100 results about "Parallel scheduling" patented technology

The invention discloses a network scanning task intelligent segmentation and load balancing method and system. The method comprises the following steps: identifying abnormal fragments by executing integrity verification on a scanning task, analyzing a task dependency relationship to determine a divisible position, and generating a segmentation scheme in combination with granularity constraint; node resource states are collected to calculate load indexes, and scheduling weights are hierarchically configured according to load levels to establish parallel scheduling channels; constructing a node cooperation group based on load difference, formulating an inter-group circulation rule to form a cooperation scheduling space, fragmenting tasks according to node capacity, and generating a parallel execution plan; monitoring the execution progress to identify overstocked nodes, diagnosing bottleneck types and planning a dredging path to execute task migration; the resource time sequence data is collected to identify the low-load time period, and the task is distributed to the optimal time slot to be executed, so that the task segmentation reasonability and the node load balancing degree are improved, and the distributed scanning execution efficiency is improved.

The invention discloses a channel coding and decoding read-write parallel scheduling method, which comprises the following steps of: inputting a base graph matrix, and sequentially arranging to obtain a read variable node sequence and a write variable node sequence of each row according to a column sequence number of a non-zero element in each row of the base graph matrix; constructing a conflict graph according to the base graph matrix; selecting the vertex with the highest saturation from the uncolored vertexes, selecting the minimum available color, and coloring the vertex; coloring each vertex, grouping the vertexes, dividing the vertexes with the same coloring color sequence number into the same group to obtain a vertex group set, and distributing a processing unit for each vertex group; constructing an operation sequence of each processing unit; conflict variable nodes of operation are eliminated; storing a reading variable node sequence and a writing variable node sequence of each operation of each processing unit; the invention further discloses a channel coding and decoding read-write parallel scheduling device. According to the technical scheme, the conflict rate can be effectively reduced, and the method can be suitable for various base graphs.

The invention provides a simulation scene multi-target calculation method and device based on parallel scheduling and a server, and relates to the technical field of industrial simulation application, and the method comprises the steps: obtaining a parameter set of a plurality of simulation tasks, and carrying out the task clustering processing of the parameter set according to the parameter similarity, and obtaining a plurality of task clusters; all the task clusters are distributed to different computing nodes, parallel solving processing of simulation tasks is carried out on the task clusters in all the computing nodes on the basis of a self-adaptive tuning model of a message passing interface, and local simulation results corresponding to all the computing nodes are obtained; and carrying out convergence and recombination processing on the local simulation results to obtain a complete simulation result data set, carrying out compression and indexed storage processing on the simulation result data set to generate a lightweight simulation result, and carrying out streaming visualization processing on the lightweight simulation result to obtain a target interactive simulation visualization result. According to the method, the simulation efficiency of simulation scene multi-target calculation can be remarkably improved.

The invention discloses a method for automatically generating a rejection defense document based on multi-agent collaboration, and belongs to the technical field of artificial intelligence. The method comprises the steps of collecting and preprocessing original multi-modal interaction data related to a payment refusing case, processing the original multi-modal interaction data into a text form to obtain an original interaction text, and storing the original interaction text in a database; performing cleaning processing on the original interaction text in the database by adopting Non-Agent to obtain semantic intermediate representation with consistent format; and based on the structured input, generating an anti-distinguishing reason by a multi-agent collaborative anti-distinguishing framework to obtain an anti-distinguishing document of the current payment refusing case. According to the invention, based on structured layering, responsibility constraint and parallel scheduling, the key defects of'hallusion caused by input noise ', 'single LLM responsibility overload' and'end-to-end time delay 'in the prior art are overcome together, so that the automatic defense signal generation system realized based on the method is remarkably improved in the aspects of compliance and fact accuracy; and practical advantages are embodied in business generalizability and engineering efficiency.

The invention provides a water conservancy model parallel scheduling calculation method, system and equipment, which are applied to a water conservancy model parallel computing scheduling system, the water conservancy model parallel computing scheduling system comprises a computing engine, a visual interaction module, a GIS map and a user interaction interface, and the water conservancy model parallel computing scheduling method comprises the following steps: obtaining a control instruction, the control instruction is sent to the calculation engine; controlling the calculation engine to interact with the visual interaction module based on the control instruction, and obtaining an interaction result; and rendering the interaction result into the GIS map. According to the method, the whole process of the computing power resources is visually displayed, the integration of an algorithm is concerned in a water conservancy model platform, construction in the computing power resources is concerned at the same time, and the computing nodes, the computing unit states, the computing process grouping information and the like are visually displayed in the computing process, so that the computing power resource construction application is assisted to be optimized and improved.

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 embodiment of the invention discloses a production change control method and device for mixed line production and a medium, and relates to the technical field of production control, and the method comprises the steps: collecting real-time production index data corresponding to a current production product in a current production line in real time under the triggering of a production change instruction issued by a production management system, obtaining a target process parameter set corresponding to the target product change product; according to the real-time production index data and the production change instruction, a production change state vector containing a production change stage label is determined, and the production change stage label comprises a production change preparation stage, a production change pre-adjustment stage and an execution stage; and based on the production change state vector and the target process parameter set, through a preset reinforcement learning agent, determining a time sequence production change action sequence, converting the time sequence production change action sequence into an equipment control instruction for instruction distribution, and realizing production change control. The sequential production change action sequence comprises a pre-adjustment action instruction, a parallel scheduling instruction and a parameter loading instruction.

The invention discloses a test sequence optimization method and device, equipment and a storage medium. Comprising the steps of obtaining function point related data, and generating a function point test sequence based on an upper confidence bound algorithm; defining a global resource occupancy matrix, and according to the function point test sequence, sequentially importing the function points, traversing and querying the global resource occupancy matrix, and generating a parallel scheduling task; and executing the parallel scheduling task, and monitoring resource conflict conditions and test failure information in real time. By comprehensively considering related data of the function points, intelligent sorting of the function points is achieved, and the problems that in a traditional sorting method, exploration and utilization are difficult to balance and global dynamic optimization is lacked are solved. By reasonably distributing parallel tasks, resource conflicts are avoided, the utilization rate of limited test resources is improved, and efficiency reduction caused by unreasonable resource planning is reduced. Resource conflicts are found and processed in time through real-time monitoring, the conflicts are prevented from influencing the test process, test failures can be quickly responded, and the test efficiency is improved.

The embodiment of the invention provides a multi-account concurrent processing management system and a multi-account concurrent processing management method, which are applied to the technical field of multi-account high-concurrency business processing, and are used for analyzing a plurality of received account operation requests into a micro-operation set, packaging micro-operations into an operation object identifier, an operation type and a dependency relationship, and storing the operation object identifier, the operation type and the dependency relationship. Comprising resource dependence, granularity dependence and business logic dependence; determining a lock execution path based on the operation object identifier, the operation type and the dependency relationship; constructing a dependency tensor model, performing conflict density mapping to form conflict density distribution, and performing tensor conflict operation on the microoperation in combination with a lock execution path to obtain a conflict intensity value; a high-conflict area is identified based on conflict density distribution, a potential deadlock graph is constructed, a micro-operation directed acyclic graph is generated, nodes with zero in-degree and out-degree are identified to form a first execution sequence, a second execution sequence is constructed according to residual micro-operations, a micro-operation execution sequence is formed through splicing, and efficient parallel scheduling and throughput capacity improvement are achieved.

The invention discloses a large model heterogeneous reasoning method and system. The method comprises the steps that target components to be unloaded and reserved components to be unloaded are determined according to calculation and storage characteristics for a large model, the target components are components which are densely stored and have the calculation strength lower than a set standard, and the reserved components are calculation-intensive components; and unloading the target component to a heterogeneous processing unit different from the GPU to execute related tasks, allocating the related tasks of the reserved component to the GPU to execute, and performing acceleration processing on the tasks executed by the heterogeneous processing unit in cooperation with a parallel scheduling strategy and a quantitative storage strategy so as to obtain a reasoning result. According to the method, the dependence on the GPU video memory and the reasoning cost are reduced, and the throughput is improved.

The invention discloses a multi-operator coarse-grained parallel scheduling method based on a heterogeneous data stream architecture, which belongs to the field of runtime systems, and comprises the following steps of: dividing a computing load of an operator with in-degree of 0 in a computing graph to be executed into a plurality of sub-tasks Task and packaging the sub-tasks into independent Task objects; constructing four priority queues which are respectively used for storing Task objects which have different preferences on four functional parts in the processing unit, putting index Task objects of operators into the corresponding priority queues according to the preferences of the operators on the functional parts and the number of instructions of the operators, and sorting the index Task objects, k (K < = 4) Tasks with complementary resource requirements are selected from the four priority queues to form a new composite operator, and the new composite operator is executed by the processing unit array; according to the method, the new operator is generated by identifying the preference of the operator function part and dynamically combining the Task, so that the utilization rate of the internal function part of the PE can be fully improved, and the operation efficiency of the whole data flow program is accelerated.

The application provides an electrical control system and parallel scheduling method of a large gear automatic welding workstation, belongs to the field of automatic welding equipment and intelligent manufacturing technology, the method controls two riveting and welding platforms to alternately perform welding, turning over and feeding and discharging through a central controller, and realizes no-waiting alternate operation by using area interlocking and pre-switching logic; combined with demonstration-free visual guidance welding and mobile base control, it ensures continuous operation of the welding robot and guarantees that multiple mobile work equipment do not interfere and collide; the central controller opens an industrial Ethernet protocol interface, collects welding parameters in real time and uploads them to an MES system, realizes data tracing and remote monitoring. The application solves the problems of low utilization rate, weak multi-variety adaptability, weak collaborative control ability and insufficient data collection and networking ability of the existing large gear welding workstation equipment.

This invention discloses a parallel optimization system for massive video processing. The system adopts a four-layer integrated parallel and collaborative processing architecture, comprising, from top to bottom: a parallel task layer for video stream access, grouping, splitting, encapsulation, and queue management; a resource abstraction layer for unified hardware modeling, status acquisition, topology construction, and capability assessment; a parallel scheduling layer for task-hardware matching, load balancing, priority scheduling, and dynamic adjustment; and an execution optimization layer for data stream localization, parallel read / write, cache optimization, and zero-copy processing. The parallel task layer, resource abstraction layer, parallel scheduling layer, and execution optimization layer form a complete parallel processing link: task input, resource awareness, accurate scheduling, and optimized execution. This invention implements concurrent rate limiting and smooth access for the input video stream to prevent traffic surges; and sets synchronization points and timing control for parallel tasks to ensure orderly output.

Apparatuses, systems, and techniques to execute CUDA programs. In at least one embodiment, an application programming interface is performed to indicate two or more blocks of threads to be scheduled in parallel.

The application provides a sparse matrix-based parallel scheduling acquisition method, system, device and medium, the method comprises the following steps: acquiring a sparse matrix; constructing an initial dependency graph according to the sparse matrix; processing the initial dependency graph through a bidirectional sub-tree to obtain a target aggregation graph; and acquiring a task dependency graph according to the target aggregation graph; the sparse matrix-based parallel scheduling acquisition method can efficiently realize the parallelization of sparse matrix calculation with a cyclic carrying dependency relationship by combining an affinity perception scheduling strategy to generate a high-performance parallel scheduling scheme through bidirectional sub-tree aggregation; the system has the same beneficial effects.

The invention provides a CPU end DLRM reasoning dynamic optimization method based on load pre-calculation, and belongs to the technical field of deep learning and computer system optimization. Comprising the following steps: step 1, dynamic core distribution: S1, constructing a heterogeneous load quantification model of an embedded search task and a Bottom-MLP task, S2, carrying out load sampling and dynamic adjustment, and S3, modifying a deep learning framework thread pool mechanism, and distributing and binding a CPU physical core for the two types of tasks to realize parallel scheduling; and step 2, dynamic prefetching: T1, determining timeliness matching and accurate prefetching principles and defining a prefetching distance as a key parameter according to embedded search three-level indirect addressing characteristics, T2, recording an iteration period, separating and aggregating calculation time and DRAM access time in an embed dingbag operator core cycle, and T3, dynamically calculating the prefetching distance and inserting a prefetching instruction to execute prefetching. According to the method, the problems of unbalanced heterogeneous task core distribution and irregular memory access prefetching failure are effectively solved, the DLRM reasoning end-to-end delay is reduced, and the CPU resource utilization rate and the cache hit rate are improved.

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 embodiment of the invention provides a large-scale low-altitude flight safety dynamic calculation parallel scheduling method and device, and is applied to the technical field of low-altitude airspace safety management. The method comprises the following steps of: gridding a low-altitude airspace based on a geographic coordinate system to generate geographic space grids, and performing grid coding on each grid; dividing each grid into a calculation area and a buffer area; the calculation area is a core operation unit and is used for real-time dynamic calculation and parallel scheduling of large-scale low-altitude flight safety, and the buffer area is a data interaction hub and is used for independently loading and caching grid data of adjacent areas; and according to the aircraft position information and the security envelope range, matching the aircraft to the calculation area to which the aircraft belongs, and routing the aircraft data to the corresponding buffer area. On the basis, inter-partition calculation decoupling can be achieved through cooperative work of the calculation area and the buffer area, the low-altitude flight safety dynamic calculation efficiency is remarkably improved, and therefore a real-time and efficient safety calculation solution is provided for a high-density low-altitude flight scene.

This invention discloses a cloud-based collaborative method and platform for scheduling and allocating resources for 3D rendering tasks, belonging to the field of data processing technology. The method includes the following steps: performing a structured analysis of the 3D rendering job to be executed, breaking it down into parallel subtasks to form asset task groups, and initially mapping them to corresponding cloud computing node sets; conducting a multi-dimensional quantitative assessment of resource impact for each asset task group during the preparation phase before entering rendering computation, obtaining a set of load risk description parameters; generating a corresponding set of resource allocation control parameters for each asset task group, and implementing concurrent access control and node preheating scheduling to form a set of warmed nodes; and performing affinity parallel scheduling processing to complete the 3D rendering task. This solves the technical problems in existing technologies where the shared channel carrying capacity during the preparation phase is not considered, leading to long GPU waiting times, delayed rendering startup, and unstable loading responses.

The invention discloses a sequence parallelization method and device of an attention module. The method comprises the following steps: receiving a block sparse attention module and hardware parameters; carrying out load division on the block sparse attention module based on the hardware parameters to generate a parallel dependency graph; performing operator-level context tiling processing on the parallel dependency graph to obtain a transformed parallel dependency graph; and carrying out operator modeling on the transformed parallel dependency graph based on preset continuous execution time of each operator to generate a CUDA flow scheduling graph. According to the method, in the calculation process of a large language model, load balanced distribution and operator-level parallel scheduling of the block sparse attention module between cross-calculation nodes and cross-graphics processor equipment can be realized. On the premise of ensuring the correctness of the calculation and communication dependency relationship, the collaborative execution efficiency of calculation and communication operation is improved, and the overall operation performance and the resource utilization rate in the multi-graphics processor cluster environment are remarkably optimized.

The invention belongs to the technical field of beverage preparation, and discloses a parallel execution control system and method of a modular beverage preparation process. Comprising the following steps: acquiring an order, and disassembling according to a manufacturing process to obtain a step node set; in the step node set, the step nodes of different orders are combined in pairs, and conflict values are obtained; establishing a first echelon during execution, establishing a first execution layer in the first echelon, and obtaining the step node with the lowest conflict degree with other step nodes as the first step node of the first execution layer; judging whether the remaining step nodes of the first echelon are classified into a first execution layer or not through the conflict value; continuously establishing a new execution layer until all nodes which do not enter the layer are stored; executing the first echelon according to an execution layer sequence, establishing an execution layer of a subsequent echelon, and triggering the execution layer sorting of the subsequent echelon according to a condition completed by step nodes; through parallel scheduling of multiple echeles and multiple execution layers driven by conflict values, the smoothness and efficiency of modular beverage preparation are improved.

This invention discloses a real-time data synchronization method based on DAG orchestration and asynchronous zero-copy architecture, comprising: constructing a directed acyclic graph (DAG) based on timestamped data and its data lineage to obtain parallelizable task nodes; sending data using a zero-copy transmission method based on io_uring based on the parallelizable task nodes to obtain zero-copy transmitted data; grouping the received data according to the source IP address and storing it in the corresponding receiving queue to obtain grouped queue data; and sorting and merging the grouped queue data according to the global timestamp using a time window multi-path merge algorithm, and using an event barrier mechanism to handle the order of data definition language instructions and data manipulation language instructions to obtain a globally ordered event stream. This invention achieves dependency-aware parallel scheduling, zero-copy asynchronous transmission, and globally ordered merging of cross-source events.

The application discloses a large number modular multiplication acceleration method and architecture based on static random access memory (SRAM) in-memory computing. According to the input multiplicand, multiplier and modulus, the modular multiplication operation is decomposed into multiple large integer multiplication operations based on the Barrett modular multiplication process. According to the components in the decomposed large integer multiplication operations as the polling operands, the components are sequentially input into the in-memory computing multiplication-addition macro array in the SRAM to perform multiplication-addition operations and generate partial sums. The partial sums are aggregated and added to obtain the execution results of the large integer multiplication operations. The execution results are subjected to a shift and truncation operation, and the results after the shift and truncation are subtracted from the modulus to obtain the modular multiplication result of the multiplicand and the multiplier with respect to the modulus. The application reduces the iteration number and resource occupation of high-bit-width modular multiplication operation through hardware reuse and parallel scheduling based on the SRAM in-memory computing and the Barrett modular multiplication process, and improves the operation efficiency and scalability.

The invention discloses a Hive data synchronization method and a Hive data synchronization device based on DataX. The method comprises the following steps: configuring hiveader plug-in parameters in a DataX framework, wherein the hiveader plug-in parameters comprise Hive connection information and an SQL (Structured Query Language) statement set; a Job.split () method is called to segment the SQL set into independent statements, and a Task instance is generated; each Task creates a temporary intermediate table in an ORC format in a pre-part () stage and executes a corresponding SQL to write a result into an HDFS (Hadoop Distributed File System); in the startRead () stage, reading temporary table data through an HDFS interface and sending the temporary table data to the Writer end; and the temporary table and the file are automatically deleted in the story () stage. According to the method, the Hive SQL native execution capacity is achieved through plug-in, complex SQL query is supported, a calculation-storage decoupling architecture is adopted, SQL-level parallel scheduling is supported, and synchronization flexibility and performance are remarkably improved.

The invention provides a self-adaptive gridding FPGA (Field Programmable Gate Array) waveform assembly line dynamic reconstruction method and system, which introduces a self-adaptive gridding algorithm, depends on modeling and safe sequence planning based on DAG (Directed Acyclic Graph), and an assembly line parallel scheduling and integrated time sequence calibration mechanism. The problems that in an existing FPGA waveform dynamic reconstruction technology, the resource fragmentation problem is caused by static partition, the system reliability risk is caused by component dependence management deficiency, and the reconstruction delay is large and the real-time performance is poor due to a serial reconstruction mode are solved. According to the method, the utilization rate of FPGA logic resources is remarkably increased, the safety and controllability of the reconstruction process are achieved, the overall reconstruction time is greatly shortened, and meanwhile the processing precision after reconstruction is guaranteed through self-adaptive time sequence calibration. And an efficient hardware dynamic reconstruction solution is provided for application scenes with extremely high requirements on real-time performance and reliability, such as software defined radio and integrated radio frequency systems.

A computer-implemented method can receive an internal representation of a transformer model, an internal representation of a device cluster, and an internal representation of a workload for execution of the transformer model on the device cluster. The method can generate a plurality of candidate execution plans based on the internal representation of the transformer model and the internal representation of the device cluster. Each candidate execution plan represents a unique parallel schedule for partitioning devices in the device cluster for parallel execution of the transformer model. The method can determine an optimal execution plan, including evaluating resource usage of the plurality of candidate execution plans based on the internal representation of the workload, and selecting, among the plurality of candidate execution plans, the optimal execution plan which yields the lowest resource usage. The evaluating includes simulating execution of the transformer model on the device cluster to process the workload.

The invention provides a computing power scheduling method and device, communication equipment, a medium and a product, and the method comprises the steps: carrying out the data preprocessing and topology awareness of a computing task, and obtaining a computing graph of the computing task; performing sub-graph division on the calculation graph to obtain calculation sub-graphs, and packaging the calculation sub-graphs into schedulable units; the calculation benefit of the schedulable unit is obtained by determining the execution delay, the power consumption and the heterogeneous coefficient of the calculation nodes which distribute the schedulable unit to different areas; constructing a task scheduling model meeting a service level target according to a maximum and minimum algorithm and the calculation benefit; the optimal allocation of the schedulable units on the computing nodes is determined through the task scheduling model, the schedulable units are scheduled to the computing nodes in parallel according to the optimal allocation, large computing tasks can be scheduled to a proper area according to delay and power consumption, the problem of hardware resource heterogeneity in the scheduling process can be solved, and the scheduling efficiency is improved. And the total power consumption of the whole task is reduced through refined scheduling.

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 database copy dynamic expansion system and method based on physical file cloning, which efficiently manages database copies in the expansion process through a physical file cloning technology and a parallel scheduling strategy of an underlying storage system, realizes elastic expansion from a single node to multiple nodes or multiple nodes to a single node of a database, and improves the expansion efficiency of the database copies. And the online service is kept uninterrupted and the data consistency is kept in the migration process.

The invention provides a cross-architecture parallel scheduling method and system for heterogeneous computing power resources. The method comprises the following steps: establishing and maintaining a resource database containing node architecture types, single-core computing efficiency and current available resources; receiving parallel computing jobs, analyzing the total computing workload of the jobs and a job completion time limit constraint set by a user, running an optimization algorithm based on the resource database, outputting a minimum core number set required by meeting the time limit constraint, and determining a corresponding candidate node group; starting a parallel process of the job on each computing node in the candidate node group, and appointing a management process on one node as a main process for coordinating the overall execution of the job; in the job execution process, the parallel process running on each node performs cross-node data exchange and synchronization through the collective communication operation of the message passing interface, and the host process is responsible for synchronous control. According to the method, seamless integration and efficient parallel computing of heterogeneous computing power resources can be realized.