Mind you, the Mill's split-stream instruction decoding means each decoder only has to deal with ~168 bytes and ~16 instructions, and there's a cache for both (I

Mind you, the Mill's split-stream instruction decoding means each decoder only has to deal with ~168 bytes and ~16 instructions, and there's a cache for both (I think?).