Hi! This is Sayuri Hasegawa. I hope enjoy reading this submission. Basically
you use the submission like this:
mv ... data/input.txt
make dump # produces data/input*.txt and others
make solve # produces analysis/ and output/
make verify # checks if the output in output/ is correct.
# often patches a last-minute error, and
# stores the solutions to archive/
make zip # produces result.7z and code.tar.gz
# from the best solutions in archive/
You require bash, GNU make, p7zip, Python (2 only, PyPy is just fine), any
optimizing C++ compiler. I've also used a modified version of Golly for fast
board generation: `grep 'Hasegawa Sayuri' golly/` for all modifications.
You can also use `./solve.sh N` and `./verify.sh N` for individually solving
and verifying single test case. See `solve.sh` for strategies used.
The entire process took about 13 minutes in MacBook Air (Mid 2011) with
4 GiB RAM. This does not include the time spent for tweaking various
parameters and algorithms. YMMV.