Replace initialization method by Isometry in StatePreparation

[ShellyGarion]

Summary

Close #12081
Replace initialization method by Isometry in StatePreparation

Details and comments

[qiskit-bot]

One or more of the the following people are requested to review this:

• @Cryoris
• @Qiskit/terra-core
• @ajavadia

[coveralls]

Pull Request Test Coverage Report for Build 8998409586

Warning: This coverage report may be inaccurate.

This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.

• For more information on this, see Tracking coverage changes with pull request builds.
• To avoid this issue with future PRs, see these Recommended CI Configurations.
• For a quick fix, rebase this PR at GitHub. Your next report should be accurate.

Details

• 21 of 21 (100.0%) changed or added relevant lines in 5 files are covered.
• 1766 unchanged lines in 119 files lost coverage.
• Overall coverage increased (+0.3%) to 89.631%

---

Files with Coverage Reduction	New Missed Lines	%
qiskit/converters/circuit_to_gate.py	1	97.44%
qiskit/circuit/classical/expr/constructors.py	1	91.82%
qiskit/circuit/library/standard_gates/r.py	1	97.62%
qiskit/circuit/controlflow/control_flow.py	1	92.86%
qiskit/transpiler/passes/layout/apply_layout.py	1	98.25%
qiskit/transpiler/target.py	1	93.74%
qiskit/primitives/statevector_sampler.py	1	99.08%
crates/qasm2/src/expr.rs	1	94.03%
qiskit/primitives/backend_sampler_v2.py	1	99.19%
qiskit/circuit/library/generalized_gates/permutation.py	1	93.1%

Totals
Change from base Build 8668339202:	0.3%
Covered Lines:	62195
Relevant Lines:	69390

---

💛 - Coveralls

[alexanderivrii]

Thanks @ShellyGarion, this looks good. Hopefully all the tests will pass now. Could you please add the release note? And also some data on the number of gates / depth / etc. comparing the old and the new methods, i.e. highlighting improvement in this PR?

[ShellyGarion]

The following code compares the number of CX gates and circuit depth of this branch and the main branch:

    def test_cx_num(self):
        for n_qubits in [5, 7, 10, 12, 15]:
            state = np.random.rand(2 ** n_qubits) + np.random.rand(2 ** n_qubits) * 1j
            state = state / np.linalg.norm(state)

            # State preparation
            qc = QuantumCircuit(n_qubits)
            qc.append(StatePreparation(state), range(n_qubits))

            from qiskit import transpile
            transpiled = transpile(qc, basis_gates=['u', 'cx'], optimization_level=0)
            cx_sp = transpiled.count_ops()['cx']

            print("num qubits=", n_qubits, "cx count StatePreparation=", cx_sp, "depth StatePreparation=", transpiled.depth())

this branch:

num qubits= 5 cx count StatePreparation= 26 depth StatePreparation= 53
num qubits= 7 cx count StatePreparation= 120 depth StatePreparation= 241
num qubits= 10 cx count StatePreparation= 1013 depth StatePreparation= 2027
num qubits= 12 cx count StatePreparation= 4083 depth StatePreparation= 8167
num qubits= 15 cx count StatePreparation= 32752 depth StatePreparation= 65505

main branch:

num qubits= 5 cx count StatePreparation= 52 depth StatePreparation= 110
num qubits= 7 cx count StatePreparation= 240 depth StatePreparation= 488
num qubits= 10 cx count StatePreparation= 2026 depth StatePreparation= 4063
num qubits= 12 cx count StatePreparation= 8166 depth StatePreparation= 16345
num qubits= 15 cx count StatePreparation= 65504 depth StatePreparation= 131024

[alexanderivrii]

Very nice, we see 2x reduction in the number of gates. Not directly related to this PR, but do you think it is worthwhile to track state preparation benchmarks via ASV (in the case that they are not there already)?

[ShellyGarion]

Very nice, we see 2x reduction in the number of gates. Not directly related to this PR, but do you think it is worthwhile to track state preparation benchmarks via ASV (in the case that they are not there already)?

I think it would be nice to add it. I think that Isometry is being tracked in: https://github.com/Qiskit/qiskit/blob/main/test/benchmarks/isometry.py, but it would be nice to add a specific case for StatePreparation.

[alexanderivrii]

LGTM! Just one small comment: I am personally in favor of removing internal code that is not likely to be ever used in the future. @ShellyGarion, you don't think that any of the removed methods in statepreparation.py will ever be useful, correct?

[alexanderivrii]

A quick run of ASV on the new statepreparation benchmark:

t have improved:

| Change   |   Before [b5c5179a] <main> |   After [4dab85c4] <pr-shelly> |   Ratio | Benchmark (Parameter)                                                                                   |
|----------|----------------------------|--------------------------------|---------|---------------------------------------------------------------------------------------------------------|
| -        |                         28 |                             17 |    0.61 | statepreparation.StatePreparationTranspileBench.track_cnot_counts_after_mapping_to_ibmq_16_melbourne(4) |
| -        |                        186 |                            111 |    0.6  | statepreparation.StatePreparationTranspileBench.track_cnot_counts_after_mapping_to_ibmq_16_melbourne(6) |
| -        |                         79 |                             47 |    0.59 | statepreparation.StatePreparationTranspileBench.track_cnot_counts_after_mapping_to_ibmq_16_melbourne(5) |
| -        |                        411 |                            231 |    0.56 | statepreparation.StatePreparationTranspileBench.track_cnot_counts_after_mapping_to_ibmq_16_melbourne(7) |
| -        |                        872 |                            460 |    0.53 | statepreparation.StatePreparationTranspileBench.track_cnot_counts_after_mapping_to_ibmq_16_melbourne(8) |

SOME BENCHMARKS HAVE CHANGED SIGNIFICANTLY.
PERFORMANCE INCREASED.