EDA-Aware RTL Generation with Large Language Models

Mubashir ul Islam, Humza Sami, Pierre-Emmanuel Gaillardon, Valerio Tenace · 2024

Read EDA-Aware RTL Generation with Large Language Models by Mubashir ul Islam, Humza Sami, Pierre-Emmanuel Gaillardon, Valerio Tenace (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-04485.txt.

Description

Mubashir ul Islam and three co-authors wrote a 2024 paper on using language models to write chip design code. The code, called RTL, describes how a chip's logic behaves. Even top models rarely produce error-free RTL on the first try. Fixing their output by hand swells the checking workload.

Their framework, AIvril2, verifies its own work in a loop. Several agents share the task, and error logs from standard design tools feed back as guidance. The agents then repair both syntax and function flaws on their own. The setup works with any language model. On a standard human-written benchmark, code quality rose by a factor of about 3.4 over prior methods. Best-case pass rates hit 77 percent for Verilog and 66 percent for VHDL. The work makes model-driven chip design more trustworthy.

Catalog record

CREATORMubashir ul Islam, Humza Sami, Pierre-Emmanuel Gaillardon, Valerio Tenace
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE26 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-04485

Browse the full library · About the Tower of Babel