Future Architectures should incorporate HPUs

Doug Burger

Abstract:

For decades, the computer architecture research community has focused mostly on fundamental infrastructure, providing innovations and breakthroughs that improved the performance and efficiency of computer systems, permitting application developers to produce progressively more exciting applications and capabilities. That paradigm is under increasing stress as we race to the end of Moore's Law. I will argue that as our field diversifies with an unavoidable and increasing focus on specialization and applications, we should include novel sensors and actuators as part of our system designs, and ultimately humans as well.

Bio:

Doug Burger is Director of the Client and Cloud Applications research group in Microsoft Research. In this role, he directs strategic research efforts aimed at producing disruptive advances in mobile client design, data center design, and the applications and experiences that span the two. He received his PhD in 1998 from the University of Wisconsin-Madison, and subsequently joined the University of Texas at Austin, where he spent ten years on the faculty. With Steve Keckler, he co-led the TRIPS project, which developed EDGE architectures and NUCA caches. The TRIPS project culminated in a working prototype ASIC containing two 16-wide out-of-order issue cores, each with a 1,024 instruction window, architected with a file-grain tiled design. He received the ACM Maurice Wilkes Award in 2006, and is a Fellow of the IEEE and the ACM. He currently serves as past Chair of ACM SIGARCH.