(More difficult issue)
When we try to map to hardware (especially for merge moves), sometimes the moves do not fit the topology of the device. For instance, consider this move sequence:
- Rook splits from a1 to b1
- Rook on a2 moves to a3.
- Rooks on a3 and b1 merge to b3.
This creates this tree:
a1 - a2 - a3
| |
b1 - b3
This "pentagon" layout will not fit on the grid. However, if we added a bonus ancillary qubit with SWAP gates at b2, it could fit in a 2x3 rectangle on the grid. We need to be able to detect these problems and correct them by adding extra qubits on one of the paths.
(More difficult issue)
When we try to map to hardware (especially for merge moves), sometimes the moves do not fit the topology of the device. For instance, consider this move sequence:
This creates this tree:
a1 - a2 - a3
| |
b1 - b3
This "pentagon" layout will not fit on the grid. However, if we added a bonus ancillary qubit with SWAP gates at b2, it could fit in a 2x3 rectangle on the grid. We need to be able to detect these problems and correct them by adding extra qubits on one of the paths.