MemFlow: Memory-Driven Data Scheduling with Datapath Co-Design in Accelerators for Large-Scale Inference Applications

Keyphrases

• Large-scale Inference 100%
• Data Scheduling 100%
• Accelerator Design 100%
• DRAM Access 66%
• Neural Network 33%
• Hardware Accelerator 33%
• Design Tools 33%
• Singular Value Decomposition 33%
• Performance Improvement 33%
• Design Methodology 33%
• Principal Coordinate Analysis (PCoA) 33%
• Limited Resources 33%
• Memory Hierarchy 33%
• Data Access 33%
• Energy-performance Tradeoff 33%
• Inference Algorithms 33%
• Scratchpad 33%
• Scratchpad Memory 33%

Computer Science

• Datapath 100%
• Participatory Design 100%
• Neural Network 25%
• Hardware Accelerator 25%
• Singular Value 25%
• Memory Hierarchy 25%
• Data Access 25%
• Large Data Set 25%
• Scratchpad Memory 25%
• Principle Component Analysis 25%