DAS-MP: Enabling High-Quality Macro Placement with Enhanced Dataflow Awareness

TL;DR

DAS-MP addresses the limited dataflow perspective in macro placement by extracting hidden macro–cell and cell–cell dataflow and integrating it into placement constraints. The method introduces a three-way dataflow model, a two-stage fine-tuning workflow that accounts for macro area and orientation, and a dataflow-guided SA/Sequence-Pair optimization. Empirical results on seven NanGate45 benchmarks show average HPWL improvements of 7.9% and substantial congestion reductions (≈82.5%), along with notable WNS/TNS improvements, while incurring modest runtime overhead. The approach is designed to be portable as a plugin for existing design flows, enabling more accurate, dataflow-aware macro placement with co-optimization of macro and standard-cell clusters.

Abstract

Dataflow is a critical yet underexplored factor in automatic macro placement, which is becoming increasingly important for developing intelligent design automation techniques that minimize reliance on manual adjustments and reduce design iterations. Existing macro or mixed-size placers with dataflow awareness primarily focus on intrinsic relationships among macros, overlooking the crucial influence of standard cell clusters on macro placement. To address this, we propose DAS-MP, which extracts hidden connections between macros and standard cells and incorporates a series of algorithms to enhance dataflow awareness, integrating them into placement constraints for improved macro placement. To further optimize placement results, we introduce two fine-tuning steps: (1) congestion optimization by taking macro area into consideration, and (2) flipping decisions to determine the optimal macro orientation based on the extracted dataflow information. By integrating enhanced dataflow awareness into placement constraints and applying these fine-tuning steps, the proposed approach achieves an average 7.9% improvement in half-perimeter wirelength (HPWL) across multiple widely used benchmark designs compared to a state-of-the-art dataflow-aware macro placer. Additionally, it significantly improves congestion, reducing overflow by an average of 82.5%, and achieves improvements of 36.97% in Worst Negative Slack (WNS) and 59.44% in Total Negative Slack (TNS). The approach also maintains efficient runtime throughout the entire placement process, incurring less than a 1.5% runtime overhead. These results show that the proposed dataflow-driven methodology, combined with the fine-tuning steps, provides an effective foundation for macro placement and can be seamlessly integrated into existing design flows to enhance placement quality.

Figures (12)

• Figure 1: The illustration shows all dataflow connections, with the uncovered dataflow connections indicated by shaded lines.
• Figure 2: The overview of the overall DAS-MP methodology. The first part is the proposed enhanced dataflow extraction process where light yellow rectangles outline all the newly proposed features in dataflow extraction. The second and third parts are the proposed fine-tuning stages where light blue rectangles list the macro specificity analysis and orientation optimization.
• Figure 3: One-hop direct dataflow connection of macro cluster-cell cluster. The strength is given based on the bit width of the extracted connections. Illustration of one-hop indirect dataflow connection from macro cluster-macro cluster. The left figure treats the cell cluster as the smallest unit, and the right figure treats the single cell instance as the smallest unit.
• Figure 4: Illustration of newly established two-hop dataflow connection macro cluster-cell cluster-cell cluster.
• Figure 5: Comparison between one-hop connected cell cluster (only macro cluster-cell cluster) and two-hop connected cell cluster (after considering macro cluster-cell cluster-cell cluster) in an example design swerv_wrapper.