A Machine Learning Approach Towards SKILL Code Autocompletion
DescriptionMoore's Law continues to increase the complexity of electronic systems, electronic design automation (EDA) must advance to meet global demand. An important example of an EDA technology is SKILL, a scripting language used to customize and extend EDA software. Recently, code generation models using the transformer architecture have achieved impressive results in academic settings and have even been used in commercial developer tools. In this work, a novel methodology for generating SKILL code is proposed and experimentally validated. More specifically, we propose novel methodologies for (i) creating an unlabeled and labeled SKILL dataset, (ii) a training strategy where T5 models pre-trained on code are fine-tuned on this custom dataset in an unsupervised and supervised manner, and (iii) evaluating synthesized SKILL code. We show that models trained using the proposed methodology outperform baselines in terms of a human-judgment score and BLEU score. Furthermore, the limitations of the methodology are discussed and directions for future work to address these limitations are suggested.
Event Type
Engineering Track Poster
TimeWednesday, July 12th5:04pm - 5:06pm PDT
LocationLevel 2 Exhibit Hall
Back-End Design
Embedded Systems
Front-End Design