Bosonic quantum computing, a type of quantum computing that uses bosonic modes as the basic information processing unit, has shown promise for various practical applications. However, the development of software and compiler optimizations for this type of computing is lagging. To address this, researchers from the University of Pennsylvania, AWS Quantum Technology, and IBM Quantum have introduced Bosehedral, an efficient compiler optimization framework for Gaussian Boson sampling on Bosonic quantum hardware. Bosehedral can reduce gate by 25 to 40 percent but maintain program fidelity of 98 to 99.9 percent, leading to significant end-to-end application performance improvement.
What is Bosonic Quantum Computing and Why is it Important?
Bosonic quantum computing, also known as continuous variable quantum computing, is a type of quantum computing that uses bosonic modes as the basic information processing unit. Unlike qubit-based quantum computing, which uses discrete variables, bosonic quantum computing uses qumodes, which have an infinite dimensional state space. This type of quantum computing has shown promise for various practical applications that are difficult for classical computing, such as graph clique, graph similarity, point process, and molecule vibrational spectra simulation.
Bosonic quantum computing is attractive for several reasons. Firstly, it has long lifetimes of qumodes, such as superconducting cavities. Secondly, it has the ability to transduce between stationary and flying information. Lastly, it has a strong built-in information encoding and processing capability to naturally and efficiently encode certain computations.
What is the Current State of Bosonic Quantum Computing?
Despite the great progress in hardware, the development of software and compiler optimizations for Bosonic quantum computing is far behind. Early efforts on programming and compilation for Bosonic quantum computing provide basic programming interfaces and operation decomposition functions, but miss program optimizations. This has hindered the full exploitation of these computing platforms.
The complexity of devising compiler optimizations for Bosonic quantum computing arises from the inherent infinite-dimensional state space of each individual qumode. All the gates manipulating the state of qumodes have infinite-dimensional gate matrices, making it nontrivial for the quantum compiler that runs on a classical computer to derive equivalent program transformations for optimization.
What is the Solution to this Challenge?
A team of researchers from the University of Pennsylvania, AWS Quantum Technology, and IBM Quantum have introduced Bosehedral, an efficient compiler optimization framework for Gaussian Boson sampling on Bosonic quantum hardware. Bosehedral overcomes the challenge of handling infinite dimensional qumode gate matrices by performing all its program analysis and optimizations at a higher algorithmic level using a compact unitary matrix representation.
Bosehedral optimizes qumode gate decomposition and logical-to-physical qumode mapping and introduces a tunable probabilistic gate dropout method. This significantly improves the performance by accurately approximating the original program with much fewer gates.
How Effective is Bosehedral?
Bosehedral can reduce gate by 25 to 40 percent but maintain program fidelity of 98 to 99.9 percent for various benchmarks and underlying architectures. Compared with the baseline Strawberry Fields, the divergence between the sampled output distribution in noisy simulation and the ideal output distribution is reduced by 26.1 percent on average. This translates to significant end-to-end application performance improvement.
What are the Key Contributions of this Research?
The researchers have proposed Bosehedral, the first efficient and effective compiler optimization framework for Gaussian Boson sampling in Bosonic quantum computing. Bosehedral overcomes the challenges of infinite dimensional gate matrices by performing program analysis and compiler optimization at a high-level representation. The team has proposed several compiler optimization algorithms for qumode gate decomposition, logical-to-physical qumode mapping, and probabilistic gate dropout for program simplification. The evaluation shows that Bosehedral can outperform baseline Bosonic quantum compilers by significantly improving the execution fidelity and the end-to-end application performance for various benchmarks and architectures.`
The article titled “Bosehedral: Compiler Optimization for Bosonic Quantum Computing” was published on February 3, 2024. The authors of this article are Junyu Zhou, Kun Wang, Yunong Shi, Ali Javadi-Abhari, and Gushu Li. The article was sourced from arXiv, a repository managed by Cornell University.