Parallel world simulation resource dynamic scheduling optimization method and system

Abstract

The invention provides a parallel world simulation resource dynamic scheduling optimization method and system. The method comprises the following steps: detecting a dependency relationship between tasks in parallel world simulation resources to obtain topological data, and quantifying resource competition intensity and sequence constraint to form an associated feature matrix; a cross-dimension optical signal transmission path is constructed through dynamic change of a programmable photon array analysis matrix, and a low-conflict resource transmission channel map is formed; synchronously processing the matrix and the channel map to generate a global task execution sequence; executing a multi-world resource occupancy operation and generating a resource occupancy relation graph; and calculating a cross-dimension resource conflict probability in the atlas, and dynamically adjusting a task execution sequence to avoid a simulation deadlock risk. According to the method, nanosecond-level dynamic resolution of cross-dimension resource conflicts is realized, the simulation deadlock rate is zeroed, and the task throughput is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of dynamic scheduling optimization, in particular to a parallel world simulation resource dynamic scheduling optimization method and system. BACKGROUND

[0002] Parallel world simulation needs to synchronize and coordinate massive heterogeneous computing tasks in multidimensional dynamic environment, facing three core challenges: resource heterogeneity and dynamics (real-time fluctuations of CPU / GPU / FPGA computing units and storage, network link states, need to dynamically match task demand and resource availability); task dependency and timeliness constraints (cross-world interaction events trigger real-time changes in task dependency topology, delays in critical path tasks will cascade to global simulation progress, and critical resources need to be supplied in real time); multi-objective optimization conflicts.

[0003] The current targeted solution is an adaptive scheduling framework based on predicted execution time. This scheme collects dynamic state information of simulation computing nodes in real time, builds a prediction model combining task characteristics, estimates the time consumption of each node executing tasks; based on the prediction results, an adaptive allocation strategy is used to dynamically allocate tasks to the optimal node, and redundant resources are reserved to respond to sudden dependency changes, in order to optimize resource utilization and task completion time. However, the core deficiency is that the prediction model is highly dependent on the quality of historical task data, and is not adaptable to new types of tasks or sudden cross-world interaction events, and real-time state collection in super-large node clusters can easily cause communication storms, leading to increased scheduling delays. SUMMARY

[0004] The present application provides a parallel world simulation resource dynamic scheduling optimization method and system to solve the problems of resource conflict detection delay, cross-dimension deadlock out of control and insufficient task scheduling throughput in the prior art.

[0005] In a first aspect, the present application provides a parallel world simulation resource dynamic scheduling optimization method, comprising:

[0006] Detecting the dependency relationship between tasks in the parallel world simulation resource to obtain task dependency topology data, and quantifying the correlation feature matrix formed between tasks due to resource competition intensity and sequence constraints based on the task dependency topology data;

[0007] Analyzing the dynamic changes of the correlation feature matrix through a programmable photon array to construct a transmission path of cross-dimension optical signals, and forming a low-conflict resource transmission channel map matching the resource competition intensity according to the transmission path;

[0008] Synchronously processing the correlation feature matrix and the low-conflict resource transmission channel map to generate a global task execution sequence;

[0009] Based on the global task execution sequence, perform multi-world resource occupation operations and record the resource occupation status, and generate a multi-world resource occupation relationship graph based on the resource occupation status.

[0010] Calculate the probability of cross-dimensional resource conflicts in the multi-world resource occupancy relationship graph, and dynamically adjust the execution order of parallel world simulation resource tasks according to the probability of cross-dimensional resource conflicts to avoid the risk of parallel world simulation deadlock.

[0011] Optionally, task dependencies are detected in parallel world simulation resources to obtain task dependency topology data, and the association feature matrix formed between tasks due to resource competition intensity and order constraints is quantified based on the task dependency topology data, including:

[0012] Scan the port connection status of each task node in the parallel world simulation resources and record the physical links for data transmission between the task nodes to form task dependency topology data describing the inter-task dependencies.

[0013] Based on the task dependency topology data, the number of task nodes sharing the same resource is counted, and the number of task nodes sharing the same resource is used as a quantitative value of resource contention intensity.

[0014] The predecessor and successor tasks of the task-dependent topology data transmission are detected, and the length of the time window from the completion of the predecessor task to the start of the successor task is measured as the duration value of the sequence constraint.

[0015] The quantified value of the resource competition intensity and the duration value of the order constraint are combined into a two-dimensional vector. At the same time, the node arrangement order in the task-dependent topology data is obtained, and all two-dimensional vectors are constructed into an association feature matrix according to the node arrangement order.

[0016] Optionally, the dynamic changes of the associated feature matrix are analyzed using a programmable photon array to construct a transmission path for cross-dimensional optical signals, and a low-conflict resource transmission channel map matching the intensity of resource contention is formed based on the transmission path, including:

[0017] The programmable photonic array receives the real-time data stream of the associated feature matrix and maps the row and column coordinates of the associated feature matrix as element values ​​to the spatial light modulation unit of the programmable photonic array.

[0018] The spatial light modulation unit dynamically adjusts the phase offset of the optical signal through the element values, and maintains a positive proportional relationship between the offset and the element values ​​to form a phase distribution map;

[0019] Identify regions in the phase distribution map where the phase of the optical signal changes continuously, and connect these regions to generate a transmission path for the cross-dimensional optical signal;

[0020] The resource contention intensity value of the coverage area of ​​the transmission path is obtained, and the resource contention intensity value is converted into an optical signal intensity adjustment coefficient. At the same time, the optical signal intensity adjustment coefficient is loaded into the transmission path.

[0021] By combining transmission paths with optical signal intensity modulation coefficients, a low-collision resource transmission channel map can be constructed.

[0022] Optionally, the correlation feature matrix and the low-collision resource transmission channel map are processed synchronously to generate a global task execution sequence, including:

[0023] The resource competition intensity value and the duration value of the sequence constraint in the correlation feature matrix are combined to form the task node feature vector, and the light intensity attribute value of the transmission path of the corresponding task node in the low conflict resource transmission channel map is extracted as the channel feature vector.

[0024] The feature vector of the task node is concatenated with the feature vector of the channel to form the feature combination of the task node, and the difference degree of the feature combination between the task node and the adjacent task nodes is calculated.

[0025] The position score of the task node in the feature combination is calculated based on the difference degree of the feature combination and the light intensity attribute value, and a global task execution sequence is generated by sorting the position scores.

[0026] Optionally, the feature is that, based on the global task execution sequence, multi-world resource occupancy operations are performed and resource occupancy states are recorded, and a multi-world resource occupancy relationship graph is generated according to the resource occupancy states, including:

[0027] Based on the global task execution sequence, each task is driven to perform an acquisition operation in the multi-world resources;

[0028] Record the hardware resource identifier and time window occupied during the execution of each task to form a resource occupancy status data unit containing the hardware resource identifier and the time window occupied;

[0029] The resource occupancy status data units of each world are integrated to form a occupancy status set, and the overlap of the occupancy time windows of the same hardware resource identifier in the occupancy status sets of different worlds is detected. When the occupancy time windows of the same resource in different worlds overlap, it is marked as a resource occupancy conflict event.

[0030] Using the world as the node and the resource occupation conflict events as the connecting edges, all nodes and connecting edges are integrated to construct a multi-world resource occupation relationship graph.

[0031] Optionally, the overlap of occupancy time windows for the same hardware resource identifier in the occupancy state sets of different worlds is detected. When the occupancy time windows for the same resource overlap in different worlds, it is marked as a resource occupancy conflict event, including:

[0032] Scan resource occupancy records with the same hardware resource identifier in different world occupancy status sets, and extract the start and end timestamps of the resource occupancy time windows in different worlds.

[0033] Calculate the difference between the start timestamps of the same hardware resource identifier in different worlds, and simultaneously calculate the difference between the end timestamps. The difference between the start timestamps and the difference between the end timestamps are superimposed as the time window overlap determination value.

[0034] When the time window overlap determination value is less than the preset threshold, it is determined that the time windows for occupying the resource by different worlds overlap.

[0035] For resource occupancy records with overlapping occupancy time windows, the resource occupancy records are treated as resource occupancy conflict events.

[0036] Optionally, the probability of cross-dimensional resource conflicts occurring in the multi-world resource occupancy graph is calculated, and the execution order of parallel world simulation resource tasks is dynamically adjusted according to the probability of cross-dimensional resource conflicts to avoid the risk of deadlock in parallel world simulation, including:

[0037] Scan the connecting edges in the multi-world resource occupation relationship graph and count the number of resource occupation conflict events represented by the connecting edges as the conflict event count;

[0038] Calculate the ratio of the conflict event count to the total number of connected edges in the graph, and use the ratio as the probability of cross-dimensional resource conflict occurring;

[0039] When a resource whose occurrence probability exceeds a preset warning threshold is detected, the position of the task node corresponding to the resource in the global task execution sequence is located.

[0040] Adjust the task order of the global task execution sequence according to the location to make the probability of occurrence lower than the risk tolerance threshold, so as to avoid the risk of deadlock in parallel world simulation.

[0041] Secondly, this application provides a dynamic scheduling and optimization system for parallel world simulation resources, comprising:

[0042] The quantization module is used to detect inter-task dependencies in parallel world simulation resources to obtain task dependency topology data, and to quantify the association feature matrix formed between tasks due to resource competition intensity and sequence constraints based on the task dependency topology data.

[0043] The construction module is used to analyze the dynamic changes of the associated feature matrix through a programmable photon array to construct a transmission path for cross-dimensional optical signals, and to form a low-conflict resource transmission channel map that matches the intensity of resource contention based on the transmission path.

[0044] The processing module is used to synchronously process the correlation feature matrix and the low-conflict resource transmission channel map to generate a global task execution sequence;

[0045] The generation module is used to perform multi-world resource occupation operations and record resource occupation status based on the global task execution sequence, and generate a multi-world resource occupation relationship map based on the resource occupation status.

[0046] The adjustment module is used to calculate the probability of cross-dimensional resource conflicts in the multi-world resource occupancy relationship graph, and dynamically adjust the execution order of parallel world simulation resource tasks according to the probability of cross-dimensional resource conflicts to avoid the risk of parallel world simulation deadlock.

[0047] Thirdly, embodiments of this application provide a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a parallel world simulation resource dynamic scheduling optimization method as described in the first aspect above.

[0048] Fourthly, embodiments of this application provide a computer storage medium storing a computer program, which, when executed by a computer, implements a method for dynamic scheduling and optimization of parallel world simulation resources as described in the first aspect.

[0049] This application's embodiments achieve intelligent scheduling of parallel world simulation resources through collaborative analysis of task-dependent topology and photon transmission paths. Specifically, dynamic analysis based on the correlation feature matrix accurately quantifies resource competition relationships between tasks; low-conflict transmission channels constructed using programmable photon arrays significantly improve resource allocation efficiency; and conflict probability calculation of multi-world resource occupancy maps effectively prevents simulation deadlock risks. This method overcomes the limitations of traditional static scheduling, achieving real-time prediction and dynamic avoidance of cross-dimensional resource conflicts, ensuring the efficient and stable operation of the parallel world simulation system, and providing an intelligent solution for resource optimization of complex computing tasks.

[0050] Furthermore, precise modeling of parallel world simulation resources is achieved through task node topology analysis and resource contention quantification. Specifically, a high-fidelity task dependency network is constructed using physical link scanning based on port connection states; two-dimensional vectorization of resource contention intensity and sequence constraint duration significantly enhances the expressive power of task association features; and a matrix construction method based on node sequence arrangement provides structured input for subsequent resource scheduling. This method overcomes the limitations of traditional dependency analysis, enabling the digital representation of complex relationships between simulation tasks, laying a data foundation for the optimal allocation of parallel world resources, and effectively improving the operational efficiency and stability of large-scale simulation systems.

[0051] Furthermore, dynamic optical signal mapping for simulation resource scheduling is achieved through a programmable photonic array. Specifically, phase offset control based on spatial light modulation units precisely constructs the optical representation of the associated feature matrix; intelligent identification of phase-continuous regions significantly improves the generation efficiency of cross-dimensional transmission paths; and a dynamic matching mechanism between optical signal intensity and resource contention intensity effectively reduces the probability of channel conflicts. This method overcomes the bandwidth limitations of traditional electronic scheduling, enabling optical processing of high-concurrency resource requests, providing a low-latency, high-throughput resource transmission solution for parallel world simulations, and significantly improving the execution efficiency of complex computational tasks.

[0052] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 The flowchart of a method for dynamic scheduling and optimization of parallel world simulation resources provided in this application is shown.

[0055] Figure 2 This paper presents a schematic diagram of the structure of a parallel world simulation resource dynamic scheduling and optimization system provided in this application.

[0056] Figure 3 A schematic diagram of the structure of a computing device provided in this application is shown. Detailed Implementation

[0057] To enable those skilled in the art to better understand the present application, the technical solution of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0058] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0059] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0060] Figure 1 A flowchart of a method for dynamic scheduling and optimization of parallel world simulation resources is provided in this application embodiment, such as... Figure 1 As shown, the method includes:

[0061] Step 101: Detect inter-task dependencies in parallel world simulation resources to obtain task dependency topology data, and quantify the association feature matrix formed between tasks due to resource competition intensity and sequence constraints based on the task dependency topology data.

[0062] In this step, inter-task dependencies refer to the mutual constraints arising from shared resources or execution order limitations during task execution. Task dependency topology data is a data structure describing the connection methods between tasks, formed by analyzing inter-task dependencies.

[0063] In this embodiment, firstly, a directed graph traversal algorithm is used to detect the dependencies between tasks in the parallel world simulation resources. The system scans the input / output resource requirements of each task node and identifies the sequential execution constraints between tasks. Secondly, the detection results are converted into task dependency topology data, and an adjacency list is used to store the dependency links between task nodes, clarifying the predecessor and successor relationships. Finally, the association feature matrix formed by the intensity of resource competition and sequence constraints between tasks is quantified. This is achieved through weighted calculation: for each task pair in the dependency topology, the competition intensity of their shared resource types is analyzed, and the rigidity of the sequence constraints is superimposed to generate an N×N numerical matrix reflecting the comprehensive correlation between tasks.

[0064] For example, in a digital twin assembly workshop of an automotive OEM, the system collects simulation task data from four parallel virtual production lines in real time. Embedded probes detect that the body welding task depends on the supply status of stamped parts, while the painting task awaits the welding quality inspection results, thus constructing dependency topology data. Based on this topology, the competitive intensity of the painting robot cluster for paint spraying air pump resources is quantified, while the sequential constraints of quality inspection and assembly tasks are identified. These dynamic relationships are encoded into an association feature matrix, where the cross values ​​of the painting task row and the quality inspection task column are marked as high-intensity sequential constraints, and the competitive intensity of paint spraying air pump resources is visualized using gradient color levels.

[0065] Step 102: Analyze the dynamic changes of the associated feature matrix through a programmable photon array to construct a transmission path for cross-dimensional optical signals, and form a low-conflict resource transmission channel map that matches the intensity of resource contention based on the transmission path.

[0066] In this step, a programmable photonic array is a programmable hardware device that processes data through optical signals. Cross-dimensional optical signals refer to optical signals transmitted across different dimensions. The transmission path is the route along which the optical signal propagates within the programmable photonic array.

[0067] In this embodiment, firstly, the dynamic changes of the correlation feature matrix are analyzed using a programmable photonic array. Utilizing the optical interference characteristics of the photonic chip, the values ​​in the matrix are mapped to optical waveguide phase modulation parameters, and the matrix is ​​updated in real-time to convert into optical signal phase differences. Secondly, a cross-dimensional optical signal transmission path is constructed. Based on the phase difference distribution, a dynamic optical switch array is activated, establishing multiple wavelength-multiplexed optical paths within the photonic chip. Each optical path corresponds to the feature correlation strength of different task groups. Finally, a low-conflict resource transmission channel map matching the resource contention intensity is formed. Optical signal-to-noise ratio analysis is used to select the transmission path with the least interference. Each channel in the output map is labeled with the task group it serves and its anti-conflict level, ensuring that high-contention task groups exclusively occupy independent channels.

[0068] Building upon the aforementioned digital twin assembly line case of an automotive OEM, the programmable photonic array receives real-time updates of the associated feature matrix. When a sudden increase in resource contention for painting tasks is detected, the photonic chip dynamically reconstructs the optical signal transmission path: allocating low-latency direct optical paths for high-priority quality inspection tasks, while simultaneously establishing detour buffer channels for the painting task group. By modulating the optical phase through an interferometer array, a low-conflict resource transmission channel map matching the current contention intensity is generated—the optical path bandwidth of the painting task group is compressed to alleviate air pump pressure, ensuring that timing constraints are not violated.

[0069] Step 103: Simultaneously process the associated feature matrix and the low-conflict resource transmission channel map to generate a global task execution sequence.

[0070] In this step, the global task execution sequence is the task scheduling order generated by comprehensively analyzing the correlation feature matrix and the low-conflict resource transmission channel map.

[0071] In this embodiment, firstly, the correlation feature matrix and the low-conflict resource transmission channel map are processed synchronously, and the task correlation data in the matrix and the channel allocation scheme in the map are input into the dynamic scheduler. Secondly, a global task execution sequence is generated, and the scheduler performs a dual-rule sorting: for task groups marked "high conflict risk" in the channel map, they are arranged in descending order of constraint strength according to the order of the correlation feature matrix; for tasks sharing the same low-conflict channel, they are arranged in ascending order of resource contention strength. Finally, a global sequence is generated, which prioritizes high-dependency and high-contention tasks to occupy channels, while distributing the time positions of conflict source tasks.

[0072] Continuing with the digital twin assembly line case study from an automotive OEM, the task scheduling engine synchronously processes the process constraints in the associated feature matrix and the physical limitations of the low-conflict resource transmission channel map. When the photon map indicates an overload in the paint spraying air pump channel, the system staggers the start times of the three painting robots and inserts assembly tasks as a buffer. In the final generated global task execution sequence, highly conflicting painting tasks are distributed across three non-contiguous time slots, while quality inspection tasks are strictly scheduled to start after the baking process is completed, achieving a dual balance between resource competition and sequential constraints.

[0073] Step 104: Based on the global task execution sequence, perform multi-world resource occupation operations and record the resource occupation status, and generate a multi-world resource occupation relationship graph according to the resource occupation status.

[0074] In this step, multi-world resource occupancy operations refer to the allocation and use of resources in multiple parallel world simulation environments. Resource occupancy status describes whether a resource is occupied and the duration of its occupation. The multi-world resource occupancy relationship graph is a graph describing resource allocation generated by recording resource occupancy statuses.

[0075] In this embodiment, firstly, multi-world resource occupancy operations are performed based on the global task execution sequence, with the simulation progressing step by step according to the sequence time. Each task requests its required resource type (such as memory blocks or processor cores) upon startup. Secondly, the resource occupancy status is recorded, and the occupancy / release timestamps and associated task IDs of each physical resource are captured in real time by a resource monitor. Finally, a multi-world resource occupancy relationship graph is generated, aggregating all resource status records. A heatmap is plotted with the time axis as the vertical axis and resource type as the horizontal axis. The density of color blocks in the graph reflects the frequency of contention for the same resource by multiple tasks at the same spatiotemporal point.

[0076] For example, when executing a global sequence, the system performs multi-world resource occupancy operations on parallel virtual production lines: while the painting robot on production line 1 is locking the air pump resource, the corresponding equipment on production line 2 is simultaneously marked as occupied. The resource occupancy status is displayed in real time on the 3D topology interface—four occupancy links of different colors extend from the painting air pump node, representing the resource request queues of the four production lines respectively. In the multi-world resource occupancy relationship graph generated based on this, the painting task links of production lines 1 and 3 intersect at the air pump node, indicating a potential conflict risk.

[0077] Step 105: Calculate the probability of cross-dimensional resource conflicts in the multi-world resource occupancy relationship graph, and dynamically adjust the execution order of parallel world simulation resource tasks according to the probability of cross-dimensional resource conflicts to avoid the risk of parallel world simulation deadlock.

[0078] In this step, cross-dimensional resource conflict refers to the conflict caused by tasks competing for resources in different dimensions. Parallel world simulation deadlock risk refers to the risk that multiple parallel world simulation tasks will be unable to continue execution due to improper resource allocation.

[0079] In this embodiment, firstly, the probability of cross-dimensional resource conflicts in the multi-world resource occupancy relationship graph is calculated, and the spatiotemporal coordinates corresponding to high-density color blocks in the heatmap are statistically analyzed. A probability model is trained using historical deadlock data to predict the resource occupancy conflict rate. Secondly, the execution order of parallel world simulation resource tasks is dynamically adjusted based on the probability of cross-dimensional resource conflicts: when the conflict probability exceeds a threshold, the scheduler rearranges the global sequence in step 103, inserting high-conflict-probability tasks into low-duty-cycle time periods or migrating them to alternative channels. Finally, an iterative feedback loop is used to avoid the risk of deadlock in parallel world simulation, ensuring that there are no long-term locked states in the resource occupancy relationship graph.

[0080] For example, in the aforementioned case study of a digital twin assembly workshop at an automotive OEM, analysis of the resource usage map revealed that the painting task on production line 1 was holding an air pump awaiting quality inspection results, while the quality inspection task on production line 3 was blocked due to waiting for the air pump to be released, creating a deadlock loop across production lines. The system calculated that the probability of this conflict had reached a threshold and immediately dynamically adjusted the task sequence: interrupting the painting task on production line 1 and inserting a buffer process, forcibly releasing the air pump resources for the quality inspection task on production line 3. After the adjustment, the deadlock probability decreased, and the two production lines resumed parallel operation. Continuous monitoring showed that dynamic scheduling improved the overall efficiency of the production lines and reduced resource conflict events.

[0081] To analyze the dependencies between parallel world simulation tasks, construct task dependency topology data, quantify resource competition intensity and sequence constraint characteristics, and form an association feature matrix to reveal the interaction patterns between tasks, providing data support for subsequent resource scheduling, and achieving accurate prediction of task execution order and conflict prediction, thereby improving the resource utilization efficiency and task collaboration capability of the simulation system, as described in step 101, the process involves detecting task dependencies in parallel world simulation resources to obtain task dependency topology data, and quantifying the association feature matrix formed by resource competition intensity and sequence constraints between tasks based on the task dependency topology data, including:

[0082] Step 201: Scan the port connection status of each task node in the parallel world simulation resource and record the physical links for data transmission between the task nodes to form task dependency topology data describing the inter-task dependencies.

[0083] In this step, the port connection status of a task node refers to the interface status of the task node in communicating with other nodes in the simulation environment. The physical link is the actual connection path used for data transmission between task nodes. Task dependency topology data is a data structure describing the dependencies between tasks, formed by analyzing port connection status and physical links.

[0084] In this embodiment, firstly, the port connection status of each task node in the parallel world simulation resource is scanned using a distributed probe tool: the system sends a handshake signal to the communication port of each task node and judges the node's activity level based on the response status. Secondly, the physical links for data transmission between task nodes are recorded using network traffic mirroring technology: a packet capture proxy is deployed at the data exchange node to capture TCP / IP packets between task nodes, extract the source address, destination address, and transmission protocol, and generate a physical link connection log. Finally, to form task dependency topology data describing the dependencies between tasks, the physical link log needs to be parsed: the predecessor and successor relationships of task nodes are identified based on the data flow direction, and the directed dependencies are transformed into a graph structure stored in an adjacency list, where nodes represent tasks and edges represent dependency directions.

[0085] Step 202: Based on the task dependency topology data, count the number of task nodes sharing the same resource, and use the number of task nodes sharing the same resource as a quantitative value of resource competition intensity.

[0086] In this step, the number of task nodes sharing the same resource refers to the total number of task nodes that need to use the same resource simultaneously in the simulation environment. The quantitative value of resource contention intensity is obtained by statistically analyzing the number of task nodes sharing the same resource, which is a competition level indicator.

[0087] In this embodiment, firstly, the number of task nodes sharing the same resource is counted based on the task dependency topology data: the adjacency list of the task dependency topology data is traversed to identify all task edges pointing to the same resource node, and the total number of tasks converging to that resource is counted. Secondly, the number of task nodes sharing the same resource is used as a quantitative value of resource contention intensity. The larger this value, the more intense the resource contention.

[0088] Step 203: Detect the predecessor and successor tasks of the task-dependent topology data transmission, and measure the length of the time window from the completion of the predecessor task to the start of the successor task as the duration value of the sequence constraint.

[0089] In this step, the predecessor task is the task that must be completed before the current task. The successor task is the task that can only start after the current task is completed. The time window length is the time interval between the completion of the predecessor task and the start of the successor task. The duration of the sequence constraint is the execution interval between the predecessor and successor tasks obtained by measuring the time window length.

[0090] In this embodiment, firstly, the predecessor and successor tasks that depend on the topology data transmission are detected: the set of edges of the topology data depends on the task is traversed to identify all task pairs with direct dependencies. Secondly, the length of the time window from the completion of the predecessor task to the start of the successor task is measured: the time difference between the predecessor task's state changing to "completed" and the successor task's state changing to "started" is calculated using the task scheduler's log timestamps. Finally, this time window length is used as the duration of the sequence constraint, reflecting the forced waiting interval of the task chain.

[0091] Step 204: Combine the quantified value of the resource competition intensity with the duration value of the order constraint into a two-dimensional vector. At the same time, obtain the node arrangement order in the task-dependent topology data, and construct all two-dimensional vectors into an association feature matrix according to the node arrangement order.

[0092] In this step, the node arrangement order is the order in which the task nodes in the task-dependent topology data are arranged. The association feature matrix is ​​a task relationship matrix formed by combining all two-dimensional vectors according to the node arrangement order.

[0093] In this embodiment, firstly, the quantified value of resource contention intensity and the duration value of sequence constraints are combined into a two-dimensional vector: for each task node, its resource contention intensity value and the duration of its sequence constraints with its direct successor task are extracted to form a vector. Secondly, the node arrangement order in the task-dependent topology data is obtained: the task-dependent topology data is processed by a topology sorting algorithm to generate a linear sequence of acyclic dependent task nodes. Finally, all two-dimensional vectors are constructed into an association feature matrix according to the node arrangement order.

[0094] To leverage the dynamic modulation characteristics of programmable photonic arrays, the associated feature matrix is ​​mapped to the phase distribution of optical signals, generating cross-dimensional optical signal transmission paths and constructing a low-conflict resource transmission channel map. Resource contention intensity is visualized through optical signal intensity adjustment, reducing the risk of resource conflicts between parallel tasks. As described in step 102, the dynamic changes of the associated feature matrix are analyzed using a programmable photonic array to construct cross-dimensional optical signal transmission paths, and a low-conflict resource transmission channel map matching the resource contention intensity is formed based on these transmission paths. This includes:

[0095] Step 301: Receive the real-time data stream of the associated feature matrix using a programmable photonic array, and map the row and column coordinates of the associated feature matrix as element values ​​to the spatial light modulation unit of the programmable photonic array.

[0096] In this step, the real-time data stream refers to the dynamic input data of the correlation feature matrix in the programmable photonic array. The spatial light modulation unit is an optical element in the programmable photonic array used to control the phase and intensity of the optical signal. Element values ​​refer to the numerical values ​​corresponding to each row and column coordinate in the correlation feature matrix.

[0097] In this embodiment, firstly, a programmable photonic array is used to receive the real-time data stream of the associated feature matrix, and the electronic signal of the matrix is ​​converted into an optical signal input through a high-speed photoelectric conversion interface. Secondly, the row and column coordinates of the associated feature matrix are mapped as element values ​​to the spatial light modulation units of the programmable photonic array: each row of the matrix corresponds to a group of modulation units in the photonic array, the row and column coordinate values ​​are converted into the physical position coordinates of the modulation unit, and the element value at that coordinate is written into the control register of the modulation unit.

[0098] Step 302: The spatial light modulation unit dynamically adjusts the phase offset of the optical signal through the element value, and maintains the offset in a positive proportion to the element value to form a phase distribution map.

[0099] In this step, the phase offset of the optical signal refers to the amount of phase adjustment during the propagation of the optical signal. The phase distribution map is a phase change spectrum formed after the phase of the optical signal is adjusted by the spatial optical modulation unit.

[0100] In this embodiment, firstly, the spatial light modulation unit dynamically adjusts the phase offset of the optical signal using element values: each modulation unit generates a corresponding driving voltage based on the element values ​​in the register. This voltage changes the alignment angle of the liquid crystal molecules, thereby controlling the phase of the optical signal passing through that unit. Secondly, the offset is kept proportional to the element value: the larger the element value, the higher the driving voltage, and the larger the deflection angle of the liquid crystal molecules, resulting in a proportional increase in the phase offset of the optical signal. Finally, a phase distribution map is formed: after all modulation units operate synchronously, the photonic array output presents a light field phase distribution map, where the phase offset in different regions visually reflects the numerical distribution of the original matrix.

[0101] Step 303: Identify regions in the phase distribution map where the phase of the optical signal changes continuously, and connect the regions to generate a transmission path for the cross-dimensional optical signal.

[0102] In this step, the region where the phase of the optical signal changes continuously refers to the optical region in the phase distribution map where the phase value transitions smoothly. The transmission path of the cross-dimensional optical signal is the optical signal propagation path formed by connecting the regions where the phase changes continuously.

[0103] In this embodiment, firstly, regions where the phase of the optical signal changes continuously in the phase distribution map are identified: an optical gradient detection algorithm is used to scan the phase distribution map and mark regions where the phase value transitions smoothly. Secondly, the regions are connected to generate a cross-dimensional optical signal transmission path: the marked continuous regions are physically connected by a dynamic optical switch array, for example, a strip region where the phase gradually changes from 0° to 90° is connected into a curved optical path, forming a signal transmission channel across different dimensions of the photonic chip.

[0104] Step 304: Obtain the resource contention intensity value of the coverage area of ​​the transmission path, convert the resource contention intensity value into an optical signal intensity adjustment coefficient, and load the optical signal intensity adjustment coefficient into the transmission path.

[0105] In this step, the resource contention intensity value refers to the numerical value quantified by the number of task nodes sharing the same resource. The optical signal intensity adjustment coefficient is an optical signal power adjustment parameter calculated based on the resource contention intensity value.

[0106] In this embodiment, firstly, the resource contention intensity value of the transmission path coverage area is obtained: by tracing back the element values ​​mapped to the modulation unit in step 301, the original resource contention intensity data corresponding to the modulation units traversed by the transmission path is extracted. Secondly, the resource contention intensity value is converted into an optical signal intensity adjustment coefficient: a coefficient is generated according to a preset conversion rule, and this coefficient determines the output power of the laser source. Finally, the optical signal intensity adjustment coefficient is applied to the transmission path: the intensity of the optical signal injected into the path is adjusted by the laser controller; for example, a strong optical signal is applied to a high contention intensity path, and a weak optical signal is applied to a low contention path.

[0107] Step 305: Combine the transmission paths with the loaded optical signal intensity adjustment coefficient to construct a low-collision resource transmission channel map.

[0108] In this step, the low-collision resource transmission channel map is an optimized optical signal transmission network formed by integrating transmission paths with loaded optical signal intensity modulation coefficients.

[0109] In this embodiment, firstly, the transmission paths with loaded optical signal intensity modulation coefficients are combined: all modulated transmission paths are integrated, and optical signals of different wavelengths are allocated to independent paths using wavelength division multiplexing (WDM). Secondly, a low-collision resource transmission channel map is constructed: a map is generated based on the relationship between optical signal intensity and path topology. Each channel in the map is labeled with its service task group, optical signal intensity level, and path physical coordinates, ultimately forming a blueprint for an optical transmission network that avoids high-contact tasks sharing channels.

[0110] To integrate the dual features of the correlation feature matrix and the low-conflict resource transmission channel map, a global task execution sequence is generated by calculating the difference in feature combinations of task nodes and their location scores. This achieves synergistic optimization of task priority and resource channel characteristics, ensuring that critical tasks obtain the optimal resource path and improving the efficiency and reliability of multi-task parallel execution. As described in step 103, the correlation feature matrix and the low-conflict resource transmission channel map are processed synchronously to generate the global task execution sequence, including:

[0111] Step 401: Combine the resource competition intensity value and the duration value of the sequence constraint in the associated feature matrix into a task node feature vector, and simultaneously extract the light intensity attribute value of the transmission path of the corresponding task node in the low-conflict resource transmission channel map as the channel feature vector.

[0112] In this step, the task node feature vector is a multi-dimensional numerical representation composed of resource contention intensity values ​​and the duration values ​​of sequence constraints. The light intensity attribute value refers to the optical signal intensity parameter of the transmission path in the low-collision resource transmission channel map. The channel feature vector is a numerical representation of the task node transmission characteristics centered on the light intensity attribute value.

[0113] In this embodiment, firstly, the resource contention intensity value and the duration value of the order constraint in the association feature matrix are combined into a task node feature vector: traversing each row of the association feature matrix, the resource contention intensity value and the duration value of the order constraint corresponding to the task node are extracted and directly concatenated into a binary vector. Secondly, the light intensity attribute value of the transmission path of the corresponding task node in the low-conflict resource transmission channel map is extracted simultaneously: based on the task node identifier marked in the map, the light signal intensity level of its transmission path is located, and this level is mapped to a numerical light intensity attribute value. Finally, the light intensity attribute value is used as the channel feature vector: the channel feature vector of each task node is a one-dimensional numerical vector, stored alongside its task node feature vector.

[0114] Step 402: Concatenate the task node feature vector and the channel feature vector to form a feature combination of the task node, and calculate the feature combination difference between the task node and its neighboring task nodes.

[0115] In this step, feature combination is a comprehensive feature representation formed by concatenating the feature vectors of task nodes and the feature vectors of channels. Feature combination dissimilarity is used to quantify the degree of numerical difference in feature combinations between adjacent task nodes.

[0116] In this embodiment, firstly, the task node feature vector and the channel feature vector are concatenated to form a feature combination for the task node: for each task node, its task node feature vector (binary) and channel feature vector (univariate) are concatenated end-to-end to form a three-dimensional vector. Secondly, the feature combination difference between the task node and its neighboring task nodes is calculated: directly adjacent nodes are identified based on task-dependent topology data, and the three-dimensional feature combination vectors of the two task nodes are compared using a cosine similarity function. The similarity result is then inverted to obtain the difference value. This difference reflects the differences in resource requirements and channel allocation between nodes.

[0117] Step 403: Calculate the position score of the task node in the feature combination based on the difference degree of the feature combination and the light intensity attribute value, and generate a global task execution sequence based on the numerical value of the position score.

[0118] In this step, the location score is a task node priority evaluation value calculated based on the difference in feature combinations and light intensity attribute values. The global task execution sequence is a task scheduling order list generated by sorting the location scores.

[0119] In this embodiment, firstly, the location score of the task node is calculated based on the feature combination difference degree and the light intensity attribute value. A scoring function is designed to fuse two inputs: the feature combination difference degree is converted into a conflict risk weight, and the light intensity attribute value is superimposed to generate a comprehensive score. Secondly, a global task execution sequence is generated based on the numerical sorting of the location scores: all task nodes are arranged in descending order of location score, and nodes with the same score are then arranged in ascending order of sequence constraint duration. Finally, a global task execution sequence is generated to ensure that high-scoring tasks are executed first, and low-scoring tasks are scheduled later.

[0120] To record multi-world resource occupancy states based on the global task execution sequence, a resource occupancy relationship graph is constructed by detecting the overlap of cross-world resource occupancy time windows. This visualizes multi-world resource conflict events, providing a structured data foundation for resource conflict analysis and deadlock prevention, and ensuring the stable operation of parallel world simulations. According to step 104, based on the global task execution sequence, multi-world resource occupancy operations are performed and resource occupancy states are recorded. A multi-world resource occupancy relationship graph is then generated based on these resource occupancy states, including:

[0121] Step 501: Drive each task to perform an acquisition operation in the multi-world resources according to the order of the global task execution sequence.

[0122] In this step, multi-world resources refer to the set of computing resources that can be invoked by tasks in a parallel-world simulation environment. Occupation operation refers to the process by which tasks request and use hardware resources.

[0123] In this embodiment, firstly, each task is driven to perform an occupation operation in the multi-world resource according to the global task execution sequence: the task scheduler activates task nodes sequentially according to the sequence time steps, and initiates an exclusive hardware resource request to the multi-world resource pool through the resource request interface. Secondly, the resource allocation module reviews the request based on a preset priority strategy, and after approval, sends a locking command to the corresponding hardware resource, enabling the task to obtain exclusive resource access. Finally, the task execution engine starts the task operation process under the resource locking state, ensuring the exclusive use of the hardware resource by the task during the occupation period.

[0124] Step 502: Record the hardware resource identifier and occupation time window occupied during the execution of each task to form a resource occupation status data unit containing the hardware resource identifier and occupation time window.

[0125] In this step, the hardware resource identifier is a unique code that identifies a specific hardware resource. The occupancy time window is the start and end time range of a task's occupation of the hardware resource. The resource occupancy status data unit is a data structure that records the resource occupancy status of a single task.

[0126] In this embodiment, firstly, the hardware resource identifier and occupation time window occupied during the execution of each task are recorded: the resource monitoring agent captures the unique identifier of the hardware resource the moment the task occupies the resource, and simultaneously starts a timer to record the start timestamp of the occupation. Secondly, when the task releases the resource, the end timestamp is captured again, forming a structured record with <hardware resource identifier, start time, end time> as the fields. Finally, a resource occupancy status data unit containing the hardware resource identifier and occupation time window is formed: each record is encapsulated as an independent data unit and stored in the time-series database according to the task execution order, wherein the time window is accurate to the millisecond level and strictly bound to the hardware resource identifier.

[0127] Step 503: Integrate the resource occupancy status data units of each world to form a occupancy status set, and detect the overlap of the occupancy time windows of the same hardware resource identifier in the occupancy status sets of different worlds. When the occupancy time windows of the same resource overlap in different worlds, it is marked as a resource occupancy conflict event.

[0128] In this step, the placeholder state set is a resource usage dataset formed by integrating multiple resource placeholder state data units. Resource occupancy conflict events refer to situations where different world tasks compete for the same hardware resources at the same time.

[0129] In this embodiment, firstly, resource occupancy status data units from each world are integrated to form an occupancy status set: a cross-world data aggregator categorizes and merges all resource occupancy status data units by world number, generating a resource occupancy set by world dimension. Secondly, the overlap of occupancy time windows for the same hardware resource identifier in the occupancy status sets of different worlds is detected: the conflict detection engine traverses all world sets, comparing the occupancy time windows of the same hardware resource identifier in different worlds; if any two time windows intersect, a conflict determination is triggered. Finally, when the occupancy time windows of the same resource overlap in different worlds, it is marked as a resource occupancy conflict event: a marked record containing the conflicting resource identifier, the conflicting world pair, and the overlapping time range is generated for each conflict event.

[0130] Step 504: Using the world as the node and the resource occupation conflict event as the connecting edge, integrate all nodes and connecting edges to construct a multi-world resource occupation relationship graph.

[0131] In this step, the multi-world resource occupation relationship graph is a resource competition relationship network graph constructed with worlds as vertices and resource conflicts as edges.

[0132] In this embodiment, firstly, using worlds as nodes and resource occupation conflict events as connecting edges: the graph builder abstracts each world as a graph node, and simultaneously converts the conflict events marked in step 503 into connecting edges. Secondly, all nodes and connecting edges are integrated: all world nodes and conflict event edges are traversed, merging multiple conflict events between nodes into a single edge (avoiding duplicate edges), and adding conflict details to each edge. Finally, a multi-world resource occupation relationship graph is constructed: based on the graph database storing node and edge data, a visual relationship network is generated, where the thickness of the edges in the graph reflects the conflict frequency, and the node size reflects the total number of conflict events in that world.

[0133] To accurately identify cross-world resource occupation conflicts, the intensity of resource competition is quantified by calculating the overlap value of time windows. A conflict detection mechanism based on timestamp differences is established to achieve automated marking and classification of resource conflict events. As described in step 503, the overlap of occupation time windows for the same hardware resource identifier in the occupation state sets of different worlds is detected. When the occupation time windows for the same resource overlap in different worlds, it is marked as a resource occupation conflict event, including:

[0134] Step 601: Scan resource occupancy records with the same hardware resource identifier in different world occupancy status sets, and extract the start and end timestamps of the resource occupancy time windows in different worlds.

[0135] In this step, the resource usage record is a data entry describing the time range during which a task uses hardware resources. The start timestamp is the precise time when the task begins using hardware resources. The end timestamp is the precise time when the task ends using hardware resources.

[0136] In this embodiment, firstly, resource occupancy records with the same hardware resource identifier are scanned across different world occupancy status sets: the conflict detection engine traverses all world resource occupancy status sets and quickly locates records with the same hardware resource identifier in different worlds using hash indexes. Secondly, the start and end timestamps of the resource occupancy time windows in different worlds are extracted: for each matching record, its time window field is parsed, and the start and end timestamps are stored in temporary cache queues respectively, forming a time-series data pool grouped by resource.

[0137] Step 602: Calculate the difference between the start timestamps of the same hardware resource identifier in different worlds, and at the same time calculate the difference between the end timestamps. The difference between the start timestamps and the difference between the end timestamps are superimposed as the time window overlap determination value.

[0138] In this step, the time window overlap determination value is a numerical index obtained by superimposing the start time difference and the end time difference of different worlds occupying the same resource.

[0139] In this embodiment, firstly, the difference between the start timestamps of the same hardware resource identifier in different worlds is calculated: from the timing data pool generated in step 601, the start timestamps of any two worlds are taken, and the absolute time difference is calculated. Secondly, the difference between the end timestamps is calculated: the absolute time difference is simultaneously calculated for the end timestamps of the same world. Finally, the difference between the start timestamps and the difference between the end timestamps are superimposed as the time window overlap determination value: the two differences are directly added to generate a scalar overlap determination value, and the larger this value is, the higher the degree of time window misalignment.

[0140] Step 603: When the time window overlap determination value is less than the preset threshold, it is determined that the time windows for occupying the resource in different worlds overlap.

[0141] In this step, the preset threshold is a pre-set critical value used to determine whether time windows overlap.

[0142] In this embodiment, firstly, the preset threshold is defined by the resource conflict tolerance strategy. Secondly, when the time window overlap determination value is less than the preset threshold: the determination value output in step 602 is compared with the threshold in real time; if the determination value is less than the threshold, an overlap flag is triggered. Finally, it is determined that the time windows for resource occupation in different world pairs overlap: a Boolean result (True / False) is output to mark the conflict risk status of the resource in the world pair.

[0143] Step 604: For resource occupancy records with overlapping occupancy time windows, the resource occupancy records are treated as resource occupancy conflict events.

[0144] In this step, resource contention events are records of situations where different world missions compete for the same hardware resources at the same time.

[0145] In this embodiment, firstly, for resource occupancy records with overlapping occupancy time windows, all resource records marked as True in step 603 are filtered. Secondly, resource occupancy records are treated as resource occupancy conflict events: a structured event object is generated for each record, containing conflicting resource identifiers, conflicting world pairs, and overlapping time ranges, which are finally aggregated into a set of resource occupancy conflict events to construct conflict edges in a multi-world resource occupancy relationship graph.

[0146] To quantify the probability of multi-world resource conflicts, establish a dynamic adjustment mechanism based on risk thresholds, and optimize the global task execution sequence in real time to avoid deadlock risks, thereby achieving adaptive resource scheduling in the parallel world simulation system and ensuring stable system operation and efficient resource utilization under complex task environments, as described in step 105, the probability of cross-dimensional resource conflicts in the multi-world resource occupancy relationship graph is calculated, and the execution order of parallel world simulation resource tasks is dynamically adjusted based on the probability of cross-dimensional resource conflicts to avoid parallel world simulation deadlock risks, including:

[0147] Step 701: Scan the connecting edges in the multi-world resource occupation relationship graph and count the number of resource occupation conflict events represented by the connecting edges as the conflict event count.

[0148] In this step, connecting edges are graphical elements representing resource conflict events in a multi-world resource occupancy graph. The conflict event count is the total number of resource conflict events obtained through statistics.

[0149] In this embodiment, firstly, the connecting edges in the multi-world resource occupation graph are scanned: all edges connecting different world nodes are traversed, with each edge representing a marked resource occupation conflict event. Secondly, the number of resource occupation conflict events represented by the connecting edges is counted: a counter is maintained for each edge, and the count is incremented whenever a resource preemption occurs between the world pairs corresponding to that edge. Finally, the accumulated value is output as the conflict event count, which reflects the total conflict frequency of a specific resource across multiple worlds.

[0150] Step 702: Calculate the ratio of the conflict event count to the total number of connected edges in the graph, and use the ratio as the probability of cross-dimensional resource conflict.

[0151] In this step, the probability of cross-dimensional resource conflicts is calculated as a conflict likelihood index by the ratio of the number of conflict events to the total number of connected edges.

[0152] In this embodiment, firstly, the ratio of the conflict event count to the total number of connected edges in the graph is calculated: sum the conflict event counts of all edges (e.g., the total number of conflicts can be 50), and divide by the total number of edges in the graph (e.g., 100 edges). Secondly, the ratio is used as the probability of cross-dimensional resource conflict (e.g., 50 / 100 = 0.5), and this probability value quantifies the overall risk level of resource contention between multiple worlds.

[0153] Step 703: When a resource whose occurrence probability exceeds a preset warning threshold is detected, locate the position of the task node corresponding to the resource in the global task execution sequence.

[0154] In this step, the preset warning threshold is a pre-defined warning line for the probability of resource conflicts. Location refers to determining the specific position of the task node corresponding to the conflicting resource in the global task execution sequence.

[0155] In this embodiment, firstly, a warning threshold is preset. Secondly, resources with an occurrence probability exceeding the warning threshold are detected: all resource types are traversed, and high-risk resources with an occurrence probability greater than the threshold are filtered out. Finally, the position of the task node corresponding to the high-risk resource in the global task execution sequence is located: the resource occupancy status data unit is traced back to determine the identifier of the task node currently occupying the resource, and then its position is indexed according to the node sorting of the global task execution sequence.

[0156] Step 704: Adjust the task order of the global task execution sequence according to the position so that the occurrence probability is lower than the risk tolerance threshold, so as to avoid the risk of deadlock in parallel world simulation.

[0157] In this step, the risk tolerance threshold is the upper limit of the maximum allowed probability of resource conflicts in the system. Parallel world simulation deadlock risk refers to the possibility that multiple parallel world tasks cannot continue execution due to resource conflicts.

[0158] In this embodiment, firstly, the task order of the global task execution sequence is adjusted according to location: for task nodes associated with high-risk resources, a rescheduling algorithm is used to move them to the back of the sequence so that they are executed during low-conflict periods. Secondly, the probability of occurrence is lowered below the risk tolerance threshold: steps 701-703 are re-executed to calculate the conflict probability of the adjusted sequence until the probability value drops below the tolerance threshold. Finally, the risk of deadlock in parallel world simulation is avoided: task order optimization eliminates concurrent preemption of the same resource across multiple worlds, avoiding system-level deadlock caused by circular waiting.

[0159] Figure 2 This application provides a schematic diagram of the structure of a parallel world simulation resource dynamic scheduling and optimization system, as shown below. Figure 2 As shown, the system includes:

[0160] The quantization module 21 is used to detect the inter-task dependencies in the parallel world simulation resources to obtain task dependency topology data, and to quantify the association feature matrix formed between tasks due to resource competition intensity and sequence constraints based on the task dependency topology data.

[0161] The construction module 22 is used to analyze the dynamic changes of the associated feature matrix through a programmable photon array to construct a transmission path for cross-dimensional optical signals, and to form a low-conflict resource transmission channel map that matches the intensity of resource contention based on the transmission path.

[0162] Processing module 23 is used to synchronously process the correlation feature matrix and the low-conflict resource transmission channel map to generate a global task execution sequence;

[0163] The generation module 24 is used to perform multi-world resource occupation operations and record resource occupation status based on the global task execution sequence, and generate a multi-world resource occupation relationship map based on the resource occupation status.

[0164] The adjustment module 25 is used to calculate the probability of cross-dimensional resource conflicts in the multi-world resource occupancy relationship graph, and dynamically adjust the execution order of parallel world simulation resource tasks according to the probability of cross-dimensional resource conflicts to avoid the risk of parallel world simulation deadlock.

[0165] Figure 2 The aforementioned parallel world simulation resource dynamic scheduling and optimization system can execute... Figure 1 The implementation principle and technical effects of the parallel world simulation resource dynamic scheduling optimization method described in the illustrated embodiment will not be repeated here. The specific methods by which each module and unit of the parallel world simulation resource dynamic scheduling optimization system in the above embodiments are described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0166] In one possible design, Figure 2 The parallel world simulation resource dynamic scheduling and optimization system of the embodiment shown can be implemented as a computing device, such as... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0167] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0168] The processing component 32 is used for the above Figure 1 The embodiment describes a method for dynamic scheduling and optimization of parallel world simulation resources.

[0169] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for dynamic scheduling and optimization of resources in parallel world simulation, characterized in that, include: In parallel world simulation resources, the dependencies between tasks are detected to obtain task dependency topology data, and the association feature matrix formed between tasks due to resource competition intensity and order constraints is quantified based on the task dependency topology data. The dynamic changes of the associated feature matrix are analyzed by a programmable photon array to construct a transmission path for cross-dimensional optical signals, and a low-conflict resource transmission channel map matching the intensity of resource contention is formed based on the transmission path. The associated feature matrix and the low-conflict resource transmission channel map are processed synchronously to generate a global task execution sequence; Based on the global task execution sequence, perform multi-world resource occupation operations and record the resource occupation status, and generate a multi-world resource occupation relationship graph based on the resource occupation status. Calculate the probability of cross-dimensional resource conflicts in the multi-world resource occupancy relationship graph, and dynamically adjust the execution order of parallel world simulation resource tasks according to the probability of cross-dimensional resource conflicts to avoid the risk of parallel world simulation deadlock.

2. The method according to claim 1, characterized in that, In parallel world simulation resources, inter-task dependencies are detected to obtain task dependency topology data. Based on this task dependency topology data, the association feature matrix formed between tasks due to resource competition intensity and order constraints is quantified, including: Scan the port connection status of each task node in the parallel world simulation resources and record the physical links for data transmission between the task nodes to form task dependency topology data describing the inter-task dependencies. Based on the task dependency topology data, the number of task nodes sharing the same resource is counted, and the number of task nodes sharing the same resource is used as a quantitative value of resource contention intensity. The predecessor and successor tasks of the task-dependent topology data transmission are detected, and the length of the time window from the completion of the predecessor task to the start of the successor task is measured as the duration value of the sequence constraint. The quantified value of the resource competition intensity and the duration value of the order constraint are combined into a two-dimensional vector. At the same time, the node arrangement order in the task-dependent topology data is obtained, and all two-dimensional vectors are constructed into an association feature matrix according to the node arrangement order.

3. The method according to claim 1, characterized in that, The dynamic changes of the associated feature matrix are analyzed using a programmable photon array to construct transmission paths for cross-dimensional optical signals, and a low-conflict resource transmission channel map matching the intensity of resource contention is formed based on these transmission paths, including: The programmable photonic array receives the real-time data stream of the associated feature matrix and maps the row and column coordinates of the associated feature matrix as element values ​​to the spatial light modulation unit of the programmable photonic array. The spatial light modulation unit dynamically adjusts the phase offset of the optical signal through the element values, and maintains a positive proportional relationship between the offset and the element values ​​to form a phase distribution map; Identify regions in the phase distribution map where the phase of the optical signal changes continuously, and connect these regions to generate a transmission path for the cross-dimensional optical signal; The resource contention intensity value of the coverage area of ​​the transmission path is obtained, and the resource contention intensity value is converted into an optical signal intensity adjustment coefficient. At the same time, the optical signal intensity adjustment coefficient is loaded into the transmission path. By combining transmission paths with optical signal intensity modulation coefficients, a low-collision resource transmission channel map can be constructed.

4. The method according to claim 1, characterized in that, The associated feature matrix and the low-collision resource transmission channel map are processed synchronously to generate a global task execution sequence, including: The resource competition intensity value and the duration value of the sequence constraint in the correlation feature matrix are combined to form the task node feature vector, and the light intensity attribute value of the transmission path of the corresponding task node in the low conflict resource transmission channel map is extracted as the channel feature vector. The feature vector of the task node is concatenated with the feature vector of the channel to form the feature combination of the task node, and the difference degree of the feature combination between the task node and the adjacent task nodes is calculated. The position score of the task node in the feature combination is calculated based on the difference degree of the feature combination and the light intensity attribute value, and a global task execution sequence is generated by sorting the position scores.

5. The method according to claim 1, characterized in that, Based on the global task execution sequence, multi-world resource occupancy operations are performed and resource occupancy states are recorded. A multi-world resource occupancy relationship graph is then generated based on the resource occupancy states, including: Based on the global task execution sequence, each task is driven to perform an acquisition operation in the multi-world resources; Record the hardware resource identifier and time window occupied during the execution of each task to form a resource occupancy status data unit containing the hardware resource identifier and the time window occupied; The resource occupancy status data units of each world are integrated to form a occupancy status set, and the overlap of the occupancy time windows of the same hardware resource identifier in the occupancy status sets of different worlds is detected. When the occupancy time windows of the same resource in different worlds overlap, it is marked as a resource occupancy conflict event. Using the world as the node and the resource occupation conflict events as the connecting edges, all nodes and connecting edges are integrated to construct a multi-world resource occupation relationship graph.

6. The method according to claim 5, characterized in that, The system detects the overlap of occupancy time windows for the same hardware resource identifier in different world occupancy state sets. When the occupancy time windows for the same resource overlap in different worlds, it is marked as a resource occupancy conflict event, including: Scan resource occupancy records with the same hardware resource identifier in different world occupancy status sets, and extract the start and end timestamps of the resource occupancy time windows in different worlds. Calculate the difference between the start timestamps of the same hardware resource identifier in different worlds, and simultaneously calculate the difference between the end timestamps. The difference between the start timestamps and the difference between the end timestamps are superimposed as the time window overlap determination value. When the time window overlap determination value is less than the preset threshold, it is determined that the time windows for occupying the resource by different worlds overlap. For resource occupancy records with overlapping occupancy time windows, the resource occupancy records are treated as resource occupancy conflict events.

7. The method according to claim 1, characterized in that, Calculate the probability of cross-dimensional resource conflicts occurring in the multi-world resource occupancy graph, and dynamically adjust the execution order of parallel world simulation resource tasks based on the probability of cross-dimensional resource conflicts to avoid the risk of deadlock in parallel world simulation, including: Scan the connecting edges in the multi-world resource occupation relationship graph and count the number of resource occupation conflict events represented by the connecting edges as the conflict event count; Calculate the ratio of the conflict event count to the total number of connected edges in the graph, and use the ratio as the probability of cross-dimensional resource conflict occurring; When a resource whose occurrence probability exceeds a preset warning threshold is detected, the position of the task node corresponding to the resource in the global task execution sequence is located. Adjust the task order of the global task execution sequence according to the location to make the probability of occurrence lower than the risk tolerance threshold, so as to avoid the risk of deadlock in parallel world simulation.

8. A parallel world simulation resource dynamic scheduling and optimization system, characterized in that, include: The quantization module is used to detect inter-task dependencies in parallel world simulation resources to obtain task dependency topology data, and to quantify the association feature matrix formed between tasks due to resource competition intensity and sequence constraints based on the task dependency topology data. The construction module is used to analyze the dynamic changes of the associated feature matrix through a programmable photon array to construct a transmission path for cross-dimensional optical signals, and to form a low-conflict resource transmission channel map that matches the intensity of resource contention based on the transmission path. The processing module is used to synchronously process the correlation feature matrix and the low-conflict resource transmission channel map to generate a global task execution sequence; The generation module is used to perform multi-world resource occupation operations and record resource occupation status based on the global task execution sequence, and generate a multi-world resource occupation relationship map based on the resource occupation status. The adjustment module is used to calculate the probability of cross-dimensional resource conflicts in the multi-world resource occupancy relationship graph, and dynamically adjust the execution order of parallel world simulation resource tasks according to the probability of cross-dimensional resource conflicts to avoid the risk of parallel world simulation deadlock.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a method for dynamic scheduling and optimization of parallel world simulation resources as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The system contains a computer program that, when executed by a computer, implements a method for dynamic scheduling and optimization of parallel world simulation resources as described in any one of claims 1 to 7.

Images