Fuchun Wei
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Fuchun Wei has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Chen Tian
Nanjing University, Nanjing, China
2 shared publications - Guihai Chen
Shanghai Jiao Tong University
2 shared publications - Jingyi Zhang
Huawei Technologies Co., Ltd., Shenzhen, China
2 shared publications - Wanchun Dou
Nanjing University
2 shared publications - Yanlin Liu
Huawei Technologies Co., Ltd., Shenzhen, China
2 shared publications - Zhiheng Hu
Huawei Technologies Co., Ltd, Hangzhou, Zhejiang, China
2 shared publications
Latest Publications
- Open Access
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 1156-1171
Squeezing Operator Performance Potential for the Ascend Architecture
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 1156-1171
Open AccessSqueezing Operator Performance Potential for the Ascend Architecture
- Open Access
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 1118-1132
Using Analytical Performance/Power Model and Fine-Grained DVFS to Enhance AI Accelerator Energy Efficiency
Recent advancements in deep learning have significantly increased AI processors' energy consumption, which is becoming a critical factor limiting AI development. Dynamic Voltage and Frequency Scaling (DVFS) stands as a key method in power optimization. However, due to the latency of DVFS control in AI processors, previous works typically apply DVFS control at the granularity of a program's entire duration or sub-phases, rather than at the level of AI operators. The advent of millisecond-level DVFS capabilities on the latest Ascend NPU platforms enables us to set frequency individually for single or multiple operators, opening up the opportunity for further enhancing energy efficiency through fine-grained DVFS control. To ensure performance is unaffected in DVFS, our work builds performance and power models for each operator. Through in-depth timeline analysis, we demonstrate that the cycle count of an operator can be modeled as a convex piecewise linear function of frequency, resulting in a performance model with an average error of 1.96%. Moreover, we build power models that incorporate temperature-dependent terms, which enhances the model's precision and results in an average error of 4.62%. Based on our performance and power models as well as the fine-grained DVFS functionality of Ascend NPU, we propose a DVFS strategy that integrates operator classification, preprocessing, and a genetic algorithm-based search. Experiments on applications including GPT-3 training achieve a reduction in AICore (the computing component within the Ascend NPU) power by 13.44% and NPU chip power by 4.95%, while limiting performance degradation to 1.76%.
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 1118-1132
Open AccessUsing Analytical Performance/Power Model and Fine-Grained DVFS to Enhance AI Accelerator Energy Efficiency
Recent advancements in deep learning have significantly increased AI processors' energy consumption, which is becoming a critical factor limiting AI development. Dynamic Voltage and Frequency Scaling (DVFS) stands as a key method in power optimization. However, due to the latency of DVFS control in AI processors, previous works typically apply DVFS control at the granularity of a program's entire duration or sub-phases, rather than at the level of AI operators. The advent of millisecond-level DVFS capabilities on the latest Ascend NPU platforms enables us to set frequency individually for single or multiple operators, opening up the opportunity for further enhancing energy efficiency through fine-grained DVFS control. To ensure performance is unaffected in DVFS, our work builds performance and power models for each operator. Through in-depth timeline analysis, we demonstrate that the cycle count of an operator can be modeled as a convex piecewise linear function of frequency, resulting in a performance model with an average error of 1.96%. Moreover, we build power models that incorporate temperature-dependent terms, which enhances the model's precision and results in an average error of 4.62%. Based on our performance and power models as well as the fine-grained DVFS functionality of Ascend NPU, we propose a DVFS strategy that integrates operator classification, preprocessing, and a genetic algorithm-based search. Experiments on applications including GPT-3 training achieve a reduction in AICore (the computing component within the Ascend NPU) power by 13.44% and NPU chip power by 4.95%, while limiting performance degradation to 1.76%.