Skip Navigation | ANU Home | Search ANU | Search FEIT | Feedback
The Australian National University
Faculty of Engineering and Information Technology (FEIT)
Department of Computer Science
Printer Friendly Version of this Document
High Performance Scientific Computing COMP4300

Lecture Notes

  • Audio of the lectures is available via webct under comp4300 (for everyone regardless of whether you are in comp4300, comp6430, and honours or PhD student). If you have problems accessing this let me know.

Lecture Notes
30: Thu May 31 Final Review Lecture

No lecture notes, no recording. We ran the CEDAM survey, reviewed the course content, and highlighted some areas that were new compared to the previous times this course had run (when it was only 24 lectures). We noted material that had been provided in labs or lectures that had not been examined as part of the assignments or mid-semester quiz. We then reviewed question 1 on the 2004 exam. The material in this question relates to what you did in lab5, and although you don't need to have read it, you would have found it helpful if you had read the paper by Mark Bull that is a link from the lab5 web page. Only one person in the class had looked at this link!! The take home message ....there will be parts of the 2007 exam that are best answers if you have read some of the associated material provided on the course web site!

Expect the final assignment back late next week.

We will have a final exam tutorial on the Tuesday before the exam, where I can take questions and hand back assignments. Details will be on the discussion web page.

Otherwise - study hard, good luck, and it's been fun!

29: Fri May 25 Global Arrays and HPF (1up)
  • shmem.c: MPI2 one-sided communication example
  • shmemf.f: Fortran front end to above
  • Global Arrays
  • HPF
  • HPC++
  • 28: Thu May 24 Putting it all together: Parallel and Grid Computing in the Climate Community, guest lecture by Jay Larson.
  • Automating Climate Science: Large Ensemle Simulations on the TeraGrid with the GRIPhyN Virtual Data System: a paper from last year by Nefedova et al. describing a grid workflow that performs a 200+ member ensemble simulation using a coupled climate model (6 pages).
  • The Earth System Grid II: Turning Climate Datasets Into Community Resources: A paper by Foster et al that talks about what the ESG is (3 pages).
  • Grid Workflow: a quick overview of the Earth System Grid (ESG) that David Bernholdt wrote for distribution at the ORNL booth at SC06 (2 pages).
  • 27: Wed May 23 Grid Computing, guest lecture by Jay Larson.
  • What is the Grid. A three point checklist, paper by Ian Foster
  • Anatony of the Grid: Enabling Scalable Virtual Organizations, paper by Ian Foster, Carl Kesselman and Steven Tuecke
  • 26: Fri May 18 MPI-2 (1up)
    25: Thu May 17 Software Distributed Shared Memory ( 1up)
  • Seminar on DSM, 4pm Monday 21st May by H'sien Wong
  • SCASH
  • Intel Cluster OpenMP
  • TreadMarks web site
  • TreadMarks paper
  • Adsmith reference site
  • Adsmith paper
  • Commercial use of Linda
  • 24: Wed May 16 Memory Consistency Models ( 1up)
    It is likely to take some time for you to fully appreciate the material presented in this lecture, but it is fundamental to concurrent/parallel programming and should make you reconsider what you did in COMP2310.
  • Shared Memory Consistency Models: A Tutorial Sarita V. Adve. Kourosh Gharachorloo
  • 23: Fri May 11 Instruction level Parallelism and GCC (open office version)
    This will be a lecture by Ben Elliston on Instruction Level Parallelism and GCC.
    22: Thu May 10 Assignment 2 discussion
    21: Wed May 9 Finish lectures on pthreads
    20: Fri May 4 OpenMP in GCC
    Guest lecture by Ben Elliston from IBM Linux Technology Centre on how GCC compiles OpenMP code.
    19: Thu May 3 The OpenMP programming model ( 1up)
  • The OpenMP web site
  • 18: Wed May 2 Shared Memory; Programming ( 1up)
    We started this last Friday, but did not get that far. I've updated the slides slightly since Friday.
    17: Fri Apr 27 Shared Memory; Programming ( 1up)
    See Chapter 8 of Wilkinson and Allen and Chapter 7 of Grama, Gupta, Karypis and Kumar.
    16: Thu Apr 26 Shared Memory; Hardware ( 1up)
    See Chapter 2 of Grama, Gupta, Karypis and Kumar.
    15: Thu Apr 5 Load Balance and Termination (1up)
    Chapter 7 of Wilkinson and Allen. Their slides are available at the web site.
  • Slides from Wilkinson and Allen
  • *: Wed Apr 4 The Quiz!
    14: Fri Mar 30 Pipeline Parallelisation (1up)
    Returning to Chapter 5 of Wilkinson and Allen. Their slides are available at the web site.
  • Slides from Wilkinson and Allen
  • 13: Thu Mar 29 We will complete our discussion of assignment 1, in particular the fast fourier transform. Also take any quesitons concerning assignment 1. If there is spare time we will move on to pipeline parallelisation (Friday Lecture Material).
    12: Wed Mar 28 Synchronous Computations and Barriers (1up)
    Jumping ahead slightly to Chapter 6 of Wilkinson and Allen since this material will be part of the lab this week. Their slides are available at the web site.
  • Slides from Wilkinson and Allen
  • *: Fri Mar 23 Overview Lecture and tour of the APAC National Facility by Jay Larson. Meet Meet at 3.15pm in the Leonard Huxley Theatre. It will take you about 15mins to walk there from the Computer Science side of campus.
    11: Thu Mar 22 Assignment Lecture (1up). You may have noticed that assignment 1 is now available in draft form. In this lecture we will go over some background material related to Fourier transforms. And, before you panic(!), to do the assignment you do not need an indepth understanding of Fourier transforms - you need to do something with a matrix vector product and a recursive algorithm.

    This assignment relates to material presented in Chapter 12 (Image Processing) of Wilkinson and Allen. Slides for this chapter of their book are available at the web site.

  • Slides from Wilkinson and Allen
  • 10: Wed Mar 21 Parallelisation via Data Partitioning (1up)
    We completed this material, although a little rushed on the last section concerning the Barnes-Hutt approach. You can read more about this and the material covered today in chapter 4 of Wilkinson and Allen. Their slides are available at the web site.
  • Slides from Wilkinson and Allen
    See also
  • Paper by Salmon and Warren on N-Body Parallelisation
  • 9: Fri Mar 16 We spent a bit of time talking about this week's lab. I put up some of the ping-pong data that I had when running with 32 processes, and related the data to the architecture of the AC. We also talked about blocking and non-blocking communications and the transition that occured in one of the lab exercises. We discussed latency and bandwidth, and how asymptotic bandwidth is only achieved at long message length.
    We also finished the discusion on embarassingly parallel problems.
    8: Thu Mar 15 Embarassingly Parallel Problems (1up)
    See chapter 3 of Wilkinson and Allen, and the slides available at their web site.
  • SETI
  • Folding@home
  • Slides from Wilkinson and Allen
  • 7: Wed Mar 14 Will complete performance models and then discuss the APAC machine and labs for this week. There are no formal notes. I will use material from the following links.
  • APAC Hardware, we will talk about the Altix hardware.
  • SGI configuration options
  • SGI numalink 4 white paper
  • Altix Data Sheet from SGI
  • SGI Global Shared Memory
  • SGI Numalink
  • APAC Course Material: we will go through the introduction to the APAC National Facility Quickly
  • The APAC User Guide
  • 6: Fri Mar 9 Performance Models (1up)
    Measuring and modeling performance. Mr Amdahl and more!

    Part of this material is taken from 6-13, and 62-74 of Wilkinson and Allen. But also from Chapter 5 of Grama, Gupta, Karypis and Kumar.

    5: Thu Mar 8 Basic Message Passing (1up)
    A quick introduction to MPI, there will be more in lab 1. Some references:
  • MPI forum
  • Online MPI-1 Book This lecture covers part of chapter 2 in Wilkinson and Allen. During the lecture we went through the MPI example on page 60 of the text. You can find this example as slide 43 of the MPI slides available at the web site associated with this book
  • Slides from Wilkinson and Allen
  • 4: Wed Mar 7 Basic Communications (1up)
    This material lays the ground work for message passing and should make you appreciate the complexities of writing a good message passing library for a given platform. See parts of chapter 2 and 4 in the book by Grama et al for more info.
    *: Fri Mar 2 No lecture due to the storm!
    *: Thu Mar 1 No lecture due to the storm!
    *: Wed Feb 28 No lecture due to the storm!
    3: Fri Feb 23 Continued from Thursday lecture, completing material on parallel hardware. See parts of chapter 2 in the book by Grama et al for more info.
    2: Thu Feb 22 Overview of Parallel Hardware (1up)
    A quick review of single CPU hardware followed by an overview of parallel hardware issues. Material presented here will be picked up on in later lectures. ...we finished single processors in this lecture. If you want to know more about single processors look at either the book by Dowd and Severance or the one by Bryant and O'Hallaron given on the reference page.
    1: Wed Feb 21 Unit Introduction (ppt)
  • APAC
  • APAC National Facility
  • Bunyip
  • The Earth Simulator