Closed danieljprice closed 8 months ago
On my laptop, running
make SETUP=testgrav phantomtest && ./bin/phantomtest ptmass
with OMP_NUM_THREADS=4 takes
<-- testing complete
total wall time = 1 min, 39.00 s (= 9.9000E+01s)
total cpu time = 4 min, 51.06 s (= 2.9106E+02s)
If we compile with MPI but not openMP then the ptmass test takes
make MPI=yes SETUP=testgrav OPENMP=no phantomtest && mpiexec -np 4 ./bin/phantomtest ptmass
...
<-- testing complete
total wall time = 3 min, 9.62 s (= 1.8962E+02s)
total cpu time = 3 min, 5.11 s (= 1.8511E+02s)
but if we combine MPI+openMP then the ptmass test just hangs...
In recent GitHub actions runs, the sink particle tests with MPI and openMP have got really slow for some reason causing the actions workflows to timeout. Unsure if this is a change from the GitHub side, or a problem from a previously merged commit, but it is blocking code merges at the moment.
Originally posted by @danieljprice in https://github.com/danieljprice/phantom/issues/483#issuecomment-1819815865