On 7/15/05, Mark Rages <markrages@gmail.com> wrote:
> I'll post my solution in a minute, once I'm satisfied it works.

Well, I think it's working and it's quitting time, so here goes:

I started by listing all the possibilities:

A > B > C choose B
A > C > B choose C
B > A > C choose A
B > C > A choose C
C > A > B choose A 
C > B > A choose B

I notice that there are really only three comparisons, if we ingnore 
equality.  (After all, if A=B, it doesn't matter which I choose.)

So I made a mapping from the comparison result to the choice:

A>B | B>C | C>A | Choose
----+-----+-----+--------
 1  |  1  |  0  | B
 1  |  0  |  0  | C
 0  |  1  |  0  | A
 0  |  1  |  1  | C
 1  |  0  |  1  | A
 0  |  0  |  1  | B

Or, in decimal and numerical order:
comparison  | choose
------------+--------
   0        | X 
   1        | B
   2        | A 
   3        | C
   4        | C
   5        | A
   6        | B
   7        | X

But there's an interesting symmetry here: 3..0 is the same as 4..7.  This will
simplify my selection code.

Here's my code.  I used one temp register for the comparison result.

gteq macro fileA,fileB
	movfw fileB		; w=B
	subwf fileA,w		; w=A-B
	;; if B>A we borrowed, so _borrow=0
	;;  ... _borrow=1 means A>B or A==B

        ;; Tack the result on the right of the comparison register.
	rlf iris_adc_temp,f	; '1'-> A>=B
	endm

abc_median macro A,B,C
	local done
	
	clrf comparison_temp

        ; do comparisons
	gteq A,B
	gteq B,C
	gteq C,A
	
	;; fold around 4
	movfw comparison_temp
	btfsc comparison_temp,2	; is >=4?
	sublw 0x07		; yes. 4->3,5->2,6->1,7->0

        skpnz                   ; handle special case A=B=C
	movlw 0x01		; in that case just choose any!

	movwf comparison_temp	
	
        ; Now select from A,B,C like this:
	;; 1 -> B
	;; 2 -> A
	;; 3 -> C

	movfw A
	btfss comparison_temp,0	; this bit clear: comparison_temp=2
	goto done

 	movfw B
	btfss comparison_temp,1	; this bit clear: comparison_temp=1
	goto done

 	movfw C
done
	endm
-- 
You think that it is a secret, but it never has been one.
  - fortune cookie

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist