Installation

Table of Contents

• Dependencies
• Installation steps
• Versions
• Custom CMake options

nda supports the usual installation procedure using CMake.

If you want to skip the installation step, you can go directly to Integration in C++ projects to see how you can integrate nda into your own C++ project by using CMake's FetchContent.

‍Note: To guarantee reproducibility in scientific calculations, we strongly recommend the use of a stable release version.

Dependencies

The dependencies of the C++ nda library are as follows:

• gcc version 12 or later OR clang version 15 or later OR IntelLLVM (icx) 2023.1.0 or later
• CMake version 3.20 or later (for installation or integration into an existing project via CMake)
• HDF5 library version 1.8.2 or later
• a working MPI implementation (openmpi and Intel MPI are tested)
• OpenMP
• a BLAS/LAPACK implementation (e.g. oneMKL or OpenBLAS)

For specific functionality, the following optional dependencies might be required:

• CUDA Toolkit for GPU support
• MAGMA for some specific matrix operations on the GPU

nda fetches and builds the following libraries automatically (unless the user explicitly tells CMake not to and to use local installations instead):

• TRIQS/cpp2py
• TRIQS/itertools
• TRIQS/mpi
• TRIQS/h5

For the Python interface, additional dependencies are required:

• Python version 3.6 or later
• numpy version 1.11.0 or later
• mako version 0.9.1 or later
• scipy (version 1.11.3 is tested but older/newer versions should work as well)

The Python interface is built with TRIQS/cpp2py. Please refer to the GitHub repository for further information.

Installation steps

1. Download the source code of the latest stable version by cloning the TRIQS/nda repository from GitHub:

   $ git clone https://github.com/TRIQS/nda nda.src

2. Create and move to a new directory where you will compile the code:

   $ mkdir nda.build && cd nda.build

3. In the build directory, call cmake including any additional custom CMake options (see below):

   $ cmake ../nda.src -DCMAKE_INSTALL_PREFIX=path_to_install_dir

   Note that it is required to specify CMAKE_INSTALL_PREFIX, otherwise CMake will stop with an error.

4. Compile the code, run the tests and install the application:

   $ make -j N
   $ make test
   $ make install

   Replace N with the number of cores you want to use to build the library.

Versions

To choose a particular version, go into the directory with the sources, and look at all available versions:

$ cd nda.src && git tag

Checkout the version of the code that you want:

$ git checkout 1.2.0

and follow steps 2 to 4 above to compile the code.

Custom CMake options

The compilation of nda can be configured by calling cmake with additional command line options

$ cmake ../nda.src -DCMAKE_INSTALL_PREFIX=path_to_install_dir -DOPTION1=value1 -DOPTION2=value2 ...

The following options are available:

Options	Syntax
Specify an installation path	-DCMAKE_INSTALL_PREFIX=path_to_install_dir
Build in Debugging Mode	-DCMAKE_BUILD_TYPE=Debug
Disable testing (not recommended)	-DBuild_Tests=OFF
Build the documentation	-DBuild_Documentation=ON
Disable Python support	-DPythonSupport=OFF
Build shared libraries	-DBUILD_SHARED_LIBS=ON
Build benchmarks	-DBuild_Benchs=ON
Test SSO memory optimizations	-DBuild_SSO_Tests=ON
Enable CUDA support	-DCudaSupport=ON
Enable MAGMA support	-DUse_Magma=ON