Bob Fehrenbach <PICLIST@MITVMA.MIT.EDU> wrote:

> I haven't tested this but adding one line and changing one branch
> destination would cause the carry to be cleared only if it were set.
> This might provide a small speed advantage.
>
>
> DIVBY3: BCF     STATUS, C           ;Omitted in first posting.
> NO_C:   RRF     TEST                ;New label
>         BNC     NODEC
>
>         SKPNZ
>         RETLW   0
>
>         DECF    TEST
>         BCF     STATUS, C          ;Added line
> NODEC:  SKPZ
>         GOTO    NO_C               ;Changed destination
>         RETLW   1

    Bob:

    Yes, this cuts about 2.5 cycles (on average) from the execution
    time.

    Good call.

> I am curious as to why anyone would want to know if a number were
> divisible by three, or is this an academic exercise?

    Bill Cornutt's answer ("It is the classic solution to the 'one
    potato, two potato' problem") would, of course, be the USUAL
    reason, but my need for the routine is a little different:

    The "divisible by 3" routine is only the first step of a much
    larger process:  I'm trying to see if the input number is EQUAL
    to 3.

    Here's how it works:

    First, I check for divisibility by 3.  If the number passes that
    test, I then make sure that it's not also divisible by 2.  After
    that, I use Eric Smith's routine for checking whether a number's
    within a certain range... I start with a range of [0-255], then
    progressively narrow the range, one step at a time, to [2-4].

    At this point, if the input number's passed all those tests, I'm
    pretty sure that it may be equal to 3.  To make ABSOLUTELY sure,
    though, I check bits 2-7 of the number individually, verifying
    that each is clear, then I check bits 0 and 1 and verify
    that they're both set.

    Since this is a very critical aplication, though, even THIS
    isn't enough.  To get that last little bit of certainty, I
    divide 4,294,967,296 by the input number.  If the answer's
    1,431,655,765, (with no remainder), I asume that the input
    number was indeed equal to 3.

    Finally, I double-check my assumption using the following code:

        XORLW   3
        BZ      ITS_EQUAL_TO_3

    -Andy

Andrew Warren - fastfwd@ix.netcom.com
Fast Forward Engineering, Vista, California
http://www.geocities.com/SiliconValley/2499