Graphics Architecture, Winter 2009 John Owens
-
- Technology
-
UC Davis course EEC277 introduces the design and analysis of the architecture of computer graphics systems. Topics include the graphics pipeline, general-purpose programmability of modern graphics architectures, exploiting parallelism in graphics, and case studies of noteworthy and modern graphics architectures.
-
- video
Parallelism
In our final content lecture, we look at how to parallelize the graphics pipeline. What is challenging about parallelizing the GPU? What are the ways we could parallelize it? We discuss the sorting taxonomy of parallelism strategies, look at different ways to communicate within a multi-node system, and analyze the taxonomy using historical graphics architectures.
-
- video
GRAMPS: A Programming Model for Graphics Pipelines and Heterogeneous Parallelism
Jeremy Sugerman from Stanford describes GRAMPS, a programming model for graphics pipelines and heterogeneous parallelism.
-
- video
Juggling the Pipeline
We turn away from a fixed-function graphics pipeline and explore what we can do with a user-programmable pipeline, where not only pipeline stages but also the structure of the pipeline can be customized. We look at Reyes, delay streams, and the programmable culling unit.
-
- video
Overflow from the Pipeline Lectures
This lecture contains the overflow from the 4 pipeline lectures, mostly the composition/display lecture.
-
- video
Composition/display
The final stage of the graphics pipeline is composition/display. In this lecture we look at antialiasing algorithms, compositing, the depth buffer, and monitors. [Note: The beginning part of this lecture is the remainder of the rasterization lecture, and this lecture spills into the overflow lecture.]
-
- video
GPU Graphics and Compute Architecture
John Nickolls, chief compute architect for NVIDIA's GPUs,
discusses NVIDIA GPU graphics and compute architecture.
Customer Reviews
Great Course for a Niche Subject
The content in this course isn't a typical area of study in computer science programs at univerisities, but it has large research and commercial implications. Often, there is a course for graphics and a computer architecture course (mainly CPU design). There isn't a whole lot of information about graphics hardware design on the internet due to the secretive nature of graphics card companies, which is kind of unfortunate. Even though this course is probably getting dated, it seemed like most of the information is still relevant (especially since Larrabee didn't take off).
First, a brief summary about what is talked about in the course. You will learn about the GPU pipeline in a more architectural way than a regular computer graphics course. There's information on GPGPUs and how to effectively use them. It seemed from the beginning of the course that that would be the heavy focus, but it turned out that a lot of the course is concerned with graphics rather than unbounded computation using GPUs. I would highly recommend watching/taking a graphics course before this course. While it's not completely necessary, it would be helpful since you would get some of the same knowledge taught from a different perspective. Taking a computer architecture course would also be recommended so that you can see how CPUs and GPUs actually have pretty big differences.
The instructor is fantastic and very knowledgeable and paces the course fairly well (although the last few lectures seemed a little rushed). The other criticism I had was that certain parts such as antialising seems a little out of place in the context they were discussed. I wouldn't have minded some example runs of some of the graphical techniques as well (for example, going through the whole pipeline start to finish with a few example meshes in a scene or something along those lines). The connection of each stage felt too much like a black box sometimes.
The students and the questions they asked were great and really helped me understand things a little better, and they didn't hold anything back when asking the guest speakers tough questions about design decisions. The guest speakers were nice to have as well since they were industry experts (minus the last one).
Looking back, the professor seemed to have high hopes for Intel's solution to graphics hardware called Larrabee, but that actually failed to come to market due to poor performance (about a year or so after this lecture took place).
Overall, I highly recommend this course. If anything, this seems to be the most coherent set of information about graphics hardware available, and it's well done. Thanks for posting this course up!
glReview
Only watched the first two so far, seems very good so far...
cann
err=-3259