Phoenix Runner

[suzanmanasreh]

All the updates in this PR:

• Top-level CMakeLists.txt
  • /common CMakeLists.txt because I added common as a subdirectory (easy re-factor, but I thought it made sense for a library)
  • cases are custom targets with their own binary directory inside /build
  • warnings if it can't find packages
  • support both module loaded PETSc and package installed PETSc (I don't think the PETSC_INCLUDE_DIRS naturally handles both, so this is messy sorry)
• Case file renames
• Kept old Makefiles to leave that option
• Readme updates
  • This is confusing since the install script that works on my account doesn't work on all Phoenix accounts.
• Working Phoenix runner
  • Had to custom install Lapack and FFTW in /packages for ml gcc/12.1.0-qgxpzk mvapich2/2.3.7-733lcv to work since the FFTW module depends on mvapich2/2.3.6-ouywal, and that one is broken inside the Phoenix runner. This install script is now under install-phoenix.sh, but it’s more of a worst-case scenario because install.sh works for me on Phoenix. install.sh does work on the ICE runner because the ICE FFTW depends on the default mvapich2/2.3.7-1 already.
  • Tests both with Makefiles and CMake. Makefile.in isn't flexible at all though (there's nothing I can do about this), so it doesn't work with all builds.

[sbryngelson]

please rebase so there aren't so many commits

[suzanmanasreh]

i took out the the line that lets you see a PR's commits on master but okay i'll try. the commit history on this is really complicated from having to add and remove my own cases.

[sbryngelson]

this has commits from 3 weeks ago?

[suzanmanasreh]

yeah i basically kept the same branch that i made other PRs from but ran a "git merge master" after pulling the PR onto master and never started a new branch, so the real differences start after #8ec18f5. sorry this is so hacky, i'm trying to resolve conflicts in the rebase now, but there's conflicts in all 96 commits, so it's too cumbersome.

[suzanmanasreh]

I just reset the commit history. Some of the Makefiles on the branch might look old, but that was because I gave all the case files unique names, so a global Makefile wasn't gonna work anymore without some wildcards. I'm hoping Makefiles can be completely removed in the future though if I'm happy with cmake. It handles everything pretty dynamically, but if you need a better explanation of the system I set up, I can try!

[suzanmanasreh]

Maybe someone else could help with the review if they're familiar with cmake? It's pretty much identical to Makefile.in and PETSc's official cmake.

[suzanmanasreh]

@sbryngelson look Phoenix runner works! I'm not putting a MacOS runner in the same PR though cause there's already too much here. The CMake works on Mac too though.