Evaluating characteristics of CUDA communication primitives on high-bandwidth interconnects

Computer Science

• Application Developer 100%
• Clock Frequency 33%
• Data Placement 66%
• data-intensive application 33%
• Graphics Processing Unit 66%
• Heterogeneous System 33%
• Learning System 33%
• Machine Learning 33%
• Measurement Methodology 33%
• Memory Bandwidth 33%
• Memory Transfer 33%
• Microbenchmarks 33%
• Power Management 33%
• Relative Performance 33%
• Synchronism 33%
• System Developer 66%
• System Interconnect 33%
• system level design 33%
• Systems and Application 33%
• Virtual Memory 33%

Keyphrases

• Data Transfer Method 25%
• Memcpy 25%