HGRP: Heterogeneous Graph Neural Network for Routability Prediction
HGRP targets routability prediction in VLSI physical design by representing circuits as heterogeneous graphs with three distinct node types—cells, nets, and routing grids. It introduces a staged message-passing pipeline that propagates information from logical topology to physical layout to spatial grid adjacency, aiming to capture the coupling between netlist structure and routing congestion. Accurate congestion estimation matters because it provides rapid feedback during placement, avoiding costly iterations through detailed routing.
HGRP presents a conceptually clean architecture, but its empirical footprint is too small to support a rigorous state-of-the-art claim. The evaluation relies on only two unseen designs from the CircuitNet benchmark, and the paper omits the directly comparable HybridNet baseline despite citing it as related work. While the ablations are internally consistent, the 18.69% relative F1 improvement over LHNN rests on a razor-thin test bed and is undermined by lower overall classification accuracy (0.8803 vs. 0.8966 for NetlistGNN). As presented, the paper is a reasonable architectural proposal rather than a rigorously validated advance.
The heterogeneous graph construction is well motivated and explicit: modeling cells, nets, and grids as typed nodes with relation-specific edges directly encodes topological connectivity, spatial mapping, and grid adjacency. The ablation study credibly supports this design. Removing grids cripples performance, which confirms that cell-level representations alone cannot capture local routing-resource competition. Likewise, staged propagation outperforms mixed aggregation because it preserves relation-specific semantics rather than blurring them.
The experimental scope is severely limited. Results are reported on just two hold-out designs from CircuitNet, so design-specific variance swamps any confidence in generalization, and no statistical testing is provided. Moreover, HybridNet [Zhao et al., 2023] is cited as a related dual-branch geometric/topological method but is excluded from Table 1 even though the authors retrained other baselines on CircuitNet. On the efficiency front, HGRP demands the highest memory (9.495 GB) and longest inference time (8.542 s) of all compared methods, yet scalability implications are never discussed. The overall classification accuracy also regresses compared with NetlistGNN (0.8803 vs. 0.8966), which the authors dismiss via class imbalance but which still signals weaker discrimination across the full grid distribution.
Within the HGRP framework, the evidence is internally coherent: grid nodes, attention, and staged propagation each improve the target metrics in the ablation study. The retraining of GPDL, NetlistGNN, and LHNN under the same dataset partition is procedurally fair. However, the comparative evidence is incomplete. The claimed 18.69% relative F1 improvement over LHNN is arithmetically correct, yet without design-level error bars or inclusion of HybridNet, it is impossible to separate architectural merit from dataset idiosyncrasy. The paper also shifts evaluation regimes between congestion (NRMSE, SSIM, F1) and wirelength (MAE, correlation), making cross-task claims of flexibility largely anecdotal.
Reproduction is partially enabled by public code and data. The GitHub repository and CircuitNet dataset are referenced, and some hyperparameters are listed. Still, critical details are missing, including batch size, training epochs, learning-rate schedule, and early-stopping criteria. The train/validation/test split is ambiguous because zero-riscy-a and zero-riscy-b are jointly described as used for validation and testing without clarifying which serves which role. Hardware is stated only as an NVIDIA A100 GPU. These gaps would force reproducers to guess implementation minutiae.
Routability prediction enables rapid evaluation of placement quality without routing, providing effective feedback for routability-driven placement optimization. Recent machine learning methods have achieved efficient routability prediction through data-driven approach. However, their prediction performance highly depends on the capability of input representations to capture essential circuit information. To address the limitations of existing circuit representations in simultaneously preserving logical topology, physical layout and routing resource information, this paper proposes HGRP, a heterogeneous graph neural network for routability prediction. HGRP models the cells, nets, and routing grids as distinct types of nodes and explicitly represents topological connectivity, spatial mapping, and grid adjacency, thereby constructing a heterogeneous graph representation that preserves multi-level circuit semantics. Based on the representation, a heterogeneous message passing mechanism is developed to enable information interaction and feature fusion among distinct types of nodes, effectively learning the complex correlations between circuit structures and routing behaviors. Experimental results on the CircuitNet dataset show that HGRP improves F1-score by 18.69\% relative improvements compared with the state-of-the-art in the congestion prediction task. The source code is publicly available at https://github.com/ShenShan123/HGPR.
Pick a starting point or write your own. Challenges run in the background, so you can keep reading while the AI investigates.
No challenges yet. Disagree with the review? Ask the AI to revisit a specific claim.