To harness the power of quantum computing, it is essential that a quantum computer provide maximal possible fidelity for a quantum circuit. To this end, much work has been done in the context of qubit routing or embedding, i.e., mapping circuit qubits to physical qubits based on gate performance metrics to optimize the fidelity of execution. Here, we take an alternative approach that leverages a unique capability of a trapped-ion quantum computer, i.e., the all-to-all qubit connectivity. We develop a method to determine a fixed number (budget) of quantum gates that, when calibrated, will maximize the fidelity of a batch of input quantum programs. This dynamic allocation of calibration resources on randomly accessible gates, determined using our heuristics, increases, for a wide range of calibration budget, the average fidelity from 70% or lower to 90% or higher for a typical batch of jobs on an 11-qubit device, in which the fidelity of calibrated and uncalibrated gates are taken to be 99% and 90%, respectively. Our heuristics are scalable, more than 2.5 orders of magnitude faster than a randomized method for synthetic benchmark circuits generated based on real-world use cases.

为了利用量子计算的力量，量子计算机必须为量子电路提供最大可能的保真度。为此，在量子位路由或嵌入的上下文中已经做了很多工作，即将电路量子位映射到基于门性能指标的物理量子位以优化执行的保真度。在这里，我们采用了一种替代方法，该方法利用了俘获离子量子计算机的独特功能，即全对全量子位连接。我们开发了一种方法来确定固定数量（预算）的量子门，当校准时，将最大化一批输入量子程序的保真度。使用我们的启发式方法确定的随机访问门上校准资源的动态分配，增加了广泛的校准预算，11-qubit 设备上典型批次的平均保真度从 70% 或更低到 90% 或更高，其中校准和未校准门的保真度分别为 99% 和 90%。我们的启发式方法是可扩展的，比基于实际用例生成的合成基准电路的随机方法快 2.5 个数量级以上。