Recursive Learning-Based Virtual Buffering for Analytical Global Placement

2025-06-07

arXiv.org

AI Summary

To address the interconnect-delay-dominated design challenges in advanced technology nodes, the high computational cost of conventional buffer insertion methods, and the lack of closed-loop electrical rule checking (ERC) feedback, this paper proposes MLBuf-RePlAce—the first open-source, learning-driven, virtual-buffer-aware global placement framework. It deeply integrates buffer prediction into the analytical RePlAce optimization flow, introduces a novel recurrent neural network to jointly model buffer type and location, and enables real-time ERC violation detection with gradient-differentiable constraints for the first time. The framework is fully integrated end-to-end into OpenROAD. Experimental results show that, within the OpenROAD flow, MLBuf-RePlAce reduces total negative slack (TNS) by 31% on average (up to 56%); in commercial flows, it achieves an average TNS reduction of 28% (up to 53%), with only a 0.2% increase in post-route power consumption.

Technology Category

• Machine Learning: Hardware-aware ML
• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Due to the skewed scaling of interconnect versus cell delay in modern technology nodes, placement with buffer porosity (i.e., cell density) awareness is essential for timing closure in physical synthesis flows. However, existing approaches face two key challenges: (i) traditional van Ginneken-Lillis-style buffering approaches are computationally expensive during global placement; and (ii) machine learning-based approaches, such as BufFormer, lack a thorough consideration of Electrical Rule Check (ERC) violations and fail to"close the loop"back into the physical design flow. In this work, we propose MLBuf-RePlAce, the first open-source learning-driven virtual buffering-aware analytical global placement framework, built on top of the OpenROAD infrastructure. MLBuf-RePlAce adopts an efficient recursive learning-based generative buffering approach to predict buffer types and locations, addressing ERC violations during global placement. We compare MLBuf-RePlAce against the default virtual buffering-based timing-driven global placer in OpenROAD, using open-source testcases from the TILOS MacroPlacement and OpenROAD-flow-scripts repositories. Without degradation of post-route power, MLBuf-RePlAce achieves (maximum, average) improvements of (56%, 31%) in total negative slack (TNS) within the open-source OpenROAD flow. When evaluated by completion in a commercial flow, MLBuf-RePlAce achieves (maximum, average) improvements of (53%, 28%) in TNS with an average of 0.2% improvement in post-route power.

Problem

• Addressing buffer porosity awareness in global placement
• Reducing computational cost of buffering during placement
• Mitigating ERC violations in learning-based buffering approaches

Innovation

• Recursive learning-based virtual buffering approach
• Open-source learning-driven global placement framework
• Addresses ERC violations during global placement