Make BOUT++ a separate package.#6
Conversation
…nding the C version of the MPI library.
|
N.B. fa6fefc can be reverted if using a version of bout with the changes from boutproject/BOUT-dev#3101 |
…es where possible.
…ffd17bc595900ac269708c008 and on bout/next at bcfde74a43c82771ff2970df980e2e450191778f).
e725b1e to
6430d09
Compare
…sions <5 that didn't use CMake.
mikekryjak
left a comment
There was a problem hiding this comment.
Great work on this @oparry-ukaea, it really seems like you've managed to capture almost all of the default behaviour properly (at least as far as I can tell). I checked by running a simulation with this version and with my existing installation and comparing the BOUT++ console dump. I am yet to compare the configure console output.
| "python", | ||
| "raja", | ||
| "scorep", | ||
| "sundials", |
There was a problem hiding this comment.
Both SUNDIALS and PETSc need to be fixed to the latest version in the CMake. Upgrading these versions is not always trivial.
There was a problem hiding this comment.
SUNDIALS now limited to versions between 2.6 and 6.7.0 as per the bout docs.
PETSc now limited to 3.7:3.23 as per the bout (5.2.0) docs (I also needed to restrict boutpp@5.0.0 to PETSc <= 3.17 to make it build).
There was a problem hiding this comment.
I think SUNDIALS 7 and PETSc 3.24 do work, although that will depend on the BOUT++ version
There was a problem hiding this comment.
Happy to change it if you're reasonably confident; it certainly seems to build fine with sundials@7.0.0, but I haven't run the tests.
… those options up to PETSc instead. +hypre and +slepc are left in to trigger the corresponding BOUT_USE definitions.
|
Think this is ready to merge. |
(main after boutproject/BOUT-spack#6 merge).
Adds a package file for BOUT++ and makes it a dependency of the Hermes-3 package.
Resolves #3.
Resolves #9.
ToDo:
- [ ] Add "coverage" variant (conflicts if compiler isn't gcc or clang); defaults to offSkipping coverage variant, for now; can be added later if necessary.