I'm not sure if there is a comprehensive timing summay but here are some different approaches and code you can compile and run on your machine: http://haskell.org/haskellwiki/Sudoku In fact, if you run them all, you can share your numbers with us and we can add them to the Sudoku page. And of course you can check the archives for any threads about it (assuming you can get to each person's code somehow). Jared. On 7/9/06, Murray Gross <mgross21@verizon.net> wrote:

I'd greatly appreciate it if someone could provide a summary of the best times recently posted for Haskell Sudoku solvers so I can compare them with some experimental code I have.

Since the Sudoku puzzle is incidental to the purpose of my code, I'd just like to find out whether or not my current code is working in the neighborhood of "the good stuff." Right now, easy puzzles go under in 4 or 5 seconds, evil puzzles range from about 12 seconds to just under 2 minutes on a 650 MHz Duron.

Thanks,

Murray Gross

_______________________________________________ Haskell mailing list Haskell@haskell.org http://www.haskell.org/mailman/listinfo/haskell

-- http://www.updike.org/~jared/ reverse ")-:"

On Sun, 9 Jul 2006, Chris Kuklewicz wrote:

I improved the speed and strength of my solver and it finishes the 36638 puzzles of http://www.csse.uwa.edu.au/~gordon/sudoku17 in 3885 seconds. It can solve all but 164 without guessing, and averages greater 9 puzzles per second on a 1.33 GHz PPC Mac OS X laptop using ghc-6.4.2

Most of those puzzles (about 91.5%) are particularly simple and go especially quickly. The next 8.1% of the puzzles use more expensive deductive methods, and the remaining 0.4% do depth first guessing.

Thanks, this is just the kind of number I was looking for. Given the differences in hardware and software, your code appears to be about 10 times as fast as what I am running now. Given that I have done no code tweaking, this tells me that my algorithms do quite nicely, even if the code is not setting any records, which puts my code right in the area that I was looking for. Once again, thanks. Murray Gross

Henrik Nilsson wrote:

Hi Dominic,

...

Does anyone know where the source code for Yampa Arcade can be found?

It's included with the (now very old) afrp-0.4 distribution available from

www.haskell.org/yampa

It's in the examples/SpaceInvaders directory.

Hope that helps!

All the best,

/Henrik

Henrik, Thanks. I tried ghc --make Main.hs on the code from http://www.cs.hmc.edu/courses/2006/spring/cs182-3/afrp but got Compiling ObjectBehavior ( ./ObjectBehavior.hs, ./ObjectBehavior.o ) ./ObjectBehavior.hs:37:56: parse error on input `->' The code in examples/SpaceInvaders for ObjectBehavior.as is identical so I assume that won't compile either. Why .as? If anyone has got this working please let me know. I tried making www.haskell.org/yampa but got [dom@tility afrp-0.4]$ make make -C src make[1]: Entering directory `/home/dom/yampa/nilsson/afrp-0.4/src' ghc -c -fglasgow-exts -O -package lang -package concurrent -o AFRPUtilities.o AFRPUtilities.hs AFRPUtilities.hs:169:46: Ambiguous occurrence `>>^' It could refer to either `AFRPUtilities.>>^', defined at AFRPUtilities.hs:106:2 or `Control.Arrow.>>^', imported from AFRP at AFRPUtilities.hs:90:0-10 So it looks like the code has rotted a bit. I'm assuming the code from hmc is in better shape (as it seems to be in a Spring '06 directory and the compilation got a few modules further) and may have a go at seeing if I can get it to compile. Dominic.