[izacus]

My experience is now a few years old, but WASM binary we had ran less than 10% slower than C++ Clang compiled native binary from the same source. And that was for rendering code which is rather CPU intensive and benefits from optimisation.

WASM is surprisingly fast.

[connicpu]

It basically adds a bitmask to every memory access, which sounds about right for a 10% slowdown.