MergePath-SpMM: Parallel Sparse Matrix-Matrix Algorithm for Graph Neural Network Acceleration

Keyphrases

• Sparse Matrices 100%
• Matrix Algorithm 100%
• Graph Neural Network 100%
• Neural Network Acceleration 100%
• Load Balancing 50%
• Hardware Accelerator 50%
• Massively Parallel 25%
• Predictive Analytics 25%
• Hardware Complexity 25%
• Parallelization 25%
• Superior Performance 25%
• Performance Scale 25%
• Robust Performance 25%
• Fine-grained Parallelism 25%
• Needs-based 25%
• Matrix Computation 25%
• Load Imbalance 25%
• Real-world Graphs 25%
• Specialized Hardware 25%
• Workload Imbalance 25%
• GPU Systems 25%
• Auto-tune 25%

Computer Science

• Parallelism 100%
• Hardware Accelerator 100%
• Graph Neural Network 100%
• Superior Performance 50%
• Load Balancing 50%
• Computation Matrix 50%
• Load Imbalance 50%
• Load-Balanced 50%
• Predictive Analytics 50%
• Computer Hardware 50%
• Graphics Processing Unit 50%
• Synchronism 50%