Canvas 3d JS Library

Last weekend I finally committed all the work I’ve been doing with porting the C3DL library to use WebGL instead of Canvas 3D. Below is a screenshot of the library rendering a COLLADA model.

However, there are still some bugs I haven’t yet addressed. One known issue is in the particle system. The particles will randomly change their color every few seconds. Another issue I found was in the Effect class. Using Effects does not work if using two canvases per page. On the other hand, models seem to load and render just fine which means their vertices, normals and texcoords all appear to be set correctly. This means the library is in a state which allows anyone with a nightly build, an SVN client and a copy of our repository to (theoretically) run all our demos!

WFM

In CDOT, I installed minefield on different systems to make sure the library works on a variety of configurations. I have the setup (minefield and an SVN checkout of C3DL) on a macbook pro, a mac mini, and a desktop (New Zealand) running both Fedora and Windows. C3DL demos are running properly, but I’ve also been hearing users are experiencing a wide variety of problems with 3.7a1pre and C3DL. I think for our next release we’ll be able to test on at least 15 different machine configurations. That sounds pretty optimistic, but I believe some of the students from Seneca’s OSD600 and DPS909 classes can help us out. This will increase the chances of the library functioning properly on most systems.

Processing.js

One of the reasons I committed as soon as I could is because students from the OSD600 and DPS909 classes may want to look at how we are using WebGL. There are a bunch of us who took on the task of injecting 3D functionality into the Processing.js library. We can either plug C3DL into Processing.js to get 3D working, then strip away the irrelevant objects or start from scratch, incrementally adding WebGL function calls into Processing.js. Dave and Cathy suggested we should take the former route, especially if we don’t want to deal with the lower level WebGL calls. I’m willing to give this a try even though getting two JavaScript libraries to talk to one another can be quite the challenge.

Teapots Per Second

We have a benchmark script which we use to keep track of C3DL’s performance. Unfortunately the only performance records we seem to have kept are very old and benchmarked rotating cubes. Because we switched to rendering teapots in the benchmark a while ago, the results can’t really be compared. We’ll need to restart the logs. Now that the port is almost complete, I’ll benchmark the script using Firefox 3.7a1pre, revert to an older version of the library and test again with Firefox 3.5 with the Canvas 3D extension and see the difference. If we dropped any frames, I’ll need to fix that. I’ll then throw the log somewhere accessible so we can keep aiming to increase the library’s performance from that point.

Next Steps

I’ll be busy mostly fixing any bugs present in the code I just committed.