Using pre-compiled binaries

A precompiled binary of DIALECT, which was compiled using github actions, is provided with the release on github.

The use of the pre-compiled binaries requires the installation of certain libraries on your computer.

GCC

As the pre-complied binaries are dynamically linked against the GCC library, it has to be installed on your system. An example for the installation of GCC on a Debian or Ubuntu machine is shown here.

sudo apt-get install gcc

Compiling the code

If you want to compile the program from source, you need to install additional programs and libraries.

Rust

Of course you need the Rust programming language itself to compile the code. The installation is explained in detail on the official site.

CMaKe

To compile certain Rust packages, you also need to install cmake.

sudo apt-get install cmake

Gfortran

In addition, the Gfortran compiler must be installed.

sudo apt-get install gfortran

Open-SSL

You need Open-SSL as some used Rust libraries depend on it. You can follow the guide shown here to install Open-SSL on linux.

To be able to compile and run the DIALECT program, it is necessary to download a linear algebra package as a requirement for the ndarray_linalg rust crate. You can choose between Intel MKL, OPENBLAS or Netlib by changing the features of ndarray_linalg in the Cargo.toml file according to the documentation (github page). In our experience, the intel MKL library yields the best performance.

Per default, the Cargo.toml files of DIALECT configure the ndarray_linalg create with "openblas-static", which should download and compile the OPENBLAS library and link it with the DIALECT package.

If you want to use the linear algebra packages that are already installed on your system, you need to change the feature of ndarray_linalg in the Cargo.toml files to "openblas-system", "intel-mkl-system" or "netlib-system".

The instructions for the installation of the libraries are provided in the following section.

Installation of LINALG packages

Intel MKL

If you want to compile DIALECT from source using the intel MKL as the linear algebra package, you will need to install it according to the following steps.

You can just download the MKL library from the Intel webpage and after the installation the environment variable $MKLROOT must be set with the setvars.sh script, which is located in the root installation directory of MKL.

source /path/to/MKL/setvars.sh

In the case of the older versions you have to execute the mklvars.sh script that is located in the installation directory of the MKL Library.

source /path/to/MKL/mklvars.sh intel64

OPENBLAS/Netlib

If you want to use OPENBLAS or netlib as the LINALG library, you need to install the BLAS and LAPACK libraries with the package manager of your operating systems.

sudo apt-get install libopenblas-dev
sudo apt-get install libblas-dev
sudo apt-get install libatlas-base-dev 
sudo apt-get install liblapack-dev

Mac users

In case of Mac users that have an apple silicon based processor, we recommend the compilation of DIALECT in conjunction with the accelerate framework.

A few changes to the Cargo.toml files are neccessary to compile DIALECT with accelerate:

ndarray-linalg = { version = "0.16", default-features = false }
blas-src = { version = "0.8", features = ["accelerate"] }

In addition, a build.rs file that contains the following line

fn main() {println!("cargo:rustc-link-lib=framework=Accelerate")}

must be created in the src folders of the DIALECT package and the dialect dynamics subpackage in order to use the accelerate framework.

dialect manual