On Fri, 7 May 1999, Scott Dattalo wrote:

> Here's another solution that gives two extra cycles for other stuff
> (like CLRWDT or testing an I/O bit)
>
>     clrf   lo
>     clrf   hi
>
> loop1
>
>     nop             ;2 free cycles executed 255 out of 256
>     nop             ;iterations
>
> loop2
>
>     decfsz  lo,f   ;or incfsz
>      goto   loop1
>     decfsz  hi,f   ;if low byte rolls over, dec hi byte,
>      goto   loop2  ;but don't branch to the beginning.

After a little churning, I found a 3 cycle solution:


    clrf    lo
    clrf    hi

loop

    decfsz  lo,f
     goto   loop

   ;lo is zero right now.

    decf    hi,f
    skpnz
     goto   done      ;or perhaps return

   ;okay, we've gotta lot of choices here
   ;
   ;we could do this:
   ;  incf  lo,f
   ;  nop           ;or clrwdt
   ;  incf  lo,f
   ;  goto  loop
   ;
   ;But I like this:

    bsf   lo,1      ;Equivalent to 2 incf's

    nop             ;Two free cycles
    nop             ;everytime lo rolls over

    goto  loop


   ;Or if you wanted more free cycles:
   ;   bsf   lo,2   ;Equivalent to 4 incf's
   ;   goto  $+1    ;Eight free cycles
   ;   goto  $+1    ;everytime lo rolls
   ;   goto  $+1    ;over
   ;   goto  $+1
   ;
   ;   goto  loop
   ;
   ;Of course, you could preset the lo byte
   ;to any value and then have the remaining
   ;cycles evailable for other stuff.
   ;

Like everything I comment, this untested :).

The time at which the lo byte is incremented does not happen exactly every
third cycle after it rolls over. In fact, some values of lo are skipped
altogether. However, the total number of cycles between the beginning of
the loop and the 'goto done' instruction is 2^16 * 3.

This kind of loop would be useful if you wanted to have a fixed delay AND
still be able to do other things too.

Scott