Journal: Bioinformatics

Article Title: GPU accelerated biochemical network simulation

doi: 10.1093/bioinformatics/btr015

Figure Lengend Snippet: Timing comparisons. ( A – C ) Time taken to simulate a given number of realisations for a single core of an Intel Core i7-975 Extreme Edition Processor 3.33 GHz (solid line) and one Tesla C2050 GPU (dashed line) for (A) the LSODA (B) the Euler–Maruyama and (C) the Gillespie algorithm, respectively. The relative speed-ups for given numbers of simulations are indicated next to the GPU simulation results. ( D ) Summary of the relative speed-up of the three different algorithms.

Article Snippet: But since in most applications of these algorithms, either in order to explore the parameter space or to perform inference, at least thousands of simulations will be needed for which the GPU outperforms the CPU even for the rather simple p53-Mdm2 model. We also compared the cuda-sim implementations of the LSODA and Gillespie algorithms with implementations in the Matlab package SBTOOLBOX2 ( ) and our Euler–Maruyama implementation with the native sde function within Matlab.

Techniques: