At 10:50 AM 29/10/97 -0500, you wrote:
>There hasn't been a "programming challenge" lately.  Here's a problem I'm
>looking at:  I have a 7 bit ASCII character which was transmitted
>serially LSB first, but received MSB first.  Thus, I need to reverse the
>order of the bits (I can't change the receiver, since other parts of the
>data are transmitted MSB first).  For example, 0011010 should come out as
>0101100.  The routine should be a subroutine which gets the input value
>in W and returns the result in W.  It is OK to assume the high bit of W
>will be 0, and the high bit of the result must be 0 (the routine works
>only on the low 7 bits)
>
>The obvious answer is a lookup table of 128 entries, but this is a big
>portion of the code space in an '84.  Speed is not important, so I'm
>going to replace the lookup table with an "analytical" method.  Here's a
>completely arbitrary scoring scheme which favors using less ROM and RAM
>(low score is better):
>
>Each ROM location used : 8 points
>Each RAM location used : 30 points
>Each instruction cycle (to convert once): 1 point
>Uses less than 25 instruction cycles: -40 points

a real dirty solution:

revbits
        movwf tmp
        andlw 0x08
        btfsc tmp,0
        iorlw 0x40
        btfsc tmp,1
        iorlw 0x20
        btfsc tmp,2
        iorlw 0x10
        btfsc tmp,4
        iorlw 0x04
        btfsc tmp,5
        iorlw 0x02
        btfsc tmp,6
        iorlw 0x01
        return

RAM 1 * 30   =  30
ROM 15 * 8   = 120
Cycle 16 * 1 =  16
< 25           -40
==================
SCORE:         126

should be Correct now :)

antti



http://avrbasic.com         -- AVR Basic Compiler
http://sistudio.com/bswfe   -- Basic Stamp Windows Front End