System and method for reducing CNOT count in Clifford+T circuits on connectivity constrained architectures
The present application recognizes the problem of reducing the CNOT-count in Clifford+T circuits on connectivity constrained architectures. Here, one can “slice” the circuit at the position of Hadamard (H) gates and “build” the intermediate portions. Two kinds of partitioning are evaluated, namely: (i) a simple method of partitioning the gates of the input circuit based on the locality of H gates, and (ii) a second method of partitioning the phase polynomial of the input circuit. The intermediate {CNOT, T} sub-circuits can be synthesized using Steiner trees, similar to the work of Nash, Gheorghiu, Mosca [NGM20] and Kissinger, de Griend [KdG19]. The following algorithms have certain procedural differences that also help to further reduce the CNOT-count. The performances of the algorithms are compared while mapping different benchmark circuits as well as random circuits to some popular architectures like 9-qubit square grid, 16-qubit square grid, Rigetti 16qubit Aspen, 16-qubit IBM QX5, 20-qubit IBM Tokyo.
1. A method of partitioning gates of circuits based on the locality of H gates in the circuits, comprising:
partitioning gates of an input circuit according to positions of H gates in the input circuit to obtain a plurality of partitioned circuits;
generating a phase polynomial of each of the partitioned circuits; and
reconstructing an output circuit from the partitioned circuits for which phase Polynomials have been generated, while maintaining connectivity constraints.
2. The method of claim 1 , wherein a linear transformation is determined for each of the partitioned circuits.
3. The method of claim 2 , further comprising determining that an overall linear transformation of the output circuit maps to a final output.
4. The method of claim 1 , further comprising applying a series of transformations such that each parity term occurs at least once in the output circuit.
5. The method of claim 4 , wherein the series of transformations comprise CNOT gates.
6. The method of claim 1 , further comprising imposing connectivity constraints by constructing Steiner trees with terminals being a set of qubits or vertices satisfying certain conditions.
7. The method of claim 1 , further comprising determining edge information and performing a series of CNOT operations to get a desired result according to the edge information.
8. The method of claim 1 , wherein a set of gates that is partitioned corresponds to the phase polynomial of the entire circuit.
9. The method of claim 8 , wherein between two H gates a phase polynomial network circuit is synthesized using gates in the circuit that realize a partial phase polynomial, including terms in the polynomial network that become incomputable after the H gate being placed at an end of the current slice.
10. The method of claim 1 , comprising synthesizing CNOT+Rz circuits using Steiner trees by:
computing a Steiner tree according to Steiner rows and terminal rows;
for each root to leaf path, flipping the role of the root and leaf, such that parity gets accumulated at the root; and
adding up the parities to obtain a desired parity, wherein a matrix is changed according to the unflipped leaf and root to reflect changes due to position of CNOT gates.
11. The method of claim 1 , wherein while synthesizing CNOT+X circuits:
(a) reducing to upper triangle, wherein at least one CNOT is not used to unnecessarily eliminate parities of terminal rows; and
(b) transposing and reducing again to upper triangle without disturbing the zeros in the now lower triangle, wherein the template or recursive procedures are not used, information about the errant rows is obtained from Steiner trees already constructed, and correction procedures are used to correct these rows.
12. A system comprising a processor and memory, the memory comprising computer executable instructions that when executed by the processor cause the system to partition gates of input circuits based on the locality of H gates in the circuits by:
partitioning gates of an input circuit according to positions of H gates in the input circuit to obtain a plurality of partitioned circuits;
generating a phase polynomial of each of the partitioned circuits; and
reconstructing an output circuit from the partitioned circuits for which phase Polynomials have been generated, while maintaining connectivity constraints.
13. The system of claim 12 , wherein a linear transformation is determined for each of the partitioned circuits.
14. The system of claim 13 , further comprising instructions for determining that an overall linear transformation of the output circuit maps to a final output.
15. System of claim 12 , further comprising instructions for applying a series of transformations such that each parity term occurs at least once in the output circuit.
16. The system of claim 15 , wherein the series of transformations comprise CNOT gates.
17. The system of claim 12 , further comprising instructions for imposing connectivity constraints by constructing Steiner trees with terminals being a set of qubits or vertices satisfying certain conditions.
18. The system of claim 12 , further comprising instructions for determining edge information and performing a series of CNOT operations to get a desired result according to the edge information.
19. The system of claim 12 , wherein a set of gates that is partitioned corresponds to the phase polynomial of the entire circuit.
20. The system of claim 19 , wherein between two H gates a phase polynomial network circuit is synthesized using gates in the circuit that realize a partial phase polynomial, including terms in the polynomial network that become incomputable after the H gate being placed at an end of the current slice.
21. The system of claim 12 , comprising instructions for synthesizing CNOT+Rz circuits using Steiner trees by:
computing a Steiner tree according to Steiner rows and terminal rows;
for each root to leaf path, flipping the role of the root and leaf, such that parity gets accumulated at the root; and
adding up the parities to obtain a desired parity, wherein a matrix is changed according to the unflipped leaf and root to reflect changes due to position of CNOT gates.
22. The system of claim 12 , comprising instructions for, while synthesizing CNOT+X circuits:
(a) reducing to upper triangle, wherein at least one CNOT is not used to unnecessarily eliminate parities of terminal rows; and
(b) transposing and reducing again to upper triangle without disturbing the zeros in the now lower triangle, wherein the template or recursive procedures are not used, information about the errant rows is obtained from Steiner trees already constructed, and correction procedures are used to correct these rows.
23. A non-transitory computer readable medium comprising computer executable instructions for partitioning gates of input circuits based on the locality of H gates in the circuits, comprising instructions for:
partitioning gates of an input circuit according to positions of H gates in the input circuit to obtain a plurality of partitioned circuits;
generating a phase polynomial of each of the partitioned circuits; and
reconstructing an output circuit from the partitioned circuits for which phase Polynomials have been generated, while maintaining connectivity constraints.