CS 651-4/551-3: Real-Time Rendering
Assignment #2

Synopsis: Add support for LOD and visibility calculations to your rendering engine.

Due: As before, this assignment has multiple components and I am requiring you to turn in a checkpoint to gauge your progress and help you stay on track:

    Oct 24, 2002 - Checkpoint: view-frustum culling and simple distance-switched LOD (50%)
    Oct 31, 2002 - Final assignment: Everything else (50%)

Details: For this assignment you will add LOD and visibility calculations to your scene-graph rendering engine.  As a base level, you are expected to implement by the checkpoint:

• Hierarchical view-frustum culling: calculate and store bounding volumes with subtrees in your scene graph, and avoid rendering off-screen objects by testing those bounding volumes against the planes of the view frustum.  Bounding volumes should be correctly updated with object transformations.  To account for animated objects, you may either pick a bounding volume large enough to contain the object at all steps of the animation, or incrementally update the volume.
• Simple distance-switched LOD: Objects may be represented by multiple LODs depending on the distance from the viewpoint to the object.

Doing the above correctly will earn you full credit on the checkpoint.  For the final assignment, you should extend either LOD support or visibility support -- you can of course extend both, but this is not necessary for full credit.  Examples of extensions you could implement include:

• Cell-portal culling - The pfPortals-style portal culling that we described in class is quite straightforward to implement; the challenge here will be to find or build scenes consisting of multiple cells connected by portals. 
• Occlusion culling using NV_OCCLUSION_QUERY - Again, conceptually straightforward.  The dangers here include killing your fill rate by "rendering" too many very large bounding boxes, and starving the GPU by forcing it to wait on the CPU to make visibility decisions.
• Terrain-specialized occlusion culling - Some terrain-specific visibility algorithms render from front-to-back with respect to the viewer, so that the terrain horizon moves up the screen as you draw, and terrain elements or objects below the horizon need not be rendered.  You would want to create a terrain and scene that showed this technique off to advantage...for example, a terrain viewed at eye-level contains a lot more occlusion than a terrain viewed from high above.
• LOD generation - create your own LODs using any of the algorithms described in class, or any other algorithm you invent or read about.  Some LOD algorithms are easy to implement, others more tricky.  Talk to me if you want to work on this.
• Continuous LOD - implement CLOD (a.k.a. View-Independent Progressive Meshing or VIPM) for fine-grained level of detail control.  The challenge here is to represent the data in a way that lends itself to efficient rendering.  The Game Programming Gems and the LOD for Games chapter of the LOD book provide some good resources here.  You will probably want to use an extremely simple algorithm to create your LOD, such as collapsing edges in order of length.
• Terrain LOD - implement a terrain LOD algorithm such as ROAM. 
• Fixed-frame rate LOD control - implement a feedback-based reactive LOD selection approach, or a predictive LOD selector that optimizes the LOD cost-benefit tradeoff as we discussed in class.  The first is so easy that it might not be enough for full credit.  For either of these approaches, you will need to have enough objects and enough LODs to give your system sufficient control over frame rate.

Resources: 3D Studio Max should now be installed on all the machines in the Olsson 002a lab.  I also have a nine-CD collection of 3DSMax models, each with 3 LODs, that should be available by tomorrow (Wednesday Oct 16).  I will send out mail with the location (it will be a Windows share).  Remember that these models are licensed intellectual property and we do not have permission to distribute them.  Maya is another modeling package; right now it does not appear to work in 002a.  If you want to use Maya we should be able to accommodate you in the graphics lab.  As usual, there are a number of valuable resources available on the web and in the various required/recommended books for the course.

Policy on code reuse: As before, you may use code from outside sources as long as you attribute exactly what you used and where you found it, and share any particularly valuable finds with the rest of the class.  You are of course welcome and expected to build on the code that you turned in for assignment 1, though you may well find that you need to rewrite or redesign some portions of the codebase. Obviously if you reuse a significant amount of code that raises the bar for the assignment, and you will be expected to do additional work to get the same grade that you would get if you did all the coding yourself.  As always, if you have any questions about code reuse or attribution, please ask me.

Platform: As before, I expect that your code will compile and run on the machines in 002a, either in Windows or Linux.  If you want to use a particular input device, talk to me first...you may have to lend it to me so I can grade your assignment.

Turning in the assignment: Put all your code and models in a folder, along with a detailed README.txt file that describes what you did, what code you wrote and what code you used from elsewhere, and how to run your assignment.  Zip up the file and e-mail it to me, or put it somewhere I can download it.  Be sure to include workspace and project files so that I can compile your code.  I will read (and grade you on) your source code, so follow good programming practices.  Naming convention: for (my) convenience, please name the zip file and the enclosing folder of your project with just your last name, e.g. "luebke.zip". 

Advice: Again, it's a big assignment in a short time frame, but if you are organized and start early you should find it easier than assignment 1.  Please talk to me if you have any questions or want to discuss your plan for the beyond-the-checkpoint extensions.  I can help you find the information you need to pull it off, steer you away from overly ambitious projects, and talk over the difficulties you may encounter.