ScalAH23: 14th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Heterogeneous Systems
Novel hybrid scalable scientific algorithms are needed with the advent of
variety of novel accelerators including graphics processing units (GPUs),
field-programmable gate arrays (FPGAs) as well as with the growth of the size
of quantum computing devices and neuromorphic chips and various artificial
intelligence (AI) specific processors. This myriad of devices requires an
unified hybrid approach that allows efficient and scalable hybrid approaches
combining classical and novel computing paradigms to be implemented at scale.
These extreme-scale heterogeneous systems require novel scientific algorithms
to hide the complexity, hide network and memory latency, have advanced
communication, and have no synchronization points where possible. With the
advent of AI in the past few years the need of such scalable mathematical
methods and algorithms for such hybrid architectures that are able to handle
data and compute intensive applications at scale becomes even more important.
Scientific algorithms for multi-petaflop and exa-flop systems also need to be
fault tolerant and fault resilient, since the probability of faults increases
with scale. Resilience at the system software and at the algorithmic level is
needed as a crosscutting effort. Key science applications require novel
mathematics and mathematical models and system software that address the
scalability and resilience challenges of current- and future-generation
extreme-scale heterogeneous high performance computing (HPC) systems.
Submission Guidelines
Authors are invited to submit manuscripts in English structured as technical
papers at a length of at least 6 letter size (8.5in x 11in) pages and not
exceeding 8 pages, including figures, tables, and references using the
double-column ACM Primary Article format. Reference style files are
available at https://www.acm.org/publications/proceedings-template
(use \documentclass[sigconf,screen,final]{acmart} in the ACM
LaTeX template).
Submitted papers must represent original unpublished research that is
not currently under review for any other conference or journal. Papers not
following these guidelines will be rejected without review and further action
may be taken, including (but not limited to) notifications sent to the heads of
the institutions of the authors and sponsors of the conference. Submissions
received after the due date, exceeding length limit, or not appropriately
structured may also not be considered. Papers should be submitted electronically
at https://submissions.supercomputing.org.
All manuscripts will be peer-reviewed and judged on correctness,
originality, technical strength, and significance, quality of presentation, and
interest and relevance to the workshop attendees. Accepted papers will be
published with the ACM SC Workshop proceedings. At least one author of an
accepted paper must register for and present the paper at the workshop.
Authors may contact the workshop program chair, Christian Engelmann at
engelmannc@ornl.gov, for more
information.
Reproducibility Initiative
As part of a major initiative that aims to increase the level of reproducibility
and replicability of results, ScalAH22 invites authors of technical papers to
submit optional appendix information that can promote better reproducibility of
computational results. Submitted Artifact Description (AD) and Artifact
Evaluation (AE) appendices should follow the SC23 conference model and included
in the submitted manuscript.
Up to 4 additional letter size (8.5in x 11in) pages are permitted for
the AD and AE appendices, but should not be used to circumvent the 8-page limit
for the technical contribution. A manuscript can not be disqualified based on
information provided or not provided in the AD/AE appendices, nor if the appendix
is not available. The availability and quality of an appendix can be used in
ranking. In particular, if two manuscripts are of similar quality, the
existence and quality of the AD/AE appendices can be part of the evaluation
process.
Important Dates
- Submission opens: June 1, 2023
- Full paper submission: August 18, 2023 (firm, no extensions)
- Notification of acceptance: September 8, 2023 (firm)
- Final paper submission: September 29, 2023 (firm)
- Workshop/conference early registration: October 13, 2023
- Workshop: November 13, 2023
Topics
Topics of interest include, but are not limited to:
- Novel scientific algorithms that improve performance, scalability, resilience and power efficiency on hybrid architectures
- Porting scientific algorithms and applications to hybrid and heterogeneous architectures (with different accelerators, hybrid classical/quantum, classical/AI accelerated, etc.)
- Crosscutting approaches (system software and applications) in addressing scalability challenges on hybrid architectures
- Naturally fault tolerant, self-healing or fault oblivious scientific algorithms for hybrid architectures
- Methods and algorithms for silent data corruption with systems at scale
- Ensuring algorithms scalability over various accelerator partitions/islands, and taking advantage where the system itself has different kinds of specialized compute nodes
Workshop Chairs
- Vassil Alexandrov, Hartree Centre, Science and Technology Facilities Council, UK
- Jack Dongarra, University of Tennessee, Knoxville, USA
- Al Geist, Oak Ridge National Laboratory, USA
- Dieter Kranzlmueller, Leibniz Supercomputing Centre and Ludwig-Maximilians-University Munich, Germany
Program Committee
- Hartwig Anzt, University of Tennessee, Knoxville, USA, and Karlsruher Institute for Technology (KIT), Germany
- Rick Archibald, Oak Ridge National Laboratory, USA
- Hans-Joachim Bungartz, Technical University of Munich, Germany
- James Elliott, Sandia National Laboratories, USA
- Nahid Emad, University of Versailles SQ, France
- Wilfried Gansterer, University of Vienna, Austria
- Yasuhiro Idomura, Japan Atomic Energy Agency, Japan
- Kirk E. Jordan, IBM T.J. Watson Research, USA
- Dieter Kranzlmueller, Ludwig-Maximilians-University Munich, Germany
- Ignacio Laguna, Lawrence Livermore National Laboratory, USA
- Paul Lin, Lawrence Berkeley National Laboratory, USA
- Kengo Nakajima, RIKEN, Japan
- Ron Perrot, University of Oxford, UK
- Yves Robert, ENS Lyon, France
- Stuart Slattery, Oak Ridge National Laboratory, USA
- Valerie Taylor, Argonne National Laboratory, USA
- Keita Teranishi, Oak Ridge National Laboratory, USA
Program
This is the preliminary workshop program and subject to change.
- 09:00-10:00 Session 1
- 09:00-9:15 Welcome: Vassil Alexandrov (Hartree Centre, Science and Technology Facilities Council, UK)..
- 09:15-10:00 Invited Talk 1: "The Legacy of ECP Software Efforts, Realized and to Come,"
Michael A. Heroux (Sandia National Laboratories, USA).
(Abstract,
Bio).
- 10:00-10:30 Coffee break (coffee provided)
- 10:30-12:30 Session 2
- 10:30-11:10 Invited Talk 2: "TBD,"
Ewa Deelman (University of Southern California Information Sciences Institute, USA).
- 11:10-11:30 Paper 1: "GPU-based LU Factorization and Solve on Batches of Matrices with Band Structure,"
Ahmad Abdelfattah,
Stanimire Tomov,
Piotr Luszczek,
Hartwig Anzt,
Jack Dongarra.
- 11:30-11:50 Paper 2: "Massively Distributed Finite-Volume Flux Computation,"
Ryuichi Sai,
Mathias Jacquelin,
Francois Hamon,
Mauricio Araya-Polo,
Randolph R. Settgast.
- 11:50-12:10 Paper 3: "Parallel Symbolic Cholesky Factorization,"
Tobias Ribizel,
Hartwig Anzt.
- 12:10-12:30 Paper 4: "Advancing the distributed Multi-GPU ChASE library through algorithm optimization and NCCL library,"
Xinzhe Wu,
Edoardo Di Napoli.
- 12:30-14:00 Lunch break (lunch on your own)
- 14:00-15:00 Session 3
- 14:00-14:40 Invited Talk 3: "The Pursuit of the Brain's Ubiquitous Stochasticity,"
Brad Aimone (Sandia National Laboratories)
(Abstract,
Bio).
- 14:40-15:00 Paper 5: "Optimization of Ported CFD Kernels on Intel Data Center GPU Max 1550 using oneAPI ESIMD,"
Mohammad Zubair,
Aaron Walden,
Gabriel Nastac,
Eric Nielsen,
Christoph Bauinger,
Xiao Zhu.
- 15:00-15:30 Coffee break (coffee provided)
- 15:30-17:30 Session 4
- 15:30-16:10 Invited Talk 4: "Innovative Supercomputing by Integrations of Simulations/Data/Learning on Oakforest-PACS II,"
Kengo Nakajima (University of Tokyo)
(Abstract,
Bio).
- 16:10-16:50 Invited Talk 5: "Title,"
Kathleen Hamilton (Oak Ridge National Laboratory).
- 16:50-17:10 Paper 6: "Task-Based Polar Decomposition Using SLATE on Massively Parallel Systems with Hardware Accelerators,"
Dalal Sukkari,
Mark Gates,
Mohammed Al Farhan,
Hartwig Anzt,
Jack Dongarra.
- 17:10-17:30 Paper 7: "Moment Representation of Regularized Lattice Boltzmann Methods on NVIDIA and AMD GPUs,"
Pedro Valero-Lara,
Jeffrey Vetter,
John Gounley,
Amanda Randles.