This walkthrough demystifies Mesa NIR’s memory optimization pipeline, including SSA promotion, deref analysis, copy propagation, and store elimination, and shows how it reduces memory traffic to boost GPU performance.
https://www.collabora.com/news-and-blog/blog/2026/05/07/optimizing-memory-access-in-nir/
#Mesa #NIR #ShaderCompiler #GPUDrivers #GraphicsDev #ShaderOptimization #Vulkan #OpenSource
• The start of graphics programming: Omg I wanna make super cool shaders.
• The first turn: Here is my realtime pathtracer with denoiser and PBR shader support
• The second turn: Okay but what if I got this to render on a potatoe? Just gotta draw some texture mapped triangles using the CPU right?
• The conclusion: Epic! ... now... why did I do any of this again ? .... Oooohhhh yeah. I wanted some simple PSX shaders to add to my submission to the PSX Horror gamejam ....
I FORGOR 💀
A question for gamedevs on the fedi out there; when you make a 3d water shader with transparency, how do you output the underwater buffer after fading it out, etc.? I may use the poll result in a blog post about it.
Please excuse the potential hashtag spam. I'd like to get this to as many relevant people as possible.
#gamedev #indiedev #fedidev #graphicsdev #shader #vfx #godotengine #unityengine #unity3d #unrealengine.
Procedural materials, represented as functional node graphs, are ubiquitous in computer graphics for photorealistic material appearance design. They allow users to perform intuitive and precise editing to achieve desired visual appearances. However, creating a procedural material given an input image requires professional knowledge and significant effort. In this work, we leverage the ability to convert procedural materials into standard Python programs and fine-tune a large pre-trained vision-language model (VLM) to generate such programs from input images. To enable effective fine-tuning, we also contribute an open-source procedural material dataset and propose to perform program-level augmentation by prompting another pre-trained large language model (LLM). Through extensive evaluation, we show that our method outperforms previous methods on both synthetic and real-world examples.
Yey another fantastic video by Sebastian Lague:
https://youtu.be/C1H4zIiCOaI
I think this is a great example of how to approach optimization problems. He takes many different approaches throughout the video and will often pause to ask "what are the results from this profiling test actually telling me?" and "is there a better route I should take right now?"
Collabora
LUZZY the bun
Wattana's Place of Brainrot
Josh Jersild
Alrecenk
Paul Keen
Bryce Dixon
leadedsolder