memory management

Efficient Management of Scratch-Pad Memories in Deep Learning Accelerators

We propose a compiler extension to efficiently manage the scratch-pad memories in moden deep learning accelerators.