Installation

System requirements

In principle, BDF supports all Unix-like platforms. However, we strongly propose you to use BDF at Linux platform together Intel Fortran and C compiler because BDF package is developed and thoroughly tested on such platforms. System requirements of BDF is summarized as following lines:

 OS: Linux, Mac, FreeBSD, Unix
 Compiler: Fortran 90 compiler such as Intel, gfortran, g95
 Support softwares: Python, make
 Parallel environments: MPI

Configure and compile BDF package

Several steps are needed to install BDF package.

1. Set fortran and C compiler.

• If you use bash shell, you can set fortran and c compiler:

$export FC=ifort
$export CC=gcc
$export CXX=g++

• Tips: If you ignored this step, the configure script will pick up the gfortran and gcc compiler as default.

2. Set blas and lapack library. The external Blas and Lapack library is needed for compling BDF. In addition, C language Blas and Lapack are required. It is easer to use Intel MKL. You only need set "MATHLIB" and "MATHINCLUDE" as

$export MATHLIB="-lmkl_intel_ilp64 -lmkl_sequential -lmkl_core -lpthread"
$export MATHINCLUDE="-I/opt/intel/mkl/include"

However, you may choose to use Blas and Lapack from netlib. You may download lapack library from http://www.netlib.org/lapack/lapack-3.6.0.tgz. Do not forget also compile "lapacke" and "cblas", which is a C interface for lapack and blas library.

Then, you need set "MATHLIB" and "MATHINCLUDE"

$export MATHLIB="-L/Users/aaa/mathlib/lapack -llapack64 -lblas64"
$export MATHINCLUDE="-I/Users/aaa/lapack-3.6.0/lapacke/include -I/Users/aaa/lapack-3.6.0/cblas/include"

And, due to netlib only offer 32-bit cblas and lapacke officially, you need either modify codes from netlib, or specific FORTINT=int:

$export MATHINCLUDE="-I/Users/aaa/lapack-3.6.0/lapacke/include -I/Users/aaa/lapack-3.6.0/cblas/include -DFORTINT='int'"

the later option can make the package compile, but may meet unexpected error when the job involves cblas and lapacke. 3.(optional)Set processor specific compiling flag. If you need to speed up the program, you can turn on specific compiling flag in config/simd.mh for example, you can turn on core-avx2 with intel compiler on intel CPU with Haswell microarchitecture or later on.

SIMDFLAGS := -march=core-avx2

CFLAGS :=$(SIMDFLAGS) $(CFLAGS)
CXXFLAGS :=$(SIMDFLAGS) $(CXXFLAGS)
F77FLAGS :=$(SIMDFLAGS) $(F77FLAGS)
F90FLAGS :=$(SIMDFLAGS) $(F90FLAGS)

4. Configure BDF package.

After setting compiler and mathematic library. You can configure BDF by command configure in BDF root directory.

$./configure

Configure support several parameters.

 --enable-debug=yes[no]     Compile BDF in debug mode.
 --enable-i8=yes[no]            Compile BDF with 64 bit integrals.
 --enable-openmp=yes[no]  Compile BDF with OpenMP supporting.
 --enable-mpi=yes[no]         Compile BDF with MPI supporting.
 --enable-parallel=yes[no]   Compiler BDF with MPI and OpenMP support.
 --enable-mkl=yes[no]   Compiler BDF withIntel MKL support.

5. Compile BDF package.

 $make

Compile parallel bdf

Some of BDF modules support parallel calculation. You can compiler BDF as parallel mode. At present, we support MPI and OpenMP.

1. MPI and OpenMP hybrid.

 $./configure  --enable-parallel=yes

2. Only MPI.

 $./configure  --enable-mpi=yes

3. Only OpenMP.

 $./configure  --enable-openmp=yes

Examples of some typical installations

1. A 64bit machine with intel compiler and MKL library.

  export FC=ifort
  export CC=gcc
  export CXX=g++
  export MATHLIB="-lmkl_intel_ilp64 -lmkl_sequential -lmkl_core -lpthread"
  export MATHINLCUDE="-I/opt/intel/mkl/include"
  ./configure --enable-i8=yes --enable-openmp=yes --enable-mkl=yes
  make

2. A Mac laptop with gfortran, gcc and user compiled lapack and blas library

  $export FC=gfortran
  $export CC=gcc
  $export CXX=g++
  $export MATHILIB="-L/Users/mike/mathlib -lblas -llapack"
  $export MATHINCLUDE="-I/Users/mike/lapack-3.6.0/lapacke/include -I/Users/mike/lapack-3.6.0/cblas/include"
  $./configure --enable-i8=yes --enable-openmp=yes
  $make

3. A Linux cluster with fort, icc, openmpi and intel MKL library.

• Check MPI flags by using command

  $mpif90 --shownme

• You may get such output.

  ifort -I/opt/openmpi/include -I/opt/openmpi/lib -L/opt/openmpi/lib -lmpi_f90 -lmpi_f77 -lmpi -ldl -lm -Wl,--export-dynamic -lrt -lnsl -lutil

• You can compile bdf package as

  $export FC=ifort
  $export CC=icc
  $export CXX=icpc
  $export MATHLIB="-lmkl_intel_ilp64 -lmkl_sequential -lmkl_core -lpthread"
  $export MATHINCLUDE="I/opt/intel/mkl/include"
  $export MPILIB="-I/opt/openmpi/include -I/opt/openmpi/lib -L/opt/openmpi/lib -lmpi_f90 -lmpi_f77 -lmpi -ldl -lm -Wl,--export-dynamic -lrt -lnsl -lutil"
  $./configure --enable-i8=yes --enable-parallel=yes --enable-mkl=yes
  $make

• If you only want to evoke MPI, just use

  $./configure --enable-i8=yes --enable-mpi=yes --enable-mkl=yes

Set system variables of BDF package

Using dynamic library

For reducing program size and easy developing, dynamic library can be used.

Branch

Just clone/pull/checkout the branch dynamic-dev. The installation is just as static library version.

List of dynamic libraries

The extra libraries libcint,libgen1int,libxcfun are dynamic. The list of other dynamic libraries can be finded in variable "DYNUTL" in database/xianest.mh

Modifying libraries

If the modified code is code of static library, please rebuild the library and the programs use this library. Just type

make

If the modified code is code of program, just rebuild the program. Just type

make

After modifying dynamic libraries, just rebuild the library. You need to type

make ISLIB=dynamic

If function calls are changed in both dynamic library and program, please rebuild the library first, then program.