BEGIN:VCALENDAR VERSION:2.0 PRODID:Linklings LLC BEGIN:VTIMEZONE TZID:America/Denver X-LIC-LOCATION:America/Denver BEGIN:DAYLIGHT TZOFFSETFROM:-0700 TZOFFSETTO:-0600 TZNAME:MDT DTSTART:19700308T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:-0600 TZOFFSETTO:-0700 TZNAME:MST DTSTART:19701101T020000 RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20260422T000610Z LOCATION:710 DTSTART;TZID=America/Denver:20231113T090000 DTEND;TZID=America/Denver:20231113T173000 UID:submissions.supercomputing.org_SC23_sess445@linklings.com SUMMARY:14th Workshop on Latest Advances in Scalable Algorithms for Large- Scale Heterogeneous Systems (ScalAH'23) DESCRIPTION:Invited Talk 5: Building Quantum Machine Learning for Real-Wo rld Applications\n\nQuantum machine learning is a rapidly growing field of quantum computing, and many deep learning models and methods have been ad apted into quantum analogues using gate-based or annealing-based platforms . These methods have been essential for uncovering subtleties in quantum learning dynamics, and t...\n\n\nKathleen Hamilton (Oak Ridge National Lab oratory (ORNL))\n---------------------\nScalAH'23 – Afternoon Break\n----- ----------------\nGPU-Based LU Factorization and Solve on Batches of Matri ces with Band Structure\n\nThis paper presents a portable and performance- efficient approach to solve a batch of linear systems of equations using G raphics Processing Units (GPUs). Each system is represented using a specia l type of matrices with a band structure above and/or below the diagonal. Each matrix is factorized using...\n\n\nAhmad Abdelfattah, Stanimire Tomov , Piotr Luszczek, Hartwig Anzt, and Jack Dongarra (University of Tennessee )\n---------------------\nScalAH'23 – Morning Break\n--------------------- \nParallel Symbolic Cholesky Factorization\n\nWe present a hybrid sequenti al/parallel symbolic Cholesky factorization algorithm that computes the sp arsity pattern of the symbolic factors in parallel. We evaluate the perfor mance on a large subset of the SuiteSparse matrix collection and multicore CPUs as well as flagship GPUs by AMD and NVIDIA, ...\n\n\nTobias Ribizel (Karlsruhe Institute of Technology (KIT)) and Hartwig Anzt (University of Tennessee, Innovative Computing Laboratory (ICL); Karlsruhe Institute of T echnology (KIT))\n---------------------\nTask-Based Polar Decomposition Us ing SLATE on Massively Parallel Systems with Hardware Accelerators\n\nWe i nvestigate a new task-based implementation of the polar decomposition on m assively parallel systems augmented with multiple GPUs using SLATE. We imp lement the iterative QR Dynamically-Weighted Halley (QDWH) algorithm, whos e building blocks mainly consist of compute-bound matrix operations, allow ...\n\n\nDalal Sukkari and Mark Gates (University of Tennessee, Innovative Computing Laboratory); Mohammed Al Farhan (King Abdullah University of Sc ience and Technology (KAUST)); and Hartwig Anzt and Jack Dongarra (Univers ity of Tennessee, Innovative Computing Laboratory)\n---------------------\ nInvited Talk 1: The Legacy of ECP Software Efforts, Realized, and to Com e\n\nThe US Department of Energy (DOE) Exascale Computing Project (ECP) is coming to an end. But the impact of the project is just beginning. ECP has produced dozens of GPU-enabled, scalable application codes and dozens of GPU-capable libraries and tools that underpin these applications. The experienc...\n\n\nMichael A. Heroux (Sandia National Laboratories)\n------ ---------------\nMoment Representation of Regularized Lattice Boltzmann Me thods on NVIDIA and AMD GPUs\n\nThe lattice Boltzmann method is a highly s calable Navier-Stokes solver that has been applied to flow problems in a w ide array of domains. However, the method is bandwidth-bound on modern GPU accelerators and has a large memory footprint. In this paper, we present new 2D and 3D GPU implementations of...\n\n\nPedro Valero-Lara, Jeffrey Ve tter, and John Gounley (Oak Ridge National Laboratory (ORNL)) and Amanda R andles (Duke University)\n---------------------\nMassively Distributed Fin ite-Volume Flux Computation\n\nDesigning large-scale geological carbon cap ture and storage projects and ensuring safe long-term CO2 containment - as a climate change mitigation strategy - requires fast and accurate numeric al simulations. These simulations involve solving complex PDEs governing s ubsurface fluid flow using implicit...\n\n\nRyuichi Sai (TotalEnergies EP Research & Technology US, LLC); Mathias Jacquelin (Cerebras Systems); Fran cois Hamon and Mauricio Araya-Polo (TotalEnergies EP Research & Technology US, LLC); and Randolph R. Settgast (Lawrence Livermore National Laborator y (LLNL))\n---------------------\nAdvancing the Distributed Multi-GPU ChAS E Library through Algorithm Optimization and NCCL Library\n\nAs supercompu ters become larger with powerful Graphics Processing Unit (GPU), tradition al direct eigensolvers struggle to keep up with the hardware evolution and scale efficiently due to communication and synchronization demands. Subsp ace eigensolvers, like the Chebyshev Accelerated Subspace Eigenso...\n\n\n Xinzhe Wu and Edoardo Di Napoli (Jülich Supercomputing Centre)\n---------- -----------\nWelcome\n\nVassil Alexandrov (Hartree Centre)\n-------------- -------\nInvited Talk 3: The Pursuit of the Brain’s Ubiquitous Stochastic ity\n\nOne of the most dramatic differences between the brain and modern c omputing systems is the ubiquitous stochasticity of neural circuits. The brain leverages noise in its biophysics to make its computations more powe rful and efficient, whereas today’s computers are designed, at great expen se, t...\n\n\nBrad Aimone (Sandia National Laboratories)\n---------------- -----\nInvited Talk 2: Living in a Heterogenous World – How Scientific Wo rkflows Bridge Diverse Cyberinfrastructure and What Can We Do Better?\n\nS cientific workflows are now a common tool used by domain scientists in a n umber of disciplines. They are appealing because they enable users to thi nk at high level of abstraction, composing complex applications from indiv idual application components. Workflow management systems (WMSs), such as Peg...\n\n\nEwa Deelman (University of Southern California)\n------------- --------\nInvited Talk 4: Innovative Supercomputing by Integrations of Si mulations/Data/Learning on Oakforest-PACS II\n\nRecently, supercomputing h as been changing dramatically. Integration/convergence of Simulation/Data/ Learning (S+D+L) is important towards Society 5.0 proposed by Japanese Gov ernment, which enables integration of cyber space and physical space. In 2 015, we started the BDEC project (Big Data & Extreme...\n\n\nKengo Nakajim a (University of Tokyo)\n---------------------\nOptimization of Ported CFD Kernels on Intel Data Center GPU Max 1550 Using oneAPI ESIMD\n\nWe descri be our experience porting FUN3D's CUDA-optimized kernels to Intel oneAPI S YCL. We faced several challenges, including the suboptimal performance of the oneAPI code on Intel's new data center GPU. The suboptimal performance of the oneAPI code was due to high register spills, memory latency,...\n \n\nMohammad Zubair (Old Dominion University); Aaron Walden, Gabriel Nasta c, and Eric Nielsen (NASA Langley Research Center); and Christoph Bauinger and Xiao Zhu (Intel Corporation)\n---------------------\nScalAH'23 – Lunc h Break\n\nTag: Algorithms, Heterogeneous Computing, Large Scale Systems\n \nRegistration Category: Workshop Reg Pass\n\nSession Chairs: Vassil Alexa ndrov (Hartree Centre, STFC); Jack Dongarra (University of Tennessee, Knox ville; Oak Ridge National Laboratory (ORNL)); Christian Engelmann (Oak Rid ge National Laboratory (ORNL)); Al Geist (Oak Ridge National Laboratory (O RNL)); and Dieter A. Kranzlmueller (Ludwig-Maxmilians-Universität München, Leibniz Supercomputing Centre (LRZ)) END:VEVENT END:VCALENDAR