1 / 120

HiPEAC :

HiPEAC Review Brussels June 25, 2010. HiPEAC :. Hi gh- P erformance and E mbedded A rchitecture and C ompilation. Granted : ICT 217068. Agenda. Agenda. Project Status Overview. Update since January , 2010. Project Status Overview. Change of project manager. WP4: Management.

verda
Download Presentation

HiPEAC :

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. HiPEAC Review Brussels June 25, 2010 HiPEAC: High- Performance and Embedded Architecture and Compilation Granted: ICT 217068

  2. Agenda Agenda

  3. Project Status Overview Update sinceJanuary, 2010 Project Status Overview

  4. Change of project manager WP4: Management Wouter De Raeve Jeroen Ongenae Transition period between June 1 and September 1

  5. Partner change WP4: Management

  6. Internships Defines topic HiPEAC Company HiPEAC student Applies to call Call for topics: Feb 3, 2010 Deadline for application: April 14, 2010 20 topics, 79 applications WP1: Mobility Company choosescandidate Approvalby SC Internship

  7. 12 Internships ARM • Improving system energy consumption through optimizing memory usage (Georgia Kouveli) • Power and performance modelling of highly scalable ARM designs (Manuel Comparetti) • Many core exploration on FPGA (Jaume Joven) • Fault-tolerant and Radiation-hardened Processor Design (MassoudMokhtarpourGhahroodi) • ScalableInterconnectionNetwork and Cache Coherenceusing AMBA bus interface (Crispin Gomez Requena) • Optimising and GeneratingOpenCL code foranEmbedded GPU Architecture (Richard Membarth and DominikGrewe) IBM • Performance tools (Vladimir Čakarević) • Parallel debugging (Lois OrosaNogueira) • Machine Learning to aidDynamicCompilation (Zheng Wang) • Split Compilation - synergy between static and dynamic analysis and compilation (Per Larsen) ST • Definition of a L1-data cache prefetch mechanism in a context of "many-core" (J. Lattore) WP1: Mobility

  8. CollaborationGrants HiPEAC student Student and host prepare a proposal HiPEAC Institution Deadline for application: May 18, 2010 34 proposals submitted (27 in 2009) 25 granted WP1: Mobility Rankingby independent reviewers Approvalby SC CollaborationGrant

  9. Collaboration grants WP1: Mobility

  10. New instrument: Conference Grants Stimulate attendance of European students at international conferences which are held in Europe Currently selected conferences: • ISCA 2010, Rennes, André Seznec (GC) • PACT 2010, Vienna, Jens Knoop (PC) • CGO 2011, Chamonix, Olivier Temam (GC) + Mike O’Boyle (PC) ISCA 2010: • Call May 7, deadline for application May 14 • 43 applications, 27 accepted • Budget: 14k€ WP1: Mobility

  11. HiPEAC Newsletter Info 23 – July 2010 (1000) Info 22 - April 2010 (837) WP3: Spreading Excellence

  12. Membership New members (sinceJanuary 2010) WP3: Spreading Excellence

  13. ACACES Summer School La Mola, Terrassa, Spain July 11-17, 2010 294 applications, 197 participants Summer school on advanced computer architecture and compilers for high-performance and embedded systems WP3: SpreadingExcellence

  14. ACACES Summer School 17 industry attendants, 8 companies 140 PhD students 59 senior attendants 31 countries 60 grants (181 requests) WP3: SpreadingExcellence

  15. ACACES 2010 courses WP3: Spreading Excellence

  16. Keynotes Insup LeeUniversity of Pennsylvania Cyber Physical Systems: The Next Computing Revolution Jesus LabartaTechnical University of Catalonia (UPC) MareIncognito: a vision towards exascale WP3: SpreadingExcellence

  17. PUMPS 2010 @ BCW @ BSC 250 applicants 100 seats

  18. HiPEAC 2011, Heraklion On schedule 86 submissions (last year: 97 submissions) 5 workshops, 2 tutorials, probably tutorials on Saturday and workshops on Sunday. Plan for a IEEE Micro special issue workshop WP3: Spreading Excellence

  19. HiPEAC booth at DATE’10 Dresden, Germany March 09-11, 2010 Over 60 booth visitors from academia and industry WP3: SpreadingExcellence

  20. HiPEAC booth at DAC’10 Anaheim, USA June 4-6, 2010 Over 73 booth visitors from academia and industry WP3: SpreadingExcellence

  21. 40 attendees WP3: Spreading Excellence

  22. Towards hipeac.pl Intend expansion of HiPEAC into new EU member states E.g. major HiPEAC meeting Oct 2009 in Wroclaw Already several Polish members (e.g. from Poznan, Krakow) Basis for future joint FP7/FP8 projects HiPEAC special session at MIXDES, Wroclaw, June 24-26 Web portal to be set up soon Contact persons: R. Leupers, Z. Chamski, Świercz • WP3: Spreading Excellence

  23. HiPEAC Awards It is necessary to improve the scientific excellence of the network by publishing more papers at top conferences. The identification of the best conferences with low numbers of European papers nd the award scheme to motivate more European paper submissions are positive steps. The review panel looks forward to the updated figures of accepted papers at the next review meeting. Recommendation 12

  24. HiPEAC Awards Financial awards (excl. Micro): • 2008: 8; 2009: 13; 2010: 9 Conference travel grants – 20 k€/year Recommendation 12

  25. Mini-Sabbaticals • It is necessary to improve the mini-sabbatical programme in order to achieve 5-10 sabbaticals per year (at the moment only 2 mini-sabbatical grants have been awarded) or to draw some conclusions if this is not possible. The mini-sabbaticals remain a weak point of the project. Only two mini-sabbaticals have been applied for. • Results from the sabbatical (Leiden to Sofia) should be presented at the formal review. Sabbaticals done: • 2009: • Prof. Rainer Leupers (Aachen) to ACE • Prof. Enrique Torre (Unizar) to U. Berkeley • 2010: • Prof. Ed Deprettere (Leiden) at Technical University Sofia, March 13 – June 12, 2010 • Prof. Stefano Crespi (Milano) at Harvard University, 2-17 June 2010 Sabbaticals approved: • 2010: • Prof. Sid Touati (University of Versailles) to TechnischeUniversitätMünchen and Leibniz Supercomputing Centre of the Bavarian Academy of Science, September 2010 • Prof. Benjamin Sahelices (Unizar) to University of Illinois at Urbana-Champaign, September 2010 • Prof. Jose Miguel-Alonso (University of Cantabria) to University of Manchester, October 2010 • [Olivier Temam (INRIA) to China] Recommendation 10/13

  26. Mini-Sabbaticals • Prof. Ed Deprettere (Univ. Leiden) at Technical University Sofia March 13 – June 12, 2010 Previous contact: the Leiden Embedded Research Center established a new joint education and research embedded systems laboratory at campus of the Technical University Sofia, called the Daedalus Laboratory. (Prof. Marin Marinov and Prof. Angel Popov) The Daedalus laboratory was officially opened on April 7, 2010. It will allow Bulgarian students at The Netherlands to come back home with a job. During the mini-sabbatical stay at Sofia, a 4-week course called “Advanced Design Methods, Techniques, and Tools for Multi-Processor Embedded Systems” has been taught and will be a regular course at Sofia starting 2011. Definition of several Masters Projects for some of the Sofia students Future joint research projects: send a proposal to one of the coming European call for projects. Recommendation 10/13

  27. HiPEAC Research results • Building a repository of the research results of the HiPEAC community is a very good idea. Progress on this should be monitored at the next review. HiPEAC technical reports are not just a repository of research results. It is an easy access tool that allows HiPEAC members to timestamp early ideas and when needed get a feedback on them. Between the pre-review and now the following can be reported: • six new technical reports have been submitted; • Plans to disseminate the technical reports outside the HiPEAC community have been worked out; • The web interface is expanded and is currently under test; • HiPEAC tech reports will be advertised widely on the following mailing lists: micro_publicity, SOCinfo, fpga-list, IEEE-TCSC, hpc-announce, Embedded_news and more; • We will keep monitoring the usage of the time stamping service and promote it within HiPEAC and outside of our network. Recommendation 9

  28. Tools Recommendation 9

  29. Recommendation 9

  30. Recommendation 9

  31. Agenda Agenda

  32. Budget & Planning WP1: Mobility WP3: Spreading Excellence WP2: Partner’s budgets WP1+2: Cluster Money Budget

  33. 5.8% 7.7% 29% underspending 15.4% Budget 71% 85% 50%

  34. WP1: Mobility Budget

  35. WP3: Spreading Excellence Budget

  36. WP2: Partner’s budget Budget

  37. Partner’s details Budget

  38. WP1+2: Cluster Money Budget

  39. Agenda Agenda

  40. Agenda Agenda

  41. EuropeanAdvancedComputingInstitute HiPEAC should investigate the feasibility of setting a permanent structure with pan-European outreach to promote research in Computing Systems via a type of "European Advanced Computing Institute". Similar structures in other research areas should be identified and analyzed. The first results of this analysis were presented at the informal review. The development of this thinking should be presented at the next formal review. Recommendation 1

  42. Situation FP8 FP7 HiPEAC1 HiPEAC2 HiPEAC? 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 HRM1 HRM2 Braga document Brussels document FP7 Call 1: 20+5 M€ FP7 Call 4: 25 M€ FP7 Call 7: 41+4 M€ FP7-8 Call ?: ??? M€

  43. Process We consulted the HiPEAC community on the future of HiPEAC. Brainstorm meeting on April 6, 2010 in Brussels Airport. Attendees were: Koen De Bosschere, Marc Duranton, Babak Falsafi, Rainer Leupers, Nacho Navarro, Mike O'Boyle, Yanos Sazeides, Per Stenström, Jürgen Teich, Mateo Valero, Sami Yehia. SC discussion on May 2 and 3, 2010 in Edinburgh. Discussion with project officer PanosTsarchopoulos on May 4, 2010. Conclusionspresented at reviewonJune 25, 2010 Recommendation 1

  44. What could a network of excellence like HiPEAC do in order to further increase the impact/visibility of our community up to the level of our American peers? What would be our most important action towards that goal?  • Let’s not always look across the Atlantic • It is today much better than 15 years ago!! • Let’s take the lead where we are stronger (embedded) • Let’s spread our own excellence first • Let’s make EU/HiPEAC a stronger brand • Let’s make EU/HiPEAC more prestigious • If we can’t beat them, let’s join them; Let’s organize events in the US • Let’s change the publication culture to a journal culture. A journal culture has more benefits for non-native speakers. Change the review system. • Let’s teach our undergraduate, graduates and PhD students how to write good papers • Let’s improve our conference. Let’s make HiPEAC no 2 conference in computer architecture. DATE is no 2 in design automation after DAC. Eurosys is on its way to become no 2 in systems. Hence, it is doable. • Let’s create visibility in existing venues • Submit to best conferences • Special issues of journals – e.g. IEEE Micro • Get in touch with key players in the US • Let’s foster excellence • There should be enough internal competition for resources • Be serious about mobility – also with the US Recommendation 1

  45. EU funding now happens through instruments like STREPs, IPs, CSAs and NoEs. Are these the right instruments to have maximal impact in our community? Do you know of other instruments that would be better suited? Could the existing instruments be optimized? If so, how?  • Current instruments are fine, except for the administrative overhead • More focused funding • More small and easy to get grants • Single PI grants (beyond ERC) • Long term investment must be in people, not in a consortium. • More money for FET-open • There is little difference between STREPS and IPs in our community. It would be better to have two types of STREPS and no more IPs. • Long term STREPs – coordinated by academics and with industry in an advisory role • Shorter term STREPs – coordinated by industry. • New instrument • Companies state research challenges • Academic groups can propose projects • Companies have an advisory role • At the end of the project, companies can have funds for 3-6 months of technology transfer • Academics should never be forced to adapt their tools to industry in a research project. • Impact requires bridging the research-product gap and this is not feasible. I would suggest some support for early development activities, i.e. during the stage that results are being evaluated by industry. There are examples of this happening in some national partners in the EU; Austria is particularly strong. However, perhaps there could be a European dimension to this. Recommendation 1

  46. One day, the funding for networks of excellence will stop. How do you see the future of the HiPEAC community after that date? Whichactivities would you definitely like to see continued, and would you be willing to pay for?  Which organizational form would be most appropriate for this new network (non-for profit organization, branch of IEEE/ACM, …)?  • No funding, no network • Conference could go on • ACACES, Journal • Cluster meetings through CSA/COST • Don’t care, other networks will emerge • Success will critically depend on volunteers Recommendation 1

  47. HiPEAC SWOT analysis • Strengths • Weaknesses • Threats • Opportunities 47

  48. HiPEAC Strengths • Strengths • Weaknesses • HiPEAC unites a large community • HiPEAC influences the European research agenda through the definition of future calls • Many new STREPS with HiPEAC members have been initiated • The HiPEAC conference is doing well • The ACACES summer school is doing well • The roadmap has impact in our community • Threats • Opportunities 48

  49. HiPEAC Weaknesses • Strengths • Weaknesses • The current cluster structure is too static and does not reflect the challenges in our domain. • It is difficult to determine the effect of all networking events in terms of actual collaborations. Collaborative research should be revitalized. • HiPEAC is missing appealing grand challenges, like Robocup, Darpa Urban Challenge, … • In comparison with other communities, we are small (especially our industrial base). • We have a weak position in the new member states. • We have not been successful in ERC so far. Even if successful, there is no guarantee that an ERC grant will tackle the HiPEAC challenges. • Threats • Opportunities 49

  50. HiPEAC Weaknesses • Strengths • Weaknesses • Link between academia and industry remains very weak in our community. • Many academic research results are not (considered) relevant for the European companies. • Companies not eager to share their tools/methodologies with academia • Companies are increasingly focusing on short term research. Universities show little interest in this kind of research. • We lack an industrial board-level network (CTO, CEO, …). • Solution? More mobility between industry and academia? Funding for technology transfer? Industry partner program? Industrial board? Listen more to industry to learn more about their needs? • Threats • Opportunities

More Related