AVR-Chat

On Thu, 09 Jun 2005 11:21:50 -0500 you wrote:

>
>>
>>It appears to require 6 shifts, six ANDs, and 5 ORs... seems unlikely
>>it would be faster than the "straightforward" implementation?  It looks
>>like a case of someone trying to be clever after they learned why a
>>quicksort is better than something like a bubble sort (what he does is
>>swap bits 0/1, 2/3, 4/5, 6/7... then 0-1/2-3, 4-5/6-7, and finally 0-
>>3/4-7).  The problem is that with only 8 bits there's not any speedup
>>yet that I can see -- just as smart general purpose sorting routines
>>will revert to something like a bubble sort of lists of less than, say,
>>10 items!
>
>Unless someone pops up from left field here with a new solution, I'm 
>going to say that we have a couple of nice solutions now, and call 
>this one "done".
>
>The table lookup is, somewhat surprisingly efficient, other than on 
>code space, especially when the table can be located on an even page.
>Interestingly enough, I think it executes in the equivalent of one 
>instruction on the PIC.  VBG!
>
>The external loopback is fun, if you have the pins available.
>
>That's the point of challenges like this, to encourage "out of the 
>box" thinking and explore ideas.
>
>
>
>So, anyone want to propose a new little problem?

This one is known but can prove a puzzler if one doesn't know it.

How do you swap the contents of 2 registers without using any other registers or
stack?
--
Ralph Hilton
http://www.ralphhilton.org
C-Meter: http://www.cmeter.org
FZAOINT http://www.fzaoint.net