StreamHPC is at HiPEAC, Berlin

hotel-smallHiPEAC is the conference on “High-performance and Embedded Architectures and Compilers” to be held on 20 to 22 January in Berlin – tickets are still available, if you can spare three days. I (Vincent Hindriksen) would like to share the program I’m doing.

Edit: I met great people here and hope to see you next year (again).


It is about “Programmability Issues for Heterogeneous Multicores” (website, conference-page). I chose this, because I think this is an important subject as we move to a language we want to lay on top of OpenCL – how needs the perfect higher-level languegae look like?


  • A Framework for Profiling and Performance Monitoring of Heterogeneous Applications (Perhaad Mistry, Yash Ukidave, Dana Schaa, David Kaeli)
  • Embedding OpenCL in GHC Haskell (Benedict R. Gaster, J. Garrett Morris)
  • Transaction Coalescing – Lowering Transactional Overheads by Merging Transactions (Vesna Smiljkovic, Srdjan Stipic, Osman Unsal, Adrian Cristal, Mateo Valero)
  • Software Reliability Enhancements for GPU Applications (Si Li, Naila Farooqui, Sudhakar Yalamanchili)
  • Locality-Aware Work Stealing on Multi-CPU and Multi-GPU Architectures (João V. F. Lima, Thierry Gautier, Nicolas Maillard, Bruno Raffin)
  • Energy-efficient Mapping of Task Collections onto Manycore Processors (Patrick Cichowski, Jörg Keller, Christoph Kessler)

Of course nice to see a big name in OpenCL here: Benedict Gaster. But look closely to all talk-titels and discover each is of interest to the GPU/Vector-programmer.


The next morning is reserved for “LPGPU Workshop on Power-Efficient GPU and Many-core Computing” (website, conference-page). I have been interested in the LPGPU-project and look forward to hear more – it has been organised by our friends at CodePlay who are focusing more on mobile since a while.


  • Trends in heterogeneous systems architectures (Simon McIntosh-Smith – University of Bristol)
  • A Framework for Modeling GPUs Power Consumption (Sohan Lal – TU Berlin)
  • An in-depth analysis of the Epiphany-IV 28nm 64-core coprocessor (Andreas Olofsson – Adapteva)
  • Practically Costless Coherence for GPUs and Manycore Accelerators (Stefanos Kaxiras – Uppsala University)
  • Architectural Numeration (Theo Drane – Imagination Technologies)
  • The Impact of Different Compiler Options on Energy Consumption (James Pallister – Embecosm/University of Bristol)

Noticed Imagination and Adapteva in the list? You see this workshop is more hardware-oriented. As I always say during trainings: architecture-knowledge is the foundation for GPU-programmers. It is also somewhat more academic, but then I’ll learn more things for the long term.


The “International Workshop on Adaptive Self-tuning Computer Systems” (conference-page) is about various ways of self-tuning. This is a change in my program, which luckily was possible. Self-tuning I find an uttermost important subject.

  • Automatic Performance Tuning and Machine Learning – sponsored by Microsoft Research (Markus Pueschel – ETHZ)
  • Application-Level Voltage and Frequency Tuning Of Multi-Phase Program on the SCC (Kenneth Berry – Florida International University, Felipe Navarro – Florida International University, and Chen Liu – Clarckson University)
  • Bio-inspired Self-Tuning Mechanism for Distributed Computing (Ichiro Satoh  – Japan National Institute of Informatics)
  • Position Paper: Evolving Advanced Neural Networks on Run-Time Reconfigurable Digital Hardware Platform (László BAKÓ  – Sapientia University Romania, Sándor-Tihamér BRASSAI – Sapientia University Romania, Lajos LOSONCZI – Lambda Communication Ltd, and László-Ferenc MÁRTON – Sapientia University Romania)
  • Adaptive OpenCL (ACL) Execution in GPU Architectures (Dan Connors, Kyle Dunn and Jeff Wiencrot – UC Denver, USA)
  • Position Paper: Weak Heterogeneity as a way of Adapting Multicores to Real Workloads (Erik Tomusk and Michael O’Boyle – University of Edinburgh, UK)
  • Position Paper: Code Specialization For Red-Black Tree Management Algorithms (Alexandre Carbon, Yves Lhuillier and Henri-Pierre Charles – CEA, LIST)
  • Sambamba: Runtime Adaptive Parallel Execution (Kevin Streit, Clemens Hammacher, Andreas Zeller and Sebastian Hack – Saarland University)

This is the most international speaker-list I saw. I wanted to make a lit and put “Adaptive OpenCL” on #1, but that would be only because I can nod a lot. I am happy I changed the program, as this list is very interesting for me (besides OpenCL professional, also a background in neural networks).


The last day is all for the “Joint ENCORE&PEPPHER Workshop on Programmability and Portability for Emerging Architectures” (website, conference-page). Like LPGPU two fantastic projects I want to hear all about.


  • Accelerator Codes Autotuning with OpenACC and OpenHMPP (François Bodin  – CAPS)
  • Algorithm Engineering for Large Data Sets (Peter Sanders – Karlsruhe Institute of Technology)
  • PEPPHER: Performance Portability and Programmability for Heterogeneous Many-core Architectures (Siegfried Benkner  – University of Vienna)
  • Optimized Composition of Performance-aware Parallel Components (Christoph Kessler – Linköping University)
  • ENCORE: Enabling technologies for a programmable many-core (Alex Ramirez – BSC)
  • Reducing GPGPU programming efforts by improving its memory efficiency (Chunyang Gou – University of Delft)
  • The Formic architecture and Myrmics runtime system (Polyvios Pratikakis & Spyros Lyberis – FORTH)

First talk is hopefully not exactly the same as the day before, as the audience is completely different. Again academic many talks, so bringing my (paper) notebook, as I assume

Khronos (Benelux) Chapter meetup

At Wednesday evening you are welcome to meet other Khronos-technology developers at this Benelux-chapter meetup – open for developers from all countries, as there are not many chapters in Europe. This is not OpenCL only, but also OpenGL, OpenGL ES, WebGL, WebCL, COLLADA, OpenVG, OpenSL ES, OpenMAX AL, OpenMAX IL, EGL, StreamInput, OpenGL SC, OpenMAX DL, OpenKODE, OpenWF and OpenML.

Khronos Chapters are a new initiative. If you want to know more about it, come over and I’ll tell you everything.

Subject: How can open standards better compete with closed formats?
: Wednesday 23 January, around 19:00.Where: Location TBD.

Ping me at or +31645400456 (sms) if you would like to attend, so I can let you know when the location has been decided or something has changed.

See you there?

But the conference is 3 days. I’m very happy to meet you in person to talk tech or projects. There are many possibilities to talk, like during the poster-session for the industry where StreamHPC will be presented (if not everything keeps going Murphy), and a social event. I also like to share notes and be kept informed about interesting talks I am going to miss.

If (mobile) internet is available, I will tweet during the workshops.

Related Posts