Using ChatGPT for developing and simulating a circuit in VHDL

This article describes the experience of using ChatGPT for developing and simulating an 8253/ 8254 counter/timer in VHDL without prior knowledge of the language.An 8253/ 8254 chip contains three counters/timers, each of which can work in six different modes of operation.The design is hierarchical and was developed in steps; initially a single counter was developed and simulated; next, a higher-level module containing three such timers was developed and a particular configuration was simulated.ChatGPT suggested the proper tools for the user's platform, the VHDL code, the testbenches, the procedure (commands) and the troubleshooting.Manual interventions were necessary for fine-tuning.The quality and accuracy of the code generated by ChatGPT were found to be proportional to the user's specifications, which implies that the specifications must be accurate.In order to get a satisfactory result, several trials (initial specifications, simulation, review, revised specifications) were needed.What made this experiment interesting was that the author did not have prior VHDL language know-how.The experience was interesting and leaded to the following findings: ChatGPT can produce code but it is not guaranteed that it won't have syntactical or logical errors; the quality of the generated code is directly dependent on the clarity of the specifications, otherwise the user wastes a lot of time correcting errors and re-defining specifications; ChatGPT is a valuable self-learning tool for both students and teachers providing personalized and interactive learning.