aparapi by Gary Frost, AMD
If you’re working on big data or other compute intensive projects, then don’t miss this opportunity to learn about Aparapi.
Aparapi is an open source project which extends Java’s promise of ‘Write Once, Run Anywhere’ to include GPU devices. At development time, Aparapi offers an API which allows data parallel algorithms to be expressed in Java. At runtime, Aparapi converts Java bytecode to OpenCL and executes the converted code on the GPU. If OpenCL is unavailable, Aparapi will automatically fall-back and execute using a Java thread pool. You can learn more about Aparapi here: http://code.google.com/p/aparapi/
Using Aparapi on the GPU, the IN-SPIRE™ Team from Pacific Northwest National Laboratory demonstrated a reduced query time for their unstructured text data source software from 650 seconds to less than 20 seconds.* For appropriate data-parallel workloads, we have seen 30x performance uplift demonstrated through the use of Aparapi.
This presentation will cover the API and will provide sample code so that you can use the GPU to achieve performance gains on your compute intensive projects. We will also talk about exciting new features and future plans.
*"Information provided courtesy of Pacific Northwest National Laboratory, run by Battelle Memorial Institute, for the U.S. Department of Energy"
Gary Frost is a Software Engineer and Principal Member of the Technical Staff at AMD, Austin, TX. He has been part of the Java Labs team for 6 years and has recently been focused on enabling Java access to GPU/APU devices.
Gary has previously presented at JavaOne, EclipseCon and OSCON as well as at various Java User Groups and has been described as “Easily a top ten in the 500 or so presenters I have seen at my local Java User Group over the past 10+ years” by from appwriter.com.
As well as being a self-confessed technology geek, Gary has recently taken up kite buggying and kite surfing. He still plays soccer (albeit slowly) and has nearly (but not quite) given up on his dream of being a rock guitarist.