Makefile.include: Difference between revisions

Latest revision as of 13:17, 18 December 2024

Writing a makefile.include file from scratch is not easy, so we suggest taking one of archetypical files that closely resembles your system as a starting point. It is necessary to customize it anyways to set appropriate paths etc. Optionally, you can enable additional features by setting precompiler flags or linking VASP to other libraries. For instance, we strongly recommend enabling HDF5 support.

Archetypical files

The templates contain information such as precompiler options, compiler options, and how to link libraries. Choose the template based on the compiler, parallelization etc. from the list below and mind the description:

Intel Composer suite and oneAPI Base + HPC toolkits for CPUs

makefile.include.oneapi
makefile.include.oneapi_omp
makefile.include.intel: Parallelized using MPI, combined with MKL.
makefile.include.intel_omp: Parallelized using MPI + OpenMP, combined with MKL.
makefile.include.intel_ompi_mkl_omp: Parallelized using OpenMPI + OpenMP, combined with MKL.
makefile.include.intel_serial: Not parallelized, strongly reduced feature-set, i.e., not suitable for production.

GNU compilers for CPUs

makefile.include.gnu: Parallelized using MPI, Free and Open-Source Software (FOSS) stack.
makefile.include.gnu_omp: Parallelized using MPI + OpenMP, FOSS stack.
makefile.include.gnu_ompi_mkl_omp: Parallelized using OpenMPI + OpenMP, combined with MKL.
makefile.include.gnu_ompi_aocl: Parallelized using OpenMPI, combined with AMD Optimizing CPU Libraries (AOCL).
makefile.include.gnu_ompi_aocl_omp: Parallelized using OpenMPI + OpenMP, combined with AOCL.

NVIDIA HPC-SDK for CPU and GPU

makefile.include.nvhpc: CPU version parallelized using MPI.
makefile.include.nvhpc_omp: CPU version parallelized using MPI + OpenMP.
makefile.include.nvhpc_ompi_mkl_omp: CPU version parallelized using OpenMPI + OpenMP, combined with MKL.
makefile.include.nvhpc_acc: Ported to GPUs using OpenACC, parallelized using MPI .
makefile.include.nvhpc_omp_acc: Ported to GPUs using OpenACC, parallelized using MPI + OpenMP.
makefile.include.nvhpc_ompi_mkl_omp_acc: Ported to GPUs using OpenACC, parallelized using OpenMPI + OpenMP, combined with MKL.

Others

makefile.include.nec_aurora

An advanced system administrator might benefit from a more detailed discussion about the precompiler options, compiler options, and how to link libraries.

Customize

Open the selected template of the archetypical files and add the required information as explained in the comments towards the end of the file. Then, add any optional feature as listed below. For more details see the list of precompiler options.

HDF5 support (strongly recommended, and mandatory for some features)

The HDF5 library is is needed for reading and writing HDF5 files such as vaspin.h5, vaspout.h5 and vaspwave.h5. The library is available for download on the HDF5 official website. To activate HDF5 support add the following in the VASP makefile.include

CPP_OPTIONS+= -DVASP_HDF5
HDF5_ROOT  ?= /path/to/your/hdf5/installation
LLIBS      += -L$(HDF5_ROOT)/lib -lhdf5_fortran
INCS       += -I$(HDF5_ROOT)/include

Available for VASP >= 6.2.0.

Warning: This is required to perform some features of VASP, e.g. electron-phonon coupling, and is required for py4vasp.

fftlib (recommended when using OpenMP)

When you plan to run VASP on multiple OpenMP threads and you are not using the FFTs from the Intel-MKL library, you should link against fftlib (included in the VASP distribution). To do so, uncomment the corresponding sections in the makefile.include.*_omp files. In makefile.include.gnu_omp, for instance, that would be:

# For the fftlib library (recommended)
CPP_OPTIONS+= -Dsysv
FCL        += fftlib.o
CXX_FFTLIB  = g++ -fopenmp -std=c++11 -DFFTLIB_THREADSAFE
INCS_FFTLIB = -I./include -I$(FFTW_ROOT)/include
LIBS       += fftlib
LLIBS      += -ldl

Wannier90 (optional)

To include the Wannier90 program, download the library from the source and compile libwannier.a.

Important: In case of Wannier90 3.x, you should compile a serial version by removing COMMS=mpi in the make.inc of Wannier90.

Then, execute make lib to build the Wannier90 library. To activate this feature set the following:

CPP_OPTIONS    += -DVASP2WANNIER90
WANNIER90_ROOT ?= /path/to/your/wannier90/installation
LLIBS          += -L$(WANNIER90_ROOT)/lib -lwannier

Mind: VASP version <= 6.1.x are compatible with Wannier90 <= 1.2. To interface VASP 6.1.x with Wannier90 2.x, set -DVASP2WANNIER90v2 instead. As of VASP 6.2.x only Wannier90 2.x and 3.x are supported.

Libxc (optional)

To include the Libxc library of exchange-correlation functionals, install the library from the source and install it. Note that to get correct results with meta-GGA functionals (see discussion at LTBOUNDLIBXC), it is necessary to use Libxc from version 5.2.0 onwards (or the master version from gitlab for the latest implemented functionals) and to compile it with the option --disable-fhc. For instance, with GNU Autotools the steps to compile Libxc are

autoreconf -i (necessary is the executable configure is not already present)
./configure --prefix=PATH/TO/LIBXC --disable-fhc
make
make install

Then, add the following in the VASP makefile.include

CPP_OPTIONS += -DUSELIBXC
LIBXC_ROOT  ?= /path/to/your/libxc/installation
LLIBS       += -L$(LIBXC_ROOT)/lib -lxcf03 -lxc
INCS        += -I$(LIBXC_ROOT)/include

Libbeef (optional)

To include the BEEF van der Waals functionals, install the library from the source on GitHub and add the following in the VASP makefile.include

CPP_OPTIONS  += -Dlibbeef
LIBBEEF_ROOT ?= /path/to/your/libbeef/installation
LLIBS        += -L$(LIBBEEF_ROOT)/lib -lbeef

DFT-D4 (optional)

To include the DFT-D4 van der Waals method, install the library from the source on GitHub and add the following in the VASP makefile.include

CPP_OPTIONS += -DDFTD4 
DFTD4_ROOT  ?= /path/to/your/dft4/installation
# version 3.6.0 built and installed with meson
LLIBS       += $(shell pkg-config --with-path=$(DFTD4_ROOT)/lib64/pkgconfig --libs dftd4)                                                                                                                                                  
INCS        += $(shell pkg-config --with-path=$(DFTD4_ROOT)/lib64/pkgconfig --cflags dftd4)    
# version 3.6.0 and loaded dftd4 module, i.e. PKG_CONFIG_PATH environment variable set correctly
#LLIBS       += $(shell pkg-config --libs dftd4)
#INCS        += $(shell pkg-config --cflags dftd4)
# for dft4-2.3.2 and older 
#LLIBS       += -L$(DFTD4_ROOT)/build -ldftd4
#INCS        += -I$(DFTD4_ROOT)/libdftd4.a.p

Mind: The compilation of the DFTD4 source code requires a Fortran compiler supporting Fortran 2018 features.

ELPA (optional)

To include the ELPA eigenvalue solvers, install the library from the source on GitLab and add the following in the VASP makefile.include

CPP_OPTIONS += -DELPA
ELPA_ROOT   ?= /path/to/your/elpa/installation
LLIBS       += -L$(ELPA_ROOT)/lib -lelpa
INCS        += -I$(ELPA_ROOT)/include/elpa-<version>/elpa
INCS        += -I$(ELPA_ROOT)/include/elpa-<version>/modules

Mind: In the above you need to replace <version> by the correct designation of your ELPA version.

libMBD (optional)

To include the library libMBD of many-body dispersion methods, install the library from the source on GitHub and add the following in the VASP makefile.include

CPP_OPTIONS += -DLIBMBD 
LIBMBD_ROOT ?= /path/to/your/libMBD/installation
LLIBS       += -L$(LIBMBD_ROOT)/build/src -lmbd
INCS        += -I$(LIBMBD_ROOT)/build/src/modules/

Mind: To run a calculation the path to libmbd.so has to be added (either in .bashrc or in the terminal): LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/your/libMBD/installation/build/src

Plugins (optional)

Create a new conda environment. Alternative environment creation packages should work, but we have not tested them.

   conda create -n vasp_plugin python=3.10

Enter the create vasp_plugin conda environment

   conda activate vasp_plugin

Navigate to the plugins directory within VASP source code,

   cd </path/to/vasp/source/code>/src/plugins

Install the VASP Python package through pip

   pip install .

Mind: Make sure to be within the conda environment when you compile VASP.

Add the following lines to your makefile.include

   CPP_OPTIONS+= -DPLUGINS
   LLIBS      += $(shell python3-config --ldflags --embed) -lstdc++
   CXX_FLAGS   = $(shell python3 -m pybind11 --includes) -std=c++11

Mind: When running VASP with the Python interface you will need to add the lib directory of your Python to LD_LIBRARY_PATH. You can do this by running export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$(python3-config --prefix)/lib

@@ Line 45: / Line 45: @@
 Then, add any optional feature as listed below. For more details see the [[Precompiler options|list of precompiler options]].
-=== HDF5 support (strongly recommended) ===
+=== HDF5 support (strongly recommended, and mandatory for some features) ===
 ----
 The HDF5 library is is needed for reading and writing HDF5 files such as {{FILE|vaspin.h5}}, {{FILE|vaspout.h5}} and {{FILE|vaspwave.h5}}.
@@ Line 57: / Line 57: @@
 Available for VASP >{{=}} 6.2.0.
+{{NB|warning|This is required to perform some features of VASP, e.g. [[:Category:Electron-phonon interactions|electron-phonon coupling]], and is required for py4vasp.}}
 === fftlib (recommended when using OpenMP) ===
@@ Line 83: / Line 84: @@
 === Libxc (optional) ===
 ----
-To include the Libxc library of exchange-correlation functionals, install the library from [https://www.tddft.org/programs/libxc/download/ the source] and install it. Note that to get correct results with meta-GGA functionals (see discussion at {{TAG|LTBOUNDLIBXC}}), it is necessary to use Libxc from version 5.2.0 onwards (or the master version from [https://gitlab.com/libxc/libxc gitlab] for the latest implemented functionals) and to compile it with the option <code>--disable-fhc</code>. For instance, with GNU Autotools the steps to compile Libxc are
+To include the Libxc library of exchange-correlation functionals, install the library from [https://libxc.gitlab.io/download/ the source] and install it. Note that to get correct results with meta-GGA functionals (see discussion at {{TAG|LTBOUNDLIBXC}}), it is necessary to use Libxc from version 5.2.0 onwards (or the master version from [https://gitlab.com/libxc/libxc gitlab] for the latest implemented functionals) and to compile it with the option <code>--disable-fhc</code>. For instance, with GNU Autotools the steps to compile Libxc are
   autoreconf -i (necessary is the executable configure is not already present)
   ./configure --prefix=PATH/TO/LIBXC --disable-fhc
@@ Line 158: / Line 159: @@
      CPP_OPTIONS+= -DPLUGINS
      LLIBS      += $(shell python3-config --ldflags --embed) -lstdc++
-     CXX_FLAGS   = -Wall -Wextra  $(shell python3 -m pybind11 --includes) -std=c++11
+     CXX_FLAGS   = $(shell python3 -m pybind11 --includes) -std=c++11
-{{NB|mind|When running VASP with the python interface you will need to add the lib directory of your python to LD_LIBRARY_PATH. You can do this somewhat easily with conda by running python3-config --ldflags, followed by export LD_LIBRARY_PATH{{=}}$LD_LIBRARY_PATH:<path-from-earlier-command>}}
+{{NB|mind|When running VASP with the Python interface you will need to add the lib directory of your Python to LD_LIBRARY_PATH. You can do this by running <code>export LD_LIBRARY_PATH{{=}}$LD_LIBRARY_PATH:$(python3-config --prefix)/lib</code>}}
 == Related articles ==