Fast Parallel Sudoku Solver
This is a fast parallel Sudoku Solver implemented using OpenMP. As solving Sudoku is a known NP-complete problem, I have used variety of heuristics to make it fast. A detailed discussion of the heuristics I have used may be found in these places:
The DesignDoc.pdf describes the algrithm I have used, the data/task distribution among the threads and the various optimisations I did to speed up the Solver.
Compilation
The code can be compiled as follows. The grid size of the sudoku which the program will be used to solved is set at compile time. minigridsize is the size of the smaller squares in the sudoku grid. So that for a 16x16 board, minigridsize=3.
$ ./make.sh <minigridsize>Default minigridsize is 3.
Running
The program can be run by passing the number of threads, thread_count, the location of the text file containing the sudoku grid.
$ ./sudoku <thread_count> <grid_file_path>Tests
A number of test grids of various sizes, and difficulty levels are present in the TestCases directory. To solve a 16x16 Sudoku grid using 2 threads
$ ./make.sh 4
$ ./sudoku 2 TestCases/16x16_board/sudoku16_input1