arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2204.10707cs.AR
收藏

Crescent:驯服内存干扰,加速深度点云分析

Crescent: Taming Memory Irregularities for Accelerating Deep Point Cloud Analytics

Yu Feng, Gunnar Hammonds, Yiming Gan, Yuhao Zhu

摘要:

点云中的3D感知正在改变未来智能机器的感知能力。然而,点云算法受到不规则的内存访问的困扰,导致内存子系统中的大量低效,这瓶颈了整体效率。本文提出了Crescent,这是一种算法-硬件协同设计系统,可以在实现高精度的同时驯服深度点云分析中的不规则性。为此,我们介绍了两个近似技术,近似邻居搜索和选择性银行冲突省略,“正规化”的DRAM和SRAM存储器访问。然而,这样做必然会引入准确性损失,我们通过将近似集成到网络训练过程中的新网络训练过程来减轻这种损失。本质上,我们的训练过程训练以特定近似设置为条件的模型,从而保持高准确性。实验表明,与优化的基线加速器相比,Crescent的性能提高了一倍,能耗降低了一半,精度损失< 1%。我们论文的代码可以在https://github.com/horizon-research/crescent上找到。

英文摘要:

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.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑