On Thursday, Apr 20th 2006 at 19:51 +0300, quoth Ilmari Karonen:

=>Robert C. Helling wrote:
=>> On Thu, 20 Apr 2006, Ilmari Karonen wrote:
=>> > Ilmari Karonen wrote:
=>> > Incidentally, here's the hardcoded-givens version.  It's indeed quite a
=>> > lot faster the generic version -- I've yet to find a puzzle that would
=>> > take more than 0.1 seconds to solve on my workstation.
=>> 
=>> Try
=>> 
=>> ....3....
=>> .15...6..
=>> 6..2..34.
=>> ...6...8.
=>> .39...5..
=>> 5.....9.2
=>> .........
=>> ...97.25.
=>> 1...5..7.
=>
=>0.3 seconds.
=>
=>> or
=>> 
=>> .98......
=>> ....7....
=>> ....15...
=>> 1........
=>> ...2....9
=>> ...9.6.82
=>> .......3.
=>> 5.1......
=>> ...4...2.
=>
=>5 seconds.
=>
=>But flipping the last one by 180 degrees makes it _really_ tough on the
=>regexp, which always solves the puzzles in left-right, top-down order. Solving
=>it that way around takes well over 3 _minutes_.

Interesting. I would expect that rotating a sudoku matrix would not have 
any impact in the difficulty of the problem. I remember from my old 
numerical analysis course (back before you young whippersnappers were 
born) we looked at something called the Hubbard or Hibbard constant for a 
matrix. The deal was that if you multiplied a matrix by its inverse, then 
there was something you would come up with that would tell you how hard it 
was to invert the matrix. Of course you needed the inverse to calculate it 
in the first place...

-- 
steveo at syslang dot net TMMP1 http://frambors.syslang.net/
Do you have neighbors who are not frambors?

Reply via email to