Or check the repository on GitHub with the benchmarks's source code.
Types added manually, C++ generated by a custom TypeScript AST visitor, heavy lifting done by Cheerp.
Benchmarking performed client side, thanks to Benchmark.js. Higher ops/sec is better.
The recap bar is filled proportionally to the probability that the generated code is more performant than the original, color coded to 🔴 = "likely worse", 🟡 = "hard to say", 🟢 = "likely better"