POET: Power-Oriented Evolutionary Tuning for LLM-Based RTL PPA Optimization

2026-03-18

AI Summary

This work addresses the challenges of ensuring functional correctness and prioritizing power optimization when leveraging large language models (LLMs) to refine RTL code for improved power, performance, and area (PPA). The authors propose POET, a novel framework that uniquely integrates differential testing with LLM-generated transformations to effectively suppress hallucinations and guarantee 100% functional correctness. Furthermore, POET introduces a power-prioritized non-dominated sorting evolutionary search mechanism that automatically converges toward low-power Pareto-optimal solutions without manual hyperparameter tuning. Evaluated on 40 designs from the RTL-OPT benchmark, POET consistently achieves the best power results across all cases while delivering competitive improvements in area and delay.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Applying large language models (LLMs) to RTL code optimization for improved power, performance, and area (PPA) faces two key challenges: ensuring functional correctness of optimized designs despite LLM hallucination, and systematically prioritizing power reduction within the multi-objective PPA trade-off space. We propose POET (Power-Oriented Evolutionary Tuning), a framework that addresses both challenges. For functional correctness, POET introduces a differential-testing-based testbench generation pipeline that treats the original design as a functional oracle, using deterministic simulation to produce golden references and eliminating LLM hallucination from the verification process. For PPA optimization, POET employs an LLM-driven evolutionary mechanism with non-dominated sorting, power-first intra-level ranking, and proportional survivor selection to steer the search toward the low-power region of the Pareto front without manual weight tuning. Evaluated on the RTL-OPT benchmark across 40 diverse RTL designs, POET achieves 100% functional correctness, the best power on all 40 designs, and competitive area and delay improvements.

Problem

• LLM hallucination
• functional correctness
• power optimization
• PPA trade-off
• RTL optimization

Innovation

• LLM-based RTL optimization
• functional correctness verification
• power-oriented evolutionary tuning
• differential testing
• Pareto-front exploration