Parallel Sudoku

This is a java implementation of a parallel sudoku solver (it computes every possible solution) based on the fork-join paradigm. It also includes a sequential backtracking implementation in order to make benchmarking and speed-up analysis possible.

The program takes text files representing a sudoku grid as input:

.....2...
..4...75.
73.9..6..
3.5..8...
.6.1...85
..92.....
.9..4.26.
2.....17.
...3.....

And outputs in the following order:

• The solution space
• The completion percentage of the input grid
• The number of possible solutions to the sudoku

Some benchmark graphs:

Usage

Compile the main class as follows:

javac Sudoku.java

Run any test by passing a text file in the specified format as parameter:

java Sudoku $(PATH_TO_FILE)

Or use the grader.sh bash script to test the algorithm on any file present in the tests folder:

./grader.sh

Features

• Fork-Join parallelization
• Sequential cutoff optimization
• Thread-halving optimization
• Bit masks optimization on candidates selection