HW-Flow-Fusion: Inter-Layer Scheduling for Convolutional Neural Network Accelerators with Dataflow Architectures

Keyphrases

• Data Reuse 100%
• Dataflow Architecture 100%
• Convolutional Neural Network Accelerator 100%
• Convolutional Neural Network 50%
• Multiple Layers 50%
• Memory Efficiency 25%
• Latency 25%
• Analytical Model 25%
• Bandwidth Utilization 25%
• Data Dependency 25%
• Power Budget 25%
• Network Layer 25%
• Scheduling Techniques 25%
• Memory Bandwidth 25%
• Efficient Scheduling 25%
• VGG16 25%
• Optimal Scheduling 25%
• Reconfigurable Accelerator 25%
• Partitioning Strategy 25%
• Resource Partitioning 25%
• Recomputation 25%
• Communication Volume 25%
• Data Movement 25%
• Model Partition 25%
• Reuse Models 25%
• Layer Sharing 25%
• Mapping Framework 25%
• ResNet18 25%
• Scheduling Framework 25%
• Spatial Array 25%
• Map Space 25%

Computer Science

• Convolutional Neural Network 100%
• Data Reuse 100%
• Data Dependency 25%
• Network Layer 25%
• Reuse Models 25%
• Analytical Model 25%
• Bandwidth Efficiency 25%
• Memory Bandwidth 25%
• Computation-To-Communication Ratio 25%