48 releases
0.28.291394 | Jul 31, 2023 |
---|---|
0.28.277227 | Jun 6, 2023 |
0.28.263081 | Mar 29, 2023 |
0.27.244707 | Dec 8, 2022 |
0.25.222597 | Jul 26, 2022 |
#353 in Math
126 downloads per month
265KB
5K
SLoC
Rust-based Quantum Development Kit Simulators
ⓘ TIP
This crate provides low-level APIs for interacting with the Rust-base simulators included in the Quantum Development Kit. If you're interested in using these simulators to run your Q# programs, please see the installation instructions at https://github.com/microsoft/qsharp-runtime/blob/main/documentation/preview-simulators.md.
This crate implements simulation functionality for the Quantum Development Kit, including:
- Open systems simulation
- Stabilizer simulation
The c_api
module allows for using the simulation functionality in this crate from C, or from other languages with a C FFI (e.g.: C++ or C#), while Rust callers can take advantage of the structs and methods in this crate directly.
Similarly, the python
module allows exposing data structures in this crate to Python programs.
Representing quantum systems
This crate provides several different data structures for representing quantum systems in a variety of different conventions:
State
: Represents stabilizer, pure, or mixed states of a register of qubits.Process
: Represents processes that map states to states.Instrument
: Represents quantum instruments, the most general form of measurement.
Noise model serialization
Noise models can be serialized to JSON for interoperability across languages. In particular, each noise model is represented by a JSON object with properties for each operation, for the initial state, and for the instrument used to implement $Z$-basis measurement.
For example:
{
"initial_state": {
"n_qubits": 1,
"data": {
"Mixed": {
"v": 1, "dim":[2 ,2],
"data": [[1.0, 0.0], [0.0, 0.0], [0.0, 0.0], [0.0, 0.0]]
}
}
},
"i": {
"n_qubits": 1,
"data": {
"Unitary": {
"v": 1,"dim": [2, 2],
"data": [[1.0, 0.0], [0.0, 0.0], [0.0, 0.0], [1.0, 0.0]]
}
}
},
...
"z_meas": {
"Effects": [
{
"n_qubits": 1,
"data": {
"KrausDecomposition": {
"v":1, "dim": [1, 2, 2],
"data": [[1.0, 0.0], [0.0, 0.0], [0.0, 0.0], [0.0, 0.0]]
}
}
},
{
"n_qubits": 1,
"data": {
"KrausDecomposition": {
"v": 1,"dim": [1, 2, 2],
"data":[[0.0, 0.0], [0.0, 0.0], [0.0, 0.0], [1.0, 0.0]]
}
}
}
]
}
}
The value of the initial_state
property is a serialized State
, the value of each operation property (i.e.: i
, x
, y
, z
, h
, s
, s_adj
, t
, t_adj
, and cnot
) is a serialized Process
, and the value of z_meas
is a serialized Instrument
.
Representing arrays of complex numbers
Throughout noise model serialization, JSON objects representing $n$-dimensional arrays of complex numbers are used to store various vectors, matrices, and tensors. Such arrays are serialized as JSON objects with three properties:
v
: The version number of the JSON schema; must be"1"
.dims
: A list of the dimensions of the array being represented.data
: A list of the elements of the flattened array, each of which is represented as a list with two entries representing the real and complex parts of each element.
For example, consider the serialization of the ideal y
operation:
"y": {
"n_qubits": 1,
"data": {
"Unitary": {
"v": 1, "dim": [2, 2],
"data": [[0.0, 0.0], [0.0, 1.0], [0.0, -1.0], [0.0, 0.0]]
}
}
}
Representing states and processes
Each state and process is represented in JSON by an object with two properties, n_qubits
and data
. The value of data
is itself a JSON object with one property indicating which variant of the StateData
or ProcessData
enum is used to represent that state or process, respectively.
For example, the following JSON object represents the mixed state $\ket{0}\bra{0}$:
{
"n_qubits": 1,
"data": {
"Mixed": {
"v": 1, "dim":[2 ,2],
"data": [[1.0, 0.0], [0.0, 0.0], [0.0, 0.0], [0.0, 0.0]]
}
}
}
Representing instruments
TODO
Known issues
- Performance of open systems simulation still needs additional work for larger registers.
- Some gaps in different conversion functions and methods.
- Stabilizer states cannot yet be measured through
Instrument
struct, only through underlyingTableau
. - Many parts of the crate do not yet have Python bindings.
- Test and microbenchmark coverage still incomplete.
Crate features
Dependencies
~7–13MB
~216K SLoC