I'd like to see that z80 library. The 8088 has more registers than the z80, but BASICA FP (at 4.77 MHz) is only very slightly faster than AppleSoft FP (at 1 MH

I'd like to see that z80 library. The 8088 has more registers than the z80, but BASICA FP (at 4.77 MHz) is only very slightly faster than AppleSoft FP (at 1 MHz). My experience is that 16 bit operations on the z80 save a lot of code size but very little time compared to doing multiple 8 bit operations. Adding a register pair to HL takes 11 cycles (15 to add to IX/IY), vs 4 cycles to add an 8 bit register to A. ADC to HL takes 15 cycles. Each byte copied from one register to another takes 4 cycles (7 for IX/IY H/L). So if you want to ADC BC,DE you're looking at 64 = 24 cycles for adding in A vs 27 cycles copying BC to HL, adding DE, and copying back. The code size is 6 bytes vs 5 bytes. On 6502 the same thing, using Zero Page, is 20 cycles and 13 bytes of code. So 6502 wins on cycles, but 6502 machines were 1-2 MHz and z80 3-4 MHz, so z80 wins on high clock speed. BUT ... how the heck do you keep two 32 bit floats, plus* some just as large temporary variables, in the 11 bytes of registers on a z80? You can get a bit more using EX and EXX but it's going to be incredibly fiddly -- and you remove any possibility of fast interrupt handling on the same machine. As soon as the z80 needs to use RAM it loses. Loading or storing a register pair to/from a fixed RAM location is 20 cycles. Indirecting via (IX/IY+n) is 19 cycles per byte. Indirecting via (HL) then INC/DEC HL is 13 cycles per byte. Really, I'd like to see that 8080/z80 FP library.