Abstract
This study investigates the performance of the EGSnrc computer code coupled with a Cell-based hardware in Monte Carlo simulation of radiation dose in radiotherapy. Performance evaluations of two processor-intensive functions namely, HOWNEAR and RANMAR-GET in the EGSnrc code were carried out basing on the 20-80 rule (Pareto principle). The execution speeds of the two functions were measured by the profiler gprof specifying the number of executions and total time spent on the functions. A testing architecture designed for Cell processor was implemented in the evaluation using a PlayStation3 (PS3) system. The evaluation results show that the algorithms examined are readily parallelizable on the Cell platform, provided that an architectural change of the EGSnrc was made. However, as the EGSnrc performance was limited by the PowerPC Processing Element in the PS3, PC coupled with graphics processing units or GPCPU may provide a more viable avenue for acceleration.
Original language | English (US) |
---|---|
Article number | 012028 |
Journal | Journal of Physics: Conference Series |
Volume | 341 |
Issue number | 1 |
DOIs | |
State | Published - 2012 |
Externally published | Yes |
Event | 25th High Performance Computing Symposium 2011, HPCS 2011 - Montreal, QC, Canada Duration: Jun 15 2011 → Jun 17 2011 |
ASJC Scopus subject areas
- General Physics and Astronomy