Replace Core State counter synchronization with host functors or device kernel calls

[LSchwiebert]

This PR replaces the synchronized Core State counter getter and setter functions with device kernel calls or host functors. These getter and setter functions were introduced in PR #2177 and #2208, which moved these counters to the device. The objective is to remove synchronization between the host and device when running an event.

Currently, this has code to revise only one synchronization -- the synchronization in the Stepper::operator()(SpanConstPrimary primaries) function -- with a functionally-equivalent executor functor. Once everyone is happy with the solution used here, it will be replicated elsewhere to replace other synchronizations. Just trying to reduce the burden of reviewing similar changes.

[github-actions]

Test summary

 5 905 files   9 496 suites   7m 58s ⏱️
 1 739 tests  1 729 ✅  10 💤 0 ❌
30 748 runs  30 589 ✅ 159 💤 0 ❌

Results for commit 5a5eb8b.

♻️ This comment has been updated with latest results.

[codecov]

Codecov Report

❌ Patch coverage is 78.57143% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 85.21%. Comparing base (05d3f69) to head (5a5eb8b).
⚠️ Report is 7 commits behind head on develop.

Files with missing lines	Patch %	Lines
src/celeritas/global/detail/SetGenerated.cc	62.50%	3 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2242      +/-   ##
===========================================
- Coverage    86.87%   85.21%   -1.66%     
===========================================
  Files         1331     1298      -33     
  Lines        42220    40671    -1549     
  Branches     13000    10595    -2405     
===========================================
- Hits         36677    34657    -2020     
- Misses        4342     4408      +66     
- Partials      1201     1606     +405     

Files with missing lines	Coverage Δ
src/celeritas/global/Stepper.cc	100.00% <100.00%> (ø)
src/celeritas/global/detail/SetGenerated.hh	100.00% <100.00%> (ø)
src/celeritas/global/detail/SetGenerated.cc	62.50% <62.50%> (ø)

... and 735 files with indirect coverage changes

🚀 New features to boost your workflow:

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

[amandalund]

Thanks @LSchwiebert! The small change set is definitely helpful for these types of algorithmic updates, and this looks like the right path forward.

For the particular case you started with (setting the number of pending tracks in the core counters when primaries are inserted) I'm actually not sure we need to update that value at all. That's my fault: in the optical loop, which uses the same CoreStateCounters struct, we do need to keep track of the number of pending tracks, so I think I added it here for completeness since at the time it was free. I don't think it's actually needed in the current stepping algorithm: it's just cleared again almost immediately.

I think it would help me a lot (and you might have already done some of this on your own) to go through each case where we're currently syncing the counters and figure out how we plan to make each one asynchronous (and confirm they are necessary). That should also give us an idea of whether we'll need a special kernel for each case or whether there's a more general, reusable approach.

[amandalund]

You should be able to do this using the KernelLauncher in corecel/sys instead of the ActionLauncher.

[LSchwiebert]

Thanks. I'll switch to that.

[LSchwiebert]

Thanks for taking a look at my PR. I'm going to remove that counter update and confirm that the tests are passing. Assuming that's the case, and I'm sure it is, I'll revert that code but save it as a template for the other similar changes.

I did go through and make a list of the places in the code where we had a sync_get_counters() call and tried to figure out which ones needed to be changed and which ones didn't. I didn't check the sync_put_counters() calls, since they are paired with a sync_get_counters() call if we are changing any counters. Or they are being used to initialize the counters and likely need to remain.

However, I didn't make a note of what changes need to be made, just where they need to be made. So, I'll go back and add that information and then share that document.

[amandalund]

That would be awesome! If you wanted to share a google doc or something ahead of time I'd be more than happy to sketch them out together (since this is mostly for my benefit :) ). Probably a lot of them will be straightforward like the example you implemented here, but there were a couple places where it wasn't obvious to me what the best approach was.

[LSchwiebert]

Closing this for now. Will reopen when the other changes are done -- or at least enough that we want to merge into the code base.