Hi,
Is there any way to stop an openmc running job without loosing few days of calculation. I remark that the state point file is not created yet.
Tarek El bardouni
Hi,
Is there any way to stop an openmc running job without loosing few days of calculation. I remark that the state point file is not created yet.
Tarek El bardouni
You’ll have to clarify what you mean by “stopping”. Stopping with the intent of restarting the job later? If you didn’t instruct OpenMC to generate statepoint files, in general there is no way to completely stop the job (i.e., terminate the executable) without losing progress. In most Linux shells, you can suspend a running executable (in bash ctrl+Z) and then restart it (fg), but you couldn’t shutdown the computer.
Paul
Thank you for your prompt reply. I asked if it’s possible to tell OpenMC to terminate the calculation normally before the end of batches in order to resume it in future (some thing like Ctrl+C for MCNP). This is useful to examine the convergence of tallies.
Tarek El Bardouni
There’s nothing exactly like that (catching an interrupt signal). I would suggest running with statepoints generated at an interval, e.g.
<state_point interval=“10” />
would create a state point file every 10 batches. You could then examine convergence that way. I’ll also point out that with version 0.7.0 of OpenMC, there is a capability to have a tally “trigger” which terminates the simulation when tallies have reached a certain criterion, e.g. relative error less than 1%.
Tarek,
Unfortunately, CTRL+C will completely stop execution without reporting any useful information about tally convergence. If you are interested in convergence information, you might take a look at the tally trigger feature:
https://mit-crpg.github.io/openmc/usersguide/input.html#trigger-element
https://mit-crpg.github.io/openmc/usersguide/input.html#tally-element
You can set one or more triggers on different tallies and instruct OpenMC to monitor their convergence towards some threshold. The uncertainty / threshold ratio will be reported in real time for the tally furthest from convergence. Perhaps this might be helpful for you.
Best,
Will Boyd
Thank you very much for the two issues; they will be helpful.
Tarek El Bardouni