Name: 15th JLESC Workshop
Start: 2023-03-21T12:00:00+01:00
End: 2023-03-23T22:00:00+01:00
Location: LaBRI

15th JLESC Workshop

from Tuesday, March 21, 2023 (12:00 PM) to Thursday, March 23, 2023 (10:00 PM)

Monday, March 20, 2023
Tuesday, March 21, 2023

12:00 PM Lunch (Salle Badiane, Haut Carré)
Lunch (Salle Badiane, Haut Carré)
12:00 PM - 1:40 PM
Room: Badiane Room

1:45 PM Welcome
Welcome
1:45 PM - 2:00 PM
Room: LaBRI Amphi
2:00 PM Keynote: Vector operations, tiled operations, distributed execution, task graphs, what next? - Samuel Thibault (INRIA)
Keynote: Vector operations, tiled operations, distributed execution, task graphs, what next?
- Samuel Thibault (INRIA)
2:00 PM - 3:00 PM
Room: LaBRI Amphi In the past decades, we have made dramatic changes in the way we express HPC computations. Vector architectures made us employ vector operations, architectures with caches made use leverage tiled kernels, and distributed systems made us express communications. More recently, notably GPU architectures made us try to embrace task-based parallelism so as to efficiently distribute work among a heterogeneous set of resources, and automatically optimize the entailed flurry of data transfers. It is thus intriguing to try to imagine what could be the next programming paradigm shift. In this talk, I will explore one the current candidates: recursive task graphs. We can indeed notice that they are currently being proposed in various task-based runtime systems. While the details differ, a lot of commonalities arise, and similar performance benefits are observed. On the application side, expressing computation recursively is also a pattern that arises commonly for expressing e.g algorithms on compressed data such as H-matrices. We will thus consider the current proposals and results, and discuss the kinds of benefits that we can expect in the long run. This can include making it easier to express complex algorithms with complex data structures, and improve the efficiency of the execution. It also opens for new perspectives in optimizing execution, which will be part of the just-starting NumPEx PEPR French project. Session chairperson: Emmanuel Jeannot
3:00 PM Break
Break
3:00 PM - 3:30 PM
3:30 PM
3:30 PM - 5:00 PM
Room: Salle Ada Lovelace
Contributions

3:30 PM Heterogeneous and reconfigurable architectures for the future of computing - Kazutomo Yoshii (Argonne National Laboratory)
3:30 PM - 5:00 PM
Room: LaBRI Amphi
Contributions

3:30 PM ExODE : Scaling the solving of Ordinary Differential Equation for Computational Biology - Jonathan Rouzaud-Cornabas (Inria / LIRIS)

3:40 PM Beyond 2D block-cyclic: extended patterns for distributed linear algebra - Lionel Eyraud-Dubois (Inria)

3:50 PM Parallel Scalable Domain Decomposition Methods in Pharmaco-Mechanical Fluid-Structure Interaction - Lea Sassmannshausen

4:00 PM SpMM, more computational intensive operation for sparse matrix in Krylov subspace methods - Atsushi Suzuki (R-CCS)

4:10 PM Batched Iterative Solvers in Plasma Fusion Simulations - Hartwig Andreas Anzt (University of Tennessee)
6:30 PM Dinner
Dinner
6:30 PM - 8:30 PM
Room: Badiane Room
6:30 PM - 7:30 PM
Contributions

6:30 PM Memory Visualization for Task-Based GEMM in PaRSEC - George Bosilca

6:30 PM Monitoring mesoscale convection simulations with nekRS using JuMonC at Scale - Mathis Bode (Forschungszentrum Jülich GmbH)

6:30 PM A Communication Module for FMSolvr - Theresa Werner

6:30 PM Memory Power Consumption on Heterogeneous Memory Systems - Andrès RUBIO PROAÑO (Riken)

6:30 PM Ginkgo — a High-Performance Portable Numerical Linear Algebra Software - Terry Cojean (Karlsruhe Institute of Technology)

6:30 PM On MPI+(task-based OpenMP) performances - Romain PEREIRA rpereira

6:30 PM libyt: a Tool for Parallel In Situ Analysis with yt - Shin-Rong Tsai (University of Illinois at Urbana-Champaign)

6:30 PM Steering Large Scale Ensemble Simulations for Online DNN Training with Adaptive Sampling - Sofya Dymchenko (INRIA and UGA)
Wednesday, March 22, 2023
9:00 AM Keynote Talk 2 - Miron Livny (University of Wisconsin)
Keynote Talk 2
- Miron Livny (University of Wisconsin)
9:00 AM - 10:00 AM
Room: LaBRI Amphi Chairperson: Franck Cappello
10:00 AM Break
Break
10:00 AM - 10:30 AM
10:30 AM
10:30 AM - 12:00 PM
Room: Salle Ada Lovelace
Contributions

10:30 AM Quantum Computing and HPC - miwako tsuji (RIKEN R-CCS)
Short Talks on Applications
Short Talks on Applications
10:30 AM - 12:00 PM
Room: LaBRI Amphi
Contributions

10:30 AM HPC challenges for ocean and atmosphere simulations - Martin Schreiber (Université Grenoble Alpes)

10:40 AM Improving IPPL: a performance portable library for grids and particles - Sriramkrishnan Muralikrishnan

10:50 AM Aevol: An experimental evolution simulator (and its mini-Apps) - Jonathan Rouzaud-Cornabas (Inria / LIRIS)

11:00 AM Extending the COMET component model to support hierarchical composite data: Aevol case study - Jerry Lacmou Zeutouo (Research Engineer at INRIA Lyon)

11:10 AM Productive Large Scale QM Calculations - William Dawson (RIKEN R-CCS)

11:20 AM Running Native HPC Applications on the Cloud - Volodymyr Kindratenko (University of Illinois at Urbana-Champaign)

11:30 AM Femtoscale Imaging of Nuclei Using High-performance Computing - Anshu Dubey (Argonne National Laboratory)
12:00 PM Lunch
Lunch
12:00 PM - 2:00 PM
Room: (both close to Forum tram stop)
2:00 PM Short Talks on Advanced Architectures
Short Talks on Advanced Architectures
2:00 PM - 3:30 PM
Room: LaBRI Amphi
Contributions

2:00 PM SciStream: Architecture and Toolkit for Streaming Data from Instruments to HPC - Rajkumar Kettimuthu (Argonne National Laboratory)

2:10 PM HomE: Enabling Homomorphic Encryption of DL, a (recently started) ERC Consolidator Grant - ANTONIO PENA (Barcelona Supercomputing Center (BSC))

2:20 PM Controlling the Energy Efficiency of HPC Nodes - A Reinforcement Learning Based Approach - Akhilesh Raj (Student Researcher at Argonne National Lab)

2:30 PM DPU Offloading with OpenMP Programming Model - MUHAMMAD USMAN (Barcelona Supercomputing Center)

2:40 PM KheOps: A Collaborative Environment for the Cost-effective Reproducibility of Edge-to-Cloud Experiments - Daniel Rosendo (INRIA)

2:50 PM Life cycle environmental impacts of HPC systems - Bill Kramer (University of Illinois)

3:00 PM Streaming hardware compressor co-design using the Chisel hardware construction language - Kazutomo Yoshii (Argonne National Laboratory)
Short Talks on Distributed Resources
Short Talks on Distributed Resources
2:00 PM - 3:30 PM
Room: Salle Ada Lovelace
Contributions

2:00 PM Towards an application-driven dynamic resource approach for HPC - Dominik Huber (Technical University of Munich (TUM))

2:10 PM Study of the folding of distributed experiments containing a distributed file system - Quentin Guilloteau

2:20 PM Dynamic resources in MPI - SERGIO ISERTE AGUT (Barcelona Supercomputing Center)

2:30 PM Optimize heterogenous storage resources use on HPC systems with simulations - Julien Monniot (INRIA)

2:40 PM Composition of Scheduling and Control-Theory Techniques - Raphael Bleuse (Univ. Grenoble Alpes, Inria)

2:50 PM Seamless Heterogeneous Memory Management Via The EcoHMEM Methodology - HATEM ELSHAZLY (Barcelona Supercomputing Center (BSC))

3:00 PM Cloud-Bursting and Autoscaling for Python-Native Scientific and AI Workflows - Tingkai Liu (University of Illinois at Urbana-Champaign)
3:30 PM Break
Break
3:30 PM - 4:00 PM
4:00 PM Project Talks on further topics
Project Talks on further topics
4:00 PM - 5:00 PM
Room: Salle Ada Lovelace
Contributions

4:00 PM Project Talk on Compression for Instruments - Robert Underwood (Argonne National Laboratory)

4:20 PM Scalable GPU-Accelerated Incremental Checkpointing of Sparsely Updated Data - Nigel Phillip Tan (University of Tennessee Knoxville)

4:40 PM Designing Flash-X, a Multiphysics Application for Exascale and Beyond - Jared O'Neal (Argonne National Laboratory)
4:00 PM - 5:30 PM
Room: LaBRI Amphi
Contributions

4:00 PM Training Deep Surrogate Models with Large Scale Online Learning - Lucas Meyer (INRIA)

4:10 PM A Feature-Driven Fixed-Ratio Lossy Compression Framework for Real-World Scientific Datasets - Sheng Di (Argonne National Laboratory)

4:20 PM JuLES: AI Super-Resolution Models for Large-Scale Simulations at Scale - Mathis Bode (Forschungszentrum Jülich GmbH)

4:30 PM High-Dimensional Performance Modeling via Tensor Completion - Edward Hutter (University of Illinois at Urbana-Champaign)

4:40 PM Malicious Peers Detection in Federated Learning - Cedric PRIGENT (INRIA)

4:50 PM Workflows for AI Model Curation and Comparison - Justin Wozniak (University of Chicago)

5:00 PM Waggle AI@Edge Computing: NSF Sage and Beyond. - Rajesh Sankaran (Argonne National Laboratory)
6:00 PM Social Event + Dinner at "La Cité du Vin"
Social Event + Dinner at "La Cité du Vin"
6:00 PM - 10:00 PM
Thursday, March 23, 2023
9:00 AM Keynote: Co-designing Self-Service Digital Twin Workflows with DIY Cluster Toolbox and DRI Leasing Federation - Sadaf Alam (University of Bristol)
Keynote: Co-designing Self-Service Digital Twin Workflows with DIY Cluster Toolbox and DRI Leasing Federation
- Sadaf Alam (University of Bristol)
9:00 AM - 10:00 AM
Room: LaBRI Amphi Per industry definition, objectives of Digital Twins (DTs) include facilitating real-time or on-demand investigation, impact studies, and recalibrating for monitoring, diagnostics and prognostics of virtual ecosystems representing real world scenarios. Efficient and secure implementation of complex workflows encompassing a wide range of experimental, observational and simulation computational and data science methods have been identified among the underpinning requirements for DTs. This talk covers two foundational building blocks for enabling complex and distributed workflows as the underlying technologies span from edge to supercomputing ecosystems. The first one is technical and the second one relates to business models. A do-it-yourself (DIY) cluster toolbox by the University of Bristol called Cluster-in-the-Cloud (CitC) has been co-designed for diverse workflows across different software defined, public cloud, IT infrastructure. In addition, these workflows rely on distributed data and compute resources, spanning from supercomputing facilities to cloud to the edge instruments. CitC enables flexible mapping and orchestration of software stacks pipelines and data-driven workflows. On the business model side, specifically for institutional, national, and internationally funded tier-n digital research infrastructure (DRI), there needs to be a resource allocation scheme managed in a federated manner across resource and service providers. This could be compared to a leasing model that can be deployed on a multi-tenant DRI. Domain-specific examples exist within global-scale science experiments and service providers consortia but not across them. A discussion on DT workflows as co-design driver use cases will be presented for realising a self-service and secure edge-cloud-supercomputing continuum.
10:00 AM Break
Break
10:00 AM - 10:30 AM
10:30 AM Break-out Session: Next-generation Numerical Linear Algebra Libraries
Break-out Session: Next-generation Numerical Linear Algebra Libraries
10:30 AM - 12:00 PM
Room: Salle Ada Lovelace
Contributions

10:30 AM BOS: Next-generation Numerical Linear Algebra Libraries - Toshiyuki Imamura (RIKEN R-CCS)
Short Talks on Interative Tools and Monitoring
Short Talks on Interative Tools and Monitoring
10:30 AM - 12:00 PM
Room: LaBRI Amphi
Contributions

10:30 AM Supercomputing in the Browser - Web-based interactive HPC-Access at JSC - Jens Henrik Goebbert (Forschungszentrum Jülich, Jülich Supercomputing Centre)

10:40 AM CharmTyles: Large-scale interactive Charm++ with Python - Aditya Bhosale (University of Illinois Urbana-Champaign)

10:50 AM Data analysis, interactive development, and the Julia Language with HPC Distributed Systems. - Aaron Saxton (NCSA)

11:00 AM Blue Waters Monitoring, Usage and Experience Data is Available - Bill Kramer (University of Illinois)

11:10 AM Advances on monitoring of supercomputers with LLview - Ilya Zhukov (Jülich Supercomputing Centre)

11:20 AM Process mapping on any topologies with TopoMatch - Emmanuel Jeannot (Inria)

11:30 AM Understanding the relation between monitoring events and topology of exascale architectures for HPC applications - Idriss Daoudi (Argonne National Laboratory)
12:00 PM Lunch either at L'Alcala or Bistrot Régent
Lunch either at L'Alcala or Bistrot Régent
12:00 PM - 2:00 PM
Room: either at L'Alcala or Bistrot Régent
2:00 PM Project Talks on I/O, Storage and Workflows
Project Talks on I/O, Storage and Workflows
2:00 PM - 3:30 PM
Room: Salle Ada Lovelace
Contributions

2:00 PM More on I/O scheduling - Frédéric Vivien (Inria)

2:20 PM Open Platforms and Reproducibility - Katarzyna Keahey (ANL)

2:40 PM LibPressio: A Unifiying Data Compression Interface for Users and Developers - Robert Underwood (Argonne National Laboratory)

3:00 PM Update on CI-HPC Project: Github2Gitlab-Integration, SSH-Gitlab-Runner - Jakob Fritz (FJZ, JSC)
2:00 PM - 3:30 PM
Room: LaBRI Amphi
Contributions

2:00 PM DASK enabled external tasks for in transit analytics with DEISA - Amal Gueroudji (French Alternative Energies and Atomic Energy Commission (CEA Sacaly))

2:10 PM Memory-Aware Scheduling of Tasks Sharing Data on Multiple GPUs with Dynamic Runtime Systems - Maxime GONTHIER (INRIA Bordeaux)

2:20 PM A Communication Module for FMSolvr - Theresa Werner

2:30 PM A task-based data-flow model for distributed and heterogeneous applications - KEVIN SALA (BARCELONA SUPERCOMPUTING CENTER (BSC))

2:40 PM Programming Heterogeneous Architectures using Hierarchical Tasks - Gwenole Lucas (Inria Bordeaux)

2:50 PM Using coroutines in a task-based runtime system - Joseph Schuchart (University of Tennessee, Knoxville)

3:00 PM Enhancing iteration performance on distributed task-based workflows - ALEX BARCELO (Barcelona Supercomputing Center)
3:30 PM Break
Break
3:30 PM - 4:00 PM
4:00 PM Project Talks on AI/ML/DL
Project Talks on AI/ML/DL
4:00 PM - 5:30 PM
Room: Salle Ada Lovelace
Contributions

4:00 PM Generating Efficient Neural Networks for Protein Diffraction Data - Georgia Wexler Channing (University of Tennessee, Knoxville)

4:20 PM Machine Learning for Predicting Flow Fields - Ando Kazutu (RIKEN Center for Computational Science)

4:40 PM Architecture and Hyperparameter Search for Super-Resolution Networks Operating on Medical Images - Xin Liu (Juelich Supercomputing Centre, Germany)

5:00 PM Leveraging Rehearsal Buffers to Enable Efficient Data-Parallel Continual Learning - Thomas Bouvier
Short Talks on Workflows, I/O and Frameworks
Short Talks on Workflows, I/O and Frameworks
4:00 PM - 5:30 PM
Room: LaBRI Amphi
Contributions

4:00 PM On Temporal I/O Behavior Characterization: Predicting I/O Phases Using Frequency Techniques - Alexis Bandet

4:10 PM CI in HPC: Working hard or hardly working? - Ivo Kabadshow (Juelich Supercomputing Centre)

4:20 PM Optimizing iterative applications using a data-flow programming model - DAVID ALVAREZ ROBERT (Barcelona Supercomputing Center)

4:30 PM Perspectives on the Versatility of a Searchable Lineage for Scalable HPC Data Management - Bogdan Nicolae (ANL)

4:40 PM On the Impact of Improving Runtime Estimates in HPC - Robin bozennec (INRIA)

4:50 PM Bringing Elasticity to HPC Storage and Data Services - Matthieu Dorier (Argonne National Laboratory)
5:30 PM Farewell
Farewell
5:30 PM - 5:45 PM
Room: LaBRI Amphi

7:00 PM Dinner at Café du Port: https://www.lecafeduport.com
Dinner at Café du Port: https://www.lecafeduport.com
7:00 PM - 10:00 PM
Room: Café du Port