The WASM build uses the 'normal' CPU core. In this core the CPU emulator is a collection of indirect calls that performs poorly in WASM due the runtime checks to verify the number of parameters and their types.
For other builds there is a dynamic code generator. A similar approach would work wonders. Assistance gratefully accepted! - see source folders under emulation/cpu for armv7, armv8 & x32 for the 30 odd method implementations.
To get native like performance a binary translator like that used for the x64 core is required