Hardware descriptions code completion based on a pre-training model
Xinyu Liang · 2021 IEEE Conference on Telecommunications, Optics and Computer Science (TOCS) · 2021
The code completion function, as a coder-friendly tool, appears in almost all IDEs. With the development of artificial intelligence technology in recent years, intelligent code completion techniques have gradually demonstrated better results than traditional code completion. Most code completion tools focus on imperative programming languages, while the development of intelligent tools in hardware description languages is far behind. In this paper, we propose a model that can be directly applied to code completion on hardware descriptions based on the GPT-2 model. The method can achieve the same level of accuracy without designing complex neural network structures. In addition, the method also has fast training speed, strong generalization ability, and the ability to capture the localness of the code.