R800 MULUB

There’s a mistake in the r table on the MAP, (hl) is not supposed to be included there. I list (hl) separately everywhere because the timing is different. I’ll change it. Currently it incorrectly claims ld (hl),(hl) exists too which it doesn’t (at least officially).

That said, the rule you mention for the bitfield encoding is still true, although some instructions like IN and OUT (and maybe also MULUB) do not accept (hl). I don’t know if MULUB A,(HL) exists, or whether it works if it does. If I had to place a bet on it, I’d say no, but it needs to be tried .

The opcode for LD (HL),(HL) would be 76, which is the opcode for HALT. The DD (ix) and FD (iy) prefixed versions probably also execute a HALT, or else are a no-op ("The Undocumented Z80 Documented" does not list it as HALT so maybe the latter).

The n, o and nn are fairly standard notations, also used by Zilog in their documentation and elsewhere, so I won’t be changing them. If anything, I do write "nn" and "oo" in the opcode column, probably I should change them to be single-letter as well. (Edit: changed it now.)

I see your point about endianness but I don’t like deviating from nn or use anything that resembles “hl”, so the reader is just going to have to be aware that the Z80 is little-endian I guess. I’ll add a note to the nn entry in the legend at the bottom.

The biggest pity is that EX DE,IX doesn't work. It would go in 10 cycles. push ix : pop de is 33 cycles.

About EXX is clear it cant exchange registers that fast, it got to be flipping one bit. One bit of the register addressing sheme.
Now I wonder whether the good old EX DE,HL of 8080 too works with such bittflip in register addressing! That EX DE,IX does not work is an indication.

Mhm and now I feel like the EX DE,HL always been a bit like an IX prefix for DE. Except that one always has to flip back with another EX DE,HL afterwards.

Would’ve been nice, but don’t forget plain old ld:

    ld e,ixl
    ld d,ixh

20 cycles.