OSDI '23

Currently Playing: OSDI '23 - Effectively Scheduling Computational Graphs of Deep Neural Networks toward Their...

OSDI '23 - Effectively Scheduling Computational Graphs of Deep Neural Networks toward Their Domain-Specific Accelerators Jie Zhao, Information Engineering University, Siyuan Feng, Shanghai Jiao Tong University, Xiaoqiang Dan, Stream Computing Inc., Fei Liu, Stream Computing Inc., Chengke Wang, Stream Computing Inc., Sheng Yuan, Stream Computing Inc., Wenyuan Lv, Stream Computing Inc., Qikai Xie, Stream Computing Inc. Fully exploiting the computing power of an accelerator specialized for deep neural networks (DNNs) calls for the synergy between network and hardware architectures, but existing approaches partition a computational graph of DNN into multiple sub-graphs by abstracting away hardware architecture and assign resources to each sub-graph, not only producing redundant off-core data movements but also under-utilizing the hardware resources of a domain-specific architecture (DSA). This paper introduces a systematic approach for effectively scheduling DNN computational graphs on DSA platforms. By fully taking into account hardware architecture when partitioning a computational graph into coarse-grained sub-graphs, our work enables the synergy between network and hardware architectures, addressing several challenges of prior work: (1) it produces larger but fewer kernels, converting a large number of off-core data movements into on-core data exchanges; (2) it exploits the imbalanced memory usage distribution across DNN network architecture, better saturating the DSA memory hierarchy; (3) it enables across-layer instruction scheduling not studied before, further exploiting the parallelism across different specialized compute units. Results of seven DNN inference models on a DSA platform show that our work outperforms TVM and AStitch by 11.15× and 6.16×, respectively, and obtains throughput competitive to the vendor-crafted implementation. A case study on GPU also demonstrates that generating kernels for our sub-graphs can surpass CUTLASS with and without convolution fusion by 1.06× and 1.23×, respectively. View the full OSDI '23 program at https://www.usenix.org/conference/osdi23/technical-sessions