Crescent: Taming Memory Irregularities for Accelerating Deep Point Cloud Analytics

TL;DR

Crescent is a co-designed system that improves deep point cloud analytics by regularizing memory accesses through approximation techniques, resulting in doubled performance and halved energy consumption with minimal accuracy loss.

Contribution

The paper introduces novel approximation techniques and a training method that together tame memory irregularities in deep point cloud processing, enhancing efficiency without significant accuracy loss.

Findings

Doubles performance compared to baseline

Halves energy consumption

Maintains < 1% accuracy loss

Abstract

3D perception in point clouds is transforming the perception ability of future intelligent machines. Point cloud algorithms, however, are plagued by irregular memory accesses, leading to massive inefficiencies in the memory sub-system, which bottlenecks the overall efficiency. This paper proposes Crescent, an algorithm-hardware co-design system that tames the irregularities in deep point cloud analytics while achieving high accuracy. To that end, we introduce two approximation techniques, approximate neighbor search and selectively bank conflict elision, that "regularize" the DRAM and SRAM memory accesses. Doing so, however, necessarily introduces accuracy loss, which we mitigate by a new network training procedure that integrates approximation into the network training process. In essence, our training procedure trains models that are conditioned upon a specific approximate setting and, thus, retain a high accuracy. Experiments show that Crescent doubles the performance and halves the energy consumption compared to an optimized baseline accelerator with < 1% accuracy loss. The code of our paper is available at: https://github.com/horizon-research/crescent.