BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260522T150153Z
LOCATION:C2/3/4 Ballroom
DTSTART;TZID=America/Chicago:20181114T083000
DTEND;TZID=America/Chicago:20181114T170000
UID:submissions.supercomputing.org_SC18_sess323@linklings.com
SUMMARY:Research Posters
DESCRIPTION:SC18 Research Posters will be on display on Tuesday, Wednesday
 , Thursday from 8:30am to 5pm in the C2/3/4 Ballroom.\n\nPortable Parallel
  Performance via Multi-Dimensional Homomorphisms\n\nAchieving portable per
 formance over different parallel architectures and varying problem sizes i
 s hard: e.g., a program optimized for multi-core CPUs on large input sizes
  can significantly differ from the same program optimized for Graphics Pro
 cessing Units (GPUs) on small sizes.\n\nWe propose an appr...\n\n\nAri Ras
 ch, Richard Schulze, and Sergei Gorlatch (University of Münster)\n--------
 -------------\nConvolutional Neural Networks for Coronary Plaque Classific
 ation in Intravascular Optical Coherence Tomography (IVOCT) Images\n\nCurr
 ently, IVOCT is the only imaging technique with the resolution necessary t
 o identify vulnerable thin cap fibro-atheromas (TCFAs). IVOCT also has gre
 ater penetration depth in calcified plaques as compared to Intravascular U
 ltrasound (IVUS). Despite its advantages, IVOCT image interpretation is ch
 ...\n\n\nChaitanya Kolluru, David Prabhu, Yanzan Gharaibeh, David Wilson, 
 and Sanjaya Gajurel (Case Western Reserve University)\n-------------------
 --\nKernel-Based and Total Performance Analysis of CGYRO on 4 Leadership S
 ystems\n\nWe present the results of an exhaustive performance analysis of 
 the CGYRO code on 4 leadership systems spanning 5 different configurations
  (2 KNL-based, 1 Skylake-based, and 2 hybrid CPU-GPU architectures). CGYRO
  is an Eulerian gyrokinetic solver designed and optimized for collisional,
  electromagnet...\n\n\nIgor Sfiligoi, Jeff Candy, and Emily Belli (General
  Atomics)\n---------------------\nSol: Transparent Neural Network Accelera
 tion Platform\n\nWith the usage of neural networks in a wide range of appl
 ication fields, the necessity to execute these efficiently on high perform
 ance hardware is one of the key problems for artificial intelligence (AI) 
 framework providers. More and more new specialized hardware types and corr
 esponding libraries a...\n\n\nNicolas Weber (NEC Laboratories Europe, NEC 
 Corporation)\n---------------------\nHIVE: A Cross-Platform, Modular Visua
 lization Ecosystem for Heterogeneous Computational Environments\n\nHPC ope
 rational environments usually have supporting computational systems for as
 sisting pre- and post-processing activities such as the visualization and 
 analysis of simulation results. A wide variety of hardware systems can be 
 found at different HPC sites, and in our case, we have a  CPU-only (x86...
 \n\n\nJorji Nonaka (Riken Center for Computational Science); Kenji Ono (Ky
 ushu University, RIKEN); Naohisa Sakamoto and Kengo Hayashi (Kobe Universi
 ty, RIKEN); Tomohiro Kawanabe and Fumiyoshi Shoji (Riken Center for Comput
 ational Science); Masahiro Fujita (LTE Inc); Kentaro Oku (Kashika Inc); an
 d Kazuma Hatta (Imagica Digitalscape)\n---------------------\nMPI/OpenMP p
 arallelization of the Fragment Molecular Orbitals Method in GAMESS\n\nIn t
 his work, we present a novel parallelization strategy for the Fragment Mol
 ecular Orbital (FMO) method in the quantum chemistry package GAMESS. The o
 riginal FMO code has been parallelized only with MPI, which limits scalabi
 lity of the code on multi-core massively parallel machines. To address thi
 ...\n\n\nVladimir Mironov (Lomonosov Moscow State University), Yuri Alexee
 v (Argonne National Laboratory), and Dmitri Fedorov (National Institute of
  Advanced Industrial Science and Technology (AIST))\n---------------------
 \nCross-Layer Group Regularization for Deep Neural Network Pruning\n\nImpr
 oving weights sparsity is a common strategy for deep neural network prunin
 g. Most existing methods use regularizations that only consider structural
  sparsity within an individual layer. In this paper, we propose a cross-la
 yer group regularization taking into account the statistics from multiple 
 ...\n\n\nShuang Gao and Xin Liu (Nvidia Corporation)\n--------------------
 -\nA Parallel-Efficient GPU Package for Multiphase Flow in Realistic Nano-
 Pore Networks\n\nSimulations of fluid flow in oil/gas shale rocks are chal
 lenging in part due to the heterogeneous pore sizes ranging from a few nan
 ometers to a few micrometers. Additionally, the complex fluid-solid intera
 ction occurring physically and chemically must be captured with high resol
 ution. To address the...\n\n\nYidong Xia (Idaho National Laboratory); Anse
 l Blumers (Brown University, Idaho National Laboratory); Zhen Li (Brown Un
 iversity); Lixiang Luo (IBM); Jan Goral (University of Utah); Matthew Andr
 ew (Carl Zeiss X-ray Microscopy Inc); Joshua Kane (Idaho National Laborato
 ry); and Yu-Hang Tang (Lawrence Berkeley National Laboratory)\n-----------
 ----------\nThe NAStJA Framework: Non-Collective Scalable Global Communica
 tions\n\nIn recent years, simulations in various areas of science and engi
 neering have proven to be very useful.  To efficiently deploy simulation c
 odes on current and future high-performance computer systems, high node le
 vel performance, scalable communication and the exclusion of unnecessary c
 alculations a...\n\n\nMarco Berghoff and Ivan Kondov (Karlsruhe Institute 
 of Technology)\n---------------------\nMGRIT Preconditioned Krylov Subspac
 e Method\n\nMGRIT re-discretize the problem with larger time-step width at
  the coarse-levels, which often cause unstable convergence. We propose a K
 rylov subspace method with MGRIT preconditioning as a more stable solver. 
 For unstable problems, MGRIT preconditioned Krylov subspace method perform
 ed better than M...\n\n\nRyo Yoda, Akihiro Fujii, and Teruo Tanaka (Kogaku
 in University)\n---------------------\nGPU Acceleration at Scale with Open
 Power Platforms in Code_Saturne\n\nCode_Saturne is a widely used computati
 onal fluid dynamics software package that uses finite-volume methods to si
 mulate different kinds of flows tailored to tackle multi-bilion-cell unstr
 uctured mesh simulations. This class of codes has shown to be challenging 
 to accelerate on GPUs as they consist o...\n\n\nSamuel Antao (IBM); Charle
 s Moulinec (Science and Technology Facilities Council, UK); Yvan Fournier 
 (EDF Research and Development); Robert Sawko, Malgorzata Zimon, and Christ
 opher Thompson (IBM); Alex Skillen (Science and Technology Facilities Coun
 cil, UK); Juan Uribe (EDF Research and Development); and David Emerson (Sc
 ience and Technology Facilities Council, UK)\n---------------------\nMachi
 ne Learning for Adaptive Discretization in Massive Multiscale Biomedical M
 odeling\n\nFor multiscale problems, traditional time stepping algorithms u
 se a single smallest time stepsize in order to capture the finest details;
  using this scale leads to a significant waste of computing resources for 
 simulating coarse-grained portion of the problem. To improve computing eff
 iciency for mul...\n\n\nChangnian Han, Prachi Gupta, Peng Zhang, Danny Blu
 estein, and Yuefan Deng (Stony Brook University)\n---------------------\nP
 erformance Evaluation of the Shifted Cholesky QR Algorithm for Ill-Conditi
 oned Matrices\n\nThe Cholesky QR algorithm, which computes the QR factoriz
 ation of a matrix, is a simple yet efficient algorithm for high-performanc
 e computing. However it suffers from numerical instability. In a recent wo
 rk, this instability has been remedied by repeating Cholesky QR twice (Cho
 leskyQR2).  ChokeskyQ...\n\n\nTakeshi Fukaya (Hokkaido University); Ramase
 shan Kannan (Arup UK); Yuji Nakatsukasa (National Institute of Informatics
 , Japan); Yusaku Yamamoto (University of Electro-Communications, Japan); a
 nd Yuka Yanagisawa (Waseda University)\n---------------------\nCapsule Net
 works for Protein Structure Classification\n\nCapsule Networks have great 
 potential to tackle problems in structural biology because of their attent
 ion to hierarchical relationships. This work describes the implementation 
 and application of a capsule network architecture to the classification of
  RAS protein family structures on GPU-based comput...\n\n\nDan A. Rosa de 
 Jesus, Julian Cuevas Paniagua, and Wilson Rivera (Lawrence Berkeley Nation
 al Laboratory, University of Puerto Rico at Mayaguez) and Silvia Crivelli 
 (Lawrence Berkeley National Laboratory)\n---------------------\nEnabling H
 igh-Level Graph Processing via Dynamic Tasking\n\nData-intensive computing
  yields irregular and unbalanced workloads, in particular on large-scale p
 roblems running on distributed systems. Task-based runtime systems are com
 monly exploited to implement higher-level data-centric programming models,
  promoting multithreading and asynchronous coordinatio...\n\n\nMaurizio Dr
 occo, Vito Giovanni Castellana, Marco Minutoli, Antonino Tumeo, and John F
 eo (Pacific Northwest National Laboratory)\n---------------------\nVeloC: 
 Very Low Overhead Checkpointing System\n\nCheckpointing large amounts of r
 elated data concurrently to stable storage is a common I/O pattern of many
  HPC applications. However, such a pattern frequently leads to I/O bottlen
 ecks that lead to poor scalability and performance. As modern HPC infrastr
 uctures continue to evolve, there is a growing...\n\n\nBogdan Nicolae and 
 Franck Cappello (Argonne National Laboratory) and Adam Moody, Elsa Gonsior
 owski, and Kathryn Mohror (Lawrence Livermore National Laboratory)\n------
 ---------------\nAn Alternative Approach to Teaching Bigdata and Cloud Com
 puting Topics at CS Undergraduate Level\n\nBig data and cloud computing co
 llectively offer a paradigm shift in the way businesses are now acquiring,
  using and managing information technology. This creates the need for ever
 y CS student to be equipped with foundation knowledge in this collective p
 aradigm and to possess some hands-on-experience...\n\n\nDebzani Deb, Muzta
 ba Fuad, and Keith Irwin (Winston-Salem State University)\n---------------
 ------\nApplying the Execution-Cache-Memory Model: Current State of Practi
 ce\n\nThe ECM (Execution-Cache-Memory) model is an analytic, resource-base
 d  performance model for steady-state loop code running on multicore proce
 ssors. Starting from a machine model, which describes the interaction betw
 een the code and the hardware, and static code analysis, it allows an accu
 rate predi...\n\n\nGeorg Hager, Jan Eitzinger, and Julian Hornich (Univers
 ity of Erlangen-Nuremberg, Erlangen Regional Computing Center); Francesco 
 Cremonesi (Swiss Federal Institute of Technology in Lausanne); Christie L.
  Alappat (University of Erlangen-Nuremberg, Erlangen Regional Computing Ce
 nter); Thoams Roehl (University of Erlangen-Nuremberg); and Gerhard Wellei
 n (University of Erlangen-Nuremberg, Erlangen Regional Computing Center)\n
 ---------------------\nToward Smoothing Data Movement Between RAM and Stor
 age\n\nWe propose to design and implement a software framework, which prov
 ides a Multilayer Buffer System (MBS) to cache in/out datasets into CPU ma
 in memory from/to slower storage media, such as parallel file systems (e.g
 ., Lustre), solid-state drive (e.g., Burst Buffer) or non-volatile RAM. Al
 though MBS ...\n\n\nTariq Alturkestani (King Abdullah University of Scienc
 e and Technology), Thierry Tonellot and Vincent Etienne (Saudi Aramco), an
 d Hatem Ltaief (King Abdullah University of Science and Technology)\n-----
 ----------------\nGPGPU Performance Estimation with Core and Memory Freque
 ncy Scaling\n\nGraphics processing units (GPUs) support dynamic voltage an
 d frequency scaling to balance computational performance and energy consum
 ption. However, simple and accurate performance estimation for a given GPU
  kernel under different frequency settings is still lacking for real hardw
 are, which is impor...\n\n\nQiang Wang and Xiaowen Chu (Hong Kong Baptist 
 University)\n---------------------\nDistributed Adaptive Radix Tree for Ef
 ficient Metadata Search on HPC Systems\n\nAffix-based search allows users 
 to retrieve data without the need to remember all relevant information pre
 cisely. While building an inverted index to facilitate efficient affix-bas
 ed search is a common practice for standalone databases and desktop file s
 ystems, they are often insufficient for high-p...\n\n\nWei Zhang (Texas Te
 ch University), Houjun Tang and Suren Byna (Lawrence Berkeley National Lab
 oratory), and Yong Chen (Texas Tech University)\n---------------------\nIm
 proving Error-Bounded Lossy Compression for Cosmological N-Body Simulation
 \n\nCosmological simulations may produce extremely large amount of data, s
 uch that its successful run depends on large storage capacity and huge I/O
  bandwidth, especially in the exascale computing scale. Effective error-bo
 unded lossy compressors with both high compression ratios and low data dis
 tortion ...\n\n\nSihuan Li (University of California, Riverside); Sheng Di
  (Argonne National Laboratory); Xin Liang and Zizhong Chen (University of 
 California, Riverside); and Franck Cappello (Argonne National Laboratory)\
 n---------------------\nTensor-Optimized Hardware Accelerates Fused Discon
 tinuous Galerkin Simulations\n\nIn recent years the compute/memory balance
  of processors has been continuously shifting towards compute. The rise of
  Deep Learning, based on matrix multiplications, accelerated this path, es
 pecially in terms of single precision and lower precision compute. An impo
 rtant research question is if this d...\n\n\nAlexander Breuer (University 
 of California, San Diego); Alexander Heinecke (Intel Corporation); and Yif
 eng Cui (San Diego Supercomputer Center)\n---------------------\nEnergy Ef
 ficiency of Reconfigurable Caches on FPGAs\n\nThe performance of a given c
 ache architecture depends largely on the applications that run on it. Even
  though each application has its best-suited cache configuration, vendors 
 of fixed HPC systems must provide compromise designs. Reconfigurable cache
 s can adjust cache configuration dynamically to ge...\n\n\nTianqi Wang (Bo
 ston University), Ang Li (Pacific Northwest National Laboratory), and Tong
  Geng and Martin Herbordt (Boston University)\n---------------------\nSpot
 SDC: an Information Visualization System to Analyze Silent Data Corruption
 \n\nAggressive technology scaling trends are expected to make the hardware
  of HPC systems more susceptible to transient faults. Transient faults in 
 hardware may be masked without affecting the program output, cause a progr
 am to crash, or lead to silent data corruptions (SDC). While fault injecti
 on studi...\n\n\nZhimin Li (University of Utah), Harshitha Menon (Lawrence
  Livermore National Laboratory), Yarden Livnat (University of Utah), Kathr
 yn Mohror (Lawrence Livermore National Laboratory), and Valerio Pascucci (
 University of Utah)\n---------------------\nTop-Down Performance Analysis 
 of Workflow Applications\n\nScientific simulation frameworks are common to
  use on HPC systems. They contain parallelized algorithms and provide vari
 ous solvers for a specific application domain. Usually, engineers execute 
 multiple steps to solve a particular problem which are often distributed o
 ver multiple jobs. Finding perfo...\n\n\nChristian Herold and Bill William
 s (Technical University Dresden)\n---------------------\nUsing Darshan and
  CODES to Evaluate Application I/O Performance\n\nBurst buffers have becom
 e increasingly popular in HPC systems, allowing bursty I/O traffic to be s
 erviced faster without slowing down application execution. The ubiquity of
  burst buffers creates opportunities for studying their ideal placement in
  the HPC topology. Furthermore, the topology of the ne...\n\n\nHarsh Kheta
 wat (North Carolina State University), Christopher Zimmer (Oak Ridge Natio
 nal Laboratory), Frank Mueller (North Carolina State University), and Sudh
 arshan Vazhkudai and Scott Atchley (Oak Ridge National Laboratory)\n------
 ---------------\nImplementing Efficient Data Compression and Encryption in
  a Persistent Key-Value Store for HPC\n\nRecently, persistent data structu
 res, like key-value stores (KVSs), which are stored in an HPC system's non
 volatile memory, provide an attractive solution for a number of emerging c
 hallenges like limited I/O performance. This paper investigates how to eff
 iciently integrate data compression and encry...\n\n\nJungwon Kim and Jeff
 rey S. Vetter (Oak Ridge National Laboratory)\n---------------------\nRepr
 oducibility as Side Effect\n\nThe ability to keep records and reproduce ex
 periments is a critical element of the scientific method for any disciplin
 e. However, the recording and publishing of research artifacts that allow 
 to reproduce and directly compare against existing research continue to be
  a challenge. In this paper, we pr...\n\n\nShu Wang, Zhuo Zhen, and Jason 
 Anderson (University of Chicago) and Kate Keahey (Argonne National Laborat
 ory, University of Chicago)\n---------------------\nCharacterizing Declust
 ered Software RAID for Enhancing Storage Reliability and Performance\n\nRe
 dundant array of independent disks (RAID) has been widely used to address 
 the reliability issue in storage systems. As the scale of modern storage s
 ystems continues growing, disk failure becomes the norm. With ever-increas
 ing disk capacity, RAID recovery based on disk rebuild becomes more costly
 , ...\n\n\nZhi Qiao and Song Fu (University of North Texas) and Hsing-bung
  Chen and Bradley Settlemyer (Los Alamos National Laboratory)\n-----------
 ----------\nJob Simulation for Large-Scale PBS-Based Clusters with the Mau
 i Scheduler\n\nFor large-scale High Performance Computing centers with a w
 ide range of different projects and heterogeneous infrastructures, efficie
 ncy is an important consideration. Understanding how compute jobs are sche
 duled is necessary for improving the job scheduling strategies in order to
  optimize cluster u...\n\n\nGeorg Zitzlsberer, Branislav Jansik, and Jan M
 artinovic (IT4Innovations, Czech Republic; Technical University of Ostrava
 , Czech Republic)\n---------------------\nEnabling Neutrino and Antineutri
 no Appearance Observation Measurements with HPC Facilities\n\nWhen fitting
  to data with low statistics and near physical boundaries, extra measures 
 need to be taken to ensure proper statistical coverage. The method NOvA us
 es is called the Feldman-Cousins procedure, which entails fitting thousand
 s of independent pseudoexperiments to generate acceptance interval...\n\n\
 nNorm Buchanan and Steven Calvez (Colorado State University); Pengfei Ding
  (Fermi National Accelerator Laboratory); Derek Doyle (Colorado State Univ
 ersity); Alex Himmel, Burt Holzman, Jim Kowalkowski, and Andrew Norman (Fe
 rmi National Accelerator Laboratory); Alex Sousa (University of Cincinnati
 ); and Marc Paterno, Saba Sehrish, Brandon White, and Christopher Green (F
 ermi National Accelerator Laboratory)\n---------------------\nDeepSim-HiPA
 C: Deep Learning High Performance Approximate Calculation for Interactive 
 Design and Prototyping\n\nWe present a data-driven technique that can lear
 n from physical-based simulations for the instant prediction of field dist
 ribution for 3D objects. Such techniques are extremely useful when conside
 ring, for example, computer aided engineering (CAE), where computationally
  expensive simulations are oft...\n\n\nAhmed Al-Jarro and Serban Georgescu
  (Fujitsu Laboratories of Europe Ltd.) and Yasumoto Tomita and Kouta Nakas
 hima (Fujitsu Laboratories Ltd)\n---------------------\nAn Efficient SIMD 
 Implementation of Pseudo-Verlet Lists for Neighbor Interactions in Particl
 e-Based Codes\n\nIn particle-based simulations, neighbour finding (i.e. fi
 nding pairs of particles to interact within a given range) is the most tim
 e consuming part of the computation. One of the best such algorithms, whic
 h can be used for both Molecular Dynamics (MD) and Smoothed Particle Hydro
 dynamics (SPH) simula...\n\n\nJames Willis (Durham University, Institute f
 or Computational Cosmology); Matthieu Schaller (Leiden Observatory); and P
 edro Gonnet (Google LLC)\n---------------------\nDevelopment of Numerical 
 Coupled Analysis Method by Air Flow Analysis and Snow Accretion Analysis\n
 \nIn this research, to take countermeasures for the snow accretion damage,
  we developed a simulator of realizing the snow accretion process in the f
 ollowing steps. Firstly, air flow analysis is performed by “Airflow simula
 tor” developed by RTRI (Railway Technical Research Institute). Secondly, t
 raject...\n\n\nKohei Murotani, Koji Nakade, Yasushi Kamata, and Daisuke Ta
 kahashi (Railway Technical Research Institute, Japan)\n-------------------
 --\nFeatherCNN: Fast Inference Computation with TensorGEMM on ARM Architec
 tures\n\nThis poster presents a fast inference computation library for ARM
  architecture named as CNNForward. CNNForward is trying to improve the eff
 iciency of inference computation for convolutional neural networks on ARM-
 based multi-core and many-core architectures using both mathematical formu
 la reconstruc...\n\n\nHaidong Lan (Shandong University), Jintao Meng (Tenc
 ent Holdings Ltd), Christian Hundt and Bertil Schmidt (Johannes Gutenberg 
 University Mainz), Minwen Deng (Tencent Holdings Ltd), Weiguo Liu (Shandon
 g University), and Yanjie Wei and Shengzhong Feng (Shenzhen Institutes of 
 Advanced Technology)\n---------------------\nFast and Accurate Training of
  an AI Radiologist\n\nThe health care industry is expected to be an early 
 adopter of AI and deep learning to improve patient outcomes, reduce costs,
  and speed up diagnosis. We have developed models for using AI to diagnose
  pneumonia, emphysema, and other thoracic pathologies from chest x-rays. U
 sing the Stanford Universi...\n\n\nLucas A. Wilson, Vineet Gundecha, Srini
 vas Varadharajan, Alex Filby, Pei Yang, and Quy Ta (Dell EMC); Valeriu Cod
 reanu and Damian Podareanu (SURFsara); and Vikram Saletore (Intel Corporat
 ion)\n---------------------\nInteractive HPC Deep Learning with Jupyter No
 tebooks\n\nDeep learning researchers are increasingly using Jupyter notebo
 oks to implement interactive, reproducible workflows. Such solutions are t
 ypically deployed on small-scale (e.g. single server) computing systems. H
 owever, as the sizes and complexities of datasets and associated neural ne
 twork models in...\n\n\nWahid Bhimji, Steven Farrell, Oliver Evans, Matthe
 w Henderson, and Shreyas Cholia (Lawrence Berkeley National Laboratory); A
 aron Vose (Cray Inc); and Mr Prabhat, Rollin Thomas, and Richard Shane Can
 on (Lawrence Berkeley National Laboratory)\n---------------------\nA Massi
 vely Parallel Evolutionary Markov Chain Monte Carlo Algorithm for Sampling
  Complicated Multimodal State SpacesState\n\nWe develop an Evolutionary Ma
 rkov Chain Monte Carlo (EMCMC) algorithm for sampling from large multi-mod
 al state spaces. Our algorithm combines the advantages of evolutionary alg
 orithms (EAs) as optimization heuristics and the theoretical convergence p
 roperties of Markov Chain Monte Carlo (MCMC) algo...\n\n\nWendy K. Tam Cho
  and Yan Liu (University of Illinois)\n---------------------\nPerformance 
 Evaluation of the NVIDIA Tesla V100: Block Level Pipelining vs. Kernel Lev
 el Pipelining\n\nAs accelerators become more common, expressive and perfor
 mant, interfaces for them become ever more important. Programming models l
 ike OpenMP offer simple-to-use but powerful directive-based offload mechan
 isms. By default, these models naively copy data to or from the device wit
 hout overlapping comp...\n\n\nXuewen Cui (Virginia Tech), Thomas R. W. Sco
 gland and Bronis R. de Supinski (Lawrence Livermore National Laboratory), 
 and Wu Feng (Virginia Tech)\n---------------------\nFull State Quantum Cir
 cuit Simulation by Using Lossy Data Compression\n\nIn order to evaluate, v
 alidate, and refine the design of a new quantum algorithm or a quantum com
 puter, researchers and developers need methods to assess their correctness
  and fidelity. This requires the capabilities of simulation for full quant
 um state amplitudes. However, the number of quantum sta...\n\n\nXin-Chuan 
 Wu (University of Chicago, Argonne National Laboratory); Sheng Di, Franck 
 Cappello, Hal Finkel, and Yuri Alexeev (Argonne National Laboratory); and 
 Frederic T. Chong (University of Chicago)\n---------------------\nOptimizi
 ng Next Generation Hydrodynamics Code for Exascale Systems\n\nStudying con
 tinuum dynamics problems computationally can illuminate complex physical p
 henomena where experimentation is too costly. However, the models used in 
 studying these phenomena usually require intensive calculations, some of w
 hich are beyond even the largest supercomputers to date. Emerging ...\n\n\
 nDana Akhmetova (KTH Royal Institute of Technology); Sumathi Lakshmirangan
 atha (University of Wyoming); Diptajyoti Mukherjee (Allegheny College); Fr
 ederick Oullet (University of Florida); Patrick Payne (Los Alamos National
  Laboratory); Nicholas Stegmeier (South Dakota State University); and Chri
 stoph Junghans, Robert Pavel, and Vinay Ramakrishnaiah (Los Alamos Nationa
 l Laboratory)\n---------------------\nOpeNNdd: Open Neural Networks for Dr
 ug Discovery: Creating Free and Easy Methods for Designing Medicine\n\nBri
 nging new medicines to patients can be prohibitively expensive in terms of
  time, cost, and resources.  This leaves many diseases without therapeutic
  interventions.  In addition, new and reemerging diseases are increasing i
 n prevalence across the globe at an alarming rate.  The speed and scale of
  ...\n\n\nBryce Kroencke (American River College); Shawn Shacterman (Unive
 rsity of California, Berkeley); Nicholas Pavini (American River College); 
 Benjamin Samudio (American River College, Sierra College); and Silvia Criv
 elli (Lawrence Berkeley National Laboratory)\n---------------------\nTunin
 g CFD Applications for Intel Xeon Phi with TAU Commander and ParaTools Thr
 eadSpotter\n\nTuning and understanding the performance characteristics of 
 computational fluid dynamics (CFD) codes on many-core, NUMA architectures 
 is challenging. One must determine how programming choices impact algorith
 m performance and how best to utilize the available memory caches, high-ba
 ndwidth memory, an...\n\n\nIzaak B. Beekman and Nicholas Chaimov (ParaTool
 s Inc); Sameer Shende and Allen D. Malony (ParaTools Inc, University of Or
 egon); Nicholas Bisek and Ryan Gosse (US Air Force Research Laboratory); a
 nd Andrew Wissink (Create AV, US Army)\n---------------------\nUnderstandi
 ng Potential Performance Issues Using Resource-Based alongside Time Models
 \n\nNumerous challenges and opportunities are introduced by the complexity
  and enormous code legacy of HPC applications, the diversity of HPC archit
 ectures, and the nonlinearity of interactions between applications and HPC
  systems. To address these issues, we propose the Resource-based Alongside
  Time (R...\n\n\nNan ding (Lawrence Berkeley National Laboratory), Victor 
 W. Lee (Intel Corporation), and Wei Xue and Weimin Zheng (Tsinghua Univers
 ity)\n---------------------\nWorkflow for Parallel Processing of Sequentia
 l Mesh Databases\n\nThis poster presents a workflow for parallel loading o
 f sequentially stored mesh databases. It can be used as a connection betwe
 en tools for the creation of complex engineering models along with paralle
 l solvers to allow broader usage of HPC by the engineering community. Scal
 ability tests show that ...\n\n\nOndřej Meca, Lubomír Říha, and Tomáš Brzo
 bohatý (Technical University of Ostrava, Czech Republic)\n----------------
 -----\nMassively Parallel Stress Chain Characterization for Billion Partic
 le DEM Simulation of Accretionary Prism Formation\n\nHerein, a novel algor
 ithm for characterizing stress chains using a large parallel computer syst
 em is presented. Stress chains are important for analyzing the results of 
 large-scale discrete element method (DEM) simulations. However, the genera
 l algorithm is difficult to parallelize especially when s...\n\n\nMikito F
 uruichi, Daisuke Nishiura, and Takane Hori (Japan Agency for Marine-Earth 
 Science and Technology)\n---------------------\nAccelerating Wave-Propagat
 ion Algorithms with Adaptive Mesh Refinement Using the Graphics Processing
  Unit (GPU)\n\nClawpack is a library for solving nonlinear hyperbolic part
 ial differential equations using high-resolution finite volume methods bas
 ed on Riemann solvers and limiters. It supports Adaptive Mesh Refinement (
 AMR), which is essential in solving multi-scale problems. Recently, we add
 ed capabilities to ...\n\n\nXinsheng Qin, Randall LeVeque, and Michael Mot
 ley (University of Washington)\n---------------------\nLarge-Message Size 
 Allreduce at Wire Speed for Distributed Deep Learning\n\nIn large-scale di
 stributed deep learning, the Allreduce operation for large messages (100 K
 B or more) is critical for gathering gradients from multiple worker nodes 
 and broadcasting the sum of the gradients to them. When the message is lar
 ge, the latency in Allreduce operation would make it difficul...\n\n\nKenj
 i Tanaka, Yuki Arikawa, Kenji Kawai, Junichi Kato, Tsuyoshi Ito, Huy Cu Ng
 o, Kazutaka Morita, Fumiaki Miura, Takeshi Sakamoto, and Satoshi Shigemats
 u (Japan Telegraph and Telephone Corporation)\n---------------------\nA Lo
 w-Communicaton Method to Solve Poisson's Equation on Locally-Structured Gr
 ids\n\nThis poster describes a new algorithm, Method of Local Corrections 
 (MLC), and a high-performance implementation for solving Poisson's equatio
 n with infinite-domain boundary conditions, on locally-refined nested rect
 angular grids.  The data motion is comparable to that of only a single V-c
 ycle of mul...\n\n\nBrian Van Straalen, Peter McCorquodale, and Phil Colel
 la (Lawrence Berkeley National Laboratory) and Christos Kavouklis (Lawrenc
 e Livermore National Laboratory)\n---------------------\nMulti-Client Deep
 IO for Large-Scale Deep Learning on HPC Systems\n\nWith the growth of comp
 utation power, leadership High-Performance Computing (HPC) systems can tra
 in larger datasets for Deep neural networks (DNNs) more efficiently. On HP
 C systems, a training dataset is on a parallel file system or node-local s
 torage devices. However, not all HPC clusters have node...\n\n\nYue Zhu, F
 ahim Chowdhury, and Huansong Fu (Florida State University); Adam Moody, Ka
 thryn Mohror, and Kento Sato (Lawrence Livermore National Laboratory); and
  Weikuan Yu (Florida State University)\n---------------------\nRedesigning
  The Absorbing Boundary Algorithm for Asynchronous High Performance Acoust
 ic Wave Propagation\n\nExploiting high concurrency, relaxing the synchrony
  of existing algorithms, and increasing data reuse have immense effect in 
 performance. We integrate the Multicore-optimized Wavefront Diamond (MWD) 
 tiling approach by Malas et al. [SIAM SISC, 2015, ACM Trans. Parallel Comp
 ut. 2017],  which takes int...\n\n\nRached Abdelkhalak and Kadir Akbudak (
 King Abdullah University of Science and Technology) and Vincent Etienne an
 d Thierry Tonellot (Saudi Aramco)\n---------------------\nEnabling Reprodu
 cible Microbiome Science through Decentralized Provenance Tracking in QIIM
 E 2\n\nIn this poster, we demonstrate the ways in which automatic, integra
 ted, decentralized provenance tracking in QIIME 2, a leading microbiome bi
 oinformatics platform, enables reproducible microbiome science. We use sam
 ple data from a recent study of arid soil microbiomes  (Significant Impact
 s of Increa...\n\n\nAhmad Turan Naimey and Christopher Keefe (Northern Ari
 zona University, Pathogen and Microbiome Institute)\n---------------------
 \nProcessing-in-Storage Architecture for Machine Learning and Bioinformati
 cs\n\nUser-generated and bioinformatics database volumes has been increasi
 ng exponentially for more than a decade. With the slowdown and approaching
  end of Moore's law, traditional technologies cannot satisfy the increasin
 g demands for processing power.   This work presents PRINS, a highly-paral
 lel in-sto...\n\n\nRoman Kaplan, Leonid Yavits, and Ran Ginosar (Israel In
 stitute of Technology)\n---------------------\nOptimization of Ultrasound 
 Simulations on Multi-GPU Servers\n\nRealistic ultrasound simulations have 
 found a broad area of applications in preoperative photoacoustic screening
  and non-invasive ultrasound treatment planing. However, the domains are t
 ypically thousands of wavelengths in size, leading to large-scale numerica
 l models with billions of unknowns. The ...\n\n\nFilip Vaverka and Matej S
 petko (Brno University of Technology, Faculty of Information Technology); 
 Bradley E. Treeby (University College London, Biomedical Ultrasound Group)
 ; and Jiri Jaros (Brno University of Technology, Faculty of Information Te
 chnology)\n---------------------\nGPU-Accelerated Interpolation for 3D Ima
 ge Registration\n\nImage registration is a key technology in image computi
 ng with numerous applications in medical imaging. Our overarching goal is 
 the design of a consistent and unbiased computational framework for the in
 tegration of medical imaging data with simulation and optimization to supp
 ort clinical decision m...\n\n\nNaveen Himthani (University of Texas, Inst
 itute for Computational Engineering and Sciences); Andreas Mang (Universit
 y of Houston); Amir Gholami (University of California, Berkeley); and Geor
 ge Biros (University of Texas, Institute for Computational Engineering and
  Sciences)\n---------------------\nOpenACC to FPGA: A Directive-Based High
 -Level Programming Framework for High-Performance Reconfigurable Computing
 \n\nAccelerator-based heterogeneous computing has become popular solutions
  for power-efficient high performance computing (HPC).  Along these lines,
  Field Programmable Gate Arrays (FPGAs) have offered more advantages in te
 rms of performance and energy efficiency for specific workloads than other
  acceler...\n\n\nSeyong Lee (Oak Ridge National Laboratory), Jacob Lambert
  (University of Oregon), Jungwon Kim and Jeffrey S. Vetter (Oak Ridge Nati
 onal Laboratory), and Allen D. Malony (University of Oregon)\n------------
 ---------\nEnabling Data Analytics Workflows Using Node-Local Storage\n\nT
 he convergence of high-performance computing (HPC) and Big Data is a neces
 sity with the push toward extreme-scale computing. As HPC simulations beco
 me more complex, the analytics need to process larger amounts of data, whi
 ch poses significant challenges for coupling HPC simulations with Big Data
  an...\n\n\nTu Mai Anh Do (University of Southern California, Information 
 Sciences Institute); Ming Jiang, Brian Gallagher, Albert Chu, and Cyrus Ha
 rrison (Lawrence Livermore National Laboratory); and Karan Vahi and Ewa De
 elman (University of Southern California, Information Sciences Institute)\
 n---------------------\nRefactoring and Optimizing Multiphysics Combustion
  Models for Data Parallelism\n\nHigh-fidelity combustion simulations combi
 ne high-resolution computational fluid dynamics numerical methods with mul
 ti-physics models to capture chemical kinetics and transport processes. Th
 ese multi-physics models can dominate the computation cost of the simulati
 on. Due to the high cost of combusti...\n\n\nChristopher Stone (US Departm
 ent of Defense HPC Modernization Program, Engility Corporation); Alexei Po
 ludnenko (Texas A&M University); and Brian Taylor (US Air Force Research L
 aboratory)\n---------------------\nSC18 Research Posters\n\nSC18 Research 
 Posters will be on display on Tuesday, Wednesday, Thursday from 8:30am to 
 5pm in the C2/3/4 Ballroom.\n\n---------------------\nHigh-Accuracy Scalab
 le Solutions to the Dynamic Facility Layout Problem\n\nThe dynamic facilit
 y layout problem (DFLP) is concerned with finding arrangements of faciliti
 es within plant locations that minimize the sum of material handling and r
 elocation costs over a planning horizon. DFLP is relevant in manufacturing
  engineering; accurate solutions can reduce operational cos...\n\n\nApan Q
 asem, Clara Novoa, Chandra Kolla, and Samantha Coyle (Texas State Universi
 ty)\n---------------------\nScript of Scripts Polyglot Notebook and Workfl
 ow System\n\nComputationally intensive disciplines such as computational b
 iology often use tools implemented in different languages and analyze data
  on high-performance computing systems. Although scientific workflow syste
 ms can powerfully execute large-scale data-processing, they are not suitab
 le for ad hoc dat...\n\n\nGao Wang (University of Chicago); Man Chong Leon
 g (Rice University); and Bo Peng (University of Texas, MD Anderson Cancer 
 Center)\n---------------------\nMATEDOR: MAtrix, TEnsor, and Deep-Learning
  Optimized Routines\n\nThe MAtrix, TEnsor, and Deep-learning Optimized Rou
 tines (MATEDOR) project develops software technologies and standard APIs, 
 along with a sustainable and portable library, for large-scale computation
 s that can be broken down into very small matrix or tensor computations. T
 he main target of MATEDOR i...\n\n\nAhmad Abdelfattah, Jack Dongarra, Stan
 imire Tomov, and Ichitaro Yamazaki (University of Tennessee) and Azzam Hai
 dar (Nvidia Corporation)\n---------------------\nBinarized ImageNet Infere
 nce in 29us\n\nWe propose a single-FPGA-based accelerator for ultra-low-la
 tency inference of ImageNet in this work. The design can complete the infe
 rence of Binarized AlexNet within 29us with accuracy comparable to other B
 NN implementations.  We achieve this performance with the following contri
 butions: 1. We comp...\n\n\nTong Geng (Boston University, Pacific Northwes
 t National Laboratory); Ang Li (Pacific Northwest National Laboratory); Ti
 anqi Wang (Boston University); Shuaiwen Leon Song (Pacific Northwest Natio
 nal Laboratory); and Martin Herbordt (Boston University)\n----------------
 -----\nA Compiler Framework for Fixed-Topology Non-Deterministic Finite Au
 tomata on SIMD Platforms\n\nAutomata traversal acceleration has been studi
 ed on various parallel platforms. Many existing acceleration methods store
  finite automata states and transitions in memory. For these designs memor
 y size and bandwidth are the main limiting factors to performance and powe
 r efficiency. Many applications,...\n\n\nMarziyeh Nourian, Hancheng Wu, an
 d Michela Becchi (North Carolina State University)\n---------------------\
 nParallel Implementation of Machine Learning-Based Many-Body Potentials on
  CPU and GPU\n\nMachine learning models can be used to develop highly accu
 rate and efficient many-body potentials for molecular simulations based on
  the many-body expansion of the total energy.  A prominent example is the 
 MB-pol water model that employs permutationally invariant polynomials (PIP
 s) to represent the ...\n\n\nYaoguang Zhai and Nathaniel Danandeh (Univers
 ity of California, San Diego); Zhenye Tan (University of California, San D
 iego; Tongji University); Sicun Gao and Francesco Paesani (University of C
 alifornia, San Diego); and Andreas W. Goetz (San Diego Supercomputer Cente
 r)\n---------------------\nMulti-GPU Accelerated Non-Hydrostatic Numerical
  Ocean Model with GPUDirect RDMA Transfers\n\nWe have implemented our “kin
 aco” numerical ocean model on Tokyo University’s Reedbush supercomputer, w
 hich utilizes the latest Nvidia Pascal P100 GPUs with GPUDirect technology
 . We have also optimized the model’s Poisson/Helmholtz solver by adjusting
  the global memory alignment and thread block conf...\n\n\nTakateru Yamagi
 shi (Research Organization for Information Science and Technology, Japan) 
 and Yoshimasa Matsumura and Hiroyasu Hasumi (University of Tokyo)\n-------
 --------------\nHardware Acceleration of CNNs with Coherent FPGAs\n\nThis 
 paper describes a new flexible approach to implementing energy-efficient C
 NNs on FPGAs. Our design leverages the Coherent Accelerator Processor Inte
 rface (CAPI) which provides a cache-coherent view of system memory to atta
 ched accelerators. Convolution layers are formulated as matrix multiplica.
 ..\n\n\nMd Syadus Sefat, Semih Aslan, and Apan Qasem (Texas State Universi
 ty)\n---------------------\nExploring Application Performance on Fat-Tree 
 Networks in the Presence of Congestion\n\nNetwork congestion, which occurs
  when multiple applications simultaneously use shared links in cluster net
 work, can cause poor communication performance, decreasing the performance
  and scalability of parallel applications. Many studies are performed whil
 e clusters also run other production workloads...\n\n\nPhilip A. Taffet (R
 ice University, Lawrence Livermore National Laboratory); Sanil Rao (Univer
 sity of Virginia, Lawrence Livermore National Laboratory); and Ian Karlin 
 (Lawrence Livermore National Laboratory)\n---------------------\nDistribut
 ed Fast Boundary Element Methods\n\nWe present a parallel implementation o
 f the fast boundary element method (BEM) for the Helmholtz equation. After
  a brief description of BEM, vectorization of the computationally most dem
 anding kernels, and shared memory parallelization, we focus on the distrib
 uted memory parallelization using a new ...\n\n\nMichal Merta, Jan Zapleta
 l, and Michal Kravcenko (Technical University of Ostrava, Czech Republic)\
 n---------------------\nWarpX: Toward Exascale Modeling of Plasma Particle
  Accelerators\n\nTurning the current experimental plasma accelerator state
 -of-the-art from a promising technology into mainstream scientific tools d
 epends critically on high-performance, high-fidelity modeling of complex p
 rocesses that develop over a wide range of space and time scales. As part 
 of the U.S. Departmen...\n\n\nMaxence Thevenet, Jean-Luc Vay, Ann Almgren,
  John Bell, Remi Lehe, Andrew Myers, Jaehong Park, Olga Shapoval, and Weiq
 un Zhang (Lawrence Berkeley National Laboratory); Lixin Ge, Mark Hogan, an
 d Cho Ng (SLAC National Accelerator Laboratory); and Dave Grote (Lawrence 
 Livermore National Laboratory)\n---------------------\nA Locality and Memo
 ry Congestion-Aware Thread Mapping Method for Modern NUMA Systems\n\nOn mo
 dern NUMA systems, the memory congestion problem could degrade performance
  more than the memory access locality problem because a large number of pr
 ocessor cores in the systems can cause heavy congestion on memory controll
 ers. In this work, we propose a thread mapping method that considers the .
 ..\n\n\nMulya Agung, Muhammad Alfian Amrizal, Ryusuke Egawa, and Hiroyuki 
 Takizawa (Tohoku University)\n---------------------\nHermes: a Multi-Tiere
 d Distributed I/O Buffering System for HDF5\n\nHigh-Performance Computing 
 (HPC) systems’ increasing ability to run data-intensive problems at larger
  scale and resolution has driven the evolution of modern storage technolog
 ies. In addition, extreme amounts of data are collected by large scientifi
 c instruments and sensor network is resulting in a ...\n\n\nHariharan Deva
 rajan (Illinois Institute of Technology, HDF Group)\n---------------------
 \nEstimating Molecular Dynamics Chemical Shift with GPUs\n\nExperimental c
 hemical shifts (CS) from solution and solid state magic-angle-spinning nuc
 lear magnetic resonance spectra provide atomic level data for each amino a
 cid within a protein or complex. However, structure determination of large
  complexes and assemblies based on NMR data alone remains challe...\n\n\nE
 ric F. Wright and Mauricio H. Ferrato (University of Delaware)\n----------
 -----------\nFlowOS-RM: Disaggregated Resource Management System\n\nA trad
 itional data center consists of monolithic-servers is confronted with limi
 tations including lack of operational flexibility, low resource utilizatio
 n, low maintainability, etc. Resource disaggregation is a promising soluti
 on to address the above issues. We propose a concept of disaggregated da..
 .\n\n\nRyousei Takano, Kuniyasu Suzaki, and Hidetaka Koie (National Instit
 ute of Advanced Industrial Science and Technology (AIST))\n---------------
 ------\nUsing Thrill to Process Scientific Data on HPC\n\nWith ongoing imp
 rovement of computational power and memory capacity, the volume of scienti
 fic data keeps growing. To gain insights from vast amounts of data, scient
 ists are starting to look at Big Data processing and analytics tools such 
 as Apache Spark. In this poster, we explore Thrill, a framewor...\n\n\nMar
 iia Karabin (Clemson University, Los Alamos National Laboratory); Xinyu Ch
 en (University of New Mexico); Supreeth Suresh (University of Wyoming); Iv
 o Jimenez (University of California, Santa Cruz); and Li-Ta Lo and Pascal 
 Grosset (Los Alamos National Laboratory)\n---------------------\nLarge Sca
 le MPI-Parallelization of LBM and DEM Systems: Accelerating Research by Us
 ing HPC\n\nCasting, solidification, and the behavior of dry, saturated, an
 d partially saturated granular media are examples of interesting and impor
 tant problems in multiple areas of civil, mechanical, and chemical enginee
 ring. For interacting particle-fluid systems, the Discrete Element Method 
 (DEM) and Latti...\n\n\nBohumir Jelinek, George Mason, John Peters, Daniel
  Johnson, and Marcus Brumfield (Mississippi State University); Alex Carril
 lo (US Army Engineer Research and Development Center); and Clay Goodman an
 d Farshid Vahedifard (Mississippi State University)\n---------------------
 \nBoosting the Scalability of Car-Parrinello Molecular Dynamics Simulation
 s for Multi- and Manycore Architectures\n\nWe present our recent optimizat
 ions of the ultra-soft pseudo-potential (USPP) code path of the ab inito m
 olecular dynamics program CPMD (www.cpmd.org). Following the internal inst
 rumentation of CPMD, all relevant USPP routines have been revised to fully
  support hybrid MPI+OpenMP parallelization. For...\n\n\nTobias Klöffel and
  Bernd Meyer (University of Erlangen-Nuremberg) and Gerald Mathias (Leibni
 z Supercomputing Centre)\n---------------------\nUPC++ and GASNet-EX: PGAS
  Support for Exascale Applications and Runtimes\n\nLawrence Berkeley Natio
 nal Lab is developing a programming system to support HPC application deve
 lopment using the Partitioned Global Address Space (PGAS) model. This work
  is driven by the emerging need for adaptive, lightweight communication in
  irregular applications at exascale.  We present an ove...\n\n\nScott B. B
 aden, Paul H. Hargrove, Hadia Ahmed, John Bachan, Dan Bonachea, Steven Hof
 meyr, Mathias Jacquelin, Amir Kamil, and Brian van Straalen (Lawrence Berk
 eley National Laboratory)\n---------------------\nCompiling SIMT Programs 
 on Multi- and Many-Core Processors with Wide Vector Units: A Case Study wi
 th CUDA\n\nThere has been an increasing interest in SIMT programming tools
  for multi- and manycore (co)processors with wide vector extensions. In th
 is work, we study the effective implementation of a SIMT programming model
  (a subset of CUDA C) on Intel platforms with 512-bit vector extensions (h
 ybrid MIMD/SIMD...\n\n\nHancheng Wu, John Ravi, and Michela Becchi (North 
 Carolina State University)\n---------------------\nLarge Scale Computation
  of Quantiles Using MELISSA\n\nQuantiles being order statistics, the class
 ical approach for their computation requires availability of the full samp
 le before ranking it. This approach is not suitable at exascale. Large ens
 embles would need to gather a prohibitively large amount of data. We propo
 se an iterative approach based on t...\n\n\nAlejandro Ribes (EDF Research 
 and Development), Théophile Terraz (French Institute for Research in Compu
 ter Science and Automation (INRIA)), Yvan Fournier and Bertrand Iooss (EDF
  Research and Development), and Bruno Raffin (French Institute for Researc
 h in Computer Science and Automation (INRIA))\n---------------------\nTens
 orfolding: Improving Convolutional Neural Network Performance with Fused M
 icrokernels\n\nConvolution layers are prevalent in many classes of deep ne
 ural networks, including Convolutional Neural Networks (CNNs) which provid
 e state-of-the-art results for tasks like image recognition, neural machin
 e translation and speech recognition. In the recent past, several techniqu
 es to improve gener...\n\n\nMichael Anderson, Evangelos Georganas, Sasikan
 th Avancha, and Alexander Heinecke (Intel Corporation)\n------------------
 ---\nImproving the I/O Performance and Memory Usage of the Xolotl Cluster 
 Dynamics Simulator\n\nXolotl is a cluster dynamics simulator used to predi
 ct gas bubble evolution in solids. It is currently being used to simulate 
 bubble formation in the plasma-facing surface within fusion reactors and t
 he nuclear fuel used in fission reactors. After observing performance prob
 lems in coupled-code simul...\n\n\nPhilip C. Roth (Oak Ridge National Labo
 ratory), Sophie Blondel (University of Tennessee), David E. Bernholdt (Oak
  Ridge National Laboratory), and Brian D. Wirth (University of Tennessee)\
 n---------------------\nMLModelScope: Evaluate and Measure Machine Learnin
 g Models within AI Pipelines\n\nThe current landscape of Machine Learning 
 (ML) and Deep Learning (DL) is rife with non-uniform frameworks, models, a
 nd system stacks but lacks standard tools to facilitate the evaluation and
  measurement of models. Due to the absence of such tools, the current prac
 tice for evaluating and comparing th...\n\n\nAbdul Dakkak, Cheng Li, and W
 en-mei Hwu (University of Illinois) and Jinjun Xiong (IBM)\n--------------
 -------\nAutomatic Generation of Mixed-Precision Programs\n\nFloating-poin
 t arithmetic is foundational to scientific computing in HPC, and choices a
 bout floating-point precision can have a significant effect on the accurac
 y and speed of HPC codes. Unfortunately, current precision optimization to
 ols require significant user interaction, and few work on the sca...\n\n\n
 Logan Moody (Lawrence Livermore National Laboratory, James Madison Univers
 ity); Nathan Pinnow (Lawrence Livermore National Laboratory, Western Washi
 ngton University); Michael O. Lam (James Madison University, Lawrence Live
 rmore National Laboratory); Harshitha Menon, Markus Schordan, and G. Scott
  Lloyd (Lawrence Livermore National Laboratory); and Tanzima Islam (Wester
 n Washington University)\n---------------------\nDetection of Silent Data 
 Corruptions in Smooth Particle Hydrodynamics Simulations\n\nSoft errors, s
 uch as silent data corruptions (SDCs) hinder the correctness of large-scal
 e scientific applications. Ghost replication (GR) is proposed herein as th
 e first SDCs detector relying on the fast error propagation inherent to ap
 plications that employ the smooth particle hydrodynamics (SPH) m...\n\n\nA
 urélien Cavelan, Florina M. Ciorba, and Ruben M. Cabezón (University of Ba
 sel)\n---------------------\nMaking Sense of Scientific Simulation Ensembl
 es\n\nScientists run many simulations with varying initial conditions, kno
 wn as "ensembles", to understand the influence and relationships among mul
 tiple parameters or ensemble members. Most of the ensemble visualization a
 nd analysis approaches and techniques focus on analyzing the relationships
  between e...\n\n\nMai Dahshan and Nicholas Polys (Virginia Tech)\n-------
 --------------\nHPC-as-a-Service for Life Sciences\n\nHPC-as-a-Service is 
 a well-known term in the area of high performance computing. It enables us
 ers to access an HPC infrastructure without a need to buy and manage their
  own infrastructure. Through this service, academia and industry can take 
 advantage of the technology without an upfront investment ...\n\n\nVaclav 
 Svaton and Jan Martinovic (Technical University of Ostrava, Czech Republic
 ); Nina Jeliazkova (IDEAconsult Ltd, Bulgaria); Vladimir Chupakhin (Jansse
 n Pharmaceutika NV); Pavel Tomancak (Max Planck Institute of Molecular Cel
 l Biology and Genetics); and Petr Vojta (Palacký University Olomouc, Czech
  Republic)\n---------------------\nSciGaP: Apache Airavata Hosted Science 
 Gateways\n\nThe goal of the Science Gateways Platform as a service (SciGaP
 .org) project is to provide core services for building and hosting science
  gateways. Over the last two years, SciGaP services have been used to buil
 d and host over twenty-five science gateways. SciGaP services support thes
 e gateways throu...\n\n\nMarlon Pierce, Suresh Marru, Eroma Abeysinghe, Su
 dhakar Pamidighantam, Marcus Christie, and Dimuthu Upeksha (Indiana Univer
 sity)\n---------------------\nRGB (Redfish Green500 Benchmarker): A Green5
 00 Benchmarking Tool Using Redfish\n\nPerformance and energy are important
  factors for supercomputers and data-centers with a trade-off between them
 . Energy efficiency metric considers both of these properties.  The Green5
 00 is a branch of Top500 project which provides a list of supercomputers b
 ased on energy efficiency. It has a manual...\n\n\nElham Hojati, Yong Chen
 , and Alan Sill (Texas Tech University) and Jon Hass (Dell Inc)\n---------
 ------------\nWhich Architecture Is Better Suited for Matrix-Free Finite-E
 lement Algorithms: Intel Skylake or Nvidia Volta?\n\nThis work presents a 
 performance comparison of highly tuned matrix-free finite element kernels 
 from the finite element library on different contemporary computer archite
 ctures, NVIDIA V100 and P100 GPUs, an Intel Knights Landing Xeon Phi, and 
 two multi-core Intel CPUs (Broadwell and Skylake).  The a...\n\n\nMartin K
 ronbichler (Technical University Munich), Momme Allalen and Martin Ohleric
 h (Leibniz Supercomputing Centre), and Wolfgang A. Wall (Technical Univers
 ity Munich)\n---------------------\nFloating-Point Autotuner for CPU-Based
  Mixed-Precision Applications\n\nIn this poster, we present the design and
  development of an autotuning tool for floating-point code. The goal is to
  balance accuracy and performance in order to produce an efficient and acc
 urate mixed-precision program. The tuner starts by maximizing accuracy thr
 ough the use of a high-precision libr...\n\n\nRuidong Gu, Paul A. Beata, a
 nd Michela Becchi (North Carolina State University)\n---------------------
 \nProgramming the EMU Architecture: Algorithm Design Considerations for Mi
 gratory-Threads-Based Systems\n\nThe decades-old memory bottleneck problem
  for data-intensive applications is getting worse as the processor core co
 unts continue to increase. Workloads with sparse memory access characteris
 tics only achieve a fraction of a system's total memory bandwidth. EMU arc
 hitecture provides a radical approach...\n\n\nMehmet E. Belviranli, Seyong
  Lee, and Jeffrey S. Vetter (Oak Ridge National Laboratory)\n-------------
 --------\nAI Matrix – Synthetic Benchmarks for DNN\n\nThe current AI bench
 marks suffer from a number of drawbacks. First, they cannot adapt to the e
 merging changes of deep learning (DL) algorithms and are fixed once select
 ed. Second, they contain tens to hundreds of applications and have very lo
 ng running time. Third, they are mainly selected from open...\n\n\nWei Wei
 , Lingjie Xu, Lingling Jin, and Wei Zhang (Alibaba Inc) and Tianjun Zhang 
 (University of California, Berkeley)\n\nRegistration Category: Tech Progra
 m Reg Pass, Exhibits Reg Pass
END:VEVENT
END:VCALENDAR