Towards Fine-Grained Online Adaptive Approximation Control for Dense SLAM on Embedded GPUs
Towards Fine-Grained Online Adaptive Approximation Control for Dense SLAM on Embedded GPUs
复制标题
面向嵌入式 GPU 上的密集 SLAM 的细粒度在线自适应逼近控制
DOI：
10.1145/3486612
复制
发表时间：
2022
影响因子：
1.4
通讯作者：
Jingweijia Tan
中科院分区：
文献类型：
--
作者：
Tiancong Bu;Kaige Yan;Jingweijia Tan
密集SLAM是嵌入式环境下的一个重要应用。然而，嵌入式平台通常无法为高精度实时密集SLAM提供足够的计算资源，即使采用gpu等高并行架构。为了解决这个问题，一个解决方案是在嵌入式gpu上设计合适的密集SLAM近似技术。在这项工作中，我们提出了两种新的近似技术，关键数据识别和冗余分支消除。本文还分析了另外两种方法的误差特性——跳环和线程逼近。然后，我们提出了一种在线自适应逼近控制器SLaPP，旨在将误差控制在可接受的阈值以下。评估结果表明，与不进行近似的情况相比，SLaPP平均可以实现2.0倍的性能加速和30%的节能。
Dense SLAM is an important application on an embedded environment. However, embedded platforms usually fail to provide enough computation resources for high-accuracy real-time dense SLAM, even with high-parallelism architecture such as GPUs. To tackle this problem, one solution is to design proper approximation techniques for dense SLAM on embedded GPUs. In this work, we propose two novel approximation techniques, critical data identification and redundant branch elimination. We also analyze the error characteristics of the other two techniques—loop skipping and thread approximation. Then, we propose SLaPP, an online adaptive approximation controller, which aims to control the error to be under an acceptable threshold. The evaluation shows SLaPP can achieve 2.0× performance speedup and 30% energy saving on average compared to the case without approximation.