Awesome

E4S Container Launcher

E4S container launcher is a tool to simplify running MPI applications in E4S containers with host libraries. This project is built upon the MPICH ABI Compatibility Initiative and CEA's Wi4MPI.

Check out the documentation here !

Usage

Running an application with E4S Container Launcher can be as simple as prefixing the launcher command by e4s-cl launch.

e4s-cl init
e4s-cl launch mpirun -np 4 ./foo

The full details on execution and configuration can be found here.

How it works

Creating containers

e4s-cl wraps around standard MPI-capable launchers to execute a small amount of pre-processing before creating containers in which to run commands. This processing includes analyzing the library environment from the host, to bind select libraries to the container in order to make them accessible to the running process.

The concept of profile

One of the core components of the tool are profiles. They contain specific information about an execution environment that can be reproduced, such as container image and technology, or files and libraries to import in the containers.

They can be accessed and modified using the e4s-cl profile commands.

$ e4s-cl profile list
============= Profile Configurations (/home/user/.local/e4s_cl/user.json) ==
+----------+------------+-------------+------------+-----------+-------+--------+
| Selected |    Name    |   Backend   |   Image    | Libraries | Files | WI4MPI |
+==========+============+=============+============+===========+=======+========+
|          |  MPICH_3.4 | singularity |       None |    16     |   1   |   No   |
+----------+------------+-------------+------------+-----------+-------+--------+

Profiles are also used as intermediate outputs of the program. e4s-cl profile detect will analyze a given binary to list all its library and file dependencies. Those results are stored into a given profile to be reused.

Installation

e4s-cl requires python3.7+ to run. If you do not have access to a compatible version, the guided installation process will download a conda interpreter to run the project.

Guided standalone installation

You can install e4s-cl and a dedicated python interpreter using the provided Makefile. The installation location can be set using the INSTALLDIR variable:

git clone https://github.com/E4S-Project/e4s-cl
cd e4s-cl
make INSTALLDIR=<prefix> install

Once the installation succeeds, add <prefix>/bin to your PATH.

python package installation

If you would rather install e4s-cl using standard python packaging facilities, you may install using pip:

git clone https://github.com/E4S-Project/e4s-cl
cd e4s-cl
pip install .

The full details on installation targets can be found here.