MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

An Open-Source Tool for GPU-Based Microscopic Monte Carlo Simulation

M Tsai12*, C Yan1 , Z Tian1 , N Qin1 , S Hung2, X Jia1 , (1) UT Southwestern Medical Center, Dallas, TX, (2) National Taiwan University, Taipei, Taiwan

Presentations

(Thursday, 8/2/2018) 1:00 PM - 3:00 PM

Room: Karl Dean Ballroom A1

Purpose: Monte Carlo (MC) simulation of radiation interaction with water medium at physics, pre-chemistry, and chemistry stages, as well as computing biologically relevant quantities such as DNA damages, are of critical importance for the understanding of microscopic basis of radiation effects. Due to large problem size and the many-body simulation problem, existing packages on a CPU platform are inefficient, with days of computation frequently seen. We have developed a GPU-based microscopic Monte Carlo simulation tool gMicroMC using advanced GPU-acceleration as well as GPU-enabled parallel processing algorithms. This abstract introduces this open-source code to the research community.

Methods: gMicroMC simulates physics stage using an interaction-by-interaction scheme to calculate initial generation of radicals in water by primary and secondary particles. After pre-chemistry stage, initial positions of all radicals are determined. Simulation of radicals’ diffusion and reactions in the chemistry stage is achieved in a step-by-step model using GPU-accelerated parallelization together with a GPU-enabled box-sorting algorithm to reduce computation of interaction pairs. DNA geometry models in a cell nuclear is constructed. Computations of various types of DNA damage in physics and chemistry stages are achieved by analyzing event positions and types relative to DNA model.

Results: Accuracy of physics stage simulation is demonstrated by computing stopping power and track length. The results agree with data in ICRU 16, ICRU 37 and produced by Geant4-DNA with ~15% difference. Difference of yield values of major radiolytic species with Geant4-DNA results is within 10%. Comparing to CPU computation, the speedups of gMicroMC package are 118.4x and 101.1x for 750KeV and 500KeV electrons, respectively. Our gMicroMC package is capable of calculating different types of DNA damages.

Conclusion: We have successfully developed a GPU-based fast microscopic MC simulation, gMicroMC. Combined accuracy and efficiency made it attractive for research. The tool will be made open to the research community.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email