ecalj Install
1. Install fortran compilars and some tools
We need following tools to generate fortran binaries. ecalj made from source codes at ecalj/SRC/main/*.f90
and ecalj/SRC/subroutines/*.f90
.
- git (Tto download the ecalj. It is convenient to upgrade your code)
- Fortran compiler (we can choose gfortran, ifort, or nvfortran)
- Math library (blas, lapack, fft). We can usually use intel-mkl.
- MPI library (open mpi works for ubuntu24 )
- cmake, make, bash, gnuplot
We can use apt to install them when ubuntu. Similar in other systems, or your system already have.
I use following versions for thinkpad T14: ubuntu 24.04 |
---|
openmpi-bin/noble,now 4.1.6-7ubuntu2 amd64 |
openmpi-common/noble,noble,now 4.1.6-7ubuntu2 all |
cmake/noble,now 3.28.3-1build7 amd64 |
make/noble,now 4.3-4.1build2 amd64 |
gfortran/noble,now 4:13.2.0-7ubuntu1 amd64 |
intel-mkl/noble,now 2020.4.304-4 amd64 |
2. Install python
and required tools using mise
TIP
For the case the default version of Python is outdated (python >3.9 needed). We will prepare the latest Python in your ./local. The mise is a package management software. We can use anaconda instead. Or you can install tools at the following step 4.
- Add the following settings to
~/.bashrc
for the automatic installation and activation ofmise
:
export PATH="$HOME/.local/bin:$PATH"
type mise > /dev/null 2>&1 || curl https://mise.run | sh
eval "$(~/.local/bin/mise activate bash)"
- Update
~/.bashrc
to installmise
:
source ~/.bashrc
- Install
python
usingmise
:
mise use python@latest -g
- Install the required
python
libraries:
pip install numpy pandas seekpath spglib pymatgen mp-api scipy plotly
3. Install and InstallTest
For ohtaka and kugui in ISSP, skip here and see here
We recommend you to check ecalj/InstallAll at first. InstallAll writes files to your $HOME/bin. Add $HOME/bin to your path. (The install directory is BINDIR = os.path.join(HOME, 'bin')
. to BINDIR = os.path.join(HOME, 'bin2')
, for example, to change install directory).
Run the following command at ecalj/
FC=ifort ./InstallAll [Options]
(We can use gfortran or nvfortran instead of ifort)
It performs compile and link followed by the install test at ecalj/SRC/TestInstall/. (testecalj.py is the script for test) If succeeded, we see 'OK! All PASSED!' at the end of tests. The compile and install test may take 5~10 minutes (usually laptop is faster).
Following options are valid Options:
-np
[value]: default: 8 specify the number of MPI parallelization in test calculation--clean
default: none delete the cache files before compiling--gpu
default: none compile the GPU and GPU-MP version