Wow, thanks for making me feel old, paper

Ok, ok, hear me out. A PDF reader that doesn't lag. I know, crazy.

@parskatt.bsky.social First Dad, now Mum, what are they going to come up with next?

As someone who routinely switches between Linux, MacOS and Windows, the fact that Win11 removed a feature to put taskbar at the top hurts a lot.

Finished documenting work on optimized wireframe rendering: precision, clipping & math details for stable, performant edges in OpenGL. Includes an analysis of NVIDIA’s wireframe paper & fixes its issues. fauder.github.io/posts/2025/1... #GraphicsProgramming #Rendering #OpenGL #ShaderDev #GameDev

If we do spatial queries in R^3, should we call the data structure kd-tree, 3d-tree, or just be honest about it and call it sorted pointcloud?

Assume that you have a "cone" represented by normal n and cos(θ), where θ is the half angle, how do you compute the intersection of the two without involving acos? I'm actually mostly interested in the binary classification (interesects / doesn't intersect)

"Inside NVIDIA GPUs: Anatomy of high performance matmul kernels", includes a great intro to GPU architecture and PTX/SASS: www.aleksagordic.com/blog/matmul

Prove me wrong: For all the 3R papers, there is no reason for not running a fusion over your results. The figures would be way more readable, and misalignment more obvious. The fuzziness of point-clouds hides it. We have the technology! 2025 vs. 2016

New blog post! In "Billions of triangles in minutes" we'll walk through hierarchical cluster level of detail generation of, well, billions of triangles in minutes. Reposts welcome! zeux.io/2025/09/30/b...

Really looking forward to people starting to up the energy of their 3D reconstruction paper titles. These guys had it figured out, i'd skip the question mark tho.

A banger paper from @dcremers.bsky.social group! Great to see that focusing on strong two-view prediction, coupled with classical back-end optimization can lead to SOTA results! ganlinzhang.xyz/vista-slam/%... Also, you don't need 80GB GPU for inference ;)

#SIGGRAPH2025 Advances in Real-Time Rendering in Games course slides for "Fast as Hell: idTech8 Global Illumination” talk by @idsoftwaretiago.bsky.social from @idsoftware.bsky.social are now online advances.realtimerendering.com/s2025/index....

meshoptimizer v0.25 is out! Featuring new simplification function that optimizes positions and attributes for appearance, experimental permissive mode to simplify faceted regions with selective seam preservation, regularization option for improved tessellation quality and deformation, multiple ...

Friends don't let friends form full Jacobians

Not sure if useful for anybody, but I wanted a resource that lists different robust functions, their derivatives and IRLS weights in one place. So, here is a desmos workspace that does just that: www.desmos.com/calculator/q...

Kinda struggling with all the Dust3R / VGGT followups. How much is the improvement due to any proposed architectural changes, and how much is it just training for a longer time?

Listening to @gameengineering.bsky.social recently, and I’m wondering how much the brilliance of places like looking glass, lionhead, mm, and rad was a function of culture, or whether they were just smarter than everybody else…

Derivatives, Gradients, Jacobians and Hessians – Oh My! New mathy blog post to try and make some scary sounding terms less scary. blog.demofox.org/2025/08/16/d...

Nice introduction (or refresher) to PBR: "Physically Based Rendering: Putting Pieces Together", Game Industry Conference 2024 presentation recording www.youtube.com/watch?v=OwyM...

Falling down the rabbit hole of text rendering. Please send help.

Worth watching presentation, also provides some background info on how ReSTIR came to be and where it is heading: "ReSTIR: Traveling the Path of Reuse" www.youtube.com/watch?v=MRg9...

Yes! Here are all the SGP grad school recordings for every year: school.geometryprocessing.org

MoGe-v2 is bananas github.com/microsoft/MoGe

Caved in and tried zed with all the llm bells and whistles. Sadly, what gets me the most is that llm based edit predictions are faster than lsp suggestions :S pyright is SO ... DANG ... SLOW Where did we go wrong?

I wonder if a side-effect of LLMs will be lesser adoptions of new programming languages, esp. languages developed after LLM-boom. While the outputs of an LLM on C / JS are impressive, trying it on something like Odin is a bit of a dumpster fire.

was looking for some slides, then stumbled upon this interesting repo/file: GPU Optimization for GameDev (md format) gist.github.com/silvesthu/50...

Another treasure trove is this list: tfpsly.free.fr/bookmarks.html although by now maybe a 4th of these links lead to defunct sites.

While we are spinning up the coding agent's doing "intern's" work, what is the game plan to train people up if there is no entry level position? Is the long term game to improve the models to completely remove necessity of coding to build software product?

Really enjoyed MP-SfM! (arxiv.org/pdf/2504.20040) Very cool problem formulation, awesome results! Also, a great excuse for a joke for like 3 people.

New blog post! "Load store conflicts", in which we look at some performance sensitive code that has surprisingly dramatic performance swings based on the compiler and the microarchitecture used. Reposts appreciated! zeux.io/2025/05/03/l...

Mirrors were a mistake. Lambertian everywhere, please.

I keep swinging between liking python and thinking that it's not a serious programming language. I also wouldn't underestimate my ability to write shitty code.

I know this makes me a dinosaur, and "transformers go brrr" probably wins given enough data, but I don't enjoy this recent where "optimization" is a dirty word.

Very glad to see we are using all this compute for important tasks

Excuse me, what is this black magic? May I interest you in some torch.optim.Adam, like God intended?

Our method can reconstruct 3D geometry from hundreds of images in a single forward pass!!! (*, **) * actually, around 72 if you have $4k GPU ** it's gonna be more like $18k GPU if you have more images

Interesting, but wrong

Trying out Mast3r-SLAM - running it with `--no-viz` - there does not seem to be a way to visualize the resulting .pt file? @rmurai0610.bsky.social @ericdexheimer.bsky.social ?

Not sure what I was expecting with 50xx stock…

On one hand, moving to end-to-end fully learned framework will likely win at the end. On the other hand if something like COLMAP fails, it likely is a correspondence problem, so at least you get some insight into what went wrong. Harder to reason about what went wrong in latent space.

Interesting to see this trend of papers throwing shade on Dust3r's optimization stage. Buddy, your transformer is optimized with the same algo.

Say what you what about nu-metal, but Roots is such a tremendous album: www.youtube.com/watch?v=vVkD...

I do wonder if there comes a time when a person actually knows what they are doing instead of flipping order of matrix multiplication and where the inverses are until it works. I hope so, but I am not there yet.

@vincentleroy.bsky.social MASt3R retrieval module's similarity matrices seem to have odd behaviour. Self-similarity (diagonal elements) are not 1.0 and they differ between different images (slightly, mean of 0.256 and std-dev of 0.014). Any intuition why that is?

The results here are great, but it is always a little sus to me that they are showing these detailed reconstructions with tons of texture. Show me how it does on an empty rooms with flat white walls, and then we talk, ok?

Introducing MASt3R-SLAM, the first real-time monocular dense SLAM with MASt3R as a foundation. Easy to use like DUSt3R/MASt3R, from an uncalibrated RGB video it recovers accurate, globally consistent poses & a dense map. With @ericdexheimer.bsky.social* @ajdavison.bsky.social (*Equal Contribution)

New week, new day, and 5 new papers for my ever growing "papers-to-read" stack.

How does one write a good related work section these days? Looking at papers coming out is seems like authors often use long list of related papers doing approximately same thing. This seems mostly like a strategy to appease reviewer, rather than providing useful related work section

How are Kernel Smoothing in statistics, Data-Adaptive Filters in image processing, and Attention in Machine Learning related? My goal is not to argue who should get credit for what, but to show a progression of closely related ideas over time and across neighboring fields. 1/n