DyCL: Dynamic Neural Network Compilation Via Program Rewriting and Graph Optimization

Abstract

The deep learning (DL) compiler serves as a vital infrastructure component to enable the deployment of deep neural networks on diverse hardware platforms such as mobile devices and Raspberry Pi. DL compiler's primary function is to translate DNN programs written in high-level DL frameworks such as PyTorch and Ten-sorFlow into portable executables. These executables can then be flexibly executed by the deployed host programs. However, existing DL compilers rely on a tracing mechanism, which involves feeding a runtime input to a neural network program and tracing the program execution paths to generate the computational graph necessary for compilation. Unfortunately, this mechanism falls short when dealing with modern dynamic neural networks (DyNNs) that possess varying computational graphs depending on the inputs. Consequently, conventional DL compilers struggle to accurately compile DyNNs into executable code. To address this limitation, we propose DyCL, a general approach that enables any existing DL compiler to successfully compile DyNNs. DyCL tackles the dynamic nature of DyNNs by introducing a compilation mechanism that redistributes the control and data flow of the original DNN programs during the compilation process. Specifically, DyCL develops program analysis and program transformation techniques to convert a dynamic neural network into multiple sub-neural networks. Each sub-neural network is devoid of conditional statements and is compiled independently. Furthermore, DyCL synthesizes a host module that models the control flow of the DyNNs and facilitates the invocation of the sub-neural networks. Our evaluation demonstrates the effectiveness of DyCL, achieving a 100% success rate in compiling all dynamic neural networks. Moreover, the compiled executables generated by DyCL exhibit significantly improved performance, running between 1.12× and 20.21× faster than the original DyNNs executed on general-purpose DL frameworks.