I know "physically based" means practically just "derived from Disney BRDF", but it still bothers me as a term. As a model it has still plenty of shortcomings in terms on how well it matches actual physics, and while true that it is more physically based than some previous shaders it feels weird place to draw the line.

Looks fantastic, shouldn't be 'too difficult' to port to web via WebAssembly. There goes my weekend! Keep up the great work :)

It's great to see projects like this, it makes programming fun. Bite sized pieces that can be easily held in the head and reasoned about.

The title here says data-driven but the one in the github says data-oriented, so which is it?

This appears to be unrelated to https://www.yoctoproject.org ?

I wonder if yocto is a trademark

They sell "yocto compatible" labels, so there is something to consider

Yocto- is a metric system prefix. I doubt that anybody can claim copyright.

"Yocto Project" is a registered trademark.

https://tmsearch.uspto.gov/

Yeah just like you cant claim "apple"

Doesn't mention CUDA. Only Intel Embree or whatever. Seems like an odd choice.

Wish it had support for Nvidia because the API looks very powerful.

I mean the whole point is being small c++ libraries for rendering. CUDA sortof defeats the whole stated goal.

Also Embree isn't really comparable to CUDA, Embree is libraries for BVH generation and traversal, CUDA is a language and tools for GPU programming...

Forget I said CUDA then. The point is that it would be nice to run it on highly parallel hardware.

Oh I assumed that Embree was some new Intel graphics thing.

I consider that a pro, no reliance on big-tech proprietary stuff. +1 for the author not going that way