Efficient sparse-dense matrix-matrix multiplication on GPUs using the customized sparse storage format

Keyphrases

• Global Memory 100%
• Dense Matrix 100%
• Sparse Matrices 100%
• Sparse Formats 100%
• Matrix multiplication 100%
• Machine Learning 50%
• NVIDIA GPU 50%
• Memory Access 50%
• Shared Memory 50%
• NVIDIA 50%
• Irregular Data 50%
• Byte 50%
• Sparse Structure 50%
• Public Dataset 50%
• Scientific Computing 50%
• Poor Performance 50%
• P100 50%
• Generated Matrix 50%
• Data Access 50%
• Coalescent 50%
• Resource-poor 50%
• Low Resource Utilization 50%
• Memory Traffic 50%
• Roofline Performance Model 50%
• Performance Optimization 50%
• Many-core Architecture 50%
• CUSPARSE 50%
• Storage Data 50%
• Memory Saving 50%
• Tesla 50%
• Memory Reuse 50%

Computer Science

• Graphics Processing Unit 100%
• Matrix Multiplication 100%
• Experimental Result 20%
• Machine Learning 20%
• Learning System 20%
• Data Structure 20%
• Shared Memory 20%
• Data Access 20%
• Resource Utilisation 20%
• Natural Sciences Computing 20%
• Memory Traffic 20%
• Roofline Performance Model 20%
• Performance Optimization 20%
• Global Memory Access 20%