What is Spack?
Spack is a
multi-platform package manager that builds and installs multiple versions and configurations of software. It works on Linux, macOS, and many supercomputers. Spack is non-destructive: installing a new version of a package does not break existing installations, so many configurations of the same package can coexist.
It also has amazing documentation!
You can install Spack locally, without root permissions and in an environment that will not affect the rest of your system. Meanwhile, it allows to be chained to installed variants provided by your system administrators.
It builds preCICE and all its dependencies from source and allows you to load the installation into your running environment when needed. It also allows you to build multiple versions and variants of preCICE, which then coexist on your system. Spack also generates module files which can be useful on clusters.
A few hints to get you started:
spack info precicedisplays the package info, versions, variants, and dependencies.
spack spec precicedisplays all the dependencies that will be built.
- Want to use a system-installed compiler (e.g. Intel)? Try
spack compiler find.
- Do you need a specific compiler version? You can build
llvmfrom source and use them to compile your software.
- Want to build something special? If you ran
spack install precice@firstname.lastname@example.org ^email@example.com build_type=RelWithDebInfo, this would install the
developversion of preCICE, with the compiler GCC 7.3.0, OpenMPI version 3.1.2 and in
- You can even edit the package/recipe using
spack edit precice.
- For more advanced usage, you can create your own package repository and use it to build your software.
Where are the packages installed?
Just inside the
Deleting this directory will remove everything.
Setting up Spack
Get and configure Spack on your laptop / supercomputer (no sudo required!):
git clone -b develop https://github.com/spack/spack.git
source spack/share/spack/setup-env.sh # Maybe put this in your ~/.bashrc
To install the latest release of preCICE run:
spack install precice ^firstname.lastname@example.org
That’s it! Spack will now automatically get and build preCICE and all of its dependencies from source. This may take a while, but you don’t need to do anything else.
spack install precice
(but keep an eye on for Boost-related compilation errors and
let us know in that case).” %}
You just installed the latest release of precice with the default configuration under
To see all the installed variants of precice that Spack knows, run the following:
spack find precice
To load the preCICE module, run:
spack load precice
You can now use preCICE normally and build any adapter following their respective instructions.
If you want to uninstall preCICE,
spack uninstall precice or delete the complete
spack/ directory to remove everything.
Installing the python bindings
To install the python bindings using Spack, run the following:
spack install email@example.com
Then to use the python bindings:
spack load firstname.lastname@example.org
Use dependencies from your system
You can instruct Spack to recognize specific dependencies that are already installed on your system, by modifying your preferences in
For example, to specify a locally installed MPI version, you could write:
Here we specify that a local install of OpenMPI version 3.1.2 exists in
buildable flag specifies that Spack is allowed to look for and build newer versions of the package if they exist instead of using the locally available one. Here we set it to
false to prevent Spack from trying to build a newer version and add unnecessary installation time.
Install preCICE without non-essential dependency extensions
You might want to opt out some default install options for some dependencies of preCICE as they can cause conflicts. Specifically, extensions of Eigen and Boost can cause errors.
To only install the essential boost libraries that are used by preCICE, you can strip away some default options of the Eigen and Boost packages:
spack install precice ^email@example.com -atomic -chrono -date_time -exception -graph -iostreams -locale -math -random -regex -serialization -signals -timer -wave ^firstname.lastname@example.org -fftw -metis -mpfr -scotch -suitesparse ^email@example.com
Note that, here, we install preCICE specifically with Boost 1.65.1 and Eigen 3.3.1. We also demand OpenMPI version 3.1.2 as this allows Spack to use the local OpenMPI install we specified in the example
packages.yaml above. This is not necessary: feel free to use any other OpenMPI version or just fully omit the
^openmpi argument to let Spack decide.
I need preCICE v1.x with Python enabled, how can I get it from Spack?
Spack has removed support for Python 2. You can still, however, use Spack 0.19.
I need more help with Spack
Look first for topics with the
spack tag on our forum on Discourse. If you don’t find anything, we will be happy to help you there!