[Run issue]: Segmentation Fault at Higher Histories

[renatoiiibolo]

Contact Details

No response

What was the issue?

I am trying to score radiation damage using DNADamageStepbyStep on the TsNucleus geometry using electrons, protons, and alpha particles. I also created a parallel world to which I appended by-voxel EnergyDeposit and Dose scorers. I am also trying to play with the DBSCAN scorer appended to the parallel world.

The runs have so far generated representative DSB distributions. For instance, 10,000 histories for the proton run results in a DSB spatial distribution expected of a 250-MeV proton irradiation; however, there are only 13-15 DSBs here. Since I need more DSBs for spatial clustering analysis, I tried increasing the histories to 12,000, but I run into a segmentation fault. If I remove the DBSCAN scorer block, I can run the 12,000 histories, and even bring it up to 15,000. But, 18,000 histories causes a segmentation fault.

Similarly, for the electron irradiation setup. It nearly always runs into a segmentation fault with DBSCAN activated. Without DBSCAN, it works fine with 12,000 histories, but it crashes and results in a segmentation fault at 15,000 histories.

What seems to be the problem here?

TOPAS-nBio category

Physics

Version

3.0

What Operative System (OS) are you running on?

MacOS

Relevant log output

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

Attestation

• I testify that I have fully read the bug reporting guideline beforehand.

[JoseRamosMendez]

Hi there,
Have you given a look at the example examples/scorers/SSBandDSbWithDBSCAN.txt in version v4.0.1? The scorer uses TOPAS-nBio/scorers/TsScoreSimpleSSBandDSBWithDBSCAN.cc, which was recently restored from version v2.0. In that scorer there's a way to pass the "damage" info to the clustering algorithm that uses DSBSCAN, without adding a DBSCAN scorer.
Best,
José

[renatoiiibolo]

Hi! Yes, I found the example informative. Right now, I already have the DNADamageStepbyStep scorer in my parameter file. I'm trying to run simulations without the DBSCAN scorer or passing the damage info to the clustering algorithm and see how the program works. I'd like to ask if the segmentation faults at higher histories is due to the beam angular spread or beam position spread setup?

Interestingly, for alpha particle runs, I can only reach higher histories (100) if I enable the DBSCAN scorer; without it, I can only reach 50 histories without segmentation faults.