Benchmarks are way too wrong. They were using a Javascript library called BarcodeReader and switched to their native that was already written in C++ using webassemly. How can you be 100% surr that BarcodeReader library doesn't step down and messing up your results? It's like saying "I'm using Angular for displaying hello messages to users and switched to native C++ using webassemly. Now my app is 100 times faster".