Meta-tracing Interpreters in WebAssembly
summary
A meta-tracing pipeline for improving interpreted programs compiled to WebAssembly, claiming offline trace capture of executions, tandem-repeat detection (DNA-sequence algorithm) to find hot user loops, and reconstruction of optimized trace fragments that remove interpreter dispatch overhead and use on-stack replacement for control transfer. Independent tools instrument wasm with host calls, emit traces, build fragments with guarded exits, and reintegrate them; prototypes exist for several phases but full end-to-end deoptimization in wasm remains unresolved.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 0:22Meta-tracing for WebAssembly
- 1:47Problem scenario and solutions
- 3:54Why trace-based optimization
- 5:41Meta-tracing Java interpreter
- 8:10Off-trace exits in WebAssembly
- 9:56System design and architecture
- 11:00Trace data structure
- 12:03Finding user loops
- 19:51Beyond interpreters
- 20:35Adapting to new interpreters
- 21:51Questions from the audience
- 23:00Comparison with partial evaluation
related talks
32mThe Amazing Trace: Enhancing Distributed WebAssembly With OTEL- Brooks Townsend
16mDebugging WebAssembly? Put some Whamm on it!
16mWasmgrind: Towards Dynamic Concurrency Analysis for Multithreaded WebAssembly
32mAll The Memory Safety Of C Combined With All The Blazing Speed Of JavaScript
40mJoy of Meta-Tracing Just-in-Time Compilation: More Than Just a VM Generator
17mWasm-R3: Record-Reduce-Replay for Realistic and Standalone WebAssembly Benchmarks
30mMechanising WebAssembly Well?
27mObserving the Future: Embracing OTEL in WebAssembly