Hi GEANT4 gurus,
I am using GEANT4 to simulate a medical linear accelerator and the goal of my project is to compute absorbed dose for radiotherapy patient treatment plans. The particles of interest are photons (<7 MeV), electrons (<7 MeV), and positrons. Currently, it’s taking a long time to calculate the dose for a single patient plan (about a week on a 2-core CPU). I have over 90 plans to run. I am interested in accelerating the computing time and asking for suggestions here. Is there a GPU acceleration of GEANT4? I am most likely going to invest in more cores and run my simulation in parallel. Are there alternative pathways to speed up simulations? Thank you so much for reading.
There is no general GPU acceleration within Geant4 at the moment. One possibility for medical applications might be the extensions/applications MPEXS and/or GGEMS, but you would need to contact the developers of these for details. A search on the usual publication databases for “Geant4 GPU” will also produce a large number of hits, so you may be able to find some useful info/contacts through that.
Also a on @weller’s suggestion to look at biasing methods.
Thank you for this suggestion. I am currently using directional bremsstrahlung splitting in my simulation. I will consider additional biasing techniques as well.
Thanks for providing this information. This is super helpful. I will look into these extensions. Have a good day!