Open manodeep opened 1 year ago
This is on master
branch, right? What's the failure mode of the tests---wrong pair counts, wrong weights, crashing, etc? Is it worth sharing the test output in case there's a pattern to which ones fail?
Good point! Here is the log from before (not from the fresh one, but should be the same) Corrfunc_master_pytest.log
None of the failures are single-threaded, so it's probably an OpenMP library issue. Maybe the same one from here? https://github.com/manodeep/Corrfunc/pull/258#issuecomment-966474906
General information
Issue description
Cloned a fresh repo on local supercomputer with Skylake cpus and installed with
python -m pip install -e .
and then ranpytest -v
Expected behavior
Tests should pass
Actual behavior
Most tests failed, there were only a handful of tests that seemed to pass
What have you tried so far?
Unsure what is going on. GH Actions on similar AVX512F cpus (may be skylake) seems to pass just fine