Find this at http://www.npac.syr.edu/users/gcf/batonrougefeb93/

Overview of HPC for Physical Sciences

This overviews status of HPCC architectures

Software approaches focussing on HPF and an Application Analysis

Problem Architectures Load Balancing and the Complex System Approach

Table of Contents for Overview of HPC for Physical Sciences

001 Overview of High Performance Computing for Physical Sciences
002 Performance of Computers as a Function of Time
003 Performance of QCD Machines
004 When will Parallel Computing Take Over  ?
005 Parallelism Implies
006 Program in Computational Science  		Implemented within current 
    academic framework
007 We have learnt that Parallel Computing Works !
008 METHODOLOGY
009 Sideways View of Concurrent Computing Mappings
010 Sample uses of Concurrent Computers 
    - 4 Nodes
011 Parallel Computing Trends to Year 2000
012 39 National Grand Challenges
013 39 National Grand Challenges (II)
014 39 National Grand Challenges (III)
015 39 National Grand Challenges (IV)
016 Most Grand Challenges are
017 ACTION Motivation
018 Applications of Interest to Industry IA
019 Applications of Interest to Industry IB
020 Applications of Interest to Industry IC
021 Pluses and Minuses of Parallel Computing
022 P C E T E C H Ñ A Result of
023 The Life History of an HPC Idea
024 Core Enabling Software Technologies
025 Core Enabling Algorithms and Components
026 Portable Scalable Languages
027 Problem Architectures
028 Software Problem Architecture Interactions
029 Why build on existing languages - especially Fortran / C
030 Fortran( C ) Plus Message Passing
031 Fortran( C ) Plus Message Passing
032 Data Parallel Fortran
033 Data Parallel Fortran ( C, C++, LISP, ADA )
034 The Origins of High Performance Fortran
035 Strategy for HPF
036 Model for Scalable Languages
037 Dynamically Triangulated Random Surfaces
038 Update Strategies for DTRS
039 N=36 Wireframe DTRS
040 N=144  Lambda =1.5 Colored DTRS
041 N=144 Lambda=1  Colored DTRS
042 Computational Issues and Code Performance Optimization
043 Sequential Computer Architecture
044 Performance of Random Surface Code
045 Parallel Algorithms
046 Parallel Grid Generation
047 What is Fortran 90?
048 HPF Language Features - 1
049 HPF Language Features - 2
050 FORTRAN-90D 
    The First Implementation of HPF 
    (NPAC, Syracuse University) 
    Current Status
051 Fortran90D Performance on Gaussian Elimination
052 Gaussian Elimination 
    Example of Fortran90D Source Code
053 Fortran90D Interface
054 HPF Directives
055 Data Alignment
056 Data Distribution
057 FORALL
058 DO INDEPENDENT
059 HPF Library
060 Intrinsic Library
061 HPF Library
062 Fortran 90 Local Routine Intrinsics
063 Does HPF need a different type of compiler ?
064 HPF Interpreter
065 Benchmarking  Set  ( Fortran90D and Fortran 77 )    
066 HPF/FORTRAN-90D Benchmark Suite
067 The final draft of the HPF Language Specification is version 1.0 -
     DRAFT, dated January 25, 1993	                                   
                      
068 Request for Public Comment on High Performance Fortran
069 Application Structure Versus Fortran90D Features
070 What applications does HPF support? 
071 Current HPF can do I .....
072 Current HPF can do II.....
073 HPF can express using FORALL
074 Classic Irregular Mesh Application
075 HPF Can Express
076 Simulation of Spin Models of Magnets
077 New Monte Carlo Algorithms
078 Magnetization for Random Field Ising Model -- Metropolis
079 Magnetization for Random Field Ising Model --  Simulated Annealing
080 Wolff cluster (bands shown in yellow) for 3 state Potts model at 
    Tc
081 Swendsen-Wang clusters (boundaries shown in black) for 3 state 
    Potts model at Tc
082 Cluster Algorithm versus Metropolis
083 Parallel Algorithms
084 Parallel Cluster Algorithms
085 MIMD algorithm on nCUBE-1
086 SIMD algorithms on CM-2
087 Autocorrelation Plots for Cluster Algorithms
088 HPF probably cannot express well
089 I do not know how to express - let alone optimize - 
090 Late breaking results
091 Large N-Body Calculations     (Quinn, Salmon, Warren)
092 Hierarchical Decomposition
093 Ncube Speed Up of Barnes Hut Algorithm
094 17x106 Body Simulation
    Diameter 250Mpc
    Quinn, Salmon, Warren, Zurek
095 The largest "galaxy" halo (137,000 bodies) from the 8.8M
     body simulation 
    (Warren, Fullagar, Quinn, Zurek)
096 8M bodies - 10 Mpc diameter 
    Final state with ~700 resolved "galaxies" 
    (Warren, Quinn, Zurek)
097 Smooth Particle Hydro simulation of the collision of two 
    "main sequence stars." 137,000 bodies (Warren,Davies)
098 Expressing Problem Structure in Language
099 The map of Problem ---> Computer is performed in two or more 
    stages
100 From Problem to Machine Space as a Mapping
101 Different Granularities of Decomposition I
102 Different Granularities of Decomposition II
103 Compiler/Interpreter Tradeoffs at Different Levels of Granularity
104 The Mapping of Heterogeneous MetaProblems onto Heterogeneous 
    MetaComputer Systems
105 Mapping of Irregular Grid
106 Finding General Maps for FortranD/HPF
107 Three Physical Optimization Methods for Allocating Data to 
    Multicomputer Nodes
108 Finite Element Mesh
109 Graph Contraction
110 Some Results of Load Balancing Studies I
111 Some Results of Load Balancing Studies II
112 Physics Analogy for Load Balancing
113 Complex System SHLSoft governed by Hamiltonian = Execution Time
114 PHYSICS ANALOGY FOR STATIC AND DYNAMIC LOAD BALANCING
115 Definition of Temperature for a Complex System
116 Particle dynamics problem on a four node system
117 Energy Structure in Physics Analogy with Multiple Minima
118 Phase Transitions in Physical model -- Scattered versus Domain 
    Decompositions