Ümit V. Çatalyürek

OSU | citeseer | DBLP
Home
Partitioning | Clustering | Scheduling | Coloring | Semantic Graphs | Middleware
Advising | BMI 731 | ECE 864 | ECE 694J
2007 | 2006 | 2005 | 2004 | 2003 | 2002 | 2001 | 2000 and earlier
PaToH | Zoltan | DataCutter | STORM | MSSG
Family | Hobbies
Research | Hardware&Game | Bargain | Others

ECE 694J - Scientific Computing on Emerging Architectures

a c b

3 credits, MWF 3:30pm, Call # 22325-7

Instructor: Prof. Umit V. Catalyurek

 

Course Objective and Learning Goals

The students will be introduced to the architecture and system design of emerging architectures. The course will focus on the reasons for the architectural idiosyncrasies introduced with these architectures, on the programming difficulties associated with these architectural qualities, on techniques for managing these difficulties, and how best to augment/alter the designs of these systems to minimize the programming cost of future revisions of the hardware.  Additionally, the course will act as a primer for students to emerging architectures, such that they become exposed to the exciting developments in the field and might conduct future research with these technologies in mind.

 

Prerequisites

Prereq: ECE 662 or CS&E 675.01 or CS&E 675.02.

Prereq by topic: C/C++, parallel programming knowledge.

 

Text

No textbook is required.

Suggested reading:

CUDA 2.0 Programming Guide

IBM SDK 3

 

Week

Topic

1

Introduction to Architectural Trends, Introduction to CUDA

2

CUDA API, Example Matrix Multiplication

3

GPU/CUDA Threading Hardware, Memory Review: Shared Memory, Data Layout, Memory Conflicts

4

CUDA Optimizations: Memory, Control Flow, Arithmetic

5

No Class; Work on homework, Paper Presentation & Term Project

6

Student Paper Presentations

7

Introduction to Cell BE & Cell SDK: Motivation, Architecture, and Logistics

8

Cell BE Programming: 'Hello World' variants, HPC Programming Models, Cell BE Tools and Libraries

9

IBM Cell Simulator, Applications, and Published papers

10

Students term projects presentations

Exam

Term project reports are due

 

 

Grading

Component

Contribution to Overall Grade

5 Homeworks

30%

Paper presentation

15%

Progress Report

10%

Term Project Report

30%

Term Project Presentation

15%

 

General Comments:

  • Lecture notes, details of assignments, drop boxes: visit http://carmen.osu.edu
  • No Class:
    • Jan 19th: Martin Luther King Day
    • The week of Feb 2nd: work on homework, paper review and term project!
    • Monday March 2nd
  • Assignments are due to beginning of the class on the due date.
    • Late Submissions: 5% after 3:30pm, additional 10% for each day

 

Assignments:

  • Homework #1: CUDA - Matrix Multiplication. Due: Fri Jan 23, 2009.
  • Homework #2: CUDA - Profiling Array Reversal. Due: Fri Feb 6, 2009.
  • Paper Review. Due: Wed Feb 11, 2009.
  • Homework #3: CUDA – Co-occurrence Matrix Computation.
  • Homework #4: Cell – Matrix Multiplication.
  • Homework #5: Cell – Co-occurrence Matrix Computation.

 

©2006 Umit Catalyurek. Last Update: July 20th, 2006