I was running the following code:
%env OPENMC_CROSS_SECTIONS /home/3fi/mcnp_introductor/mcnp_endfb70/cross_sections.xml
geometry = openmc.Geometry.from_xml(’/home/3fi/mcnp_introductor/openmc/geometry.xml’)
geom = openmc.Geometry(geometry.root_universe)
point = openmc.stats.Point((0, 0, 500))
source = openmc.Source(space=point)
settings = openmc.Settings()
settings.source = source
settings.particles = 1000
settings.run_mode = ‘fixed source’
settings.batches = 1
And I got this error:
Can anyone please help me to understand why it has this error?
There are a number of things could be happening here. From the output, it’s hard to say.
One thing that might help is to run the
openmc executable from the command line instead of the Python interpreter. Useful information to help debug can sometimes be filtered out of the Python output. Increasing the verbosity setting of the run might provide more information as well.
Hi @pshriwise ,
I was running with verbosity 10 and in CLI with this command: $ openmc -s 1 -g
It shows segmentation fault as shown below:
Any idea what is wrong here?
I’m honestly not sure. I was expecting more of an indication as to what was going wrong. Is there any chance you can share this model so I can dig into it further?
@pshriwise Here are the files. I converted the mcnp geometry to opnmc format. I am using the void geometry for now just to track the particles. For cross section I am using ENDF/B-VII.0 from LANL-Based Data Libraries | OpenMC
geometry.xml (617.1 KB)
materials.xml (20.4 KB)
settings.xml (294 Bytes)
I would really appreciate your help
@Fahima I’m not 100% sure why the segfault is happening, but I think it may be the following sequence of events:
- Particles are being lost as they are transported
- The lost particle limit is hit on one OpenMP thread, which proceeds to call
fatal_error in the program, which calls
- The call to
std::exit starts tearing down memory
- Meanwhile, other OpenMP threads are still trying to execute and eventually hit some piece of memory that no longer exists, causing a segfault.
If you run with a single thread (
openmc -s 1), you don’t get a segfault but you’ll see a bunch of errors about lost particles. I would recommend following the advice given in our documentation about how to diagnose geometry errors.
@paulromano I was trying to debug with only one thread. When I was trying with 1E4 particles it run successfully. However, when I run it with 1E5 it shows segment fault after a while. It just show segment fault even though I was running with the command
openmc -s 1. I am again attaching all of my files here if you can help with anything I would really appreciate.
geometry.xml (618.8 KB)
materials.xml (20.4 KB)
settings.xml (327 Bytes)