NWChem compilation (GCC, Intel MPI, internal BLAS)

nwchem-make.sh

#!/bin/bash

source /opt/intel/mkl/bin/mklvars.sh intel64
source /opt/intel/impi/2018.4.274/intel64/bin/mpivars.sh

export FC=gfortran
export _FC=mpif90
export CC=gcc
export _CC=mpicc

export LD=/usr/bin/ld

export NWCHEM_TOP=/home/redoak/NWChem/nwchem-6.8
export NWCHEM_TARGET=LINUX64
export NWCHEM_MODULES="all"
export LARGE_FILES=TRUE
export USE_64TO32=y
export ARMCI_NETWORK=MPI-PR

export USE_MPI=y
export USE_MPIF=y
export USE_MPIF4=y

export USE_INTERNALBLAS=y

make nwchem_config |& tee config.log
make -j8 |& tee make.log

To enable MKL:

• Remove:
  export USE_INTERNALBLAS=y

• Add:
  export BLASOPT="-L${MKLROOT}/lib/intel64 -Wl,--no-as-needed -lmkl_scalapack_ilp64 -lmkl_gf_ilp64 -lmkl_sequential -lmkl_core -lmkl_blacs_intelmpi_ilp64 -lpthread -lm -ldl"

• Between make nwchem_config... and make -j8 ..., add:
  make 64_to_32 |& tee lts.log

Gives ~11% performance boost on a simple benzene molecule CCSD(T) job using an Azure H8 instance.