On Fri, 11 Feb 2000, James Newton wrote:

> 4. I'm worried that the comments I've added (of the form W= and A2=) on each
> line of John Payson's 16 bit Binary to BCD conversion code may be incorrect.
> Specifically: Am I getting the subtraction by addition idea correct? How can
> comf do a negate when it does a ones complement vice a twos complement? Math
> is not my strongest point. see:
> http://techref.massmind.org/microchip/math/radix/b2bu-16b5d.htm

Hi James,

after you sent me the (private) message the other day, I did spend a few
minutes running through the comments.

        movf    NumH,w  ;w  = A3*16+A2
        andlw   $0F     ;w  = A2
        addwf   Hund,f  ;B2 = A3-46 + A2 = A3+A2-46
>>>     addwf   Hund,f  ;B2 = 2*(A3+A2-46) = 2A3+2A2-92
        addwf   Ones,f  ;B0 = A3+4 + A2 = A3+A2+4

should be:
        addwf   Hund,f  ;B2 = 2*A2+A3-46


And

>>>     rlf     Tens,f  ;B1 = 2*(4A2+A1-92) = 8A2+2A1-184
>>>     rlf     Ones,f  ;B0 = 2*(A3+A2+A1+4) = 2A3+2A2+2A1+8
        comf    Ones,f  ;B0 = -2A3-2A2-2A1-8
        rlf     Ones,f  ;B0 = 2*(-2A3-2A2-2A1-8) = -4A3-4A2-4A1-16

 -92 << 1  does not equal -184

-92 = 0xa4
 and 0xa4 << 1 is 0x48

Then the next line is going to pick up the carry. So these two lines
should read instead:

>>>     rlf     Tens,f  ;B1 = 2*(4A2+A1+0xa4) = 8A2+2A1+0x48
>>>     rlf     Ones,f  ;B0 = 2*(A3+A2+A1+4)+1 = 2A3+2A2+2A1+9

These are the only two errors that immediately popped out. However, since
the results of these calculations are used further down in the code, the
errors are propogated. There may be a bug silently lurking in this code -
I hadn't noticed the C bit thing before...


Scott