{numlib}[foss/2024a] Hypre v2.31.0, MUMPS v5.7.3, PETSc v3.22.0, ParMETIS v4.0.3, SCOTCH v7.0.5, SuiteSparse v7.8.2, SuperLU_DIST v9.0.0

[maxim-masterov]

(created using eb --new-pr)

Note that ParMETIS was moved from glaros.dtc.umn.edu to GitHub. The former URL is not working anymore and the new GitHub repository contains no tags or releases (see this issue). After some browsing, I found the versioned tarball among the static downloads for CP2K, so I'm using it to get the package. Let me know if you have a better idea about where to fetch it from.

Depends on:

• mpi4py {lib}[gompi/2024a] mpi4py v4.0.0 #21563

[maxim-masterov]

removed SuiteSparse-7.8.2-foss-2024a-METIS-5.1.0.eb because it was recently merged with #21417

[smoors]

Note that ParMETIS was moved from glaros.dtc.umn.edu to GitHub. The former URL is not working anymore and the new GitHub repository contains no tags or releases (see this issue). After some browsing, I found the versioned tarball among the static downloads for CP2K, so I'm using it to get the package. Let me know if you have a better idea about where to fetch it from.

you could take the tarball from this commit, which is the initial import of v4.0.3:
KarypisLab/ParMETIS@d90a2a6
https://github.com/KarypisLab/ParMETIS/archive/d90a2a6.tar.gz

i compared the contents with the original 4.0.3 tarball; the only important difference is that the commit does not contain the metis sources. This can be fixed by adding METIS as a dependency for ParMETIS, which is a cleaner approach in any case. Nix also uses the same commit: NixOS/nixpkgs#329102

but the CP2K tarball has the same checksum as the original, so i'm not against just using that.

let us know what you decide?

[maxim-masterov]

I think it's better to use the static tarball for now, until KarypisLab/ParMETIS#27 and KarypisLab/ParMETIS#26 are resolved.