BEGIN:VCALENDAR VERSION:2.0 PRODID:Linklings LLC BEGIN:VTIMEZONE TZID:America/New_York X-LIC-LOCATION:America/New_York BEGIN:DAYLIGHT TZOFFSETFROM:-0500 TZOFFSETTO:-0400 TZNAME:EDT DTSTART:19700308T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:-0400 TZOFFSETTO:-0500 TZNAME:EST DTSTART:19701101T020000 RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20210402T160555Z LOCATION:Track 8 DTSTART;TZID=America/New_York:20201112T100000 DTEND;TZID=America/New_York:20201112T183000 UID:submissions.supercomputing.org_SC20_sess214@linklings.com SUMMARY:11th Workshop on Latest Advances in Scalable Algorithms for Large- Scale Systems DESCRIPTION:Workshop\n\nReplacing Pivoting in Distributed Gaussian Elimina tion with Randomized Techniques\n\nLindquist, Luszczek, Dongarra\n\nGaussi an elimination is a key technique for solving\ndense, non-symmetric system s of linear equations. Pivoting is\nused to ensure numerical stability but can introduce significant\noverheads. We propose replacing pivoting with recursive butterfly\ntransforms (RBTs) and iterative refinement. RBTs use\ nan...\n\n---------------------\nA Survey of Singular Value Decomposition Methods for Distributed Tall/Skinny Data\n\nSchmidt\n\nThe Singular Value Decomposition (SVD) is one of the most important matrix \nfactorizations, enjoying a wide variety of applications across numerous \napplication dom ains. In statistics and data analysis, the common applications of \nSVD su ch as Principal Components Analysis (PCA) and linear regression...\n\n---- -----------------\nPerformance Analysis of a Quantum Monte Carlo Applicati on on Multiple Hardware Architectures Using the HPX Runtime\n\nWei, Chatte rjee, Huck, Hernandez, Kaiser\n\nThis paper describes how we successfully used the HPX programming model to port the DCA++ application on multiple a rchitectures that include POWER9, x86, ARM v8, and NVIDIA GPUs. We describ e the lessons we can learn from this experience as well as the benefits of enabling the HPX in the application ...\n\n---------------------\nRecursi ve Basic Linear Algebra Operations on TensorCore GPU\n\nZhang, Karihaloo, Wu\n\nEncouraged by the requirement of high speed matrix computations and training deep neural networks, TensorCore was introduced in NVIDIA GPU\nto further accelerate matrix-matrix multiplication. It supports very fast ha lf precision general matrix matrix multiplications (GEMMs), which is aroun d 8x faster...\n\n---------------------\nAn Integer Arithmetic-Based Spars e Linear Solver Using a GMRES Method and Iterative Refinement\n\nIwashita, Suzuki, Fukaya\n\nIn this paper, we develop a (preconditioned) GMRES solv er based on integer arithmetic, and introduce an iterative refinement fram ework for the solver. We describe the data format for the coefficient matr ix and vectors for the solver that is based on integer or fixed-point numb ers.\nTo avoid overflow ...\n\n---------------------\nHigh-Order Finite El ement Method Using Standard and Device-Level Batch GEMM on GPUs\n\nBeams, Abdelfattah, Tomov, Dongarra, Kolev...\n\nWe present new GPU implementatio ns of the tensor contractions arising from basis-related computations for high-order finite element methods. We consider both tensor and non-tensor bases. In the case of tensor bases, we introduce new kernels based on\na series of fused device-level matrix multiplicat...\n\n-------------------- -\nImplementation and Numerical Techniques for One Eflop/s HPL-AI Benchmar k on Fugaku\n\nImamura, Kudo, Nitadori, Ina\n\nOur performance benchmark o f HPL-AI on the supercomputer Fugaku was awarded the 55th Top500. The effe ctive performance was 1.42 EFlop/s, and the world's first achievement to e xceed the wall of exascale in a floating-point arithmetic benchmark. Becau se HPL-AI is brand new and has no reference code fo...\n\n---------------- -----\nRevisiting Exponential Integrator Methods for HPC with a Mini-Appli cation\n\nShanks\n\nIn this work we look at employing communication-avoidi ng techniques commonly used in Krylov methods in the context of exponentia l integrators for the solution of stiff partial differential equations. We choose an exponential integrator method based on polynomial approximation s, as compared to those ...\n\n---------------------\nScalA – Introduction : 11th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems\n\nAlexandrov, Dongarra, Geist, Engelmann\n\nNovel scalable scient ific algorithms are needed to enable key science applications to exploit t he computational power of large-scale systems. These extreme-scale algori thms need to hide network and memory latency, have very high computation/c ommunication overlap and minimal communication and have n...\n\n---------- -----------\nTwo-Stage Asynchronous Iterative Solvers for Multi-GPU Cluste rs\n\nNayak, Cojean, Anzt\n\nGiven the trend of supercomputers accumulatin g much of their compute power in \nGPU accelerators composed of thousand s of cores and operating in streaming\nmode, global synchronization poin ts become a bottleneck, severely confining \nthe performance of applicat ions. In consequence, asynchronous m...\n\n---------------------\nScalA – Closing\n\nAlexandrov\n\n---------------------\nScalA – Break\n\n\n\n----- ----------------\nScalA – Break\n\n\n\n---------------------\nKeynote 3: E CP – Recent Experiences in Porting Complex Applications to Accelerator-Bas ed Systems\n\nSiegel\n\nhe U.S. Department of Energy's Exascale Computing Project (ECP) represents a broad effort to enable mission critical science and engineering on next generation HPC systems. As part of this, ECP incl udes 24 application development teams spanning a broad range of science an d engineering domains. The t...\n\n---------------------\nA Fast Scalable Iterative Implicit Solver with Green's Function-Based Neural Networks\n\nI chimura, Fujita, Hori, Maddegedara, Ueda...\n\nBased on the Green's functi ons that reflect mathematical properties of partial differential equations (PDE), we developed a novel preconditioner using neural networks (NNs) wi th high accuracy and small computational cost for improving the convergenc e property of an iterative implicit solver. As the ...\n\n---------------- -----\nScalA – Break\n\n\n\n---------------------\nScalA – Keynote: Perfor mance Evaluation of the Supercomputer "Fugaku" and A64FX Manycore Processo r\n\nSato\n\nWe have been carrying out the FLAGSHIP 2020 to develop the Ja panese next-generation flagship supercomputer, Post-K, named “Fugaku”. In the project, we have designed a new Arm-SVE enabled processor, called A64F X, as well as the system, including interconnect, with the industry partne r, Fujitsu. The p...\n\n---------------------\nScalA – Keynote: High Perfo rmance Data Analytics and Some Applications\n\nEmad\n\nIn most areas of sc ience, data production is now faster than compute capabilities. The comput ational modeling and data analysis associated with high-performance comput ing techniques are used to make these huge amounts of data effectively tal k. In this talk, we highlight some challenges in the ecosys...\n\n\nTag: A lgorithms, Extreme Scale Computing, Performance/Productivity Measurement a nd Evaluation, Scalable Computing, Scientific Computing\n\nRegistration Ca tegory: Workshop Reg Pass END:VEVENT END:VCALENDAR