[PoC] Reduce the number of reshapes to improve Statevector performance

[t-imamichi]

Summary

This PR is a PoC to study the performance issue to resolve #14028.
This is made for discussion.

• Added flatten option to control the flatten operation in Statevector. _evolve_operator
  • Changed to apply flatten only at the beginning and the end of Statevector._evolve_instruction
• Simplified the contraction with np.tensordot
  • It does not have any performance improvement though

Details and comments

import timeit

import numpy as np
from qiskit import __version__ as qiskit_vesrion
from qiskit.circuit.library import EfficientSU2
from qiskit.quantum_info import Statevector

print(f"{qiskit_vesrion=}")

n_qubits = 18

qiskit_circuit = EfficientSU2(n_qubits, entanglement="pairwise", reps=n_qubits)
qiskit_circuit.assign_parameters(np.ones(qiskit_circuit.num_parameters), inplace=True)

number = 1
execution_time = timeit.timeit(lambda: Statevector(qiskit_circuit), number=number)
print(f"Average execution time: {execution_time / number:.6f} seconds")

main branch a6d6584

qiskit_vesrion='2.1.0.dev0+a6d6584'
Average execution time: 1.764300 seconds

this PR

qiskit_vesrion='2.1.0.dev0+313d92d'
Average execution time: 1.373686 seconds

[coveralls]

Pull Request Test Coverage Report for Build 13893388124

Details

• 15 of 15 (100.0%) changed or added relevant lines in 1 file are covered.
• 4 unchanged lines in 2 files lost coverage.
• Overall coverage increased (+0.02%) to 88.14%

---

Files with Coverage Reduction	New Missed Lines	%
crates/accelerate/src/unitary_synthesis.rs	1	94.69%
crates/qasm2/src/lex.rs	3	92.98%

Totals
Change from base Build 13864210106:	0.02%
Covered Lines:	72730
Relevant Lines:	82516

---

💛 - Coveralls