一句话：提出算法-架构协同设计的 CIM 加速器 ASDR，通过动态采样和颜色/密度解耦近似，实现低功耗实时 NeRF 渲染。
神经辐射场（NeRF）能生成逼真图像与视频，但主流神经渲染模型推理延迟高、功耗大，且访存不规则、计算开销显著。存内计算（CIM）有望缓解访存瓶颈并降低功耗。本文提出算法-架构协同设计 ASDR，一种面向高效神经渲染的 CIM 加速器。算法层引入两种优化：在线感知不同像素渲染难度的动态采样，以减少访存与计算；解耦并近似颜色与密度的体渲染，以降低 MLP 开销。架构层设计基于 ReRAM 的 CIM 架构及高效数据映射与复用微架构。实验表明，在图形渲染任务中，ASDR 相比最先进 NeRF 加速器和 Xavier NX GPU 分别最高取得 9.55× 与 69.75× 加速，PSNR 仅损失 0.1。
Neural Radiance Fields (NeRF) offer significant promise for generating photorealistic images and videos. However, existing mainstream neural rendering models often fall short in meeting the demands for immediacy and power efficiency in practical applications. Specifically, these models frequently exhibit irregular access patterns and substantial computational overhead, leading to undesirable inference latency and high power consumption. Computing-in-memory (CIM), an emerging computational paradigm, has the potential to address these access bottlenecks and reduce the power consumption associated with model execution. To bridge the gap between model performance and real-world scene requirements, we propose an algorithm-architecture co-design approach, abbreviated as ASDR, a CIM-based accelerator supporting efficient neural rendering. At the algorithmic level, we propose two rendering optimization schemes: (1) Dynamic sampling by online sensing of the rendering difficulty of different pixels, thus reducing access memory and computational overhead. (2) Reducing MLP overhead by decoupling and approximating the volume rendering of color and density. At the architecture level, we design an efficient ReRAM-based CIM architecture with efficient data mapping and reuse microarchitecture. Experiments demonstrate that our design can achieve up to 9.55× and 69.75× speedup over state-of-the-art NeRF accelerators and Xavier NX GPU in graphics rendering tasks with only 0.1 PSNR loss.
@inproceedings{Liu0ZWSG025,
title = {ASDR: Exploiting Adaptive Sampling and Data Reuse for CIM-based Instant Neural Rendering},
author = {Fangxin Liu and Haomin Li and Bowen Zhu and Zongwu Wang and Zhuoran Song and Haibing Guan and Li Jiang},
booktitle = {ASPLOS},
year = {2025},
doi = {10.1145/3676642.3736117}
}