DOCSLIB.ORG

CS 258 Parallel Computer Architecture Lecture 2

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text

Load more

Review CS 258 • Industry has decided that Multiprocessing is the Parallel Computer Architecture future/best use of transistors Lecture 2 – Every major chip manufacturer now making MultiCore chips • History of microprocessor architecture is parallelism Convergence of Parallel Architectures – translates area and density into performance • The Future is higher levels of parallelism – Parallel Architecture concepts apply at many levels January 28, 2008 – Communication also on exponential curve Prof John D. Kubiatowicz • Proper way to compute speedup – Incorrect way to measure: http://www.cs.berkeley.edu/~kubitron/cs258 » Compare parallel program on 1 processor to parallel program on p processors – Instead: » Should compare uniprocessor program on 1 processor to parallel program on p processors 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.2 History Plan for Today • Parallel architectures tied closely to programming • Look at major programming models models – where did they come from? – Divergent architectures, with no predictable pattern of – The 80s architectural rennaisance! growth. – What do they provide? – Mid 80s renaissance – How have they converged? • Extract general structure and fundamental issues Application Software Systolic System Arrays Software SIMD Architecture Systolic SIMD Message Passing Arrays Generic Dataflow Architecture Shared Memory Message Passing Dataflow Shared Memory 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.3 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.4 Programming Model Shared Memory ⇒ Shared Addr. Space Processor • Conceptualization of the machine that programmer Processor uses in coding applications Processor – How parts cooperate and coordinate their activities – Specifies communication and synchronization operations Processor • Multiprogramming Shared Memory – no communication or synch. at program level Processor • Shared address space Processor – like bulletin board • Message passing Processor – like letters or phone calls, explicit point to point Processor • Data parallel: • Range of addresses shared by all processors – more regimented, global actions on data – All communication is Implicit (Through memory) – Implemented with shared address space or message passing – Want to communicate a bunch of info? Pass pointer. • Programming is “straightforward” – Generalization of multithreaded programming 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.5 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.6 Historical Development Adding Processing Capacity • “Mainframe” approach – Motivated by multiprogramming I/O P devices – Extends crossbar used for Mem and I/O – Processor cost-limited => crossbar P Mem Mem Mem Mem I/O ctrl I/O ctrl – Bandwidth scales with p I/O C – High incremental cost I/O C Interconnect Interconnect » use multistage instead M M M M • “Minicomputer” approach Processor Processor – Almost all microprocessor systems have bus – Motivated by multiprogramming, TP • Memory capacity increased by adding modules – Used heavily for parallel computing I/O I/O – Called symmetric multiprocessor (SMP) C C M M • I/O by controllers and devices – Latency larger than for uniprocessor • Add processors for processing! – Bus is bandwidth bottleneck – For higher-throughput multiprogramming, or parallel » caching is key: coherence problem $ $ programs – Low incremental cost P P 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.7 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.8 Shared Physical Memory Shared Virtual Address Space • Any processor can directly reference any location • Process = address space plus thread of control – Communication operation is load/store • Virtual-to-physical mapping can be established so – Special operations for synchronization that processes shared portions of address space. • Any I/O controller - any memory – User-kernel or multiple processes • Multiple threads of control on one address space. • Operating system can run on any processor, or all. – Popular approach to structuring OS’s – OS uses shared memory to coordinate – Now standard application capability (ex: POSIX threads) • Writes to shared address visible to other threads • What about application processes? – Natural extension of uniprocessors model – conventional memory operations for communication – special atomic operations for synchronization » also load/stores 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.9 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.10 Structured Shared Address Space Cache Coherence Problem R? Machine physical address space Virtual address spaces for a collection of processes communicating W R? via shared addresses Pn pr i vat e $4 $4 $4 $4 Load Pn Common physical P2 addresses P1 Write-Through? P0 0 4 St or e 1 5 Miss P2 pr i vat e Shared portion 2 6 of address space 3 7 P1 pr i vat e Private portion of address space P0 pr i vat e • Caches are aliases for memory locations • Does every processor eventually see new value? • Add hoc parallelism used in system code • Tightly related: Cache Consistency • Most parallel applications have structured SAS – In what order do writes appear to other processors? • Same program on each processor • Buses make this easy: every processor can snoop on – shared variable X means the same thing to each thread every write – Essential feature: Broadcast 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.11 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.12 Engineering: Intel Pentium Pro Quad Engineering: SUN Enterprise CPU P-Pro P-Pro P-Pro Interrupt 256-KB module module module controller L2 $ CPU/mem Bus interface P P cards $ $ P-Pro bus (64-bit data, 36-bit addr ess, 66 MHz) $2 $2 Mem ctrl PCI PCI Memory Bus interface/switch bridge bridge controller PCI I/O MIU cards PCI bus PCI bus 1-, 2-, or 4-way Gigaplane bus (256 data, 41 address, 83 MHz) interleaved DRAM I/O cards Bus interface – All coherence and multiprocessing glue in processor module SBUS SBUS SBUS 100bT, SCSI 100bT, • Proc + mem card - I/O card 2 FiberChannel – Highly integrated, targeted at high volume – 16 cards of either type – All memory accessed over bus, so symmetric – Low latency and bandwidth – Higher bandwidth, higher latency bus 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.13 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.14 Quad-Processor Xeon Architecture Scaling Up M M M °°° Omega Network Network General Network Network $ $ °°° $ M $ M $ °°° M $ P P P P P P “Dance hall” Distributed memory – Problem is interconnect: cost (crossbar) or bandwidth (bus) – Dance-hall: bandwidth still scalable, but lower cost than crossbar » latencies to memory uniform, but uniformly large – Distributed memory or non-uniform memory access (NUMA) » Construct shared address space out of simple message • All sharing through pairs of front side busses (FSB) transactions across a general-purpose network (e.g. read- – Memory traffic/cache misses through single chipset to memory request, read-response) – Example “Blackford” chipset – Caching shared (particularly nonlocal) data? 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.15 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.16 Stanford DASH The MIT Alewife Multiprocessor • Clusters of 4 processors share 2nd-level cache • Up to 16 clusters tied together with 2-dim mesh • 16-bit directory associated with every memory line – Each memory line has home cluster that contains DRAM – The 16-bit vector says which clusters (if any) have read copies – Only one writer permitted at a time P P P P L1-$ L1-$ L1-$ L1-$ • Cache-coherence Shared Memory – Partially in Software! L2-Cache – Limited Directory + software overflow • User-level Message-Passing • Rapid Context-Switching • Never got more than 12 clusters (48 processors) • 2-dimentional Asynchronous network working at one time: Asynchronous network probs! • One node/board • Got 32-processors (+ I/O boards) working 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.17 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.18 Engineering: Cray T3E AMD Direct Connect External I/O P Mem $ Mem ctrl and NI XY Switch Z – Scale up to 1024 processors, 480MB/s links – Memory controller generates request message for non-local references • Communication over general interconnect – No hardware mechanism for coherence – Shared memory/address space traffic over network » SGI Origin etc. provide this – I/O traffic to memory over network – Multiple topology options (seems to scale to 8 or 16 processor chips) 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.19 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.20 What is underlying Shared Memory?? Message Passing Architectures Network • Complete computer as building block, including I/O – Communication via Explicit I/O operations M $ M $ °°° M $ P P P • Programming model – direct access only to private address space (local memory), – communication via explicit messages (send/receive) Systolic Network Arrays SIMD • High-level block diagram Generic – Communication integration? Architecture Message Passing » Mem, I/O, LAN, Cluster M $ M $ °°° M $ – Easier to build and scale than SAS Dataflow P P P Shared Memory • Programming model more removed from basic • Packet switched networks better utilize available hardware operations link bandwidth than circuit switched networks – Library or OS intervention • So, network passes messages around! 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.21 1/28/08 Kubiatowicz CS258 ©UCB Spring 2008 Lec 2.22 Message-Passing Abstraction Evolution of Message-Passing Machines • Early machines: FIFO on each link – HW close to prog. Model; Match ReceiveY
Recommended publications
  • Real-Time Visualization of Aerospace Simulations Using Computational Steering and Beowulf Clusters

    Real-Time Visualization of Aerospace Simulations Using Computational Steering and Beowulf Clusters

    The Pennsylvania State University The Graduate School Department of Computer Science and Engineering REAL-TIME VISUALIZATION OF AEROSPACE SIMULATIONS USING COMPUTATIONAL STEERING AND BEOWULF CLUSTERS A Thesis in Computer Science and Engineering by Anirudh Modi Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy August 2002 We approve the thesis of Anirudh Modi. Date of Signature Paul E. Plassmann Associate Professor of Computer Science and Engineering Thesis Co-Advisor Co-Chair of Committee Lyle N. Long Professor of Aerospace Engineering Professor of Computer Science and Engineering Thesis Co-Advisor Co-Chair of Committee Rajeev Sharma Associate Professor of Computer Science and Engineering Padma Raghavan Associate Professor of Computer Science and Engineering Mark D. Maughmer Professor of Aerospace Engineering Raj Acharya Professor of Computer Science and Engineering Head of the Department of Computer Science and Engineering iii ABSTRACT In this thesis, a new, general-purpose software system for computational steering has been developed to carry out simulations on parallel computers and visualize them remotely in real-time. The steering system is extremely lightweight, portable, robust and easy to use. As a demonstration of the capabilities of this system, two applications have been developed. A parallel wake-vortex simulation code has been written and integrated with a Virtual Reality (VR) system via a separate graphics client. The coupling of computational steering of paral- lel wake-vortex simulation with VR setup provides us with near real-time visualization of the wake-vortex data in stereoscopic mode. It opens a new way for the future Air-Traffic Control systems to help reduce the capacity constraint and safety problems resulting from the wake- vortex hazard that are plaguing the airports today.
  • Computer Hardware

    Computer Hardware

    Computer Hardware MJ Rutter mjr19@cam Michaelmas 2014 Typeset by FoilTEX c 2014 MJ Rutter Contents History 4 The CPU 10 instructions ....................................... ............................................. 17 pipelines .......................................... ........................................... 18 vectorcomputers.................................... .............................................. 36 performancemeasures . ............................................... 38 Memory 42 DRAM .................................................. .................................... 43 caches............................................. .......................................... 54 Memory Access Patterns in Practice 82 matrixmultiplication. ................................................. 82 matrixtransposition . ................................................107 Memory Management 118 virtualaddressing .................................. ...............................................119 pagingtodisk ....................................... ............................................128 memorysegments ..................................... ............................................137 Compilers & Optimisation 158 optimisation....................................... .............................................159 thepitfallsofF90 ................................... ..............................................183 I/O, Libraries, Disks & Fileservers 196 librariesandkernels . ................................................197
  • Parallel Computer Architecture

    Parallel Computer Architecture

    Parallel Computer Architecture Introduction to Parallel Computing CIS 410/510 Department of Computer and Information Science Lecture 2 – Parallel Architecture Outline q Parallel architecture types q Instruction-level parallelism q Vector processing q SIMD q Shared memory ❍ Memory organization: UMA, NUMA ❍ Coherency: CC-UMA, CC-NUMA q Interconnection networks q Distributed memory q Clusters q Clusters of SMPs q Heterogeneous clusters of SMPs Introduction to Parallel Computing, University of Oregon, IPCC Lecture 2 – Parallel Architecture 2 Parallel Architecture Types • Uniprocessor • Shared Memory – Scalar processor Multiprocessor (SMP) processor – Shared memory address space – Bus-based memory system memory processor … processor – Vector processor bus processor vector memory memory – Interconnection network – Single Instruction Multiple processor … processor Data (SIMD) network processor … … memory memory Introduction to Parallel Computing, University of Oregon, IPCC Lecture 2 – Parallel Architecture 3 Parallel Architecture Types (2) • Distributed Memory • Cluster of SMPs Multiprocessor – Shared memory addressing – Message passing within SMP node between nodes – Message passing between SMP memory memory nodes … M M processor processor … … P … P P P interconnec2on network network interface interconnec2on network processor processor … P … P P … P memory memory … M M – Massively Parallel Processor (MPP) – Can also be regarded as MPP if • Many, many processors processor number is large Introduction to Parallel Computing, University of Oregon,
  • Programming Languages, Database Language SQL, Graphics, GOSIP

    Programming Languages, Database Language SQL, Graphics, GOSIP

    b fl ^ b 2 5 I AH1Q3 NISTIR 4951 (Supersedes NISTIR 4871) VALIDATED PRODUCTS LIST 1992 No. 4 PROGRAMMING LANGUAGES DATABASE LANGUAGE SQL GRAPHICS Judy B. Kailey GOSIP Editor POSIX COMPUTER SECURITY U.S. DEPARTMENT OF COMMERCE Technology Administration National Institute of Standards and Technology Computer Systems Laboratory Software Standards Validation Group Gaithersburg, MD 20899 100 . U56 4951 1992 NIST (Supersedes NISTIR 4871) VALIDATED PRODUCTS LIST 1992 No. 4 PROGRAMMING LANGUAGES DATABASE LANGUAGE SQL GRAPHICS Judy B. Kailey GOSIP Editor POSIX COMPUTER SECURITY U.S. DEPARTMENT OF COMMERCE Technology Administration National Institute of Standards and Technology Computer Systems Laboratory Software Standards Validation Group Gaithersburg, MD 20899 October 1992 (Supersedes July 1992 issue) U.S. DEPARTMENT OF COMMERCE Barbara Hackman Franklin, Secretary TECHNOLOGY ADMINISTRATION Robert M. White, Under Secretary for Technology NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY John W. Lyons, Director - ;,’; '^'i -; _ ^ '’>.£. ; '':k ' ' • ; <tr-f'' "i>: •v'k' I m''M - i*i^ a,)»# ' :,• 4 ie®®;'’’,' ;SJ' v: . I 'i^’i i 'OS -.! FOREWORD The Validated Products List is a collection of registers describing implementations of Federal Information Processing Standards (FTPS) that have been validated for conformance to FTPS. The Validated Products List also contains information about the organizations, test methods and procedures that support the validation programs for the FTPS identified in this document. The Validated Products List is updated quarterly. iii ' ;r,<R^v a;-' i-'r^ . /' ^'^uffoo'*^ ''vCJIt<*bjteV sdT : Jr /' i^iL'.JO 'j,-/5l ':. ;urj ->i: • ' *?> ^r:nT^^'Ad JlSid Uawfoof^ fa«Di)itbiI»V ,, ‘ isbt^u ri il .r^^iytsrH n 'V TABLE OF CONTENTS 1.
  • ECE 590: Digital Systems Design Using Hardware Description Language (VHDL) Systolic Implementation of Faddeev's Algorithm in V

    ECE 590: Digital Systems Design Using Hardware Description Language (VHDL) Systolic Implementation of Faddeev's Algorithm in V

    Project Report ECE 590: Digital Systems Design using Hardware Description Language (VHDL) Systolic Implementation of Faddeev’s Algorithm in VHDL. Final Project Tejas Tapsale. PSU ID: 973524088. Project Report Introduction: = In this project we are implementing Nash’s systolic implementation and Chuang an He,s systolic implementation for Faddeev’s algorithm. These two implementations have their own advantages and drawbacks. Here in this project report we first see detail of Nash implementation and then we will go for Chaung and He’s implementation. The organization of this report is like this:- 1. First we take detail idea about what is systolic architecture and how it can be used for matrix multiplication and its advantages and disadvantages. 2. Then we discuss about Gaussian Elimination for matrix computation and its properties. 3. Then we will see Faddeev’s algorithm and how it is used. 4. Systolic arrays for MATRIX TRIANGULARIZATION 5. We will discuss Nash implementation in detail and its VHDL coding. 6. Advantages and disadvantage of Nash systolic implementation. 7. Chaung and He’s implementation in detail and its VHDL coding. 8. Difficulties chased in this project. 9. Conclusion. 10. VHDL code for Nash Implementation. 11. VHDL code for Chaung and He’s Implementation. 12. Simulation Results. 13. References. 14. PowerPoint Presentation Project Report 1: Systolic Architecture: = A systolic array is composed of matrix-like rows of data processing units called cells. Data processing units (DPU) are similar to central processing units (CPU)s, (except for the usual lack of a program counter, since operation is transport-triggered, i.e., by the arrival of a data object).
  • On the Efficiency of Register File Versus Broadcast Interconnect For

    On the Efficiency of Register File Versus Broadcast Interconnect For

    On the Efficiency of Register File versus Broadcast Interconnect for Collective Communications in Data-Parallel Hardware Accelerators Ardavan Pedram, Andreas Gerstlauer Robert A. van de Geijn Department of Electrical and Computer Engineering Department of Computer Science The University of Texas at Austin The University of Texas at Austin fardavan,[email protected] [email protected] Abstract—Reducing power consumption and increasing effi- on broadcast communication among a 2D array of PEs. In ciency is a key concern for many applications. How to design this paper, we focus on the LAC’s data-parallel broadcast highly efficient computing elements while maintaining enough interconnect and on showing how representative collective flexibility within a domain of applications is a fundamental question. In this paper, we present how broadcast buses can communication operations can be efficiently mapped onto eliminate the use of power hungry multi-ported register files this architecture. Such collective communications are a core in the context of data-parallel hardware accelerators for linear component of many matrix or other data-intensive operations algebra operations. We demonstrate an algorithm/architecture that often demand matrix manipulations. co-design for the mapping of different collective communication We compare our design with typical SIMD cores with operations, which are crucial for achieving performance and efficiency in most linear algebra routines, such as GEMM, equivalent data parallelism and with L1 and L2 caches that SYRK and matrix transposition. We compare a broadcast bus amount to an equivalent aggregate storage space. To do so, based architecture with conventional SIMD, 2D-SIMD and flat we examine efficiency and performance of the cores for data register file for these operations in terms of area and energy movement and data manipulation in both GEneral matrix- efficiency.
  • Computer Architectures

    Computer Architectures

    Parallel (High-Performance) Computer Architectures Tarek El-Ghazawi Department of Electrical and Computer Engineering The George Washington University Tarek El-Ghazawi, Introduction to High-Performance Computing slide 1 Introduction to Parallel Computing Systems Outline Definitions and Conceptual Classifications » Parallel Processing, MPP’s, and Related Terms » Flynn’s Classification of Computer Architectures Operational Models for Parallel Computers Interconnection Networks MPP’s Performance Tarek El-Ghazawi, Introduction to High-Performance Computing slide 2 Definitions and Conceptual Classification What is Parallel Processing? - A form of data processing which emphasizes the exploration and exploitation of inherent parallelism in the underlying problem. Other related terms » Massively Parallel Processors » Heterogeneous Processing – In the1990s, heterogeneous workstations from different processor vendors – Now, accelerators such as GPUs, FPGAs, Intel’s Xeon Phi, … » Grid computing » Cloud Computing Tarek El-Ghazawi, Introduction to High-Performance Computing slide 3 Definitions and Conceptual Classification Why Massively Parallel Processors (MPPs)? » Increase processing speed and memory allowing studies of problems with higher resolutions or bigger sizes » Provide a low cost alternative to using expensive processor and memory technologies (as in traditional vector machines) Tarek El-Ghazawi, Introduction to High-Performance Computing slide 4 Stored Program Computer The IAS machine was the first electronic computer developed, under
  • Systolic Computing on Gpus for Productive Performance

    Systolic Computing on Gpus for Productive Performance

    Systolic Computing on GPUs for Productive Performance Hongbo Rong Xiaochen Hao Yun Liang Intel Intel, Peking University Peking University [email protected] [email protected] [email protected] Lidong Xu Hong H Jiang Pradeep Dubey Intel Intel Intel [email protected] [email protected] [email protected] Abstract an SIMT (single-instruction multiple-threads) programming We propose a language and compiler to productively build interface, and rely on an underlying compiler to transparently high-performance software systolic arrays that run on GPUs. map a wrap of threads to SIMD execution units. If data need to Based on a rigorous mathematical foundation (uniform re- be exchanged among threads in the same wrap, programmers currence equations and space-time transform), our language have to write explicit shuffle instructions [19]. has a high abstraction level and covers a wide range of ap- This paper proposes a new programming style that pro- plications. A programmer specifies a projection of a dataflow grams GPUs as building software systolic arrays. Systolic ar- compute onto a linear systolic array, while leaving the detailed rays have been extensively studied since 1978 [15], and shown implementation of the projection to a compiler; the compiler an abundance of practice-oriented applications, mainly in implements the specified projection and maps the linear sys- fields dominated by iterative procedures [32], e.g. image and tolic array to the SIMD execution units and vector registers signal processing, matrix arithmetic, non-numeric applica- of GPUs. In this way, both productivity and performance are tions, relational database [7, 8, 10, 11, 16, 17, 30], and so on.
  • CS252 Lecture Notes Multithreaded Architectures

    CS252 Lecture Notes Multithreaded Architectures

    CS252LectureNotes MultithreadedArchitectures Concept Tolerateormasklongandoftenunpredictablelatencyoperationsbyswitchingtoanothercontext, whichisabletodousefulwork. SituationToday–Whyisthistopicrelevant? ILPhasbeenexhaustedwhichmeansthreadlevelparallelismmustbeutilized ‹ Thegapbetweenprocessorperformanceandmemoryperformanceisstilllarge ‹ Thereisamplereal-estateforimplementation ‹ Moreapplicationsarebeingwrittenwiththeuseofthreadsandmultitaskingisubiquitous ‹ Multiprocessorsaremorecommon ‹ Networklatencyisanalogoustomemorylatency ‹ Complexschedulingisalreadybeingdoneinhardware ClassicalProblem 60’sand70’s ‹ I/Olatencypromptedmultitasking ‹ IBMmainframes ‹ Multitasking ‹ I/Oprocessors ‹ Cacheswithindiskcontrollers RequirementsofMultithreading ‹ Storageneedtoholdmultiplecontext’sPC,registers,statusword,etc. ‹ Coordinationtomatchaneventwithasavedcontext ‹ Awaytoswitchcontexts ‹ Longlatencyoperationsmustuseresourcesnotinuse Tovisualizetheeffectoflatencyonprocessorutilization,letRbetherunlengthtoalonglatency event,letLbetheamountoflatencythen: 1 Util Util=R/(R+L) 0 L 80’s Problemwasrevisitedduetotheadventofgraphicsworkstations XeroxAlto,TIExplorer ‹ Concurrentprocessesareinterleavedtoallowfortheworkstationstobemoreresponsive. ‹ Theseprocessescoulddriveormonitordisplay,input,filesystem,network,user processing ‹ Processswitchwasslowsothesubsystemsweremicroprogrammedtosupportmultiple contexts ScalableMultiprocessor ‹ Dancehall–asharedinterconnectwithmemoryononesideandprocessorsontheother. ‹ Orprocessorsmayhavelocalmemory M M P/M P/M
  • Dynamic Adaptation Techniques and Opportunities to Improve HPC Runtimes

    Dynamic Adaptation Techniques and Opportunities to Improve HPC Runtimes

    Dynamic Adaptation Techniques and Opportunities to Improve HPC Runtimes Mohammad Alaul Haque Monil, Email: [email protected], University of Oregon. Abstract—Exascale, a new era of computing, is knocking at subsystem, later generations of integrated heterogeneous sys- the door. Leaving behind the days of high frequency, single- tems such as NVIDIA’s Tegra Xavier have taken heterogeneity core processors, the new paradigm of multicore/manycore pro- within the same chip to the extreme. Processing units with cessors in complex heterogeneous systems dominates today’s HPC landscape. With the advent of accelerators and special-purpose diverse instruction set architectures (ISAs) are present in nodes processors alongside general processors, the role of high perfor- in supercomputers such as Summit, where IBM Power9 CPUs mance computing (HPC) runtime systems has become crucial are connected to 6 NVIDIA V100 GPUs. Similarly in inte- to support different computing paradigms under one umbrella. grated systems such as NVIDIA Xavier, processing units with On one hand, modern HPC runtime systems have introduced diverse instruction set architectures work together to accelerate a rich set of abstractions for supporting different technologies and hiding details from the HPC application developers. On the kernels belonging to emerging application domains. Moreover, other hand, the underlying runtime layer has been equipped large-scale distributed memory systems with complex network with techniques to efficiently synchronize, communicate, and map structures and modern network interface cards adds to this work to compute resources. Modern runtime layers can also complexity. To efficiently manage these systems, efficient dynamically adapt to achieve better performance and reduce runtime systems are needed.
  • Research Article a Systolic Array-Based FPGA Parallel Architecture for the BLAST Algorithm

    Research Article a Systolic Array-Based FPGA Parallel Architecture for the BLAST Algorithm

    International Scholarly Research Network ISRN Bioinformatics Volume 2012, Article ID 195658, 11 pages doi:10.5402/2012/195658 Research Article A Systolic Array-Based FPGA Parallel Architecture for the BLAST Algorithm Xinyu Guo,1 Hong Wang,2 and Vijay Devabhaktuni1 1 Electrical Engineering and Computer Science Department, The University of Toledo, MS.308, 2801 W. Bancroft Street, Toledo, OH 43607, USA 2 Department of Engineering Technology, The University of Toledo, MS.402, 2801 W. Bancroft Street, Toledo, OH 43606, USA Correspondence should be addressed to Hong Wang, [email protected] Received 23 May 2012; Accepted 25 July 2012 Academic Editors: F. Couto, B. Haubold, and J. T. L. Wang Copyright © 2012 Xinyu Guo et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. A design of systolic array-based Field Programmable Gate Array (FPGA) parallel architecture for Basic Local Alignment Search Tool (BLAST) Algorithm is proposed. BLAST is a heuristic biological sequence alignment algorithm which has been used by bio- informatics experts. In contrast to other designs that detect at most one hit in one-clock-cycle, our design applies a Multiple Hits Detection Module which is a pipelining systolic array to search multiple hits in a single-clock-cycle. Further, we designed a Hits Combination Block which combines overlapping hits from systolic array into one hit. These implementations completed the first and second step of BLAST architecture and achieved significant speedup comparing with previously published architectures.
  • R00456--FM Getting up to Speed

    R00456--FM Getting up to Speed

    GETTING UP TO SPEED THE FUTURE OF SUPERCOMPUTING Susan L. Graham, Marc Snir, and Cynthia A. Patterson, Editors Committee on the Future of Supercomputing Computer Science and Telecommunications Board Division on Engineering and Physical Sciences THE NATIONAL ACADEMIES PRESS Washington, D.C. www.nap.edu THE NATIONAL ACADEMIES PRESS 500 Fifth Street, N.W. Washington, DC 20001 NOTICE: The project that is the subject of this report was approved by the Gov- erning Board of the National Research Council, whose members are drawn from the councils of the National Academy of Sciences, the National Academy of Engi- neering, and the Institute of Medicine. The members of the committee responsible for the report were chosen for their special competences and with regard for ap- propriate balance. Support for this project was provided by the Department of Energy under Spon- sor Award No. DE-AT01-03NA00106. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the organizations that provided support for the project. International Standard Book Number 0-309-09502-6 (Book) International Standard Book Number 0-309-54679-6 (PDF) Library of Congress Catalog Card Number 2004118086 Cover designed by Jennifer Bishop. Cover images (clockwise from top right, front to back) 1. Exploding star. Scientific Discovery through Advanced Computing (SciDAC) Center for Supernova Research, U.S. Department of Energy, Office of Science. 2. Hurricane Frances, September 5, 2004, taken by GOES-12 satellite, 1 km visible imagery. U.S. National Oceanographic and Atmospheric Administration. 3. Large-eddy simulation of a Rayleigh-Taylor instability run on the Lawrence Livermore National Laboratory MCR Linux cluster in July 2003.