In-Depth Training Course on Parallelizing Derivative Pricing and Risk Models using GPU Technology

Presented by CANdiensten and Tech-X Corporation.

Course Objective

This course provides an introduction to parallezing derivative pricing and risk models using CUDA technology from NVIDIA which allows to use Graphic Processing Units (GPU) for general purpose calcuation. The use of GPU allows to speed up the calculation by a factor up to 300 compared to a calculation on traditional CPUs.

Target Audience

This course is developed for people interested in developing CUDA code using the tools available from NVidia.

Delivery Type

Courses are delivered as instructor-led classes in computer classroom facilities. Course topics are presented with alternating sessions of lectures and exercises. Additional online training information is available. All classes feature low student-teacher ratios.

Syllabus

  • The CUDA Programming Model
  • Using, Designing and Implementing Parallel Algroithms
  • The CUDA Memory Model and Memory Optimization
  • GPU processing of large data sets
  • Multi GPU developments, Integrating CUDA in OpenMP
  • Using the Profiler, Debugger and other Tools
  • Case Studies
    • Monte Carlo Pricing of American Options
    • Pricing CDOs
    • Exposure estimation
    • Speeding up a Value-at-Risk calculation
    • and more

Course Materials

Each attendee will be provided with course material and will have access to the latest version of a CUDA development environment. Course materials are distributed in print and on CD-ROM, and are yours to keep; a computer equipped with GPUs installed is available for use during class.

Prerequisites

Course attendees are expected to have familiarity with C/C++ and some basic understanding of finance.

dates

19 October 2010, 3 days

price

EURO 2400 excluding VAT

by

Peter Messmer, Ph.D.

Dr. Messmer received his Diploma and Ph.D in physics from ETH Zürich, Switzerland and was trained in parallel computing at the Swiss Center of Scientific Computing (CSCS) and the Edinburgh Parallel Computing Center (EPCC). His research interests are in computational plasma physics and parallel computing with a particular focus on emerging high-performance architectures. Since joining Tech-X in 2002, he has been working on parallel plasma simulation codes and was responsible both for implementing new physics models as well as optimizing the code for a broad range of parallel computers, including petascale systems.

The large computational demand of these simulations led him to investigate alternate architectures for accelerating scientific computations, including FPGAs and GPUs. He is the project lead of GPULib, a library of general purpose vector algorithms on GPUs usable from within high-level languages like IDL and MATLAB.