QPTest: Automated Testing of Quantum Computing Platforms

量子计算正在成为一种新的计算范式，它允许比传统计算更有效地解决一些算法挑战。执行量子程序需要量子编程平台，该平台实现了量子编程语言，提供了优化编译器和执行环境，可以模拟程序或在量子硬件上运行程序。与传统的编译器和运行时引擎类似，这些平台是程序执行的关键基础设施。提高这一关键基础设施的可靠性对于实现主流量子编程至关重要。然而，量子编程平台的自动化测试技术目前大多尚未开发。简单地采用传统的编译器测试技术是不切实际的，因为量子编程遵循一种完全不同的编程范式。例如，量子程序由不同的基本构建块组成，如量子比特、门和电路，与传统计算不同，程序的行为本质上是概率性的。本提案中提出的项目将通过解决三个正交的挑战，为自动测试量子计算平台创建第一个全面的框架：(1)由于目前量子程序相对较少，我们将探索自动程序生成器，例如，使用专门设计用于产生可能有效的量子程序的语法，并使用由执行期间观察到的量子位分布指导的生成算法。(2)由于量子计算平台缺乏正式定义的测试预言机，我们将设计特定于量子的元态测试原语，同时解决相同量子程序的不同执行可能产生不同结果的问题。(3)最后，为了帮助理解和调试在测试平台中检测到的编程错误，我们将创建新的程序约简技术，由于量子程序固有的概率性质，这是非平凡的。我们计划将我们的想法应用于几个现实世界的量子计算平台，如Qiskit和Circ。如果成功，QPTest项目将为自动测试量子计算平台提供基础，并产生类似于传统编译器测试的早期工作的影响。

Quantum computing is emerging as a novel computing paradigm, which allows for addressing some algorithmic challenges significantly more efficiently than traditional computing. Executing a quantum program requires a quantum programming platform, which implements a quantum programming language, offers an optimizing compiler, and an execution environment that either simulates the program or runs it on quantum hardware. Similar to traditional compilers and runtime engines, these platforms are the critical infrastructure on which programs execute. Increasing the reliability of this critical infrastructure is crucial to enable mainstream quantum programming. However, automated testing techniques for quantum programming platforms currently are mostly unexplored. Simply adapting traditional compiler testing techniques is impractical, as quantum programming follows a fundamentally different programming paradigm. For example, quantum programs are composed of different basic building blocks, such qubits, gates, and circuits, and unlike in traditional computing, the behavior of a program is inherently probabilistic. The project presented in this proposal will create the first comprehensive framework for automatically testing quantum computing platforms – by addressing three orthogonal challenges: (1) As there currently are relatively few quantum programs, we will explore automated program generators, e.g., using a grammar specifically designed to yield likely valid quantum programs and using a generation algorithm guided by the distributions of qubits observed during executions. (2) Motivated by the lack of formally defined test oracles for quantum computing platforms, we will design quantum-specific meta-morphic testing primitives, while addressing the problem that different executions of the same quantum program may yield different results. (3) Finally, to help understand and debug programming errors detected in the tested platforms, we will create novel program reduction techniques, which is non-trivial due to the inherently probabilistic nature of quantum programs. We plan to apply our ideas to several real-world quantum computing platforms, such as Qiskit and Circ. If successful, the QPTest project will provide a foundation for automatically testing quantum computing platforms, and have an impact similar to early work on traditional compiler testing.