Mechanism for task abstraction and task management in quantum operating system

Abstract

The invention provides a task abstraction and task management mechanism in a quantum operating system, which comprises the following steps: (1) quantum task abstraction: abstracting a quantum computing task into a quantum process and designing a quantum process control identifier to facilitate management and resource allocation of the quantum operating system; (2) quantum task management: performing management and resource allocation on the quantum process according to information in the quantum process identifier; and (3) designing a quantum process user interface, providing a Quantum library function and Qufork and Quexec system calling, and converting a quantum computing task required to be computed by a user into a quantum process for an operating system to manage. The mechanism provides a brand new normal form different from the existing quantum cloud service. By abstracting the quantum tasks into the quantum processes, an operating system can uniformly manage the quantum tasks, the fidelity of the quantum tasks is improved, and efficient concurrency is achieved. In addition, the mechanism can be deeply combined with quantum computing resource management, and development of a quantum operating system is promoted.

Description

Technical fields:

[0001] This invention relates to the field of quantum computing, specifically to a task abstraction and task management mechanism in a quantum operating system. Background technology:

[0002] With the continuous development of quantum computing technology, quantum computers are gradually demonstrating potential speedup advantages in some classical computational problems, such as large integer prime factorization and unordered data search. However, quantum computing resources are often expensive. Current superconducting quantum chips can only integrate a limited number of physical qubits, insufficient to support the execution of large-scale quantum algorithms. Furthermore, due to factors such as crosstalk and decoherence between physical qubits, the fidelity of quantum programs is difficult to guarantee. Improving the fidelity of quantum programs is a widespread research goal in the field of quantum computing.

[0003] Superconducting quantum computers are among the most promising quantum computing technologies and have attracted widespread attention. However, due to the low-temperature storage characteristics of superconductivity, quantum computers are not yet widespread. Existing quantum computing resources often exist in the form of "quantum cloud services." This means that users submit their desired quantum computing tasks to the cloud service platform of quantum hardware vendors (Google, IBM, etc.), which then transmits the data to a real quantum computer for execution and to obtain the results. Cloud service platforms typically schedule quantum tasks according to the order in which they are submitted and the user's priority.

[0004] Classical operating system (OS) technologies can be drawn upon for abstracting and managing quantum tasks. An operating system is indispensable system software that supports the efficient execution of concurrent computing tasks and the full utilization of computing resources in a computer system. Classical operating systems typically abstract computing tasks into processes or threads as the execution units of a program. The operating system is responsible for allocating necessary computing resources, such as CPU time, memory, and I / O devices, to each process. Process scheduling is controlled by the operating system's scheduler, which determines which task can obtain processor access based on certain strategies (such as time-slice round-robin, priority scheduling, etc.) and switches execution between different tasks through context switching mechanisms. To achieve concurrent execution, the operating system also provides synchronization and mutual exclusion mechanisms (such as semaphores, locks, etc.) to ensure that multiple tasks can safely share resources without conflicts. Furthermore, the operating system is responsible for task lifecycle management, including task creation, destruction, pausing, and resumption. Through these task abstraction and management mechanisms, the operating system ensures the rational allocation and efficient execution of resources in a multi-tasking environment, maximizing the utilization of computing resources while providing a stable user experience.

[0005] However, there are significant differences between quantum programs and classical programs. For example... Figure 1As shown, quantum circuits are the basic representation of quantum programs. The execution of quantum programs relies on qubits and quantum gate operations within the quantum circuit and is constrained by the fundamental principles of quantum mechanics. Quantum programs primarily utilize the superposition and entanglement properties of qubits. However, during measurement, quantum states collapse, losing their superposition information. Furthermore, due to the no-cloning theorem, quantum states cannot be copied. This prevents quantum programs from freely copying data and restoring context like classical programs. In addition, the execution of quantum programs is highly dependent on the physical properties of quantum hardware, such as the error rate of quantum gate operations, the decoherence time of qubits, and the interactions between qubits. For example, IBM's Washington quantum chip has 127 physical qubits, with a qubit consistency maintenance time of 101.6 microseconds, while the average time required to execute a single quantum gate operation is 550.41 nanoseconds. Based on this hardware characteristic, theoretically, when the depth of a quantum program exceeds approximately 180, the decoherence effect of the qubits will cause the computational results to become invalid. These differences necessitate that the task abstraction and management of quantum operating systems consider more characteristics related to quantum programs. Summary of the Invention:

[0006] To this end, this invention proposes a task abstraction and task management mechanism in a quantum operating system, used to hide the complexity of quantum programs and quantum hardware from the user, manage concurrent quantum tasks, and improve the fidelity of quantum computing tasks. This invention consists of three parts: (1) quantum task abstraction, (2) quantum task management, and (3) quantum process user interface design. The specific contents of each part are as follows:

[0007] (I) Quantum Task Abstraction

[0008] Step 1: Quantum Process Abstraction Mechanism. To enable the operating system to manage quantum tasks, this invention abstracts quantum tasks into quantum processes. The function of a quantum process is to execute a specified circuit and return the result of the quantum circuit execution to the parent process. Based on the differences between quantum programs and classical programs, the quantum operating system can allocate physical qubit resources to the quantum process according to the circuit information it is to execute.

[0009] Step 2: Design of Quantum Process Control Identifiers. The information stored in the Process Control Identifier (PCB) is the basis for the operating system to manage the corresponding process. This invention designs a specific quantum process identifier for quantum processes, storing information related to the quantum circuits that the process will execute. The information stored in the quantum process control identifier includes two categories: quantum circuit information and quantum process runtime information. The quantum circuit information includes the number of qubits, the quantum circuit depth, and the number of multi-qubit gates. This information is mainly used to select the quantum processing unit (QPU) capable of supporting the execution of the quantum process. The quantum runtime information mainly records the QPU executing the quantum process and the mapping relationship between logical qubits and physical qubits, used for the release and management of quantum computing resources.

[0010] (II) Quantum Task Management

[0011] Step 3: Quantum Process State Management Mechanism. The states of a quantum process include newly created, waiting, ready, running, and terminated. Whether a quantum process can execute depends on the state of quantum computing resources within the current system, including the number of idle physical qubits and the quantum state decoherence time. When a quantum process is created, if there is no QPU in the current system that meets the execution conditions of the quantum process, the quantum process will enter a waiting state. When computing resources capable of executing the quantum process become available, the quantum process will enter a ready state and execute. After the quantum circuit completes execution, the quantum process returns the execution result to its parent process and then enters a terminated state. Figure 2 This is a schematic diagram of state transitions in a quantum process.

[0012] Step 4: Quantum Process Mapping Management Mechanism. Quantum bit resources are categorized as robust or non-robust. Figure 3 The topology of the IBMQ London quantum chip is shown in the diagram. Nodes in the diagram represent physical qubits, and their values ​​represent the readout error rate of that qubit. Edges in the diagram represent connections between physical qubits, and their weights represent the CNOT gate error rate. The CNOT gate error rate is a major factor affecting the fidelity of quantum programs. Given limited quantum resources, the quantum operating system prioritizes mapping quantum processes with high CNOT gate density (calculated as the number of CNOT gates in the quantum program / the number of logical qubits) to robust physical qubit regions, ensuring both fidelity and efficient concurrency in quantum programs. The robust set of physical qubits is obtained through a community discovery algorithm.

[0013] (III) Quantum Process User Interface Design

[0014] Step 5: Qufork System Call. This invention provides a new system call, Qufork, which provides an interface for creating quantum processes in a quantum operating system. Qufork is responsible for creating the quantum process and writing the information of the quantum circuit to be executed into the quantum process's control identifier, so that the quantum operating system can allocate resources and schedule the quantum process.

[0015] Step 6: The Quexec system call. This invention provides a new system call, Quexec. Qufork and Quexec are often used together. Qufork creates a quantum process, and Quexec replaces the code space of that quantum process, enabling it to execute the corresponding quantum circuits on the QPU and pass the execution results to the parent process. The Qufork and Quexec system calls abstract quantum computing tasks into quantum processes, facilitating the operating system's management of quantum computing tasks.

[0016] Step 7: Quprocess library function. This invention provides a new Quprocess library function. Its parameter is the quantum circuit to be executed by the user, and it parses the quantum circuit information. The Quprocess library function creates a new quantum process by calling Qufork and Quexec, and executes the user-provided quantum circuit. The return value of the Quprocess function is the execution result of the quantum circuit. The Quprocess library function provides an interface for users to utilize quantum computing resources.

[0017] This invention provides a task abstraction and management mechanism in a quantum operating system, comprising: (1) quantum task abstraction, which abstracts quantum computing tasks into quantum processes and designs quantum process control identifiers to facilitate management and resource allocation by the quantum operating system; (2) quantum task management, which manages quantum processes and allocates resources based on the information in the quantum process identifiers; and (3) quantum process user interface design, which provides Quprocess library functions and Qufork and Quexec system calls to transform the quantum computing tasks required by the user into quantum processes for management by the operating system. This mechanism proposes a new paradigm that differs from existing "quantum cloud services". By abstracting quantum tasks into quantum processes, the operating system can uniformly manage quantum tasks, improve the fidelity of quantum tasks, and achieve high-efficiency concurrency. In addition, this mechanism can be deeply integrated with quantum computing resource management to promote the development of quantum operating systems. Attached image description:

[0018] Figure 1 The quantum circuit diagram (search 01) corresponding to Grover's algorithm.

[0019] Figure 2 This is a schematic diagram of state transitions in a quantum process.

[0020] Figure 3 This refers to the topology of the IBMQ London quantum chip.

[0021] Figure 4 This is a schematic diagram of task abstraction and task management in a quantum operating system proposed in this invention.

[0022] Figure 5 This is a flowchart of an embodiment of the quantum-classical hybrid neural network training method of the present invention.

[0023] Figure 6 This is a demonstration diagram of a training embodiment of the quantum-classical hybrid neural network of the present invention. Specific implementation methods:

[0024] To make the above features and effects of the present invention clearer and easier to understand, specific embodiments are described below, and detailed descriptions are provided in conjunction with the accompanying drawings.

[0025] 1. Operating Environment. This invention is based on the open-source software development kit Qiskit and the Linux kernel to implement task abstraction and task management mechanisms in a quantum operating system. Qiskit is used to simulate the execution of quantum circuits.

[0026] 2. Specific steps. Figure 4 This is a design diagram of a task abstraction and task management mechanism in a quantum operating system according to the present invention. Figure 5 This is a flowchart of one embodiment of the present invention. Figure 6 This is a demonstrative diagram of one embodiment of the present invention. The present invention mainly includes the following steps:

[0027] Step 1, Quantum Process Abstraction Mechanism. In Figure 5 In the flowchart shown, the forward propagation process of the quantum circuitry in the quantum-classical hybrid neural network is defined as the quantum task to be computed. When training or using this neural network for inference, its forward propagation process is abstracted as a quantum process, which is scheduled and resource-allocated by the quantum operating system.

[0028] Step 2, Quantum Process Control Identifier Design. In Figure 5 In the flowchart shown, the number of qubits, the number of CNOT gates, and the line depth of the quantum circuit used are written into the quantum process control identifier.

[0029] Step 3, Quantum Process State Management Mechanism. Figure 5 In the flowchart shown, based on the information written into the quantum process control identifier, the quantum operating system queries a QPU that meets the conditions for the quantum process. Figure 5 In the process shown, all QPUs are simulated using a simulator. If no suitable QPU is found, the quantum process enters a waiting state.

[0030] Step 4: Quantum Process Mapping Management Mechanism. This invention employs a community discovery algorithm to construct a robust set of qubits in the quantum chip. This invention prioritizes allocating a robust set of qubits as the initial mapping region for each quantum process.

[0031] Step 5, Qufork system call. Figure 4 In the flowchart shown, for each forward propagation of the quantum-classical hybrid neural network, the present invention calls the Qufork system call once to create a quantum process for the computational task.

[0032] Step 6: Quexec system call. The Quexec system call is used in conjunction with the Qufork system call described above. This invention uses the Quexec system call to replace the code space of the quantum process created by Qufork, enabling it to execute the corresponding quantum circuit computation task.

[0033] Step 7, Quprocess library functions. Figure 4 In the flowchart shown, the user submits the quantum task to be executed through the Quprocess library function. The Quprocess library function, while parsing the quantum circuit, integrates the Qufork and Quexec system calls and returns the circuit execution result to the user.

Claims

1. A task abstraction and task management mechanism in a quantum operating system, characterized in that, include: (1) Quantum task abstraction: Quantum computing tasks are abstracted into quantum processes and quantum process control identifiers are designed to facilitate the management and resource allocation of the quantum operating system; (2) Quantum task management: Quantum processes are managed and resources are allocated according to the information in the quantum process identifiers; (3) Quantum process user interface design: Quprocess library functions, Qufork system call, and Quexec system call are provided to convert the quantum computing tasks that users need to compute into quantum processes for the operating system to manage.

2. The task abstraction and task management mechanism in a quantum operating system as described in claim 1, characterized in that, Quantum task abstraction includes: Step 1, Quantum Process Abstraction Mechanism: A process is the basic unit of operating system scheduling and resource allocation. A quantum program is abstracted into a quantum process. The function of a quantum process is to execute a specified path and return the result of the quantum path execution to the parent process. Step 2, Quantum Process Control Identifier Design: This involves designing a specific quantum process identifier for the quantum process, storing information related to the quantum circuit to be executed by the process. The information stored in the quantum process control identifier includes two types: quantum circuit information and quantum circuit runtime information. The quantum circuit information includes the number of qubits, quantum circuit depth, and number of multi-qubit gates. The quantum circuit information is mainly used to select a quantum processing unit (QPU) that can support the execution of the quantum process. The quantum circuit runtime information mainly records the QPU executing the quantum process and the mapping relationship between logical qubits and physical qubits, which is used for the release and management of quantum computing resources.

3. The task abstraction and task management mechanism in a quantum operating system as described in claim 1, characterized in that, Quantum task management includes: Step 3, Quantum Process State Management Mechanism: The states of a quantum process include new, waiting, ready, running, and terminated. When a quantum process is created, if no QPU that meets the execution conditions of the quantum process can be found, the quantum process will enter the waiting state. When the computing resources that can execute the quantum process are available, the quantum process will enter the ready state and run. After the quantum circuit is completed, the quantum process will return the circuit execution result to the parent process and then enter the terminated state. Step 4, Quantum Process Mapping Management Mechanism: When quantum resources are limited, the quantum operating system prioritizes mapping quantum processes with high CNOT gate density to robust physical qubit regions to ensure the fidelity and high-efficiency concurrency of quantum programs. The robust physical qubit set is obtained through a community discovery algorithm. The CNOT gate density is calculated as the number of CNOT gates in the quantum program / the number of logical qubits.

4. The task abstraction and task management mechanism in a quantum operating system as described in claim 1, characterized in that, The quantum process user interface design includes: Step 5, Qufork system call: This includes providing a new system call, Qufork, to provide an interface for creating quantum processes in the quantum operating system; Qufork is responsible for creating the quantum process and writing the information of the quantum circuit to be executed into the quantum process's control identifier; Step 6, Quexec system call: including the provided system call Quexec; Qufork and Quexec are often used together; Qufork creates a quantum process, and Quexec replaces the code space of the quantum process, enabling it to execute the corresponding quantum circuit on the QPU and pass the circuit execution result to the parent process; Step 7, Quprocess library functions: This includes providing new Quprocess library functions, which accept the quantum circuit to be executed by the user and parse the quantum circuit information; the Quprocess library functions create new quantum processes through the Qufork and Quexec system calls and execute the quantum circuit corresponding to the quantum process; the return value of the Quprocess function is the execution result of the quantum circuit.

Images