mirror of https://github.com/Qiskit/qiskit.git
![]() These tests all have a built-in assumption that if the trivial layout produces a routed circuit, it will automatically be used at optimisation levels 2 and 3. This is not formally the case; only optimization levels 0 and 1 set that by default. An implementation detail of the current VF2Layout pass causes it to short-circuit out on the first possible solution without scoring, if the number of virtual qubits matches the number of physical qubits. This allowed these tests to pass without any flakiness, even though some of them used unseeded randomness in `GenericBackendV2`. This short-circuiting in VF2Layout is based on a false premise that all "complete" isomorphic layouts must have the same score, but this is not actually correct, and follow-on modifications to the VF2 pass might change this. |
||
---|---|---|
.. | ||
__init__.py | ||
test_compiler.py | ||
test_transpiler.py |