By Rachel Robey
The distinction honors the impact of the TRIPS architecture, “one of the most complex microprocessor prototypes ever built in academia.”
Congratulations to Department of Computer Sciences (CS) Professor Karu Sankaralingam and alumnus Doug Burger PhD’98 (Microsoft), who recently received the 2024 ISCA Influential Paper Award at this year’s International Symposium on Computer Architecture (ISCA) in Buenos Aires, Argentina. Alongside Sankaralingam and Burger, co-authors Stephen W. Keckler (NVIDIA), Ramadass Nagarajan (Intel), Haiming Liu (Google), Changkyu Kim (Meta), Jaehyuk Huh (KAIST), and Charles R. Moore are co-recipients of the award.
Presented annually, the ISCA Influential Paper Award recognizes a standout paper from the ISCA Conference fifteen years prior. Winning papers are those which have demonstrated “the most impact on the field (in terms of research, development, products or ideas) during the intervening years.”
Dubbed the “test of time award,” the ISCA Influential Paper Award is jointly administered by the Association for Computing Machinery Special Interest Group on Computer Architecture (ACM SIGARCH) and the Institute of Electrical and Electronics Engineers Computer Society Technical Committee on Computer Architecture (IEEE-CS TCCA).
TRIPS: A new standard in microchip architecture
Originally published in 2003, “Exploiting ILP, TLP, and DLP with the polymorphous TRIPS architecture” proposed the polymorphous TRIPS architecture, a microprocessor architecture “which can be configured for different granularities and types of parallels.” The work addressed the issue of design fragility, a phenomenon “in which an architecture tailored for one type of workload would perform poorly when processing a different type of workload.” At the time of publication, polymorphous computing was just emerging as a potential methodology for addressing fragility.
“Our primary goal was to demonstrate that a polymorphous architecture — one chip design capable of targeting very different types of workloads — negated the need for numerous designs tuned to specific classes of workloads,” says Sankaralingam. “Today, the ideas from the TRIPS project are seen in how memory systems are built in modern microprocessors. Almost all modern chips have some form of polymorphous memory system.”
TRIPS was developed at the University of Texas-Austin while Sankaralingam, Nagarajan, Liu, Kim, and Huh were PhD students researching under co-advisors and project leads Keckler and Burger. Burger describes it as “one of the most complex microprocessor prototypes ever built in academia” — at the time of publication, it held a never-before-attempted 1,024 instructions in flight.
In a recent retrospective, Sankaralingam reflected on the broad impact of the TRIPS project: “We showed that this partitioning approach results in a less fragile architecture by using polymorphous mechanisms to yield high performance for both coarse and fine-grained concurrence.”
Congratulations again to Karu, Doug, and the co-authors.