Guest Editor’s Introduction: Trends in

Hewlett-Packard University Memorial Auditorium, August 8-10, 1993. In the five years since its inception, the annual symposium has grown into one of the most popular and fun conferences for chip designers, architects, and users, The informal atmosphere at the beautiful Stanford University campus in late summer, coupled with technical presentations on the latest “product-oriented” chips, seem to be a winning combination for attracting chip enthusiasts not only from nearby Silicon Valley companies and Bay Area universities but also from around the world. Hot Chips has been used as a model for starting new conferences such as the first Hot Interconnects Symposium that was held just before Hot Chips V last August. Speakers at Hot Chips V also commented on the pleasure (and pressure!) of having their presentations attended by most of the thousand or so attendees, due to the single-track program format. The submissions to Hot Chips symposiums usually give a good snapshot of the trends in microprocessor design. For Hot Chips V, the program committee, cochaired by Teresa Meng and myself, had the pleasure of selecting from numerous excellent submissions. The final program included the latest highperformance microprocessors, as in all previous Hot Chips symposiums. The program also reflected a definite industry trend toward more highly integrated chips. Here, the microprocessor designer takes on the role of a microsystem designer involved in systemwide cost and performance optimizations rather than just processor performance optimizations. Another trend is an exciting merging of the fields of communications, computers, and consumer electronics: Computers will be used more as communications tools than merely as computation tools. We could see this trend in a number of “multimedia information processing chips” presented in the areas of graphics, video, and audio processing. It was also clearly reflected in the highly popular evening panel session on “Compelling Applications for Computing in the Year 2000,” where distinguished panelists discussed the changing computing landscape. Yet another trend is the growing appreciation of the importance of the design validation of complex chips and chip systems, as evidenced by the well-received session on this topic. As usual, the program also included a few research projects from both industry and academia that explored the limits of either high-power designs, low-power designs, or alternative design approaches. The articles in this special issue of ZEEEZvficro represent some of these trends in microprocessor design. “Designing the TFP Microprocessor” by Hsu represents the quest for ever higher performing microprocessor chips. The focus of this design is to increase the performance of floating-point-intensive scientific computations with large amounts of data, traditionally executed on scientific supercomputers. This results in a design with four custom VLSI chips: integer, floatingpoint, and two custom cache tag chips. The chips have a modest frequency of 75 MHz: The approach toward supercomputer performance is through processor organization techniques rather than through exploiting process technology. Several interesting features have been implemented to achieve high performance, such as four-way superscalar execution, decou-