Hello. This is a basic stage, so everything works in the first plane, no extra threads of execution. Thus, if you use some heavy behaviours like "a-antialiasing" or "d-depth of field" it can take a long to complete and the program won't respond until finished. Each key you use activates one behaviour, this key again deactivates it.

Each behaviour activated and every extra level of the sphere flake involve more time consuming. Nevertheless with one light it could be rasonably fast.

I would suggest use the "4" key to reach the fourth level of the sphere flake and hit twice the "L" to turn on three lights (maybe it is best viewed with three lights that with only one), maybe then "r" for reflects.

These are the keys what you can interact with the program:
- a/A. Antialiasing
- d/D. Depth of field
- r/R. Reflection
- l/L. Increase the number of lights (max=3)
- k/K. Decrease the number of lights (min=1)
- 1,2,3,4. The level of iterations for the sphere flake

Well, on my 1GHz CPU, it is pretty slow. But looks great.

Cool, good job!

Cool, nice work. Any details on they type of ray tracing occurring? Did you set a maximum amount of ray bounces? Are you planning on implementing transparency? Do you have a geometry accelerator structure like a BVH, or just a list of spheres (Intel's Embree details some of these)? Single precision for your rays or double (I think Revolution said he noticed artifacts with single precision when ray tracing a long time ago). Also, is the local illumination model Blinn?

Of course, if you want to keep any of the implementation details private, completely understand. I can take down the screenshot upload even.

I'll be attempting to implement probably a path tracer (ray-tracer vs path-tracer differences still confuse me) in a couple months with at least a backup version that runs on the x86 processor. I'm thinking the soft shadows will be difficult and an old link I had on a Cook Torrance BRDF implementation by Pixar unfortunately no longer works. When I get around to trying this, I'll definitely read up more on the different reflection model options (probably Cook-Torrance's).

When running your .exe, the different options were pretty quick, but the 'a' and 'd' options, like you said, took a while. I didn't have time to let them finish, but they seemed stuck. My CPU for each option would stay around 26% utilization and then come back down ('a' and 'd' remained at this %). Mem is about at 2.5 MB.

Congrats on getting this done, you should be proud![/img]

It worked fine in Windows XP, just somewhat slow here too (P4 3.33 GHz).
Very nice avcaballero.

Well done. I made similar demo (ray traced rotated spheres) for Menuetos. If you want, take a look.
BIN+SRC - http://macgub.co.pl/menuet/ray01.zip
SCREENSHOT - http://macgub.co.pl/menuet/ray01_big.jpg