Open cpommranz opened 1 year ago
Hi @cpommranz. Your comment helped me a lot in my own situation. I don't have voxelized phantoms, but I noticed the same behaviour starting, at least one year, ago. Probably, a bit more. Using the _opt3 it works fine for me, too. Have you find the source of this regression?
Kind Regards, Nikos
Hi @NikEfth. I tested two more configurations:
So, this seems to be a regression in the Geant4 v11.1 series. Please note, that GATE v9.3 is only tested with the Geant4 v11.1 series. It compiles fine against Geant4 v11.2.0, but I haven't run the benchmarks.
Thank you. I did as you said and installed geant 11.2 (for the tag v11.2) but unfortunately, the memory leak persists. I think that it is much slower than previously, but I will need to run a full sim to know for sure.
GATE/Geant4 enters an infinite loop when using the emstandard_opt4 physics list for voxelized phantoms. An infinite number of steps with step length 0 and dE 0 is produced (see tracking output below). Geant4 does not detect the stuck particle and does not try to force push it, as it typically does to unstuck particles.
Log output:
The issue is gone when switching to the emstandard_opt3 physics list. I cannot reproduce this issue with GATE v9.1 and Geant4 v10.7.2, so it seems to be a recent regression. I'm currently not sure if this issue is related to GATE or rather Geant4 itself. I will investigate further. A minimal working example is available here: minimal_working_example.zip. Please be prepared to kill the process before running the example, as the vast amount of Hits produced in the infinite loop quickly eats up memory.