Comment (by Ken Hui):

Thanks for the suggestion on the MoL parameter setting, I will give a try on that.

You mentioned that OpenMP might give different result due to the unfixed order of operations, but I have got exactly the same output data (I have been comparing is mainly the h+ hx in mp_Psi4_l2_m2_rXXX.00.asc and the rho_maximum in hydrobase-rho.scalars.asc) with no discrepancy at all if I ran the simulation again with the same number of core setting, does that mean the problem is not related to OpenMP?

For all simulations I have done, I start the simulation from scratch each time (reading from the BNS initial data all over again), so probably the issue I am having is not related to the checkpoint/recovery part.

Anyway thanks again for the suggestion, I will try re-run the simulations to see if I still got very different output using different core number.

--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2516/different-simulation-result-with-different