Pjbomb2

#GPU #graphics folks, say I want a GPU-side concurrent hash table, and the API only supports 32-bit atomics. It's easy if keys are also 32-bit, but what if I want some complex keys (say, a struct)? Can I do this and how?

Packing is used in a couples places in truetrace
And each emissive triangle is it’s own light source, as the triangle itself is a light

At its core, the bit trails are just a string a 1’s and 0’s that tell me which direction to traverse the tree at any given branch to reach a specific triangle(those triangles are at the “leaves” of the tree)

These bittrails that I talk about here in the origional post are used to traverse the light acceleration structure while gaurenteeing the triangle we will end at in the tree
So I can compute the correct weight quickly for any emissive triangle randomly hit by bouncing rays

The structure makes sampling emissive triangles far better/higher quality for Next Event Estimation
However, rays can also randomly hit an emissive triangle by chance, in which case we need to compute a weight so that NEE still is correct when combined with naive
Another reply incomming

So this is a bit of a specialized case for bitfield
They have a LOT of uses, like bitflag packing(which I use for my materials, I can pack 32 booleans into 1 uint of space)
For this specifically, it’s used for the light acceleration structure I use
Need more space, so this will be several replies…

Oh also have this

Have this

So actually there were 2 major bugs that were 4+ months old that caused massive amounts of fireflies
First one was I forgot a “- 1” in my tlas bitfield traversal
Second was I was double counting mesh count contribution to the pdf, this ones over a year old

So it’s a bit old now but i found a major bug with my light acceleration structure TLAS, and testing adding bitfield for BLAS(left is old, right is new)
The bitfield is now fully implemented(with a fallback for depth > 32) and various other improvements have been made

DX12 Bindless Plugin for Unity3D is now open source, and you can use it in your projects! Can be useful for writing a custom GPU Driven Renderer. (github below)

Some more shots by Yanus now that I’ve fixed some caustic stuff(was missing the pdf calculation for my guiding CDF…)
100% pathtraced and rendered in unity using purely TrueTrace

Some progress on implementing Projected Displacement Mapping, still have the issue with triangle boundaries, which is partly due to triangle winding order…

Has anyone ever implemented "Projective Displacement Mapping for Ray Traced Editable Surfaces"?
If so, can I get some help? I cant solve these offsets between triangle boundaries for the life of me and I am totally stuck again(this is exactly what stopped me last time),any tips?

Just stumbled upon this ancient PcPersective article on real-time ray tracing.. In 2008. :)
pcper.com/2008/06/ray-...
It details the story of nine students (Karim, Wussie, Rick, Jan, Rutger, Trevor, Mathijs, Roel, Wilco) who worked in various roles on CPU-ray traced games using the Arauna engine.

Some images showing off my improvements to my photon mapper I have integrated into truetrace
The photon mapper is up on the Dev branch on the github

Forgot to post these
From some improvements to my caustic photon mapper tacked into truetrace last week

Computer graphics image of a classical building with intricate geometric details in ornaments. Rendering and scene statistics

Anpther major update to our cluster lod ray tracing sample.
Zorah scene available as glTF
raw.githubusercontent.com/nvpro-sample...
BLAS caching github.com/nvpro-sample... BLAS merging github.com/nvpro-sample...

Special thanks to Adam Marrs, Pyarelal Knowles for ideas and Chris Perrella for zorah.

TinyBVH 1.6.7 is now available on the main branch. Changes:
* TinyBVH now only needs C++11 (tiny_ocl.h needs C++17)
* Full-sweep SAH builder is now threaded
* Fix for crash in BuildAVX
Still zero dependencies. :)
And: The repo reached 1k stars!
Link: github.com/jbikker/tiny...

I am glad to attend #HPG2025 where I just presented our paper "Fused Collapsing for Wide BVH Construction", co-authored with Mathias Paulin. We propose a fast build algorithm for wide BVHs that directly computes a wide hierarchy without additional collapsing pass. Webpage: wbrbr.org/publications...

Rendering nerds! Check out our latest work "Vector-Valued Monte Carlo Integration Using Ratio Control Variates" that has just gotten the best paper award at SIGGRAPH 2025. This paper presents a method that reduces variance of a wide range of rendering and diff. rendering tasks with negligible cost.

A Fluorescent Material Model for Non-Spectral Editing & Rendering

Belcour Laurent, Fichet Alban, Barla Pascal
(Intel Corp. France, Inria Bordeaux)

arxiv.org/abs/2505.19672

Yoooo what?? I thought opacity micro maps were kinda locked down black box to an extent, or needed dedicated hardware to be faster!

O wow, ReSTIR keeps giving. Awesome stuff.

What’s the speed vs a pure bvh like?

Saw that "The Witcher 4" UE5 tech demo with the voxel trees for fast rendering at a distance?
I figured... that should be easy to do in TinyBVH. :) So now TinyBVH can do voxel meshes. Attached video: CPU-only, switching between BVH and voxels.
Code in tiny_bvh_foliage.cpp: github.com/jbikker/tiny...

Now available: A blog post on (S)BVH optimization.
The article describes how to improve beyond SBVH quality, often significantly. Results are compared to full-sweep, binned SAH and the H-PLOC.
In short: Off-line BVH construction can double ray tracing performance.
jacco.ompf2.com/2025/05/20/b...

Added tri presplitting thanks to BoyBayKiller
Before vs after tri presplitting(splitting triangles before doing ANY building, vs SBVH which splits DURING build), only affects SWRT
Tri count jumped from 2832120 triangles to 3348607 triangles