# Crescent: Taming Memory Irregularities for Accelerating Deep Point Cloud Analytics

arXiv: https://arxiv.org/abs/2204.10707 · HTML: https://mlsanity.com/paper/2204.10707/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 22 Apr 2022.

## Built on by (most cited first, 8 of 8)

- [TorchSparse++: Efficient Training and Inference Framework for Sparse Convolution on GPUs](https://mlsanity.com/paper/2311.12862/) (2023)
- [SPADE: Sparse Pillar-based 3D Object Detection Accelerator for Autonomous Driving](https://mlsanity.com/paper/2305.07522/) (2023)
- [Selectively Dilated Convolution for Accuracy-Preserving Sparse Pillar-based Embedded 3D Object Detection](https://mlsanity.com/paper/2408.13798/) (2024)
- [Thales: Formulating and Estimating Architectural Vulnerability Factors for DNN Accelerators](https://mlsanity.com/paper/2212.02649/) (2022)
- [FastPoint: Accelerating 3D Point Cloud Model Inference via Sample Point Distance Prediction](https://mlsanity.com/paper/2507.23480/) (2025)
- [FractalCloud: A Fractal-Inspired Architecture for Efficient Large-Scale Point Cloud Processing](https://mlsanity.com/paper/2511.07665/) (2025)
- [Spira: Exploiting Voxel Data Structural Properties for Efficient Sparse Convolution in Point Cloud Networks](https://mlsanity.com/paper/2511.20834/) (2025)
- [FlashFPS: Efficient Farthest Point Sampling for Large-Scale Point Clouds via Pruning and Caching](https://mlsanity.com/paper/2604.17720/) (2026)

## Abstract (arXiv)

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.