Course: Parallel Programming and Cluster Computing at OU - August 8

Skip forward to main content

Leandro Avila

Kenneth Craft

Andrew Fitz Gibbon

Tom Murphy

Henry Neeman

Charlie Peck

Enrol me in this course

	Parallel Programming and Cluster Computing at Oklahoma University August 9 - August 15, 2009 Additional Session Survey

1	Sunday, August 9 Introduction: Who Are We, and Why Are We Here? - Charlie The SC Education Program portion of the annual SC Conference focuses on engaging educators and students across the K-20 spectrum in learning about the latest technologies and applications for advancing scientific discovery. The educators are supported in their endeavors to integrate emerging technologies and methodologies into the preparation of the future workforce. The SC Education effort is supported by a close collaboration with several other organizations, namely the National Computational Science Institute, the Shodor Foundation, and TeraGrid. SC Education Program What the Heck is Supercomputing? - Henry This talk will help set the stage for terms and concepts we'll be delving deeper into later in the week. Also, everything you need to know about supercomputing will be shown on a single slide. Intro. to Supercomputing (PDF)

2	Monday, August 10 Time to initialize a drop of water - Tom This talk examines aspects of modeling by doing the back of an envelope calculation to see how long it would take to initialize molecular data in a drop of water. Many simplifications were made. Excel spreadsheet for drop of water calculation Desert Island Analogy - Henry Desert Island Presentation HHeelllloo_WWoorrlldd - Hello_World with MPI - Charlie This talk will walk you through a hello_world program using MPI Hello_World - C Version Hello_World - Fortran Version Greetings - C Version Greetings - C Version With Error Handling Greetings - Fortran Version Class coding of the area under the curve problem - Tom & Fitz We developed the class code trap.c to calculate the area under the curve. Trapezoid Calculation Specification Base-line Trapezoid code

3	Tuesday, August 11 Tyranny of the Storage Hierarchy - Henry Fundamentally, just two issues govern performance: the storage hierarchy, and parallelism. This presentation focuses on the implications of a fundamental reality: fast implies expensive implies small, and slow implies cheap implies large. Topics include: what is the storage hierarchy?; registers; cache; RAM; the relationship between cache and RAM; cache hits and misses; cache lines; cache mapping strategies (direct, fully associative, set associative); cache conflicts; write-through vs. write-back; locality; tiling; hard disk; I/O strategies; virtual memory. A key point in the broader context of HEC is that parallel performance can be difficult to predict or understand without a thorough grounding in the storage hierarchy. Storage Hierarchy Presentation (PDF) Storage Hierarchy Presentation (PPT) What is a Cluster (PDF) What is a Cluster (PPT) Tiling Excercise - C Version Tiling Excercise - Fortran Version C Dynamic array(UPDATED) More MPI, Including Collective Communication - Henry We talk in more detail about some of the MPI plumbing, like tags and types. Also, an introduction to a few more MPI calls that we haven't yet talked about. More MPI Presentation MPI Collective Operations Lab MPI Blocking and Non-Blocking Sends - Tom More on the various benefits and pitfalls of point-to-point communication MPI Send/Receive Blocked/Unblocked Presentation Deadlock Code Examples Deadlock Code - Tom's Version in C Deadlock Code - Henry's Version in C Deadlock Code - Henry's Version in Fortran

4	Wednesday, August 12 BCCD - Bootable Cluster CD Self contained, non destructive and automated cluster environment tool for education. BCCD (Documentation & Downloads) Introduction to OpenMP - Tom Scratching the surface of OpenMP with some key concepts and some code to torture OpenMP Presentation hello.c for.c vers.c ranksize.c timing.c

5	Thursday, August 13 Multicore Madness - Henry Multicore Madness (PDF) Multicore Madness (PPT) GPGPU: Numbercrunching on Your Graphics Card - Henry GPGPU Talk (PDF) GPGPU Talk (PPT) Debugging Parallel Programs - Charlie Debugging parallel programs presentation eratosthenes-serial.c eratosthenes-mpi.c eratosthenes-openmp.c

6	Friday, August 14 Instruction Level Parallelism - Henry Instruction Level Parallelism (PDF) Instruction Level Parallelism (PPT) LittleFe - Charlie LittleFe Overview LittleFe assembly instructions Stupid Compiler Tricks - Henry Stupid Compiler Tricks (PDF) Stupid Compiler Tricks (PPT)

7	Other Resources

The Parallel Programming & Cluster Computing workshop focuses on techniques and tools for parallel computing. Much of this workshop concentrates on distributed parallelism (MPI); in addition, shared memory parallelism (OpenMP), instruction level parallelism, Graphics Processing Unit parallelism and hybrid shared/distributed parallelism are also explored. Participants will learn about developing, debugging, profiling and tuning of parallel applications across a variety of architectures, using tools from a variety of sources, including GNU, Intel, TotalView, and the Bootable Cluster CD. The material is designed for undergraduate faculty from a variety of disciplines who would like to add parallel computing to their undergraduate teaching and research. In addition, undergraduate and graduate students are encouraged to attend alongside a sponsoring faculty member. The workshop is hands-on, with exercises in both programming and curriculum development.

Instructors

Administration

Topic outline

Workshop Description

You are here

Instructors

Administration

Topic outline

Workshop Description