Date(s) - 18/03/2013
1:15 pm - 2:15 pm
High-Level Parallel Computing in Finance
— Haskell Case Studies within HIPERFIT —
This talk will discuss parallelisation patterns and implementation variants for computations relevant to the financial industry. As part of the research centre HIPERFIT for Functional High-Performance Financial Information Technology, we are analysing and prototyping computation-intensive application kernels for pricing and financial modelling, to explore the design space for parallelisation and optimisation. Functional languages like Haskell with parallelism extensions provide a good abstraction level to reason about parallel programs without getting lost in implementation details.
We will give a brief introduction to parallel Haskell dialects to put the work into context, and then provide an overview of the research centre and its main directions. In the main part, we will discuss common parallelisation patterns found in financial application kernels, captured as “parallel skeletons” (library functions which describe algorithmic patterns in higher-order functions with a ready-made parallel implementation).
Multicore and GPGPUs receive increasing attention in a number of application areas with high computational demand, including the financial industry. However, identifying, allocating and controlling parallelism in massively parallel hardware is challenging, and programmers need to “think parallel” from the start in order to effectively exploit modern parallel computers. The functional paradigm enables to analyse and design applications for parallel computing from the start, and programming profits from a rich type system and libraries to express algorithms in a mathematical way.
Dr. Jost Berthold is Assistant Professor at DIKU, the Department of Computer Science at the University of Copenhagen. He obtained a doctorate degree from the University of Marburg in 2008, working on concepts and implementation of parallel Haskells, in particular the language Eden. His research interests cover the spectrum of parallel functional programming: programming concepts and parallelisation patterns, language constructs, and questions related to their implementation and pragmatics. Since 2011, Dr. Berthold works at the core of the HIPERFIT research centre which applies advanced programming languages and parallel functional programming to problems from financial mathematics.