Multiple processor version of a Monte Carlo code for photon transport in turbid media

Abstract

Although Monte Carlo (MC) simulations represent an accurate and flexible tool to study the photon transport in strongly scattering media with complex geometrical topologies, they are very often infeasible because of their very high computation times. Parallel computing, in principle very suitable for MC approach because it consists in the repeated application of the same calculations to unrelated and superposing events, offers a possible approach to overcome this problem. It was developed an MC multiple processor code for optical and IR photon transport which was run on the parallel processor computer CRAY-T3E (128 DEC Alpha EV5 nodes, 600 Mflops) at CINECA (Bologna, Italy). The comparison between single processor and multiple processor runs for the same tissue models shows that the parallelization reduces the computation time by a factor of about N , where N is the number of used processors. This means a computation time reduction by a factor ranging from about 10 2 (as in our case where 128 processors are available) up to about 10 3 (with the most powerful parallel computers with 1024 processors). This reduction could make feasible MC simulations till now impracticable. The scaling of the execution time of the parallel code, as a function of the values of the main input parameters, is also evaluated.