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

TL;DR

DyCL is a novel compilation approach that transforms dynamic neural networks into multiple static sub-networks with a control flow model, enabling existing DL compilers to efficiently compile and execute DyNNs with improved performance.

Contribution

DyCL introduces program analysis and transformation techniques to compile dynamic neural networks by splitting them into static sub-networks and synthesizing a control flow host module.

Findings

Achieved 100% success in compiling all tested DyNNs.

Generated executables run between 1.12x and 20.21x faster than original DyNNs.

Significantly improves compilation success and runtime performance for DyNNs.

Abstract

DL compiler's primary function is to translate DNN programs written in high-level DL frameworks such as PyTorch and TensorFlow 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 \tool, a general approach that enables any existing DL compiler to successfully compile DyNNs. \tool 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, \tool 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, \tool 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 \tool, achieving a 100\% success rate in compiling all dynamic neural networks. Moreover, the compiled executables generated by \tool exhibit significantly improved performance, running between $1.12\times$ and $20.21\times$ faster than the original DyNNs executed on general-purpose DL frameworks.