|
(302) Improved performance of a PBPK model of endocrine disruption through parallelization of computer codes running on a high-performance computer. Dixon, Kenneth1, Smith, Philip1, Oldham, Janae1, Robinson, Jeffery1, 1 ABSTRACT- Predicting the effects of endocrine modulating substances (EMSs) on individual animals, much less populations of any wildlife species is a daunting task. Mathematical modeling and computer simulation may be the only possible way to gain an understanding of the dynamics of EMS exposure and effects in the environment. We developed a model to assist in the estimation of risk to a population of wildlife from exposure to an EMS using the Matlab programming language. It is based on the compartmental approach to physiologically-based pharmacokinetic (PBPK) modeling. The model predicts the effects of an EMS on an individual animal and its offspring. Several individuals are simulated to obtain population level effects. Using a time-step of one hour, the model can take several hours to run on a Pentium III PC. As the model becomes more complex (with additional endocrine systems) it will take even longer to run long-term simulations. This makes model calibration and testing a time-consuming process. One approach to improving model performance is to convert the program to parallelized Fortran or C code in which separate parts of the program run on individual processors in a high-performance computer. We examined the performance of parallelized C++ and Fortran codes by scaling from two to 56 processors on a SGI/Cray Origin 2000 computer. Key words: supercomputers, high-performance computing, parallel computing, PBPK models |
