Performance engineering for a tall & skinny matrix multiplication kernels on GPUs

Keyphrases

• Matrix multiplication 100%
• Performance Engineering 100%
• NVIDIA 50%
• GPGPU 50%
• Maximum Performance 50%
• CuBLAS 50%
• Square Matrix 50%
• Potential Performance 50%
• Mapping Scheme 50%
• Parallelization 50%
• Code Generation 50%
• Volta 50%
• General Matrix 50%
• Matrix Size 50%
• Generation Approach 50%
• Domain of Interest 50%
• Roofline Model 50%

Computer Science

• Multiplication Kernel 100%
• Matrix Multiplication 100%
• Autotuning 50%
• Code Generation 50%
• Parallelization 50%
• Thread Distribution 50%
• Mapping Scheme 50%
• Domain 50%
• Roofline Model 50%

INIS

• matrices 100%
• performance 100%
• kernels 100%
• engineering 100%
• implementation 50%
• libraries 16%
• mapping 16%
• distribution 16%
• range 16%
• size 16%