Computing value at risk in OpenCL on the graphics processing unit

Abstract

We present our work on computing the value at risk (VaR) of a large hypothetical portfolio in the OpenCL programming model on an AMD FirePro V7900 graphics processing unit (GPU). In the computation of the VaR we follow the delta-gamma Monte Carlo approach. The value change of the portfolio within a short time period is approximated by the sum of a linear delta component and a non-linear gamma component. To approximate the distribution of the value change of the portfolio we generate a large number scenarios. From each scenario a loss or gain of the portfolio is calculated by the delta-gamma approximation. All these potential losses and gains are then sorted, from which an appropriate percentile is chosen as the VaR. We implemented this algorithm in OpenCL. The details are discussed and the experimental results are reported.