GPU-accelerated simulations of quantum annealing and the quantum approximate optimization algorithm

Abstract

We use a GPU-accelerated version of the massively parallel Jülich universal quantum computer simulator (JUQCS-G) to benchmark JUWELS Booster, a GPU cluster with 3744 NVIDIA A100 Tensor Core GPUs. Using JUQCS-G, we study the relation between quantum annealing (QA) and the quantum approximate optimization algorithm (QAOA). We find that a very coarsely discretized version of QA, termed approximate QA, performs surprisingly well in comparison to the QAOA. It can either be used to initialize the QAOA, or to avoid the costly optimization procedure altogether. Furthermore, we study the scaling of approximate QA for exact cover problems from 30 to 40 qubits. We find that the case with largest discretization error performs most favorably, surpassing the best result obtained from the QAOA.

Fast quantum circuit simulation using hardware accelerated general purpose libraries

“…We start from the assumptions that: a) in its simplest incarnation a quantum circuit simulator is a matrix vector multiplication software, b) the supporting libraries should be as easy as possible to install on consumer and specialised hardware; c) the performance is important for large qubit numbers when memory and communication overhead seem to be the bottleneck (ie. [20] for a very recent discussion of GPU simulators and the associated communication overhead). Our goal is to analyse the performance of simulating quantum circuits with GPUs.…”

QuOp_MPI: a framework for parallel simulation of quantum variational algorithms

“…which together form the parity-conserving mixing operator U parity (t) = e −it Blast e −it Beven e −it Bodd (16) that mixes probability amplitude between subgraphs of equal parity as illustrated in Figure 2. By initialising |ψ 0 QAOAz in a quantum state that satisfies the parity constraint, probability amplitude is constrained to S .…”

“…Classical numerical simulation plays a key role in the development of QVAs. Through simulation of the idealised quantum dynamics, researchers are able to study QVAs independently of implementation-specific hardware constraints and at scales that still exceed the functional limitations of current quantum hardware [16]. To assist with these efforts we have developed QuOp MPI (Quantum Optimisation with MPI) [17], which provides a flexible framework for the design and classical simulation of QVAs.…”

“…There is currently significant interest in developing tools for the simulations of QVAs in a high-performance computing setting. Recent examples include TensorFlow Quantum a software framework for quantum machine learning [18] and the Jülich universal quantum computer simulator [16]. Both of these utilise GPU acceleration, with the latter being targeted at distributed GPU clusters.…”

Benchmarking Advantage and D-Wave 2000Q quantum annealers with exact cover problems

“…We assess the progress in quantum annealing technology by benchmarking both Advantage and D-Wave 2000Q with exact cover problems. The exact cover problem is an NP-complete problem [40] that has become a prominent application to study quantum annealing [41][42][43][44][45] and gate-based quantum computing [46][47][48][49][50]. In our case, the exact cover problems are derived from the tail assignment problem [51] (see [49] for more information) and represent simplified aircraft scheduling scenarios.…”