Keynote
We have the pleasure of welcoming three keynote speakers at CompSys'19.
| Monday June 3rd, 11:30-12:30 |
| Shared-Memory Heterogeneous Computing |
| By Dr. Peter Hofstee |
| IBM Research, Austin, TX |
| Abstract Shared-memory heterogeneous computing is gaining ground, but poses challenges to the programming community that are at least as large as those that resulted from the single- to multi-core transition. This talk covers some of the architectural foundations, drawing some lessons from the Cell Broadband Engine ( and the Roadrunner supercomputer ) more than a decade ago, then looks at how this approach is working out in recent systems with shared memory combinations of CPUs and GPUs and/or FPGAs, and finally examines how it is likely to evolve. |
| Short bio Dr. Peter Hofstee is best known for his contributions to Heterogeneous computing as the chief architect of the Synergistic Processor Elements in the Cell Broadband Engine processor used in the Sony PlayStation 3, and the first supercomputer to reach sustained Petaflop operation. After returning to IBM research in 2011 he has focused on optimizing the system roadmap for big data, analytics, and cloud, including the use of accelerated compute. His early research work on coherently attached reconfigurable acceleration on POWER7 paved the way for the new coherent attach processor interface on POWER8. Peter Hofstee is an IBM Master Inventor with more than 100 issued patents and a member of the IBM Academy of Technology. |
| Tuesday June 4th, 13:30-14:30 |
|
| By Prof. Simon Portegies Zwart |
| Observatory, Leiden University |
| Abstract Energy and momentum are conserved in Newton's laws of gravitation.
Numerical integration of the equations of motion should comply to
these requirements in order to guarantee the correctness of a
solution, but this turns out to be insufficient. The steady growth of
numerical errors and the exponential divergence, renders numerical
solutions over more than a dynamical time-scale meaningless. Even
time reversibility is not a guarantee for finding the definitive
solution to the numerical few-body problem. As a consequence,
numerical N-body simulations produce questionable results. Using
brute force integrations to arbitrary numerical precision I will
demonstrate empirically that the statistics of an ensemble of resonant
3-body interactions is independent of the precision of the numerical
integration, and conclude that, although individual solutions using
common integration methods are unreliable, an ensemble of approximate
3-body solutions accurately represent the ensemble of true solutions. |
| Short bio Since 2009 Simon Portegies Zwart leads an interdisciplinary research team on Computational Astrophysics at the Sterrewacht Leiden (CAstLe). This team is currently composedof 1 software engineer, 2 postdoctoralresearchers, 6 PhD students and 4 MSc students. The aim of this team is to study the universe by means ofsimulation. The specific areas of research in astrophysics include the evolution of exotic planetary systems,the evolution of binary (and higher order multiple) stars, and the dynamical evolution of dense stellar systemssuch as globular clusters and galactic nuclei. From a computational point of view the research group aimsat the development of simulation environments for solving the equations for gravitational dynamics, stellarstructure and evolution, hydrodynamics and radiative transfer. Calculations are performed on computers builtby the research group and equipped with GRAPE hardware or graphical processing units (GPU) but also usingsupercomputers and grids. |
| Wednesday June 5th, 9:30-10:30 |
| Riding on the Edge: The Fun, Risks, and Lessons |
| By Dr. Aaron Ding |
| ESS Group, TU Delft |
| Abstract This talk will introduce three initiatives on edge computing, namely FLAMeS, LocalVLC, and FADES, intended for edge analytics, communication and offloading. Besides sharing first-hand experience on system development and insights, the talk will reveal several pitfalls and lessons learned through live cases. The goal is twofold: 1) to disclose blind spots and interesting directions in the edge domain that deserve further investigations, and 2) to share observations on doing research in a “buzzword” domain, especially what may hinder us from transferring the (mostly fun) system work into solid scientific outcome.TBD |
| Short bioDr. Aaron Ding is a Tenure-Track Assistant Professor at the Department of Engineering Systems and Services in TU Delft. He has over 12 years of R&D experience across EU, UK and USA. Prior to joining TU Delft, he has worked at TU Munich in Germany, at Columbia University in USA, at University of Cambridge in UK, and at University of Helsinki in Finland. He obtained his MSc and PhD both with distinction from the Department of Computer Science (Birthplace of Linux) at University of Helsinki. His PhD was supervised by Prof. Sasu Tarkoma and Prof. Jon Crowcroft at University of Cambridge. Aaron's research focuses on edge computing, IoT architecture and distributed networking services. He is a two-time recipient of Nokia Foundation Scholarships, and awarded the Best Paper of ACM EdgeSys, ACM SIGCOMM Best of CCR and PhD fellowships from the Academy of Finland and University of Helsinki. Aaron is the founder of ACM EdgeSys and co-chaired IEEE HotPOST. He is the co-founder of FCG series, a joint initiative to promote active collaborations between top European and Asian research institutes, including Cambridge, Delft, Helsinki, Leuven, London, Munich, Oslo, Stockholm, MPI, EPFL, IMDEA, HKUST, Peking, Tsinghua, and Fudan. |
