Periodically storing actor results

[jasqs]

Dear All,

Is there any possibility to store actor results periodically during the simulation? For instance, to have insight into simulation statistics for very long runs. The GATE 9.* offered such a functionality, and I could have defined, e.g., storing every 60 s. Is it possible now?

Thanks for help

[dsarrut]

Hi Jan! This is not available yet. It should be feasible with the new Gate10 architecture, but require some work and this was not a priority. If someone is interested to implement it, I can help/guide to do it.

Rough sketch:

• internal cpp mechanism on GateVActor to store start time
• if the user defined a period, EndOfEvent should check time and period (must be enable for the actor)
• if period is over, trigger a call to python function
• python function must be defined by user (or could be built in)
• unclear how to transfer intermediate computation from cpp to py in a generic way

[dobrienphd]

I would like to add my voice to support the implementation of this feature. I ran into this as a problem when trying to generate a very large phase space file. At the end of a very long simulation it attempted to write the entire phase space file at once and ran out of memory on my 32GB RAM system. As a result it crashed and the incomplete file was corrupted. So for the use case of scoring large phase space files this would be really helpful!

Thanks

[dsarrut]

Yes, indeed, it will be very useful. However, for the moment I dont have enough human ressources to do this task properly. If anyone want to dig into this, it is with great pleasure that we can guide, help and support. Just drop me an email and we can chat by zoom to discuss technical aspects.

[dobrienphd]

I appreciate the quick response! Please know that my comment was not meant as a complaint, just as a vote for this issue when it comes time to triage upcoming changes. I understand the limitations on resources.