HPC 2012

The Second Annual Front Range High Performance Computing Symposium, August 13-14 2012, brought together designers and users of high performance computing systems from universities, government laboratories, and industry along the front range. Topic areas discussed included research, education, and best practices. The symposium was organized by the Front Range Consortium on Research Computing and hosted by Colorado State University’s Information Science and Technology Center (ISTeC, istec.colostate.edu).

 

 

 

FRCRC HPC 2012

AUGUST 13 • MONDAY

 

   
  7:30AM –  8:30AMGRegistrationMorgan Library
  8:30AM –  8:45AMGIntroduction and WelcomeBSB 131
  8:45AM –  9:45AMKWhat Blue Waters is Already Teaching Us about Petascale Computing BSB 131
Speakers: Bill Kramer

Blue Waters, NSF's sustained petascale system, is one of the first systems developed using "co-design" principles. Over the past several years, the Blue Waters team has not just been deploying one of the most balanced and powerful systems in the world, but we have worked with over 25 science and engineering teams, partners and vendor to explore a range of issues that are required to enable real sustained performance for the worlds most challenging computational problems. This presentation will give an overview of the Blue Waters Project and share some of the challenges, issues and solutions that are necessary to make real petascale performance achievable.

View Presentation Slides

  9:45AM – 10:00AMGCoffee BreakBSB 131
 10:00AM – 10:20AMTDevelopment of a Parallel Compositional Reservoir Simulator Clark C146
Speakers: Jeff Brown, Panruo Wu, Zizhong Chen, Hossein Kazemi
 10:00AM – 10:20AMCGPU Progress in Sparse Matrix Solvers for ISV Software Clark C251
Speakers: Stan Posey
 10:00AM – 10:20AMHCampus Bridging Using a Science-DMZ, gridFTP and Globus Online Services Clark C144
Speakers: Thomas Hauser, Daniel Milroy, Conan Moore, Jazcek Braden
 10:20AM – 10:30AMTSimulation for Solving 3D Seismic Wave Propagation Equations Using the Discontinuous Galerkin Method in Parkfield, California Clark C146
Speakers: Feng Chen, Po Chen
 10:20AM – 10:40AMCCoupled Mesoscale Microscale Model for Wind Resource Estimation and Turbine Aerodynamics Using an Overset Approach Clark C251
Speakers: Harish Gopalan, Christopher Gundling, Jayanarayan Sitaraman, Jeffry Mirocha, Wayne Miller
 10:20AM – 10:40AMHExploring Energy Efficient Application Specific Architectures Clark C144
Speakers: Louis Rabiet, Patricia Quinton, Sanjay Rajopadhye
 10:40AM – 11:00AMTThe IRC Ecosystem Modeling System: Design Issues and Risks for Portable, Scalable, and Concurrent Software and Data Clark C146
Speakers: Thomas Hilinski
 10:40AM – 11:00AMCTwo Parameter Estimation Approaches for Stochastic Differential Equation Systems Using Parallel Computing with R Clark C251
Speakers: Libo Sun, Chihoon Lee, Jennifer Hoeting
 10:40AM – 11:00AMHTime-Utility Functions for Evaluating Resource Allocations in Heterogeneous HPC Systems Clark C144
Speakers: LUis Briceno, Bhavesh Khemka, H.J. Siegel, Anthony A. Maciejewski, Christopher Groer, Gregory Koenig, Gene Okonski, Steve Poole
 11:00AM – 11:20AMTEvaluation of Real-Forecast WRF Scaling Up to Tens of Thousands of Cores Clark C146
Speakers: Davide Del Vento, Mike Page, Raffaele Montuoro, Mark Lubin, Scott McMillan
 11:00AM – 11:20AMCConservative Flux form Galerkin FEM Using GPGPU Computation Clark C251
Speakers: Larry Bush
 11:00AM – 11:20AMHImproving CUDASW++, a Parallelization of Smith-Waterman for CUDA Enabled Devices Clark C144
Speakers: Sanjay Rajopadhye, Douglas Hains, Zach Cashero, Mark Ottenberg, Wim Bohm
 11:20AM – 11:40AMTHigh Resolution Physics-Based Hydrologic Modeling of the Upper Colorado River Basin: The CI-WATER High-Performance Computing Model Clark C146
Speakers: Fred Ogden, Craig Douglas, Scott Miller, Ye Zhang, Li Deng, Wencong Lai
 11:20AM – 11:40AMCRefinement Strategy with Conservative Space-Time Coupling for the Lattice-Boltzmann Method Clark C251
Speakers: S.M.J. Guzik, Xingeng Gao, Todd Weisgraber, Berni Alder, Phillip Colella
 11:20AM – 11:40AMHEfficiently Identifying Underperforming Nodes in Large Clusters Clark C144
Speakers: Thomas Hauser, Jazcek Braden, Monte Lunacek
 11:40AM – 12:00AMTHigh-resolution Ocean Modeling of the Coral Triangle Region Clark C146
Speakers: Joanie Kleypas, Frederic Castruccio, Enrique Curchitser, Zack Powell
 11:40AM – 12:00PMCSimplifying Data Management for HPC with Globus Online Software-as-a-Service Clark C251
Speakers: Steven Tuecke
 11:40AM – 12:00PMHScan Detection and Parallelization in ‘Inherently Sequential’ Nested Loop Programs Clark C144
Speakers: Sanjay Rajopadhye, Yun Zou
 12:00PM – 12:20PMTGrowing Time and Length Scales in Supercooled Liquids Clark C146
Speakers: Elijah Flenner, Grzegorz Szamel
 12:00PM – 12:20PMCGlobally Distributed Datacenters: A Collaborative Peer-to-Peer Approach Clark C251
Speakers: H.M.N Dilum Bandara, Anura Jayasumana
 12:00PM – 12:20PMHHeterogeneous Energy and Makespan-Constrained DAG Scheduling Clark C144
Speakers: H.J. Siegel, Anthony A. Maciejewski, B. Dalton Young, Sudeep Pasricha, Jay Smith
 12:20PM –  1:45PMGLunch- Sponsored by Appro "Why Appro Supercomputers for Research and Education" Morgan Library
Speakers: David Barkai
  1:45PM –  2:45PMKThe Steep Hill to Exascale BSB 131
Speakers: Craig Stunkel

On the road from Petascale to Exascale computing, we are one-third of the way to the destination, having breached the 10 Petaflops barrier. However, this was the easiest part of the journey, and it could be accomplished with relatively conventional architectural and technology advancements. The road ahead is steep, because systems must become 20 times more energy efficient than today’s most efficient systems, but also because of dueling challenges such as resiliency. To complete this steeper portion of the journey, we require innovation at every level of the system, from semiconductor fabrication and packaging to circuit design to processor and system architecture. These “hardware” advances must be paired with innovation in execution models, programming languages, algorithms, and applications to fully realize dramatic gains in energy efficiency and reliability while scaling performance to tens of millions of threads. In this talk we will examine some promising ideas for achieving the Exascale goal, and we will also discuss the usefulness of this “grand” challenge for future non-Exascale systems. To illustrate current progress and to highlight potential paths forward, we will use IBM’s recently introduced Blue Gene/Q system as a case study.

  2:45PM –  3:30PMPPoster PresentationsMorgan Library

University of Wyoming  

Jared Baker       Computational Investigation of the Sensitivity of Spoiler Attachment on Wind Turbine Blades

Hannah Jang-Condell       Disk-O-PaRTY: Disk-Optimized Parallel Radiative Transfer for Young Stellar Objects

 

Colorado School of Mines  

Sophia Wolfenden       Computing the Central Valley: Modeling a Complex Hydrologic System on Multiple Processors

Hannah Menke       Brine and CO2 Leakage Rates in the Intermediate Zone as a result of Geologic Sequestration

Brian Davis       Initial Convergence Study: FEA of SiC Shell of Nuclear Fuel

Conrad Rohleder       Revealing the Binding Specificity of Alpha-Conotoxin MII for Neuronal Nicotinic Acetylcholine Receptors

 

Colorado State University  

Andrew Stone       An Active Library for Specifying Semi-Regular Grid Topologies

Ryan Friese       Analyzing the Trade-offs Between Minimizing Makespan and Minimizing Energy Consumption in a Heterogeneous Environment

Xinfeng Gao*      A fourth-order finite-volume method for compressible Navier-Stokes equations with adaptive mesh refinement

Bhavesh Khemka       A Performance Comparison of Resource Allocation Policies in Distributed Computing  Environments with Random

Failures

Mark A. Oxley       Energy Aware Stochastically Robust Static Resource Allocation

Tim Hansen       Dual-Stage Framework for Robust Scheduling of Scientific Applications in Heterogeneous Environments

 

University of Colorado  

Tim Dunn       Parallel Load Balancing Serial Jobs

Chris Chronopoulos       Parallel ADI with Pre-Propagation and Message Vectorization

 

LLNL  

Stephen Guzik*        A programming model framework based on the Rose compiler to advance heterogeneous computing for engineering

applications.

 

NREL  

Antti-Pekka Hynninen*       Improving CHARMM Molecular Dynamics performance

 

Assimilation Monitoring Project   

Alan Robertson*       How To Assimilate a Million Servers and Not Get Indigestion

 

* = not student   

 

  3:30PM –  4:30PMPPoster SessionMorgan Library
  4:30PM –  5:15PMBChoices in Cyberinfrastructure Morgan Library 147
Speakers: Thomas Hauser, Tim Kaiser, Pat Burns, Anke Kamrath, Craig Tierney, Tim Kuhfuss, Ilene Carpenter

Many-core processors, networking, and novel storage technologies are changing rapidly, and new paradigms like cloud and grid computing continue to transform the way we access high performance computing resources. In this BOF, the FRCRC member institutions will discuss these emerging trends. What’s the right balance? What’s the right approach? Join the debate.

 

 

  4:30PM –  5:15PMBComputational Science in the Classroom Morgan Library 175
Speakers: Bryan Shader

What do instructors need in terms of improvements to HPC environments to support classroom HPC, and what funding and collaborative opportunities are there in this space? What roles are there for distance learning, now and in the future?

 

  4:30PM –  5:15PMBEnergy Research Morgan Library 174
Speakers: Wesley Jones

 

This BOF session will provide opportunities for like-minded attendees to share interests and discuss collaborative opportunities in modeling and simulation for renewable energy research.

 

  4:30PM –  5:15PMBMany Core Developer Forum Morgan Library 173
Speakers: Richard Loft

Interested in emerging technologies such as GP-GPU’s, MIC and FPGA’s? This BOF will explore the landscape and discuss what the needs and collaborative opportunities are for developers/researchers interested in these disruptive technologies.

 

  5:15PM –  6:00PMBData Management Morgan Library 173
Speakers: Anke Kamrath

Got (too much) data? This BOF will discuss data-related issues including best practices for managing the data lifecycle, and tips on moving, analyzing publishing, and yes, deleting data! We'll be asking folks from across the FRCRC to share their data management challenges and solutions and there will be time for the audience to engage in discussions on these topics.  Participants are encouraged to share their efforts in:

  • Data Software Solutions (Tools, Databases, Workflows)
  • Storage Solutions (Local Computing Resources, Use of Cloud Resources)
  • Type of data being managed and/or serviced. Data lifecycle management.
  • Biggest Challenges, Innovations you'd like to share

 

  5:15PM –  6:00PMBEarth System Science Morgan Library 174
Speakers: Richard Loft

This group will provide opportunities for like-minded researchers to share interests and discuss collaborative opportunities in computational geoscience research.

 

  5:15PM –  6:00PMBFuture Directions of the FRCRC Morgan Library 147
Speakers: Thomas Hauser

The Front Range Consortium for Research Computing (FRCRC) is a group of universities and government labs located near a region of the Rocky Mountains known as the Front Range. All member institutions rely on high performance computing to meet their strategic research goals. The mission of the FRCRC is to connect local scientists, students, and HPC experts to improve the use of HPC across the Front-Range, as well as to increase visibility of its members and their HPC activities within the national and global science and engineering community. What do you want FRCRC to focus on? Join the discussion!

 

  5:15PM –  6:00PMBHPC Needs of Students Morgan Library 175
Speakers: Timothy Dunn, Dalton Young

What do students needs in terms of HPC training, system access and support? This BoF will explore these issues from a student’s perspective.

 

  6:00PM –  7:30PMGReception/Awards- Sponsored by IBMMorgan Library

AUGUST 14 • TUESDAY

  7:30AM –  8:30AMGRegistration/BreakfastMorgan Library
  8:45AM – 10:30AMALagged Fibonacci: Parallel Random Number Generator Morgan Library 110Q
Speakers: Pat Burns

This tutorial will begin by presenting an overview of the desirable attributes of random number generators. Usage of pseudorandom numbers in simulations will be covered, with comments and input from the participants. The structure of Lagged Fibonacci Random Number Generators (LFG’s) will be covered, along with strategies for initialization. The attributes of Lagged Fibonacci Random Number Generators (LFG’s) that make them particularly useful for simulations will be covered, along with strategies for avoiding pitfalls. A hands-on exercise using a serial LFG will be conducted. Parallel implementations will be covered next, and performance results from MPI and OpenMP implementations will be presented. A hands-on exercise using a parallel LFG will conclude the tutorial.

 

Students will be expected to grasp pseudo-code, and will be given Fortran and C codes to implement in the hands-on exercises. Students should bring their own laptops, and may use an account on a parallel HPC system, or be given an account on the Cray XT6m at CSU for the parallel exercise.

 

  8:45AM – 10:30AMIHybrid MPI/Open MP Morgan Library 173
Speakers: Monte Lunacek

The Message Passing Interface (MPI) is the standard specification for supporting parallel programming in a distributed environment.  OpenMP is an application programming interface (API) for parallel programming in a shared-memory environment.  Sometimes it is more efficient to combine these two approaches into a program where OpenMP is used within each node and MPI is used across the nodes.

 

In this tutorial, we’ll discuss strategies for writing hybrid parallel programs and demonstrate when this type of approach is advantageous, and when a purely MPI approach is best.  We’ll also show how using a simple MPI load-balancing framework can quickly create a loosely-coupled hybrid program from an existing OpenMP application. 

 

  8:45AM – 10:30AMIProfiling with Tau Morgan Library 203
Speakers: George Carr

The Tuning and Analysis Utilities Performance System (TAU) is a toolkit supporting performance analysis of Fortran, C, C++, JAVA, Python, and GPU applications. The toolkit is available on almost all HPC (and some other) computational platforms facilitating a portable, apples to apples performance methodology. TAU supports analysis of MPI and OpenMP applications, call trees, source instrumentation, compiler based instrumentation, memory usage analysis and includes graphical analysis tools. 

TAU, like MPI, has a large number of options but you only need to know a few to gain a lot of value from the use of TAU. With most applications, you can be up and running with results in hand within a couple hours (plus the run time of your application). This quickstart will introduce a cookbook approach to using TAU.

 

 

  8:45AM – 10:30AMIUsing Globus Online Morgan Library 147
Speakers: Steven Tuecke
  8:45AM – 10:30AMNHPC 101 Morgan Library Event Hall
Speakers: Tim Kaiser

This tutorial will cover all three time slots.

 

In the first session we will discuss the importance of parallel computing to high performance computing. We will by example, show the basic concepts of parallel computing. The advantages and disadvantages of parallel computing will be discussed. We will present an overview of current and future trends in HPC hardware.

 

The second session will provide an introduction to MPI, the most common package used to write parallel programs for HPC platforms. As tradition dictates, we will show how to write "Hello World" in MPI. Attendees will be shown how to and allowed to build and run relatively simple examples on a consortium resource.

 

The third session will briefly discuss other important HPC topics. This will include a discussion of OpenMP, hybrid programming, combining MPI and OpenMP. Some computational libraries available for HPC will be highlighted. We will briefly mention parallel computing using graphic processing units (GPUs).

 

 10:30AM – 10:45AMGBreakMorgan Library
 10:45AM – 12:30AMNHadoop 102 Part I Morgan Library 110Q
Speakers: Karina Hauser

This workshop will cover two time slots.  The overall goal of the workshop is to provide an overview about Hadoop, an open source software framework for large scale data processing.

 

Part 1 will give a short introduction about the Hadoop framework, where it came from, the differences to traditional HPC and its applications in different industries. We will create a single node Hadoop installation on a Linux system (PCs will be provided or bring your own Laptop). 

 

 10:45AM – 12:30PMAAdvanced Batch Scripting for Parallel Environments Morgan Library 203
Speakers: Tim Kaiser

This tutorial will cover a number of "tricks" for batch scripting for parallel environments.  We will concentrate on the PBS/Torque/Maui/Moab family of scheduling environments but most of the subjects will be applicable to SGE, LoadLeveler and other environments. 

 

Some of the topics covered will include: 

  • Chaining jobs and using job dependencies
  • Running multiple instances of parallel and serial applications, both at the same time and in series
  • Explicit mapping of tasks to nodes for:
    • o Varying task count and layout per node    
    • o MPMD programming    
    • o Mappings for hybrid applications 
  • Automatic notifications and archiving of scripts 
  • On the fly creation of directory structures 
  • Using scheduler and shell variables  

 

 10:45AM – 12:30PMIOpen ACC Morgan Library 173
Speakers: Mark Ebersole

OPEN-ACC is a directive based approached to programming GPU's and other many-core architectures. Modern GPUs have grown past their graphics heritage and evolved into the world's most successful parallel computing architecture.  The introduction of this talk will briefly cover where the GPU came from and how it turned into this processing powerhouse.  We will then touch on the programming model used to access this computational power using the CUDA C programming language.  From there we'll move into an alternate method of programming the GPU using the OpenACC compiler directive standard and go through some example code in C and Fortran.  Finally, we'll cover some of the extremely powerful profiling tools, debugging tools, and libraries available to help you get the most out of accelerating your application on the GPU.

This tutorial will give attendees a basic grounding in using this programming technique in their applications.

 

30 minutes - NVIDIA GPU Architecture/programming models
30 minutes - Putting OPEN-ACC Directives in your code.
30 minutes - Libraries and Profiling/Debugging Tools

 

 10:45AM – 12:30PMISystem Performance Modeling Morgan Library 147
Speakers: H.J. Siegel

This tutorial considers a heterogeneous parallel computing system and corresponding workload being investigated by the Extreme Scale Systems Center (ESSC) at Oak Ridge National Laboratory (ORNL). The ESSC is part of a collaborative effort between the Department of Energy (DOE) and the Department of Defense (DoD) to deliver research, tools, software, and technologies that can be integrated, deployed, and used in both DOE and DoD environments. The heterogeneous parallel system and workload described here are representative of a prototypical computing environment being studied as part of this collaboration. Each task can exhibit a time-varying importance or utility to the overall enterprise. In this system, an arriving task has an associated priority and timeliness. The priority is used to describe the importance of a task, and timeliness is used to describe how soon the task must be executed. These two metrics are combined to create a utility function curve that indicates how valuable it is for the system to complete a task at any given moment. This research focuses on using time-utility functions to generate a metric that can be used to compare the performance of different resource schedulers in a heterogeneous parallel computing system. This tutorial will include: (a) a mathematical model of a heterogeneous parallel computing system where tasks arrive dynamically and need to be assigned based on their priority, timeliness, utility characteristic class, and task execution type; (b) the use of priority and timeliness to generate time-utility functions that describe the value a task has at any given time; (c) the derivation of a metric based on the total utility gained from completing tasks to measure the performance of the computing environment; and (d) a comparison of the performance of resource allocation heuristics in this environment.

 

 10:45AM – 12:30PMNHPC 101 Part II Morgan Library Event Hall
Speakers: Stève Toléqué

This tutorial will cover all three time slots.

 

In the first session we will discuss the importance of parallel computing to high performance computing. We will by example, show the basic concepts of parallel computing. The advantages and disadvantages of parallel computing will be discussed. We will present an overview of current and future trends in HPC hardware.

 

The second session will provide an introduction to MPI, the most common package used to write parallel programs for HPC platforms. As tradition dictates, we will show how to write "Hello World" in MPI. Attendees will be shown how to and allowed to build and run relatively simple examples on a consortium resource.

 

The third session will briefly discuss other important HPC topics. This will include a discussion of OpenMP, hybrid programming, combining MPI and OpenMP. Some computational libraries available for HPC will be highlighted. We will briefly mention parallel computing using graphic processing units (GPUs).

 

 10:45AM – 12:30PMNVisualization with VisIT Morgan Library 171
Speakers: Thomas Hauser

This tutorial will provide an overview on the capabilities of the free interactive parallel visualization and graphical analysis tool VisIt .

 

VisIt contains a rich set of visualization features so that you can view your data in a variety of ways. It can be used to visualize scalar and vector fields defined on two- and three-dimensional (2D and 3D) structured and unstructured meshes. VisIt was designed to handle very large data set sizes in the terascale range and yet can also handle small data sets in the kilobyte range. Users can quickly generate visualizations from their data, animate them through time, manipulate them, and save the resulting images for presentations. VisIt was developed by the Department of Energy (DOE) Advanced Simulation and Computing Initiative (ASCI) to visualize and analyze the results of terascale simulations.

 

Attendees are encouraged to install VisIT to be able to follow along in this tutorial:

https://wci.llnl.gov/codes/visit/download.html

 

 

 12:30PM –  1:30PMGLunchMorgan Library
  1:30PM –  3:15PMAHigh Throughput Computing on HPC Systems Morgan Library 173
Speakers: Monte Lunacek

It is often desirable to use an HPC system to run several serial jobs (e.g. parameter scans).  In this tutorial, we will discuss two efficient ways of accomplishing this.  One of the methods we'll discuss uses iPython's architecture for parallel and distributed computing.  We'll discuss the different types of parallel applications that can be developed and show some examples of task-farming.  The other method we present is a load balancing tool that we have developed that allows you to specify your jobs in a text file that is then read and executed across the resource you request.  You can generate the text file using any scripting language you like.  Using either tool reduces the load on the scheduler.

 

  1:30PM –  3:15PMIUsing Earth System Modeling Framework Morgan Library 147
Speakers: Cecilia DeLuca, Gerhard Theurich

The Earth System Modeling Framework (ESMF) is the leading U.S. modeling framework for high performance weather, climate and related applications.  It can be used in multiple ways: 1) to create interoperable component-based modeling systems; 2) as a source of remapping, time management, metadata handling, and other functions that can be used independently of the component constructs; and 3) as a fast, file-based offline generator of interpolation weights for many different kinds of grids.  Grid remapping software includes options for regional or global domains; bilinear, higher order, or conservative interpolation methods; logically rectangular or unstructured grids; masking; and a variety of pole treatments.  A parallel python interface to these grid remapping functions (ESMP) is also available.  This tutorial will introduce the main features of ESMF and lead the user through a few easy ways to get started. 

 

  1:30PM –  3:15PMIWorkflow Management in HPC Morgan Library 203
Speakers: Chris Harrop

The day when a scientist could conduct his or her numerical modeling and simulation research by writing, running, and monitoring the progress of a modest Fortran code or two, is quickly becoming a distant memory. It is a fact that researchers now often have to make hundreds or thousands of runs of a numerical model to get a single result. In addition, each end-to-end "run" of the model often entails running many different codes for pre- and post-processing in addition to the model itself. The codes that comprise the end-to-end modeling system often have complex interdependencies that dictate the order in which they can be run. And, in order to run the end-to-end system efficiently, concurrency must be used whenever possible. Furthermore, these codes are usually run on high performance machines that are notoriously difficult for scientists to use, and which tend to suffer from various types of failures. As machines get larger and larger, the failure rate of software and components increases commensurately. Managing the execution of the end-to-end system is often difficult even for a single run of the end-to-end system. Multiply that by the thousands of runs needed to accomplish a scientific experiment, and you have a very challenging situation. For simulations that must run reliably in real-time, the situation is almost hopeless. A Workflow Management System solves this problem by providing two things, a means by which to describe the various codes that need to be run, along with their runtime requirements and interdependencies, and a software engine for reliably automating the execution of the workflow.  In this Tutorial I will describe how NOAA/ESRL's Workflow Manager can be used to simplify the execution of large-scale simulation experiments.  

 

  1:30PM –  3:15PMNHadoop 102 Part II Morgan Library 110Q
Speakers: Karina Hauser

This workshop will cover two time slots.  The overall goal of the workshop is to provide an overview about Hadoop, an open source software framework for large scale data processing. 

Part 2 will use simple hands-on exercises to illustrate the following Hadoop components: The Hadoop Distributed File System (HDFS), MapReduce, HBase, Hive and Pig. The examples with focus on big data analytics in business.

 

  1:30PM –  3:15PMNHPC 101 Part III Morgan Library Event Hall
Speakers: Tim Kaiser

This tutorial will cover all three time slots.

 

In the first session we will discuss the importance of parallel computing to high performance computing. We will by example, show the basic concepts of parallel computing. The advantages and disadvantages of parallel computing will be discussed. We will present an overview of current and future trends in HPC hardware.

 

The second session will provide an introduction to MPI, the most common package used to write parallel programs for HPC platforms. As tradition dictates, we will show how to write "Hello World" in MPI. Attendees will be shown how to and allowed to build and run relatively simple examples on a consortium resource.

 

The third session will briefly discuss other important HPC topics. This will include a discussion of OpenMP, hybrid programming, combining MPI and OpenMP. Some computational libraries available for HPC will be highlighted. We will briefly mention parallel computing using graphic processing units (GPUs).