Quantum classical hybrid computing architecture
Through the quantum classic hybrid computing architecture, including quantum cloud platform and task scheduling server, the challenge of existing cloud platforms being unable to be directly applied to quantum computing is solved, and the effective management and resource scheduling of quantum computing tasks is realized, ensuring the smooth execution of tasks and Resource utilization efficiency.
Patent Information
- Application Number
- CN202411101696.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-11-29
AI Technical Summary
The existing cloud platforms are mainly aimed at classical computing designs and cannot be directly applied to quantum computing, which leads to challenges in the management and resource scheduling of quantum computing tasks. It is necessary to design a quantum classic hybrid computing architecture to effectively manage and run quantum computing tasks.
A classic quantum hybrid computing architecture was designed, including a quantum cloud platform, a task scheduling server, a quantum simulator and a quantum computer. User requests were obtained through the quantum cloud platform, the task scheduling server performed quantum circuit optimization and shear, and assigned tasks to quantum simulation. The task pool of the machine or quantum computer and the task schedule is performed through machine learning models to predict execution time and accuracy.
It realizes effective management and operation of quantum computing tasks, ensures the smooth execution of user tasks, and improves the utilization efficiency and computing power of quantum computing resources.
Smart Images

Figure CN119047593A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of quantum computing technology, and in particular, to a quantum-classical hybrid computing architecture. Background Art
[0002] There are essential differences between quantum computers and classical computers in computing principles and architectures, which leads to significant differences in task processing and resource management between the two. Classical computers rely on logical operations of binary bits, while quantum computers use the properties of quantum bits (qubits) such as quantum state superposition and entanglement to achieve exponential growth in computing power. Although quantum computers have shown computing potential that surpasses classical computers in specific problems, such as solving large-scale combinatorial optimization and quantum chemical simulation, their practical application still faces challenges.
[0003] Quantum computing can be run through a quantum simulator or a quantum computer, where the quantum simulator is implemented by a classical computer. The programming and task scheduling of quantum computing are completely different from those of classical computers. Existing cloud platforms are mainly designed for classical computing and cannot be directly applied to quantum computing. Therefore, in order to effectively manage and run quantum computing tasks and schedule computing resources between quantum simulators and quantum computers, a quantum-classical hybrid computing architecture specifically designed for quantum computing is required. Summary of the invention
[0004] The purpose of the present disclosure is to provide a quantum-classical hybrid computing architecture, which is designed to allocate quantum computing resources of quantum simulators and quantum computers for quantum computing task scheduling, and to manage and run quantum computing tasks.
[0005] In order to achieve the above objectives, the present disclosure provides a quantum classical hybrid computing architecture, the architecture comprising:
[0006] Quantum cloud platform;
[0007] A task scheduling server, connected to the quantum cloud platform for communication;
[0008] A quantum simulator, connected in communication with the task scheduling server;
[0009] A quantum computer, in communication with the task scheduling server;
[0010] in:
[0011] The quantum cloud platform is used to obtain a quantum computing request submitted by a user, wherein the quantum computing request includes a quantum computing task, identity authentication information, and computer type information that is allowed to run the quantum computing task. The quantum cloud platform is also used to determine whether the user has the authority to submit a quantum computing request based on the identity authentication information, and when it is determined that the user has the authority, send the quantum computing task and the computer type information to the task scheduling server;
[0012] The task scheduling server is used to perform quantum circuit optimization and quantum circuit cutting on the quantum computing task to obtain a quantum computing subtask, and when the computer type information represents a quantum simulator, the quantum computing subtask is assigned to a first task pool of the quantum simulator, and when the computer type information represents a quantum computer, the quantum computing subtask is assigned to a second task pool of the quantum computer;
[0013] The quantum simulator is used to obtain quantum computing subtasks from the first task pool for execution and return the execution results to the task scheduling server, and the quantum computer is used to obtain quantum computing subtasks from the second task pool for execution and return the execution results to the task scheduling server;
[0014] The task scheduling server is also used to obtain the calculation result of the quantum computing task according to the execution result, and return the calculation result to the quantum cloud platform. The quantum cloud platform is also used to output the calculation result to the corresponding user.
[0015] Optionally, allocating the quantum computing subtask to a second task pool of a quantum computer comprises:
[0016] Extracting quantum program features of the quantum computing subtask;
[0017] Inputting the quantum program features into a trained preset machine learning model to obtain the predicted execution time and predicted execution accuracy of quantum computing subtasks on each quantum computer;
[0018] Among quantum computers whose predicted execution accuracy is greater than or equal to a preset threshold and whose number of quantum bits is greater than or equal to the number of quantum bits required to run the quantum computing subtask, determine the quantum computer with the shortest predicted execution time, and assign the quantum computing subtask to the second task pool of the quantum computer with the shortest predicted execution time.
[0019] Optionally, the preset machine learning model is any one of a multivariate linear regression model, a multi-output random forest regression model or a multi-output gradient boosting regression model.
[0020] Optionally, the quantum program features include the number of quantum circuit bits, the number of quantum circuit measurements, the number of single quantum logic gates without parameters, the number of single quantum logic gates with parameters, the number of double quantum logic gates without parameters, the number of double quantum logic gates with parameters, the number of multi-quantum logic gates without parameters, the number of multi-quantum logic gates with parameters, the number of conditional check instructions, and the number of loop instructions.
[0021] Optionally, when the second task pool is not empty, the task scheduling server is further used to map the head quantum computing subtask in the second task pool to the quantum bit of the quantum computer corresponding to the second task pool, determine the remaining number of free quantum bits of the quantum computer, and search the second task pool for other quantum computing subtasks whose required number of quantum bits is less than or equal to the number of free quantum bits, and then map the quantum computing subtask to the remaining free quantum bits of the quantum computer corresponding to the second task pool, so that the quantum computer obtains the quantum computing subtask mapped to its own quantum bit from the second task pool for execution.
[0022] Optionally, when the second task pool is empty, the task scheduling server schedules quantum computing subtasks from the second task pools of other non-empty quantum computers to the empty second task pool.
[0023] Optionally, when the first task pool is not empty, the task scheduling server uses the total number of quantum bits of the quantum simulator corresponding to the first task pool minus the number of quantum bits required for the head quantum computing subtask in the first task pool to obtain the remaining available quantum bits of the quantum simulator, and searches for a quantum computing subtask whose required number of quantum bits is less than or equal to the number of available quantum bits among other quantum computing subtasks in the first task pool, and determines that the quantum computing subtask is run together with the head quantum computing subtask on the quantum simulator corresponding to the first task pool.
[0024] Optionally, when the first task pool is empty, the task scheduling server schedules quantum computing subtasks from the first task pools of other non-empty quantum simulators to the empty first task pool.
[0025] Optionally, the quantum computer is also used to load the interface to parse the quantum computing subtask, convert the quantum computing subtask into a quantum instruction linked list, and fill it into a designated memory space according to the instruction sequence, and the FPGA driver controls the microwave transmitter to execute the quantum circuit.
[0026] Optionally, the quantum cloud platform is further configured to output information prompting the user that the user does not have the authority when it is determined that the user does not have the authority.
[0027] Through the above technical solution, the quantum cloud platform sends the quantum computing tasks and computer type information submitted by authorized users to the task scheduling server. The task scheduling server optimizes and cuts the quantum circuits of the quantum computing tasks to obtain quantum computing subtasks, and assigns the quantum computing subtasks to the first task pool of the quantum simulator or the second task pool of the quantum computer according to the computer type information, so that the quantum simulator and the quantum computer can obtain the quantum computing subtasks from their respective task pools for execution, and return the execution results to the task scheduling server. The task scheduling server obtains the calculation results of the quantum computing tasks according to the execution results of the quantum computing subtasks, and returns the output to the quantum cloud platform. In this way, the management and operation of quantum computing tasks are realized through the quantum-classical hybrid computing architecture, ensuring the smooth execution of users' quantum computing tasks.
[0028] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings:
[0030] Figure 1 It is an application scenario diagram of a quantum-classical hybrid computing architecture according to an exemplary embodiment.
[0031] Figure 2 The figure is a block diagram of a task scheduling server according to an exemplary embodiment.
[0032] Figure 3 It is a block diagram of a quantum classical hybrid computing architecture according to an exemplary embodiment. DETAILED DESCRIPTION
[0033] The specific implementation of the present disclosure is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the present disclosure, and is not used to limit the present disclosure.
[0034] Figure 1 is an application scenario diagram of a quantum classical hybrid computing architecture according to an exemplary embodiment, see Figure 1 , the quantum classical hybrid computing architecture includes:
[0035] Quantum cloud platform;
[0036] A task scheduling server, connected to the quantum cloud platform for communication;
[0037] A quantum simulator, connected in communication with the task scheduling server;
[0038] A quantum computer, in communication with the task scheduling server;
[0039] in:
[0040] The quantum cloud platform is used to obtain a quantum computing request submitted by a user, wherein the quantum computing request includes a quantum computing task, identity authentication information, and computer type information that is allowed to run the quantum computing task. The quantum cloud platform is also used to determine whether the user has the authority to submit a quantum computing request based on the identity authentication information, and when it is determined that the user has the authority, send the quantum computing task and the computer type information to the task scheduling server;
[0041] The task scheduling server is used to perform quantum circuit optimization and quantum circuit cutting on the quantum computing task to obtain a quantum computing subtask, and when the computer type information represents a quantum simulator, the quantum computing subtask is assigned to a first task pool of the quantum simulator, and when the computer type information represents a quantum computer, the quantum computing subtask is assigned to a second task pool of the quantum computer;
[0042] The quantum simulator is used to obtain quantum computing subtasks from the first task pool for execution and return the execution results to the task scheduling server, and the quantum computer is used to obtain quantum computing subtasks from the second task pool for execution and return the execution results to the task scheduling server;
[0043] The task scheduling server is also used to obtain the calculation result of the quantum computing task according to the execution result, and return the calculation result to the quantum cloud platform. The quantum cloud platform is also used to output the calculation result to the corresponding user.
[0044] The quantum cloud platform is a platform that uses cloud computing technology to provide quantum computing resources and services. Users can build quantum circuits through code programming or quantum simulator graphical user interface. In possible implementations, users can also submit quantum circuit information that meets the quantum cloud platform's quantum circuit construction standards in the form of files to build quantum circuits. Figure 1 as well as Figure 3 The quantum cloud platform 110 can be accessed through a computer terminal, which may include a Python programming module and a drag-and-drop graphical programming module. The Python programming module is used for users to edit quantum circuits corresponding to quantum computing tasks using the Python programming language, and the drag-and-drop graphical programming module is used for users to drag quantum logic gates to the target position of the quantum circuit to construct the quantum circuit.
[0045] The task scheduling server is a server used to schedule and allocate quantum computing tasks. It allocates quantum computing tasks to appropriate quantum simulators and / or quantum computers according to preset rules. There can be one or more task scheduling servers, such as Figure 1 As shown, the quantum cloud platform 110 is connected to the task scheduling server 121 and the task scheduling server 122, and multiple task scheduling servers work together. The task scheduling server may include a quantum circuit optimization module, a quantum circuit cutting module, and a quantum circuit topology mapping module, which are respectively used to implement corresponding functions.
[0046] See also Figure 2 The task scheduling server may include a processor 1922, which may be one or more, and a memory 1932 for storing a computer program executable by the processor 1922. The computer program stored in the memory 1932 may include one or more modules each corresponding to a set of instructions.
[0047] In addition, the task scheduling server 1900 may further include a power supply component 1926 and a communication component 1950, wherein the power supply component 1926 may be configured to perform power management of the electronic device 1900, and the communication component 1950 may be configured to implement communication of the electronic device 1900, for example, wired or wireless communication. In addition, the electronic device 1900 may further include an input / output (I / O) interface 1958. The electronic device 1900 may operate based on an operating system stored in the memory 1932, such as Windows Server™, Mac OS X™, Unix™, Linux™, etc.
[0048] A quantum simulator is a software simulator used to simulate quantum computing. It uses the resources of a classical computer to build a virtual architecture to simulate the operation of a quantum program corresponding to a specific problem. Figure 3 Quantum simulators simulate quantum computing based on central processing units (CPUs) and graphics processing units (GPUs). Their computing power is limited by the computing power of the CPUs and GPUs. Quantum computers are computing devices designed based on the principles of quantum mechanics. Based on their implementation principles, they are divided into different types, such as optical quantum computers, ion trap quantum computers, superconducting quantum computers, and semiconductor quantum computers. In a quantum-classical hybrid computing architecture, there can be one or more quantum simulators and quantum computers, as shown in Figure 1 , the quantum simulator may include quantum simulator 131 and quantum simulator 132 , and the quantum computer may include quantum computer 133 and quantum computer 134 .
[0049] Specifically, the quantum cloud platform can be accessed by users through computer terminals, and then users can create quantum computing tasks on the quantum cloud platform through computer terminals, such as creating quantum computing tasks through Python programming modules and drag-and-drop graphical programming modules. Quantum computing tasks are computing tasks used to perform quantum computing, and may include information such as quantum circuits and the number of times quantum circuits are executed. Users can also enter or select computer type information that quantum computing tasks are allowed to run through computer terminals, such as different numbers, letters, or phrases to represent quantum simulators or quantum computers. After the user logs in to his or her account on the quantum cloud platform, the quantum cloud platform can also obtain the authentication information of the corresponding account to identify the user. After the user builds the quantum computing task locally on the computer terminal and enters the computer type information, the computer obtains the user's authentication information and packages it together with the quantum computer task and computer type information as a quantum computing request and sends it to the back-end processing platform of the quantum cloud platform, so that the quantum cloud platform obtains the quantum computing request.
[0050] After obtaining a quantum computing request, the quantum cloud platform can compare the identity authentication information therein with the stored user permission list, for example, by using the identity authentication information to look up a table in the user permission list, and according to the table lookup result, it can be confirmed whether the user has the permission to submit a quantum computing request. When it is determined that the user has the permission, the quantum cloud platform sends the quantum computing task and computer type information to the task scheduling server. The quantum cloud platform is also used to output information for prompting the user that the user does not have the permission when it is determined that the user does not have the permission, such as by outputting and displaying relevant prompt information on the user's terminal computer screen.
[0051] After receiving the quantum computing task and computer type information, the task scheduling server first performs quantum circuit optimization and quantum circuit cutting on the quantum computing task. Quantum circuit optimization is to optimize the quantum logic gates of the quantum circuit in the quantum computing task to reduce the number of quantum bits, the number and depth of quantum logic gates, etc., such as merging or rearranging or even canceling some quantum logic gates. Quantum circuit cutting is to split the quantum circuit of the quantum computing task to obtain multiple quantum computing subtasks including quantum subcircuits, so that the quantum computing subtasks can be arranged to be executed on multiple quantum simulators and / or quantum computers, thereby improving the execution efficiency of the quantum computing task. According to the computer type information corresponding to the quantum computing task, when it characterizes that the quantum computing task is allowed to run on a quantum simulator, the quantum computing subtask after the quantum computing task is cut is assigned to the first task pool of the quantum simulator, and when it characterizes that the quantum computing task is allowed to run on a quantum computer, the quantum computing subtask after the quantum computing task is cut is assigned to the second task pool of the quantum computer.
[0052] For quantum simulators, when they are idle after completing the last quantum computing subtask, they can check the first task pool corresponding to them, and obtain the quantum computing subtask execution from it, and send the execution result to the task scheduling server after obtaining it; for quantum computers, when they are idle after completing the last quantum computing subtask, they can check the second task pool corresponding to them, and obtain the quantum computing subtask execution from it, and send the execution result to the task scheduling server after obtaining it. Different quantum simulators or quantum computers have different execution speeds for different quantum computing subtasks, so some quantum simulators or quantum computers may be idle. At this time, the quantum computing subtasks of the corresponding task pools of different quantum simulators or quantum computers can be intelligently scheduled and allocated to minimize the situation where some quantum simulators or quantum computers are idle, thereby improving the overall operating efficiency of the architecture.
[0053] After receiving the execution structure of the quantum computing subtask, the task scheduling server can obtain the calculation result of the quantum computing task according to the execution results of the quantum computing subtasks belonging to the same quantum computing task, for example, the execution results can be merged. The task scheduling server can assign the same identification code to the quantum computing subtasks of the same quantum computing task, and then identify the quantum computing subtasks belonging to the same quantum computing task according to the identification code; it can also record the quantum computer and the position of its quantum bits executed by the quantum computing subtasks of the same quantum computing task, and then identify the quantum computing subtasks belonging to the same quantum computing task according to the position of the execution result in a quantum operation result. For example, a 5-qubit quantum computing task is cut into a 3-qubit and a 2-qubit quantum computing subtask, which are respectively assigned to qubits 0-2 and qubits 9-10 of a quantum computer. Assuming that the quantum computer has a total of 11 qubits, and returns the calculation result "00100101011" after one operation, the corresponding execution results 0-2 "001" and 9-10 "11" are the execution results of their quantum computing subtasks, and then it can be determined that they belong to the same quantum computing task according to their positions.
[0054] After the task scheduling server obtains the calculation result according to the execution result, it returns the calculation result to the quantum cloud platform, and the quantum cloud platform outputs the calculation result to the corresponding user, for example, the calculation result can be displayed on the user's computer terminal screen. The task scheduling server can package the calculation result and its corresponding identity authentication information and send it to the quantum cloud platform, so that the quantum cloud platform can determine the corresponding user according to the identity authentication information.
[0055] Through the above technical solution, the quantum cloud platform sends the quantum computing tasks and computer type information submitted by authorized users to the task scheduling server. The task scheduling server optimizes and cuts the quantum circuits of the quantum computing tasks to obtain quantum computing subtasks, and assigns the quantum computing subtasks to the first task pool of the quantum simulator or the second task pool of the quantum computer according to the computer type information, so that the quantum simulator and the quantum computer can obtain the quantum computing subtasks from their respective task pools for execution, and return the execution results to the task scheduling server. The task scheduling server obtains the calculation results of the quantum computing tasks according to the execution results of the quantum computing subtasks, and returns the output to the quantum cloud platform. In this way, the management and operation of quantum computing tasks are realized through the quantum-classical hybrid computing architecture, ensuring the smooth execution of users' quantum computing tasks.
[0056] Optionally, allocating the quantum computing subtask to a second task pool of a quantum computer comprises:
[0057] Extracting quantum program features of the quantum computing subtask;
[0058] Inputting the quantum program features into a trained preset machine learning model to obtain the predicted execution time and predicted execution accuracy of quantum computing subtasks on each quantum computer;
[0059] Among quantum computers whose predicted execution accuracy is greater than or equal to a preset threshold and whose number of quantum bits is greater than or equal to the number of quantum bits required to run the quantum computing subtask, determine the quantum computer with the shortest predicted execution time, and assign the quantum computing subtask to the second task pool of the quantum computer with the shortest predicted execution time.
[0060] Specifically, when the task scheduling server allocates quantum computing subtasks between quantum computers, it considers that different quantum computers may have different characteristics, so it intelligently allocates quantum computing subtasks according to the needs and the characteristics of quantum computers. First, the quantum program features of the quantum computing subtasks are extracted. The quantum computing tasks can be written in quantum programming languages such as OpenQASM, which include descriptions of features such as quantum logic gates of quantum circuits, and then the quantum program features can be extracted. For example, in one possible implementation, the quantum program features include the number of quantum circuit bits, the number of quantum circuit measurements, the number of single quantum logic gates without parameters, the number of single quantum logic gates with parameters, the number of double quantum logic gates without parameters, the number of double quantum logic gates with parameters, the number of multi-quantum logic gates without parameters, the number of multi-quantum logic gates with parameters, the number of conditional check instructions, and the number of loop instructions.
[0061] After extracting the quantum program features, the quantum program features are input into the trained preset machine learning model for prediction, and the preset machine learning model outputs the predicted execution time and predicted execution accuracy of the quantum computing subtasks corresponding to the quantum program features on each quantum computer. For the quantum computing subtask, the predicted execution accuracy of each quantum computer output by the preset machine learning model is compared with the preset threshold, and the quantum computer corresponding to the predicted execution accuracy greater than or equal to the preset threshold is found, and the number of quantum bits required for the quantum computing subtask to run, that is, the number of quantum bits in the quantum subcircuit, is compared with the number of quantum bits possessed by the quantum computer, and the quantum computer with the number of quantum bits greater than or equal to the required number of quantum bits is found. Among the quantum computers that meet these two conditions, the quantum computer with the shortest predicted execution time of the quantum computing subtask is found, so as to assign the quantum computing subtask to the second task pool of the quantum computer.
[0062] For example, for a quantum computing subtask, after inputting the preset machine learning model, the predicted execution time, predicted execution accuracy and the number of qubits of each quantum computer of quantum computers 1 to 4 are obtained as follows: quantum computer 1: 40s, 95%, 32 qubits; quantum computer 2: 60s, 98%, 40 qubits; quantum computer 3: 55s, 98%, 6 qubits; quantum computer 4: 20s, 92%, 12 qubits. Assuming that the preset threshold is 95% and the required number of qubits is 10, the quantum computers with predicted execution accuracy greater than or equal to 95% and qubits greater than or equal to 10 are determined to be quantum computer 1 and quantum computer 2, and then the quantum computer with the shortest predicted execution time is determined to be quantum computer 1, and then the quantum computing subtask is assigned to the second task pool of quantum computer 1. It should be noted that qubits are equivalent to quantum bits.
[0063] Optionally, the preset machine learning model is any one of a multivariate linear regression model, a multi-output random forest regression model, or a multi-output gradient boosting regression model. All three models are supervised machine learning models, whose output values are continuous, and are also regression models, which can be used to output the predicted execution time and predicted execution accuracy corresponding to multiple quantum computers. In order to train the preset machine learning model, sample data and label data can be prepared in advance. The sample data can be quantum program features of different quantum computing subtasks, and the label data can be its actual execution time on the quantum computer and the actual predicted execution accuracy. 80% of the sample data and label data are retained for training, and the remaining 20% are used for testing and evaluation.
[0064] Optionally, when the second task pool is not empty, the task scheduling server is further used to map the head quantum computing subtask in the second task pool to the quantum bit of the quantum computer corresponding to the second task pool, determine the remaining number of free quantum bits of the quantum computer, and search the second task pool for other quantum computing subtasks whose required number of quantum bits is less than or equal to the number of free quantum bits, and then map the quantum computing subtask to the remaining free quantum bits of the quantum computer corresponding to the second task pool, so that the quantum computer obtains the quantum computing subtask mapped to its own quantum bit from the second task pool for execution.
[0065] Unlike classical computing, quantum computing on quantum computers requires mapping of quantum circuits, specifying specific quantum bits on the quantum computer to run quantum circuits. If only a single quantum circuit is run, when the number of quantum bits on the quantum computer is greater than the number of quantum bits on the quantum circuit, other quantum bits are idle during the operation of a single quantum circuit, which will cause idle computing resources. Therefore, consider mapping multiple quantum computing subtasks to the quantum bits of the quantum computer at the same time, and parallelizing multiple quantum computing subtasks to improve computing efficiency.
[0066] Specifically, when the second task pool is not empty, the quantum computer only executes the quantum computing subtasks from the second task pool corresponding to itself. The task scheduling server maps the head quantum computing subtask in the second task pool to the quantum bit of the corresponding quantum computer to determine the execution position of the quantum computing subtask. The head quantum computing subtask is the quantum computing subtask at the head of the second task pool in terms of execution order. After the quantum computer completes the current quantum calculation, it should execute the head quantum computing subtask in order. For example, for the second task pool in first-in-first-out order, the quantum computing subtask that enters first is the head quantum computing subtask.
[0067] After mapping, the remaining number of free qubits of the quantum computer is calculated, that is, the number of qubits of the quantum computing subtasks that have not been mapped. Then, other quantum computing subtasks that meet the requirements can be searched in order in the second task pool. When the number of qubits of a quantum computing subtask is less than or equal to the number of free qubits, it is determined that the quantum computing subtask meets the requirements, and then the quantum computing subtask can be mapped to the remaining free qubits of its quantum computer, and then the step of determining the remaining number of free qubits of the quantum computer can be returned to execute until the qubits of the quantum computer are all allocated or the number of qubits of the quantum computing subtasks in the second task pool is greater than the number of free qubits of the quantum computer. In this case, multiple quantum computing subtasks may be mapped on the quantum computer at the same time. When the quantum computer runs a calculation, these quantum computing subtasks can be executed simultaneously, which improves the computing efficiency.
[0068] For example, if there are quantum computing subtasks 1-5 in the second task pool, the number of quantum bits required are 6, 5, 5, 3, and 12 respectively. If the number of quantum bits of the quantum computer corresponding to the second task pool is 10, then after mapping the 6 quantum bits of the first quantum computing subtask 1 to the quantum computer, there are 4 free quantum bits remaining. Then continue to search in quantum computing subtasks 2-5, map quantum computing subtask 4 to the quantum computer, and there is 1 free quantum bit remaining. Continue to search for quantum computing subtasks that do not meet the conditions for mapping. Therefore, when the quantum computer performs quantum computing next time, quantum computing subtasks 1 and 4 are executed at the same time.
[0069] During the calculation process, some quantum computers may be idle while others may be busy. This will cause uneven distribution of computing resources and poor overall computing efficiency. To improve the overall computing efficiency of the architecture, the computing resources of quantum computers are automatically allocated based on the scheduling strategy.
[0070] Optionally, when the second task pool is empty, the task scheduling server schedules quantum computing subtasks from the second task pools of other non-empty quantum computers to the empty second task pool.
[0071] Specifically, when the second task pool is empty, the task scheduling server can schedule quantum computing subtasks in other non-empty second task pools to the empty second task pool, so that the originally empty second task pool becomes non-empty, and then its corresponding quantum computer can obtain quantum computing subtasks from it to execute, so as to balance the load of each quantum computer.
[0072] Similar to the case of quantum computers, in order to improve the utilization efficiency of quantum simulators, the quantum computing subtasks of the first task pool of the quantum simulator are intelligently scheduled to improve the utilization efficiency of a single quantum simulator and the computing efficiency of the entire architecture.
[0073] Optionally, when the first task pool is not empty, the task scheduling server uses the total number of quantum bits of the quantum simulator corresponding to the first task pool minus the number of quantum bits required for the head quantum computing subtask in the first task pool to obtain the remaining available quantum bits of the quantum simulator, and searches for a quantum computing subtask whose required number of quantum bits is less than or equal to the number of available quantum bits among other quantum computing subtasks in the first task pool, and determines that the quantum computing subtask is run together with the head quantum computing subtask on the quantum simulator corresponding to the first task pool.
[0074] Since quantum simulators do not have real qubits, there is no need to map quantum circuits to specific qubits in quantum simulators like in quantum computers. Instead, quantum computing subtasks that are to be executed in parallel can be arranged. When the first task pool is not empty, that is, there are quantum computing subtasks to be executed, the first quantum computing subtask is the next task to be executed by the quantum computer. Based on the resources required, that is, the number of qubits required, the remaining idle computing resources of the quantum simulator are determined, and other quantum computing subtasks are arranged to be calculated in parallel with the first quantum computing subtask.
[0075] First, the task scheduling server uses the total number of qubits of the quantum simulator to subtract the number of qubits of the team-leading quantum computing subtask to obtain the remaining available qubits, and then determines the quantum computing subtasks in the same first task pool whose qubits are less than the available qubits, and then recalculates the remaining available qubits of the quantum simulator, and returns to the step of searching for quantum computing subtasks whose required qubits are less than or equal to the available qubits, until the remaining available qubits are 0 or the number of qubits of the quantum computing subtasks in the first task pool are all greater than the available qubits of the quantum simulator. The determined quantum computing subtask and the team-leading quantum computing subtask will be scheduled to be executed on the quantum simulator together when the quantum simulator performs the next operation.
[0076] For example, if there are quantum computing subtasks ae in the first task pool, the number of quantum bits required are 6, 5, 5, 3, and 12 respectively. If the quantum bit number of the quantum simulator corresponding to the first task pool is 10, the number of quantum bits 10 minus the number of quantum bits 6 of the first quantum computing subtask a in the team is subtracted, and the remaining number of available quantum bits is 4. Then, the quantum computing subtask that meets the conditions is searched for in the first task pool as quantum computing subtask d, and the remaining number of available quantum bits can be calculated as 4-3=1. At this time, there is no quantum computing subtask that meets the conditions and can be run in parallel. Therefore, it is determined that quantum computing subtask a and quantum computing subtask d are run in parallel on the quantum simulator. The next time the quantum simulator performs an operation, quantum computing subtask a and quantum computing subtask d are operated on at the same time.
[0077] Optionally, when the first task pool is empty, the task scheduling server schedules quantum computing subtasks from the first task pools of other non-empty quantum simulators to the empty first task pool.
[0078] When the first task pool is empty, the task scheduling server can schedule quantum computing subtasks in other non-empty first task pools to the empty first task pool, so that the originally empty first task pool becomes non-empty, and then its corresponding quantum simulator can obtain quantum computing subtasks from it to execute, so as to balance the load of each quantum simulator and improve the overall computing efficiency of the architecture.
[0079] In a possible implementation, the CPU and GPU in the classical computer where the quantum simulator is located can also be used to process classical computing tasks. The task scheduling server can input the classical computing tasks into the trained machine learning model to predict the execution time of the classical computing tasks on each CPU and GPU, and then the classical computing tasks can be assigned to the CPU or GPU with the shortest execution time, thereby improving the processing efficiency of the classical computer for classical computing tasks. The machine learning model can be any one of a multivariate linear regression model, a multi-output random forest regression model, or a multi-output gradient boosting regression model. When training the machine learning model, sample data and label data can be prepared in advance. The sample data can be the classical program features of different classical computing tasks, and the label data can be the actual execution time of the tasks running on each CPU and GPU. 80% of the sample data and label data are retained for training, and the remaining 20% are used for testing and evaluation.
[0080] Optionally, the quantum computer is also used to load the interface to parse the quantum computing subtask, convert the quantum computing subtask into a quantum instruction linked list, and fill it into a designated memory space according to the instruction sequence, and the FPGA driver controls the microwave transmitter to execute the quantum circuit.
[0081] Specifically, the loading interface is responsible for receiving and processing the input of quantum computing subtasks, and converting quantum computing subtasks into a form that can be processed by the quantum chip of the quantum computer. Quantum instructions are quantum logic gate operations that act on quantum bits to implement specific quantum computing processes. Converting quantum computing subtasks into quantum instruction lists means creating an ordered list containing all quantum logic gates arranged in the order of execution. In a quantum computer, a dedicated memory space is required to store the quantum instructions to be executed. The quantum instruction list generated above is stored in the corresponding memory location for subsequent execution. FPGA (Field-Programmable Gate Array) is a programmable integrated circuit that is often used to implement high-speed and complex digital signal processing tasks. In the field of quantum computing, FPGA can be used to control operations acting on quantum bits, such as the generation and adjustment of microwave signals, and to achieve the state evolution of quantum bits by generating microwave signals of specific frequencies.
[0082] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings; however, the present disclosure is not limited to the specific details in the above embodiments. Within the technical concept of the present disclosure, a variety of simple modifications can be made to the technical solution of the present disclosure, and these simple modifications all fall within the protection scope of the present disclosure.
[0083] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.
[0084] In addition, various embodiments of the present disclosure may be arbitrarily combined, and as long as they do not violate the concept of the present disclosure, they should also be regarded as the contents disclosed by the present disclosure.
Claims
1. A quantum-classical hybrid computing architecture, It is characterized in that The architecture includes: Quantum cloud platform; A task scheduling server, connected to the quantum cloud platform for communication; A quantum simulator, connected in communication with the task scheduling server; A quantum computer, in communication with the task scheduling server; in: The quantum cloud platform is used to obtain a quantum computing request submitted by a user, wherein the quantum computing request includes a quantum computing task, identity authentication information, and computer type information that is allowed to run the quantum computing task. The quantum cloud platform is also used to determine whether the user has the authority to submit a quantum computing request based on the identity authentication information, and when it is determined that the user has the authority, send the quantum computing task and the computer type information to the task scheduling server; The task scheduling server is used to perform quantum circuit optimization and quantum circuit cutting on the quantum computing task to obtain a quantum computing subtask, and when the computer type information represents a quantum simulator, the quantum computing subtask is assigned to a first task pool of the quantum simulator, and when the computer type information represents a quantum computer, the quantum computing subtask is assigned to a second task pool of the quantum computer; The quantum simulator is used to obtain quantum computing subtasks from the first task pool for execution and return the execution results to the task scheduling server, and the quantum computer is used to obtain quantum computing subtasks from the second task pool for execution and return the execution results to the task scheduling server; The task scheduling server is also used to obtain the calculation results of the quantum computing task according to the execution results, and return the calculation results to the quantum cloud platform. The quantum cloud platform is also used to output the calculation results to the corresponding users.
2. The architecture according to claim 1, It is characterized in that The allocating the quantum computing subtasks to a second task pool of the quantum computer comprises: Extracting quantum program features of the quantum computing subtask; Inputting the quantum program features into a trained preset machine learning model to obtain the predicted execution time and predicted execution accuracy of quantum computing subtasks on each quantum computer; Among quantum computers whose predicted execution accuracy is greater than or equal to a preset threshold and whose number of quantum bits is greater than or equal to the number of quantum bits required to run the quantum computing subtask, determine the quantum computer with the shortest predicted execution time, and assign the quantum computing subtask to the second task pool of the quantum computer with the shortest predicted execution time.
3. The architecture according to claim 2, It is characterized in that The preset machine learning model is any one of a multivariate linear regression model, a multi-output random forest regression model or a multi-output gradient boosting regression model.
4. The architecture according to claim 3, It is characterized in that The quantum program features include the number of quantum circuit bits, the number of quantum circuit measurements, the number of single quantum logic gates without parameters, the number of single quantum logic gates with parameters, the number of double quantum logic gates without parameters, the number of double quantum logic gates with parameters, the number of multi-quantum logic gates without parameters, the number of multi-quantum logic gates with parameters, the number of conditional check instructions, and the number of loop instructions.
5. The architecture according to claim 1, It is characterized in that When the second task pool is not empty, the task scheduling server is further used to map the head quantum computing subtask in the second task pool to the quantum bit of the quantum computer corresponding to the second task pool, determine the remaining number of free quantum bits of the quantum computer, and search the second task pool for other quantum computing subtasks whose required number of quantum bits is less than or equal to the number of free quantum bits, and then map the quantum computing subtask to the remaining free quantum bits of the quantum computer corresponding to the second task pool, so that the quantum computer obtains the quantum computing subtask mapped to its own quantum bit from the second task pool for execution.
6. The architecture according to claim 5, It is characterized in that When the second task pool is empty, the task scheduling server schedules quantum computing subtasks from the second task pools of other non-empty quantum computers to the empty second task pool.
7. The architecture according to claim 1, It is characterized in that When the first task pool is not empty, the task scheduling server uses the total number of quantum bits of the quantum simulator corresponding to the first task pool to subtract the number of quantum bits required by the team-leading quantum computing subtask in the first task pool to obtain the remaining number of available quantum bits of the quantum simulator, and searches for a quantum computing subtask whose required number of quantum bits is less than or equal to the number of available quantum bits among other quantum computing subtasks in the first task pool, and determines that the quantum computing subtask is run together with the team-leading quantum computing subtask on the quantum simulator corresponding to the first task pool.
8. The architecture according to claim 7, It is characterized in that When the first task pool is empty, the task scheduling server schedules quantum computing subtasks from the first task pools of other non-empty quantum simulators to the empty first task pool.
9. The architecture according to claim 1, It is characterized in that The quantum computer is also used to load the interface to parse quantum computing subtasks, convert the quantum computing subtasks into quantum instruction linked lists, and fill them into the specified memory space according to the instruction sequence. The FPGA driver controls the microwave transmitter to execute the quantum circuit.
10. The architecture according to claim 1, It is characterized in that The quantum cloud platform is also used to output information prompting the user that the user does not have the authority when it is determined that the user does not have the authority.
Citation Information
Cited By
Quantum computing task processing method and device and storage medium
CN120851231A
Quantum computing task processing method and apparatus, and storage medium
CN120851231B
