Proof-of-concept / requirement-setting code for a future project (not hard to guess what it is)
99% of the credit for this should go to Herbert Kociemba's solver algorithm and supporting python libraries. Without it, it would be a far-fetched dream to solve 70,000 Rubik's cubes to less than 24 moves each in 90 minutes on 2019 mid-tier laptop hardware while the user plays Disco Elysium at the same time or something.
Primarily an aid in selecting what goals the solver should have set (on specific compute hardware), to optimize for total (solve + move) time.
Install requirements with pip3 install -r requirements.txt
On first run, the underlying solver library needs to populate some quite large lookup tables. This may take half an hour or more.
When remoted into systems to run this characterization, it is reccomended to use the options -nr to save the output to a file for transfer with scp/sftp, instead of trying to get X11/whatever forwarding working.
example program output graph
Recommended to use with pypy to get (in my experience) about 30-40% faster solves. If doing so, you may need to install the package pypy3-tk.
usage: characterize.py [-h] [-n] [-f FULLCSV] [-s SUMMARYCSV] [--minDepth MINDEPTH] [--maxDepth MAXDEPTH] iterations timeSecondsPerMove
positional arguments:
iterations | number of cubes to solve per goal depth
timeSecondsPerMove | time of a 180deg turn in seconds, used for plotting time estimates (worst-case)
options:
-h, --help | shows help message and exits
-n, --noPlot | disables output graph
-r, --toFile | stores output graph to 'output.png' file in project root
-f FULLCSV, --fullCsv FULLCSV |
outputs a csv file of every simulated solve to a given filepath
-s SUMMARYCSV, --summaryCsv SUMMARYCSV |
outputs a summary csv (same data as displayed during iterative solving phase) to a given filepath
--minDepth MINDEPTH | minimum goal depth to sample, defaults to 19
--maxDepth MAXDEPTH | maximum goal depth to sample, defaults to 25