Improved Parallel Randomized Quasi Monte Carlo Algorithm of Asian Option Pricing on MIC Architecture

Abstract

High-dimensional Option pricing, which plays an important role in complex financial activities, presents a great computational challenge in practice. Randomized Quasi Monte Carlo (RQMC) algorithm is of practical significance for forecasting option prices or other finance derivatives. In this paper, we present an improved parallel RQMC algorithm to forecast Asian option prices using Many Integrated Core (MIC) architecture. The improved algorithm employs novel data structure, independent random generator, vectorization technology, and data alignment. Numerical experiments were conducted on MIC architecture and the parallel performance was then analyzed. A speedup of 1.37 was achieved on MIC over CPU. Efficiency of 70.85% was achieved by using 64 OpenMP threads of a MIC card. An average speedup of 3.38 can be obtained by mixing the CPU and MIC computation in comparison with a single core of the CPU. Ample evidences proved the RQMC algorithm can benefit enormously from MIC architecture.