My solution seemed to be a little bit too slow, so I’ve tried to squeeze it. Bitsets were used only to store the matrix, so I’ve tried to check if asking about an edge will be faster on normal array. Finally, the only differences between the first and the last submission are some optimizations in the top of the code. I didn’t care about incorrect attempts, as the wouldn’t change my position.

My debug does nothing without an exact compilation flag. You might remember it from a puzzle for viewers from the last ICPC finals.

orz gennady