Navigation :

Specific architectures

This page needs updating!

Here the general instructions for building Octopus are supplemented by instructions for some specific large supercomputers. Included also is how to configure the optional ETSF_IO library, and how you can run the testsuite in parallel.

Generic Ubuntu 10.10

Install the following packages:

 * Required: gfortran, gcc, make, automake, m4, libtool, libgsl0-dev, libblas-dev, liblapack-dev, libfftw3-dev
 * Optional: mpi-default-dev, libgd2-xpm-dev, libsparskit-dev, libnetcdf-dev, libtrilinos-dev

[Unfortunately you should not use libblacs-mpi-dev, libscalapack-mpi-dev packages because they can give incorrect results. To use BLACS and ScaLAPACK (optional) you must build them yourself.]

First download libxc, extract with tar xzf, enter the resulting directory, and execute:

 ./configure --prefix=`pwd` CC=gcc FC=gfortran FCCPP="/lib/cpp -ansi -freestanding" FCFLAGS="-O3 -ffree-line-length-none" CFLAGS=-O3
 make install
 make check

For octopus, download and extract, and execute the following (inserting the path where you put libxc):

 ./configure --prefix=`pwd` CC=gcc CXX=g++ FC=gfortran FCCPP="/lib/cpp -ansi -freestanding" FCFLAGS="-O3 -ffree-line-length-none" CFLAGS=-O3 --with-libxc-prefix=''LIBXC_PATH''
 make install

If you are using MPI, replace the compilers by CC=/usr/bin/mpicc CXX=/usr/bin/mpicxx FC=/usr/bin/mpif90 and add –enable-mpi.

To build ETSF_IO (optional), libnetcdf-dev is required.

 ./configure --prefix=`pwd` CC=gcc FC=gfortran FCFLAGS="-O3 -ffree-line-length-none" CFLAGS="-O3" --with-netcdf-ldflags="-lnetcdff"
 make install

Add –with-etsf-io-prefix="$DIR/etsf_io-1.0.4" to the Octopus configure line, where $DIR/etsf_io-1.0.4 is where you installed ETSF_IO.

(Aug 2018)

United States

Sequoia/Vulcan (IBM Blue Gene/Q)

Stampede

Edison

Cori

Lonestar

Lawrencium

Europe

Curie

PFFT 1.0.5 and BigDFT 1.6.0

The Curie supercomputer, owned by GENCI and operated into the TGCC by CEA, is the first French Tier0 system open to scientists through the French participation into the PRACE research infrastructure.

Curie is offering 3 different fractions of x86-64 computing resources for addressing a wide range of scientific challenges and offering an aggregate peak performance of 2 PetaFlops.

General information: https://www-hpc.cea.fr/en/complexe/tgcc-curie.htm

Compilation with PFFT version 1.0.5 and LibISF taken from BigDFT 1.6.0 (previously compiled):

 #!/bin/bash
 
 module load c/intel
 module load fortran/intel
 module load bullxmpi
 module load mkl
 module load gsl/1.14
 
 WPREFIX=$WORKDIR/software
 PREFIX=$HOME/software
 export ISF_HOME="$WORKDIR/software/libisf"
 
 export CC=mpicc
 export CFLAGS="-xHost -O3 -m64 -ip -prec-div -static-intel -sox -I$ISF_HOME/include"
 export FC=mpif90
 export FCFLAGS="-xHost -O3 -m64 -ip -prec-div -static-intel -sox -I$PREFIX/fftw/3.3/include -I$ISF_HOME/include"
 
 export LIBS_BLAS="${MKL_LIBS}"
 export LIBS_FFT="-Wl,--start-group -L$PREFIX/pfft/1.0.5/lib -L$PREFIX/fftw/3.3/lib -lpfft -lfftw3 -lfftw3_mpi -Wl,--end-group"
 export LDFLAGS=$LDFLAGS" -L$ISF_HOME/lib -lisfsolver -lrt"
  
 ../configure --prefix=$WPREFIX/octopus/superciliosus_pfft \
   --disable-openmp --with-libxc-prefix=$WPREFIX/libxc/2.0.x \
   --disable-gdlib --with-gsl-prefix=/usr/local/gsl-1.14 \
   --enable-mpi --enable-newuoa \
   --with-pfft-prefix=$PREFIX/pfft/1.0.5

BigDFT was configured like this:

 #!/bin/bash 
 
 export FC=mpif90
 
 ../configure --with-ext-linalg="-lmkl_intel_lp64 -lmkl_sequential -lmkl_core -lm" \
   --with-ext-linalg-path="-L/usr/local/Intel_compilers/c/composer_xe_2011_sp1.7.256/mkl/lib/intel64" \
   --prefix=$WORKDIR/software/bigdft

PFFT 1.0.7 and BigDFT 1.7.0

 #!/bin/bash

 module load c/intel
 module load fortran/intel
 module load bullxmpi
 module load mkl
 module load gsl/1.14
 
 WPREFIX=$WORKDIR/poisson/software
 WLPREFIX=$WORKDIR/poisson/local
 export ISF_HOME="$WLPREFIX/libisf"
 export FFTW_HOME="$WLPREFIX/fftw-3.3.3"
 export PFFT_HOME="$WLPREFIX/pfft-1.0.7-alpha"

 export CC=mpicc
 export CFLAGS="-xHost -O3 -m64 -ip -prec-div -static-intel -sox -I$ISF_HOME/include"
 export FC=mpif90
 export FCFLAGS="-xHost -O3 -m64 -ip -prec-div -static-intel -sox -I$FFTW_HOME/include -I$ISF_HOME/include"
 
 export LIBS_BLAS="${MKL_LIBS}"
 export LIBS_FFT="-Wl,--start-group -L$PFFT_HOME/lib -L$FFTW_HOME/lib -lpfft -lfftw3 -lfftw3_mpi -Wl,--end-group"
 
 export LDFLAGS=" -L/usr/lib64 -L$ISF_HOME/lib -lPSolver-1 -lrt"
 export LD_LIBRARY_PATH=$ISF_HOME/lib:$LD_LIBRARY_PATH
 
 ../configure --prefix=$WPREFIX/octopus/superciliosus_pfft \
   --disable-openmp --with-libxc-prefix=$WPREFIX/libxc/11066 \
   --disable-gdlib --with-gsl-prefix=/usr/local/gsl-1.14 \
   --enable-mpi --enable-newuoa \
   --with-pfft-prefix=$PFFT_HOME

LibISF compilation:

 #!/bin/bash
 
 module load c/intel
 module load fortran/intel
 module load bullxmpi
 module load mkl
 module load gsl/1.14
 
 export FC=mpif90
 
 ../configure --with-ext-linalg="-lmkl_intel_lp64 -lmkl_sequential -lmkl_core -lm" \
    --with-ext-linalg-path="-L/usr/local/Intel_compilers/c/composer_xe_2011_sp1.7.256/mkl/lib/intel64" \
    --prefix=$WORKDIR/poisson/local/libisf \
    --disable-libbigdft --disable-binaries --enable-dynamic-libraries

FFTW 3.3.3 compilation script is in https://www-user.tu-chemnitz.de/~mpip/software/install_fftw-3.3.3_gcc.sh and it is changed as follow:

 < myprefix=$HOME/local
 --- 
 > module load c/intel
 > module load fortran/intel
 > module load bullxmpi
 > module load mkl
 > module load gsl/1.14
 > 
 > myprefix=$WORKDIR/poisson/local
 23c29,30
 < wget https://www.fftw.org/fftw-$FFTW_VERSION.tar.gz
 ---
 > - wget https://www.fftw.org/fftw-$FFTW_VERSION.tar.gz
 > cp ../fftw-$FFTW_VERSION.tar.gz .
 876c883
 < ./configure --prefix=$INSTALLDIR --enable-mpi --enable-openmp --enable-shared \
 ---
 > ./configure --prefix=$INSTALLDIR --enable-mpi --enable-openmp --enable-shared --disable-shared \

PFFTW 1.0.7 compilation script is in https://www-user.tu-chemnitz.de/~mpip/software/install_pfft-1.0.7-alpha_gcc.sh and it is changed as follow:

 < myprefix=$HOME/local
 ---
 > 
 > module load c/intel
 > module load fortran/intel
 > module load bullxmpi
 > module load mkl
 > module load gsl/1.14
 > 
 > myprefix=$WORKDIR/poisson/local
 24c31,32
 < wget https://www.tu-chemnitz.de/~mpip/software/pfft-$PFFT_VERSION.tar.gz
 ---
 > -wget https://www.tu-chemnitz.de/~mpip/software/pfft-$PFFT_VERSION.tar.gz
 > cp ../pfft-$PFFT_VERSION.tar.gz .

Fermi/Juqueen

IBM Blue Gene/Q, Cineca, Italy

General information: https://www.hpc.cineca.it/content/ibm-fermi-user-guide

The exactly the same script has been tested in the Juqueen, Forschungszentrum Jülich, Jülich Supercomputing Centre (JSC), Jülich, Germany. General information: https://www.fz-juelich.de/ias/jsc/EN/Expertise/Supercomputers/JUQUEEN/JUQUEEN_node.html

 #/bin/bash
 
 module load gsl
 module load lapack
 module load blacs
 module load scalapack 
 
 export PFFT_HOME="$HOME/local/pfft-threads-1.0.5-alpha"
 export FFTW3_HOME="$HOME/local/fftw-threads-3.3.2"
 export LD_LIBRARY_PATH=$PFFT_HOME/lib:$FFTW3_HOME/lib:$LD_LIBRARY_PATH 
 
 export FC_INTEGER_SIZE=8
 export CC_FORTRAN_INT=int
 export CC=mpicc
 export CFLAGS="-g -Wl,-relax -O3 -I$PFFT_HOME/include -I$FFTW3_HOME/include"
 export FC=mpif90
 export FCFLAGS=$CFLAGS" -qxlf90=autodealloc -qessl -qsmp=omp "
 export LIBS_BLAS="-lesslsmpbg -L/opt/ibmmath/essl/4.4/lib -lesslbg"
 export LIBS_LAPACK="-L$LAPACK_LIB -llapack"
 
 export LIBS_FFT="-L$FFTW3_HOME/lib/ -lfftw3_mpi -lfftw3 -lm"
 echo "GSL DIR = $GSL_HOME "
 echo "PFFT DIR = $PFFT_HOME "
 export LDFLAGS=$LDFLAGS" -R/opt/ibmcmp/lib/bg/bglib \
    -L/bgsys/drivers/V1R4M2_200_2010-100508P/ppc/gnu-linux/powerpc-bgp-linux/lib \
    -L/bgsys/drivers/V1R4M2_200_2010-100508P/ppc/gnu-linux/lib/gcc/powerpc-bgp-linux/4.1.2 \
    -L/opt/ibmcmp/xlf/bg/11.1/bglib \
    -L/opt/ibmcmp/xlmass/bg/4.4/bglib \
    -L/opt/ibmcmp/xlsmp/bg/1.7/bglib \
    -L/bgsys/drivers/V1R4M2_200_2010-100508P/ppc/runtime/SPI \
    -L/bgsys/drivers/V1R4M2_200_2010-100508P/ppc/comm/sys/lib \
    -L/bgsys/drivers/V1R4M2_200_2010-100508P/ppc/comm/default/lib \
    -L/usr/local/zlib/v1.2.3/lib \
    -L/bgsys/drivers/ppcfloor/runtime/SPI \
    -L$PFFT_HOME/lib -L$FFTW3_HOME/lib \
    -lpfft -lfftw3_mpi -lfftw3 -lm \
    -ldl -lxl -lxlopt -lrt -lpthread"
 
 SVN_VERS=$(svn info .. | grep Revision | awk '{print $2}')
 echo $SVN_VERS
 
 ../configure --prefix=$HOME/software/octopus_omp \
    --with-libxc-prefix=$HOME/software/libxc_new \
    --with-fft-lib="$FFTW3_HOME/lib/libfftw3.a  $FFTW3_HOME/lib/libfftw3_omp.a" \
    --host=powerpc32-unknown-linux-gnu \
    --build=powerpc64-unknown-linux-gnu \
    --disable-gdlib \
    --disable-f90-forall \
    --with-blas="-L/opt/ibmmath/lib64 -lesslbg" \
    --with-lapack="-L$LAPACK_LIB -llapack" \
    --with-gsl-prefix=$GSL_HOME \
    --enable-mpi --enable-openmp

To build FFTW3. Small changes are made to the script taken from M. Pippig web site https://www-user.tu-chemnitz.de/~mpip/software.php?lang=en It has a patch and is build with MPI and multithreaded (multithreaded may not work in Juqueen):

 #!/bin/sh -e 
 
 myprefix=$HOME/local
 FFTW_VERSION="3.3.2"
 INSTALLDIR="$myprefix/fftw-threads-${FFTW_VERSION}"
 TMP="tmp-mine-fftw-${FFTW_VERSION}" 
 
 # bash check if directory exists
 if [ -d $TMP ]; then
         echo "Directory $TMP already exists. Delete it? (y/n)" 
 	read answer 
 	if [ ${answer} = "y" ]; then 
 		rm -rf $TMP 
 	else
 		echo "Program aborted."
 		exit 1
 	fi
  fi 
 
 mkdir $TMP && cd $TMP 
 
 wget https://www.fftw.org/fftw-$FFTW_VERSION.tar.gz
 gzip -dc fftw-$FFTW_VERSION.tar.gz | tar xvf -
 cd fftw-$FFTW_VERSION
 
 # fix bug in fftw alltoall routine that causes deadlocks
 patch -b -p0 <<\EOF
 --- mpi/transpose-alltoall.c
 +++ mpi/transpose-alltoall.c
 @@ -223,8 +223,8 @@
 	  sbo[pe] = (int) (pe * (b * p->tblock) * vn);
 	  rbs[pe] = (int) (db * bt * vn);
 	  rbo[pe] = (int) (pe * (p->block * bt) * vn);
 -	  if (sbs[pe] != (b * p->tblock) * vn
 -	      || rbs[pe] != (p->block * bt) * vn)
 +	  if (dbt != p->tblock
 +	      || db != p->block)
 	       equal_blocks = 0;
      }
      pln->send_block_sizes = sbs;
 EOF
 
 ./configure --build=powerpc64-bgq-linux-gnu --host=powerpc64-bgq-linux \
 --prefix=$INSTALLDIR --enable-mpi --enable-openmp --enable-threads --disable-shared \
 CC=mpixlc_r F77=mpixlf90_r MPICC=mpixlc_r MPICXX=mpixlcxx_r MPIFC=mpixlf90_r MPILIBS=" " \
 CFLAGS='-O3 -g' \
 FCFLAGS='-O3 -g' \ 
 
 make -j 4
 make install

To build PFFT, also taken from https://www-user.tu-chemnitz.de/~mpip/software.php?lang=en

 #!/bin/sh -e 
 
 myprefix=$HOME/local
 PFFT_VERSION=1.0.5-alpha
 FFTW_VERSION=3.3.2
 INSTDIR=$myprefix/pfft-threads-$PFFT_VERSION
 FFTWDIR=$myprefix/fftw-threads-$FFTW_VERSION
 TMP="tmp-mine-pfft-$PFFT_VERSION" 
 
 # bash check if directory exists
 if [ -d $TMP ]; then
         echo "Directory $TMP already exists. Delete it? (y/n)"
 	read answer 
 	if [ ${answer} = "y" ]; then 
 		rm -rf $TMP 
 	else
 		echo "Program aborted."
 		exit 1
 	fi
 fi
 
 mkdir $TMP && cd $TMP
 
 wget https://www.tu-chemnitz.de/~mpip/software/pfft-$PFFT_VERSION.tar.gz
 gzip -dc pfft-$PFFT_VERSION.tar.gz | tar xvf -
 cd pfft-$PFFT_VERSION
 ./configure --build=powerpc64-bgq-linux-gnu --host=powerpc64-bgq-linux \
 --prefix=$INSTDIR --with-fftw3=$FFTWDIR \
 MPICC=mpif90 MPICXX=mpicxx MPIFC=mpif90 CC=mpicc FC=mpif90 \
 CFLAGS='-O3 -g -qmaxmem=-1' \
 FCFLAGS='-O3 -g -qmaxmem=-1' \
 --disable-shared
 
 make -j 4
 make install

Jugene

MareNostrum II

MareNostrum III

Corvo

Corvo is the computational cluster of the Nano-bio Spectroscopy Group. It consists of 1504 processor cores and 5284 GiB of RAM connected by an Infiniband network.

Script to build LIBFM library included in the Scafacos library (modified Scafacos svn version 1920, this is valid since version 9887 of Octopus):

 ./configure --enable-fcs-solvers=fmm,direct --enable-fcs-fmm-comm=armci \
   --enable-fcs-fmm-unrolled --enable-fcs-fmm-max-mpol=40 \
   CFLAGS=-O3 CXXFLAGS=-O3 FCFLAGS=-O3 --disable-doc CXX=mpic++ CC=mpicc \
   FC=mpif90 --prefix=$HOME/Software/scafacos \
   LDFLAGS=-L/opt/scalapack/1.8.0/lib -L/opt/blacs/1.1/lib \
   -L/opt/gotoblas2/1.08/lib -L/opt/netcdf/4.0.1/lib -L/opt/lapack/3.2/lib \
   -L/opt/etsf_io/1.0.2/lib --enable-fcs-int=int --enable-fcs-float=double \
   --enable-fcs-integer=integer --enable-fcs-real=real*8 
 
 make
 make install

The scripts to compile FFTW 3.3.2 and PFFT 1.0.5

 #!/bin/sh -e 
 
 myprefix=$HOME/local
 FFTW_VERSION="3.3.2"
 INSTALLDIR="$myprefix/fftw/${FFTW_VERSION}"
 TMP="tmp-fftw-${FFTW_VERSION}"
 
 # bash check if directory exists
 if [ -d $TMP ]; then
        echo "Directory $TMP already exists. Delete it? (y/n)"
        read answer
        if [ ${answer} = "y" ]; then
                rm -rf $TMP
        else
                echo "Program aborted."
                exit 1
        fi
 fi
 
 mkdir $TMP && cd $TMP
 
 wget https://www.fftw.org/fftw-$FFTW_VERSION.tar.gz
 gzip -dc fftw-$FFTW_VERSION.tar.gz | tar xvf -
 cd fftw-$FFTW_VERSION
 
 # fix bug in fftw alltoall routine that causes deadlocks
 patch -b -p0 <<\EOF
 --- mpi/transpose-alltoall.c
 +++ mpi/transpose-alltoall.c
 @@ -223,8 +223,8 @@
           sbo[pe] = (int) (pe * (b * p->tblock) * vn);
           rbs[pe] = (int) (db * bt * vn);
           rbo[pe] = (int) (pe * (p->block * bt) * vn);
 -         if (sbs[pe] != (b * p->tblock) * vn
 -             || rbs[pe] != (p->block * bt) * vn)
 +         if (dbt != p->tblock
 +             || db != p->block)
               equal_blocks = 0;
      }
      pln->send_block_sizes = sbs;
 EOF
 
 ./configure --prefix=$INSTALLDIR --enable-mpi --enable-openmp --enable-threads --disable-shared \
 CC="mpicc" F77="mpif90" MPICC="mpicc" \
 CFLAGS="-O3" FFLAGS="-O3" \
 MPILIBS=" "
 
 make -j 4
 make install

 #!/bin/sh -e
 
 myprefix=/home/local
 PFFT_VERSION=1.0.5-alpha
 FFTW_VERSION=3.3.2
 INSTDIR=$myprefix/pfft/$PFFT_VERSION
 FFTWDIR=$myprefix/fftw/$FFTW_VERSION
 TMP="tmp-pfft-$PFFT_VERSION"
 module unload fftw/3.2.2
 
 # bash check if directory exists
 if [ -d $TMP ]; then
        echo "Directory $TMP already exists. Delete it? (y/n)"
        read answer
        if [ ${answer} = "y" ]; then
                rm -rf $TMP
        else
                echo "Program aborted."
                exit 1
        fi
 fi
 
 mkdir $TMP && cd $TMP
 export CC=mpicc
 export FC=mpif90 
 
 wget https://www.tu-chemnitz.de/~mpip/software/pfft-$PFFT_VERSION.tar.gz
 gzip -dc pfft-$PFFT_VERSION.tar.gz | tar xvf -
 cd pfft-$PFFT_VERSION
 ./configure --prefix=$INSTDIR --with-fftw3=$FFTWDIR --disable-shared
 
 make install

This is the scrip to compile Octopus with PFFT and FMM support (through Scafacos library):

 module add gcc
 module add ifort
 module add gsl
 module add etsf_io
 module add lapack
 module add netcdf
 module add gotoblas2
 module add mpibull2 
 module add blacs
 module add scalapack
 
 export CC=mpicc
 export FC="mpif90"
 
 # It is important that "/opt/intel/Compiler//11.1/038/bin/intel64/" not to be in the PATH
 export PATH=/opt/mpi/mpibull2-1.3.9-14.s/bin:/opt/netcdf/4.0.1/bin:/opt/etsf_io/1.0.2/bin:/opt/gsl//1.13/bin:/opt/intel/composer_xe_2011_sp1.10.319//bin/intel64/:/opt/gcc//4.4/bin:/opt/slurm/bin:/usr/kerberos/bin:/opt/cuda//bin:/usr/local/bin:/bin:/usr/bin
 
 export LIBFM_HOME="$HOME/Software/scafacos"
 export PFFT_HOME="$HOME/local/pfft-1.0.5-alpha"
 export FFTW3_HOME="$HOME/local/fftw-3.3.2"
 export LIBS_PFFT="-L$PFFT_HOME/lib -L$FFTW3_HOME/lib -lpfft -lfftw3_mpi -lfftw3"
 export CFLAGS=" -I$PFFT_HOME/include -I$FFTW3_HOME/include -I$LIBFM_HOME/include"
 export LDFLAGS=$LDFLAGS" -L$LIBFM_HOME/lib -L$PFFT_HOME/lib -L$FFTW3_HOME/lib -lpfft -lfftw3_mpi -lfftw3 -lm "
 export FCFLAGS=$CFLAGS
 export CPPFLAGS=" -I$LIBFM_HOME/include "
 export LD_LIBRARY_PATH=$LIBFM_HOME/lib:$LD_LIBRARY_PATH

 ../configure --enable-mpi --with-libxc-prefix=$HOME/Software/libxc_9882 \
    --prefix=$HOME/Software/octopus \
    --with-blacs=/opt/blacs/1.1/lib/libblacs.a --with-scalapack \
    --with-libfm="-L$LIBFM_HOME/lib -lfcs4fortran -lfcs -lfcs_direct -lfcs_fmm -lfcs_near -lfcs_gridsort -lfcs_common" \
    --disable-openmp

LaPalma

LaPalma is a supercomputer of the IAC. It is a part of the old MareNostrum II computer. It was not possible to compile with XL compilers, so GNU 4.6.1 version is used.

Previously to Octopus, Libxc (2.2), FFTW (3.3.4), GSL (1.16) and OpenBLAS (0.2.13) were required to compile.

GSL

 export LD_LIBRARY_PATH=/usr/lib:/gpfs/apps/MPICH2/slurm-2.5.6/64/lib/:/gpfs/apps/GCC/4.9.2/lib64/
 export CC="/gpfs/apps/MPICH2/mx/default/64/bin/mpicc -cc=/gpfs/apps/GCC/4.9.2/bin/gcc"
 export FC="/gpfs/apps/MPICH2/mx/default/64/bin/mpif90 -fc=/gpfs/apps/GCC/4.9.2/bin/gfortran"
 
 ./configure --prefix=$HOME/local/gsl/gnu --enable-static --disable-shared

OpenBLAS

 make PREFIX=$HOME/local/openblas/4.6.1 BINARY=64 CC=/gpfs/apps/GCC/4.6.1/bin/gcc FC=/gpfs/apps/GCC/4.6.1/bin/gfortran

Libxc

 export PATH=/gpfs/apps/MPICH2/mx/default/64/bin:$PATH
 export LD_LIBRARY_PATH=/gpfs/apps/MPICH2/mx/default/64/lib:/gpfs/apps/MPICH2/slurm/64/lib
 
 export LD_LIBRARY_PATH=/gpfs/apps/GCC/4.6.1/lib64/:$HOME/local/openblas/lib:$LD_LIBRARY_PATH
 export CC="/gpfs/apps/MPICH2/mx/default/64/bin/mpicc -cc=/gpfs/apps/GCC/4.6.1/bin/gcc"
 export FC="/gpfs/apps/MPICH2/mx/default/64/bin/mpif90 -fc=/gpfs/apps/GCC/4.6.1/bin/gfortran"
 
 export CFLAGS="-m64 -O3"
 export FCFLAGS=$CFLAGS
 
 ./configure --prefix=$HOME/local/libxc/gnu/4.6.1/2.2

Octopus

 export PATH=/gpfs/apps/MPICH2/mx/default/64/bin:$PATH
 export LD_LIBRARY_PATH=/gpfs/apps/MPICH2/mx/default/64/lib:/gpfs/apps/MPICH2/slurm/64/lib
 
 export LD_LIBRARY_PATH=/gpfs/apps/GCC/4.6.1/lib64/:$HOME/local/openblas/4.6.1/lib:$LD_LIBRARY_PATH
 export CC="/gpfs/apps/MPICH2/mx/default/64/bin/mpicc -cc=/gpfs/apps/GCC/4.6.1/bin/gcc"
 export FC="/gpfs/apps/MPICH2/mx/default/64/bin/mpif90 -fc=/gpfs/apps/GCC/4.6.1/bin/gfortran"
 
 export CFLAGS="-m64 -O2 -I $PWD/external_libs/isf/wrappers "
 export FCFLAGS=$CFLAGS
 
 ../configure \
    --with-blas=$HOME/local/openblas/4.6.1/lib/libopenblas.so.0 \
    --with-lapack=/gpfs/apps/LAPACK/lib64/liblapack.a \
    --disable-gdlib \
    --with-gsl-prefix=$HOME/local/gsl/gnu \
    --with-fft-lib=$HOME/local/fftw/3.3.4/lib/libfftw3.a \
    --with-libxc-prefix=$HOME/local/libxc/gnu/4.6.1/2.2 \
    --prefix=/gpfs/projects/ehu31/software/octopus/gnu/4.6.1 \
    --enable-mpi --disable-f90-forall

XBES

Magerit (CeSViMa)

Magerit is a cluster located in “Centro de Supercomputación y Visualización de Madrid”. It consists on 245 nodes eServer BladeCenter PS702 with 2 Power7 processors each of 8 cores at 3'3 GHz (422,4 GFlops) and with 32 GB de RAM. Here the “optimal” configuration script which uses GNU compilers with ESSL IBM library.

 module purge
 module load gcc/4.4
 module load openmpi/1.6.3
 
 export LIBSDIR='/sw/openmpi/'
 
 export SOFTWARE_DIR="$HOME/SOFTWARE/exec"
 
 export FC=mpif90
 export CC=mpicc
 export CXX=mpicxx
 
 export STDFLAGS="-O3 -mcpu=power7 -mtune=power7"
 
 export FCFLAGS=$STDFLAGS
 export CFLAGS=$STDFLAGS
 export CXXLAGS=$STDFLAGS
 
 export C_TYPE="gnu64-4.4-essl"
 
 export branch=`awk '{print $2}' ../.git/HEAD`
 export branch=`basename $branch`
 
 INSTALLDIR="$SOFTWARE_DIR/octopus/git-$C_TYPE/$branch"
 
 #GSL  1.16 --compiled with gcc 4.4.6
 export GSL_HOME="$LIBSDIR/GSL/1.16/"
 
 #LAPACK gcc 4.4
 export LAPACK_HOME="$LIBSDIR/LAPACK/3.4.2-gnu64-4.4"
 
 #PARMETIS gcc 4.7.2
 export PARMETIS_HOME="$LIBSDIR/METIS/PARMETIS-4.0.3/"
 
 #METIS gcc 4.7.2
 export METIS_HOME="$LIBSDIR/METIS/METIS-5.1.0/"
 
 #FFTW3.3.4 gcc 4.4
 export FFTW_HOME="$LIBSDIR/FFTW/3.3.4-gnu64-4.4"
 export LD_LIBRARY_PATH=$FFTW_HOME/lib:$LD_LIBRARY_PATH
 
 #SCALAPACK gcc 4.4.6
 export SCALAPACK_HOME="$LIBSDIR/SCALAPACK/2.0.2-gnu64-4.4/"
 export LD_LIBRARY_PATH=$SCALAPACK_HOME/lib:$LD_LIBRARY_PATH
 
 #LIBXC
 export LIBXC_HOME="$SOFTWARE_DIR/libxc/3.0.0-$C_TYPE/"
 
 #LIBVDWXC
 export LIBVDWXC_HOME="$SOFTWARE_DIR/libvdwxc/0.2.0-$C_TYPE/"
 export CPPFLAGS="-I$LIBVDWXC_HOME/include -DHAVE_LIBVDWXC"
 #export LIBS=-lvdwxcfort
 export LDFLAGS="-L$LIBVDWXC_HOME/lib -lvdwxcfort -Wl,-rpath=$LIBVDWXC_HOME/lib"
 export LD_LIBRARY_PATH=$LIBVDWXC_HOME/lib:$LD_LIBRARY_PATH
 
 export LD_LIBRARY_PATH=/opt/ibmcmp/lib64/:/opt/ibmcmp/xlsmp/2.1/lib64/:/opt/ibmcmp/xlf/13.1/lib64/:$LD_LIBRARY_PATH
 
 export LIBS_BLAS=" -L$SCALAPACK_HOME/lib -L$LAPACK_HOME/lib -lscalapack -llapack -lessl -lblacssmp -L$FFTW_HOME/lib -lfftw3 -lfftw3_threads -lfftw3_mpi / opt/gnu/gcc/4.4.6/lib64/libgfortran.a -L/opt/ibmcmp/xlf/13.1/lib64/ -lxlf90_r -lxlfmath -lxl -L/opt/ibmcmp/xlsmp/2.1/lib64/ -lxlomp_ser -Wl,--rpath /opt/ibmcmp/xlf/13.1/lib64/ /opt/ibmcmp/xlf/13.1/lib64/libxl.a -R/opt/ibmcmp/lib64/"
 
 make distclean
 make clean
 ../configure --prefix=$INSTALLDIR \
     --disable-gdlib --disable-gsltest \
     --with-libxc-prefix=$LIBXC_HOME \
     --with-fftw-prefix=$FFTW_HOME \
     --with-gsl-prefix=$GSL_HOME \
     --with-metis-prefix=$METIS_HOME\
     --with-parmetis-prefix=$PARMETIS_HOME \
     --enable-mpi \
     --enable-openmp

MPCDF systems (draco, cobra, raven)

Draco, Cobra and Raven are the HPC machines of the Max-Planck Compute and Data Facility in Garching. On these, Octopus can be built with the following script:

 #! /bin/bash -l
 # compilation script for octopus
 # needs to be executed in a subfolder (e.g. _build) of the root directory
 #
 # if you want to skip the configure step, call as ./build_octopus.sh noconfig
 # if you want to skip the configure step and the dependency checking,
 #   call as ./build_octopus.sh noconfig nodep
 
 if [[ ! -f ../configure.ac ]]; then
   echo "Error! Please execute this script in a subfolder of the root directory."
   exit 1
 fi
 
 compiler=intel
 #compiler=gnu
 cuda=yes
 
 echo Using $compiler compiler.
 [[ $cuda == yes ]] && echo Building with CUDA support.
 
 module purge
 if [[ $compiler == intel ]]; then
   # newest intel version
   module load intel/19.1.2 impi/2019.8 mkl/2020.2
   export CC=mpiicc
   export FC=mpiifort
   export CXX=mpiicpc
   export CFLAGS="-O3 -xCORE-AVX512 -qopt-zmm-usage=high -fma -ip"
   export FCFLAGS="$CFLAGS"
   export CXXFLAGS="$CFLAGS"
   export MKL="-L${MKLROOT}/lib/intel64 -lmkl_scalapack_lp64 -lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core -lmkl_blacs_intelmpi_lp64 -liomp5 -lpthread -lm -ldl"
 elif [[ $compiler == gnu ]]; then
   module load gcc/9 impi/2019.8 mkl/2020.2
   export CC=mpicc
   export FC=mpif90
   export CXX=mpicxx
   export CFLAGS="-O3 -march=skylake-avx512 -g"
   export FCFLAGS="$CFLAGS"
   export CXXFLAGS="$CFLAGS"
   export MKL="-L${MKLROOT}/lib/intel64 -lmkl_scalapack_lp64 -lmkl_gf_lp64 -lmkl_gnu_thread -lmkl_core -lmkl_blacs_intelmpi_lp64 -lgomp -lpthread -lm -ldl"
 else
   echo "Compiler $compiler unknown."
   exit 1
 fi
 
 module load gsl hdf5-serial netcdf-serial libxc metis parmetis
 export LDFLAGS="-Xlinker -rpath=$MKL_HOME/lib/intel64:$GSL_HOME/lib:$NETCDF_HOME/lib:$ELPA_HOME/lib:$METIS_HOME/lib:$PARMETIS_HOME/lib"
 
 if [[ $cuda == yes ]]; then
   module load cuda/10.2
   CUDA_FLAGS="--enable-cuda --enable-nvtx --with-cuda-prefix=$CUDA_HOME"
   LDFLAGS="$LDFLAGS:$CUDA_HOME/lib64"
 else
   CUDA_FLAGS=""
 fi
 module list
 
 export INSTALLDIR=$PWD/installed
 
 if [[ "$1" != "noconfig" ]]; then
   pushd .. && autoreconf -i && popd
   ../configure $CUDA_FLAGS \
     FCFLAGS_FFTW="-I$MKLROOT/include/fftw" \
     FCCPP="cpp -ffreestanding" \
     --prefix=$INSTALLDIR \
     --enable-mpi --enable-openmp \
     --disable-gdlib \
     --with-gsl-prefix="$GSL_HOME" \
     --with-libxc-prefix="$LIBXC_HOME" \
     --with-blas="$MKL" \
     --with-lapack="$MKL" \
     --with-blacs="$MKL" \
     --with-scalapack="$MKL" \
     --with-netcdf-prefix="$NETCDF_HOME" \
     --with-metis-prefix="$METIS_HOME" \
     --with-parmetis-prefix="$PARMETIS_HOME" \
     || exit 1
 fi
 
 echo "\n\nBuilding octopus...\n"
 
 if [[ "$2" == "nodep" ]]; then
   make NODEP=1 -j20 && make NODEP=1 install || exit 1
 else
   make -j20 && make install || exit 1
 fi
 
 mkdir -p $INSTALLDIR/.build.doc/
 cp -f config.log $INSTALLDIR/.build.doc/
 cp -f $0 $INSTALLDIR/.build.doc/
 
 echo "... done"

Specific architectures

Generic Ubuntu 10.10

United States

Europe

PFFT 1.0.5 and BigDFT 1.6.0

PFFT 1.0.7 and BigDFT 1.7.0

prev next