m-schuetz

m-schuetz / CuRast

Public

Cuda-Based Software Rasterization for Billions of Triangles

76
0
100% credibility
Found Apr 24, 2026 at 76 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
C++
AI Summary

CuRast is a CUDA-based software rasterization pipeline for efficient real-time rendering of massive triangle datasets with billions of triangles, using a multi-stage approach that handles small to large triangles without precomputed LODs or acceleration structures.

Star Growth

See how this repo grew from 76 to 76 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is CuRast?

CuRast is a C++ viewer using CUDA-based software rasterization to render billions of triangles from GLB/GLTF models in real-time. It tackles massive, dense meshes from photogrammetry—like Zorah's 13.5 billion triangles or Venice's 400 million—without precomputed LODs or acceleration structures, via a pipeline that efficiently handles tiny-to-large triangles. Drag-and-drop files into the app for instant viewing on RTX GPUs, with options for compressed loading to fit in VRAM.

Why is it gaining traction?

It crushes Vulkan by 2-5x on huge models (12x for instanced draws), proving compute shaders beat fixed-function pipelines for triangle soups. No preprocessing means quick tests on raw scans, and benchmarks like 67ms for 13B triangles hook anyone chasing Nanite-style extremes without game-engine overhead.

Who should use this?

Graphics researchers benchmarking rasterizers on photogrammetry data, 3D reconstruction devs visualizing unoptimized scans, or engine hackers prototyping software pipelines for billion-triangle scenes.

Verdict

Grab it if you're on CUDA/Windows with an RTX card and need to render absurd triangle counts—results impress. At 1.0% credibility (76 stars, Windows-only, basic docs), it's raw research code; fork and contribute for Linux/multi-mesh support.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.