Alexandre Lamure

Very proud to work on #StarWarsEclipse!

There's a neat short paper from Eurographics this year that evaluates how well microfacet shading models actually predict measured reflectance. The answer is: not nearly as well as one would hope! (1/3) ssteinberg.xyz/2026_surface_s…

Introducing AMD DGF SuperCompression gpuopen.com/learn/introduc…

youtu.be/Xolb8DcrbLc The recorded EUROGRAPHICS 2026 talk for our wave tracing paper. Very honored that this paper received the Günter Enderle Best Paper Honorable Mention award. Collaboration with Matt Pharr (NVIDIA).

@Dispatch_Graph I simply meant evaluating the occlusion of a ray that goes through some volumetric stuff like fog. Unless your volumetric is analytic fog or super cheap to raymarch, there's no easy way to account for it in real-time GI. But that could be used for the placement and animation too

@AlexandreLamure Wdym by that? Volumetric visibility as in it could be used to e.g. prevent volumetric smoke propagation through solide without a baking pass? Because that could be very interesting indeed

My idea, a 🧵 I named it NVF — Neural Visibility Fields Warning: potentially shitty idea. proceed with caution

@Dispatch_Graph Another nice benefit of NVFs would be the ability to query volumetric visibility

@Dispatch_Graph This paper kinda explored this gpuopen.com/download/HPG20… Using NNs makes it time-constant and super easy to scale for different configs, that's the biggest win imo

Subtle things to make your shaders better: 1) Anti-aliasing - Real life is not limited to pixels so unless you're specifically going for a pixelated style, you should be anti-aliasing everything! mini.gmshaders.com/p/antialiasing

Will present the new input encoding method "PEPS: Positional Encoding Projected Sampling" at i3D. In the paper, we show that 25% better compression. If you have NTC implementation with MLP, you can add this on top of it and see improvements. arxiv.org/abs/2604.24167

Graphics Programming weekly - Issue 438 - April 27h, 2026 jendrikillner.com/post/graphics-…

Here are some techniques I discovered through 14 years of shader programming:

While my vacation days coming to the end, finished a blog post on modern rendering culling techniques 😜 I tried to keep it practical: not a deep dive, but a broad overview of the main methods, tradeoffs, and how they fit together in a real renderer. Frustum culling, occlusion, GPU-driven rendering, meshlets, Nanite, light culling, etc. krupitskas.github.io/posts/modern_c… #gamedev #rendering #dx12

Ultra-fast Screen-Space Refractions and Caustics via Newton’s Method  Chase Mayer, Ulf Assarsson, Erik Sintorn Paper: jcgt.org/published/0015… Code (MIT): github.com/TheManTheMythT… Abstract We introduce a novel screen-space refraction method that replaces standard ray marching with Newton’s method for root finding. For each pixel, we define a function as the difference between the refracted ray depth and the G-buffer depth, and iteratively refine the intersection using the tangent plane of the current G-buffer point. This achieves rapid convergence, typically within only a few iterations, thereby reducing per-pixel cost. We further analyze common failure cases, such as depth discontinuities or poor initial guesses, and propose practical strategies to detect and then mitigate or resolve them. We also demonstrate our approach on refracted light rays to accelerate screen-space caustics.

Behind the pretty frames of #PRAGMATA. Breaking down many tricks and hidden gems of the renderer behind @CAPCOM_RandD's RE Engine. i hope you enjoy & find this few months worth of freetime article useful... i had FUN playing the demo and digging every bit of it, & looking forward for the game official release next week!!!! mamoniem.com/behind-the-pre… @PRAGMATAgame @PRAGMATA_JP @CAPCOM_RandD

Useful table summarising the various ways to pass resource descriptors to a shader via a root signature (from …tic.graphicsprogrammingconference.com/public/2025/sl…)

Real-time Neural Six-way Lightmaps Wei Li, Hanxiao Sun, Tao Huang, Haoxiang Wang, Tongtong Wang, Zherong Pan, Kui Wu (Tsinghua University, China; LIGHTSPEED, Australia) arxiv.org/abs/2604.03748 Abstract Participating media are a pervasive and intriguing visual effect in virtual environments. Unfortunately, rendering such phenomena in real-time is notoriously difficult due to the computational expense of estimating the volume rendering equation. While the six-way lightmaps technique has been widely used in video games to render smoke with a camera-oriented billboard and approximate lighting effects using six precomputed lightmaps, achieving a balance between realism and efficiency, it is limited to pre-simulated animation sequences and is ignorant of camera movement. In this work, we propose a neural six-way lightmaps method to strike a long-sought balance between dynamics and visual realism. Our approach first generates a guiding map from the camera view using ray marching with a large sampling distance to approximate smoke scattering and silhouette. Then, given a guiding map, we train a neural network to predict the corresponding six-way lightmaps. The resulting lightmaps can be seamlessly used in existing game engine pipelines. This approach supports visually appealing rendering effects while enabling real-time user interactivity, including smoke-obstacle interaction, camera movement, and light change. By conducting a series of comprehensive benchmarks, we demonstrate that our method is well-suited for real-time applications, such as games and VR/AR.

@Dispatch_Graph Love it! It's super well-explained 🙂 The diction is a bit fast, but it's fine and the subtitles are correct if necessary

I made a video where I talk about the different things I like about Metal! Link in description

Great collection of extensive articles on the architecture of a large number of (older) consoles, good read! copetti.org/writings/conso…

It seems indeed to work like this. I'm very surprised they tried this approach so early instead of using a G-buffer. We'll see how that works for a real real-time version

The blog post "Virtual Memory Tricks" by Niklas Gray explores clever, underutilized ways to leverage operating system virtual memory features (via APIs like VirtualAlloc or mmap) to solve common programming problems efficiently. Some of the key techniques discussed are .. • reserving extremely large virtual address spaces for arrays (e.g., billions of elements) without consuming physical RAM until pages are actually accessed, enabling simple, fast fixed-size data structures like global game object arrays or huge lookup tables. • using virtual memory reservations to generate guaranteed-unique application-wide IDs (by returning pointers from reserved blocks) • implementing an "end-of-page allocator" that places allocations at page boundaries to catch out-of-bounds or use-after-free bugs via immediate access violations, and avoiding fragmentation issues since virtual memory allows non-contiguous physical pages. • implementing a gapless ring buffer that uses shared mappings to handle wraparound seamlessly without extra boundary logic. Overall, the author argues that directly controlling virtual memory unlocks powerful, scalable, and safer memory management patterns beyond conventional malloc usage. link to the blog post: ruby0x1.github.io/machinery_blog…

My pleasure to announce Gamma Studios' website, alongside the first of a series of technical articles on color grading and tonemapping: gammastudios.tech/technical-colo… Give it a read if you want a detailed analysis on why some games raise blacks so much to make an OLED look like an LCD! I'd love to hear your opinions on my technical findings and conclusions!