Cicero: Addressing Algorithmic and Architectural Bottlenecks in Neural Rendering by Radiance Warping and Memory Optimizations

TL;DR

CICERO significantly accelerates neural rendering on mobile GPUs by optimizing algorithms and memory access, achieving up to 28x speed-up and substantial energy savings with minimal quality loss.

Contribution

The paper introduces CICERO, a novel approach combining algorithmic reductions and memory optimizations to address performance bottlenecks in neural rendering on resource-limited devices.

Findings

8.0x speed-up and 7.9x energy saving on a mobile GPU

28.2x speed-up and 37.8x energy reduction with a DNN accelerator

Less than 1.0 dB PSNR quality loss

Abstract

Neural Radiance Field (NeRF) is widely seen as an alternative to traditional physically-based rendering. However, NeRF has not yet seen its adoption in resource-limited mobile systems such as Virtual and Augmented Reality (VR/AR), because it is simply extremely slow. On a mobile Volta GPU, even the state-of-the-art NeRF models generally execute only at 0.8 FPS. We show that the main performance bottlenecks are both algorithmic and architectural. We introduce, CICERO, to tame both forms of inefficiencies. We first introduce two algorithms, one fundamentally reduces the amount of work any NeRF model has to execute, and the other eliminates irregular DRAM accesses. We then describe an on-chip data layout strategy that eliminates SRAM bank conflicts. A pure software implementation of CICERO offers an 8.0x speed-up and 7.9x energy saving over a mobile Volta GPU. When compared to a baseline with a dedicated DNN accelerator, our speed-up and energy reduction increase to 28.2x and 37.8x, respectively - all with minimal quality loss (less than 1.0 dB peak signal-to-noise ratio reduction).