Workshop at the Collegium Helveticum Prof. Dr. Satoshi Matsuoka (Tokyo Institute of Technology) Cambrian Explosion of Computing in the Post-Moore Era The so-called “Moore’s Law”, by which the performance of the processors will increase exponentially by factor of 4 every 3 years or so, is slated to be ending in 10-15 year timeframe due to the lithography of VLSIs reaching its limits around that time, and combined with other physical factors. Although there are multitudes of work trying to address this imminent and ultimate threat, most of the work are point solutions to limited scope of problems domains, and moreover, lack the holistic view on (1) how to advance computing on some conventional algorithms such as PDE solvers that has no apparent acceleration techniques, and (2) create a total system architectural stack of post-Moore systems in a holistic fashion, from devices and hardware, abstracted by system software and programming models and languages, and optimized according to the device characteristics with new algorithms and applications that exploit them. Such systems will have multitudes of varieties according to the matching characteristics of applications to the underlying architecture, leading to what can be metaphorically described as Cambrian Explosion of computing systems. For example, the promising new parameter in place of the transistor count is the perceived increase in the capacity and bandwidth of storage, driven by device, architectural, as well as packaging innovations: DRAM-alternative Non-Volatile Memory (NVM) devices, 3-D memory and logic stacking evolving from VIAs to direct silicone stacking, as well as next-generation terabit optics and networks. Here, exploiting the memory and bandwidth capacities will instead be the acceleration methodology. However, such shift in compute-vs-data tradeoffs would not exactly be return to the old vector days, since other physical factors such as latency will not change when spatial communication is involved in X-Y directions, as well as other parameters, and system software, programming, algorithms, and applications will have to cope with the new parametric changes. Such will be the case for all means of acceleration, such as neuromorphic computing, quantum computing, etc., and we need to advance research, especially in algorithms and computing systems, to address such changes as a whole. Prof. Dr. Steve Furber (University of Manchester) SpiNNaker – biologically-inspired massively-parallel computing The SpiNNaker (Spiking Neural Network Architecture) project has delivered a machine incorporating half-a-million ARM processor cores designed primarily for brain-modelling applications, but it can also be used to explore how biological inspiration may advance the capabilities of artificial neural networks and machine learning in the future. Prof. Dr. Frank Wilhelm-Mauch (Universität des Saarlandes) Quantum computers – models, applications, platforms Quantum computing uses the unique properties of quantum physics to speed up some computational tasks. A core principle for this acceleration is their ability to operate fully parallel operations on a single instance of quantum hardware. There are three currently pursued to approaches to realize this promise: Circuit-based quantum advantage for problems, universal fault-tolerant quantum computing, and quantum annealing. These are suited for different types of problems and have different requirements on hardware, reflected in a different status of implementation. There is a wealth of implementation candidates for quantum computing from atomic and solid-state physics – currently led by trapped ions and superconducting circuits with a few interesting runners-up such as semiconductor platforms. The current understanding of these platforms and related platforms allows for speculations about the future development of quantum computing including a range of milestones that could be reached in the next decade. Prof. Dr. Sven Leyffer (Argonne National Laboratory) Overcoming the Power Wall by Exploiting Application Inexactness Energy and power consumption are major limitations to continued scaling of computing systems. Inexactness, where the quality of the solution can be traded for energy savings, has been proposed as an approach to overcoming those limitations. In the past, however, inexactness necessitated the need for highly customized or specialized hardware. The current evolution of commercial off-the-shelf (COTS) processors facilitates the use of lower-precision arithmetic in ways that reduce energy consumption. We study these new opportunities, using the example of an inexact Newton algorithm for solving nonlinear equations. We also describe a set of techniques that use reduced precision to improve the quality of the computed result by reinvesting the energy saved by reduced precision. Prof. Dr. Edith Law (University of Waterloo) Crowdsourcing as a Tool for Research and Public Engagement Science is increasingly data-intensive; yet, many research tasks involving the collection, annotation and analysis of data are too complex to be fully automated. The idea of research-oriented crowdsourcing is to engage online workers or volunteers to contribute and process data towards an academic inquiry. In this talk, I will discuss the variety of research questions that arise from the process of designing crowdsourcing systems for research and public engagement, and illustrate our approach through a description of several ongoing projects on CrowdCurio, our experimentation platform for crowdsourcing. Prof. Dr. Tim Palmer (University of Oxford) Scale-dependent reduced precision for weather and climate modelling The development of stochastic parametrisation for weather and climate models has opened the possibility of working with significantly reduced numerical precision in weather and climate modelling, thereby reducing data transport inside computers. Results using the spectral IFS show the ability to run globally at single precision without degradation with 40% saving in run time. Using emulators it is shown possible to run at almost half precision for large parts of the IFS dynamical core. Work with FPGAs shows the ability to code reduced precision in hardware. I will conclude with some speculative remarks about the possible use of neural nets run at half precision and trained on output from parametrisations.