commutation analysis lead to numpy.core._exceptions._ArrayMemoryError with large mct #10488
Description
Environment
- Qiskit Terra version: 0.43.1 meta package, terra 0.24.1
- Python version: 3.10
- Operating system: docker continuumio/miniconda3
What is happening?
When transpiling (level 2) a circuit with a large multi-cx gate (see doc: here) with a large numebr of qubits (e.g. 28) the commutation analysis pass crashes.
How can we reproduce the issue?
Run this python script:
from qiskit import QuantumCircuit, transpile
qc = QuantumCircuit(29)
qc.mct(list(range(28)), qc.num_qubits - 1)
transpile(qc, optimization_level=2)Produces this output and error:
Traceback (most recent call last):
File "myfile.py", line 7, in <module> transpile(qc, optimization_level=2)
File "...qiskit/compiler/transpiler.py", line 380, in transpile _serial_transpile_circuit(
File "...qiskit/compiler/transpiler.py", line 462, in _serial_transpile_circuit result = pass_manager.run(circuit, callback=callback, output_name=outpu
t_name) File "...qiskit/transpiler/passmanager.py", line 537, in run return super().run(circuits, output_name, callback)
File "...qiskit/transpiler/passmanager.py", line 231, in run return self._run_single_circuit(circuits, output_name, callback)
File "...qiskit/transpiler/passmanager.py", line 292, in _run_single_circuit result = running_passmanager.run(circuit, output_name=output_name, call
back=callback) File "...qiskit/transpiler/runningpassmanager.py", line 125, in run dag = self._do_pass(pass_, dag, passset.options)
File "...qiskit/transpiler/runningpassmanager.py", line 169, in _do_pass dag = self._do_pass(required_pass, dag, options)
File "...qiskit/transpiler/runningpassmanager.py", line 173, in _do_pass dag = self._run_this_pass(pass_, dag)
File "...qiskit/transpiler/runningpassmanager.py", line 227, in _run_this_pass pass_.run(FencedDAGCircuit(dag))
File "...qiskit/transpiler/passes/optimization/commutation_analysis.py", line 75, in run does_commute = self.comm_checker.commute(
File "...qiskit/circuit/commutation_checker.py", line 136, in commute operator_2 = Operator(op2, input_dims=(2,) * len(qarg2), output_dims=(2
,) * len(qarg2)) File "...qiskit/quantum_info/operators/operator.py", line 85, in __init__ self._data = self._init_instruction(data).data
File "...qiskit/quantum_info/operators/operator.py", line 610, in _init_instruction op = Operator(np.eye(dimension))
File "...numpy/lib/twodim_base.py", line 215, in eye m = zeros((N, M), dtype=dtype, order=order)
numpy.core._exceptions._ArrayMemoryError: Unable to allocate 2.00 EiB for a
n array with shape (536870912, 536870912) and data type float64What should happen?
I would expect the optimizer to skip the pass if too large to optimize and leave it unoptimized.
Any suggestions?
I would skip the optimization pass when the number of qubits is too large (precise threshold to be determined based on ram memory of the current machine).