DOCSLIB.ORG

Beyond Earth Simulator

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Beyond Earth Simulator Beyond Earth Simulator International Computing for the Atmospheric Sciences Symposium (iCAS2017) 海洋研究開発機構 September 11, 2017 地球情報基盤センターMakoto Tsukakoshi Information 情報システム部Systems Department Center for Earth Information Science and塚越 Technology眞 JAMSTEC Japan Agency for Marine-Earth Science and Technology 2 The main seven research and development issues during the third mid-term plan During the third mid-term plan, we set and address the seven research and development issues with all our strength due to promote strategic and focused research and development based on the national and social needs. Exploring untapped Ocean drilling – submarine resources Getting to know the Earth from beneath the seabed Detecting signals of global Information Science - environmental Predicting the Earth's change future by simulations Understanding seismogenic zones, and contributing to disaster mitigation Marine Bioscience - Exploring the unknown Construction of research base extreme biosphere to solve to spawn the ocean frontier the mystery of life 3 Typhoon-marine interaction using nonstatic atmospheric wave ocean coupling model Typhoon Vera in 1959 by Kazuhisa Tsuboki (2015) Typhoon-Ocean Interaction Study Using the Coupled Atmosphere-Ocean Non-hydrostatic Model: With Careful Consideration of Upper Outflow Layer Clouds of Typhoon 4 • Earth Simulator ： - Developed by NASDA, JAERI and JAMSTEC (563.4* Oku Yen) - Operation started on March 2002 and immediately recognized as #1 supercomputer - “It will keep #1 for 2 years in peak and 5 years in effective performance” by H. Miyoshi *inclusing facilities and building 5 Earth Simulator has changed its role when “K computer” is completed Council for Science and Technology Policy (CSTP) Report on Nov 25, 2005: After the completion of Leading Edge High Performance Supercomputer (i.e. K-Computer), other existing supercomputers should carry their roles in each fields, according to their purpose and performance , for example, Earth Simulator, should responsible for Marine and Earth Science on the base of accumulated application software assets, and thus they should meet growing demand for HPC in our country. 既存のスーパーコンピュータは、本プロジェクトで開発する計算機システムの完成後も、例えば地球シミュレータについて は、これまでに蓄積してきたアプリケーションソフトウェアを活用して効果的・効率的に海洋地球科学分野を主とした計算を 担当するなど、それぞれが得意とする各分野を中心として、目的や性能に応じた計算に応するとの役割分担の下に、今後 も増大が予想される我が国の計算機需要に応えることとしている。 HPCI Strategy - MEXT ES K-computer Operated as national infrastructure thorugh HPCI* ＊ “Innovative High Performance Computing Infrastructure” initiative ES 6 3 Generation of Earth Simulator since 2002 System for prediction and creation of Marine-Earth Science data and information for Society Proof of Feasibility of Utilize Simulation, Prediction and data for Society Foundation of Marine- Earth Simulation Simulation and Prediction Challenge new subjects in Marine-Earth Science 10 PFLOPS Flagship supercomputer System for Marine-Earth Science June 2015 for All Science Earth Simulator Peak/Effective Performance 10x 1.3PFLOPS / 320TB 1 PFLOPS “Computenik” (Jack Donguarra, NY Times, April 20, 2002) March 2009 ３ Earth Simulator (ES2) Peak Performance x 131TFLOPS / 20TB March 2002 100 TFLOPS Earth Simulator (1st) 40TFLOPS / 10TB About 1.5MW* 8x to 10x Effective Performance based on Application Programs* 10 TFLOPS About 3MW* * Single Program Turn Around Time Performance: 2.1x to 2.8x Number of Programs: 4x = System Throughput 8x to 10 x 1 TFLOPS About 5MW* * Effective Power Consumption 7 JAMSTEC Information System Outline SINET JAMSTEC backbone network Science Information NETwork 4 10Gbps ×2 （1G～10Gbps） User terminals Earth Simulator backbone Earth Simulator (NEC SX-ACE) Visualization Server 40Gbps ×2 Service servers Mass Storage System Large Scale Shared Memory User terminals Server （SGI UV2000) Super computer systems (HPE Apollo 6000, 380 node in 2018) (ICE-X, UV2000, SX-9) 8 Earth Simulator Outline Earth Simulator JAMSTEC Backbone network Main Computer Storage Computing Nodes (5120) + Internode Frontend Server Network /home 87.6 TB 4 servers Model: NEC SX-ACE /data Total Performance 4.7 PB - Number of nodes 5,120 - Peak Performance 1.31 PFLOPS - Memory BandwidthBandwidth 1.31PB/s - Memory size 320 TB Node Performance Work - Number of CPU 1(4cores) 13.5 PB - Peak Performance 256 GFLOPS Large Scale Shared (64GFLOPS x 4cores) Double Precision Memory Server（UV2000) 17PB - Memory Bandwidth 256 GB/s Pre and Post System Mass Storage System - Memory size 64 GB 9 Benchmark Test showed ES’ｓ value The same #CPU (sockets) DOUBLES the performance with ASIS code. 2.8X with further tuning. (Performance Ratio vs. ES2) Both ES2 and SX-ACE deploy same #CPU This means : Effective Throughput Performance is > 10x (TAT – 2 to 2.8x, #of jobs 4x) Average (TUNE) 2.8 Tuning Indeed, FLOPs counts increased by 12x 260 EFLOPs(executed in FY2012) to Average (ASIS) 2.1 3,119EFLOPs(executed in FY2016) 温暖化MIROC 温暖化MIROC 全球大気NICAM Specfem3D地震 RSGDX地震 ミクロ大気MSSG 海洋ODA/同化 10 Effective Resource Comparison with K-computer K Computer Usage in FY2015 Atomic Energy Mathematics Fusion (1.7%) (1.0%) Earth Material Science Particle Physics Engineering Biology Science Chemistry Life Science Astronomy (24.3%) (22.0%) （18.8%） (16.4%) (15.5%) Others (0.3%) ~2/3(estimation) Equivalent to 28% of Next Earth Simulator Earth ~28% Science 90.3% Earth Simulator in FY2016 11 Earth Simulator Resource Allocation/Usage（FY2016） Industry & Innovation 3.7% 0.7% External JAMSTEC Math Sci & 5.4% proposed proposed Engineering Life Science project (27) 20% project (23) 40% Earth Internal, Earthquake, 20.3% Tsunami 30% 70.0% Environmental 10% Strategic project with Change Government special support (6) contract (7) Fee based usage(8) Earth Simulator Usage by Earth Simulator Resource Application [node hour] Allocation (71 projects) [node hour] Outside Japan Industry Government Agency University Institutes using Earth Simulator 12 Earth Simulator Operation Highlights • Stable Operation in FY2016 • Power Efficiency - Availability 99.86%, - Improved 30% from the previous ES2(NEC - Utilization 89.07% SX-9) - Total 3,119EFLOPs * executed in FY2016 - Continued effort to optimized the cooling “ ” (7.5% of total peak performance : Total Executed FLOPs/Total Power* *:including facilities): 24x366xall nodes) 171TFLOPs/kwh in 2016 - improved 9.2% from 2015 *Using 2016 Yellowstone CPU hours/24x366 (81%), 3,119EF requires 2.3% of peak performance on Cheyanne 13 (proposal) Use actual FLOPs per watt to evaluate total energy efficiency of computing centers Efficiency should be “actually achieved Performance/Total energy consumed” Peak or Linpack Performance : Don’t represent real work loads (“Green500”) HPCG or other Benchmark suites : Better but do not fit each center’s AP spectrum (each centers should have been optimized to its own application work loads.) Benchmarks show only capability. Do not represent the center’s operational efforts. Power consumption must be actual and total include cooling. Evaluate the energy efficiency by “Executed FLOPs per watt” ! Where, the energy is measured total inclusive amount. (Example) In 2016 (JAN-DEC), Earth Simulator (Entire System including cooling facilities) has actually achieved 178,956 GFLOPS/Kwh (7.6% of theoretical peak x 100% non- stop operation) Using HPCG/peak performance ratio (5.2%) and annual GWh from the annual report 2014, efficiency of “K-Computer” is estimated 109,221 GFLOPS/Kwh 14 Goal: “Marine-Earth Informatics” 1. Produce large and accurate data data from observation, experimentation, simulation data of various form (structured/unstructured, images, texts) • Building large and precise data sets by accurate observation and simulation applying the latest data handling method. 2. Create information from large and diverse data • Development of a computing method to deal with large data with high performance • Applying ML, pattern recognition, analysis of precursor characteristic of rare events 3. Present new application and use of information • New use case of information for government and industry • New application area 4. Develop ICT technology to handle large and complex data • Integrating Simulation and Data Analysis • Interactive real time visualization • Leading-edge High performance platforms and devices 15 database for Policy Decision making for Future climate change (d4PDF) • d4PDF is a database including both historical and +4K future climate simulation results, and provides probabilistic Information on climate change in extreme events by high- resolution large ensemble simulations with a 60km AGCM and 20km NHRCM. • 2.6M node hours* and 2PB output on Earth Simulator in 2015 by MRI and MIROC team. *9% of entire resources in 2015 MRI-AGCM Historical Simulations 60km 6,000(AGCM)+3,000(RCM) years ・ ・ 100 members (AGCM) 60 years (1950-2010) ・50 members (NHRCM) downscaling (SST perturbations representing observational error) +4K Future Climate Simulations 5,400 years (AGCM+RCM) ・ ・ 90 members (AGCM) 60 years (1950-2010) ・90 members (NHRCM) (6 ΔSST from 6 CMIP5 models x 15 SST perturbations) Histogram of daily mean precipitation at Tokyo (60km AGCM) (a) Present day (b) Changes in the +4K world MRI-NHRCM 10yr Frequency(%) 20km 100yr Ratio of freq. (+4K/present) freq. of Ratio Daily precipitation (mm/day) Daily precipitation (mm/day) Shiogama,H. (2016) Database for Policy Decision making for Future climate change (d4PDF) 16 Four-dimensional Variational Ocean ReAnalysis for the Western North Pacific over 30 years (FORA-WNP30) •First-ever dataset covering the western North Pacific (15-65N, 117E -160W) over the last three decades
Load more

Recommended publications
  • User's Manual
    Management Number:OUCMC-super-技術-007-06 National University Corporation Osaka University Cybermedia Center User's manual NEC First Government & Public Solutions Division Summary 1. System Overview ................................................................................................. 1 1.1. Overall System View ....................................................................................... 1 1.2. Thress types of Computiong Environments ........................................................ 2 1.2.1. General-purpose CPU environment .............................................................. 2 1.2.2. Vector computing environment ................................................................... 3 1.2.3. GPGPU computing environment .................................................................. 5 1.3. Three types of Storage Areas ........................................................................... 6 1.3.1. Features.................................................................................................. 6 1.3.2. Various Data Access Methods ..................................................................... 6 1.4. Three types of Front-end nodes ....................................................................... 7 1.4.1. Features.................................................................................................. 7 1.4.2. HPC Front-end ......................................................................................... 7 1.4.1. HPDA Front-end ......................................................................................
    [Show full text]
  • DMI-HIRLAM on the NEC SX-6
    DMI-HIRLAM on the NEC SX-6 Maryanne Kmit Meteorological Research Division Danish Meteorological Institute Lyngbyvej 100 DK-2100 Copenhagen Ø Denmark 11th Workshop on the Use of High Performance Computing in Meteorology 25-29 October 2004 Outline • Danish Meteorological Institute (DMI) • Applications run on NEC SX-6 cluster • The NEC SX-6 cluster and access to it • DMI-HIRLAM - geographic areas, versions, and improvements • Strategy for utilization and operation of the system DMI - the Danish Meteorological Institute DMI’s mission: • Making observations • Communicating them to the general public • Developing scientific meteorology DMI’s responsibilities: • Serving the meteorological needs of the kingdom of Denmark • Denmark, the Faroes and Greenland, including territorial waters and airspace • Predicting and monitoring weather, climate and environmental conditions, on land and at sea Applications running on the NEC SX-6 cluster Operational usage: • Long DMI-HIRLAM forecasts 4 times a day • Wave model forecasts for the North Atlantic, the Danish waters, and for the Mediterranean Sea 4 times a day • Trajectory particle model and ozone forecasts for air quality Research usage: • Global climate simulations • Regional climate simulations • Research and development of operational and climate codes Cluster interconnect GE−int Completely closed Connection to GE−ext Switch internal network institute network Switch GE GE nec1 nec2 nec3 nec4 nec5 nec6 nec7 nec8 GE GE neci1 neci2 FC 4*1Gbit/s FC links per node azusa1 asama1 FC McData FC 4*1Gbit/s FC links per azusa fiber channel switch 8*1Gbit/s FC links per asama azusa2 asama2 FC 2*1Gbit/s FC links per lun Disk#0 Disk#1 Disk#2 Disk#3 Cluster specifications • NEC SX-6 (nec[12345678]) : 64M8 (8 vector nodes with 8 CPU each) – Desc.
    [Show full text]
  • Performance Modeling the Earth Simulator and ASCI Q
    PAL CCS-3 Performance Modeling the Earth Simulator and ASCI Q Darren J. Kerbyson Adolfy Hoisie, Harvey J. Wasserman Performance and Architectures Laboratory (PAL) Los Alamos National Laboratory April 2003 Los Alamos PAL CCS-3 “26.58Tflops on AFES … 64.9% of peak (640nodes)” “14.9Tflops on Impact-3D …. 45% of peak (512nodes)” “10.5Tflops on PFES … 44% of peak (376nodes)” 40Tflops 20Tflops Los Alamos PAL Talk Overview CCS-3 G Overview of the Earth Simulator – A (quick) view of the architecture of the Earth Simulator (and Q) – A look at its performance characteristics G Application Centric Performance Models – Method of comparing performance is to use trusted models of applications that we are interested in, e.g. SAGE and Sweep3D. – Analytical / Parameterized in system & application characteristics G Models can be used to provide: – Predicted performance prior to availability (hardware or software) – Insight into performance – Performance Comparison (which is better?) G System Performance Comparison (Earth Simulator vs ASCI Q) Los Alamos PAL Earth Simulator: Overview CCS-3 ... ... RCU RCU . RCU AP AP AP Memory Memory Memory 0-7 0-7 0-7 ... Node0 Node1 Node639 640x640 ... crossbar G 640 Nodes (Vector Processors) G interconnected by a single stage cross-bar – Copper interconnect (~3,000Km wire) G NEC markets a product – SX-6 – Not the same as an Earth Simulator Node - similar but different memory sub-system Los Alamos PAL Earth Simulator Node CCS-3 Crossbar LAN network Disks AP AP AP AP AP AP AP AP 0 1 2 3 4 5 6 7 RCU IOP Node contains: 8 vector processors (AP) 16GByte memory Remote Control Unit (RCU) I/O Processor (IOP) 2 3 4 5 6 0 1 31 M M M M M M M .
    [Show full text]
  • Project Aurora 2017 Vector Inheritance
    Vector Coding a simple user’s perspective Rudolf Fischer NEC Deutschland GmbH Düsseldorf, Germany SIMD vs. vector Input Pipeline Result Scalar SIMD people call it “vector”! Vector SX 2 © NEC Corporation 2017 Data Parallelism ▌‘Vector Loop’, data parallel do i = 1, n Real, dimension(n): a,b,c a(i) = b(i) + c(i) … end do a = b + c ▌‘Scalar Loop’, not data parallel, ex. linear recursion do i = 2, n a(i) = a(i-1) + b(i) end do ▌Reduction? do i = 1, n, VL do i_ = i, min(n,i+VL-1) do i = 1, n s_(i_) = s_(i_) + v(i_)* w(i_) s = s + v(i)* w(i) end do end do end do s = reduction(s_) (hardware!) 3 © NEC Corporation 2017 Vector Coding Paradigm ▌‘Scalar’ thinking / coding: There is a (grid-point,particle,equation,element), what am I going to do with it? ▌‘Vector’ thinking / coding: There is a certain action or operation, to which (grid-points,particles,equations,elements) am I going to apply it simultaneously? 4 © NEC Corporation 2017 Identifying the data-parallel structure ▌ Very simple case: do j = 2, m-1 do i = 2, n-1 rho_new(i,j) = rho_old(i,j) + dt * something_with_fluxes(i+/-1,j+/-1) end do end do ▌ Does it tell us something? Partial differential equations Local theories Vectorization is very “natural” 5 © NEC Corporation 2017 Identifying the data-parallel subset ▌ Simple case: V------> do j = 1, m |+-----> do i = 2, n || a(i,j) = a(i-1,j) + b(i,j) |+----- end do V------ end do ▌ The compiler will vectorise along j ▌ non-unit-stride access, suboptimal ▌ Results for a certain case: Totally scalar (directives): 25.516ms Avoid outer vector loop:
    [Show full text]
  • 2020 Global High-Performance Computing Product Leadership Award
    2020 GLOBAL HIGH-PERFORMANCE COMPUTING PRODUCT LEADERSHIP AWARD Strategic Imperatives Frost & Sullivan identifies three key strategic imperatives that impact the automation industry: internal challenges, disruptive technologies, and innovative business models. Every company that is competing in the automation space is obligated to address these imperatives proactively; failing to do so will almost certainly lead to stagnation or decline. Successful companies overcome the challenges posed by these imperatives and leverage them to drive innovation and growth. Frost & Sullivan’s recognition of NEC is a reflection of how well it is performing against the backdrop of these imperatives. Best Practices Criteria for World-Class Performance Frost & Sullivan applies a rigorous analytical process to evaluate multiple nominees for each award category before determining the final award recipient. The process involves a detailed evaluation of best practices criteria across two dimensions for each nominated company. NEC excels in many of the criteria in the high-performance computing (HPC) space. About NEC Established in 1899, NEC is a global IT, network, and infrastructure solution provider with a comprehensive product portfolio across computing, data storage, embedded systems, integrated IT infrastructure, network products, software, and unified communications. Headquartered in Tokyo, Japan, NEC has been at the forefront of accelerating the industrial revolution of the 20th and 21st © Frost & Sullivan 2021 The Growth Pipeline Company™ centuries by leveraging its technical knowhow and product expertise across thirteen different industries1 in industrial and energy markets. Deeply committed to the vision of orchestrating a better world. NEC envisions a future that embodies the values of safety, security, fairness, and efficiency, thus creating long-lasting social value.
    [Show full text]
  • Hardware Technology of the Earth Simulator 1
    Special Issue on High Performance Computing 27 Architecture and Hardware for HPC 1 Hardware Technology of the Earth Simulator 1 By Jun INASAKA,* Rikikazu IKEDA,* Kazuhiko UMEZAWA,* 5 Ko YOSHIKAWA,* Shitaka YAMADA† and Shigemune KITAWAKI‡ 5 1-1 This paper describes the hardware technologies of the supercomputer system “The Earth Simula- 1-2 ABSTRACT tor,” which has the highest performance in the world and was developed by ESRDC (the Earth 1-3 & 2-1 Simulator Research and Development Center)/NEC. The Earth Simulator has adopted NEC’s leading edge 2-2 10 technologies such as most advanced device technology and process technology to develop a high-speed and high-10 2-3 & 3-1 integrated LSI resulting in a one-chip vector processor. By combining this LSI technology with various advanced hardware technologies including high-density packaging, high-efficiency cooling technology against high heat dissipation and high-density cabling technology for the internal node shared memory system, ESRDC/ NEC has been successful in implementing the supercomputer system with its Linpack benchmark performance 15 of 35.86TFLOPS, which is the world’s highest performance (http://www.top500.org/ : the 20th TOP500 list of the15 world’s fastest supercomputers). KEYWORDS The Earth Simulator, Supercomputer, CMOS, LSI, Memory, Packaging, Build-up Printed Wiring Board (PWB), Connector, Cooling, Cable, Power supply 20 20 1. INTRODUCTION (Main Memory Unit) package are all interconnected with the fine coaxial cables to minimize the distances The Earth Simulator has adopted NEC’s most ad- between them and maximize the system packaging vanced CMOS technologies to integrate vector and density corresponding to the high-performance sys- 25 25 parallel processing functions for realizing a super- tem.
    [Show full text]
  • Shared-Memory Vector Systems Compared
    Shared-Memory Vector Systems Compared Robert Bell, CSIRO and Guy Robinson, Arctic Region Supercomputing Center ABSTRACT: The NEC SX-5 and the Cray SV1 are the only shared-memory vector computers currently being marketed. This compares with at least five models a few years ago (J90, T90, SX-4, Fujitsu and Hitachi), with IBM, Digital, Convex, CDC and others having fallen by the wayside in the early 1990s. In this presentation, some comparisons will be made between the architecture of the survivors, and some performance comparisons will be given on benchmark and applications codes, and in areas not usually presented in comparisons, e.g. file systems, network performance, gzip speeds, compilation speeds, scalability and tools and libraries. KEYWORDS: SX-5, SV1, shared-memory, vector systems averaged 95% on the larger node, and 92% on the smaller 1. Introduction node. HPCCC The SXes are supported by data storage systems – The Bureau of Meteorology and CSIRO in Australia SAM-FS for the Bureau, and DMF on a Cray J90se for established the High Performance Computing and CSIRO. Communications Centre (HPCCC) in 1997, to provide ARSC larger computational facilities than either party could The mission of the Arctic Region Supercomputing acquire separately. The main initial system was an NEC Centre is to support high performance computational SX-4, which has now been replaced by two NEC SX-5s. research in science and engineering with an emphasis on The current system is a dual-node SX-5/24M with 224 high latitudes and the Arctic. ARSC provides high Gbyte of memory, and this will become an SX-5/32M performance computational, visualisation and data storage with 224 Gbyte in July 2001.
    [Show full text]
  • Recent Supercomputing Development in Japan
    Supercomputing in Japan Yoshio Oyanagi Dean, Faculty of Information Science Kogakuin University 2006/4/24 1 Generations • Primordial Ages (1970’s) – Cray-1, 75APU, IAP • 1st Generation (1H of 1980’s) – Cyber205, XMP, S810, VP200, SX-2 • 2nd Generation (2H of 1980’s) – YMP, ETA-10, S820, VP2600, SX-3, nCUBE, CM-1 • 3rd Generation (1H of 1990’s) – C90, T3D, Cray-3, S3800, VPP500, SX-4, SP-1/2, CM-5, KSR2 （HPC ventures went out） • 4th Generation (2H of 1990’s) – T90, T3E, SV1, SP-3, Starfire, VPP300/700/5000, SX-5, SR2201/8000, ASCI(Red, Blue) • 5th Generation (1H of 2000’s) – ASCI，TeraGrid，BlueGene/L，X1, Origin,Power4/5, ES, SX- 6/7/8, PP HPC2500, SR11000, …. 2006/4/24 2 Primordial Ages (1970’s) 1974 DAP, BSP and HEP started 1975 ILLIAC IV becomes operational 1976 Cray-1 delivered to LANL 80MHz, 160MF 1976 FPS AP-120B delivered 1977 FACOM230-75 APU 22MF 1978 HITAC M-180 IAP 1978 PAX project started (Hoshino and Kawai) 1979 HEP operational as a single processor 1979 HITAC M-200H IAP 48MF 1982 NEC ACOS-1000 IAP 28MF 1982 HITAC M280H IAP 67MF 2006/4/24 3 Characteristics of Japanese SC’s 1. Manufactured by main-frame vendors with semiconductor facilities (not ventures) 2. Vector processors are attached to mainframes 3. HITAC IAP a) memory-to-memory b) summation, inner product and 1st order recurrence can be vectorized c) vectorization of loops with IF’s (M280) 4. No high performance parallel machines 2006/4/24 4 1st Generation (1H of 1980’s) 1981 FPS-164 (64 bits) 1981 CDC Cyber 205 400MF 1982 Cray XMP-2 Steve Chen 630MF 1982 Cosmic Cube in Caltech, Alliant FX/8 delivered, HEP installed 1983 HITAC S-810/20 630MF 1983 FACOM VP-200 570MF 1983 Encore, Sequent and TMC founded, ETA span off from CDC 2006/4/24 5 1st Generation (1H of 1980’s) (continued) 1984 Multiflow founded 1984 Cray XMP-4 1260MF 1984 PAX-64J completed (Tsukuba) 1985 NEC SX-2 1300MF 1985 FPS-264 1985 Convex C1 1985 Cray-2 1952MF 1985 Intel iPSC/1, T414, NCUBE/1, Stellar, Ardent… 1985 FACOM VP-400 1140MF 1986 CM-1 shipped, FPS T-series (max 1TF!!) 2006/4/24 6 Characteristics of Japanese SC in the 1st G.
    [Show full text]
  • Hardware-Oblivious SIMD Parallelism for In-Memory Column-Stores
    Hardware-Oblivious SIMD Parallelism for In-Memory Column-Stores Template Vector Library — General Approach and Opportunities Annett Ungethüm, Johannes Pietrzyk, Erich Focht Patrick Damme, Alexander Krause, High Performance Computing Europe Dirk Habich, Wolfgang Lehner NEC Deutschland GmbH Database Systems Group Stuttgart, Germany Technische Universität Dresden [email protected] Dresden, Germany fi[email protected] ABSTRACT be mapped to a variety of parallel processing architectures Vectorization based on the Single Instruction Multiple Data like many-core CPUs or GPUs. Another recent hardware- (SIMD) parallel paradigm is a core technique to improve oblivious approach is the data processing interface DPI for query processing performance especially in state-of-the-art modern networks [2]. in-memory column-stores. In mainstream CPUs, vectoriza- In addition to parallel processing over multiple cores, tion is offered by a large number of powerful SIMD exten- vector processing according to the Single Instruction sions growing not only in vector size but also in terms of Multiple Data (SIMD) parallel paradigm is also a heavily complexity of the provided instruction sets. However, pro- used hardware-driven technique to improve the query performance in particular for in-memory column-store gramming with vector extensions is a non-trivial task and 1 currently accomplished in a hardware-conscious way. This database systems [10, 14, 21, 26, 27]. On mainstream implementation process is not only error-prone but also con- CPUs, this vectorization is done using SIMD extensions nected with quite some effort for embracing new vector ex- such as Intel's SSE (Streaming SIMD extensions) or AVX tensions or porting to other vector extensions.
    [Show full text]
  • Performance Evaluation of Supercomputers Using HPCC and IMB Benchmarks
    Performance Evaluation of Supercomputers using HPCC and IMB Benchmarks Subhash Saini1, Robert Ciotti1,Brian T. N. Gunney2, Thomas E. Spelce2, Alice Koniges2, Don Dossa2, Panagiotis Adamidis3, Rolf Rabenseifner3, Sunil R. Tiyyagura3, Matthias Mueller4, and Rod Fatoohi5 1Advanced Supercomputing Division NASA Ames Research Center Moffett Field, California 94035, USA {ssaini, ciotti}@nas.nasa.gov 4ZIH, TU Dresden. Zellescher Weg 12, D-01069 Dresden, Germany 2Lawrence Livermore National Laboratory [email protected] Livermore, California 94550, USA {gunney, koniges, dossa1}@llnl.gov 5San Jose State University 3 One Washington Square High-Performance Computing-Center (HLRS) San Jose, California 95192 Allmandring 30, D-70550 Stuttgart, Germany [email protected] University of Stuttgart {adamidis, rabenseifner, sunil}@hlrs.de Abstract paradigm has become the de facto standard in programming high-end parallel computers. As a result, the performance of a majority of applications depends on the The HPC Challenge (HPCC) benchmark suite and the performance of the MPI functions as implemented on Intel MPI Benchmark (IMB) are used to compare and these systems. Simple bandwidth and latency tests are two evaluate the combined performance of processor, memory traditional metrics for assessing the performance of the subsystem and interconnect fabric of five leading interconnect fabric of the system. However, simple supercomputers - SGI Altix BX2, Cray X1, Cray Opteron measures of these two are not adequate to predict the Cluster, Dell Xeon cluster, and NEC SX-8. These five performance for real world applications. For instance, systems use five different networks (SGI NUMALINK4, traditional methods highlight the performance of network Cray network, Myrinet, InfiniBand, and NEC IXS).
    [Show full text]
  • Optimizing Sparse Matrix-Vector Multiplication in NEC SX-Aurora Vector Engine
    Optimizing Sparse Matrix-Vector Multiplication in NEC SX-Aurora Vector Engine Constantino Gómez1, Marc Casas1, Filippo Mantovani1, and Erich Focht2 1Barcelona Supercomputing Center, {first.last}@bsc.es 2NEC HPC Europe (Germany), {first.last}@EMEA.NEC.COM August 14, 2020 Abstract Sparse Matrix-Vector multiplication (SpMV) is an essential piece of code used in many High Performance Computing (HPC) applications. As previous literature shows, achieving ecient vectorization and performance in modern multi-core systems is nothing straightforward. It is important then to revisit the current state- of-the-art matrix formats and optimizations to be able to deliver high performance in long vector architectures. In this tech-report, we describe how to develop an ef- ficient implementation that achieves high throughput in the NEC Vector Engine: a 256 element-long vector architecture. Combining several pre-processing and ker- nel optimizations we obtain an average 12% improvement over a base SELL-C-σ implementation on a heterogeneous set of 24 matrices. 1 Introduction The Sparse Matrix-Vector (SpMV) product is a ubiquitous kernel in the context of High- Performance Computing (HPC). For example, discretization schemes like the finite dif- ferences or finite element methods to solve Partial Dierential Equations (PDE) pro- duce linear systems with a highly sparse matrix. Such linear systems are typically solved via iterative methods, which require an extensive use of SpMV across the whole execution. In addition, emerging workloads from the data-analytics area also require the manipulation of highly irregular and sparse matrices via SpMV. Therefore, the e- cient execution of this fundamental linear algebra kernel is of paramount importance.
    [Show full text]
  • Supercomputers – Prestige Objects Or Crucial Tools for Science and Industry?
    Supercomputers – Prestige Objects or Crucial Tools for Science and Industry? Hans W. Meuer a 1, Horst Gietl b 2 a University of Mannheim & Prometeus GmbH, 68131 Mannheim, Germany; b Prometeus GmbH, 81245 Munich, Germany; This paper is the revised and extended version of the Lorraine King Memorial Lecture Hans Werner Meuer was invited by Lord Laird of Artigarvan to give at the House of Lords, London, on April 18, 2012. Keywords: TOP500, High Performance Computing, HPC, Supercomputing, HPC Technology, Supercomputer Market, Supercomputer Architecture, Supercomputer Applications, Supercomputer Technology, Supercomputer Performance, Supercomputer Future. 1 e-mail: [email protected] 2 e-mail: [email protected] 1 Content 1 Introduction ..................................................................................................................................... 3 2 The TOP500 Supercomputer Project ............................................................................................... 3 2.1 The LINPACK Benchmark ......................................................................................................... 4 2.2 TOP500 Authors ...................................................................................................................... 4 2.3 The 39th TOP500 List since 1993 .............................................................................................. 5 2.4 The 39th TOP10 List since 1993 ...............................................................................................
    [Show full text]