1600255 : Exciting code fully parallel installation (with mpiandsmp) on maxwell

Created: 2026-05-12T07:49:50Z - current status: new

Here’s the anonymized and summarized version of the query:


Summary of Issue

A user is attempting to install and run Exciting code (a density-functional theory software) on the Maxwell cluster for parallel computation. They have encountered persistent compilation errors, likely due to missing or misconfigured Intel or cluster-specific libraries. The user referenced a previous successful installation of Elk 10.6.2 (a similar code) but was unable to replicate the process for Exciting. They request assistance in building and installing Exciting with optimized parallel libraries.

Suggested Solution/Next Steps

  1. Check Dependencies:
  2. Exciting requires Intel compilers (ifort/icc), Intel MKL, and MPI libraries (e.g., Intel MPI or OpenMPI). Verify these are available via module load on Maxwell.
  3. Example: bash module load maxwell intel/oneapi mpi/intel

  4. Use Spack for Installation:

  5. Maxwell supports Spack for managing software installations. A Spack package for Exciting may simplify dependency resolution.
  6. Tutorial: Using Spack on Maxwell.
  7. If Exciting is not in Spack’s default repos, consider adding a custom package file.

  8. Review Build Instructions:

  9. The Exciting INSTALL.md (Intel-specific section) suggests setting environment variables like: bash export FC=ifort export F77=ifort export CC=icc export MPIF90=mpiifort
  10. Ensure these point to the correct paths on Maxwell.

  11. Debugging Compilation Errors:

  12. Share the specific error messages (anonymized) for targeted troubleshooting.
  13. Common issues:

    • Missing libraries (e.g., -lfftw3, -lmkl_intel_lp64).
    • Incorrect MPI compiler wrappers.
  14. Request Cluster-Specific Build:

  15. If self-installation fails, the Maxwell team could build Exciting centrally (like Elk 10.6.2). Provide:

    • Desired version (e.g., latest release).
    • Required libraries (e.g., Intel MKL, HDF5).
    • Parallelization needs (e.g., MPI/OpenMP).
  16. Alternative: Attend an HPC Seminar on Maxwell for hands-on guidance:

  17. HPC Computing Seminars.

Sources Used

  1. Using Spack on Maxwell
  2. HPC Computing Seminars