Set up A100 environment on ExCL

[sethrj]

This adds scripts for the somewhat dated but still functional milan0 system, which has two A100s. ROOT is not yet available on the system: only the non-hidden presets defined in milan0.json should work.

Hint for those working on multiple systems sharing the same home directory over NFS: use git worktree add ../celeritas-$(hostname -s) to create a local checkout that shares the same git directory, allowing you to git push in one and git pop in the other (for example).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.89%. Comparing base (d66098c) to head (bd8f5bf).
⚠️ Report is 1 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2147      +/-   ##
===========================================
- Coverage    84.91%   84.89%   -0.02%     
===========================================
  Files         1273     1273              
  Lines        44667    44667              
  Branches     16611    16611              
===========================================
- Hits         37927    37922       -5     
- Misses        4756     4758       +2     
- Partials      1984     1987       +3     

see 5 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

Test summary

 5 726 files   9 212 suites   18m 29s ⏱️
 2 078 tests  2 052 ✅  26 💤 0 ❌
32 019 runs  31 890 ✅ 129 💤 0 ❌

Results for commit bd8f5bf.

♻️ This comment has been updated with latest results.

[LSchwiebert]

On the release build, the following three tests failed because the CPU and GPU results differ. The reldeb build is fine. It's unlikely to be due to your patch, though.

    234 - celeritas/geo/Geometry:SimpleCms* (Failed)        gpu unit
    235 - celeritas/geo/Geometry:TestEm3* (Failed)          gpu unit
    236 - celeritas/geo/Geometry:ThreeSpheres* (Failed)     gpu unit

None of the -orange builds are working. They are all generating errors like the following:

/usr/bin/ld: CMakeFiles/orange_final.dir/cmake_device_link.o: in function __cudaRegisterLinkedBinary_8f27958b_8_SOA3D_cu_a998c327_3337491': /tmp/tmpxft_00098591_00000000-3_cmake_device_link.reg.c:79: undefined reference to __fatbinwrap_8f27958b_8_SOA3D_cu_a998c327_3337491'

The default and full builds aren't working, but I assume that's expected.

[sethrj]

@LSchwiebert Oops thanks for that! Looks like we're pulling in G4VG when we shouldn't be, and that's causing CUDA RDC to activate 😬 An ordering issue meant that VecGeom was still on, which leads to conflicts with the -WL,-z,-defs.

[LSchwiebert]

The release build is still failing these three tests because the CPU and GPU are producing different results, but I'm approving the PR because I don't think that's causing these failures.

    232 - celeritas/geo/Geometry:SimpleCms* (Failed)        gpu unit
    233 - celeritas/geo/Geometry:TestEm3* (Failed)          gpu unit
    234 - celeritas/geo/Geometry:ThreeSpheres* (Failed)     gpu unit

The remaining builds are working, except for full and default.

[LSchwiebert]

The various builds are all working, except as noted in my comment.

[sethrj]

@Rashika-Gupta or @stognini or @elliottbiondo : I updated the environment so that it stores caches in /scratch/${USER} for better performance; it works for me but I'd appreciate another check besides Loren :)