Simulation freezes when using weight windows (version 0.14)

User Support
1

Dear Community,

When running a simulation with weight windows in OpenMC version 0.14, the simulation freezes after a certain number of batches. Changing the seed also changes the batch where the simulation stops.
The same simulation runs fine in version 0.13.

Any suggestions? Thanks in advance!

2

I have met the same issue. I have 22 threads. The simulation froze in a certain batch and only one CPU ran at that time. In 0.13 I also met this issue in some of my models containing wws. No idea what’s the cause.

3

Maybe particles at one thread split a lot but particles on other threads have already reached the end of life, so finally only one thread is running which seems stuck?

Is it possible to redistribute split particles to other threads?

4

Installing from source solves my issue. This probably relates to this bug which has been fixed in develop branch.

5

Thanks for your answers Kingyue. Unfortunately, even after installing the very latest version of the code, the problem still exists. I experienced the same situation as you: out of 48 cores used to run the simulation, only 2 were actually working when the simulation froze. I’ll let you know what I find out.

6

Hi folks, was there any update on how you solved this? I’m currently experiencing this on the latest version.

7

To update everyone, after some testing… I’ve discovered that to solve this you should set the update interval to something higher than 1. more commonly 5-10. Then you can set the amount of realizations to do, I did 3 to test, but have done with 5-10 again and had no problem

8

I’m reviving this discussion since I’m currently running into the same issues. I’m using version 0.15.2 but have noticed that using weight windows often causes my simulations to freeze as described in this thread. This occurs for me when I’m generating the weight windows or if I’m simply using a weight windows file that was generated in a previous run. This is an issue for using previously-generated weight windows because I can not use the solution detailed above since I don’t have an update interval.

This tends to be more likely for larger amounts of particles and batches, but I can’t figure out how to fix it. I set settings.max_history_splits to 1,000 but that has not solved the problem.

9

Update: I reviewed a few other posts with similar issues and came across this one. Particle lost without warning/error or being killed, stalling the run - #4 by openMC_guy

One solution that worked in that case was changing the seed. I tried this in my case as well (settings.seed = 2) and now my simulation does not freeze. I don’t know why this works, but it is a good enough solution for now.