Integer square root for ATmega328p

[jabcam]

(This is also a test to see if I got the formatting correct:)

Here comes a sqrt-routine, written in AVR-assembler for speed, takes some 90 cycles to run, i.e. 5.6 µs on Arduino@16 MHz.

\ square root of unsigned cell. 
: sqrt ( u --- u )
    [ R16 $00 ldi,         ]
    [ R17 $80 ldi,         ]
    [ begin,               ]
    [     R16 R17   eor,   ]
    [     R16 R16   mul,   ]
    [     R0  R24   cp,    ]
    [     R1  R25   cpc,   ]
    [     lo, not, if,     ]    
    [         R16 R17 eor, ]
    [     then,            ]
    [     R17 lsr,         ]
    [ eq, until,           ]
    [ R24 R16 movw,        ]
;

[Mikael Nordman]

4 sqrt gives one as a result. It should be two. I added one to the input number to get correct results for 2 4 9 16 etc.

Here is the code written with the new tabledriven assembler which I published today.

\ square root of unsigned cell.
: sqrt ( u --- u )
  as: adiw r24 1
  as: ldi r16 $00
  as: ldi r17 $80
  as: begin
  as:   eor r16 r17
  as:   mul r16 r16
  as:   cp  r0  r24
  as:   cpc r1  r25
  as:   if  sh
  as:     eor r16 r17
  as:   then
  as:   lsr r17
  as: until eq
  as: movw  r24 r16
;