Challenges of mapping financial analytics to many-core architecture

Summary form only given. In the past 20 years there has been an explosive growth of the variety of traded financial instruments, from European and American options to a more complex, alas ill-fated, credit derivatives. The rapid increase in computational power coupled with the use of mathematical tools for valuing these instruments and estimating the risk has given rise to the discipline of computational finance. Multi- and many-core architectures present significant potential for performance gains in financial applications. Recent years have seen an emergence of a variety of such designs, from general-purpose multi-cores, to GPGPU-, ASIC- and FPGA- style accelerates. To efficiently utilize hundreds of gigaflops offered by such systems requires serious optimization and parallelization effort from an application programmer. This can be a significant deterrent to a quant, traditionally focused on development, validation and deployment speed of a given pricing model, rather than optimizing model implementation for higher performance. In this talk I will describe several recent parallel hardware and software platforms for accelerating financial analytics. I will also discuss the impact that many-core era will have on financial industry and implications for both quants as well as parallel platform designers.