high performance computing on graphics processing units: hgpu.org

Stochastic Differential Equations are important in many models of various physical or artificial phenomena. To get meaningful results it is desirable to solve the initial value numerical integration problem for a sufficiently large ensemble of realizations. Each element of the ensemble has the same form, thus exposing inherent data-parallelism. We implemented a cross-platform library written in C++ and CUDA that exploits data-parallelism by integrating all realizations in parallel using CUDA, and then computing the properties of the solution using CUDA again. This offers a great speed up over a sequential approach while keeping the overall algorithm for arriving at results essentially the same.