Canvas 3d JS Library

I had the feeling after I did most of the porting from Canvas 3D to WebGL the performance dropped. To investigate, I ran our benchmark on two systems. The benchmark is simply a script which renders textured teapots in the canvas tag. On each machine I ran the benchmark using Firefox 3.5 with the Canvas 3D extension and then on Firefox 3.7a1pre which uses WebGL. Here are the results:

CDOT Desktop (New Zealand)
Windows XP Pro
Intel Core 2 Quad CPU @ 2.4GHz
3.24GB RAM
GeForce 8600 GT 512MB

Teapots	Canvas 3D FPS		WebGL FPS
	static	rotating	static	rotating
1	64	64	64	64
9	64	63	64	63
49	30	23	32	26
100	16	13	18	15

Macbook Pro
OS X (10.5.8)
2.2GHz Intel Core 2 Duo
2GB DDR2 RAM
GeForce 8600M GT 128MB

Teapots	Canvas 3D FPS		WebGL FPS
	static	rotating	static	rotating
1	55	55	43	39
9	16	15	11	9
49	4	3	2	2
100	2	2	1	1

The performance on the desktop is improved when using WebGL, but is reduced on the laptop. This could be happening for a variety of reasons, which I’ll need to carefully think over. But one important question which keeps bothering me is “Is this a fair test?”. One frame a second is terrible performance, but when you’re rendering 100 3D objects in a browser, it’s a bit unfair to just state that it’s too slow. I’d love to improve the performance so 100 teapots renders at 30FPS, but realistically that’s probably not going to happen. So, the question is by what factor can we speed up rendering? Frankly, I’m not entirely sure, but I’ll be taking a look at other WebGL demos and comparing to see what kind of performance they are able to provide. If anyone has ideas or suggestions, I’d love to hear them!

As a final note, I think to generate more accurate results, the benchmark needs to be modified or otherwise another more thorough benchmark needs to be created. One which runs at least a series of 100 tests. Each test could include an additional teapot to render. If this is done, graphing the results would much more meaningful. If anyone, especially OSD600 or DPS909 students is interested in doing this, let us know!