I got some news about the benchmarks by Hadrien Cambazard.

Last year, there have been a competition about CSP solvers.
More information here:
http://www.cril.univ-artois.fr/CPAI08/

The program Abscon 112v3 was able to solve all the problems except problem 6
(and problem 1, since it's not solvable) !
Here are the benchmarks with the 'words' list:

| abscon112v3 |
=============================================================
instance | NODES TIME RES |
=============================================================
crossword-m1-words-23-02 | 935.00 9.31 SAT |
crossword-m1-words-23-03 | 61251.00 118.81 SAT |
crossword-m1-words-23-04 | 16638.00 64.98 SAT |
crossword-m1-words-23-05 | 1745.00 11.01 SAT |
crossword-m1-words-23-06 | - |
crossword-m1-words-23-07 | 2396.00 12.35 SAT |
crossword-m1-words-23-08 | 11063.00 24.97 SAT |
crossword-m1-words-23-09 | 102622.00 292.17 SAT |
crossword-m1-words-23-10 | 6731285.00 8244.00 UNSAT |
=============================================================

+-------------------------------+
| SUMMARY SELF - 9 instances |
+-------------------------------+
| #solved | #errors | TIME | NODES |
========================================================================
abscon112v3 | 8 / 9 88% | 0 | 1097.20 | 865991.87 |
========================================================================


Problem 6 is unsolved after 10 hours of computation.

Now, I feel pitiful with my program :-(

JC





>> Here is the tenth grid:
>> ______*______*___*_____
>> ______*______*_________
>> _____________*_________
>> ___*________*______*___
>> _____*_____*_____*_____
>> ____*_____*_______*___*
>> **____*_________*______
>> _______*_______*_______
>> ________*_____*________
>> _________*___*______***
>> _____*______*______*___
>> ____*______*______*____
>> ___*______*______*_____
>> ***______*___*_________
>> ________*_____*________
>> _______*_______*_______
>> ______*_________*____**
>> *___*_______*_____*____
>> _____*_____*_____*_____
>> ___*______*________*___
>> _________*_____________
>> _________*______*______
>> _____*___*______*______
>>
>
>
>


[Non-text portions of this message have been removed]