arXiv subjects

Rengang Zhang

Publications and source records attributed to Rengang Zhang.

2 recordsLinked to original sources

arXiv2026

E-ALS: Approximation-Potential-Aware E-Graph Rewriting for Approximate Logic Synthesis

Approximate logic synthesis (ALS) improves circuit power, performance, and area by trading exact correctness for bounded functional error. However, existing structural ALS methods largely overlook structural bias: even functionally equivalent netlists can expose markedly different approximation opportunities and lead to substantially different outcomes under the same downstream ALS flow. Our experiments show that this effect can induce final area gaps of up to 42.77%. To unlock this opportunity, we propose E-ALS, an e-graph-based framework for approximation-aware structural search. E-ALS combines Function-Reduced Saturation, an ALS-coupled surrogate, search-based extraction, and budget-guided refinement to identify approximation-friendly equivalent structures. Experiments on well-established arithmetic and logic benchmarks show that E-ALS achieves additional area reductions of 3.2 and 7.0 percentage points under maximum Hamming-Distance and Error-Distance constraints, respectively. Code is available in https://github.com/ZenuSunB/Ecompile.git.

cs.ET↗

arXiv2025

InF-ATPG: Intelligent FFR-Driven ATPG with Advanced Circuit Representation Guided Reinforcement Learning

Automatic test pattern generation (ATPG) is a crucial process in integrated circuit (IC) design and testing, responsible for efficiently generating test patterns. As semiconductor technology progresses, traditional ATPG struggles with long execution times to achieve the expected fault coverage, which impacts the time-to-market of chips. Recent machine learning techniques, like reinforcement learning (RL) and graph neural networks (GNNs), show promise but face issues such as reward delay in RL models and inadequate circuit representation in GNN-based methods. In this paper, we propose InF-ATPG, an intelligent FFR-driven ATPG framework that overcomes these challenges by using advanced circuit representation to guide RL. By partitioning circuits into fanout-free regions (FFRs) and incorporating ATPG-specific features into a novel QGNN architecture, InF-ATPG enhances test pattern generation efficiency. Experimental results show InF-ATPG reduces backtracks by 55.06\% on average compared to traditional methods and 38.31\% compared to the machine learning approach, while also improving fault coverage.

cs.AR↗