DGRL-Hardware

Here is the github repository for 'A Benchmark for Directed Graph Representation Learning in Hardware Designs'

For more details, please refer to our documents.

Methods Included:

The combinations of 1) GNN backbones/ Graph Transformers, 2) message passing direction, 3) and Magnetic Laplacian Positional Encoding:

• GNN backbones/ Graph Transformers

  DGCN, DiGCN, MagNet, GCN, GIN(E), GAT, GPS-T, GPS-P.

• Message Passing Direction

  undirected(-) , directed (DI), and bidirected (BI)

• Magnetic Laplacian Positional Encoding

  node PE (NPE), and edge PE (EPE)

Datasets Includes:

• High-Level Synthesis (HLS)
• Symbolic Reasoning (SR)
• Pre-routing Timing Prediction (TIME)
• Computational Graph (CG)
• Operational Amplifiers (AMP)