ABSTRACT

In this chapter, we will argue that there is a parallelism imperative: quants must learn to write effective parallel code in order to take advantage of future computing hardware. We provide a grounding in the basic computer science and hardware considerations needed to explore effective parallelism in more depth. These are, fortunately, far simpler than the typical financial mathematics encountered in computational finance. We spend the bulk of the chapter applying parts of these foundational points in working a detailed example of coding a nontrivial early exercise LMM problem on the GPU, which highlights the key issues of writing highly parallel code. The results clearly highlight the parallelism imperative—quants who leverage parallel execution well can gain a significant advantage over competitors who do not.