List of Invited Panelists for New Commodity Architectures

Workshop Program

List of Invited Panelists for New Commodity Architectures

ATI Technologies Inc.
Mark Segal
The new X1K series of GPUs from ATI provide many features useful for high-performance numerical computation including fine-grained flow control and a tunable latency-hiding memory subsystem. The top-of-the-line is the X1900 XT, which has 48 ALU processors operating in parallel. Modifying graphics APIs could expose the features of these and other GPUs more clearly, improving the performance and speeding development of non-graphics application.

IBM Corporation
Ashwini Nanda
Cell BE Based Systems and Applications
The Cell BE processor developed jointly by Sony, Toshiba and IBM primarily for next generation game consoles, packs a high level of floating point, vector and integer streaming performance in one chip that is an order of magnitude greater than traditional commodity microprocessors. IBM Cell Blade products are high volume building blocks for high performance, scale out servers, and targeted toward a variety of interactive digital media, real time, streaming and supercomputing applications. This talk will discuss our Cell based systems concepts, vision and several early application prototypes and results.

Intel Corporation
Pradeep K. Dubey
Teraflops for the Masses: Killer Apps of Tomorrow
The wave of digitization is all around us. While none of us has a crystal ball to predict the future “killer app” (any new application with universal appeal), it is our belief that the next round of applications will be about solving the data explosion problem for end-users, a problem of growing concern for both enterprise and home users. As one may recall, about a decade ago multimedia processing became feasible on general-purpose processors, and soon gained significant prominence due to its impact on platform architecture and usage models. We are now at a similar juncture. Processor performance has reached a level where we will soon be able to do good enough real-time simulations of some natural physical phenomena and processes on mass computing platforms. This offers unprecedented innovation opportunities (and challenges) for a computing platform optimized for this class of applications.

NVIDIA Corporation
Mark Harris
General-purpose computation on GPUs has been a popular topic for academic research for several years, driven by the high parallelism, increasing programability, and low cost of GPUs relative to other processors. On the latest NVIDIA Quadro and GeForce GPUs, this research is now starting to mature to the point of usability in commercial and industrial applications. Through partnerships with software developers, NVIDIA is helping bring applications to market in diverse areas including medical imaging, finite element simulation, and physics simulation in games. Along the way, we're learning important lessons about building software for heterogeneous systems of multiple processors. In the newly emerging market of commodity parallel computing, efficient applications must use all computational resources available in a system, including multiple GPUs and multiple CPU cores.

Sun Microsystems
Rao Shoaib
Niagara: A chip multithreaded processor
Niagara is the new generation of processor from SUN which has multicore and multiple H/W threads per core. The current shipping versions are 8 core with 4 threads per core. Future versions will also have the NIC on chip as well and with Solaris running on this processor, we can have interesting ways of virtualizing the chip and network stack. The challanges are getting the traditional applications to take advantage of this architecture.