US ATLAS HL-LHC computing discussion

[Torre Wenaus <wenaus AT gmail.com>]

Hi All,

Some updates on the agenda...

https://indico.cern.ch/event/742090/

We will have short welcoming talks from CSI (Deputy Director Frank Alexander) and US ATLAS (US ATLAS Operations Manager Srini Rajagopalan) giving the motivations and background for the workshop. Frank will then get the workshop started with an introduction to CSI.

We then have a series of talks and discussion periods that introduce the computing and software domains of ATLAS, and the challenges they present for adapting to new architectures and the HL-LHC. Guidance on the talks is that they should open with an overview of the domain for a non-specialist audience, and then in the latter part of the talk drill down to the status and potential for reengineering (MT, GPU utilization, applying ML, …). Purpose of the talks is to share understanding of the software, physics objectives, characteristics of existing apps, performance analysis, exascale porting potential.

Then we'll hear the perspectives of an experienced ATLAS physicist Peter Loch, bringing an analysis user perspective and a knowledge of the challenges presented by prospective porting targets like fast simulation; and ATLAS physicist Amir Farbin with extensive experience in ML and GPU utilization who will bring us interesting and very relevant ideas and proposals. 

We then have a series of CSI software and technology talks and surrounding discussion, being organized by Nick D'Imperio of CSI. He's arranging a series of talks with plenty of discussion time.

Eric has suggested that this CSI block should come earlier, on the first day, thoughts on that? My thinking was that with the ATLAS talks first, the CSI material could be tuned to the needs and opportunities emerging from the ATLAS talks.

Comments appreciated!

  Torre

On Wed, Jul 11, 2018 at 11:17 AM Torre Wenaus <wenaus AT gmail.com> wrote:

Hi,

I've fleshed out the agenda for the workshop

https://indico.cern.ch/event/742090/

following the draft in the googledoc.

Comments and further content are welcome

  Torre

On Wed, Jul 4, 2018 at 2:26 PM Torre Wenaus <wenaus AT gmail.com> wrote:

Hi all,

Welcome to the US ATLAS HL-LHC mailing list. The first thing we'll use it for is planning the BNL CSI workshop July 25-27 2018. Please invite people to join the list, it's open.

Our planning googledoc for the workshop is here… (editing is on)

https://docs.google.com/document/d/1DQM-8h7eFSyvGIh23aFphNv5V35DnSZYv4NjSfiL9H0/edit?usp=sharing

I've added in the doc a strawman draft agenda based on our discussions, please comment here or in the document:

• Wed: the landscape, the problems, discussions on potential help and collaboration

• As discussed on June 14, a series of talks on specific aspects of ATLAS software: describing the software domain, describing the status and potential for reengineering (MT, GPU utilization, applying ML, …), plus generous discussion time. Share understanding of the codes, physics objectives, characteristics of existing apps, performance analysis, porting potential

• Wed am  9-1

• CSI intro  25+10’

• ATLAS computing challenges towards HL-LHC - Torre Wenaus  25+10’

• Tracking software - Ed Moyse, Paolo Calafiura  30+30’

• Calorimeter software - Walter Lampl  30+30’

Wed pm  2-6

• Simulation software (fast and full) - Heather Gray  40+40’

• I/O and persistency software - David Malon  30+30’

• Workflow and dataflow software & issues - Alexei Klimentov, Kaushik De, Torre Wenaus  40’+40’

• (Event generation?)

Thu: zeroing in on specific proposals, activities and collaborations we can pursue. Perspectives on the problems

• Thu am  9-12:30

• Core software, perspectives on reengineering - Scott Snyder, Vakho Tsulaia  30+30’

• Perspectives of a user and analyst - Peter Loch  30+30’

• Proposals - Amir Farbin  30’+30’

Thu lunch - bring in sandwiches for hackathon lunch

Thu pm 2-6

• CSI software & technology talks, discussion, demo, hacking

• trying out a few new tools, look at programming models

Discussion on participation in the GPU hackathon at BNL in Sep  60’

First hackathon session

Fri: hands-on hackathon & brainstorming

• ATLAS software build & instrumentation hackathon

• Software reengineering prospects brainstorm

• what software components and workflows are most amenable to attack for exascale

• potential for the codes on different architectures, look for porting problems to attack

• looking at the codes with rewrite potential in mind, develop strategies for porting, if we can do sufficient homework beforehand.

Next steps, what to pursue and with what priorities

  Torre