Originally reported byZachary Hafen (Bitbucket: zhafen, GitHub: zhafen)
Running parallelized galaxy finding (using galaxy_find.ParticleTrackGalaxyFinder with n_processors>1) produces results that don't agree with serial galaxy finding. In particular the test in tests/test_galaxy_find.py::TestParticleTrackGalaxyFinderParallel fails.
I checked the serial results, and added some additional checks to make sure that we get the expected results at multiple snapshots, but all passed, so the serial code is still working as expected, and matches what would happen if we used GalaxyFinder by hand.
This is difficult to debug, because it requires using parallel debugging tools. However, since serial galaxy finding still works, and is sufficiently fast, this is not critical.
zhafen
commented
7 years ago
Originally reported by Zachary Hafen (Bitbucket: zhafen, GitHub: zhafen)
Running parallelized galaxy finding (using
galaxy_find.ParticleTrackGalaxyFinderwithn_processors>1) produces results that don't agree with serial galaxy finding. In particular the test intests/test_galaxy_find.py::TestParticleTrackGalaxyFinderParallelfails.I checked the serial results, and added some additional checks to make sure that we get the expected results at multiple snapshots, but all passed, so the serial code is still working as expected, and matches what would happen if we used
GalaxyFinderby hand.This is difficult to debug, because it requires using parallel debugging tools. However, since serial galaxy finding still works, and is sufficiently fast, this is not critical.