一句话：提出复用-更新排序算法与配套硬件加速器 Neo，利用帧间高斯排序的时序冗余，大幅降低 3DGS 渲染的排序开销与显存带宽压力。
3D 高斯泼溅（3DGS）是实现沉浸式 AR/VR 体验的重要技术，但其在 Meta Ray-Ban Display、Google Android XR Glasses 等资源受限设备上的落地取决于能否实时渲染，现有方案在高分辨率下帧率不足。作者通过分析指出，3DGS 渲染流水线中的排序阶段因对内存带宽需求极高而成为主要瓶颈。为此，论文提出 Neo：一种复用-更新排序算法，利用连续帧之间高斯排序结果的时序冗余，通过跟踪并增量更新高斯的深度次序来替代从头重新排序，从而减少冗余计算与带宽压力；同时设计了针对该算法优化的硬件加速器。实验表明，Neo 相比最先进的边缘 GPU 与 ASIC 方案吞吐分别最高提升 12.4 倍与 5.5 倍，DRAM 流量分别降低 94.6% 与 81.4%，使高质量、低延迟的端侧 3D 渲染更为可行。
While 3D Gaussian Splatting (3DGS) has emerged as a promising technique for immersive AR/VR experiences, its practical adoption critically depends on whether real-time rendering can be achieved on resource-constrained devices, such as Meta Ray-Ban Display and Google Android XR Glasses. However, existing solutions struggle to achieve high frame rates, especially for high-resolution rendering. Our analysis identifies the sorting stage in the 3DGS rendering pipeline as the major bottleneck due to its high memory bandwidth demand. This paper presents Neo, which introduces a reuse-and-update sorting algorithm that exploits temporal redundancy in Gaussian ordering across consecutive frames and devises a hardware accelerator optimized for this algorithm. By efficiently tracking and updating Gaussian depth ordering instead of re-sorting from scratch, Neo significantly reduces redundant computations and memory bandwidth pressure. Experimental results show that Neo achieves up to 12.4x and 5.5x higher throughput than state-of-the-art edge GPU and ASIC solution, respectively, while reducing DRAM traffic by 94.6% and 81.4%. These improvements make high-quality and low-latency on-device 3D rendering more practical.
DOI 原文 · arXiv · PDF（开放获取） · DBLP
@inproceedings{abs-2511-12930,
title = {Neo: Real-Time On-Device 3D Gaussian Splatting with Reuse-and-Update Sorting Acceleration},
author = {Changhun Oh and Seongryong Oh and Jinwoo Hwang and Yoonsung Kim and Hardik Sharma and Jongse Park},
booktitle = {ASPLOS},
year = {2026},
doi = {10.1145/3779212.3790192}
}