PRAHA CVUT
EG.ORG
tutorial 9
T9  Programming the Cell BE for High Performance Graphics

Date: Monday, September 3rd, Time: 09:00 - 12:30, Location: Lecture Room 133

Organizers:  Michael McCool (RapidMind Inc.)
Bruce D'Amora (IBM T.J. Watson Research Center)
 
Speakers:  Michael McCool (RapidMind Inc.)
Bruce D'Amora (IBM T.J. Watson Research Center)


Abstract:

This tutorial will explore programming the Cell BE processor to achieve high performance for graphics applications using the IBM SDK or the RapidMind platform. The presentation will provide an overview of the Cell BE architecture emphasizing HW features exploted by the software and programming models most commonly used when creating or porting applications to the Cell BE Additionally, programming examples detailing the use of both the IBM and RapidMind development environments will be given. The speakers will provide an overview of performance profiling tools as well as tips and techniques for achieving maximum performance.

  


Extended Summary

The material will be divided into two parts, of 90 minutes each.

Bruce D'Amora will present the Cell BE architecture, with an emphasis on the features that distinguish it from other multicore processors: its synergistic processor elements, explicit local memory, DMA transfers, and built-in mechanisms for interprocessor communication and message passing. He will then discuss key programming models most relevent to digital media workoads, review the software development tools provided in the IBM SDK, and walk through a few code examples highlighting where and how the Cell BE architecture is leveraged to improve overall performance.

Michael McCool will then present the RapidMind Developement Platform, which provides a mechanism for programming the Cell BE at a high level from a single-threaded C++ program while taking advantage of all nine cores available in the Cell BE processor. This system uses a unique interface that enables user-directed dynamic code generation for the synergistic processor elements on the SPEs while compiling out the overhead of C++.

Both presenters will discuss techniques for tuning performance for the Cell BE as part of their presentations. Examples will be drawn from the areas of simulation, rendering, and image processing, among others.

A background in software development using C and C++ would be helpful, along with a basic background in computer graphics. The presentation will be targetted at developers and researchers interested in using the Cell BE for developing advanced parallel graphics applications and building high-performance research prototypes.


Speakers' Background

Michael McCool has been an Associate Professor at the University of Waterloo since 1994 and is also Co-Founder and Chief Scientist of RapidMind, Inc. He has published papers in SIGGRAPH, Graphics Hardware, the Eurographics Symposium on Rendering, Graphics Interface, ACM Transactions on Graphics, Computer Graphics Forum, Symposium on Interactive 3D Graphics and Games, Journal of graphics tools, and elsewhere. Research interests include interval and Monte Carlo numerical methods, optimization, simulation, sampling, real-time computer graphics, scientific and medical visualization, vision, image processing, and VLSI hardware design. He has published results on real-time BDRF approximation and representation, rasterization, compressed texture mapping hardware, real- time shading languages, parallel programming, shader metaprogramming, sampling, noise reduction, shadows, polyhedral splines, rendering of vector graphics, and others. He continues to perform research into high-performance parallel computing applications and programming technology in his role as Chief Scientist at RapidMind. Michael McCool has degrees in both Computer Engineering (B.A.Sc. with Math option, 1989, Sir Sandford Medal, Waterloo) and Computer Science (M.Sc. in 1991 and Ph.D. in 1995, Toronto).

Bruce D’Amora is a Senior Technical Staff Member and Digital Media Solutions Architect in the Emerging Systems Software group at the IBM T.J. Watson Research Center. He is currently focusing on the design of Cell Broadband Engine-based platforms to accelerate applications used for creating digital animation and visual effects. Mr. D’Amora has provided numerous talks over the last several years focusing on the uses and programmability of the Cell Broadband Engine for accelerating the creation of digital content. Mr. D’Amora was previously the Chief Software Architect for the IBM Graphics development group in Austin, Texas where he lead the OpenGL development efforts from 1991 to 2000. He holds Bachelor’s degrees in Microbiology and Applied Mathematics from the University of Colorado. He also holds a Masters degree in Computer Science from National Technological University.



« Back
Last update : 12.05.2011

If you have any comments regarding this webpage please contact EG07 staff.