ALPS Installation on Mac/Linux from Spack Packages

Spack Installation

If you plan to use a parallel version of ALPS or run ALPS in a supercomputer environment, besides the source installation you can also install it through the Spack package management tool. You are encouraged to check out the details about ALPS package in Spack and the Spack documentation.

Spack is able to determine the ALPS dependencies and will install the correct versions required by ALPS. The installation is also local to each user in their own home directory, so it does not affect other users in the same computer cluster.

Installation Steps

First, Spack needs to be cloned from the GitHub repository:

git clone --depth=2 https://github.com/spack/spack.git

This creates a directory called spack. You can source the appropriate script for your shell.

For bash, zsh, and sh users:

. spack/share/spack/setup-env.sh

For csh and tcsh users:

source spack/share/spack/setup-env.csh

Next, Spack needs to find all available compilers in your computer:

spack compiler find

This command will enable Spack to detect all available compilers in the system and create a file packages.yaml under the hidden .spack directory in your home directory. It can be viewed/edited with the command:

spack config edit packages

Alternatively, you can view the available compilers through:

spack compilers

The minimum compiler versions required are gcc@10.5.0 and clang@13.0.1.

We can take a look at ALPS information page in Spack:

spack info alps

It shows the package dependencies of ALPS. All the related packages will be automatically installed through Spack’s package management system.

Finally, let us install ALPS!

spack install alps

To use ALPS, we need to load the package:

spack load alps

Spack Installation on Supercomputer Clusters

If you need to install ALPS on a supercomputer cluster, we recommend submitting a batch job to install ALPS through a job node instead of running the above commands on a login node. Since the installation sometimes takes a long time, using login node to install ALPS could affect other users of the cluster.

We have successfully installed ALPS on the following supercomputer clusters: NCSA Delta (Illinois), PSC Bridges (Pittsburgh), Purdue Anvil, SDSC Expanse (San Diego), TACC Stampede3 (Texas). Please read their documentation about how to submit a batch job. Below is one sample batch script we used to install ALPS on the NCSA Delta supercomputer cluster.

#!/bin/bash
#SBATCH --mem=16g
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=1
#SBATCH --cpus-per-task=1    # <- match to OMP_NUM_THREADS
#SBATCH --partition=cpu      # <- or one of: gpuA100x4 gpuA40x4 gpuA100x8 gpuMI100x8
#SBATCH --account=bdhb-delta-cpu    # <- match to a "Project" returned by the "accounts" command
#SBATCH --job-name=installALPS
#SBATCH --time=05:00:00      # hh:mm:ss for the job
#SBATCH --constraint="scratch"
#SBATCH -e slurm-%j.err
#SBATCH -o slurm-%j.out
### GPU options ###
##SBATCH --gpus-per-node=2
##SBATCH --gpu-bind=none     # <- or closest
##SBATCH --mail-user=you@yourinstitution.edu
##SBATCH --mail-type="BEGIN,END" See sbatch or srun man pages for more email options

. spack/share/spack/setup-env.sh

spack install alps

Walkthrough Video

Spack Installation of ALPS in WSL.