Workshop on Concurrency in the many-Cores Era

21 Nov 2011, 09:00 → 22 Nov 2011, 18:00 US/Central

"Race Track" at Wilson Hall (WH7XO) (Fermilab, Wilson Hall)

(See below under Material: Minutes a summary of the discussion sessions including a list of "demonstrators" people volunteered to develop.)

The goals of the workshop are:
Understand the requirements and constraints of current and future experiments to efficiently exploit future many-core processor architectures. Assess the interest for a joint development of a generic software framework or a collection of services.
Determine the degree of overlap between the needs and medium to long term plans of the experiments Fermilab and CERN are supporting. Identify a list of potential technologies that would allow the effective use of upcoming hardware landscape and ensure that all the relevant solutions are being explored.
Identify commonalities and synergies that lend themselves to collaboration. Decide on communication model.
Video Conferencing will be available through EVO:
Community:  Universe
Meeting Access Information:
- Meeting URL
    http://evo.caltech.edu/evoNext/koala.jnlp?meeting=MvM2Ml2e2uDeDI929eDe9v
- Phone Bridge
    ID: 432 8690

Minutes ConcurrencyNotes-Final.pdf

Monday, 21 November

Requirements and needs from existing and upcoming experiments.

Talks by experiments/users describing the complexity of the problems and giving an idea where they could benefit from a parallel system and/or leverage many cores and illustrated by their current solutions and their limitations.

Convener: Mr Jim Kowalkowski (Fermilab)

1 Welcome

Speaker: Mr Philippe Canal (FERMILAB)

Slides ConcurrencyWorkshopIntro.pdf

2 Intensity Frontier Experiment

Speaker: Dr Brian Rebel (Fermilab)

Slides Concurency.pdf neutrinos_many_cores.pdf

3 Nova , g-2

Speaker: Dr Andrew Norman (Fermilab)

4 GaudiMT

Speaker: Charles Leggett

Slides AthenaMT.pdf

10:35 Coffee Break

Hardware technology and trends and how it is planned to be done in other domains

Talks on the upcoming hardware systems for both consumer and HPC systems.

Convener: Alfio Lazzaro

5 Parallel accelerator simulations: past, present and future

Speaker: Dr James Amundson (Fermilab)

Slides amundson-multicore.pdf

6 HPC news from SuperComputing

Speaker: Amitoj Singh

Slides ManyCoreHardwareChallenges.pdf ManyCoreHardwareChallenges.pptx

7 Update from Intel Corp.

Speaker: Chris Newburn (Intel Corp.)

12:30 Lunch

Requirements and needs from existing and upcoming experiments.

Talks by experiments/users describing the complexity of the problems and giving an idea where they could benefit from a parallel system and/or leverage many cores and illustrated by their current solutions and their limitations.

Convener: Mr Jim Kowalkowski (Fermilab)

8 CMS

Speaker: Elizabeth Sexton-Kennedy (FNAL)

Slides ConcurrencyStatusCMS.pdf

9 athenaMP

Speaker: Vakho Tsulaia

Slides AthenaMP.pdf

Evaluation of technologies, designs ideas and prototypes

Presentations by the various developing groups describing what their current plans or ideas are in the domain of multi-threading and/or to leverage many cores, including results on possible technology

Convener: David Malon (Argonne)

10 Simulation and many cores.

Speaker: Mr Philippe Canal (FERMILAB)

Slides simulation_and_many_cores_Nov_2011.pdf

15:30 Coffee Break

Evaluation of technologies, designs ideas and prototypes

Presentations by the various developing groups describing what their current plans or ideas are in the domain of multi-threading and/or to leverage many cores, including results on possible technology

Convener: David Malon (Argonne)

11 I/O worker processes using shmem

Speaker: Dr Peter van Gemmeren (ANL)

Slides AtlasMpIo.pdf AtlasMpIo.pptx

12 Sharing C++ Objects in Linux

Speaker: Roberto Vitillo

Slides sharing.pdf

Tuesday, 22 November

Evaluation of technologies, designs ideas and prototypes

Presentations by the various developing groups describing what their current plans or ideas are in the domain of multi-threading and/or to leverage many cores, including results on possible technology

Convener: David Malon (Argonne)

13 Evaluation of a virtualization based solution

Speaker: Predrag Buncic (CERN)

Slides VirtualizationAndManyCore.pdf VirtualizationAndManyCore.ppt

14 Event data processing frameworks for the future

Speaker: Markus Frank (CERN)

Slides Frameworks.Chicago.V2.pdf Frameworks.Chicago.V2.ppt

15 GO, tasks and dataflow data from athena

Speaker: Dr Sebastien Binet (IN2P3/LAL)

Slides 20111122-binet-go-fwk.pdf

10:30 Coffee Break

Evaluation of technologies, designs ideas and prototypes

Presentations by the various developing groups describing what their current plans or ideas are in the domain of multi-threading and/or to leverage many cores, including results on possible technology

Convener: David Malon (Argonne)

16 Using libdispatch While Reading a ROOT File

Speaker: Dr Christopher Jones (Fermilab)

Slides libdispatch_and_ROOT.pdf

17 Future of Art

Speaker: Dr Marc Paterno Paterno (Fermilab)

Slides mt-art-talk.pdf

18 re-engineering frameworks for concurrency

Speakers: Dr Benedikt Hegner (CERN), Dr Pere Mato (CERN)

Slides ConcurrentFramework.pdf ConcurrentFramework.pptx

12:30 Lunch

Discussion on the Future of Hardware

Convener: Alfio Lazzaro

Discussion on future development

• Overlap/commonality between experiments and experts
• Areas/ideas that are encouraging but not being explored
• Areas/common work that 2 or more groups can actively collaborate on
• Concrete action items toward exploring/prototyping ideas including names associated with them.
• Timelines.
• Communication model: how to work together, share results (meetings, web/twikis, email lists)

Desired product of the workshop: records of the conclusions, decisions and agreement reached during these discussions.

Convener: Dr Pere Mato (CERN)

19 Discussion on the Future Hardware

Speaker: Dr Alfio Lazzaro (CERN openlab)

Slides lazzaro.pdf

15:30 Coffee Break

Discussion on future development

• Overlap/commonality between experiments and experts
• Areas/ideas that are encouraging but not being explored
• Areas/common work that 2 or more groups can actively collaborate on
• Concrete action items toward exploring/prototyping ideas including names associated with them.
• Timelines.
• Communication model: how to work together, share results (meetings, web/twikis, email lists)

Desired product of the workshop: records of the conclusions, decisions and agreement reached during these discussions.

Convener: Dr Pere Mato (CERN)