Molcas Forum

roma28

Member

Registered: 2021-08-15

Posts: 6

Configuring GlobalArrays to use with MKL and OpenMPI

I'm trying to install parallel version of OpenMolcas on our HPC cluster.

I'm using Intel MKL and OpenMPI.

I use the following options which I took from Molcas manual

MPICC=mpicc MPIFC=mpifc ./configure --prefix=$GAROOT --enable-i8 --with-openib --with-blas8="-L$MKLROOT/lib/intel64 -lmkl_intel_ilp64 -lmkl_intel_thread -lmkl_core -liomp5 -lpthread -lm" --with-scalapack8="-L$MKLROOT/lib/intel64 -lmkl_scalapack_ilp64 -lmkl_intel_ilp64 -lmkl_intel_thread -lmkl_core -lmkl_blacs_opnempi_ilp64 -liomp5 -lpthread -lm"

to configure GA, then compile it and run tests using MPI.
They finishes without any error.

Then I compile OpenMolcas

cmake -DLINALG=MKL -DMKLROOT=/opt/intel/composerxe/mkl -DMPI=ON  -DGA=ON  -DMPIEXEC_EXECUTABLE=/opt/shared/openmpi/4.1.1-gcc.sles12sp5/bin/mpiexec -DMPI_C_COMPILER=mpicc -DMPI_Fortran_COMPILER=mpifort -DPYTHON_EXECUTABLE=/opt/shared/anaconda/anaconda3/bin/python ../OpenMolcas-master

and run verification.

The test crushes with memory error in GA and ARMCI (which is a part of GA).

So, could anyone who has an experience in compiling parallel version of OpenMolcas help me?