PyQCS is a Quantum Computing Simulator built for physics. It currently features two simulator backends for different purposes, and a way to construct circuits in a relatively readable manner.
By default PyQCS employs a relatively slow simulator backend using dense state vectors. Note that any implementation using dense state vectors shows exponential growth in the number of qbits. Therefore this simulator backend is limited to qbit numbers below 30 for reasonable performance. For simulations requiring more qbits we recommend using a high performance framework, such as GPT's QIS module.
The second backend uses a graphical state representation (see for instance
arXiv:quant-ph/0504117) which
allows for the simulation of stabilizer states and -circuits. The graphical
simulator is considerably faster, in particular it does not exhibit exponential
growth in the number of qbits. The graphical states are available as
pyqcs.graph.state.GraphState.
Unlike other simulators PyQCS focuses on the state: Users start from a state, modify the state (using circuits) and then either look at the state or sample from the state. This direct access to the state is useful when debugging circuits or when considering physical problems. However, it can slow down compuations.
To do some computation one has to build a quantum circuit and apply it to a state.
States are created using pyqcs.State.new_zero_state(<number of qbits>).
Circuits are built from the fundamental gates (see Built-in Gates) by joining them
together using the | operator:
from pyqcs import H, CX, X circuit = H(0) | CX(1, 0) | X(1)
The usage of the | is in analogy to the UNIX pipe: gates are applied from left to
right. This is in agreement with the Feynman quantum circuit diagrams.
Note: the circuit above would have the following matrix representation:
X_1 CZ_{1,0} H_0
Applying a circuit to a state is done using multiplication:
from pyqcs import State state = State.new_zero_state(2) resulting_state = circuit * state
New in v2.2.0 is the circuitpng function that allows displaying circuits as PNGs
(using a pdflatex implementation and imagemagick):
from pyqcs import H, CX, circuitpng circuit = (H(1) | H(2)) | CX(2, 1) | (H(1) | H(2)) circuitpng(circuit)
PyQCS currently has the following gates built-in:
X- Pauli-X or NOT gate. Flips the respective qbit.
H- Hadamard gate.
CX- CNOT (controlled NOT) gate. Flips the act-qbit, if the control-qbit is set.
R- R, Rz or R_phi, the rotation gate. Rotates the respective qbit around a given angle.
M- Measurement gate: this gate measures the respective gate, collapsing the wave function and storing the result in the classical part of the state.
Z- Pauli-Z gate.
S- Clifford-S gate.
CZ- Controlled Z gate.
Starting from version 3.0.0 PyQCS has a pure C++ backend omitting the previously used numpy arrays. The python package uses handwritten adapters to this backend.
The backend can be used as a stand-alone library. It can be built and installed
using the src/backend/meson.build file. Its usage is a bit different from
what one would usually expect from a simulator: Operations like measurement are
not implemented explicitly but should be implemented by the user using a random
number generator and the provided compute_amplitude methods.
Besides that the code should be pretty much self-explaining. See the adapter code for some ideas how to use the backend.
- Add pretty printers for states.
- Write lot's of documentation.
- Add more tests.
- Add a noise model.
- Add a way to export circuits to GPT's QIS module.
| [1] | Real quantum computers have an intrinsic time evolution. This is omitted in PyQCS and reintroduced for error simulation. PyQCS therefore operates on a discrete quasi-time with every time-site being before or after a gate application. |