MICRO-28 Final Program


Message from the General Chair

On behalf of the entire organizing committee, it is my pleasure to welcome you to the 28th Annual IEEE/ACM International Symposium on Microarchitecture.

In recent years MICRO has become the leading forum for work on instruction level parallelism. This year interest in the conference has continued to grow; and we had the strongest response yet to the call for papers, with 90 submissions. Of these, 22 were accepted as long papers and 15 as short papers. The reviewing process was extremely rigorous, with each paper receiving nearly six independent reviews (5.6 on average).

In order to complement the strong slate of technical papers, this year we have chosen executives from the computer industry to be our Keynote and Invited Speakers. We are hoping they will be able to provide a different perspective on the industry and its direction. I would like to thank our keynote speaker Richard Baum (IBM Fellow and Vice President), Thomas Jermoluk (President and C.O.O., SGI), and Fred Pollack (Intel Fellow and Director) for agreeing to present their views at MICRO-28. Their visions will be of great interest in their own right.

This year, we also have incorporated a day of tutorials in the conference, featuring three 2 1/2 hour presentations on the microarchitecture design rationale of three important new microprocessors (Intel P6, HAL SPARC64, and IBM PowerPC 604e). The presentations will be given by technical experts from the chip design teams. The tutorials are free of charge to the registrants of MICRO-28.

For the first time the Micro conference has come to the upper midwest and to lovely Ann Arbor, Michigan. Ann Arbor has been selected as the host city because of its central location and proximity to the many outstanding universities in the region.

In addition to the excellent slate of presentations, MICRO is famous for its opportunities for personal interaction with leading researchers in the area. I welcome you all to Ann Arbor, and I look forward to personally greeting each and every one of you!

Trevor Mudge
General Chair, MICRO-28

===============================================================================

Tuesday, November 28

12:00-20:00 Pre-Registration
Campus Inn
18:30-20:00 Welcoming Reception
Campus Inn

Wednesday, November 29

08:00-08:15 Welcoming Remarks:
Trevor Mudge, General Chair
Kemal Ebcioglu, Program Chair

08:15-09:15 Keynote Address: Richard I. Baum
IBM Fellow and Vice President, Systems Architecture and Performance, Systems Technology and Architecture Division, IBM Corporation
Title: "The Coming Decade of Technology-Driven Discontinuities" (0 bytes)

09:15-09:30 Break

09:30-10:45 Session 1: Branch Prediction I
Session Chair: Yale Patt, University of Michigan
"Performance Issues in Correlated Branch Prediction Schemes" , Nicolas Gloy, Michael D. Smith, Cliff Young - Harvard
"Dynamic Path-Based Branch Correlation" , Ravi Nair, IBM T.J. Watson Research Center
"The Predictability of Libraries" , Brad Calder, Dirk Grunwald - University of Colorado, Boulder, Amitabh Srivastava, DEC Western Research Lab

10:45-11:00 Break

11:00-12:15 Session 2: ILP Compilation I
Session Chair: Guang Gao, McGill University
"The Performance Impact of Incomplete Bypassing in Processor Pipelines ", Pritpal S. Ahuja, Douglas W. Clark, Anne Rogers - Princeton University
"Efficient Instruction Scheduling Using Finite State Automata" , Vasanth Bala - HP Labs, Norman Rubin, DEC
"Critical Path Reduction for Scalar Programs" , Michael Schlansker, Vinod Kathail - HP Labs

12:15-13:45 Lunch

13:45-15:00 Session 3: Memory System Issues
Session Chair: Andrew Wolfe, Princeton University
"A Limit Study of Local Memory Requirements Using Value Reuse Profiles", Andrew S. Huang, John P. Shen - Carnegie Mellon University
"Zero-Cycle Loads: Microarchitecture Support for Reducing Load Latency", Todd M. Austin, Gurindar S. Sohi - University of Wisconsin, Madison
"A Modified Approach to Data Cache Management", Gary Tyson - UC Riverside, Matthew Farrens, John Matthews - UC Davis, Andrew Pleszkun - University of Colorado, Boulder

15:00-15:15 Break

15:15-16:15 Session 4: Topics in Software Pipelining and ILP Compilation
Session Chair: Bob Rau, HP Labs
"Petri Net versus Modulo Scheduling for Software Pipelining", V.H. Allan, U.R. Shah, K.M. Reddy - Utah State University
"Modulo Scheduling with Multiple Initiation Intervals", Nancy J. Warter-Perez, Noubar Partamian - California State University, Los Angeles
"Effective Scheduling of Directed Acyclic Graphs under Register Constraints ", Balas Natarajan, Michael Schlansker - HP Labs
"Improving Instruction-Level Parallelism by Loop Unrolling and Dynamic Memory Disambiguation", Jack W. Davidson, Sanjay Jinturkar - University of Virginia

17:00-19:00 Welcoming Reception
Campus Inn

20:30-23:00 Business Meeting
Campus Inn

Thursday, November 30

08:00-09:00 Invited Speaker: Thomas A. Jermoluk
President and Chief Operating Officer, Silicon Graphics, Inc.
Title: "Microprocessors -- Not a Lowest Common Denominator Commodity" (6672252 bytes)

09:00-09:15 Break

09:15-10:05 Session 5: Data flow and Multithreading Architectures
Session Chair: Jean-Luc Gaudiot, University of Southern California
"Self-Regulation of Workload in the Manchester Data-Flow Computer", John R. Gurd, David F. Snelling - University of Manchester (UK)
"The M-Machine Multicomputer", Marco Fillo, Stephen W. Keckler, William J. Dally, Nicholas P. Carter, Andrew Chang, Yevgeny Gurevich, Whay S. Lee - MIT AI Laboratory and Laboratory for Computer Science

10:05-10:20 Break

10:20-11:35 Session 6: ILP Compilation II
Session Chair: Jim Dehnert, Silicon Graphics, Inc.
"Region-Based Compilation: An Introduction and Motivation", Richard E. Hank, Wen-mei W. Hwu - University of Illinois, B. Ramakrishna Rau - HP Labs
"An Experimental Study of Several Cooperative Register Allocation and Instruction Scheduling Strategies", Cindy Norris, Lori L. Pollock - University of Delaware
"Register Allocation for Predicated Code", Alexandre E. Eichenberger, Edward S. Davidson - University of Michigan

11:35-11:50 Break

11:50-12:20 Session 7: Analysis of Branching Architecture
Session Chair: Ed Davidson, University of Michigan
"Partial Resolution in Branch Target Buffers", Barry Fagin - U.S. Air Force Academy, Kathryn Russell - Lockheed Sanders, Inc.
"A System Level Perspective on Branch Architecture Performance", Brad Calder, Dirk Grunwald - University of Colorado, Boulder, Joel Emer - DEC

12:20-13:45 Lunch

13:45-14:35 Session 8: Software and Hardware Object Code Translation
Session Chair: Wen-Mei Hwu, University of Illinois
"Dynamic Rescheduling: A Technique for Object Code Compatibility in VLIW Architectures", Thomas M. Conte, Sumedh W. Sathaye - North Carolina State University
"Improving CISC Instruction Decoding Performance Using a Fill Unit", Mark Smotherman, Manoj Franklin - Clemson University

14:35-15:00 Break

15:00-15:45 Session 9: Data Prefetching
Session Chair: Gearold Johnson, National Technological University
"SPAID: Software Prefetching In Pointer and Call-Intensive Environments", Mikko H. Lipasti, William J. Schmidt, Steven R. Kunkel, Robert R. Roediger - IBM
"An Effective Programmable Prefetch Engine for On-Chip Caches", Tien-Fu Chen - National Chung Cheng University (Taiwan)
"Cache Miss Heuristics and Preloading Techniques for General-Purpose Programs", Toshihiro Ozawa - Fujitsu Laboratories Ltd., Shin'ichiro Nishizaki - Fujitsu Social Science Laboratory Ltd., Yasunori Kimura - Fujitsu Laboratories Ltd. (Japan)

15:45-16:00 Break

16:00-17:30 Panel
Organizer: Jim Bondi - Texas Instruments
Title: "The Best Road to Higher Performance: VLIW vs SMP vs Superscalar vs Multiscalar vs ..."
Panelists:
19:45-20:15 Conference Dinner
Gandy Dancer

Friday, December 1st

08:00-09:00 Invited Speaker: Fred Pollack
Intel Fellow; Director of Measurement, Architecture, and Planning, Microprocessor Products Group, Intel Corporation
Title: "High Performance PC Directions" (6005233 bytes)

09:00-09:15 Break

09:15-10:00 Session 10: Branch Prediction II
Session Chair: Bob Colwell, Intel
"Alternative Implementations for Hybrid Branch Predictors", Po-Yung Chang, Eric Hao, Yale N. Patt - University of Michigan
"Control Flow Prediction with Tree-Like Subgraphs for Superscalar Processors", Simonjit Dutta, Manoj Franklin - Clemson University
"The Role of Adaptivity in Two-level Adaptive Branch Prediction", Stuart Sechrest, Chih-Chieh Lee, Trevor N. Mudge - University of Michigan

10:00-10:15 Break

10:15-11:00 Session 11: Multithreading and Superscalar Design studies
Session Chair: Hans Mulder, Intel
"Design of Storage Hierarchy in Multithreaded Architectures", Lucas Roh - Argonne National Laboratory, Walid A. Najjar - Colorado State University
"An Investigation of the Performance of Various Instruction-Issue Buffer Topologies", Stephan Jourdan, Pascal Sainrat, Daniel Litaize - Institut de Recherche en Informatique de Toulouse, Universite Paul Sabatier (France)
"Decoupling Integer Computation in Superscalar Processors", Subbarao Palacharla, J.E. Smith - University of Wisconsin, Madison

11:00-11:15 Break

11:15-12:30 Session 12: Architectural Features for ILP
Session Chair: Steve Melvin, Zytek
"Exploiting Short-Lived Variables in Superscalar Processors", Luis A. Lozano C., Guang R. Gao - McGill University (Canada)
"Partitioned Register File for TTAs", Johan Janssen, Henk Corporaal - Delft University of Technology (Netherlands)
"Disjoint Eager Execution: An Optimal Form of Speculative Execution", Augustus K. Uht, Vijay Sindagi - University of Rhode Island

12:30-13:45 Conference Luncheon

13:45-15:00 Session 13: Software Pipelining II
Session Chair: Stanley Habib, City University of New York
"Unrolling-Based Optimizations for Software Pipelining", Daniel M. Lavery, Wen-mei W. Hwu - University of Illinois
"Stage Scheduling: A Technique to Reduce the Register Requirements of a Modulo Schedule", Alexandre E. Eichenberger, Edward S. Davidson - University of Michigan
"Hypernode Reduction Modulo Scheduling", Josep Llosa, Mateo Valero, Eduard Ayguade, Antonio Gonzalez - Universitat Politecnica de Catalunya. Barcelona (Spain)

15:00-15:15 Closing Remarks

Saturday, December 2

Tutorials:
Organizer - Yale Patt, University of Michigan

09:00-10:30: SPARC64
Niteen Patkar, HaL Computer

10:45-12:15: P6
Glenn Hinton, Intel

13:15-14:45: PowerPC 604e
Marvin Denman, IBMOTO

15:00-??:00: Discussion
Free for all questions, more discussion, etc.

??:01 : Adjourn
(?? depends on the energy level, flight time, etc.)
*******************************************************************************

                           Conference at a glance:

Tuesday, November 28

6:30- 8:00 Welcoming Reception

Wednesday, November 29

08:00-08:15 Welcoming Remarks: Trevor Mudge, Kemal Ebcioglu
08:15-09:15 Keynote Address: Richard I. Baum, IBM
09:30-10:45 Session 1: Branch Prediction I
11:00-12:15 Session 2: ILP Compilation I
13:45-15:00 Session 3: Memory System Issues
15:15-16:15 Session 4: Topics in Software Pipelining and ILP Compilation
20:00-22:00 Business Meeting

Thursday, November 30

08:00-09:00 Invited Speaker: Thomas A. Jermoluk, SGI
09:15-10:05 Session 5: Data flow and Multithreading Architectures
10:20-11:35 Session 6: ILP Compilation II
11:50-12:20 Session 7: Analysis of Branching Architecture
13:45-14:35 Session 8: Software and Hardware Object Code Translation
15:00-15:45 Session 9: Data Prefetching
16:00-17:30 Panel: "The Best Road to Higher Performance: VLIW vs SMP vs Superscalar vs Multiscalar vs ..."
19:00-21:30 Conference Dinner

Friday, December 1st

08:00-09:00 Invited Speaker: Fred J. Pollack, Intel
09:15-10:00 Session 10: Branch Prediction II
10:15-11:00 Session 11: Multithreading and Superscalar Design studies
11:15-12:30 Session 12: Architectural Features for ILP
12:30-13:30 Conference Luncheon
13:45-15:00 Session 13: Software Pipelining II
15:00-15:15 Closing Remarks

Saturday, December 2

Tutorials: Intel P6, HaL SPARC64, HP PA8000
Organizer: Yale Patt, University of Michigan
*******************************************************************************

Information on the Keynote and Invited Speakers in MICRO-28:

Richard I. Baum (keynote speaker)
IBM Fellow and Vice President, Systems Architecture and Performance
Systems Technology and Architecture Division
IBM Corporation

Dr. Baum's current responsibilities in IBM are leading the creation of strategy, design and architecture for server systems and microprocessors, and the performance modeling and analysis for microprocessor systems. These include the design and analysis of future products and the development of underlying technologies needed for them.

Dr. Baum joined IBM as a Senior Associate Programmer in the MVS Design group in the Poughkeepsie Programming Center in 1976, and advanced to Development Manager in 1979, and then to Senior Programmer Manager in 1982. From 1984 to 1985 he was Technical Assistant to the Poughkeepsie Site General Manager. In 1985, he was promoted to Manager of Systems Technology in the Poughkeepsie Laboratory. After assuming responsibility for Central System Architecture, High End Processor Performance and Future Systems Design, Baum became Director of System Architecture and Design in 1989, and was appointed an IBM fellow while in this assignment in 1991. In 1993, Dr. Baum was appointed Assistant General Manager, Systems Technology, and in 1994, he was appointed Vice President, Systems Architecture and Performance, his current position.

He received a BS in Engineering Physics from Cornell University in 1971, and a PhD in Computer Science in 1975 from Ohio State University.

----------
Thomas A. Jermoluk
President and Chief Operating Officer
Silicon Graphics, Inc.

Reporting directly to Edward R. McCracken, Chairman and CEO, Mr. Jermoluk is in charge of developing and delivering SGI's entire range of computing systems. He is responsible for the high and low-end IRIS workstation product groups, research and development, business development, administration, finance, business systems, operations, corporate manufacturing, customer support and marketing.

Since joining Silicon Graphics in 1986, Mr. Jermoluk has worked on the design of a new CPU and bus architecture and the corresponding UNIX operating system redesign for bringing new levels of performance to the workstation class of machines. Previously, he was with Hewlett-Packard Company as a research and development section manager, responsible for hardware and software development on a new RISC-based computer. Prior to H-P, he was manager of systems software at Bell Laboratories in the UNIX development lab. Mr. Jermoluk has an M.S. in computer science and a B.S. in computer science/electrical engineering from Virginia Tech.

----------
Fred J. Pollack
Intel Fellow
Director of Measurement, Architecture, and Planning
Microprocessor Products Group
Intel Corporation

Fred Pollack has worked at Intel for 18 years. He is currently the Director of a group that is responsible for all x86 platform architecture and performance analysis. He also directs the planning for Intel's future x86 microprocessors. Before this, he was the manager of the P6 Architecture. In January of 1993, he was promoted to an Intel Fellow, one of 9 in the company.

******************************************************************************

                             MICRO-28 COMMITTEES

GENERAL CHAIR:		PROGRAM CHAIR:		PUBLICITY CHAIR:
  Trevor Mudge            Kemal Ebcioglu	  Matthew Farrens
  U. Michigan		  IBM T.J. Watson	  U.C. Davis

STEERING COMMITTEE:
Richard Belgard, Consultant		Gearold Johnson, National Tech'l U.
James Bondi, Texas Instruments		Hans Mulder, Intel 
Matthew Farrens, UC Davis		Yale Patt, Michigan
Wen-mei Hwu, Illinois			Andrew Wolfe, Princeton

PROGRAM COMMITTEE:
Vicki Allan, Utah State			Richard Belgard, Consultant
David Bernstein, IBM Haifa (Israel)	Jim Bondi, Texas Instruments
Bob Colwell, Intel Corp.		Henk Corporaal, Delft U. (Netherlands)
Jim Dehnert, Silicon Graphics		Josh Fisher, HP Labs
Mike Flynn, Stanford			Guang Gao, McGill U. (Canada)
Jean-Luc Gaudiot, USC			Rajiv Gupta, Pittsburgh
Stanley Habib, CUNY			Martin Hopkins, IBM T.J. Watson
Wen-mei Hwu, Illinois			Monica Lam, Stanford
Bill Mangione-Smith, UCLA		Steve Melvin, Zytek
Jaime Moreno, IBM T.J. Watson		Alex Nicolau, UC Irvine
Yale Patt, Michigan			Bob Rau, HP Labs
Vivek Sarkar, IBM Software Solutions	John Shen, Carnegie-Mellon
Gabriel Silberman, IBM T.J. Watson	Jim Smith, Wisconsin
Mike Smith, Harvard			Mary Lou Soffa, Pittsburgh
Andrew Wolfe, Princeton