lisyarus

lisyarus, 6 days ago to random

I'm slowly returning to the development of my village building game, and what's a better way to do that other than making a new devlog! :3

https://www.youtube.com/watch?v=iE3p30PfiVI

lisyarus, 11 days ago to random

So I was casually re-deriving some stuff from the VNDF sampling paper, and stumbled upon the humble "Jacobian of the reflection operator", tried computing it, and how I'm in a deep rabbit hole. Anyone got any ideas? :)

https://math.stackexchange.com/questions/4912592/jacobian-of-the-vector-reflection-operator

lisyarus, 11 days ago to random

It seems that most of my problems with transparent BSDF were due to poor sampling strategy (I used uniform spherical sampling, as a baseline). After nudging VNDF to work with transmission, and using a more sane material description, I finally have a thin-glass orb :)

Though it gets darker with high roughness, so it needs some more work.

lisyarus, 12 days ago to random

Still trying to implement proper transparent bsdf in my raytracer (instead of hard-coding the refraction ray). Multiple sources and implementations use different formulas and neither of them works 😆

Also I have no idea what's happening on the screenshot

lisyarus, 14 days ago to random

Another weird raytracing bug! The black patches on the border are expected - these are artifacts due to having an extremely low-poly metallic surface with interpolated normals.

The white circles, however, are probably some funny floating-point business

lisyarus, 15 days ago to random

Implemented a proper brdf (Cook-Torrance + GGX) and vndf sampling, now I can do some nicer materials! Here's a shiny ball and a golden dragon:

image/png

lisyarus, 15 days ago to random

Oh no, there's some weird bug in the raytracer (this should be a fully smooth diffuse sphere)

lisyarus, 15 days ago to random

I managed to fix the GPU hanging issues but the "optimized" BVH traversal in compute shader still works slower than the unoptimized one 😭

In other news, I tried the Möller–Trumbore ray-triangle intersection algorithm, and it too turned out to be a bit slower than my extremely direct algorithm that just solves a linear system:

image/png

lisyarus, 16 days ago to random

Third day debugging GPU workgroup atomics to squeeze more perfomance from my raytracer, but it still hangs :(

Here's a cool debug visualization though :)

lisyarus, 19 days ago to random

Made quite a few optimizations to my WebGPU pathtracer:

Ordered BVH traversal: 15 -> 19 FPS
Store vertex indices instead of triangle ID's in BVH: 19 -> 20 FPS
Don't use indexed geometry: 20 -> 24 FPS
Separate vertex positions from other attributes: 24 -> 27 FPS

Not bad

lisyarus, 20 days ago to random

So I was accumulating Monte Carlo samples using good old blending here, now I've switched to a compute shader explicitly blending into a storage texture and it is ~10x slower, interesting

lisyarus, 20 days ago to random

Managed to hack together a BVH and render this 10k-vertex Stanford bunny. At 1k resolution it takes several seconds per sample, so I guess I'll focus on optimizing that.

lisyarus, 20 days ago to random

GPU Monte Carlo path tracing goes brrrrrrr

This looks much smoother in reality, I had to massage vsync to prevent tearing when recording this

video/mp4

lisyarus, 21 days ago to random

Soooooo guess what project I've decided to do next

lisyarus, 22 days ago to random

GPU folks, say I'm writing a bare-bones non-realtime GPU raytracer (in compute, w/o rtx kernels). Does it make sense to store geometry as non-indexed, to skip fetching triangle data and fetch vertices directly?

lisyarus, 23 days ago to random

New blog post: Demystifying multiple importance sampling

It is a simple thing that turns out to be confusing to a lot of us rendering engineers. I try to explain it in detail here with path-tracing examples!

https://lisyarus.github.io/blog/posts/multiple-importance-sampling.html

lisyarus, 23 days ago to random

Quite a lot of people asked me to add an RSS feed to my blog, so here it is: https://lisyarus.github.io/blog/feed.xml

lisyarus, 24 days ago to random

I've completely remade my blog! Changed the design, improved the look of formulas, fixed broken links and typos, etc.

I was thinking of making a thread with my favourite articles, but that would be about a dozen of them, so I'll just drop the blog link here :)

https://lisyarus.github.io/blog/index.html

lisyarus, 29 days ago to random

I'm in the process of completely redesigning my blog. I used Jekyll and tbh I never liked it - I often needed to use raw html for some stuff anyway. Now transitioning to a pure html+css+js site, like some caveman :)

image/png
image/png

lisyarus, 1 month ago to random

So apparently my student and I found two separate errors in the glTF reference BRDF specification, yikes.

lisyarus, 1 month ago to random

I was reading about resource/asset packing file formats game engines use, and it seems that the answer always is "ad-hoc engine-specific binary format".

I'm starting to think that I'll just slap a zip file in here and be happy.

lisyarus, 1 month ago to random

I was really confused that the original Veach 97 paper that introduced multiple importance sampling describes something not directly applicable to pathtracing, with some unclear-how-to-use heuristics

I ended up re-deriving MIS myself and verifying it in several other sources

Now it turns out the pathtracing-relevant part is in the very end of the section, and only needs the simplest heuristic, and is equivalent to what I've derived earlier

That's a relief :)

lisyarus, 1 month ago to random

Another raytracing test: the Sibenik cathedral!

lisyarus, 1 month ago to random

Another cute raytracing test - a room from @morgan3d's archive (https://casual-effects.com/data):

lisyarus, 1 month ago to random

My next (and final) lecture in the raytracing course will be about adding textures, so my plan for today is to slap some colors on this bad boi