PrefixAgent: An LLM-Powered Design Framework for Efficient Prefix Adder Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zuo, Dongsheng, Zhu, Jiadong, Luo, Yang, Ma, Yuzhe
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913932386500608
author Zuo, Dongsheng
Zhu, Jiadong
Luo, Yang
Ma, Yuzhe
author_facet Zuo, Dongsheng
Zhu, Jiadong
Luo, Yang
Ma, Yuzhe
contents Prefix adders are fundamental arithmetic circuits, but their design space grows exponentially with bit-width, posing significant optimization challenges. Previous works face limitations in performance, generalization, and scalability. To address these challenges, we propose PrefixAgent, a large language model (LLM)-powered framework that enables efficient prefix adder optimization. Specifically, PrefixAgent reformulates the problem into subtasks including backbone synthesis and structure refinement, which effectively reduces the search space. More importantly, this new design perspective enables us to efficiently collect enormous high-quality data and reasoning traces with E-graph, which further results in an effective fine-tuning of LLM. Experimental results show that PrefixAgent synthesizes prefix adders with consistently smaller areas compared to baseline methods, while maintaining scalability and generalization in commercial EDA flows.
format Preprint
id arxiv_https___arxiv_org_abs_2507_06127
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PrefixAgent: An LLM-Powered Design Framework for Efficient Prefix Adder Optimization
Zuo, Dongsheng
Zhu, Jiadong
Luo, Yang
Ma, Yuzhe
Hardware Architecture
Artificial Intelligence
Prefix adders are fundamental arithmetic circuits, but their design space grows exponentially with bit-width, posing significant optimization challenges. Previous works face limitations in performance, generalization, and scalability. To address these challenges, we propose PrefixAgent, a large language model (LLM)-powered framework that enables efficient prefix adder optimization. Specifically, PrefixAgent reformulates the problem into subtasks including backbone synthesis and structure refinement, which effectively reduces the search space. More importantly, this new design perspective enables us to efficiently collect enormous high-quality data and reasoning traces with E-graph, which further results in an effective fine-tuning of LLM. Experimental results show that PrefixAgent synthesizes prefix adders with consistently smaller areas compared to baseline methods, while maintaining scalability and generalization in commercial EDA flows.
title PrefixAgent: An LLM-Powered Design Framework for Efficient Prefix Adder Optimization
topic Hardware Architecture
Artificial Intelligence
url https://arxiv.org/abs/2507.06127