Plenary Presentations from the Physics and Engineering Issues in Adiabatic/Reversible Classical Computing Visioning Workshop
Below you can find selected pre-recorded plenary presentations from the CCC’s visioning workshop on Physics & Engineering Issues in Adiabatic/Reversible Classical Computing. Videos are sorted by the day they were shown, which corresponds to the topic area which they fall under: physics, technology, and architecture or high-level (tools / algorithms / systems / apps.). The workshop was held virtually October 5th – 9th, 2020 to address the physics & engineering challenges in adiabatic/ reversible classical computing. Read more about the workshop’s motivation below or jump to videos:
- October 5th – Introduction
- October 5th – Physics
- October 6th – Technology
- October 7th – Architecture and High-Level
It has become widely recognized that today’s approach to general digital computation, which is based on standard combinational and sequential digital architectures constructed out of standard (irreversible) Boolean logic elements implemented using CMOS (complementary metal/oxide/semiconductor) transistor technology, is approaching fundamental physical limits to further improvements on its energy efficiency and power-limited performance. The final (2015) edition of the International Technology Roadmap for Semiconductors (ITRS), as well as recent editions of its successor roadmap, the International Roadmap for Devices and Systems (IRDS), suggest that a practical limit will be reached by around the year 2030. By the end of the CMOS roadmap, logic signal energies at the gate of a minimum-sized transistor simply cannot decrease much further without running afoul of fundamental limits on efficiency and stability arising from thermal fluctuations. Even moving to “Beyond CMOS” switching devices cannot improve this situation very much, since the same fundamental thermodynamic limits still apply.
Thus, there is an increasing need to explore new fundamental paradigms for the engineering implementation of general computing systems (at all scales from tiny embedded devices to large-scale supercomputers and data centers) in search of novel concepts for computation that can transcend the above limits that are inherent to the traditional irreversible digital paradigm. The space of ideas that have been considered include a variety of concepts for “physical” computing (computing that leverages fundamental physics to do computing in a more direct way than in the traditional digital paradigm), including various analog and stochastic computing concepts, as well as quantum computing (for problems amenable to quantum speedups).
This workshop gathered the research community in this field, laid a common foundation of existing state-of-the-art knowledge, and the participants are currently drafting a comprehensive workshop report that can make the case for a major new initiative effectively to federal-level decision-makers.
Learn more about the workshop on the workshop webpage.
Workshop Introduction
Speaker: Mark Hill (Microsoft) and Michael P. Frank (Sandia National Laboratories)
Presentation Title: Workshop Introduction
Abstract: The Computing Community Consortium (CCC) held a virtual workshop the week of Oct. 5-9 to address the physics & engineering challenges in adiabatic/ reversible classical computing. This workshop gathered the research community in this field to lay a common foundation of existing state-of-the-art knowledge and work together to prepare a comprehensive workshop report that can make the case for a major new initiative effectively to federal-level decision-makers.
Download presentation slides here.
Michael P. Frank: Fundamental Physics of Reversible Computing -- An Introduction
Speaker: Michael P. Frank (Sandia National Laboratories)
Presentation Title: Fundamental Physics of Reversible Computing — An Introduction
Abstract: The concept of using reversible computing to circumvent fundamental physical limits on energy efficiency has historical roots going all the way back the 1961 work of Landauer, and was shown theoretically workable by Bennett in 1973. But, over the last 59 years, relatively little attention has been paid, from a fundamental physics perspective, to the question of just how energy efficient, as a function of speed, practical physical implementations of reversible computing can be made to be. To finally answer this question in a definitive way is becoming an increasingly important task as the conventional, non-reversible computing paradigm approaches its limits. Recruiting the physics community to turn increased attention to solving this and related problems is one of the major motivations to this workshop. In this talk, we kick off the Fundamental Physics session by giving an overview of what’s already known about the fundamental physics of reversible computing, and highlighting some important research challenges in this area.
Norm Margolus: Quantum Limits on Classical Computation
Speaker: Norm Margolus (MIT)
Presentation Title: Quantum Limits on Classical Computation
Abstract: Energy and momentum define the maximum speed and resolution of distinct quantum evolution. Macroscopically, this makes classical evolution effectively a finite-state computation — with basic mechanical quantities becoming counts of distinct states. Microscopically, this governs what can be achieved by wavefunction evolution manipulating classical information — which is much less delicate than quantum information. This suggests that atomic-scale elements can perform classical logic, implementing a crystalline model of computation resembling classical mechanics.
Download presentation slides here.
Neal Anderson: Physical Information and Fundamental Energy Limits in Computation
Speaker: Neal Anderson (University of Massachusetts Amherst)
Presentation Title: Physical Information and Fundamental Energy Limits in Computation
Abstract: In this talk I will discuss fundamental energy limits for classical computing processes and their transparent connection to physical law. I will first show how a simple generalization of Landauer’s limit follows from quantum dynamics and established entropic inequalities alone. I will then highlight the connection between energy dissipation and the irreversibility of information loss in the simple Landauer erasure scenario used for this proof, and highlight the limitations of this scenario and potential and existing objections to such proofs. Finally, I will sketch more sophisticated proofs of energy limits for scenarios far more complex and general than Landauer erasure—both computationally and physically—that overcome or sidestep these limitations and objections.
Karpur Shukla: Foundations of the Lindbladian Approach to Adiabatic and Reversible Computing
Speaker: Karpur Shukla (Brown University)
Presentation Title: Foundations of the Lindbladian Approach to Adiabatic and Reversible Computing
Abstract: More general and robust expressions for the fundamental energy limits of classical reversible operations can be derived from nonequilibrium quantum thermodynamics. In this talk, I’ll briefly review the developments that have been made on the nonequilibrium Landauer principle, as well as the concepts of thermal operations, Kraus operators, and GKSL (Lindbladian) dynamics. I’ll then discuss recent analysis that has been done on GKSL dynamics with multiple asymptotic states, including their geometric properties, and discuss how these can be applied to classical reversible operations in order to derive fundamental dissipation bounds.
Ed Fredkin: Reversible Computing
Speaker: Ed Fredkin (Carnegie Mellon University)
Presentation Title: Reversible Computing
Abstract: The mighty NAND gate was known to be universal. RCA, the World’s largest electronics company, built large commercial computers out of nothing but 2 input NAND gates. Every other successful computer companies designed computers with a variety of different circuits. JIworked at BBN with Minsky and McCarthy. Together they knew more about computers, theoretically and practically than anyone. They assured me that Reversible Computing was impossible but the proof was too complex for me to understand. I wasn’t convinced.
Michael P. Frank: Device & Circuit Technologies for Reversible Computing--An Introduction
Speaker: Michael P. Frank (Sandia National Labs)
Presentation Title: Device & Circuit Technologies for Reversible Computing–An Introduction
Abstract: Over the decades, a wide variety of concepts for the physical engineering implementation of reversible computing have been proposed. To date, the most highly developed approaches are based on adiabatically driven microelectronic switching circuits using semiconducting and superconducting technologies. Less well-developed, but emerging, are approaches based on the ballistic propagation and elastic interaction of localized information-bearing degrees of freedom. In this talk, I survey various approaches, discuss their pros and cons, and suggest general requirements that novel approaches should try to meet.
Sarah Frost Murphy: Quantum dot Cellular Automata: A Brief Introduction
Speaker: Sarah Frost-Murphy (Gem State Informatics)
Presentation Title: Quantum dot Cellular Automata: A Brief Introduction
Abstract: Quantum-dot cellular automata or QCA is a device family that represents data as charge configuration rather than charge flow. It is a relatively mature novel device with technology and architecture being developed simultaneously since the mid-1990s and 2000s. This talk gives an introduction to the device basics and how QCA lends itself to physically and logically reversible computation.
Download presentation slides here.
Ralph Merkle: Molecular Mechanical Computing
Speaker: Ralph Merkle (Institute for Molecular Manufacturing)
Presentation Title: Molecular Mechanical Computing
Abstract: A new theoretical model for mechanical computing is shown that requires only two basic parts: links and rotary joints. These basic parts are combined into two main higher level structures: locks and balances. These suffice to create all necessary combinatorial and sequential logic required for a Turing-complete computational system. A molecular implementation of this proposal should enable 10^21 FLOPS/Watt. The talk covers the material from the following two papers: Mechanical Computing Systems Using Only Links and Rotary Joints Ralph C. Merkle, Robert A. Freitas Jr., Tad Hogg, Thomas E. Moore, Matthew S. Moses, and James Ryley Journal of Mechanisms and Robotics, Dec 2018, 10(6): 061006 https://doi.org/10.1115/1.4041209 Evaluating the friction of rotary joints in molecular machines Tad Hogg, Matthew S. Moses and Damian G. Allis Molecular Systems Design & Engineering, 2017, 2, 235-252 https://doi.org/10.1039/C7ME00021A
Download presentation slides here.
Joseph Friedman: Scalable Reversible Computing with Skyrmion Billiard Balls
Speaker: Joseph Friedman (University of Texas at Dallas)
Presentation Title: Scalable Reversible Computing with Skyrmion Billiard Balls
Abstract: Reversible skyrmion logic leverages magnetic skyrmions in the first nanoscale realization of conservative logic, providing a vision for energy-efficient computation. In this system, magnetic skyrmions propagate through a two-dimensional ferromagnetic structure while performing reversible logic operations at the gate junctions. A simple global clock enables direct cascading with the potential for scalable high-speed low-power reversible Boolean and quantum computing.
Michael P. Frank: Architectural, Algorithmic, and Systems Engineering Issues for Reversible Computing
Speaker: Michael P. Frank (Sandia National Laboratories)
Presentation Title: Architectural, Algorithmic, and Systems Engineering Issues for Reversible Computing
Abstract: The reversible computing paradigm has long-term implications that extend beyond the device and circuit levels, eventually impacting all aspects of computer design. Critical to the continuing development of the reversible approach will be the appropriate consideration, by architects and system engineers, of various overheads, scaling relations, and design trade-offs that come into play. In this talk, we survey the various issues that impact scaling, and briefly review some architectural, algorithmic and higher-level techniques that can help to address them.
Gregory Snider: It’s Time for Adiabatic Computing
Speaker: Gregory Snider (Notre Dame)
Presentation Title: It’s Time for Adiabatic Computing
Abstract: The method of use energy in computation can no longer be simply ignored. This talk will examine the Landauer principle and approaches to computation that build upon it. An approach to reversible adiabatic computing using CMOS transistors will be presented, along with a discussion of possible devices for future computation.
Download presentation slides here.
Noboyuki Yoshikawa: Reversible Quantum – Flux-Parametron: Practical Superconductor Reversible Logic
Speaker: Noboyuki Yoshikawa (Yokohama National University, Japan)
Presentation Title: Reversible Quantum – Flux-Parametron: Practical Superconductor Reversible Logic
Abstract: N/A
Himanshu Thapliyal: Energy Recovery Computing for Low-Energy and Secure IoT Devices
Speaker: Himanshu Thapliyal (University of Kentucky)
Presentation Title: Energy Recovery Computing for Low-Energy and Secure IoT Devices
Abstract: N/A
Download presentation slides here.
Robert Wille: Design Automation for Reversible and Adiabatic Circuits
Speaker: Robert Wille (Johannes Kepler University, Linz)
Presentation Title: Design Automation for Reversible and Adiabatic Circuits
Abstract: N/A
Jayson Lynch: Reversible Algorithms
Speaker: Jayson Lynch (MIT)
Presentation Title: Reversible Algorithms
Abstract: A brief summary of some of the work in reversible algorithms and suggested directions for research in efficient reversible algorithms.
Download presentation slides here.
Erik DeBenedictis: Quantum plus Classical Computation
Speaker: Erik DeBenedictis (Zettaflops, LLC)
Presentation Title: Quantum plus Classical Computation
Abstract: Quantum information technology matured a lot in the last year, leading to the long-term prospect that general, or quantum plus classical, computers will become the most capable and energy efficient. However, the technology that interfaces between the quantum and classical information domains has different requirements from both qubits and CMOS. A careful analysis indicates that reversible technology is ideally suited for this purpose. Thus, reversible technology may be destined to enter the mainstream as a enabler for processing quantum information in lieu of the often cited objective of its restarting Moore’s law.
Download presentation slides here.
