Compiling a local version of FHI-aims¶

Having a local installation of FHI-aims can be useful to test keywords and develop new features. This guide walks you through the steps to compile your own installation of FHI-aims with the Anaconda Math Kernel Library (MKL).

Prerequisites¶

You need to:

have access to the FHI-aims gitlab server to download the source code
an installation of Anaconda.
a working gcc compiler (on Unix or WSL, try sudo apt-get install build-essential gdb)

Step-by-Step-Guide¶

Install CMake, the MKL library and an openMPI compiler:

conda install -c conda-forge cmake 
conda install -c anaconda mkl 
conda install -c conda-forge openmpi-mpifort 

Clone the FHI-aims repository:

git clone https://aims-git.rz-berlin.mpg.de/aims/FHIaims.git

Enter the directory where you cloned the repository. Then:

mkdir build
cd build
cp ../initial_cache.example.cmake initital_cache.conda.cmake

Edit the initital_cache.conda.cmake too look similar to this:

set(CMAKE_Fortran_COMPILER "mpifort" CACHE STRING "" FORCE)
set(CMAKE_Fortran_FLAGS "-O2 -ffree-line-length-none" CACHE STRING "" FORCE)
set(Fortran_MIN_FLAGS "-O0 -ffree-line-length-none" CACHE STRING "" FORCE)

set(CMAKE_C_COMPILER "gcc" CACHE STRING "" FORCE)

set(CMAKE_C_FLAGS "-O2 -funroll-loops -std=gnu99" CACHE STRING "" FORCE)
set(LIB_PATHS "~/miniconda3/pkgs/mkl-2020.2-256/lib" CACHE STRING "" FORCE)   # <-- this line is important
set(LIBS "mkl_intel_lp64 mkl_sequential mkl_core" CACHE STRING "" FORCE)

set(USE_MPI ON CACHE BOOL "" FORCE)
set(USE_SCALAPACK OFF CACHE BOOL "" FORCE)
set(USE_LIBXC ON CACHE BOOL "" FORCE)
set(USE_HDF5 OFF CACHE BOOL "" FORCE)
set(USE_RLSY ON CACHE BOOL "" FORCE)

Find the path to the MKL library files.

Depending on the type of Anaconda installation you have and the operating system, this might differ from user to user. The files should be somewhere in the /pkgs/mkl-versionnumber/lib directory. Inside this folder, there should be many shared objects .so, such as libmkl_intel_lp64.so.

Put this path in the initital_cache.conda.cmake in the set(LIB_PATHS "/here/put/the/path" CACHE STRING "" FORCE) line.
Build & compile

Enter the build directory, then:
```
cmake -C initial_cache.conda.cmake ..
```
This will run CMake to prepare all files for linking and checks if the compilers work correctly. If everything configured correctly, run:
```
make -j 4
```
Where -j 4 specifies the number of cores to use to speed up the compilation. This might take a while depending on your machine.
Finishing up

When everything is done, you should have the aims.VERSION.mpi.x binary in the build directory. Just test it by running
```
mpirun aims.*.mpi.x
```
To see that everything works. Afterwards, move it to the bin/ directory in the FHI-aims folder and add it to your path:
```
mv aims.*.mpi.x ../bin
```
And add theses line to your ~/.bashrc:
```
export AIMS_EXECUTABLE="path/to/aims.201103.mpi.x"
export ASE_AIMS_COMMAND="mpirun $AIMS_EXECUTABLE"
```