On the Influence of Initial Qubit Placement During NISQ Circuit Compilation
Abstract
Noisy IntermediateScale Quantum (NISQ) machines are not faulttolerant, operate few qubits (currently, less than hundred), but are capable of executing interesting computations. Above the quantum supremacy threshold (approx. 60 qubits), NISQ machines are expected to be more powerful than existing classical computers. One of the most stringent problems is that computations (expressed as quantum circuits) have to be adapted (compiled) to the NISQ hardware, because the hardware does not support arbitrary interactions between the qubits. This procedure introduces additional gates (e.g. SWAP gates) into the circuits while leaving the implemented computations unchanged. Each additional gate increases the failure rate of the adapted (compiled) circuits, because the hardware and the circuits are not faulttolerant. It is reasonable to expect that the placement influences the number of additionally introduced gates. Therefore, a combinatorial problem arises: how are circuit qubits allocated (placed) initially to the hardware qubits? The novelty of this work relies on the methodology used to investigate the influence of the initial placement. To this end, we introduce a novel heuristic and cost model to estimate the number of gates necessary to adapt a circuit to a given NISQ architecture. We implement the heuristic (source code available on github) and benchmark it using a standard compiler (e.g. from IBM Qiskit) treated as a black box. Preliminary results indicate that cost reductions of up to 10\% can be achieved for practical circuit instances on realistic NISQ architectures only by placing qubits differently than default (trivial placement).
 Publication:

arXiv eprints
 Pub Date:
 November 2018
 DOI:
 10.48550/arXiv.1811.08985
 arXiv:
 arXiv:1811.08985
 Bibcode:
 2018arXiv181108985P
 Keywords:

 Quantum Physics;
 Computer Science  Emerging Technologies;
 Computer Science  Performance
 EPrint:
 accepted at QTOP 2019, to appear in Volume 11413 of the Lecture Notes in Computer Science series