Exploring Missed Optimizations in WebAssembly Optimizers

Zhibo Liu, Dongwei Xiao, Zongjie Li, Shuai Wang, Wei Meng

Abstract: The prosperous trend of deploying complex applications to web browsers has boosted the development of WebAssembly (wasm) compilation toolchains. Software written in different high-level programming languages are compiled into wasm executables, which can be executed fast and safely in a virtual machine. The performance of wasm executables depends highly on compiler optimizations. Despite the prosperous use of wasm executables, recent research has indicated that real-world wasm applications are slower than anticipated, suggesting deficiencies in wasm optimizations. This paper aims to present the first systematic and in-depth understanding of the status quo of wasm optimizations. To do so, we present DITWO, a differential testing framework to uncover missed optimizations (MO) of wasm optimizers. DITWO compiles a C program into both native x86 executable and wasm executable, and differentiates optimization indication traces (OITraces) logged by running each executable to uncover MO. Each OITrace is composed with global variable writes and function calls, two performance indicators that practically and systematically reflect the optimization degree across wasm and native executables. Our analysis of the official wasm optimizer, wasm-opt, successfully identifies 1,293 inputs triggering MO of wasm-opt. With extensive manual effort, we identify nine root causes for all MO, and we estimate that fixing discovered MO can result in a performance improvement of at least 17.15%. We also summarize four lessons from our findings to deliver better wasm optimizations.