Not So Fast: Analyzing the Performance of WebAssembly vs. Native Code

on YouTube ↗20m

" summary

Built Browsix-Wasm to run unmodified Unix programs in the browser and benchmarked them using SPEC CPU; WebAssembly binaries run about 45% slower on Firefox and 55% slower on Chrome, with peak slowdowns of 2.08× and 2.5×. Root causes include JIT codegen limits (poor register allocation and missed optimizations), reserved registers for GC roots, and mandatory stack/indirect-call checks; engineering fixes (e.g., file-append buffering) cut filesystem overhead ~5× and lowered that component to ~0.2%–1.1% of runtime.

AI-written summary · about

" chapters (11)

1. 0:00The evolution of web speed
2. 1:42Introduction to WebAssembly
3. 3:28WebAssembly architecture
4. 4:35Benchmark selection challenges
5. 6:12Implementing browser system calls
6. 7:52Performance optimization
7. 8:54Benchmark infrastructure
8. 10:03JIT compilation analysis
9. 12:17Performance counter results
10. 14:34Summary of findings
11. 15:24Q&A and conclusion

" related talks

• 21mHit the Ground Running with WebAssemblyRobert Aboukhalil · OpenJS World 2021
• 39mFast future with WebAssemblyMilica Mihajlija · Armada JS 2019
• 32mWebAssembly vs JavaScript: Which Is Faster?Alexandr Skachkov · OdessaJS 2018
• 53mWebAssembly – To the browser and beyond!Patrick Hamann · performance.now() 2019
• 17mA case for Web AssemblyElvina Valieva · BDX I/O 2018
• 29mWebAssembly - the Final FrontierAlbert Brand · AmsterdamJS Conference 2017
• 44mWebAssemblyNick Bray · GOTO Copenhagen 2015
• 30mNative Speed on the Web: JavaScript and asm.jsAlon Zakai · Strange Loop 2013