Abstract:Modern electronic design automation (EDA) flows depend on both implementation and signoff tools to perform timing-constrained power optimization (TCPO) through Engineering Change Orders (ECOs), which involve gate sizing and threshold-voltage ( V th )-assignment of standard cells. However, the signoff ECO optimization is highly time-consuming, and the power improvement is hard to predict in advance. Ever since the industrial benchmarks released by the ISPD-2012 gate-sizing contest, active research has been conducted extensively to improve the optimization process. Nonetheless, previous works were mostly based on heuristics or analytical methods whose timing models were oversimplified and lacked of formal validations from commercial signoff tools. In this paper, we propose ECO-GNN, a transferable graph-learning-based framework, which harnesses graph neural networks (GNNs) to perform commercial-quality signoff power optimization through discrete V th -assignment. One of the highlights of our framework is that it generates tool-accurate optimization results instantly on unseen netlists that are not utilized in the training process. Furthermore, we propose a subgraph approximation technique to improve training and inferencing time of the proposed GNN model. We show that design instances with non-overlapping subgraphs can be optimized in parallel so as to improve the inference time of the learning-based model. Finally, we implement a GNN-based explanation method to interpret the optimization results achieved by our framework. Experimental results on 14 industrial designs, including a RISC-V-based multi-core system and the renowned ISPD-2012 benchmarks, demonstrate that our framework achieves up to 14X runtime improvement with similar signoff power optimization quality compared with Synopsys PrimeTime , an industry-leading signoff tool.

An efficient leakage power optimization framework based on reinforcement learning with graph neural network

Timing-Driven Technology Mapping Approximation Based on Reinforcement Learning

ECO-GNN: Signoff Power Prediction using Graph Neural Networks with Subgraph Approximation

GCN-RL Circuit Designer: Transferable Transistor Sizing with Graph Neural Networks and Reinforcement Learning

An Open-Source AMS Circuit Optimization Framework Based on Reinforcement Learning—From Specifications to Layouts

Leakage Power Optimization for Clock Network Using Dual-Vth Technology

Leakage Power and Circuit Aging Cooptimization by Gate Replacement Techniques

Gate Replacement Techniques For Simultaneous Leakage And Aging Optimization

Power-gating-aware Scheduling with Effective Hardware Resources Optimization.

Logic Synthesis Optimization Sequence tuning using RL-based LSTM and Graph Isomorphism Network

Min-cut Based Leakage Power Aware Scheduling in High-Level Synthesis

Leakage-Power-Aware Scheduling with Dual-Threshold Voltage Design

Simultaneous Fine-grain Sleep Transistor Placement and Sizing for Leakage Optimization

Optimizing Routerless Network-on-Chip Designs: An Innovative Learning-Based Framework

NBTI-aware Adaptive Minimum Leakage Vector Selection Using a Linear Programming Approach

Leakage Power Reduction Through Dual Vth Assignment Considering Threshold Voltage Variation

Leakage Power Aware Scheduling In High-Level Synthesis

Statistical Timing and Power Optimization of Architecture and Device for FPGAs

Investigation of Dynamic Leakage-Suppression Logic Techniques Crossing Different Technology Nodes from 180 Nm Bulk CMOS to 7 Nm FinFET Plus Process

Study on leakage power estimation and reduction methodology of CMOS circuit

Localized On-Chip Power Delivery Network Optimization via Sequence of Linear Programming