The deep learning (DL) compiler serves as a vital infrastructure component, enabling the deployment of deep neural networks on diverse hardware platforms such as mobile devices and Raspberry Pi.
Its primary function is to translate DNN programs written in high-level DL frameworks like PyTorch and TensorFlow into portable executables. These executables can then be flexibly executed by the deployed host programs. However, existing DL compilers rely on the traced mechanism, which involves feeding a runtime input to a neural network program and tracing its execution path 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 input provided. Consequently, conventional DL compilers struggle to accurately compile DyNNs into executable code. To address this limitation, we propose DyCL, a flexible approach that enables existing DL compilers 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 programs during the compilation process. Specifically, DyCL applies program analysis and 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 API 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 $1.12\times$ and $20.21\times$ faster than the original DyNNs executed on general-purpose DL frameworks.