GPU610/DPS915

GPU610/DPS915 | Student List | Group and Project Index | Student Resources | Glossary

Please help make this page resourceful for all GPU610/DPS915 students to use!

Course Descriptions

GPU610 - Parallel Programming Fundamentals

• Modern GPU (Graphics Processing Unit) technology supports massively parallel computations, which complements the serial processing capabilities of CPU technology. This course teaches students how to read, write, and debug programs that use both CPU and GPU technology. Students learn to reorganize existing programs into serial code that runs on the CPU and parallel code that runs on the GPU. Students also study cases that have benefited from CPU+GPU programming.
• Course Outline

DPS915 - Introduction to Parallel Programming

• Modern GPU (Graphics Processing Unit) technology supports massively parallel computations, which complements the serial processing capabilities of CPU technology. This course teaches students how to read, write, and debug programs that use both CPU and GPU technology. Students learn to reorganize existing programs into serial code that runs on the CPU and parallel code that runs on the GPU. Students also study cases that have benefited from CPU+GPU programming and develop a CPU+GPU application for a client.

• Course Outline

Common Material

External Links

• Course Web Site – Lecture Notes
• Course Web Site – Timeline
• Class Samples

The Workshops

• The workshops provide timely opportunities to implement some of the material covered during the lectures. Each workshop is graded and all submissions are through Moodle.
• Detail Specifications
  1. Initial Assessment
  2. Linear Algebra using BLAS
  3. Device Query and Selection
  4. A Simple Device Operation
  5. Dot Product
  6. Matrix Product
  7. Matrix Product using Thrust
  8. Matrix Product using cuBlas
  9. Matrix Product using Streams
  10. CUDA to OpenCL

The Project

The course project is a three-stage, team assignment. Each team consists of 3 members. In the first stage your team evaluates 6 applications and selects 3 for continued work. The evaluation includes profiling to identify the hot spots in each application. Each team member is responsible for 2 of the candidate applications. The second stage refactors the applications to use the GPU, including shared memory. The third and final stage optimizes the performance. Each team presents the results of its work during the final week of the semester. Detail Specifications Selection and Assessment GPU Programming Optimization Grading The penalty for late submission is 30% of the assignment mark. The penalty for resubmission, in the event that the original submission was not workable is 50%. The due dates are posted in Moodle. All submissions are to be made through Moodle.

Evaluation

• Assignment 30%
• Workshops 20%
• Test 20%
• Exam 30%

Final Submission Requirements

• Under construction

Resources

• Software Support
  • Microsoft Visual Studio 2010 Pro
  • CUDA Toolkit
  • NSight Visual Studio Edition
  • NSight Eclipse Edition
• Wikis
  • How To edit Wiki pages
  • How To edit Wiki Cheatsheet
• Subversion
  • Subversion (SVN)
  • Download Page of TortoiseSVN
  • TortoiseSVN Documentation
  • SVN book at red-bean.com or download the PDF from here.
  • AnkhSVN - Free Visual Studio SVN Integration Alternative To VisualSVN