Improve logger when using either CUDA or MPI

celeritas-project / celeritas

Celeritas is a new Monte Carlo transport code designed to accelerate scientific discovery in high energy physics by improving detector simulation throughput and energy efficiency using GPUs.

https://celeritas-project.github.io/celeritas/

Other

64 stars 35 forks source link

Improve logger when using either CUDA or MPI #1482

Closed sethrj closed 2 weeks ago

sethrj commented 2 weeks ago

Add a "null logger" so that the logger can accept (but ignore) anything passed to it. I intend to port this to VecGeom to suppress the "invalid use of __host__ function in HD code" warnings.
Remove dependence of logger on MPI. This should fix a link error @stognini saw and a build error I saw.

github-actions[bot] commented 2 weeks ago

Test summary

3 371 files 5 209 suites 3m 53s :stopwatch: 1 578 tests 1 550 :white_check_mark: 28 :zzz: 0 :x: 17 399 runs 17 333 :white_check_mark: 66 :zzz: 0 :x:

Results for commit daa89924.

:recycle: This comment has been updated with latest results.

sethrj commented 2 weeks ago

@pcanal note that I extended the scope of the PR to refactor the logger internals not to have any MPI usage: instead the default setters are what determines whether the local process has a handler or a "null" handler. I think the previous iteration wasn't right, and it was hard to tell because of the structure.