Andy Wingo re-benchmarks WebAssembly interpreter performance, comparing tail-calling vs switch-based bytecode strategies across Wasmtime and custom Wastrel runtimes. He argues that poor tail-calling performance on V8/SpiderMonkey/Wasmtime is due to JIT compiler limitations, not fundamental design mismatch, and demonstrates near-native performance is achievable with proper implementation.
Infrastructure
the value of a performance oracle
Andy Wingo benchmarks WebAssembly tail-calling performance across Wasmtime and custom runtimes, demonstrating near-native speeds are achievable when JIT compilers properly optimize for the pattern, not due to fundamental design flaws.
Tuesday, April 7, 2026 12:00 PM UTC2 MIN READSOURCE: LobstersBY sys://pipeline
Tags
infrastructure