94 results about "Task parallelism" patented technology

The invention provides a multi-task parallel processing method for user problems under an AI platform, and belongs to the technical field of digital data processing of the AI platform. Task resource requirements are accurately calculated through a video memory pre-estimation function and a memory pre-estimation function, a resource consumption mode of concurrent execution is analyzed through a multi-task resource prediction model; the load capacity of the system is evaluated based on a video memory utilization rate gain index, an optimal task segmentation strategy is determined through a data set splitting degree calculation function, multi-stage video memory sub-pools are constructed to realize differentiated resource management, and a Nash equilibrium point of resource allocation is solved by adopting a data set video memory allocation game model. Collaborative optimization allocation of resources is achieved through the video memory and memory coupling allocation equation set, the task state is monitored in real time in the multi-task concurrent execution process, the resource allocation weight is dynamically adjusted, and the technical problem that the system resource utilization rate is low during AI platform multi-task parallel execution is solved.

The invention provides an MCM-GPU-oriented resource storage optimization method, device and equipment, and the method comprises the steps: carrying out the static analysis of an operator-level data flow diagram of a calculation task in a compiling stage, and extracting the data dependence relation and life cycle characteristics of each operator; generating a storage optimization scheme based on the data dependency relationship and life cycle characteristics of each operator; in the operation stage, task and data collaborative allocation is carried out according to a data prefetching strategy and a data replacement strategy, task and data collaborative allocation and the calculation task are executed in parallel, and a data flow diagram and an operator dependency relationship of the task are analyzed by utilizing a compiling period; in combination with the data locality characteristics of the MCM-GPU multi-level storage architecture and the concurrency requirements of different modules for data access, an appropriate GPU module is selected to carry out data optimization management, so that frequent data exchange between a GPU and a CPU is reduced, and the performance potential of the MCM-GPU architecture in a super-large-scale task is brought into full play.

The invention discloses a CAE (Computer Aided Engineering) workflow control subsystem and method based on cloud computing, and belongs to the technical field of crossing of computer simulation and cloud computing, and the subsystem comprises a workflow configuration analysis module which is used for converting a CAE workflow configuration file possibly containing cyclic dependence into a directed acyclic graph state machine through strong connectivity component analysis; the workflow state machine operation module is used for dynamically deriving task groups which can be executed in parallel in the next stage in the state machine according to the current state; and the workflow scheduling module is used for submitting the tasks to a cloud solving platform in parallel and realizing fault recovery through state persistence. Through deep fusion of a graph theory method and a cloud computing architecture, automatic scheduling, efficient parallel execution and reliable fault tolerance of the CAE workflow are realized, and the computing resource utilization rate and the system reliability are remarkably improved.

The invention discloses a GPU cluster resource allocation method and device, equipment and a storage medium, and relates to the technical field of GPU cluster resource allocation. According to the method, a hardware performance description vector is generated by fusing static hardware parameters and dynamic micro-benchmark test data, the limitation that the GPU performance is represented only by depending on the static parameters is broken through, and the actual operation capacity of different architecture GPUs in a heterogeneous cluster is accurately matched; task feature vectors are generated by extracting task calculation features and resource demands, and quantitative description of reasoning task resource consumption features is achieved; information of hardware, tasks and load dimensions is integrated through a machine learning model, performance degradation characteristics during multi-task parallel are effectively captured, and the accuracy of execution time prediction is improved; scheduling decision operation including candidate node screening, comprehensive cost evaluation and resource reservation backfilling is executed in combination with the predicted execution time, and the task execution efficiency and cluster load balancing are both considered.

The invention provides an efficient operation and maintenance software architecture setting method for multi-task parallel scheduling, and relates to the technical field of operation and maintenance automation. The method comprises the steps that resource state data of operation and maintenance nodes in an available area are collected in real time; dynamically calculating a comprehensive load index of each available area operation and maintenance node based on the resource state data to obtain a calculation result; according to the calculation result, identifying a high-load available area with the comprehensive load index exceeding a preset threshold value and a low-load available area with the comprehensive load index lower than the preset threshold value; and based on the high-load available area, performing priority grading on operation and maintenance tasks to be processed in the high-load available area according to task types to form a high-priority core task queue and a low-priority conventional task queue. According to the invention, accurate matching and efficient utilization of cross-regional resources are realized.

The invention provides an asynchronous parallel-based service processing method, system and device and a storage medium, and the method comprises the steps: initializing a service context object according to a service type identifier in a request; a preprocessing rule script corresponding to the service type is loaded and executed through a dynamic rule engine, and parameter verification, service pre-check and context initialization are carried out; a plurality of asynchronous processing rule scripts corresponding to the service types are loaded and executed in parallel, each asynchronous processing rule script corresponds to one service task and is packaged into an asynchronous task, the asynchronous tasks are executed in parallel, and a result is written into a service context object; and loading and executing a result processing rule script corresponding to the service type through a dynamic rule engine, aggregating results of the different step tasks, and generating and returning a service response after applying a final service logic. Parallel processing can be realized through a three-stage asynchronous processing architecture, multi-core CPU resources are fully utilized, and the throughput and response speed of the system are improved.

The embodiment of the invention provides an inference task processing method and an inference server system. The method comprises the following steps: determining a plurality of calculation tasks from a plurality of reasoning tasks; configuring task labels corresponding to the operator types for the plurality of calculation tasks; a plurality of target operators are determined from a computing power unit of a computing power group module according to the task labels, parallel processing is conducted on the computing tasks through the target operators, and each target operator processes the computing tasks of the task labels corresponding to the operator type of the target operator. Through the embodiment of the invention, the problem of low reasoning efficiency of the reasoning server in the related technology can be solved, so that the effect of improving the reasoning efficiency of the reasoning server is achieved.

The invention provides an asynchronous traceable API system for a large language model performance benchmark test, and belongs to the field of natural language processing and artificial intelligence. User-defined configuration is received through an API, an asynchronous thread is adopted to execute a task, and a task state and a result are persistently stored; multi-task parallelism, state traceability and performance data integrated output are achieved, the test automation degree and efficiency of the large language model are remarkably improved, and the method is suitable for model research and development, verification and operation and maintenance scenes.

The invention provides a time window perception priority task scheduling method and system for a detection process, and relates to the field of task scheduling and intelligent optimization, and the method comprises the following steps: S1, inputting data of a to-be-scheduled task set; s2, priority scoring: based on the priority weights of the tasks and the width of the time window, calculating a comprehensive sorting factor, and performing preliminary sorting on all the tasks according to the comprehensive sorting factor to form a scheduling preliminary sorting sequence; s3, time window adjustment: on the basis of scheduling the initial arrangement sequence, detecting the task arrangement sequence one by one by utilizing a time window perception adjustment mechanism until the time window constraint requirement is met; and S4, outputting a complete scheduling result. According to the method, scheduling sequence optimization, resource conflict relief and urgent task priority response can be realized in a multi-task parallel environment on the basis of a score sorting and time window adaptive adjustment mechanism, and the method has good universality and expansibility and is suitable for intelligent scheduling optimization requirements of multiple types of businesses.

The embodiment of the invention provides a multi-node parallel acceleration method, device and equipment based on a high-performance computing platform and an RoCE network and a medium, and relates to the technical field of cluster computing, and the method comprises the following steps: constructing a communication stack based on the RoCE network, and realizing an interconnection transmission network between nodes from a sending end to a receiving end; a computing task and a communication transmission task of the high-performance computing platform are executed through a CPU core binding technology of the high-performance computing platform; based on the transparent large page function, processing a memory allocation request initiated by an application program of an operating system and a request of the application program of the operating system for accessing a memory through a virtual address; before a main computing task of the high-performance computing platform starts, pre-processing is executed through multiple threads, open type multi-processing threads are started, the main computing task is executed through the multiple open type multi-processing threads, and a next round of pre-processing task is started in parallel. According to the scheme, parallel acceleration of multiple nodes in cluster calculation is realized.

The application discloses a kind of electromechanical-electromagnetic transient hybrid simulation calculation method and system, combine Kubernetes management cluster and the actual calculation demand of electromechanical-electromagnetic transient hybrid simulation, the computing task of containing numerous computing cases is distributed to each work node, the computing task of each work node can be effectively balanced, while the computing task of single case is allocated to the same CPU, from the two angles of task parallel and sub-network parallel, efficiency is improved, the electromechanical-electromagnetic transient hybrid simulation time consumption is reduced as a whole, and the electromechanical-electromagnetic transient hybrid simulation efficiency of multiple computing cases is improved.

The invention relates to the technical field of laboratory control, in particular to a multitask parallel experiment system and method for a laboratory, and the system comprises a task management module which is used for receiving and splitting a plurality of experiment plans submitted by a user, and obtaining a plurality of experiment tasks corresponding to the experiment plans; the task constraint module is used for performing condition constraint on the experiment task according to the constraint rule to obtain a constraint experiment task with a constraint condition; the material detection module is used for acquiring material container resource information of the experiment platform, performing material container resource detection on the constraint experiment task according to the material container resource information, and generating a resource satisfaction signal; and the multi-task scheduling module is used for constructing a multi-thread scheduling strategy based on the thread pool management mechanism and the resource satisfaction signal, and scheduling the execution module to execute a plurality of experiment plans according to the multi-thread scheduling strategy. The utilization rate of experimental resources can be optimized, so that the execution efficiency of experimental tasks and the stability of experimental operation are improved.

The invention discloses a multi-AGV cooperative scheduling system for realizing load balancing. The multi-AGV cooperative scheduling system comprises a task management module, a state monitoring module, a load balancing scheduling module, an AGV cooperative scheduling module, an edge server optimization module, a task compression module, a task parallel processing module and a performance evaluation module. Wherein the load balancing scheduling module adopts a mode of combining first-stage weighted polling initial allocation and second-stage optimal allocation based on multi-agent deep reinforcement learning, the AGV collaborative scheduling module adopts an AxTD3 method to divide an Actor network into two parallel sub-networks, and joint training is carried out by sharing a Critic network. The edge server optimization module takes execution time deviation as a state, constructs a lightweight state vector in combination with a server isomerism index, and introduces federated learning and multi-Critic network architecture. According to the invention, adaptive collaborative scheduling of multiple AGVs in an edge computing environment is realized, and the system load balancing capability, the resource utilization efficiency and the task processing response speed are effectively improved.

The application discloses a public health performance evaluation task parallel scheduling method based on priority sorting, which comprises the following steps: acquiring multi-source information in a public health performance evaluation process, splitting the public health performance evaluation process into multiple evaluation task units, constructing task dependency relationship, and generating a task basic feature set; calculating the comprehensive priority of each evaluation task based on the task basic feature set; screening schedulable tasks according to the task dependency relationship, combining the priority and task relevance to generate a parallel task batch, determining the allocation relationship between the tasks in the parallel candidate task batch and the computing nodes, executing the evaluation tasks in parallel according to the allocation relationship, and obtaining the task execution result; feeding back and correcting the task scheduling parameters according to the task execution result, judging the rescheduling, and outputting the performance evaluation result by summarizing the completed tasks. The method can realize ordered parallel scheduling of tasks, improve the evaluation efficiency and resource utilization rate, and guarantee the accuracy and reliability of the evaluation result.

The application discloses a multi-granularity and multi-level computing power scheduling method in an end-side computing power network, which comprises the following steps: deploying an online deep reinforcement learning agent and a system state monitoring module on a base station side; detecting and collecting channel states and computing resources and sending them to the deep reinforcement learning agent; outputting a task segmentation strategy of each user and a scheduling decision of each subtask; a task-type device (TD) segments tasks according to the instruction and schedules the tasks to target computing nodes according to the instruction for parallel processing; the task-type device (TD) receives the computing results of all the subtasks scheduled out in parallel; and the scheduling method segments the tasks of each user in multiple granularities according to current system state information, including channel gain, idle resources provided by a resource-type device (RD) and resources allocated to each user by an edge computing server, and schedules the segmented subtasks to suitable computing nodes in parallel, so that the execution time delay of an application program is further shortened by taking advantage of parallel execution.

The invention discloses a database parallel window aggregation method based on a column storage format and a streaming execution engine. The method comprises the following steps: distributing a hash table group for a window function; dividing to-be-aggregated query result set data into a plurality of segments, and distributing the segments to a plurality of grouped tasks for concurrent execution; after all the grouping tasks are completed, a merging and sorting task is created; after the merging sorting task is completed, creating a plurality of aggregation tasks according to grouping fields of the merging hash table, and performing lock-free concurrent execution; after all aggregation tasks are completed, dividing query result set data into a plurality of segments, distributing the segments to a plurality of connection tasks, performing concurrent execution, and performing hash connection combination on the query result set data and aggregation results obtained by the aggregation tasks by the connection tasks to obtain a final aggregation result of the window function; when a plurality of window functions exist, the window functions with the same window definition multiplex the same hash table group, the window functions with different window definitions use independent hash table groups, and grouping, sorting, aggregation and connection tasks of different hash table groups are executed concurrently. According to the method, the performance of database window aggregation operation pre-sequencing and grouping can be improved, aggregation operation is divided into sub-tasks according to grouping, and the sub-tasks are executed in parallel to improve the degree of parallelism.

The invention relates to the technical field of low codes, in particular to a low-code service configuration engine and a configuration method thereof. The device comprises a relation unit execution module used for defining and executing an input-operation-output relation unit; the JSON configuration analysis module is used for analyzing a configuration file in a JSON format, the configuration file comprises version information, variable initialization configuration and task configuration, and system behaviors are defined in a declarative configuration mode; the hardware abstract interface module is used for providing an operation interface irrelevant to hardware and realizing complete abstraction of hardware operation through a callback mechanism; the backup storage module is used for storing and backing up the configuration file by adopting two storage slots, and each storage slot is provided with a configurable storage space; and the real-time multi-task execution module is used for task scheduling and execution based on RTOS and supports multi-task parallel execution, conditional branching and loop control. According to the invention, the development complexity of the industrial control system is reduced, and non-technical personnel can quickly configure complex control logic.

The invention belongs to the technical field of databases, and particularly discloses a multi-mode time series data processing method and system.The method comprises the steps that a row storage table and a column storage table are established according to a table establishing request, and an association mapping relation between the row storage table and the column storage table is established; receiving time sequence data to be inserted, extracting tag data and index data, determining a target data node based on tag data hash calculation, writing the tag data into the row storage table, and writing the index data into a target time sequence data block of the column storage table; performing partition cutting and scanning on the stored time series data based on the screening condition, and generating a query result by splitting query tasks and performing parallel execution and merging results based on a cross-mode query requirement; and determining the influenced time sequence data block based on the operation request, and updating the label data and the index data according to the type of the operation request, or deleting the marked invalid data. According to the method and the device, low-redundancy storage, high-concurrency read-write and quick query analysis of the time series data can be realized.

This disclosure relates to a method for performing task scheduling and related products, wherein the related products include a task scheduler, an artificial intelligence processor, a device, a board, and a computer-readable storage medium. The device may be included in a computing processing unit of a combined processing apparatus, which may include one or more data processing units. The aforementioned combined processing apparatus may also include interface devices and other processing units. The computing processing unit interacts with other processing units to jointly complete user-specified computational operations. The combined processing apparatus may also include a storage device connected to the device and other processing units respectively, for storing data from the device and other processing units. The solution of this disclosure can optimize scheduling operations and achieve parallel processing of tasks in multi-tasking scenarios.

The embodiments of the present specification provide a batch task processing method and a user behavior point query method, comprising distributing tasks to corresponding computing clusters and disaster recovery clusters according to the identification interval of the identification information of each task, partition scheduling the batch tasks through multiple computing clusters, and synchronously scheduling and processing the tasks by using the computing clusters and the disaster recovery clusters, thereby improving the parallel capability and processing efficiency of the tasks. Moreover, through odd and even task fragmentation processing of each identification interval, the computing clusters and the disaster recovery clusters can schedule and process different task fragments respectively, thereby avoiding resource waste and data conflicts caused by scheduling and processing the same task by both, ensuring load balancing of the computing clusters and the disaster recovery clusters, and additionally, without the need to reconfigure the task fragments of the computing clusters and the disaster recovery clusters in the case of identification interval configuration changes, thereby improving the stability and efficiency of batch task processing.

The present application belongs to the technical field of heterogeneous processor machine code compilation, and particularly relates to a multi-task parallel heterogeneous processor machine code compilation generation method. The method comprises the following steps: defining a loadable heterogeneous task format based on a protobuf protocol and defining a device task format in a self-defined manner to describe binary machine code for task execution; parsing a model compiled by an MLIR compiler through an MLIR parsing module to extract model structure, weight parameters and task dependency compilation information required for processing unit execution; mapping and generating a binary file containing complete register configuration for describing task execution through processing of an MLIR compiler backend; and constructing a task dependency graph based on the producer-consumer dependency relationship between tasks. The present application realizes multi-task parallel scheduling and resource optimization, thereby reducing redundant occupation of computing resources.

The invention discloses a workflow decision-making system based on LLM, multiple agents and vectorization retrieval, and belongs to the technical field of big data processing. The system comprises a data source layer, a data processing and filtering layer, an intelligent agent engine layer and an application layer which work cooperatively in sequence, a data purification-intelligent scheduling-deep mining three-layer cooperative computing architecture is constructed, specifically, the data source layer is used for providing multi-source heterogeneous data, the data processing and filtering layer is used for providing multi-source heterogeneous data, and the intelligent agent engine layer is used for providing multi-source heterogeneous data. The data source layer is used for executing intelligent preprocessing and mixed filtering operation on data input by the data source layer, the agent engine layer is a multi-agent collaborative execution engine, the application layer is used for receiving an analysis result of the agent engine layer, and the system adopts a mass data processing mechanism which is hierarchically isolated through context and comprises data hierarchical storage and task isolation analysis. Incremental analysis and multi-task parallelism are achieved, and professional knowledge requirements of different service scenes are met through a dynamic modular cue word mechanism.

This invention discloses a data collaborative processing system and method based on multi-business data. The system includes: a thematic data product management module for binding specified data sources to business themes and creating report templates for those themes using the specified data sources, grouping dimensions, and statistical dimensions; a dual-path distribution module for generating a copy of the report template that supports individual modification for each authorized user; distributing conditional rules for the business theme to all authorized users; these conditional rules restrict operations on the report template content; and a conditional rendering module for providing an interactive interface based on the report template copy and the conditional rules for the business theme; receiving report generation requests input by authorized users through the interactive interface and sending these requests to an asynchronous task engine module; and the asynchronous task engine module for generating multiple reports in parallel through asynchronous tasks. This invention avoids data cross-contamination and version conflicts, improving the accuracy of data acquisition.

The application relates to a data processing method and device, computer equipment and a storage medium. The method comprises the following steps: obtaining a target task, obtaining more than one data set corresponding to the target task in parallel, determining an instruction stream corresponding to the target task, the instruction stream comprising more than one target instruction triggering sequence determination, triggering various target instructions in the triggering sequence in a loop, and taking each data set in the more than one data set as a triggering object in sequence when each target instruction is triggered, until an operation result corresponding to the target task is obtained by executing the triggered target instructions. The method can improve the data processing efficiency.

The application discloses an operation and maintenance command processing method and device, equipment, a storage medium and a program product, relates to the technical field of server operation and maintenance, and comprises the following steps: converting an operation and maintenance command into a node in an operation and maintenance command operation behavior graph, recording the association relationship between metadata in the graph, reducing the dependence on manual experience, and avoiding configuration errors, improper command sequences and the like through structured graph analysis; a risk prediction model based on a graph neural network can utilize the information of operation and maintenance command metadata and the association relationship in the graph to perform accurate risk probability calculation on a command to be executed in a dynamic and complex environment, break through the limitations of traditional static rules; a safety control strategy is determined through the risk probability, real-time risk assessment and intervention suggestions can be provided before the command is executed, system interruption, data anomaly and the like can be effectively prevented, and the dynamic operation and maintenance environment of multiple platforms, multiple roles and multiple tasks in parallel is adapted.

The invention provides a flexible lamp strip virtual debugging method and system based on digital twinning, and belongs to the technical field of digital twinning. The method comprises the steps of dynamically adjusting the resource occupancy rate of a computing unit and generating rendering configuration by judging if the fluctuation amplitude of a predicted frame rate sequence exceeds a preset threshold value, balancing the large-scale array debugging computing power demand and the debugging resource occupancy of a single lamp strip, and avoiding the waste of computing resources; by monitoring the ratio of the viewport resolution rendering data volume to the preset frame buffer area capacity, the overflow risk of the buffer area is avoided; by determining a rendering task priority execution sequence, multi-task parallel conflicts are coordinated. Therefore, dynamic allocation of virtual resources, multi-task ordered scheduling and accurate parameter management are realized, the resource utilization rate is remarkably improved, rendering delay is reduced, the debugging period is shortened, and the consistency and delivery quality of flexible lamp strip products are guaranteed.

The invention relates to an industrial-grade general intelligent test hardware platform, which is characterized in that a traditional fixed function device is split into a plurality of independent detachable and replaceable function modules through modular design, so that the test requirements of various types of signals are flexibly met, free combination and emerging signal expansion are supported, and the problems of single function and expansibility are solved; a unified interface and a protocol adaptation layer are defined, one-button intelligent wiring is matched, the wiring process is simplified, manual errors are avoided, and the problems of interface disorder and operation complexity are solved; test parameters are automatically optimized based on an embedded intelligent control unit and an adaptive algorithm, multi-task parallel processing and real-time data analysis are supported, and efficiency and precision are greatly improved; a hierarchical open software system is constructed, a unified drive API is provided, multi-platform access, user-defined extension and cloud collaboration are supported, the limitation of closed software is broken, and the personalized and long-term development requirements of enterprises are met. The invention is suitable for the technical field of embedded product test systems.

The invention discloses a data governance method and device, equipment, a storage medium and a computer program product, which are used for solving the problems that an existing data governance scheme cannot dynamically adjust a governance process, the execution efficiency of each stage is low, and parallel execution of governance tasks is difficult to realize. The method comprises the following steps: constructing a data governance graph structure according to acquired configuration information; determining a data governance task corresponding to each processing node in the data governance graph structure, and determining a dependency relationship among the data governance tasks according to edges in the data governance graph structure; according to the dependency relationship, performing task arrangement on each data governance task, and generating a governance task scheduling strategy; and executing each data governance task in an asynchronous parallel mode according to a governance task scheduling strategy to obtain a task processing result, and storing the task processing result and the corresponding data governance stage in a shared storage area in an associated manner.

The invention discloses an efficient large-scale distributed parallel computing deduction simulation platform and a design method thereof, belongs to the field of simulation, and solves the problem of how to support large-scale real-time human-in-the-loop deduction simulation. The efficient large-scale distributed parallel computing deduction simulation platform is composed of a reusable model development service module, a parallel simulation scheduling service module, a large-scale parallel computing engine module, a high-throughput data acquisition service module and a scenario situation service module. According to the method, multi-task parallel large-scale human-in-loop entity deduction simulation can be effectively supported, and meanwhile, mass simulation data can be efficiently and rapidly collected, processed, stored and visually displayed in a pushing situation.

The invention provides a security operation and maintenance processing method, device and equipment based on multi-agent cooperation, and the method comprises the steps: generating a task analysis result through the execution intention recognition and entity extraction of a security operation and maintenance request, disassembling a security operation and maintenance task into a plurality of sub-tasks with a dependency relationship according to the task analysis result, and distributing corresponding agents. An inference model is selected based on the complexity score difference of each sub-task to optimize resource allocation, meanwhile, the sub-task execution result and the intermediate analysis result of each agent are managed in a unified manner through sharing a cognitive state, and a conflict detection and resolution process is executed after a candidate conclusion is generated through multi-agent collaborative inference. Therefore, the consistency of a final output conclusion is ensured while the multi-step task parallel processing efficiency is improved.