This may seem rather obvious to some of you, but the fastest way I know
to compute parity is to XORWF Parity,F inline with my serial output
routine (assuming a firmware bit-bang engine).  Since the data to be
output is being shifted to the right in my code loop, Bit 0 of Parity
contains the desired result of all N bits when all data bits have been
shifted out (generates odd parity).  I then shift Bit 0 of Parity out as
the final bit before the stop bit.  Since the XORWF is inline, it takes
only 1 cycle  to compute parity per bit.  Ya can't do it any faster than
that.

Sorry, but the code is part of a commercial product - and I don't have
authorization to disclose the actual serial bit-bang routine here.  But
jeez, you get the gist of it: the concept couldn't be simpler.
(Ceeeeripes Man!  It's 1 lousy instruction in your output loop! ;-)

Cheers,

Chuck

> -----Original Message-----
> From: Scott Dattalo [SMTP:sdattalo@UNIX.SRI.COM]
> Sent: Friday, March 13, 1998 5:27 PM
> To:   PICLIST@MITVMA.MIT.EDU
> Subject:      Parity Challenge
>
> It's been a while since there's been a challenge. And since there's
> been a bunch talk about RS232 stuff, I thought it might be appropriate
> if we had a really fast parity generator for those software
> bit-banging
> UARTs.
>
> So here's the challenge:
> Given the variable X, write a routine that computes the odd-parity
> of X.
>
> Here's a routine that's slow/looped:
>
>         CLRF    parity
> LOOP:
>         CLRC
>         RRF     X,F             ;Get LSB and put it in the carry
>         SKPNC                   ;If LSB (i.e. C) is set
>           INCF  parity,F        ;then count it
>         MOVF    X,F
>         SKPZ
>           GOTO    LOOP
>
> The least significant bit of 'parity' contains the parity of X.
>
> The best I could do executes in 7 cycles and trashes X.
>
> Scott