Software or microcode emulation works pretty well.
So it would be faster and more efficient when sticking to the new subset and Nx slower then using the emulation path.
So it would be faster and more efficient when sticking to the new subset and Nx slower then using the emulation path.
Most architectures other than x86 have fixed sized machine instructions now, making decoding fast and predictable.