PLDI 2024
Mon 24 - Fri 28 June 2024 Copenhagen, Denmark
Mon 24 Jun 2024 13:55 - 14:10 at Iceland - Embedded Systems Chair(s): Jian-Jia Chen

Mixed-precision quantization can reduce the computational requirements of Deep Neural Network (DNN) models with minimal loss of accuracy. As executing mixed-precision DNN models on Neural Processing Units (NPUs) incurs significant under-utilization of computational resources, Precision-Scalable NPUs (PSNPUs) which can process multiple low-precision layers simultaneously have been proposed. However, the under-utilization still remains significant due to the lack of adequate scheduling algorithms to support multiple mixed-precision models on PSNPUs. Therefore, in this paper, we propose a dynamic programming-based scheduling algorithm for the operations of multiple mixed-precision models. Our scheduling algorithm finds the optimal execution plan that exploits the precision-scalable MACs to improve the end-to-end inference latency of mixed-precision models. We evaluate the performance of this algorithm in terms of hardware utilization, inference latency, and schedule search time compared to baseline scheduling algorithms. The experimental results show 1.23\texttimes{} inference latency improvements over the baseline algorithms within the allowed minutes.

Mon 24 Jun

Displayed time zone: Windhoek change

13:40 - 15:20
Embedded SystemsLCTES at Iceland
Chair(s): Jian-Jia Chen TU Dortmund University
13:40
15m
Talk
SmartVisor: User-Friendly Hypervisor for Mobile RobotsRemote
LCTES
Guanyu Chen Zhejiang University, Pan Lv Zhejiang University, Hong Li Zhejiang University, Guoqing Yang Zhejiang University
13:55
15m
Talk
Orchestrating Multiple Mixed Precision Models on a Shared Precision-Scalable NPU
LCTES
Kiung Jung Yonsei University, Seok Namkoong Yonsei University, Hongjun Um Hanyang University, Hyejun Kim Yonsei University, Youngsok Kim Yonsei University, Yongjun Park Yonsei University
14:10
15m
Talk
WoCA: Avoiding Intermittent Execution in Embedded Systems by Worst-Case Analyses with Device States
LCTES
Phillip Raffeck Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Johannes Maier Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Peter Wägemann Friedrich-Alexander University Erlangen-Nürnberg (FAU)
14:25
10m
Break
Break - 10 minutes
LCTES

14:35
15m
Talk
Unmasking the Lurking: Malicious Behavior Detection for IoT Malware with Multi-label Classification
LCTES
Ruitao Feng SMU, Sen Li Tianjin University, Sen Chen Tianjin University, Mengmeng Ge Nanyang Technological University, Xuewei Li Tianjin University, Xiaohong Li Tianjin University
14:50
15m
Talk
TWFuzz: Fuzzing Embedded Systems with Three WiresRemote
LCTES
Zhongwen Feng Chang' an University, Junyan Ma Chang'an University
15:05
15m
Talk
OpenMP-RT: Native Pragma Support for Real-Time Tasks and Synchronization with LLVM under Linux
LCTES
Brayden McDonald North Carolina State University, Frank Mueller North Carolina State University, USA