Improving performance of a Computational Neural Network Library

In order to convert a physical problem into a computational one, there should be libraries to support the execution of the program. Insilico is a library used to simulate network of neurons. The simulator numerically integrates a system of coupled non-linear ordinary differential equation (ODE). This study is to provide a stepwise optimization mechanism for the sequential code of this library into a highly parallelized one which can exploit the available processor capabilities and which in turn will effectively reduce the time and increase the efficiency of the application. The one under study is an API to simulate neurons of the olfactory system whose code is highly data dependent with a number of integral calculations. We propose a way of optimizing the application by parallelizing it through OpenMP and efficiently using the massive thread and data parallelism available on the same.