CS-rscher/liberal-elitist/ally/musician. STEM/C++/raytrace/hard-scifi/caffè/thai/snark/math. Truth & niceness are the most important things. DMs open!
Oregon
Joined July 2015
- Tweets194
- Following58
- Followers516
- Likes738
I don't really do the social media thing very much anymore, and I don't anticipate posting much here in the foreseeable future.
You can keep in touch at: bsky.app/profile/geometrian.…
Or, as always, what I'm up to on my website: geometrian.com/
1
244
Agatha Mallett retweeted
I made holographic chocolate!
Way easier than I thought - and it’s 100% pure bittersweet, no coatings. When you catch the light at the right angle, it’s mesmerizing.
Spreading the raytracing religion!
👁️ ☀️
↘️ ↗️
🫖
(My square at the Forest Grove sidewalk chalk art festival last weekend. Yes; I am *absolutely* dorky enough to make over-the-top pathtracing propaganda.)
One thing that has always bothered me is that the Phong BRDFs don't maintain energy—so I wrote a paper (with @cem_yuksel) about how to fix them!
Please enjoy this comparison from our paper, and check out my lightning summary video at the project homepage:
graphics.geometrian.com/rese…
ALT Figure 5 in our paper.
Our HPG 2020 paper/talk, "Efficient Adaptive Deferred Shading with Hardware Scatter Tiles" is up:
graphics.geometrian.com/rese…
We give tips to vendors about implementing our HPG '18 paper "Deferred Adaptive Compute Shading" in hardware. Core to the idea is interlocked, on-chip tiles.
ALT Scatter tile (Fig. 3).
ALT Scatter tile shading order (Fig. 4).
ALT DACS vs. VRS image quality comparison (Fig. 6 partial). The globally optimal VRS shading rates were used, even!
ALT Screenshot from presentation, discussing scatter tiles.
Preprint + supplemental is up for our @EGSympRendering paper
"Spectral Primary Decomposition for Rendering with sRGB Reflectance":
geometrian.com/research/inde…
Spectral reflectance for any sRGB triple without any colorimetric error!
(E.g. for using RGB assets in spectral renderers.)
ALT Title, author, teaser for paper.
People tried this basis approach before, obviously. But they got the math wrong. Wherefore literally dozens of authors have claimed its impossibility over 20+ years—just about as long as sRGB has even existed.
Therefore we also thoroughly review How Do You Even Colorimetry, Bruh
1
4
One thing I absolutely do not want to get lost:
If you've got basic raytracing knowledge and want to learn spectral rendering, check out our sample renderer! It's intended for teaching purposes—simple, commented to death, and algorithmically sensible.
github.com/imallett/simple-s…
ALT Render of Cornell box with CIE 2006 std. observer and using sRGB and spectral data.
Nice paper on making accurate orthonormal bases:
teapaw.org/JCGT_announce/status/8…
jcgt.org/published/0006/01/0…
With SIMD, you can make it > 2x faster!
Twitter ate the impl given in prev tweet, but that's ok because I optimized it even harder and it runs even faster now (still about 2⨯, keeping pace with compiler improvements from past 2 years).
Intel 990X with JCGT version as baseline:
GCC: 5.1ns -> 2.9ns
MSVC: 8.3ns -> 3.5ns
ALT https://gcc.godbolt.org/z/b8V1Ts
1
I made a distribution pathtracer in WebGL.
• Runs on your phone
• Dynamic light / camera
• Raytraced reflections and refractions
• Indirect illumination
• Realtime convergence (no denoising)
Yes, I am cheating.
geometrian.com/programming/w…
Thanks everyone for a great HPG & SIGGRAPH 2018! I think I am finally caught up on everything, so if I've missed you feel free to ping again. In-particular, if you have questions about Deferred Adaptive Compute Shading, email me! We'd love to see production-ready implementations!
I (re)invented a faster SSE cross-product implementation.
What's weird is that fused-multiply-add instructions don't help—and actually seem to *hurt*—performance? Maybe not true when `inline`d, or on newer than Haswell?
See my code and analysis here:
geometrian.com/programming/t…
ALT image of cross-product code
1
7
Check out my HPG 2018 Paper, "Deferred Adaptive Compute Shading"!
It's a drop-in replacement for checkerboard rendering that typically gives 3–5⨯ better quality/perf and is actually *easier* to implement! Gain is actually *larger* with harder shading!
geometrian.com/research/inde…
ALT Teaser figure for DACS.
Machine learning is a magical tool that can solve anything.
Input data: .obj files
Output data: .jpeg files
We can solve graphics. Let's go.
2
5
Hacking a bit on my solar system simulator recently. Accurately predicts positions/orientations of astronomical bodies for millennia into past/future.
Recently updated the graphics. Atmospheric shader runs *in constant time* (no raymarching). Improvements to do, though.
ALT Sun and Earth
ALT Earth
ALT Sun
