They mention the improvement of 128x by a new algorithm. Can someone explain how they changed the algorithm?
Thanks!

You can find the concerning paper here.

Unfortunately, it's not freely accessible.

Well could someone with IEEE access paste the part of the algorithm or put it on rapidshare or something? If we're participating in this project, its only fair we know what we're doing.
Thanks!

Gary McGuire has proven that there is no 3x3 Sudoku with 16 clues with a unique solution!

What's even better is that you can find the program and the article describing the improvements on his original "checker" algorithm here.

So why is this project still running then if the problem has been resolved?

It can't hurt to have a different source to confirm this result so this project still has some reason to continue.

Yes, McGuire recently claimed to solve the minimum Sudoku problem: No 16-clue puzzles exists.
Regardless of this claim, we still decide to continue the BOINC project.
As Afmob said, it does not hurt to finish the project, using a different version, to confirm the result.

I fully agree with you..

Nonetheless, I agree with Josh. The improvement should be posted online, e.g. the article should be freely accessible. Gary McGuire did this and he also made his programm available to everyone.

Sorry we thought it is avaliable.
The paper can be download here:
http://sudoku.nctu.edu.tw/sudoku.pdf

Thank you!

I used your paper and McGuire's paper to make my own version of the checker and I would like to know how fast it is in comparison.

To estimate the improvement you used 100 random primitive grids. Can you please provide me those 100 grids?

The 100 grids can be downloaded here:
http://sudoku.nctu.edu.tw/benchmark.txt

Once again, thank you!

how fast can your program do, afmob?

Sorry for the late response.

My program needs about 3.737 seconds to analyse all 100 grids of the benchmark. I think one reason it's slower is the fact that I've written the code in C++ without using direct assembler commands.