Advanced Parallel Graphics Number: CSD 15-668 Location: WEH 5320 Instructor: Adrien Treuille TA: Duncan Boehle Time: TR 1:30 - 2:50pm

Summary

This course covers the use of programmable GPU hardware to solve hard graphics problems. The goal is to teach techniques to parallelize algorithms using CUDA—while working on fun projects, and practicing quick problem solving and public presentation skills.

Overview

The class is made up of group discussions and lectures. Each class starts with a question—essentially challenging you and your classmates to solve the problem better than the existing techniques. Then, the lecture will discuss the simulation topic and how it relates to parallel GPU programming. Grading is computed as follows: (5%) Project 1. (5%) Project 2. (10%) Project 3. (25%) Project 4. (10%) Project 5. (10%) Project 6. (25%) Project 7. (10%) Class Presentation / Participation. Grading for late projects must be negotiated with me before the project is due.

Project 1 Winners

P1 contest winners: Prettiest fractal viewers (Click on any image to see the full resolution) 1st Place (tied) -- Nikita Nesterov 1st Place (tied)-- Andrew Wesson 2nd Place -- Matthew Kaemmerer

Project 7 Winners

P7 results: Most original/interesting research Ross Adelman: 5 Computers viewing the same rigid body simulation Download full-resolution video here Alfred Barnat: Distributed SPH simulation Download full-resolution video here

Syllabus

Note: This syllabus may change during the course. Keep checking back.   Data Parallelism Name: Overview Date: Tue 01/11 Slides: Lecture 01 Slides Name: Data Parallelism Date: Thu 01/13 Slides: Lecture 02 Slides Name: No Class (Adrien out of Town) Date: Tue 01/18 Name: No Class (Adrien out of Town) Date: Thu 01/20 Name: Graphics Architectures Date: Tue 01/25 Slides: Lecture 03 Slides Notes: Pixel Buffer Objects and CUDA Project: Project 1 Assigned: Fractals Project: Project 1 Description Project: Project 1 Skeleton Code Name: Linear Systems Date: Thu 01/27 Slides: Lecture 04 Slides Name: GPU Caching Date: Tue 02/01 Slides: Lecture 05 Slides Project: Project 1 Due Name: Poisson Blending Date: Thu 02/03 Slides: Lecture 06 Slides Project: Project 2 Assigned: Blending Project: Project 2 Description Project: Project 2 Extra Credit Project: Project 2 PBO Manual Project: Project 2 Skeleton Code Name: Graph Cuts Date: Tue 02/08 Slides: Lecture 07 Slides Project: Graph Cuts on CUDA Project: Max Flow Name: Graph Cut Image Analysis Date: Thu 02/10 Project: Project 4 Assigned Project: Project 4 Description Name: More Graph Cuts! Date: Tue 02/15 Slides: Lecture 09 Slides Project: Project 2 Due Name: Prefix Sums Date: Thu 02/17 Slides: Lecture 10 Slides Notes: Additional Prefix Sums Reading Name: Parallel Sorting Date: Tue 02/22 Slides: Lecture 11 Slides Name: Fun GPU Tricks Date: Thu 02/24 Name: P4 Presentations Date: Tue 03/01 Name: P4 Presentations Date: Thu 03/03 Name: No Class (Spring Break!) Date: Tue 03/08 Name: No Class (Spring Break!) Date: Thu 03/10   Amortized Graphics Name: No Class (Adrien out of Town) Date: Tue 03/15 Name: Amortization Architecture Date: Thu 03/17 Slides: Lecture 16 Slides Name: Light Fields Date: Tue 03/22 Slides: Lecture 17 Slides Project: Project 5 Assigned: Light Fields Project: Project 5 Description Project: Project 5 Skeleton Code Name: Dataflow Parallelism Date: Thu 03/24 Slides: Lecture 18 Slides Name: Queuing Theory Date: Tue 03/29 Project: Project 6 Assigned: Amortized Light Fields Project: Project 6 Skeleton Code Name: Precomputed Radiance Transfer Date: Thu 03/31 Slides: Lecture 20 Slides Name: Image + Video Compression Date: Tue 04/05 Slides: Lecture 21 Slides Project: Project 5 & 6 Due Project: Project 7 Assigned Project: Project 7 Description Name: Model Reduction Date: Thu 04/07 Name: Guest Talk: GPGPU Programming in cl.oquence Date: Tue 04/12 Name: No Class (Carnival) Date: Thu 04/14 Name: No Class P7 Preparation day Date: Tue 04/19 Name: P7 Presentations Date: Thu 04/21 Name: P7 Presentations Date: Tue 04/26 Name: P7 Presentations Date: Thu 04/28