HPC 2023
High Performance Computing
State of the Art, Emerging Disruptive Innovations and
Future Scenarios
An
International Advanced Workshop
June
26 – 30, 2023, Cetraro, Italy
|
Programme
Committee |
|||
|
LUCIO
GRANDINETTI (Chair) Department
of Computer Engineering, Electronics, and Systems University
of Calabria – UNICAL and Center of
Excellence for High Performance Computing ITALY JAMES
AHRENS Los
Alamos National Laboratory Information
Science and Technology Institute Los
Alamos, NM USA FRANK
BAETKE EOFS European
Open File System Organization formerly Hewlett
Packard Enterprise Munich GERMANY RUPAK
BISWAS NASA Exploration
Technology Directorate High End
Computing Capability Project NASA Ames
Research Center Moffet
Field, CA USA SUSAN COPPERSMITH Head
School of Physics University
of New South Wales Sydney Sydney AUSTRALIA GIUSEPPE
DE PIETRO National
Research Council of Italy Director
ICAR -Institute for High Performance Computing and Networks Naples ITALY SUDIP
DOSANJH Director National
Energy Research Scientific Computing Center Lawrence
Berkeley National Laboratory Berkeley,
CA USA WOLFGANG
GENTZSCH The
UberCloud Regensburg GERMANY and Sunnyvale,
CA USA VLADIMIR
GETOV Distributed
and Intelligent Systems Research Group School of
Computer Science and Engineering University
of Westminster London UNITED
KINGDOM VICTORIA
GOLIBER D-Wave
Systems Inc. GERMANY
and USA HIROAKI
KOBAYASHI Architecture
Laboratory Department
of Computer and Mathematical Sciences Graduate
School of information Sciences Tohoku
University JAPAN SATOSHI
MATSUOKA Director RIKEN Center for Computational Science Kobe and Department of Mathematical and Computing Sciences Tokyo Institute of Technology Tokyo JAPAN KEVIN OBENLAND Quantum Information and Integrated Nanosystems Lincoln Laboratory Massachusetts Institute of Technology MIT Boston, MA USA PER OSTER Director Advanced Computing Facility CSC-IT Center for Science Espoo FINLAND VALERIO PASCUCCI Center for Extreme Data Management, Analysis and Visualization and Scientific Computing and Imaging Institute School of Computing, University of Utah and Laboratory Fellow, Pacific Northwest National Laboratory USA KRISTEN PUDENZ Director Advanced Research Programs Atom Computing Berkeley, California USA DANIEL REED Department of Electrical and Computer Engineering School of Computing University of Utah Salt Lake City, Utah USA MARK SAFFMAN INFLEQTION Quantum Technologies and University of Wisconsin-Madison USA THOMAS STERLING President & CSO Simultac LLC Bloomington, IN formerly AI Computing Systems Laboratory (AICSL) School of Informatics, Computing, and Engineering Indiana University Bloomington, IN USA WILLIAM TANG Princeton University Dept. of Astrophysical Sciences, Princeton Plasma Physics Laboratory and Center for Statistics and Machine Learning (CSML) and Princeton Institute for Computational Science & Engineering
(PICSciE) Princeton University USA MICHELA TAUFER The University of Tennessee Electrical Engineering and Computer Science Dept. Knoxville, TN USA |
|||
|
|
|||
|
L. GRANDINETTI (Co-chair) ITALY |
T. LIPPERT (Co-chair) GERMANY |
||
|
M. ALBAALI (OMAN) |
J. DONGARRA (U.S.A.) |
W. GENTZSCH (GERMANY) |
R. Biswas (USA) |
Sponsors
|
AMAZON WEB SERVICES |
|
|
|
ATOM COMPUTING |
|
|
|
CEREBRAS |
|
|
|
|
||
|
CSC Finnish Supercomputing Center |
|
|
|
CSCS Swiss National Supercomputing Centre |
|
|
|
CSIR Council for Scientific and Industrial Research - South Africa |
|
|
|
DIRAQ |
|
|
|
EOFS |
|
|
|
Hewlett Packard Enterprise |
|
|
|
INFLEQTION Quantum Technologies |
|
|
|
Juelich Supercomputing Center, Germany |
|
|
|
LENOVO (t.b.c.) |
|
|
|
National Research Council
of Italy - ICAR - Institute for High Performance Computing and Networks |
|
|
|
NEC |
|
|
|
NEXT SILICON |
|
|
|
NVIDIA |
|
|
|
PARTEC |
|
|
|
PSIQUANTUM |
|
|
|
SAMBANOVA SYSTEMS |
|
|
|
University of Calabria Department of Computer Engineering, Electronics, and Systems |
|
|
Media Partners
|
HPCwire is a news portal and weekly newsletter covering the
fastest computers in the world and the people who run them. As the trusted
source for HPC news since 1987, HPCwire has served as the publication
of record on the issues, challenges, opportunities, and conflicts relevant to
the global High Performance Computing space. Its reporting covers the
vendors, technologies, users, and the uses of high performance, AI- and
data-intensive computing within academia, government, science, and industry. Subscribe now at www.hpcwire.com. |
|
About insideHPC Founded in 2006, insideHPC is a global publication
recognized for its comprehensive and insightful coverage across the HPC-AI
community, linking vendors, end-users and HPC strategists. insideHPC has a
large and loyal audience drawn from public and private companies of all
sizes, government agencies, research centers, industry analysts and academic
institutions. In short: the buyers and influencers of HPC, HPC-AI and
associated emerging technologies. THE News Analysis Site for HPC Insiders: Written
and edited by seasoned technology journalists, we’re all about HPC and AI,
offering feature stories, commentary, news coverage, podcasts and video
interviews with HPC’s leading voices. Like the evolving HPC ecosystem,
insideHPC’s coverage continually expands into emerging focus areas to better
serve our growing readership and advertising base, insideHPC in 2023 will
deliver an updated format and new spotlight coverage of of enterprise HPC,
HPC-AI, exascale (and post-exascale) supercomputing, quantum computing, cloud
HPC, edge computing, High Performance Data Analytics and the geopolitical
implications of supercomputing. |
|
UberCloud
provides Cloud With One Click – a fully automated, secure, on-demand,
browser-based and self-service Engineering Simulation Platform for engineers
and scientists to build, compute, and analyze their engineering simulations.
Our unique HPC software containers facilitate software packaging and
portability, simplify access and use of any public, privale, hybrid, and
multi-cloud resources, and ease software maintenance and support for
end-users, IT teams, and their cloud service providers. Please follow UberCloud
on LinkedIn and contact
us for performing a Proof of Concept in the Cloud. |
|
Speakers JAMES AHRENS Director of the Information Science
Technology Institute Los Alamos National Laboratory Los Alamos, NM USA FRANK
BAETKE EOFS European
Open File System Organization GERMANY RUPAK
BISWAS NASA Exploration
Technology Directorate High End
Computing Capability Project NASA Ames
Research Center Moffet Field, CA USA SERGIO BOIXO GOOGLE Quantum Artificial Intelligence Laboratory,
Google AI Santa Barbara, CA USA FERNANDO BRANDAO California Institute of Technology
(Caltech) and Director Quantum Applications at Amazon-AWS Los Angeles, CA USA RONALD BRIGHTWELL SANDIA National Laboratories Center for
Computing Research Albuquerque, NM USA JERRY CHOW IBM Fellow and Director of Quantum
Infrastructure IBM Quantum T. J. Watson Research Center Yorktown Heights, NY USA MARCUS DOHERTY Co-Founder & Chief Scientific Officer Quantum Brilliance and Australian National University ACT, Canberra AUSTRALIA SUDIP DOSANJH Director National Energy Research
Scientific Computing Center Lawrence Berkeley National Laboratory Berkeley, CA USA DANIELE DRAGONI Leonardo S.p.A. High Performance Computing Lab. Genova ITALY ANDREW DZURAK School of Electrical Engineering &
Telecommunications University of New South Wales Sydney, Australia and Australian Research Council and Founder&CEO of Diraq Sydney, NSW AUSTRALIA WOLFGANG GENTZSCH The UberCloud Regensburg GERMANY and Sunnyvale, CA USA VLADIMIR GETOV Distributed and Intelligent Systems
Research Group School of Computer Science and Engineering University of Westminster London UNITED KINGDOM VLAD GHEORGHIU Institute for Quantum Computing, University
of Waterloo and SoftwareQ Inc, Waterloo Waterloo, Ontario CANADA JUSTIN GING Atom Computing Berkeley, California USA ROBERT HOEKSTRA Extreme Scale Computing Computing Research Center Sandia National Laboratories Albuquerque, NM USA TOSHIYUKI
IMAMURA RIKEN Center for Computational Science Kobe JAPAN NOBUYASU ITO RIKEN Center for Computational Science Kobe JAPAN HIROAKI KOBAYASHI Architecture Laboratory Department of Computer and Mathematical
Sciences Graduate School of information Sciences Tohoku University JAPAN Technical
Director, Poznan Supercomputing and Networking Center POLAND SALVATORE MANDRA Senior Research Scientist and Task Lead Quantum Artificial Intelligence Lab (QuAIL) KBR, Inc. NASA, Ames Research Center CA, USA STEFANO MARKIDIS KTH Royal Institute of Technology Computer Science Department / Computational
Science and Technology Division Stockholm SWEDEN MARTIN MUELLER SambaNova Systems Inc Palo Alto, CA USA JOSH MUTUS Rigetti Computing Director Quantum Devices USA/CANADA YUICHI NAKAMURA Executive Professional, NEC Corporation JAPAN KEVIN OBENLAND Quantum Information and Integrated
Nanosystems Lincoln Laboratory Massachusetts Institute of Technology MIT Boston, MA USA PER OSTER Director Advanced Computing Facility CSC-IT Center for Science Espoo FINLAND VALERIO PASCUCCI Center for Extreme Data Management,
Analysis and Visualization and Scientific Computing and Imaging Institute School of Computing University of Utah, Salt Lake City and US DOE Pacific Northwest National
Laboratory USA NICOLAI PETKOV Faculty of Science and Engineering,
Intelligent Systems University of Groningen Groningen THE NETHERLANDS VALERIO RIZZO EMEA Head
of AI & Subject Matter Expert for Lenovo ITALY MARK SAFFMAN INFLEQTION Quantum Technologies and University of Wisconsin-Madison USA THOMAS SCHULTHESS CSCS Swiss National Supercomputing Centre Lugano and ETH Zurich SWITZERLAND PETE SHADBOLT Co-founder PsiQuantum Corp. Palo Alto, California USA NVIDIA Santa
Clara, CA USA THOMAS STERLING President & CSO Simultac LLC Bloomington, IN formerly AI Computing Systems Laboratory (AICSL) School of Informatics, Computing, and Engineering Indiana University Bloomington, IN USA FRED STREITZ Center for Forecasting and Outbreak
Analytics (CFA/CDC) USA and National AI Research Resource Task Force
(NAIRR-TF) USA and Lawrence Livermore National Laboratory (LLNL/DOE) Livermore, California USA SERGII STRELCHUK Department of Applied Mathematics and
Theoretical Physics and Centre for Quantum Information and Foundations University of Cambridge Cambridge UK WILLIAM TANG Princeton University Dept. of Astrophysical
Sciences Princeton Plasma Physics Laboratory and Center for Statistics and Machine Learning
(CSML) and Princeton Institute for Computational
Science & Engineering (PICSciE) Princeton University USA MICHELA TAUFER The University of Tennessee Electrical Engineering and Computer Science
Dept. Knoxville, TN USA MIWAKO TSUJI RIKEN Center for Computational Science Kobe JAPAN ERIC VAN HENSBERGEN ARM Research Austin, TX USA NATALIA VASSILIEVA Cerebras Systems Sunnyvale, CA USA ANDREW WHEELER HPE Fellow & VP Hewlett Packard Labs Fort Collins, CO USA |
Workshop
Agenda
Monday,
June 26th
|
Session |
Time |
Speaker/Activity |
|
9:45 – 10:00 |
Welcome
Address |
|
|
State of the art and
future scenarios |
||
|
|
10:00 – 10:30 |
T.
STERLING |
|
|
10:30 – 11:00 |
S. DOSANJH Towards a Unified Infrastructure
for Computation, Experimental Data Analysis and AI |
|
11:00 – 11:30 |
COFFEE BREAK |
|
|
|
11:30 – 12:00 |
A.
WHEELER |
|
12:00 – 12:30 |
R.
HOEKSTRA |
|
|
12:30 – 12:45 |
CONCLUDING
REMARKS |
|
|
|
Emerging Computer
Systems and Solutions |
|
|
|
17:00 – 17:30 |
Y.
NAKAMURA Simulated annealing at HPC VS
Quantum annealing, qualitative and quantitative analysis |
|
|
17:30 – 18:00 |
V.
GETOV New Frontiers in Energy-Efficient
Application-Architecture Co-Design of Multi-Core Processors |
|
|
18:00 – 18:30 |
E.
VAN HENSBERGEN Addressing Heterogeneity
and Disaggregation in Future Ecosystems |
|
18:30 – 19:00 |
COFFEE BREAK |
|
|
|
19:00 – 19:30 |
v.
rizzo |
|
19:30 – 19:45 |
CONCLUDING REMARKS |
Tuesday,
June 27th
|
Session |
Time |
Speaker/Activity |
|
|
Advances in HPC
Technology and Systems, Architecture and Software |
|||
|
|
9:30 – 10:00 |
F.
STREITZ |
|
|
|
10:00 – 10:30 |
J.
AHRENS To Exascale and Beyond:
Accomplishments and Challenges for Large Scale Scientific Visualization |
|
|
|
10:30 – 11:00 |
T.
IMAMURA |
|
|
|
11:00 – 11:30 |
COFFEE BREAK |
|
|
|
11:30 – 12:00 |
H.
KOBAYASHI Potential and Limitations of
Quantum Annealing as an Accelerator for Conventional HPC |
|
|
|
12:00 – 12:30 |
R.
BRIGHTWELL Evaluation of HPC Workloads
Running on Open-Source RISC-V Hardware |
|
|
|
12:30 – 12:45 |
CONCLUDING
REMARKS |
|
|
|
BIG DATA
Processing: Challenges and Perspectives |
||
|
|
17:00 – 17:30 |
V.
PASCUCCI The National Science Data
Fabric: Democratizing Data Access for Science and Society |
|
|
|
17:30 – 18:00 |
g.
shainer Addressing HPC/AI Performance
Bottlenecks with BlueFIeld Data Processing Units |
|
|
|
18:00 – 18:30 |
P.
OSTER Accelerated Computing with
EuroHPC LUMI - a Research Infrastructure for Advanced Computing |
|
|
|
18:30 – 19:00 |
COFFEE BREAK |
|
|
|
19:00 – 19:30 |
F.
BAETKE |
|
|
19:30 – 19:45 |
CONCLUDING REMARKS |
||
Wednesday,
June 28th
|
Session |
Time |
Speaker/Activity |
|
AI on HPC Platforms |
||
|
|
9:15 – 9:40 |
W.
TANG Impact of Advances in
HPC/AI/Machine Learning on Fusion Energy Development |
|
|
9:40 – 10:05 |
M.
MUELLER Advanced Use Cases of
Reconfigurable Dataflow Architecture in Science |
|
|
10:05 – 10:30 |
N.
VASSILIEVA Training Large Language
Models on Cerebras Wafer Scale Clusters |
|
|
10:30 –
10:55 |
M.
MORAES Molecular Dynamics + Machine
Learning =Deeper Insight for Drug Discovery |
|
|
10:55 – 11:25 |
COFFEE BREAK |
|
The QUANTUM COMPUTING
Promises 1 |
||
|
|
11:30 – 11:55 |
J.
GING |
|
|
11.55 – 12:20 |
A.
DZURAK |
|
|
12:20 –
12:45 |
K.
OBENLAND Developing and Analyzing
Quantum Computing Circuits for Applications in Physical Science |
|
|
12:45 – 13:00 |
CONCLUDING REMARKS |
|
|
The QUANTUM
COMPUTING Promises 2 |
|
|
|
17:00 – 17:25 |
S.
MANDRA |
|
|
17:25 – 17:50 |
S.
STRELCHUK |
|
|
17:50 – 18:15 |
V.
GHEORGHIU |
|
|
18:15 – 18:45 |
COFFEE
BREAK |
|
|
18:45 – 19:10 |
M.
DOHERTY |
|
|
19:10 – 19:35 |
M.
TSUJI |
|
|
19:35 – 19:45 |
COFFEE BREAK |
Thursday,
June 29th
|
Session |
Time |
Speaker/Activity |
|
|
The Quantum
Computing Promises 3 |
|
||
|
|
9:30 – 9:55 |
J.
CHOW |
|
|
|
9:55 – 10:20 |
S.
BOIXO |
|
|
|
10:20 – 10:45 |
J.
MUTUS |
|
|
|
10:45 – 11:10 |
M.
SAFFMAN |
|
|
|
11:10 – 11:40 |
COFFEE BREAK |
|
|
|
11:40 – 12:05 |
F.
BRANDAO |
|
|
|
12:05 – 12:30 |
P.
SHADBOLT A manufacturable platform for
fault-tolerant photonic quantum computing |
|
|
|
12:30 – 12:45 |
CONCLUDING REMARKS |
|
|
The Quantum
Computing Promises 4 |
|
||
|
|
17:00 – 17:25 |
D.
DRAGONI QUANTUM COMPUTING at Leonardo:
an industrial end-user standpoint |
|
|
|
17:25 – 17:50 |
N.
ITO |
|
|
|
17:50 – 18:15 |
K. KUROWSKI t.b.a. |
|
|
18:15 – 18:45 |
COFFEE BREAK |
|
|
|
18:45 -19:45 |
PANEL DISCUSSION “The
Intersection of Quantum Computing and HPC” Chairperson: Rupak BISWAS, NASA Ames Research Center Panelists: t.b.a. During the past several
decades, supercomputing speeds have gone from Gigaflops to Teraflops, to
Petaflops and Exaflops. As the end of Moore’s law approaches, the HPC
community is increasingly interested in disruptive technologies that could
help continue these dramatic improvements in capability. This interactive
panel will identify key technical hurdles in advancing quantum computing to
the point it becomes useful to the HPC community. Some questions to be
considered: · When will
quantum computing become part of the HPC infrastructure? · What are the
key technical challenges (hardware and software)? · What HPC
applications might be accelerated through quantum computing? |
||
|
Is the “belle
époque” of classical High Performance Computer Systems coming at the end? |
|||
Friday,
June 30th
|
Session |
Time |
Speaker/Activity |
|
Key Projects, Novel
Developments and Challenging Applications |
||
|
9:30 – 10:00 |
M.
TAUFER Building Trust in Scientific
Applications through Data Traceability and Results Explainability |
|
|
10:00 – 10:30 |
W.
GENTZSCH |
|
|
|
10:30– 11:00 |
T.
SCHULTHESS Piz Daint on Alps: a modern
day view of extreme computing and data in science |
|
11:00 – 11:30 |
COFFEE BREAK |
|
|
|
11:30 – 12:00 |
N.
PETKOV Machine learning based prediction of excess
returns of stocks |
|
|
12:00 – 12:30 |
S.
MARKIDIS Plasma-PEPSC: Enabling
Exascale Simulations for Plasma Science Grand Challenges |
|
12:30 – 12:45 |
CONCLUDING REMARKS |
Chairpersons
WOLFGANG
GENTZSCH
The
UberCloud
GERMANY
AND USA
NextSilicon
ISRAEL
Simultac LLC
USA
University of Westminster
UNITED KINGDOM
Sandia National Laboratories
USA
INFLEQTION and University of Wisconsin
USA
Architecture Laboratory
Tohoku University
JAPAN
NASA Ames Research Center
USA
KTH Royal Institute of Technology
SWEDEN
VLADIMIR
GETOV
University
of Westminster
UNITED
KINGDOM
|
“The Intersection of Quantum Computing and HPC” Thursday, June 29th 18:45 -19:45 Chairperson: Rupak BISWAS, NASA Ames Research Center Panelists: Sergio
Boixo, Jerry Chow, Nobuyasu Ito (t.b.c.), Pete Shadbolt, William Tang, Andrew
Wheeler During the past several decades, supercomputing speeds have gone from
Gigaflops to Teraflops, to Petaflops and Exaflops. As the end of Moore’s law
approaches, the HPC community is increasingly interested in disruptive
technologies that could help continue these dramatic improvements in
capability. This interactive panel will identify key technical hurdles in
advancing quantum computing to the point it becomes useful to the HPC
community. Some questions to be considered:
|
Abstracts
|
To Exascale and Beyond: Accomplishments and
Challenges for Large Scale Scientific Visualization James
Ahrens Los Alamos National
Laboratory, Los Alamos, NM, USA Short Abstract Highlighting accomplishments from exascale
visualization projects and presenting a vision of how to support visual
analysis for the evolving modern scientific process. Long Abstract Visualization plays a critical role in the
scientific understand of the massive streams of data from scientific
simulations and experiments. Continued growth in performance and availability
of large scale supercomputing resources (e.g. exascale soon and faster over
the next decade) enables both increasing simulation resolutions and an
increasing number of and breadth of simulation ensemble runs. In the modern
scientific process these simulation ensembles are verified for correctness
and then validated with experimental ensembles to increase our overall
scientific knowledge. Effective visualization of the verification and
validation (V&V) process is a significant challenge. Additional
challenges include the significant gap between supercomputer processing and
data storage speeds. In this talk, I will highlight current accomplishments
from the U.S. Exascale Computing Project to address these challenges include
high-dimensional visual analysis, comparative visualization, in situ
visualization, portable multi-threaded visualization algorithms, and
automated techniques. I will present a vision of a set of needed initiatives
to support the visual understanding of the complex and evolving modern
scientific process. Bio Dr. James Ahrens is the director of the
Information Science Technology Institute at Los Alamos National Laboratory.
He is also the Department of Energy Exascale Computing Project (ECP) Data and
Visualization lead for seven storage, data management and visualization
projects that will be a key part of a vibrant exascale supercomputing
application and software ecosystem. His research interests include
visualization, data science and parallel computing. Dr. Ahrens is author of
over 120 peer reviewed papers and the founder/design lead of ParaView, an
open-source visualization tool designed to handle extremely large data.
ParaViewis broadly used for scientific visualization and is in use at
supercomputing and scientific centers worldwide. Dr. Ahrens received his B.S.
in Computer Science from the University of Massachusetts at Amherst in 1989
and a Ph.D. in Computer Science from the University of Washington in 1996.
Dr. Ahrens is a member of the IEEE and the IEEE Computer Society. Contact him
at ahrens@lanl.gov. |
|
Open-Source for HPC and AI - The File System
Example Frank
Baetke EOFS, European Open File
System Organization, Germany This talk will address aspects of parallel
filesystems which are key components of today’s HPC infrastructures but seem
to be a bit underrepresented in discussions compared to hardware
architectures of current and future high-end systems. Three open-source filesystems used today in
production-ready HPC systems will be used as examples of open-source designs.
Development and test concepts that have been classified as the “Cathedral”
and the “Bazaar” approach will be discussed. It will also be shown to what extend we will see
either competition with proprietary offerings or cooperation potential to
cover different and sometimes overlapping requirements. The role of academic
environments and curricula with a focus on filesystems specifically and on
HPC middleware in general will also be covered. Note that this talk is not a comparison of features,
roadmaps and performance numbers of different products and offerings. |
|
Sergio Boixo Google, Quantum
Artificial Intelligence Laboratory, Google AI The Google Quantum AI
group's long-term objective is to develop a fault-tolerant universal quantum
computer. Last year, we experimentally demonstrated that quantum error
correction begins to improve performance as the number of qubits in a
scalable error-correcting code increases. At the same time, our experimental
quantum processor outperformed state-of-the-art supercomputers in one
specific benchmark. I will also review some recent scientific demonstrations. |
|
Building a Concatenated Bosonic Logical Qubit Fernando Brandao California Institute of Technology (Caltech) and
Director Quantum Applications at Amazon-AWS, Los Angeles, CA, USA I’ll discuss recent
progress towards building a logical qubit concatenating a bosonic qubit with
an outer quantum error correcting code. |
|
Evaluation of HPC Workloads Running on
Open-Source RISC-V Hardware Ronald
Brightwell SANDIA National
Laboratories Center for Computing Research, USA The emerging RISC-V ecosystem has the potential to
improve the speed, fidelity, and quality of hardware/software co-design
R&D activities. However, the suitability of the RISC-V ecosystem for
co-design targeting HPC use cases is not yet well understood. This talk
examines the performance of several HPC benchmark workloads running on
simulated open-source RISC-V cores under the FireSim FPGA-accelerated
simulation tool. To provide a realistic and reproducible HPC software stack,
we ported the Spack package manager to RISC-V and used it to build our
workloads. Our key finding is that each of the RISC-V cores evaluated can run
complex HPC workloads executing for many trillions of instructions under
simulation with rates of approximately 1/50th real-time. Additionally, we
provide a baseline set of performance results for comparison in future
studies. Our results highlight the readiness of the RISC- V ecosystem for
performing open co-design activities for HPC. We expect performance to
improve as co-design activities targeting RISC-V increase and the RISC-V
community continues to make advancements. |
|
The next wave of computing, quantum-centric
supercomputing Jerry M. Chow IBM Fellow and Director of Quantum Infrastructure,
IBM Quantum, T. J. Watson Research Center, USA The last few years have
witnessed a strong evolution in quantum computing technologies, moving from
research labs to an unprecedented access by the general public via the cloud.
Recent progress in quantum processor size, speed, and quality, have cleared
the picture towards a long- term vision in computing, where quantum
processors will play a key role in extending the computational reach of
supercomputers. In this talk I will describe how modularity will enable
scaling, and how quantum communication will increase computational capacity.
All this orchestrated by a hybrid cloud middleware for quantum with seamless
integration of classical and quantum workflows in an architectural construct
that we call quantum-centric supercomputer. |
|
Hybrid computing using a diamond quantum
computer directly integrated into a supercomputer and the pathway to massive
parallelization and hybridization Marcus Doherty Chief Scientist, Quantum Brilliance, Canberra,
Australia Quantum Brilliance is the
world’s largest diamond quantum computing company. Quantum Brilliance is
exploiting the remarkable ability for diamond qubits to operate in ambient
conditions to pursue the development of quantum accelerators: compact quantum
computers that are the same size, weight and power as CPUs/GPUs and uniquely
capable of massive parallelization in high-performance computing (HPC) and
deployment in edge computing. Quantum Brilliance is unique in already
integrating desktop-sized quantum computers into supercomputing systems, and
its release of the software development kit Qristal that is specific designed
for mass parallelization and deep hybridization of classical and quantum
computers. In this presentation, I
will first report key lessons learned from integrating quantum computers into
HPC systems as well as a demonstration of hybrid computational chemistry
using such systems, which attained the decisive goal of chemical accuracy. I
will then introduce the concept of Quantum Utility and how this informs which
applications quantum accelerators will deliver the greatest and earliest
advantage in HPC and edge computing. I will finally outline the pathway
forward for further miniaturizing diamond quantum computers, whilst
simultaneously increasing their qubit numbers and further engineering their
integration and hybridization with HPC systems. |
|
Towards a Unified Infrastructure for
Computation, Experimental Data Analysis and AI Sudip S.
Dosanjh Lawrence Berkeley
National Laboratory, USA NERSC’s mission is to accelerate scientific
discovery at the U.S. Department of Energy (DOE) Office of Science through
high performance computing and data analysis. NERSC supports the largest and
most diverse research community of any supercomputing facility within the
U.S., providing large-scale, state-of-the-art computing for unclassified
research programs in alternative energy sources, environmental science,
materials research, astrophysics and other science areas related to DOE’s
science mission. Data-intensive computing has been of growing
importance at NERSC. Considerably more data is transferred to NERSC than away
from NERSC. Experimental facilities are being inundated with data due to
advances in detectors, sensors and sequencers — in many cases these
instruments are improving at a rate even faster than Moore’s law for
semiconductors. Scientists are finding it increasingly difficult to analyze
these large scientific data sets and, as a consequence, they are often
transferring data to supercomputing centers like NERSC. Examples range from
cosmology to particle physics to biology. Berkeley Lab is partnering with
other institutions to create a Superfacility for Science through advanced
networking, the development of new supercomputing technologies and advances
in software and algorithms. The goal
is to integrate experimental and observational facilities and supercomputing
centers through the ESnet network. Supercomputers at NERSC are increasingly being designed
to complex workflows that combine computation, experimental data analysis and
AI. This presentation discusses some of the workflows driving the design of
NERSC-10 which will be deployed in 2026 and their architectural implications.
We have also started pathfinding for NERSC-11 which will be deployed in
2030+. It is possible that this system will be even more heterogenous than
previous systems and may include an array of different accelerators. |
|
QUANTUM COMPUTING at Leonardo: an industrial
end-user standpoint Daniele
Dragoni Leonardo S.p.A.,
High Performance Computing Lab, Genova, ITALY Quantum Computing is an
emerging paradigm that offers the potential to solve complex problems that
are considered intractable within the classical/digital computing domain.
Although no quantum advantage has been yet demonstrated on practical
problems, many industries have already started to investigate the potential
benefits associated with this technology in an attempt to gain competitive
advantages in their sector of reference. In this talk, I will
present the approach Leonardo is taking to assess in concrete the
potentialities and limitations associated with QC in the aerospace, security,
and defense sector. I will discuss our positioning with respect to QC from an
industrial end-user perspective, introducing examples of activities and use
cases that we are currently pursuing via combined HPC-QC methodologies as
part of a national strategy. |
|
Quantum Processing based on Silicon-CMOS
technology Andrew Dzurak UNSW, Sydney, Australia Diraq, Sydney, Australia In this talk I will
discuss the advantages and challenges facing the development of quantum
computers employing spin-based quantum processors that can be manufactured
using industry-standard silicon CMOS technology. I will begin by discussing
the development of SiMOS quantum dot qubits, including the demonstration of
high-fidelity single-qubit gates [1], the first demonstration of two-qubit
logic gate [2], and assessments of silicon qubit fidelities [3,4]. I will
then explore the technical issues related to scaling a CMOS quantum processor
[5] up to the millions of qubits that will be required for fault-tolerant QC,
including demonstrations of silicon qubit operation above one kelvin [6] and
the use of global microwave fields capable of controlling millions of qubits
[7]. References [1] M. Veldhorst et
al., Nature Nanotechnology 9, 981 (2014). [2] M. Veldhorst et
al., Nature 526, 410 (2015). [3] H. Yang et al.,
Nature Electronics 2, 151 (2019). [4] W. Huang et al.,
Nature 569, 532 (2019). [5] M. Veldhorst et
al., Nature Communications 8, 1766 (2017). [6] H. Yang et al.,
Nature 580, 350 (2020). [7] Vahapoglu et
al., Science Advances 7, eabg9158 (2021). |
|
Latest Trends and
Developments in Cloud HPC Wolfgang Gentzsch The UberCloud Regensburg, Germany and Sunnyvale,
CA, USA Market analysts like
Hyperion and Intersect360 Research predict a continuous growth of Cloud HPC
of 20% annually, while on-premise HPC is growing 6.8% over the coming years.
Another study among 740 engineers found that 24% of
respondents are using the cloud for engineering simulation today, with
another 24% planning to use it over the next 12 months. And because HPC is
now more and more entering relatively new fields such as digital twins, big
data analytics, machine learning, natural language processing, edge
computing, predictive maintenance, and more, Hyperion at ISC stated that
Cloud HPC will also benefit from this trend aynd they predicted that 2024
will be another tipping point for Cloud HPC. There are several reasons
for this accelerating trend. In our presentation, we will discuss the
following: ease of access and use of Cloud HPC resources; flexible cloud
services are more agile than on-premise hardware; energy cost for
hyperscalers is generally cheaper than for on-premise data centers; fastest
processors e.g. from AMD and NVIDIA are easier available in the cloud, and
specific processors especially for AI are only available in the cloud; high
failure rates from Do-It-Yourself can be dramatically reduces when
collaborating with cloud infrastructure and experienced cloud services providers;
and (last but not least) there is a growing number of published Cloud HPC
success stories. Finally, we will demonstrate several of these items with one
specific Cloud HPC technology which provides all these benefits. |
|
New Frontiers in Energy-Efficient
Application-Architecture Co-Design of Multi-Core Processors Vladimir
Getov Distributed and
Intelligent Systems Research Group, University of Westminster, London, U.K. Over the last two decades, further developments of
computer architecture and microprocessor hardware have been hitting the
so-called “energy wall” because of their excessive demands for more energy.
Subsequently, we have been ushering in a new era with electric power and
temperature as the primary concerns for scalable computing. Therefore,
significantly reducing the energy consumption for data processing and
movement has been the most important challenge towards achieving higher
computer performance at exascale level and beyond. This is a very difficult
and complex problem which requires revolutionary disruptive methods with a
stronger integration among hardware features, system software and
applications. Moreover, the interplay between power, temperature and
performance adds another layer of complexity to this already difficult group
of challenges. Since existing methodologies and tools are limited
by hardware capabilities and their lack of information about the application
code, a promising approach is to consider together the characteristics of
both the processor and the application-specific workload. Indeed, it is
pivotal for hardware to expose mechanisms for optimizing dynamically consumed
power and thermal energy for various workloads and for reducing data motion,
a major component of energy use. Therefore, our thermal energy model is based
on application-specific parameters such as consumed power, execution time,
and equilibrium temperature as well as hardware-specific parameters such as
half time for thermal rise or fall. As observed with both out-of-band and
in-band instrumentation and monitoring infrastructures on our experimental
cluster, the temperature changes follow a relatively slow capacitor-style
charge-discharge process. Therefore, we use the lumped thermal model that
initiates an exponential process whenever there is a change in processor’s
power consumption. In our recent work we have also been investigating
the use of barium titanate-based materials for building an intelligent thin
film thermoelectric (TFTE) converter attached to a multi-core processor with
dynamic workload management. Reviewing and comparing the specific properties
of barium titanate-based materials confirms the potential to achieve a rapid
heating-cooling cycle and therefore recover substantial wasted heat per unit
time. Building upon these initial results, the ongoing and future research
efforts involve the development of a novel tuning methodology and the
evaluation of its advantages in real use cases. Early experiments demonstrate
the efficient use of the model for analyzing and significantly improving the
application-specific balance between power, temperature, and performance. |
|
What does it take to run a quantum algorithm? Vlad Gheorghiu Institute for Quantum Computing, University of
Waterloo and SoftwareQ Inc, Waterloo, Canada Software engineers know
well that asymptotically optimal algorithms can be outperformed by alternatives
in practice; the O(n log n) time algorithm for integer multiplication is not
necessarily the best algorithm for multiplying 64-bit integers. With that in
mind: Does a known quantum algorithm outperform its classical counterpart in
practice? E.g., does Grover search outperform classical exhaustive search for
some given objective function? And if so, how much of an advantage does it
provide? A satisfactory answer will depend on future technological progress.
Nevertheless, we can begin to estimate the cost of particular quantum
circuits using current proposals for quantum architectures. In this talk I
will discuss the resources required for quantum computation using the surface
code and how to realistically estimate the 'quantum advantage' provided by a
quantum algorithm. |
|
Advances in Atomic Array Quantum Computing Justin
Ging Atom Computing, USA Scalability is key for quantum computer design and
implementation. Atom Computing's atomic array devices offer advances toward
large gate model quantum computers. We will show results from multiple atomic
array quantum processors supporting progress toward NISQ algorithm
implementation and error correction to carry us beyond the NISQ era. |
|
The Promise of Neuromorphic Computing Robert
Hoekstra Sandia National
Laboratories, USA Neuromorphic computing (NMC) is an emerging paradigm
that aims to emulate the brain’s architecture and algorithms to achieve
transformational computational capabilities at brain-like (~20 Watt)
low-power. While NMC potentially provides a path for AIs with more human-like
capabilities, the primary value to DOE in the near-to-medium term is the
promise of an extremely low-power alternative to conventional computing
approaches. NMC algorithms have recently been developed that enable efficient
and effective AI and scientific computing. The AI applications are of
particular interest in edge computing applications, where NMC can provide a
more efficient, robust and low-power alternative to conventional AI
algorithms and accelerators. In scientific computing, NMC is potentially
valuable for more diverse computing workloads and can deliver both power and
speed advantages due to its extremely parallel non-von Neumann architecture.
These include applications highly relevant to DOE’s broad scientific
computing missions including Monte Carlo sampling for solving stochastic
differential equations, graph analytics, and discrete optimization. |
|
Numerical challenging and Libraries from
Large-scale capacity computing to capability computing on Fugaku Toshiyuki
Imamura RIKEN Center for
Computational Science, Japan Fugaku is a platform that enables comprehensive and
all-encompassing support for state-of-the-art scientific computing and AI to
quantum computation, utilizing various toolchains as part of the Arm
ecosystem. The role of numerical libraries is to show shortcuts to problems
with high accuracy and speed for complex algorithms. Our team has been
developing an eigenvalue calculation library and deepening cooperation with
the materials science field. This library enables diagonalization of dense
matrices of one million dimensions using "Fugaku" and assists in
more precise analysis than approximations in some spaces for large-scale
capability computing. However, another point of view from capacity computing
is accelerated by our libraries, for example, batched eigensolver, and so on.
These technologies definitely consolidate an advanced quantum simulator
project running in our center. We are also deepening our research on
computational precision with partner organizations and promoting an approach
to cooperatively improve both computational accuracy and computation time by
skilfully utilizing mixed-precision arithmetic. In this session, we will
report on the status of our eigenvalue solver EigenExa, also highlighting
batch processing libraries, FFTE-C, various mixed-precision arithmetic
libraries, and the deepening of numerical libraries that support large-scale
applications. |
|
HPC-QC hybrid challenge on the Fugaku Nobuyasu
Ito RIKEN Center for
Computational Science, Japan Current status of the HPC-QC
hybrid activities in the R-CCS, and RIKEN Quantum activities will be
overviewed, together with a focus on QC simulation on the Fugaku. A
state-vector simulator, “braket”[1], has been developed for HPC[2,3] and it
is now tuning up to the Fugaku. It makes 40-qubits scale simulation easily
and execution elapse time is about one second or less per gate, and it will
reach to 48 qubits simulation in double precision using full nodes of the
Fugaku. As an example, estimation of ground-state energy of spin-1/2
Heisenberg chain is to be shown up to 40 spins using 41 circuits of 40
qubits, which implies exact calculation of 1641 qubits circuits. Another QC
simulator using tensor network method also be developed and it will reach
10,000 qubits simulation with moderately entangled circuit. |
|
Potential and Limitations of Quantum
Annealing as an Accelerator for Conventional HPC Hiroaki
Kobayashi Department of Computer
and Mathematical Sciences, Graduate School of information Sciences, Tohoku
University, JAPAN In this talk, I will be presenting our on-going
project entitled Quantum-Annealing Assisted Next-Generation HPC
Infrastructure. In this project, we
try to realize transparent accesses to not only classical HPC resources with
heterogeneous computing platforms such as x86 and vector accelerator, but
also Quantum Computing one in a unified fashion. In this project, I am focusing on the different
types of annealing machines, quantum annealing machines and its inspired
ones. Through the evaluation using
combinatorial clustering as a benchmark program, I will be discussing
potential and limitations of annealing machines as an accelerator for
conventional HPC infrastructure. |
|
Improved Simulations of Random Quantum
Circuits Salvatore Mandra Senior Research Scientist and Task Lead, Quantum Artificial Intelligence Lab (QuAIL), KBR, Inc., NASA, Ames Research Center, CA, USA In the past few years,
numerical techniques to classically simulate quantum circuits, in particular
random circuit sampling (RCS), have steadily improved. In my presentation, I
will present our latest RCS result [arXiv:2304.11119], with particular attention
to the numerical simulation of Sycamore-like circuits using tensor network
contraction and matrix product states. |
|
Plasma-PEPSC: Enabling Exascale Simulations
for Plasma Science Grand Challenges Stefano Markidis KTH Royal Institute of Technology, Computer
Science Department / Computational Science and Technology Division,
Stockholm, Sweden Plasma-PEPSC (Plasma
Exascale-Performance Simulation CoE) aims to bring plasma science to new
frontiers through the power of exascale computing and extreme-scale data
analytics. Our project focuses on maximizing the parallel performance and
efficiency of four flagship plasma codes—BIT, GENE, PIConGPU, and
Vlasiator—to address critical challenges in plasma science. By leveraging
algorithmic advancements in load balancing, resilience, and data compression,
along with programming model and library developments such as MPI,
accelerator and data movement APIs, and in-situ data analysis, we want to
enable unprecedented simulations on current and future exascale platforms.
Plasma-PEPSC adopts an integrated HPC software engineering approach, ensuring
the deployment, verification, and validation of extreme-scale kinetic plasma
simulations that can serve as a community standard. We employ a continuous
and integrated co-design methodology, collaborating closely with the EPI
Processor, accelerator design and development, and European quantum computing
initiatives. In this presentation, I will provide an overview of
Plasma-PEPSC, highlighting our objectives, methodologies, and anticipated
impact. I will showcase the advancements in plasma science made possible by
our project and our collaborative efforts to drive innovation and
community-wide adoption of our optimized plasma codes. |
|
Molecular Dynamics
+ Machine Learning =Deeper Insight for Drug Discovery Mark
Moraes Leader Engineering
Group, D.E. Shaw Research, USA At D. E. Shaw Research, we have designed and built
Anton, a massively parallel special-purpose architecture for molecular
dynamics simulation. Now in its third generation, our Anton 3 machines
achieve simulation speeds at least 100-fold faster than the fastest
general-purpose supercomputers on a wide range of biomolecular systems. Anton machines are an essential foundational
technology for our group’s scientific and drug discovery efforts, which we
further augment with deep learning molecular models to identify and optimize
drug candidates, and to correct systematic errors in quantum-mechanical
approximations. This talk will describe how we co-design hardware,
software and molecular models to enable both research and drug discovery. |
|
Advanced Use Cases of Reconfigurable Dataflow
Architecture in Science Martin
Mueller SambaNova Systems Inc.,
USA SambaNova Systems developed a novel approach to
process neural-network like artificial intelligence challenges of nearly
arbitrary size and at low latency. This session will briefly introduce you to
the company and its “Reconfigurable Dataflow Architecture”. It focusses on
explaining example scientific use cases and examples from real-life
customers, including very low-latency scenarios and application of large
language models to research problems. |
|
Algorithm specific resource estimates for
fault tolerant applications on superconducting qubits architectures Josh Mutus Rigetti Computing, Director Quantum Devices,
Usa/Canada I will describe in detail
what a fault tolerant quantum computer (FTQC), based on superconducting
qubits, might look like. We have developed an architectural model for such a
machine informed by the need to create a framework for benchmarking generic
quantum algorithms. By applying methodologies developed in measurement based
quantum computing, to separate algorithms into Clifford+T, we developed a
microarchitecture with specialized elements and detailed resource estimates.
The resulting resource estimates have reduced overheads compared to existing
techniques. The software tooling that accompanies this architecture allows us
to compute the space, time, energy requirements to execute FTQC algorithms,
and allows us to examine the tradeoffs between possible embodiments of the architecture. Funding
acknowledgment: “The views, opinions
and/or findings expressed are those of the author(s) and should not be
interpreted as representing the official views or policies of the Department
of Defense or the U.S. Government.
This research was developed with funding from the Defense Advanced
Research Projects Agency under Agreement HR00112230006.” |
|
Simulated annealing at HPC VS Quantum
annealing, qualitative and quantitative analysis Yuichi
Nakamura Executive Professional,
NEC Corporation, Japan A combinational optimization problem is found at a
lot of social problems to be solved. According to “no free lunch theory”,
although the best way is to be developed and applied the special exclusive
method for the target problems, we have to solve many problems and problems
are customizable by social conditions. Then, instead of the special exclusive
methods which is needed huge times to be developed, a general method which
can be obtained relatively good results for many problems, should be applied
to the target social problems. One of general methods is annealing. Annealing
can be solved many kinds of combinational optimization problems like as multi
dimension problems or etc. There are many approaches for annealing classical
and quantum based. In this talk, evaluations from various investigation
between simulated annealing at HPC VS quantum annealing is presented. |
|
Developing and Analyzing Quantum Computing
Circuits for Applications in Physical Science Kevin Obenland Quantum Information and Integrated Nanosystems,
Lincoln Laboratory, Massachusetts Institute of Technology MIT, USA Quantum computing has the
potential to fundamentally speedup the solution to numerous problems from
physical science. Examples include: quantum chemistry, electronic structure
systems, material science, and fluid/plasma dynamics. Small-scale examples of
quantum computers exist today, but machines that can provide a meaningful
impact on physical science will need to be much larger than today’s machines.
Much like classical computers, these future machines will require programs
(or circuits) to run on them. These programs will need to be efficient, easy
to construct, and applicable across numerous problem instances. In this talk
I will describe the process of constructing quantum circuits for these future
machines that will hopefully provide “quantum advantage” over classical
machines. I will describe a software library (pyLIQTR) that we are developing
at MIT Lincoln Laboratory and show how it can be used to understand the cost
of creating circuit implementations as well as the quantum computing
resources required for specific problems. |
|
Accelerated Computing with EuroHPC LUMI - a
Research Infrastructure for Advanced Computing Per Oster CSC - IT Center for Science Ltd. Since the installation of
the first phase of LUMI in September 2021 it has been a constant evolution of
the system. The second phase with installation of AMD GPUs in 2022 took the
system to a 3rd place on the TOP500. LUMI consists of a diverse
set of resources to accommodate advanced workflows including accelerated
computing with GPUs and quantum computers. The use of AMD GPUs has turned out
to be less of a hurdle than anticipated. Accelerated computing in form of
quantum computing (QC) is around the corner and LUMI has been connected to
two quantum computers in Finland and Sweden, respectively. All to
develop how QC and HPC can be
integrated and to give researchers a chance to explore the possibilities of
QC. This talk will present how LUMI is exploiting accelerated computing and
evolving into an research infrastructure for advanced computing supporting a
very diverse set of applications such as, digital twins, training of large
language models, and complex life science workflows. |
|
The National Science Data Fabric:
Democratizing Data Access for Science and Society Valerio
Pascucci John R. Parks Endowed
Chair, University of Utah, Professor, School of Computing Faculty, Scientific
Computing, and Imaging Institute, Director, Center for Extreme Data
Management Analysis and Visualization (CEDMAV), USA Effective use of data management techniques to
analyze and visualize massive scientific data is a crucial ingredient for the
success of any experimental facility, supercomputing center, or
cyberinfrastructure that supports data-intensive science. This is
particularly true for high-volume/high-velocity datasets and
resource-constrained institutions. However, universal data delivery remains
elusive, limiting the scientific impact of these facilities. This talk will present the National Science Data
Fabric (NSDF) testbed, which introduces a novel trans-disciplinary data
fabric integrating access to and use of shared storage, networking, computing,
and educational resources. The NSDF technology addresses the key data
management challenges in constructing complex streaming workflows that take
advantage of data processing opportunities that may arise while data is in
motion. This technology finds practical use in many research and industrial
applications, including materials science, precision agriculture, ecology,
climate modeling, astronomy, connectomics, and telemedicine. Practical use
cases include the real-time data acquisition from an Advanced Photon Source
(APS) beamline to allow remote users to monitor the progress of an experiment
and direct integration in the Materials Commons community repository. Full
integration with Python scripting facilitates the use of external libraries
for data processing. For example, hundreds of terabytes of climate modeling
data from NASA can be easily distributed and visualized with a Jupyter
notebook that I will demonstrate live. Overall, this leads to building flexible data
streaming workflows for massive models without compromising the interactive
nature of the exploratory process, the most effective characteristic of
discovery activities in science and engineering. The presentation will be
combined with a few live demonstrations including running Jupyter notebooks
that show (i) how hundreds of terabytes of NASA climate data from the cloud
can be easily distributed and visualized on any computer and (ii) how
undergraduate students of a minority-serving institution (UTEP) can be
provided with real-time access to large-scale materials science data normally
used only by established scientists in well-funded research groups. Bio Valerio Pascucci is the Inaugural John R.
Parks Endowed Chair, the founding Director of the Center for Extreme Data
Management Analysis and Visualization (CEDMAV), a Faculty of the Scientific
Computing and Imaging Institute, and a Professor of the School of Computing
of the University of Utah. Valerio has received the 2022 IEEE VGCT
Visualization Technical Achievement Award and the 2022-2023 Distinguished
Research Award (DRA) from the University of Utah and has been inducted into
the IEEE VGTC Visualization Academy in 2022. Valerio is also the President of ViSOAR LLC, a
University of Utah spin-off, and the founder of Data Intensive Science, a
501(c) nonprofit providing outreach and training to promote the use of
advanced technologies for science and engineering. Valerio's research
interests include Big Data management and analytics, progressive
multi-resolution techniques in scientific visualization, discrete topology,
and compression. Valerio is the coauthor of more than two hundred refereed
journal and conference papers and was an Associate Editor of the IEEE
Transactions on Visualization and Computer Graphics. |
|
How AI is unlocking the potential of the
Metaverse Valerio Rizzo EMEA Head of AI & Subject Matter Expert for
Lenovo, Italy The session will focus on
Lenovo business strategy and technological approach to the area of AI and
Metaverse. In this session we will also cover current state of art of both
technologies and each one contribute to the empowerment and enablement of the
other. Speaker bio and
LinkedIn profile: EMEA Head of AI
& Subject Matter Expert for Lenovo, Dr. Valerio Rizzo is a key member of
an expert team of Artificial Intelligence, Machine Learning and Deep Learning
specialists operating within the EMEA field sales organization and its
business development team. He is a recognized expert in the fields of
neuroscience and neurophysiology with 10 years of track record in brain
research made between Italy and USA. https://www.linkedin.com/in/valerio-rizzo-phd |
|
Circuit model quantum computing with neutral
atom arrays Mark Saffman Infleqtion, Inc., and University of
Wisconsin-Madison, USA Neutral atom arrays have
demonstrated remarkable progress in the last few years to the point
where they are a competitive platform for scalable circuit model quantum
computing. Progress on improving gate fidelities, design of multi-qubit gate
operations, low-crosstalk mid-circuit measurements, and the introduction of
neural network based signal analysis for improved performance wil be
presented. |
|
Piz Daint on Alps: a modern day view of
extreme computing and data in science Thomas Schulthess CSCS Swiss National Supercomputing Centre, Lugano
and ETH, Zurich, SWITZERLAND High Performance Computing
(HPC), i.e. scientific computing where performance matters, has been somewhat
intimidating for non-experts. With the recent surge of machine learning, HPC
technologies have found massive adoption in the commercial software world,
which in turn allows us to make better use of extreme-scale computing and
data in scientific workflows. Alps is how we call the new supercomputing
infrastructure at CSCS, on which we are providing a practical synthesis of
cloud-native and HPC, as well as AI technologies for science. We will discuss
our plans, opportunities to better deal with large-scale scientific data and
its analysis, and what we believe are the main investment that (domain)
science communities should be making now. |
|
A manufacturable platform for fault-tolerant
photonic quantum computing Pete
Shadbolt Co-founder
PsiQuantum Corp., Palo Alto, California, USA PsiQuantum is developing a
large-scale, fault-tolerant quantum computer based on integrated photonic
components – originally developed for optical networking. In this talk we
will describe a manufacturable platform for fault-tolerant photonic quantum
computing. This includes a manufacturing capability for integrated photonic
chips incorporating low-loss silicon nitride waveguides, spontaneous
single-photon sources and superconducting nanowire single-photon detectors,
as well as new results on optical switching and modulation using beyond-state-of-the-art
electro-optic thin films. We will also describe subsystem prototypes
including high-performance qubits, pseudo-number-resolving single-photon
detectors, cryogenic opto-electronic packaging, cryogenic control
electronics, high-performance qubit interconnects, and novel high-power
cryogenic cooling systems. We will describe recent performance improvements
as well as outstanding technical challenges. This talk will also cover future
directions, including an overview of the full system architecture and recent
progress on fault-tolerant algorithms and quantum applications. |
|
Addressing HPC/AI Performance Bottlenecks with
BlueFIeld Data Processing Units Gilad
Shainer NVIDIA, Santa
Clara, CA, USA AI and scientific
workloads demand ultra-fast processing of high-resolution simulations,
extreme-size datasets, and highly parallelized algorithms. As these computing
requirements continue to grow, the traditional GPU-CPU architecture further
suffers from imbalance computing, data latency and lack of parallel or
pre-data-processing. The introduction of the Data Processing Unit (DPU)
brings a new tier of computing to address these bottlenecks, and to enable,
for the first-time, compute overlapping and nearly zero communication
latency. The session will deliver a deep dive into DPU computing, and how it
can help address long lasting performance bottlenecks. Performance results of
a variety of HPC and AI applications will be presented as well. |
|
Active Memory Architecture Simultac LLC, USA The US IARPA AGILE research program was undertaken
within the last year to develop innovative HPC architectures for large-scale
graph-based applications. Six collaborative performer teams were selected
with the goal to create novel computer architectures and designs capable of
achieving orders of magnitude performance improvement for data analytics and
dynamic graph-driven computation. As part of this aggressive program,
Simultac LLC is developing the Active Memory Architecture (AMA), a radical
departure from conventional practices to circumvent legacy challenges and
exploit emerging opportunities. AMA Is a message-driven memory-centric non
von Neumann scalable architecture guided by the new HPX* distributed runtime
system based on a derivative of the shared-memory ParalleX execution model.
This presentation will describe at depth the AGILE AMA system structure,
semantics, and dynamic asynchronous operational control methods. The small
Fonton compute-cell, which is a highly replicated smart memory bank, is to be
detailed with some early A-SST simulation results to convey advances for
irregular time-varying graph processing. This research is driven by the AGILE
industry benchmarks and specified workflows and will be evaluated by the
IARPA program’s “Test and Evaluation” teams supported by DOE National
Laboratories. Questions and comments from the participants will be welcome
throughout the discussion. |
|
Fred Streitz Center for Forecasting and Outbreak Analytics
(CFA/CDC), USA The powerful combination
of high-performance computing (HPC) and Machine Learning (ML) has been
especially fruitful in the area of computational biology, where the addition
of ML techniques has helped ameliorate the lack of clear mechanistic models
and often poor statistics which has impeded progress in our understanding. I
will discuss the status of the ADMIRRAL (AI-Driven Machine-learned
Investigation of RAS-RAF Activation Lifecycle) Project, which is
investigating the behavior of an oncogenic protein in the context of a
cellular membrane. I will present our progress in the development of a novel
hybrid ML/HPC approach that exploits machine-learned latent spaces to
substantially advance molecular dynamics simulations. *This work was performed under the auspices
of the U.S. Department of Energy (DOE) by Lawrence Livermore National
Laboratory (LLNL) under Contract DE-AC52-07NA27344 and under the auspices of
the National Cancer Institute (NCI) by Frederick National Laboratory for
Cancer Research (FNLCR) under Contract 75N91019D00024. This work has been
supported by the NCI-DOE Collaboration established by the U.S. DOE and the
NCI of the National Institutes of Health. |
|
Simulating quantum circuits using efficient
tensor network contraction algorithms with subexponential upper bound Sergi Strelchuk Department of Applied Mathematics and Theoretical
Physics and Centre for Quantum Information and Foundations University of
Cambridge, UK We derive a rigorous upper
bound on the classical computation time of finite-ranged tensor network
contractions in d ≥ 2 dimensions. By means of the Sphere Separator
Theorem, we are able to take advantage of the structure of quantum circuits
to speed up contractions to show that quantum circuits of single-qubit and
finite-ranged two-qubit gates can be classically simulated in subexponential
time in the number of gates. In many practically relevant cases this beats
standard simulation schemes. Moreover, our algorithm leads to speedups of
several orders of magnitude over naive contraction schemes for
two-dimensional quantum circuits on as little as an 8 × 8 lattice. We obtain
similarly efficient contraction schemes for Google’s Sycamore-type quantum
circuits, instantaneous quantum polynomial-time circuits and non-homogeneous
(2+1)-dimensional random quantum circuits. |
|
Impact of Advances in HPC/AI/Machine Learning
on Fusion Energy Development William
Tang Princeton University, Center
for Statistics and Machine Learning (CSML) and Princeton Plasma Physics
Laboratory (PPPL), USA William Tang of Princeton University is a
Professor in Astrophysical Sciences, Participating Faculty at the Center for
Statistics and Machine Learning (CSML), Executive Committee member for the
Princeton Institute for Computational Science & Engineering (PICSciE),
and Principal Research Physicist at the Princeton Plasma Physics Laboratory
(PPPL) – where he served as Chief Scientist from 1998 to 2008. A Fellow of the American Physical Society
and recipient of awards including the 2018 NVIDIA Global Impact Award, he has
authored over 200 journal publications, is currently the Principal
Investigator (PI) of the new AURORA Exascale Early Science Project at Argonne
National Laboratory. A co-author of
the NATURE (April 2019) article on “Predicting Tokamak Disruptions Using Deep
Learning at Scale,” Prof. Tang’s Ph.D. students including recipients of the
US Presidential Early Career Award for Scientists and Engineers in 2000 and
2005. He has recently presented an
invited talk on AI/ML/HPC-enabled Digital Twins and chaired the associated
featured session at the international HPC conference (ISC-2023) in Hamburg,
Germany (May 22-25, 2023). Stimulating progress in delivering accurate
predictions in science and industry featuring the impact of HPC/AI/Machine
Learning methods are now enabling data-driven discoveries that are essential
for realizing the potential of fusion energy. As emphasized, for example, by
the 2022 US White House Summit on developing “a bold decadal vision for
commercial Fusion Energy” [1] accelerating the fusion energy development
timeline to meet the climate challenge will rely heavily on the scientific
and engineering advances being driven by HPC together with advanced
statistical methods featuring artificial intelligence/deep learning/machine
learning (AI/DL/ML). An especially time-urgent problem is the need to
reliably predict and avoid largescale “major disruptions” in MFE (magnetic
fusion energy) tokamak systems such as DIII-D in San Diego, the EUROFUSION
Joint European Torus (JET), and the international ITER device scheduled to
produce 500 MW of fusion power by the mid-30s -- while hopefully maximizing
such production. This mission requires innovation in the development of
improved data-driven and model-based approaches that maximize plasma
performance in existing experiments with impact on optimizing operational
scenarios. Encouraging advances include the deployment of
recurrent and convolutional neural networks in the Princeton's Deep Learning
Code "FRNN" that enabled the first adaptable predictive DL model
for carrying out efficient "transfer learning" accurate predictions
of disruptive events across different tokamak devices [2]. The demonstrated successful validation of
FRNN software on a huge observational FES database provides still stronger
evidence that deep learning approaches using large scale classical
supercomputers can predict disruptions with unprecedented accuracy. More
recent publications have further shown that this AI/DL capability can provide
not only the “disruption score,” as an indicator of the probability of an
imminent disruption but also a “sensitivity score” in real-time to indicate
the underlying reasons for the predicted disruption – i.e., “explainable AI.”
[3]. Moreover, detailed information for a plasma control
system (PCS) can now be provided to improve disruptions avoidance in
near-real-time to help optimize plasma performance. In particular, the application of these
AI/DL methods for real-time prediction and control has recently been further
advanced with the introduction of a surrogate model/HPC simulator
("SGTC") [4]. SGTC models satisfy compatibility requirements of a
plasma control system (PCS) and deliver inference times on order of
milliseconds (ms) and can deliver results 5 order of magnitude faster than
the validated first-principles-based global particle-in-cell GTC code runs on
advanced leadership computing HPC systems!
These capabilities are now leading to exciting avenues for moving from
passive prediction to active control and ultimately, to the optimization of
the design for a first-of-a-kind fusion pilot plant (FPP) -- aided by the
exciting introduction in a recent invited talk on AI/ML/HPC-enabled Digital
Twins at the international HPC conference (ISC-2023) in Hamburg, Germany (May
22-25, 2023) [5]. References [2] Julian Kates-Harbeck, Alexey Svyatkovskiy,
and William Tang, "Predicting Disruptive Instabilities in Controlled
Fusion Plasmas Through Deep Learning," NATURE 568, 526 (2019) [3] WilliamTang, Ge Dong, Jayson Barr, Keith
Erickson, Rory Conlin, Dan Boyer, Julian Kates-Harbeck, Kyle Felker, Cristina
Rea, N. C. Logan, et al., “Implementation of Ai/Deep Learning Disruption
Predictor into a Plasma Control System,” arXiv preprint arXiv:2204.01289,
2021; updated version with “Explainable AI/ML Focus”
in CONTRIBUTIONS TO PLASMA PHYSICS, Special Issue dedicated to Machine
Learning , accepted for publication (April, 2023) [4] Ge Dong, et al., Deep Learning-based
Surrogate Model for First-principles Global Simulations of Fusion Plasmas,
NUCLEAR FUSION 61 126061 (2021). [5] William Tang, et al., “Fusion Digital Twin
Tokamak Enabled by AI-Machine Learning,” Proceeding of the International
Supercomputing Conference, ISC-2023, Hamburg, Germany, to be published
(2023). |
|
Building Trust in Scientific Applications
through Data Traceability and Results Explainability Michela Taufer Dongarra Professor, University of Tennessee
Knoxville To trust findings in computational
science, scientists need workflows that trace the data provenance and support
results explainability. As workflows become more complex, tracing data
provenance and explaining results become more challenging. In this talk, we
propose a computational environment that automatically creates a workflow
execution’s record trail and invisibly attaches it to the workflow’s output,
enabling data traceability and results explainability. Our solution
transforms existing container technology, includes tools for automatically
annotating provenance metadata, and allows effective movement of data and
metadata across the workflow execution. We demonstrate the capabilities of
our environment with the study of SOMOSPIE, an earth science workflow. This
workflow uses machine learning modeling techniques to predict soil moisture
values from the 27 km resolution satellite data down to higher resolutions
necessary for policy-making and precision agriculture. By running the
workflow in our environment, we can identify the causes of different accuracy
measurements for predicted soil moisture values in different resolutions of
the input data and link different results to different machine learning
methods used during the soil moisture downscaling, all without requiring scientists
to know aspects of workflow design and implementation. |
|
Quantum HPC Hybrid Computing Platform toward
Cooperative Computation of Classical and Quantum Computers Miwako Tsuji RIKEN Center for Computational Science, Kobe,
JAPAN Quantum computers (QC) are
systems based on the principles of quantum theory. Quantum computers have
been expected to play an important role in the fields where classical
computers should show little growth. On the other hand, there should be
requirements of significant computational capabilities of supercomputers to
make use of quantum computers. Here, we present an
overview and plan for the quantum HPC hybrid computing platform in RIKEN R-CCS,
which is a new project in RIKEN to make use of quantum computational
technologies by the integration with high end supercomputers. We focus a programming
environment to support the cooperative computation of quantum computers and
supercomputers, which offloads some kernels in an application to quantum
computers based on their characteristics using a remote procedure call (RPC).
We also discuss the role of supercomputers in enhancing the development and
evolution of quantum computers, such as circuit optimization, circuit
cutting/knitting, and error collection/mitigation. |
|
Addressing
Heterogeneity and Disaggregation in Future Ecosystems Eric Van
Hensbergen ARM Research The quest for higher performance, energy efficiency,
and lower total cost of ownership have driven a trend towards increasing
heterogeneity in computing environments. Within the Arm ecosystem this has
materialized in heterogeneity both within system-on-chip (SoC) designs as
well as disaggregation of computing resources within data centers. Within an SoC, Arm has long supported
heterogenous environments in IoT and mobile segments, but the movement of the
supply chain towards chiplet based packaging has created opportunities within
the data center and HPC markets, creating new opportunities for domain
specialization without the dramatic increase in costs associated with producing
custom silicon. Meanwhile
disaggregated computing in terms of computational storage (CSD), smart NICs
(DPU), and CXL memory pooling provide different opportunities for placing
compute near different resources.
These hybrid computing models present new challenges in how to
effectively use them with existing application ecosystems. This talk will discuss the variety of
options available for hybrid computing models within the Arm ecosystem and
present some of the different options for how we are looking to make use of
these hybrid technologies more seamless to software. |
|
Training Large Language Models on Cerebras
Wafer Scale Clusters Natalia
Vassilieva Cerebras Systems, USA Large Language Models (LLMs) are shifting “what’s
possible”, but require massive compute and massive complexity of distributed
training across thousands of accelerators with traditional hardware. Cerebras
Wafer Scale Clusters make training LLMs faster and easier compared to GPUs
due to near-perfect linear scaling and simple data-parallel distribution
strategy for models of any size. In this talk we will share our experience
and insights from training various LLMs, including open-sourced family of
Cerebras-GPT models, on the Cerebras hardware. |
|
Andrew
Wheeler Hewlett Packard Labs,
USA Exascale computing excellence was engineered
performance by design – posing and addressing the challenge of building
supercomputers within power and cost budgets to perform a quintillion math
operations a second. In this new post-Exascale era of converged Analytics,
HPC, and AI - “workflows are the new applications”. Building supercomputers for end-to-end
workflows is going to be about engineering capability that is dynamic to (i)
availability/accessibility, (ii) efficiency (code portability, energy), and
(iii) configurability for performance. This HPE approach enables
architectural creativity to deliver flexible consumption models of
Exaflop-seconds for analytics, Exaflop-hours for HPC codes, and
Exaflop-months for AI models. This talk will cover the technical vision, the
challenges, and focus areas of research and development. |
