Take a break and look at some preview videos.
http://www.youtube.com/channel/UCS1wsF-S2dzbBlxFeq2h_0A
Mitsuba may get some really cool stuff :eyebrowlift:
I wonder how the method in this sampling paper differs from the sampling algorithm Cycles has now. Can it be used for true, image-wide adaptive sampling like seen with Lukas Stockner’s patch or is it another iteration of even sampling?
If it is superior to Sobol and Multi-Jitter (in terms of the rate of convergence time-wise), it might be worth something for Sergey or DingTo to look into.
The method in that paper is untested in production and INCREDIBLY complicated as far as sampling algorithms go. It’s unlikely we’ll see it tackled in Blender any time soon, unless we manage to get a rendering pHd holder on the Blender team. There are much better and easier places to look at speeding up Cycles, like BVH building and traversal, before looking into exotic sampling methods like that.
Is the reel still in progress or is it hosted somewhere the public can view yet?