CS 575 Supercomputing for the Sciences

Text: High Performance Computing, Second Edition,
Kevin Dowd & Charles Severance, O'Reilly & Associates, 1998
http://www.oreilly.com/catalog/hpc2/index.html
Text Table of Contents

Class Calendar September/October/November/December

Assignments for Class [3Nov03 upd with Deadline Dates]

Email repository Fall 03

What are Supercomputers?

1986 ByteBuyer SDSC cover story on this widely circulated local flyer (now called Computer Edge) gives details on the Cray X-MP, the first supercomputer at San Diego Supercomputer Center. (thanks to Keith Thompson) local copy. Since 1986 was nearly two decades ago, you are likely to have this level computing power on your desktop at home these days.

IEEE History of Computing [celebrating 50 years of computers, 1996]

You may find the following, online discussion useful as an overview to the field of Scientific Computing.

Scientific Computing overview by Lloyd Fosdick & Elizabeth Jessup (University of Colorado, Boulder)

This Class (CS 575 Supercomputing):

Prerequisites are a programming background in C or Fortran 90

Problem solving skills will be enhanced through computational experiments that will involve both individual and group (collaborative) learning exercises. This emphasizes how science is done these days - involving collaboration, investigation and communcation.

The computers at SDSC are fast number crunching machines whose floating point performance is exceptional. Single-processor and multiple-processor architectures will be introduced, with parallel programming using MPI, the Message Passing Interface.

Care is taken to

* introduce UNIX (at a functional level),

* develop students' familiarity with local computing resources (Rohan) and software tools (timers, profilers, Answerbook, and more),

* develop students' understanding of the concept of performance and how to measure and interpret it

Assignments (3Nov03 - updated with dates)

Beginning Calendar for Class Activities

1. September 2003
2. October 2003
3. November 2003
4. December 2003

8Sept03 (Lecture in BAM 251):

Monday's lecture will cover Chapter 2 Ch2 HPC of text, which you should read beforehand.
You should also read the first pages of Chapter 6 on Timing. We will discuss Chap. 6 "Profiling" later in the semester.

10Sept03 (Lab in BAM 110):

Class will be held in Xterm lab BAM110 where your class ROHAN accounts will be distributed and we will begin the introduction to computing the HPC way.

Lab 1 Introduction to Xterms and Timers:
This exercise includes the start of your first computational experiment, which is due on 24 Sept 03. You will be expected to use the web-page template and gather appropriate data using the sample program provided in lab.

15Sept03 (Lecture in BAM 251):

More on performance today (Performance for HPC), to connect our text Chapters 2 and 6 with our lab exercise last week. We will also discuss Problem Solving Skills, which includes programming exploration, validating computed data, writing skills, and conclusion-driven experimentation.

17Sept03 (Lab in BAM 110):

Ensure clarity for first computational experiment and report.

22Sept03 Ch.3 Memory

The new version of Moore's Law is used to describe the growth in memory accompanied by its reduced cost. How is memory organized in a HPC. What is memory stride?

24Sept03 Information Overload

Class materials are made available from the class web site. You, as a student at SDSU, may also have your own personal web page on Rohan. Student Home Pages at SDSU

How does a person come to value this large amount of information? What are the clues that can guide the informed Internet user?

29Sept03 Ch 4 Floating Point Numbers

The representation of numbers and characters and other forms of data on digital media has evolved since the days of punch cards. Analog data was a basis for measurements, which, in turn, had to be stored. For example, the thermometer can measure body temperature which can be entered on your medical records file by the technician.

01Oct03 - BAM 110 Lab3: Floating Point and the Compiler

To continue our exploration of floating point numbers, we will work through some hands-on examples that will begin the preparation for your second computational report

06Oct03 Ch 5 What a Compiler Does

08Oct03 Diffusion Model for Performance Report 2 (Due: 29Oct03)

We will discuss the background information you will need to successfully explore your computational experiment and report. The programs you will build your next experiment on will be made available on Monday. It is important to establish the background in today's lecture.

13 October 03 (Columbus Day)

We will discuss Chapter 6 of our text today, Ch 6 Timing and Profiling, with links to online Sun Documents you may find useful. In lecture we also present a sample code to use for the gprof example, which are the codes for the Statement of Second Report.

15 October 03

Returning the report for the first computational experiment, it is appropriate to examine the instructor's Gradine Rubric. We also want to discuss Chapter 7 Eliminating Clutter

20 October 03

Guest lecture - on What is Data Intensive Computing?. Stewart must travel to Washington D.C. Kirsten Barber, Computer Resource Specialist

22 October 03

Ch8 Loop Optimization. Loops are the first step towards parallelism.

---

SDSU Campus closed 27-29 Oct 03 Due to County Fires

---

3 November 03

Ch9 Understanding Parallelism.

Today, the groups will self-select themselves to begin preparation for class discussion on Computer Network Security on 12 November 03. Form groups of three (3) and make plans for the topic you wish to prepare and present in 15 minutes. You should find class members with schedules compatible with yours and skills complementary to yours since you will be asked to work together on the group Rohan coding assignment and final group Blue Horizon coding assignment.

5 November 03 Linear Systems and Parallel Programming

Begin development of group programming assignment using MPI on Rohan. Highlight lectures on Ch 10 (Shared-Memory Multiprocessors), Ch 11 (Programming Shared-Memory Multiprocessors)

10 November 03

Section IV Text: Scalable Parallel Processing. Lecture will introduce Ch 12 Large-Scale Parallel Computing along with MPI on Rohan.
Ch 13 Language Support for Performance and Ch 14, Message-Passing Environments - MPI (Message-Passing Interface) will also be introduced, briefly.

NPACI Parallel Computing Institute - Stommel Model to complement the recent email Group Report AutoParallel (10Nov03)

12 November 03 15 minute Group presentation on topic related to Computer Network Security

17 November 03 Visualization and High Performance Computing (Guest Lecture by Jeff Sale, Staff Scientist, NPACI Ed Center on Computational Science & Engineering)

19 November 03 Groups work together on their group class project which will use MPI on Rohan.

24 November 03
How are groups progressing? Last three weeks of lecture, then final project due midnight 15 Dec 03

• Group Report AutoParallel Assignment - Report due December 3, 2003 (10Nov03)
• SUN SPARC Optimization (referenced in 10nov03 email) Our course will be using "Multiple Processors in Shared Mode" at the end of SUN Chapter 3 document.
• Ch 13 Language Support for Performance
• Tour San Diego Supercomputer Center - 3pm 12Dec03 - class voted
• RockStar at SC2003 128-node 700 GFLOP Supercomputer built from scratch in Phoenix. Congratulations to Prof. Phil Papadopoulos and grad students from UCSD/SDSC.

26November 03

• Logging into Blue Horizon, as you finish up your Rohan explorations. NPACI Hot Page built on the GridPort Secure Technologies
• More on MPI: Ch 14, Message-Passing Environments - MPI (Message-Passing Interface) will be introduced

01 December 03

The Computing Continuum, Guest Lecture by Dr. Sid Karin, Founder of the San Diego Supercomputer Center, in HH 130.