Introduction

Spack is a flexible package manager designed for scientific software. It supports multiple compiler versions, dependency resolution, build variants, and installation into isolated environments. This guide provides step-by-step instructions on how to search for, inspect, and install libraries and tools using Spack.

From source with Spack

Visit the Spack Project

Spack contains packages for all of the products listed in the E4S 25.06 Full Release category. General instructions for building software with Spack can be found at the Spack website. Questions concerning building those packages are deferred to the associated package development team.

Some quickstart instructions are below.

1. Set Up Spack

1.1 Clone Spack

git clone https://github.com/spack/spack.git
cd spack

1.2 Add Spack to your shell

. share/spack/setup-env.sh

To load Spack automatically on startup, add the above line to your shell profile (e.g., ~/.bashrc, ~/.zshrc).

Keeping Spack up to date ensures access to the latest package recipes:

spack update
spack install --refresh

3. Finding Libraries and Tools

spack search <name_or_keyword>

Example:

spack search mpi

spack search kokkos

3.3 List all packages

spack list

3.3 List all packages in E4S

spack list --tag e4s

4. Inspecting Package Details

4.1 Show detailed information

spack info <package>

This displays:

  • Description
  • Available versions
  • Variants (options)
  • Dependencies
  • Maintainers

Example:

spack info hdf5

4.2 View available versions

spack versions <package>

5. Customizing Builds with Variants

Most scientific software has optional features. Use + to enable and ~ to disable features.

Example: Enable MPI and threading in HDF5

spack install hdf5 +mpi +threadsafe

Example: Disable shared libraries

spack install trilinos ~shared

6. Selecting Compilers

6.1 List available compilers

spack compilers

6.2 Add a new compiler

spack compiler find

6.3 Install with a specific compiler

spack install kokkos %gcc@12.2.0

7. Managing Dependencies

Spack resolves dependencies automatically. To view the dependency tree:

spack spec <package>

Example:

spack spec kokkos +cuda

8. Installing Packages

8.1 Basic installation

spack install <package>

#E# 8.2 Install a specific version

spack install fftw@3.3.10

8.3 Install with variants and compiler

spack install petsc +mpi +debug %gcc@11.3.0

9. Using Installed Packages

9.1 View installed packages

spack find
spack find -d   # include dependencies

9.2 Load a package into your environment

spack load <package>

9.3 Unload a package

spack unload <package>

Spack environments let you manage sets of packages together.

10.1 Create an environment

spack env create myenv
spack env activate myenv

10.2 Add and install packages

spack add trilinos +mpi
spack add kokkos +cuda
spack install

10.3 View environment packages

spack find

10.4 Deactivate environment

spack env deactivate

11. Using Binary Caches (Faster Installs)

To see available prebuilt binaries:

spack buildcache list

Spack will use binaries automatically if available.

12. Troubleshooting

Issue Solution
Build failed spack -d install <pkg> for debug output
Wrong compiler used Specify %compiler explicitly
Dependency mismatch Check with spack spec <pkg>
Clean build spack clean --all

13. Advanced Tips

  • Search by tags: 
    spack list --tags=math
  • Find reverse dependencies: 
    spack dependents <package>
  • Get help: 
    spack help
spack help install

14. Summary Workflow

# Find a package
spack search <name>

# Inspect it
spack info <name>

# Customize & verify
spack spec <name +options %compiler@version>

# Install
spack install <name +options %compiler@version>

# Load for use
spack load <name>

15. Additional Resources

  • Official documentation: https://spack.readthedocs.io
  • Package index: https://packages.spack.io
  • Tutorials: https://spack.io/tutorials
  • GitHub: https://github.com/spack/spack

Conclusion

Spack makes it easy to discover, configure, and install complex scientific software stacks. By understanding how to search for packages, inspect build options, control compilers and dependencies, and manage environments, you can consistently build powerful and reproducible software configurations.