BenoitP

Jax's true calling: Ray-Marching renderers on WebGL

https://benoit.paris/posts/jax-ray-marcher/

JAX's true calling: Ray-Marching renderers on WebGL

JAX’s true calling: Ray-Marching renderers in Python on WebGL Demo (move your mouse/thumb across the image) Why, though? Well, I’ve been drooling over this tool the cool kids use, and wondering how I can join the gang. It’s called JAX. It’s got GPU accelerated functions over n-dimensional arrays. And built-in compile-time differentiability of these!? Auto-vectorization?? And you just have to do like with numpy. What’s not to like? Go home APL!

Apr 1 at 7:36pmWeb
Show threadVHRangerApr 1
Pytorch is such a maddening mess of half implemented research features in a state of Heisen-deprecation, Jax becomes more appealing to me by the day.
Show threadcorndogeApr 1
Moving my thumb across the image causes the ball and cube graphic to disappear to black and then scrolls the page. Firefox on iOS
Show threadakoboldfryingApr 1

Me too, Chrome on Android.

I like the concept of applying Jax to SDF sphere tracing :)

Show threaddvtApr 1

> the thing JAX was truly meant for: a graphics renderer

I mean, just like ray-tracing, SDF (ray-marching) is neat, but basically everything useful is expensive or hard to do (collisions, meshes, texturing etc.). I mean mathy stuff is easier (rotations, unions/intersections, function composition, etc.) but 3D is usually used in either modeling software or video games, which care more about the former than they do the latter.

Show threadArchit3chApr 1

Games and simulations are typically stateful, I'm not sure the functional purity of JAX is a good fit.

Also, what's the story for JAX + WebGL when it comes to targeting hardware-accelerated ray tracing?

Show threadvatsachakApr 1
Yeah GPU compilers will be used for way more things than AI because parallel = good