Portfolio Optimization with Multi-Core GPUs

Abstract

Creativity and complexity of the investment process have also been evolving with the enhancements in the technology and this enables investors to move to more inspirational methodologies and techniques from the conventional investment concepts. Diversification of the investment horizon ensures the optimal expectations for the investors. However the critical factor is the selection of the best investment plan from infinite possibilities available in the modern financial and capital markets. Portfolio optimization is one of the widely used diversification technique which is based on the mathematical theory of optimization. The mathematical structure of the optimizer formulation and the dimension of the data matrices used in this mechanism make the process inefficient in the usual CPU based computing. This attempt is to utilize the computation power of GPU to implement a faster and efficient portfolio optimization tool and the achievement is the implementation of a tool which employs GPU processing power conjunction with the CPU. Simultaneous exploitation of GPU and CPU with the aid of proper parallelism techniques optimises the efficiency of complex computations. GPU s capability in managing and scheduling threads ease the process of parallel functioning in calculations. Also the extremely lightweight GPU threads are less costly to create and switch context. Single instruction multiple data is dial for this kind of computation. These mechanics accelerate the processing time of the portfolio optimizer by fifteen times compared to the conventional CPU based optimizer.