Need large simulation models for Benchmarking

[Rishabh]

Dear All,

I am an M.Sc. Embedded Systems student interested in accelerating the performance (reducing simulation time) of OpenEMS. I am currently looking into the possibilities in optimizing with CPUs and/or GPUs.

In order to do insightful performance analysis, a few simulation models would be required which can be used as benchmarks.

As users and developers of the tool, at least some of you would have models that you would like to run faster. I would be grateful if you could share them.

Feel free to reply here or send me a private message if you have any questions.


Sincerely,
Rishabh

[thorsten]

I think the very first step would be to get something running and make sure you still get the same results. I guess the tutorials are a good place to start with that.
As for bench marking I would maybe create a script that generates e.g. simple cavities with certain growing mesh dimensions as a first arbitrary test and run this for different counts of threads to get a good impression of how openEMS e.g. currently scales. I think I had such a script some time and I think we even discussed it here in the forum. I will have a look around...

Edit: Found it, viewtopic.php?f=3&t=137

regards
Thorsten

[Hale_812]

Ok, I had three months simulating a 2x2x2 m space with Tx/Rx antennas and scattering objects. The goal was to analyze a performance of cm wave radar in transition from near to far field, as well as to see the evolution of reflected pulse depending on changing back-scattering conditions. It was really painful waiting for weeks until the simulation completes (see my earlier post about fail-safe state save-restore feature request viewtopic.php?f=6&t=696)

[Rishabh]

I guess the tutorials are a good place to start with that.

I did profiling on some of the tutorials. I couldn't run the MRI tutorials though. My understanding is that the virtual human models are not open source. Please correct me if this is not the case.

Edit: Found it, viewtopic.php?f=3&t=137

I could not find any script attached by you. I am not sure if I missed it or if it is indeed not in that thread.


Regards,
Rishabh

[thorsten]

My understanding is that the virtual human models are not open source.

No they are not, but you could get access to them if you do e.g. public research e.g. at a university. But I don't think it is necessary for what you want to do.

I could not find any script attached by you. I am not sure if I missed it or if it is indeed not in that thread.

No I never attached it and now I cannot find it any more

I'll keep looking, but maybe you just create on?

regards
Thorsten