In early October the Computing Community Consortium (CCC) held a virtual workshop on Physics & Engineering Issues in Adiabatic/Reversible Classical Computing. This workshop was organized by Michael (Mike) P. Frank (Sandia National Labs), Tom Conte (Georgia Tech), Erik DeBenedictis (Zettaflops LLC), Jayson Lynch (University of Waterloo), Karpur Shukla (Brown University), Robert Wille (Johannes Kepler University Linz) with support from CCC Systems and Architecture task force members Mark Hill (Microsoft) and Sujata Banerjee (VMWare). It brought together over 40 participants with backgrounds in physics, engineering, and computer architecture to address the challenges that must be overcome to realize practical adiabatic/reversible classical computing.
A workshop report, summarizing the discussions and conclusions from the workshop, is in progress, and it will lay a common foundation of existing state-of-the-art knowledge in the area, detail current gaps in understanding, and outline a research agenda to overcome those gaps. Be sure to keep an eye on the CCC Blog, where will share the report once it’s available. In the meanwhile, recorded plenary presentations from the workshop are now available on the CCC website here. Watch the introductory session, delivered by former CCC Council Chair Mark Hill and workshop organizer Mike Frank, below:
As described in the video, Mike Frank proposed a workshop on reversible computing following the CCC’s January 2019 workshop on “Thermodynamic Computing” (TDC), which explored new computing paradigms inspired by the application of the principles of thermodynamics (and in particular, the modern non-equilibrium/stochastic approaches to thermodynamics). During the breakout sessions of the TDC workshop, Frank proposed a related “priority research direction” on “Physics and Engineering of Reversible Computing Hardware[1].” As discussed in the resulting TDC workshop report, reversible computing can be considered to be the historically first conceptualization of thermodynamic computing, having first been considered in as early as 1961 by Rolf Landauer.[2]
During the opening session Frank also argued that the time is right to push for increased research fundings in reversible computing, especially because industry can no longer rely on Dennard Scaling and Moore’s Law for continued improvements to computing efficiency and are searching for new solutions. Frank added that “there is a significant opportunity here to make some breakthroughs in computing efficiency if we put a concerted effort to both refining existing technology concepts and developing new device and circuit level technology concepts,” and that there is opportunity to advance the field “both through better understanding of fundamental theory as well as more intensive development of reversible architecture and algorithms.”
The workshop was organized around three main topic areas: fundamental physics like Landauer’s principle and physical arguments for why reversible computing is needed; device and circuit-level technology; and architecture and other high-level topics, such as tools, algorithms, systems, and applications. Each of these topics was covered during one of the first three days of the workshop (October 5th-7th). Plenary presentations related to the topic of the day were streamed in the morning (Pacific time), and were followed by a live-Q&A and breakout sessions where the participants could discuss the topic in more detail and select position papers were presented. The final two days of the workshop (October 8th – 8th) featured more breakout sessions to refine and expand upon the previous discussions and drafting of the workshop.
Learn more about this workshop and watch the recorded plenary presentations here.
[1] M. P. Frank, “Priority Research Direction: Physics & Engineering of Reversible Computing Hardware,” January 2019. [Online]. Available: https://cfwebprod.sandia.gov/cfdocs/CompResearch/docs/Frank_Reversible_Computing_2v1.pdf. [Accessed 20 September 2019].
[2] R. Landauer, “Irreversibility and Heat Generation in the Computing Process,” IBM Journal of Research and Development, vol. 5, no. 3, pp. 183-191, 1961.
