Course: “Advanced GPU Supercomputing for HFT”

Course: “Advanced GPU Supercomputing for HFT” is a 3-day, hands-on course for HFT, algorithmic trading and statistical arbitrage practitioners to be held in midtown Manhattan, New York, on Mon-Wed 21st to 23rd February 2011 (new dates!).

Summary: GPUs (Graphics Processing Units) are capable of processing data at phenomenal speeds. Speed up your data and compute intensive applications by x10, x100 and more by applying this exciting new technology. If you are in the business of HFT, algo trading or stat arb, this course will show you how to apply cutting-edge GPU supercomputing technology now to gain an advantage.

You can download the course flyer and registration form in PDF format by clicking on the image below.

Course info and registration form

Course info and registration form

What you will learn:

  • Gain in-depth knowledge of state-of-the-art GPU supercomputing technology
  • Hands-on, you will be writing code throughout the course
  • How to apply GPU supercomputing now
  • Advanced GPU tips & techniques
  • Thinking in parallel, for algo & code design
  • Use GPUs to help design trading strategies & test them
  • Use GPUs for trading strategy execution
Who should attend?
  • This course is for HFT, algorithmic trading & statistical arbitrage practitioners and those considering a move into these areas
  • Trading strategy and algorithm designers
  • Trading strategy and algorithm testers
  • Execution engine designers & programmers
  • Traders, coders, data experts, HFT system administrators responsible for deployment and maintenance of electronic trading systems

Prerequisites:

  • Basic C or C++ programming skills
  • A Windows or Mac or Linux laptop with WiFi connection (we will provide a GPU compute cluster)
  • The course is primarily CUDA based, but has elements of OpenCL. Both APIs use C and C++.

About the speaker – Andrew Sheppard:

Andrew is a former chief quant at a New York based hedge fund (see his bio on the “About” page) and has many years experience in quantitative finance. He is already a published author with the respected technical publisher O’Reilly (www.oreilly.com), and is also the author of their forthcoming book “GPU Programming”. Check him out at “GPU Supercomputing for Finance”.

Course Overview:

DAY 1 – GPU TECHNOLOGY

 

  • GPU supercomputing technology
  • Up and running with GPU (tools & code)
  • Basics of GPU debugging
  • Taking advantage of parallelism (including parallel patterns)
  • “Thinking in Parallel”: How to write parallel code and algorithms
  • Stream processor threads
  • Memory management
  • Performance and profiling
  • GPU libraries: using and building your own

DAY 2 – STRATEGY / ALGO DEVELOPMENT AND TESTING

  • A model system for HFT / algo trading based on GPUs
  • Dealing with Tick Data
  • Level 2 market data and other data
  • GPU Numerics: BLAS, LAPACK, RNG …
  • Streaming data in real-time
  • Basic strategy implementation (VWAP …)
  • Complex Event Processing (CEP)
  • Iterate & find profitable strategies faster

DAY 3 – STRATEGY / ALGO EXECUTION AND MONITORING

  • Where GPUs can make a difference
  • Monitoring and instrumentation – you can only optimize what you measure
  • Efficient data management for GPUs
  • Speed versus precision tradeoffs
  • Latency and determinism on GPUs
  • Handling errors and robustness
  • Real-time risk and compliance
Follow

Get every new post delivered to your Inbox.