MPI job will now abort. Regards Jinzhen (Jenny) Chen From: "Chen,Jinzhen" <JChen24@mdanderson.org> Dear Dr. Poulios, Thank you so much for your so detailed information. It is definitely very helpful. I will follow your steps and try again. Best Regards Jinzhen Chen From: Konstantinos Poulios <logari81@googlemail.com>
Dear Chen, Below you can see how I have compiled the parallel version of GetFEM with MUMPS on our cluster (Scientific linux 7). Depending on the specific versions of compilers and mpi, some parts of these steps need to be modified. The basic idea
that worked for me is to use the 3 environment variables, LDFLAGS, CPPFLAGS and LIBS to pass any external dependencies not detected by the configure script (header files, library directories, libraries respectively). I also used the --with-blas= configure
option to link to my compiled openblas (which also includes lapack). (The use of the PYTHONPATH variable is specific to a bug in our cluster's mpi4py module.) I hope this helps. You will probably need one more fix not included in my instructions, when you come to the final step of building the getfem python interface, but let's take that once you get there. Best regards Kostas module load python3/3.8.11 On Tue, Nov 30, 2021 at 8:51 PM Chen,Jinzhen <JChen24@mdanderson.org> wrote:
The information contained in this e-mail message may be privileged, confidential, and/or protected from disclosure. This e-mail message may contain protected health information (PHI); dissemination of PHI should comply with applicable federal and state laws. If you are not the intended recipient, or an authorized representative of the intended recipient, any further review, disclosure, use, dissemination, distribution, or copying of this message or any attachment (or the information contained therein) is strictly prohibited. If you think that you have received this e-mail message in error, please notify the sender by return e-mail and delete all references to it and its contents from your systems.
|
[Prev in Thread] | Current Thread | [Next in Thread] |