HDT is a one-day workshop that aims at bridging between theory and design of hardware. It puts special emphasis on the multitude of connections between the theory of distributed computing and hardware design, apparent in topics such as time and causality, synchronization, and fault-tolerance. The goal is to present questions and results at the forefront of current research and spark a discussion between the fields. The program consists of several invited talks, which target a general CS audience.
HDT will be held on Friday, October 18, 2019 in Budapest, Hungary, and will be co-located with DISC 2019. The workshop is chaired by Moti Medina and Andrey Mokhov. For information on registration, please go to the DISC 2019 website.
Venue: Hotel Danubius Health Spa Resort Margitsziget (special rates can be found in the Accommodation link of DISC).
Registration: registration for the HDT workshop (including all other Friday DISC's co-located workshops) can be made in here. Early registration fees are available until the end of August.
Some pictures from the workshop: from DISC website (among other workshops), and more here.
Past HDT workshops:
Bar-Ilan University, Israel.
Ran Gelles received his B.Sc. (Summa Cum Laude) and M.Sc. from the Technion–Israel Institute of Technology, in 2003 and 2009, respectively. He received his PhD in 2014 from the University of California, Los Angeles (UCLA). Between 2014-2016 he served as a Postdoctoral Research Associate at Princeton University, New-Jersey.Starting 2016, Ran is with the Faculty of Engineering at Bar-Ilan University. Ran has been a visiting research fellow at:CWI, Amsterdam (2011); AT&T, New-Jersey (2012); and Princeton University (2013).We consider fault-tolerant boolean formulas in which the output of a faulty gate is short- circuited to one of the gate’s inputs. A recent result by Kalai et al. [FOCS 2012] converts any boolean formula into a resilient formula of polynomial size that works correctly if less than a fraction 1/6 of the gates (on every input-to-output path) are faulty. We improve the result of Kalai et al., and show how to efficiently fortify any Boolean formula against a fraction 1/5 of short-circuit gates per path, with only a polynomial blowup in size. We additionally show that it is impossible to obtain formulas with higher resilience and sub-exponential growth in size.
Towards our results, we consider interactive coding schemes when noiseless feedback is present; these produce resilient Boolean formulas via a Karchmer-Wigderson relation. We develop a coding scheme that resists up to a fraction 1/5 of corrupted transmissions in each direction of the interactive channel. We further show that such a level of noise is maximal for coding schemes with sub-exponential blowup in communication. Our coding scheme takes a surprising inspiration from Blockchain technology.
Joint work with Mark Braverman, Klim Efremenko, and Michael A. Yitayew
Bar-Ilan University, Israel.
Osnat Keren received her M.Sc. degree in Electrical Engineering from the Technion-Israeli Institute of Technology and her Ph.D. degree from Tel-Aviv University, Israel in 1988 and 1999 respectively. Between 1988 and 1994 she held a chip design and senior DSP engineer position at National Semiconductor, and between 1999 and 2003 she has been the Chief Scientist at Millimetrix Broadband Networks. Since 2004, Dr. Keren has been with the Faculty of Engineering at Bar-Ilan University, Israel. Her present research interests include coding theory, hardware security, spectral methods in logic design, and countermeasures against invasive and non-invasive side channel attacks.The intensive use of portable devices such as smart phones and wearable medical devices has heightened the need to protect the sensitive information processed and stored on these devices. The weakest points, in terms of security, are the intra-communication links within the device, which are exposed to fault injection attacks. In Globally Asynchronous Locally Synchronous (GALS) systems, communication over these parallel point-to-point links is asynchronous. An attacker can tamper with the transmitted data by changing the propagation time of the signals and/or by injecting glitches which are additional pulses in between the transmitted signals.
It is impossible to prevent attacks, but it is possible to increase the hardware’s immunity by adding countermeasures that can detect malfunctioning with high probability, or make the circuit fault tolerant, i.e., enable it to function correctly in the presence of faults.
In this talk we'll present new methods for designing reliable and secure communication across asynchronous channels that suffer from an arbitrary number of skews and glitches.
Max-Planck-Institut für Informatik, Department 1: Algorithms and Complexity, Germany.
Christoph received a diploma degree in mathematics from the University of Bonn in 2007 and a Ph.D. degree from ETH Zurich in 2011. After postdoc positions at the Hebrew University of Jerusalem, the Weizmann Institute of Science, and MIT, he became group leader at MPI for Informatics in 2014.He received the best paper award at PODC 2009, the ETH medal for his dissertation, and in 2017 an ERC starting grant.In recent years, we've been studying to what extent metastability can be handled without resorting to synchronizers or analog design. The main goal is to avoid synchronizer delay, bearing the promise of faster - and thus more accurate - mixed signal control loops for tasks requiring a very small reaction time, e.g. adaptive voltage control or clock synchronization. In this paradigm, metastability is treated as a third state of the "digital" logic, which can be contained using logical masking. An additional useful tool are masking registers; in a nutshell, these use high- or low-threshold inverters at their output to effectively transform metastability into late transitions. I will present an overview of both positive and negative results regarding such circuits, highlighting the utility as well as known limitations of these techniques. I will also touch some open questions and future research.
Presentation slides: PPT.
A model of processes that interact via asynchronous wires carrying Boolean signals is presented. In this model, modules, called processes, can be made arbitrarily complex, can maintain local memory and can have an arbitrary number of inputs and outputs. A variety of circuit models can be represented by networks of signalling processes. It is shown that in a network of signalling processes consisting solely of single-ouput processes and forks, every module is an eventual C element. Consequently, the computational power of such a network is severely limited. This establishes that the celebrated C-element property of DI circuits follows solely from the fact that single output modules communicate over stable asynchronous wires. Conversely, it is shown that any Boolean function can be implemented using four- input/two-output processes where every process is either one gate (single output) or a pair of gates (two output).
This is joint work with Rajit Manohar from Yale.
Presentation slides: PDF.
Vienna University of Technology, Austria.
Ulrich Schmid is full professor and head of the Embedded Computing Systems Group at the Institut for Computer Engineering at TU Vienna. He studied computer science and mathematics and also spent several years in industrial electronics and embedded systems design. Ulrich Schmid authored and co-authored numerous papers in the field of theoretical and technical computer science and received several awards and prizes, like the Austrian START-prize at 1996. His current research interests focus on the mathematical analysis of fault-tolerant distributed algorithms and real-time systems, with special emphasis on their application in systems-on-chips and networked embedded systems.We present some of the results and open challenges in the development of a purely digital model for asynchronous circuits, which shall enables accurate and fast dynamic timing analysis and formal verification of such designs. In our research, we primarily consider advanced delay models, where the input-to-output delays depend on previous inputs also. Such models shall not only be accurate, but also realistic (we coined the term faithful for their conjunction), in the sense that the behavior of circuits described in the model is exactly, i.e., within the modeling accuracy, the same as the behavior of the corresponding real circuit. We proved that all the existing standard delay models, including pure and inertial delays, are not faithful, and proposed the involution delay model as the only faithful alternative known so far. In this talk, we will present an overview of our results and report on some of the problems we are currently working on in this area.
Presentation slides: PDF.
Newcastle University, Newcastle upon Tyne, United Kingdom.
Alex Yakovlev, DSc (2006, Newcastle), PhD (1982, St. Petersburg, Russia) is Professor of Computer Systems Design at the School of Engineering, Newcastle University, where he has been working since 1991. At Newcastle, since 2000 he is Head of Microsystems Group and Founder of Asynchronous Systems Lab, with 60 PhD alumni. He is an international pioneer of asynchronous circuit design and design automation, for which he was elected to Fellow of IET in 2015, Fellow of IEEE in 2016 and Fellow RAEng in 2017. In 2011-2013 he was a Dream Fellow of EPSRC when he introduced an idea of energy-modulated computing. In 2018 He was awarded an IET Achievement Medal for contributions in electronic engineering. His most recent research interests are in exploring computing at the speed of light and using clocks to measure data rather than a synchronisation tool.In this talk we will look at digital circuits from the viewpoint of electrical circuit theory, i.e. as loads to power sources. Such circuits, especially when they are asynchronous can be seen as voltage controlled oscillators. Their switching behaviour, including their operating frequency is modulated by the supply voltage. Interestingly, in the reverse direction if they are driven by external event sources, their switching frequency determines their inherent impedance which itself makes them ideal potentiometers or voltage dividers. Such circuits can be stacked like non-linear resistors in series and parallel, and lend themselves to interesting theoretical and practical results, such as RC circuits with hyperbolic capacitor discharges and designs of dynamic frequency mirrors.
Presentation slides: PDF.