Local-GS: Accelerating 3D Gaussian Splatting via Tile-Local Warp Coherence

2026-06-15

AI Summary

This work addresses the inefficiency in 3D Gaussian splatting caused by irregular Gaussian distributions, which induce warp divergence and redundant computations on GPUs, thereby degrading rendering performance. To tackle this issue, the paper introduces warp coherence—a concept previously unexplored in this domain—and proposes a tile-level coherent rendering paradigm tailored for SIMT architectures. The method achieves plug-and-play optimizations across three stages: precomputing shared parameters, performing warp-level culling of invalid Gaussians, and enabling branch-free pixel blending. These innovations collectively enhance GPU utilization while preserving image fidelity. Experimental results demonstrate significant speedups of up to 7.76× across multiple datasets without compromising visual quality.

Abstract

3D Gaussian Splatting (3DGS) has significantly advanced real-time novel view synthesis by representing scenes as dense collections of anisotropic 3D Gaussian primitives. However, the irregular spatial distribution of Gaussians often leads to poor GPU utilization, as warp divergence and redundant computation degrade rendering performance. To address this, we present Local-GS, a warp-coherent rendering paradigm that, organizes Gaussian primitives with respect to SIMT (Single Instruction, Multiple Threads) execution boundaries rather than scene geometry. Specifically, we propose three warp-coherent stages: a hoisting stage that precomputes shared parameters at tile level, a culling stage that discards warps with no contribution, and a blending stage that replaces per-pixel branching with a uniform instruction stream. Across extensive benchmarks on multiple datasets, Local-GS improves efficiency without compromising quality. As a plug-and-play optimization, it provides additional performance gains to all tested baselines, culminating in a $7.76\times$ speedup on Deep Blending scenes.

Problem

• 3D Gaussian Splatting
• warp divergence
• GPU utilization
• real-time rendering
• SIMT execution

Innovation

• warp coherence
• 3D Gaussian Splatting
• SIMT optimization
• tile-based rendering
• real-time rendering