Pete Shadbolt 3d6684b7f9 | 8 years ago | |
---|---|---|
abp | 8 years ago | |
bin | 8 years ago | |
doc | 8 years ago | |
examples | 8 years ago | |
tests | 8 years ago | |
.bumpversion.cfg | 8 years ago | |
.gitignore | 8 years ago | |
MANIFEST.in | 8 years ago | |
README.md | 8 years ago | |
TODO.mkd | 8 years ago | |
makefile | 8 years ago | |
setup.py | 8 years ago |
Python port of Anders and Briegel’ s method for fast simulation of Clifford circuits. You can read the full documentation here.
It's easiest to install with pip
:
$ pip install --user abp
Or clone and install in develop
mode:
$ git clone https://github.com/peteshadbolt/abp.git
$ cd abp
$ python setup.py develop --user
$ python setup.py develop --user --prefix= # Might be required on OSX
abp
comes with a tool to visualize graph states in a web browser. It uses a client-server architecture.
First, run abpserver
in a terminal:
$ abpserver
Listening on port 5000 for clients..
Then browse to http://localhost:5001/
. Alternatively, abpserver -v
will automatically pop a browser window.
Now, in another terminal, use abp.fancy.GraphState
to run a Clifford circuit:
>>> from abp.fancy import GraphState
>>> g = GraphState(range(10))
>>> for i in range(10):
... g.act_hadamard(i)
...
>>> g.update()
>>> for i in range(9):
... g.act_cz(i, i+1)
...
>>> g.update()
And you should see a visualization of the state.
abp
has a bunch of tests. You can run them all with nose
:
$ nosetests
53 tests run in 39.5 seconds (53 tests passed)
Currently I use some reference implementations of chp
and graphsim
which you won't have installed, so some tests will be skipped. That's expected.