> Finally, while running the fools-gold preset, I encountered a Microsoft
> Freecell deal that was intractable with Atomic moves in Freecell Solver and
> ended up consuming over 70% of my machine's RAM.
>
> The deal is No. 30857:
>
> {{{{{{{{{
> 2C AS 3H TC 7D 8H QD
> AC 2S 7S 7C KD QH TS
> 8S 9C 3C 2H 6D 4S JC
> KS KC 5C 8C QS 9H QC
> 4H 4D KH 6C 3S JS
> 5D 2D 9D 6S AD JD
> AH TD 8D JH 4C 9S
> 6H 3D TH 5H 7H 5S
> }}}}}}}}}
>
> Here is what my investigation yielded:
>
> 1. "./fc-solve 30857.board" solves it in 1,431 iterations. (It's a meta-moves
> preset)
>
> 2. "./fc-solve -l gi 30857.board" solves it in 944 iterations. A meta-moves
> preset too.
>
> 3. "./fc-solve -l fg --freecells-num 3 30857.board" (an atomic moves preset
> with the Freecells Number restricted to 3 freecells) solves it in 1,444,370
> iterations.
>
> 4. Freecell Solver version 2.8.x also runs into a problem with it, so it's not
> a recent regression.
>
> 5. "./fc-solve -to ABCDE01 30857.board" (atomic moves and not a meta-scan)
> solves it in 1,230,350 moves. "--freecells-num 3" seems to make it worse.
>
> 6. The game is reported to be unsolvable with 2 freecells.
>
> 7. I played the game by hand and it wasn't too difficult.
>
> 8. Tom Holroyd's Patsolve solves it after:
>
> {{{
> 6.72user 0.03system 0:06.83elapsed 98%CPU (0avgtext+0avgdata 0maxresident)k
> 0inputs+8outputs (0major+6965minor)pagefaults 0swaps
> }}}
>
> with the default preset. With -S it is solved after:
>
> {{{
> 0.08user 0.00system 0:00.09elapsed 85%CPU (0avgtext+0avgdata 0maxresident)k
> 0inputs+8outputs (0major+439minor)pagefaults 0swaps
> }}}
>
> I assume FCELL.COM does not have a problem with it either.
>
Just to see, I keyed 30857 into FFA and triggered FCELL.COM to solve it.
It came back in the blink of an eye with a 55-step solution.
-Gary Campbell
Received on Sat Jun 06 2009 - 11:03:38 IDT