Conveners
New Architectures, Portability, and Sustainability: Application Software
- Caterina Doglioni (Lund University (SE))
New Architectures, Portability, and Sustainability: Processing Frameworks
- Alessandra Forti (University of Manchester (GB))
New Architectures, Portability, and Sustainability: Validation and Accounting
- Michel Jouvin (Université Paris-Saclay (FR))
Dr
Charles Leggett
(Lawrence Berkeley National Lab (US))
5/11/20, 4:06 PM
- Different hardware available and foreseen
- Features of main APIs for heterogeneous projects.
- Projects that are looking at this topic and plans
- Experiences so far
Dr
Guilherme Amadio
(CERN)
5/11/20, 4:44 PM
We discuss attempts to develop libraries for adapting detector simulation (mainly) to new architectures, giving an overview of these codes, prospects for porting to other architectures and what has been learned in the process.
- HEP specific libraries: VecCore, VecGeom, VecMath
- Specific hardware features and their likely impact on HEP codes
Anton Poluektov
(Aix Marseille Univ, CNRS/IN2P3, CPPM, Marseille, France)
5/11/20, 5:22 PM
The use of Google TensorFlow as a general purpose parallel computing engine, using TensorFlowAnalysis as an example
Attila Krasznahorkay
(CERN),
Christopher Jones
(Fermi National Accelerator Lab. (US)), Dr
Kyle Knoepfel
(Fermi National Accelerator Laboratory)
5/12/20, 4:00 PM
How experiment frameworks are adapting to heterogeneous landscapes
Antonio Perez-Calero Yzquierdo
(Centro de Investigaciones Energéti cas Medioambientales y Tecno),
Doug Benjamin
(Argonne National Laboratory (US)),
Ricardo Brito Da Rocha
(CERN)
5/12/20, 5:00 PM
Scheduling for heterogeneous resources; multi-node slots (HPCs)
Domenico Giordano
(CERN),
Markus Schulz
(CERN),
Michele Michelotto
(Università e INFN, Padova (IT))
5/13/20, 4:00 PM
- How do we value heterogeneous resouces?
- What jobs would be used ?
- Can their efficiencies be considered the same?
Adriano Di Florio
(Universita e INFN, Bari (IT)),
Ines Ochoa
(Columbia University (US))
5/13/20, 5:00 PM
- How to validate on multiple architectures?
- What's 'the same physics'?
- How to do this practically, testing each architecture and automating