Natural Language to Verilog: Design of a Recurrent Spiking Neural Network using Large Language Models and ChatGPT

Paola Vitolo,George Psaltakis,Michael Tomlinson,Gian Domenico Licciardo,Andreas G. Andreou

2024-10-23

Abstract:This paper investigates the use of Large Language Models (LLMs) and natural language prompts to generate hardware description code, namely Verilog. Building on our prior work, we employ OpenAI's ChatGPT4 and natural language prompts to synthesize an RTL Verilog module of a programmable recurrent spiking neural network, while also generating test benches to assess the system's correctness. The resultant design was validated in three simple machine learning tasks, the exclusive OR, the IRIS flower classification and the MNIST hand-written digit classification. Furthermore, the design was validated on a Field-Programmable Gate Array (FPGA) and subsequently synthesized in the SkyWater 130 nm technology by using an open-source electronic design automation flow. The design was submitted to Efabless Tiny Tapeout 6.

Hardware Architecture,Artificial Intelligence

What problem does this paper attempt to address?

The problem this paper attempts to address is the generation of hardware description code (Verilog) using large language models (LLM) and natural language prompts to accelerate and simplify the process of complex hardware design. Specifically, the authors utilized OpenAI's ChatGPT4 and natural language prompts to synthesize an RTL Verilog module of a programmable recurrent spiking neural network (RSNN) and generated a testbench to evaluate the correctness of the system. The main objectives of the paper include: 1. **Generating hardware description code**: Using LLM to generate complex hardware description code, reducing the time and errors associated with manual code writing. 2. **Design verification**: Verifying the generated design through three simple machine learning tasks (XOR, Iris classification, and MNIST handwritten digit classification). 3. **Implementation and synthesis**: Verifying the design on FPGA and synthesizing it using an open-source electronic design automation (EDA) flow in SkyWater 130 nm technology, and submitting it to Efabless Tiny Tapeout 6. Through these objectives, the paper aims to demonstrate the potential of LLM in hardware design, particularly in generating complex hardware architectures and testbenches. This can significantly shorten the design and verification time, improving the efficiency of hardware design.

Natural Language to Verilog: Design of a Recurrent Spiking Neural Network using Large Language Models and ChatGPT

Natural Language to Verilog: Design of a Recurrent Spiking Neural Network using Large Language Models and ChatGPT

Designing Silicon Brains using LLM: Leveraging ChatGPT for Automated Description of a Spiking Neuron Array

Benchmarking Large Language Models for Automated Verilog RTL Code Generation

VeriGen: A Large Language Model for Verilog Code Generation

Data is all you need: Finetuning LLMs for Chip Design via an Automated design-data augmentation framework

Natural language is not enough: Benchmarking multi-modal generative AI for Verilog generation

SpikeLLM: Scaling up Spiking Neural Network to Large Language Models via Saliency-based Spiking

Chip-Chat: Challenges and Opportunities in Conversational Hardware Design

Evaluating Large Language Models for Automatic Register Transfer Logic Generation via High-Level Synthesis

Large Language Model for Verilog Generation with Golden Code Feedback

Language Modeling on a SpiNNaker 2 Neuromorphic Chip

RTLLM: An Open-Source Benchmark for Design RTL Generation with Large Language Model

The implementation of a Deep Recurrent Neural Network Language Model on a Xilinx FPGA

Evaluating LLMs for Hardware Design and Test

Revisiting VerilogEval: Newer LLMs, In-Context Learning, and Specification-to-RTL Tasks

Exploring and Characterizing Large Language Models For Embedded System Development and Debugging

Advanced Large Language Model (LLM)-Driven Verilog Development: Enhancing Power, Performance, and Area Optimization in Code Synthesis

A Multi-Expert Large Language Model Architecture for Verilog Code Generation

OPL4GPT: An Application Space Exploration of Optimal Programming Language for Hardware Design by LLM