Spack

This repo contains the configuration, build and run instructions for the spack installation on ARCHER2.

Using spack

Loading spack

The modules are installed as development modules on ARCHER2.

module use /work/y07/shared/archer2-lmod/apps/dev

The user spack module can be loaded with

module load spack

while the central cse module ( for installing centraly supported packages ) can be loaded with

module load spack-epcc

You will also need to source the spack environment with

source $SPACK_ROOT_EPCC/source.sh

Useful commands

• spack find: view installed packages. See spack find -h for options.
• spack install ${SPEC} : to install a specifick package name.
• spack compilers : show availabe compilers
• spack list : shows all packages available in the repository

Creating packages

A scaffold for a spack package can be created by typing

spack create --force --name hello_world https://github.com/lucaparisi91/hello_world/archive/refs/tags/v1.0.tar.gz

A few examples can be found in the custom_packages subdirectory.

Installation of spack

Fir set the environmet variable to the folder where you want to install spack

export `${SPACK_ROOT_EPCC}=... root directory of the spack installation ...`

Then install spack with

wget https://github.com/spack/spack/archive/refs/tags/v0.21.2.tar.gz
tar -zxvf v0.21.2.tar.gz

This will create a folder called spack-v0.21.2 in your installation directory. Set an environment variable which points to this folder .

export `${SPACK_ROOT}=${SPACK_ROOT_EPCC}/spack-v0.21.2`

You can now use source $SPACK_ROOT/share/spack/setup-env.sh to setup the spack environment.

Configuration of spack

These repo contains configurations for two installations of Spack

1. A centrally available installation. All packages installed trough spack with this config will be saved in a centralled saved directory on y07 . This installation is meant to be used by cse only to provide centrally installed packages to other users. The installation can be loaded using module load spack-epcc. The configuration files are present in the central_install subdirectory.

2. A user installation. All packages installated trough spack will be installed in a local directory for the user. By default this will be in /work/<project>/<project>/<user>/.spack . This installation points to the central installation and all packes installed in the central installation are available as well. The configuration files are present in the user_install subdirectory.

For installing the central installation, you will need to copy the contents of the files to ${SPACK_ROOT_EPCC} and the content of config subdirectory to ${SPACK_ROOT}/etc/spack.

cp -r  config/*.yaml ${SPACK_ROOT}/etc/spack
cp -r archer2repo ${SPACK_ROOT_EPCC}
cp source.sh ${SPACK_ROOT_EPCC}
cp -r environments ${SPACK_ROOT_EPCC}

To set the proper environment, source the source.sh file

source ${SPACK_ROOT_EPCC}/source.sh

These will make use of the ${SPACK_ROOT_EPCC} and ${SPACK_ROOT} environment variables.

In order to create the user installation repeat the whole installation and configuration process, but in a different ${SPACK_ROOT_EPCC} folder and use the user_install configuration files instead of central_install. You will also need to point the user installation to the central installation by modifying the upstreams.yaml file in the user_install/config subdirectory.

These configuration can be overriden by the user by calling spack with -C custom_config_folder option, where custom_config_folder is a directory containing .yaml configuration files for spack.

The archer2repo folder contains patches for broken packages on ARCHER2. Currently ( 09-04-24 ) a patch to cmake ( to findBlas.cmake , so that libsci is recognized ) and to quantum espresso are present.

Verified software

The current status of spack packages.

software	spack package	builds	runs	compiler
CASTEP	No
Code_Saturne	No
CP2K	Yes	Yes
Py-ChemShell/Tcl-ChemShell
FHI-aims	No
Gromacs	Yes	Yes		gnu
Lammps	Yes	Yes		gnu
namd	Yes	No ( manual fetching of source code required)
nektar++
nemo	No
nwchem	Yes
onetep	no
openfoam	Yes	Yes
quantum espresso	Yes	Yes ( patched )	Yes	gnu
vasp	yes
crystal	no
petsc
scotch	Yes
trilinos
parmetis
pytorch	Yes	no ( metadata generation failed)		gnu
tensorflow