GNNIE: GNN Inference Engine with Load-balancing and Graph-specific Caching

Keyphrases

• Caching 100%
• Feature Vector 100%
• Load Balancing 100%
• Inference Engine 100%
• Graph Neural Network 100%
• Vertex Feature 100%
• Neural Network Inference 100%
• High Sparsity 100%
• Load Graph 100%
• High Power 50%
• High Cost 50%
• Adjacency Matrix 50%
• Weight Vector 50%
• Unbalanced Load 50%
• Load Redistribution 50%
• CPU-GPU 50%
• Power-law Degree Distribution 50%
• Caching Scheme 50%
• Random Memory Access 50%
• Variable Sparsity 50%
• Vertex Neighbourhood 50%
• DRAM Access 50%

Computer Science

• Sparsity 100%
• Load Balancing 100%
• Inference Engines 100%
• Network Inference 100%
• Graph Neural Network 100%
• Feature Vector 100%
• Random Access Memory 50%
• Adjacency Matrix 50%
• Degree Distribution 50%