BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210527T135010Z
LOCATION:Digital
DTSTART;TZID=Europe/Stockholm:20200623T130000
DTEND;TZID=Europe/Stockholm:20200623T230000
UID:isc_hpc_ISC High Performance 2020_sess341@linklings.com
SUMMARY:Research Paper Session
DESCRIPTION:Research Paper\n\nResearch Paper Proceeddings:\n\n\nhttps://ww
 w.isc-hpc.com/research-papers-2020.html\n\nOffsite Autotuning Approach\n\n
 Seiferth, Korch, Rauber\n\nAutotuning techniques are a promising approach 
 to minimize the otherwise tedious manual effort of optimizing scientific a
 pplications for a specific target platform. Ideally, an autotuning approac
 h is capable of reliably identifying the most efficient implementation var
 iant(s) for a new target system...\n\n---------------------\nPattern-Aware
  Staging for Hybrid Memory Systems\n\nArima, Schulz\n\nThe ever increasing
  demand for higher memory performance and—at the same time—larger memory c
 apacity is leading the industry towards hybrid main memory designs, i.e., 
 memory systems that consist of multiple different memory technologies. Thi
 s is driven by the fact that no single memory technology is...\n\n--------
 -------------\nPredicting Job Power Consumption based on RJMS Submission D
 ata in HPC Systems\n\nSAILLANT, WEILL, MOUGEOT\n\nPower-aware scheduling i
 s a promising solution to the resource usage monitoring of High-Performanc
 e Computing facility electrical power consumption. This kind of solution n
 eeds a reliable estimation of job power consumption to feed the Resources 
 and Jobs Management System at submission time. Availab...\n\n-------------
 --------\nTimemory: Modular Performance Analysis for HPC\n\nMadsen\n\nHPC 
 has undergone a significant transition toward heterogeneous architectures.
  This transition has introduced several issues in code migration to suppor
 t multiple frameworks for targeting the various architectures. In order to
  cope with these challenges, projects such as Kokkos and LLVM create abstr
 ...\n\n---------------------\nLoad-balancing Parallel Relational Algebra\n
 \nKumar, Gilray\n\nRelational algebra (RA) comprises a basis of important 
 operations, sufficient to power state-of-the-art reasoning engines for Dat
 alog and related logic-programming languages. Parallel RA implementations\
 ncan thus play a significant role in extracting parallelism inherent in a 
 wide variety of analytic...\n\n---------------------\nSparse Linear Algebr
 a on AMD and NVIDIA GPUs -- The Race is on\n\nTsai, Cojean, Anzt\n\nEffici
 ently processing sparse matrices is a central and performance-critical par
 t of many scientific simulation codes. Recognizing the adoption of manycor
 e accelerators in HPC, we evaluate in this paper the performance of the cu
 rrently best sparse matrix-vector product (SpMV) implementations on high..
 .\n\n---------------------\nReinit++: Evaluating the Performance of Global
 -Restart Recovery Methods For MPI Fault Tolerance\n\nGeorgakoudis, Guo, La
 guna\n\nScaling supercomputers comes with an increase in failure rates due
  to\nthe increasing number of hardware components. In standard practice,\n
 applications are made resilient through checkpointing data and\nrestarting
  execution after a failure occurs to resume from the latest\ncheckpoint. H
 owever, re-deplo...\n\n---------------------\nDGEMM using Tensor Cores, an
 d Its Accurate and Reproducible Versions\n\nMukunoki, Ozaki, Ogita, Imamur
 a\n\nThis paper proposes a method for implementing dense matrix multiplica
 tion on FP64 (DGEMM) and FP32 (SGEMM) using Tensor Cores on NVIDIA's graph
 ics processing units (GPUs). Tensor Cores are special processing units tha
 t perform 4x4 matrix multiplications on FP16 inputs with FP32 precision, a
 nd return...\n\n---------------------\nTeaMPI—Replication-based Resilience
  without the (Performance) Pain\n\nSamfass, Weinzierl, Hazelwood, Bader\n\
 nIn an era where we can not afford to checkpoint frequently,\nreplication 
 is a generic way forward to construct numerical simulations\nthat can cont
 inue to run even if hardware parts fail. Yet, replication\noften is not em
 ployed on larger scales, as naı̈vely mirroring a computation onc
 e effect...\n\n---------------------\nEnabling Execution of a Legacy CFD M
 ini Application on Accelerators Using OpenMP\n\nNompelis, Jost, Koniges, D
 aley, Eder...\n\nWe describe the process and outcome of our efforts to por
 t a legacy Fortran benchmark code to heterogeneous GPU-accelerated computi
 ng architectures using OpenMP. The benchmark code is one of the multi-zone
  NAS Parallel Benchmarks (NPB-MZ) called SP-MZ. This ``mini-app'' mimics t
 he\ncomputation and da...\n\n---------------------\nSolving Acoustic Bound
 ary Integral Equations Using High Performance Tile Low-Rank LU Factorizati
 on\n\nAlharthi, AlOmairy, Akbudak, Chen, Ltaief...\n\nWe design and develo
 p a new high performance implementation of a fast direct LU-based solver u
 sing low-rank approximations on massively parallel systems. The LU factori
 zation is the first and most time-consuming step toward solving systems of
  linear equations in the context of analyzing acoustic sca...\n\n---------
 ------------\nTime Series Mining at Petascale Performance\n\nRaoofy, Karls
 tetter, Yang, Trinitis, Schulz\n\nThe mining of time series data plays an 
 important role in modern information retrieval and analysis systems. In pa
 rticular, the identification of similarities within and across time series
  has garnered significant attention and effort over the last few years. Fo
 r this task, the class of matrix profi...\n\n---------------------\nSimpli
 fying Communication Overlap in OpenSHMEM Through Integrated User-Level Thr
 ead Scheduling\n\nRahman, Ozog, Dinan\n\nOverlap of communication with com
 putation is a key optimization for high performance computing (HPC) applic
 ations. In this paper, we explore the usage of user-level threading to ena
 ble productive and efﬁcient communication overlap and pipelining. W
 e extend OpenSHMEM with integrated user-level...\n\n---------------------\
 nRunning a Pre-Exascale, Geographically Distributed, Multi-Cloud Scientifi
 c Simulation\n\nSfiligoi, Würthwein, Riedel, Schultz\n\nAs we approach the
  Exascale era, it is important to verify that the existing frameworks and 
 tools will still work at that scale. Moreover, public Cloud computing has 
 been emerging as a viable solution for both prototyping and urgent computi
 ng. Using the elasticity of the Cloud, we have thus put in p...\n\n-------
 --------------\nFootprint-Aware Power Capping for Hybrid Memory Based Syst
 ems\n\nArima, Hanawa, Trinitis, Schulz\n\nHigh Performance Computing (HPC)
  systems are facing severe limitations in both power and memory bandwidth/
 capacity. By now, these limitations have been addressed individually: to i
 mprove performance under a strict power constraint, power capping, which s
 ets power limits to components/nodes/jobs, is ...\n\n---------------------
 \nOpportunities for Cost Savings with In-Transit Visualization\n\nKress, L
 arsen, Choi, Kim, Podhorszki...\n\nWe analyze the opportunities for in-tra
 nsit visualization to provide cost savings compared to in-line visualizati
 on. We begin by developing a cost model that includes factors related to b
 oth in-line and in-transit which allows comparisons to be made between the
  two methods. We begin by developing a ...\n\n---------------------\nUnder
 standing HPC Benchmark Performance on Intel Broadwell and Cascade Lake Pro
 cessors\n\nAlappat, Hofmann, Hager, Fehske, Bishop...\n\nHardware platform
 s in high performance computing are constantly getting more complex to han
 dle even when considering multicore CPUs alone. Numerous features and conf
 iguration options in the hardware and the software environment that are re
 levant for performance are not even known to most application...\n\n------
 ---------------\nSHARP Streaming-Aggregation Hardware Design and Evaluatio
 n\n\nGraham, Shainer, Levi, Bloch\n\nThis paper introduces a new hardware-
 based Mellanox Scalable Hierarchical Aggregation and Reduction Protocol SH
 ARP reduction capability added to Mellanox’s HDR InﬁniBand switches
  called SHARP Streaming-Aggregation. This capability is designed to achiev
 e reduction bandwidths similar to those o...\n\n---------------------\nCom
 munication-Aware Hardware-Assisted MPI Overlap Engine\n\nBayatpour, Hashmi
  Maqbool, Chakraborty, Kandadi suresh, Ghazimirsaeed...\n\nOverlap of comp
 utation and communication is critical for good application-level performan
 ce. Modern high-performance networks offer Hardware-assisted tag matching 
 and rendezvous offload to enable communication progress without involving 
 the host CPU. However, hardware-based offload cannot be used in...\n\n----
 -----------------\nHyPar-Flow: Exploiting MPI and Keras for Scalable Hybri
 d-Parallel DNN Training with TensorFlow\n\nAwan, Jain, Anthony, Subramoni,
  Panda\n\nTo reduce the training time of large-scale Deep Neural Networks,
  scientists have started to explore parallelization strategies like data-p
 arallelism, model-parallelism, and hybrid-parallelism. While data-parallel
 ism has been extensively studied and developed, several problems exist in 
 realizing mode...\n\n---------------------\nDesynchronization and Wave Pat
 tern Formation in MPI-Parallel and Hybrid Memory-Bound Programs\n\nAfzal, 
 Hager, Wellein\n\nAnalytic, first-principles performance modeling of\ndist
 ributed-memory parallel codes is notoriously imprecise. Even for\napplicat
 ions with extremely regular and homogeneous\ncompute-communicate phases, s
 imply adding communication time to\ncomputation time does often not yield 
 a satisfactory prediction ...\n\n---------------------\nShared-Memory Para
 llel Probabilistic Graphical Modeling Optimization\n\nPerciano, Heinemann,
  Camp, Lessley, Bethel\n\nThis work examines performance characteristics o
 f multiple shared-memory implementations of a probabilistic graphical mode
 ling (PGM) optimization code, which forms the basis for an advanced, state
 -of-the art image segmentation method. The work is motivated by the need t
 o accelerate scientific image ...\n\n---------------------\nUsing High-Lev
 el Synthesis to Implement the  Matrix-Vector Multiplication on FPGA\n\nPal
 azzari, Marongiu\n\nThis work presents how to implement the Matrix-Vector 
 Multiplication (MVM) onto FPGA through the QuickPlay High-Level Synthesis 
 flow. The motivations arise from the Adaptive Optics field, where the MVM 
 is the core of the real-time control algorithm which controls the mirrors 
 of a telescope to compen...\n\n---------------------\nFASTHash: FPGA-based
  High Throughput Parallel Hash Table\n\nYang, Kuppannagari, Srivastava, Ka
 nnan, Prasanna\n\nHash table is a fundamental data structure that provides
  efficient data store and access. It is a key component in AI applications
  which rely on building a model of the environment using observations and 
 performing lookups on the model for newer observations. In this work, we d
 evelop FASTHash, a “tr...\n\n---------------------\nSemi-automatic Assessm
 ent of I/O Behavior by Inspecting the Individual Client-Node Timelines\n\n
 Betke, Kunkel\n\nHPC applications with suboptimal I/O behavior interfere w
 ith well-behaving applications and lead to increased application runtime. 
 In some cases, this may even lead to unresponsive systems and unfinished j
 obs. HPC monitoring systems can aid users and support staff to identify  p
 roblematic behavior a...\n\n---------------------\nScaling Genomics Data P
 rocessing with Memory-Driven Computing\n\nBecker, Worlikar, Agrawal, Schul
 tze, Ulas...\n\nResearch is increasingly becoming data-driven and natural 
 sciences are not the exception. In biology and medicine, we can observe an
  exponential growth of data that allows us to gain novel insights that wou
 ld not be possible without data science and increasing efforts for structu
 red collections from...\n\n---------------------\nEmbedding Algorithms for
  Quantum Annealers with Chimera and Pegasus Connection Topologies\n\nZbind
 en, Baertschi, Djidjev, Eidenbenz\n\nWe propose two new algorithms – Sprin
 g-Based MinorMiner (SPMM) and Clique-Based MinorMiner (CLMM) – which take 
 as input the connectivity graph of a Quadratic Unconstrained Binary Optimi
 zation (QUBO) problem and produce as output an embedding of the input grap
 h on a host graph that models the topolog...\n\n\nTag: Pre-Recorded
END:VEVENT
END:VCALENDAR

