Problem with photon-neutron transport

Hello there,
I’m a master student from Italy and I would like to share an issue that I am having with OpenMC. I’m running a photon-neutron transport calculation (via .photon_transport = True) of a PWR benchmark on my personal laptop (MacBook Pro, RAM 8 GB). The calculation crashes because the memory occupation of the calculation is excessive, and It saturate the memory of my computer. The same calculation runs perfectly on Linux. What are the possible causes of this problem?

Hi and welcome Maria,

I would start by reducing the numbers of tallies.
Then reducing the number of cells (and/or lattice elements).

These two are usually what consumes the most memory (RAM).
Notably large, fine mesh tallies.

You can also execute OpenMC with -s 1, which will disable parallelization and marginally reduce memory requirements, but will also drastically cut the performance.

Cheers,
Perry

Hi Perry,

Thank you for your answer and for your suggestion.
I have run a second calculation using just a single PIN, and unfortunately I’m observing, at a certain moment of the calculation, the same steep increase of the memory occupation.
In order to give you an idea, It is like if the calculation is generating too many photons, which is not normal for the simple problem I am studying.

I’ve tried compiling OpenMC using gcc-13 on my MacBook, but I’m receiving an error related with the arm64 architecture.

Thank you,

Maria

Sounds odd.
Can you share your inputs (.xmls)? Are you using depletion? (can be heavy on RAM)
I could run it and see how much RAM it requires on my system, or what else could be going awry.

Compiling the code always has a number of nuances on different systems.
Seems compiling on Apple M1 and arm64 can be tricky:
(OpenMC installtion Mac Apple M1 chip)

Perry

Hello Perry,

No, I’m not using depletion and thank you for the link to the previous discussion on the forum.

Here is the link to my test model:

Thank you for your help,

Maria

Hi Maria,

on my system, your problem runs to completion without any issue or modification.
It consumes ~250 MB of RAM, which is relatively low. It requires this from the start and never increases further.

You mentioned that it also ran for you on Linux.
I don’t have any experience with Macs. My guess is that the OS is not releasing the RAM memory for use by OpenMC (sometimes OS’s will keep it reserved, paged or such).

You could try adding Z boundaries, but since it’s running on mine I don’t think that’s the issue.

Perry

Hello Perry,

I’ve tried adding Z boundaries but it does not seem to make any improvement.
Do you have any tips on how to release the RAM memory?

I also tried to reinstall Openmc following the suggestions made on the previous forum discussion, but my calculation continues to saturate RAM memory.

Maria

Hi Maria,

No, unfortunately I do not have experience with Macs.

Hi Perry,
Thank you for your help anyways.

Maria