abort criterion

helpful tips or ideas for unconventional solutions
Post Reply
WTMuser
Posts: 31
Joined: Thu Jan 26, 2012 9:33 am
anti_bot: 333

abort criterion

Post by WTMuser » Tue Apr 01, 2014 1:53 pm

Dear Bernd,

we would like to perform MICRESS-calculations in rapid succession (directly operated by a MATLAB-script). To ensure a fast and stable process we are looking for an abort criterion in order to stop and to skip calculations which display numerical problems.

Is there any possibility to define an abort criterion (e.g. CPU-time < tcrit)?

Kind regards
Martin

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: abort criterion

Post by Bernd » Tue Apr 01, 2014 5:17 pm

Hi Martin,

that is an interesting question!

At present, there is only one criterion for a conditional stop of MICRESS, namely the "end_at_temperature". This option is meant for simulations where temperature is part of the result (use of "lat_heat" or 1d_temp), and it makes sense to stop at a certain temperature which may occur at different times.
In principle, one could similarly define other criteria which help you to decide whether a job is still doing serious work or not, but this of course would depend on the concrete case. If you have any specific idea, please tell us, and we can consider to implement that in future!

Apart from that, you could explicitly analyse the MICRESS output files from the MATLAB script in order to find reasonable abort criteria, and explicitly kill the corresponding MICRESS process. But I do not have experiences with MATLAB scripts...

Best wishes

Bernd

Post Reply