Clifford+T estimator - estimate a single probability of an n qubit quantum circuit consisting of Clifford gates, arbitrary single-qubit diagonal gates gates and a w qubit computational basis measurement.
This is an implementation of the algorithms reported in https://arxiv.org/abs/2101.12223
You will require
- python3
- numpy
- (optional) qiskit - required if you want to test the probabilities our code computes against qiskit predictions
- GCC (or equivalent C compiler - see notes below)
The setup.py file will compule and link the C code for you. A sufficient setup is (when executed from the directory containing the setup.py file):
python3 ./setup.py build_ext --inplace
This setup will compile the code and build the extension module in the current directory.
The code has only been tested with the following versions
- Linux 4.19.0-13-amd64 SMP Debian 4.19.160-2 (2020-11-28) x86_64
- gcc (Debian 8.3.0-6) 8.3.0
- Python 3.7.3
- numpy version 1.18.5
We make use of some gcc extensions, so other C compilers are not recommended.
In the estimate algorithm code we store CH-forms as binary matrices where a row of a matrix is a single C object of type uint_bitarray_t. If you require fewer than 64 qubits in the estimate code then you should leave the typedef
typedef uint_fast64_t uint_bitarray_t;
if you require between 64 and 127 qubits in the estimate code then you should replace this with the typedef
typedef unsigned __int128 uint_bitarray_t;
Doing so will make the code substantially (roughly 4 times, based on our testing) slower. Note that the number of qubits used in the estimate code is upper bounded by the number of non-Clifford gates in your circuit (and may be substantially less in some cases).
We provide implementations of the following algorithms, as defined in https://arxiv.org/abs/2101.12223
- Compress - which is insensitive to the difference between different single-qubit phase gates
- Compute - an implementation suitable only for pi/4 T gates
- Estimate - an implementation suitable only for pi/4 T gates
- Estimate - an implementation suitable for arbitrary phase gates
We also provide some routines to make using these more convenient. The script small_example.py contains examples for basic usage of the Compress, Compute and Estimate algorithms with pi/4 T gates and should run in a few seconds. The script fixed_probability_circuits.py contains more advanced usage with some of our convenience routines, and uses the implementation of the Estimate algorithm with arbitrart single qubit diagonal gates.