Benchmarking WebAssembly multithreading: performance, scalability, and browser-specific challenges

Abstract

WebAssembly introduces high-performance execution for web applications, but its concurrency model relies on JavaScript Web Workers rather than native threading, leading to unique performance characteristics. This study benchmarks WebAssembly’s multithreading capabilities using OpenCV.js across Chromium, Firefox, and Node.js, analyzing execution time, scalability, and platform-specific behavior. Results show that WebAssembly threading can improve performance for parallel workloads but exhibits undesirable characteristics due to the asynchronous nature of Web Workers, which can affect efficiency in certain scenarios. Varying the size of the pre-spawned worker pool demonstrates that pre-spawning workers reduces delays. However, maintaining a worker pool may not always be desirable, and a large pool can lead to resource overcommitment, thereby affecting efficiency. Additionally, differences in browser implementations result in distinct scalability trends, with Firefox exhibiting better multi-threaded scaling than Chromium in certain cases. While WebAssembly enables multithreading, its reliance on browser-specific worker management introduces inconsistencies that limit performance predictability. These findings highlight the need for further optimization and standardization in WebAssembly’s concurrency model to improve cross-platform scalability and efficiency in high-performance web applications.