Tuesday November 8th, 2011 Meeting - Austin Java Users Group

. Aparapi: "Write once run anywhere" can now include the GPU
by
Gary Frost

Abstract

This presentation shows how Aparapi, an API for expressing data parallel workloads in Java, can extend Java's promise of 'Write Once, Run Anywhere' to include GPU devices. Existing Java OpenCL bindings require developers to code data parallel algorithms in OpenCL, provide explicit buffer transfers and execution requests at runtime, and if OpenCL were unavailable, code a separate Java implementation and have an appropriate fallback strategy. Aparapi allows developers to code against a simple Java data parallel API. At runtime Aparapi attempts to execute on the GPU by converting bytecode to OpenCL; if OpenCL is unavailable Aparapi will fall back to executing using a Java thread pool. We contrast Aparapi with other OpenCL Java bindings, describe how it works, and walk through some real world examples. We also discuss how to determine whether Aparapi is a viable option for your application.

About the Speaker

Gary Frost is a PMTS Software Engineer at AMD where he has been involved in the development of tools to help optimize the performance of Java Virtual Machines. More recently Gary has been developing tools and techniques which will allow Java developers to take advantage of the huge compute potential of GPU devices.


Location & Time:

Meeting from 7 - 9 PM Li'l Tex Auditorium in the Commons Building (# 137) located at 10100 Burnet Road, near the southwest corner of Burnet Road and Braker Lane on the University of Texas J. J. Pickle Research Campus.


Sponsorship

The Austin Java User's Group is always looking for sponsors to help keep this organization active by providing meeting locations, guest speakers and financial support. Please see our sponsors link for more info.