The New Sudoku Players' Forum

[janho]

Thanks a lot!

I will give them a try as soon as I find time (probably not before next week) and I will let you know the results...

[AndrewLindsay]

Hello Group,

Mathemagics, out of interest, are you able to share the source code (powerbasic) for the solver that you benchmarked? I'm intrigued how it is written and would love to see the algorithms and how they're implemented.

best regards

Andrew

[Mathimagics]

.
Hi Andrew,

I'm happy to share that code, if you can send me a PM to give me an email address to send it to.

Cheers

[AndrewLindsay]

Excellent, thanks. All received, just looking through to understand it now.

[creint]

Do you have a testcase for easy kakuro's with a larger size like 50x50 or 100x100?
My solver is probably fast but has currently only one (limited) strategy for calculations.

[Mathimagics]

Do you have a testcase for easy kakuro's with a larger size like 50x50 or 100x100?

No, I haven't tried to generate anything larger than 36 x 24, since that is the limit for one physical page as far as I'm concerned.

Puzzle generation is so much harder than puzzle solving, I have found, but as I have recently re-worked my generator I will try and generate larger sizes.

Here is a 36 x 24 puzzle:

Hidden Text: Show

Code: Select all

982.679...521..87156...
421.135...952.7483956..
37..5879.87435162.8975.
.68...8269.7583246.8961
.412.83.8793.729586.893
..4827139.2163.795862..
48.63549718.71.5149837.
962413.54869.4613275.62
8967542..7456983.14..46
.513.97.15.87...47.798.
...93.35698741.42.6985.
51..62179.642318.238.97
876395.2315.8439.514923
957421.4537912..98..612
...5879...983..6412583.
8479..37214.92786431...
521..78954..5489761.698
968.945876...695874.241
713.5164328..14723..123
...17245639.23641..9387
.9258376..421...5186...
987..92..9536281.374561
814795.1957.7895.695784
62.968.378694.69825..95
.6985.12.63758421.29...
.47..83...23.97.21.347.
63..71.2731845..7542183
25.4935162.61234.836597
.718.245.15.37582614.69
..891467.8957.7943658..
976.279816.1879.89.169.
6124.6793512.6851...97.
.9587.84572319.9361..89
..4987362.342...597.918
...16423..578...978.857


[horybori]

I'm just curious if you could test some puzzles against my solver if you have some time. I would be interested in the results. universal-sudoku-solver-t32972.html

[Mathimagics]

I'm just curious if you could test some puzzles against my solver if you have some time.

I have provided many of the benchmark Kakuro puzzles (see last post on page 1 of this thread) so you can test them yourself.

For me to run the tests I'd need explicit instructions on how to submit these puzzles to your solver …

Cheers
MM

[Mathimagics]

.
Time for an update!

Thanks to creint, we now know that even the hardest puzzles in my benchmark set are quickly solved with SAT:

Code: Select all

 Puzzle             Rating   |    DFS    |   SAT
 ----------------------------+-----------+--------
 KP2424_Harder       4.235   |   4.392s  |  1.02s
 KP2424_Killer       4.267   |   3.844s  |  0.62s
 KP2424_Diabolical   4.406   |   4.392s  |  0.23s


Notes:

• DFS refers to the fastest time achieved by my solver (CM-B*)

• SAT times are for MiniSAT (this is much faster than Sat4J)

• times include uniqueness testing