Just for fun... I think this should be on the efficient frontier of 8 x 8 -> 16 bit multipliers, at an average cost of around 43 cycles and memory of 1330 byt

Just for fun... I think this should be on the efficient frontier of 8 x 8 -> 16 bit multipliers, at an average cost of around 43 cycles and memory of 1330 bytes or so. Compare 45.5 cycles and 1580 bytes for the mult66.a fastest in the collection linked. lmul0 = $06 ; pointer into square table low lmul1 = $08 ; pointer into square table high labs0 = $0a ; pointer into table of abs(i-1) prod_low = $0c \* = $0200 ; Align tables to start of page abstable !for i, 0, 255 { !byte <(abs(i-1)) } squaretable1_lsb !for i, 0, 511 { !byte <((i*i)/4) } squaretable1_msb !for i, 0, 511 { !byte >((i*i)/4) } ; multiply ; ; f(x) = x*x/4 ; ; return f(X+Y) - f(|Y-X|) ; ; 8 bit x 8bit unsigned multiply, 16 bit result ; ; On Entry: ; A: multiplier ; Y: multiplicand ; On Exit: ; (prod_low, A): product mult sta lmul0 sta lmul1 eor #ff sta labs0 ldx (labs0),Y ;X = abs(Y-A) lda (lmul0),Y sec sbc squaretable1_lsb,X sta prod_low lda (lmul1),Y sbc squaretable1_msb,X rts ;call this once to initialise high bytes of pointers to table mult_init lda #>abstable sta labs0+1 lda #>squaretable1_lsb ; high byte (#2 in this instance) sta lmul0+1 lda #>squaretable1_msb ; high byte (#4 in this instance) sta lmul1+1 rts