Cult Of The Bound Variable
http://www.boundvariable.org/task.shtml
Universal Machine development
Python
Using pypy to speedup execution:
pypy3 ./um.py ./umz/sandmark.umz | tee ./umz/sandmark-myoutput.txt
Codex processing
Running the UM on codex.umz proposed to "dump UM data"
um.py umz/codex.umz
self-check succeeded!
enter decryption key:
(\b.bb)(\v.vv)06FHPVboundvarHRAk
decrypting...
ok
LOADING: 9876543210
== CBV ARCHIVE ==
VOLUME ID 9
Choose a command:
p) dump UM data
x) exit
?
Trying to save dump with:
cat umz/codex.ins | pypy3 um.py umz/codex.umz - > umz/dump.out
It took me a while to understand why this was not working! I'm using Python 3, so sys.stdout.write(chr(self.reg[c])) or print(chr(self.reg[c]),end="") are not outputting bytes to stout as in Python 2, but something that depends on the terminal encoding. I can probably leave the terminal output as it is, but add an option to dump the output as bytes to a file:
cat umz/codex.ins | pypy3 um.py umz/codex.umz -d
And it works!
UMIX
Let the fun begin...
pypy3 um.py umz/dump.um
12:00:00 1/1/19100
Welcome to Universal Machine IX (UMIX).
This machine is a shared resource. Please do not log
in to multiple simultaneous UMIX servers. No game playing
is allowed.
Please log in (use 'guest' for visitor access).
;login:
C++
After some digging of the UMIX system, I realised my python implementation is too slow to allow efficient attempts to solve of the various puzzles (for instance, the adventure game takes forever to load, it is thus impossible to automatize the searach for a correct sequence of actions). Therefore I am now trying to develop a um machine in C++.
To compile:
make
To run the bentchmark tests:
./um ./umz/sandmark.umz
Benchmark
Laptop: MacBook Pro 2.3 GHz Quad-Core i7
| Implementation | time to execute sandmark.umz on my laptop |
|---|---|
| Python (PyPy) | 5412.57s user 47.87s system 98% cpu 1:32:34.61 total |
| C++ | 2713.64s user 16.22s system 99% cpu 45:56.45 total |
Saving UM state
The C++ UM implementation is indeed faster than the Python one, but one thing that could help (in both cases!) would probably be to save the UM status and start from there to cut on loading and initialization time. I will try to implement this in python since pickle makes things easy.
um.py <program.umx> -i <input_status.sav> -o <output_status.sav>