284 results about "Thread count" patented technology

The embodiment of the invention provides a thread pool operation state monitoring method and device, and the method comprises the steps: dividing the fluctuation range of the ratio of the number of active threads to the number of core threads into a high-load interval, a normal interval and a low-load interval, and calculating the optimization coefficient of the core threads based on the interval distribution characteristics; calculating the variance of the ratio of the total thread count to the maximum thread count, and extracting the adjustment coefficient of the maximum thread through the fluctuation characteristics of a variance curve; taking a difference value between the task growth rate and the processing rate as a calibration factor of the queue capacity; respectively comparing the evaluation results of the three parameters with a preset reasonable interval boundary to generate a parameter rationality score; counting the performance improvement effect of each preset adjustment strategy from a historical adjustment record; calculating a current adjustment step length based on the reference strategy; recording the adjustment suggestions and the calculation basis thereof into a version library, and setting corresponding effective conditions; according to the application, the efficiency and accuracy of monitoring the running state of the thread pool can be improved.

The invention discloses a mobile terminal offline bidirectional data synchronization method and system, and relates to the technical field of data synchronization processing, and the method comprises the following steps: initializing a whole table synchronization mechanism, and establishing a file mapping structure; a user performs data operation in an offline state, records data operation details and stores the data operation details in a local cache; when the network connection is recovered, a whole table synchronization mechanism is started, data compression and partitioning are performed, a synchronization request is sent to a server, and segmented data transmission is performed; after receiving the data, the server performs data storage and management by using a file mapping mechanism; detecting potential conflicts by utilizing a multi-version parallel management mechanism, carrying out conflict resolution, merging data with local data after the conflict resolution, and confirming that synchronization is completed to a server; after it is confirmed that all data are synchronized, local and global timestamps are updated, and meanwhile, the optimal thread count is dynamically determined by adopting a multi-thread technology. According to the invention, the efficiency, reliability and accuracy of data synchronization of the mobile terminal are remarkably improved.

Systems and methods for optimizing thread allocation in a model serving system include estimating a batch size for inference requests. An optimal configuration is then determined that defines a number of inference instances, a number of threads per inference instance, and a sub-batch size per inference instance for processing a batch of inference requests of the batch size using intra-operator parallelism that minimizes average per-batch latency. The optimal configuration is determined with reference to a plurality of predetermined model profiles that define single-inference average batch latencies for different combinations of thread counts and batch sizes, the predetermined model profiles being used as input to a dynamic programming algorithm that identifies optimal configurations that minimize the average per-batch latency.

The invention relates to a lightweight asynchronous message processing method and system for a high-concurrency scene, and the method comprises the steps: receiving a message request of a client, carrying out the preprocessing of the message request, and generating a to-be-processed message; distributing the to-be-processed message to a multi-level priority queue based on a priority scheduling mechanism; dynamically adjusting the number of core threads and the maximum number of threads of the thread pool according to the resource state data, and forming a thread pool resource configuration scheme of the current scheduling period; dynamically calculating the message processing quantity of the current scheduling batch based on the message accumulation quantity of each priority queue, the thread pool resource configuration scheme, the CPU utilization rate in the resource state data and the JVM heap memory utilization rate; and based on the out-of-heap memory buffer area reference, assigning the to-be-processed message object to a corresponding processing thread to execute a service processing logic until the processing operation of the messages in all the scheduling units is completed. The method has the effect of improving the resource adaptability of the asynchronous message processing architecture.

The invention provides a high-speed NTT method and system based on GPU matrix-thread collaborative optimization. The method comprises the steps that an input twiddle factor is converted into a matrix form; transmitting the updated twiddle factor matrix and the input array to a graphics processor; according to the input array and the size of the twiddle factor matrix, the number of blocks in the grid and the number of threads included in each block are set, so that each block is responsible for the operation of multiplying one row by one column, and each thread in the block completes the operation of multiplying one pair of data in the corresponding row-by-column process to obtain a group of intermediate results; and obtaining output data corresponding to the block according to an intermediate result in the block, and sending the output data to a corresponding position of an output matrix to serve as a multiplication result. According to the method, NTT calculation is reconstructed into parallel matrix-vector operation, the maximum utilization of GPU calculation resources is realized through the combination of an efficient thread allocation strategy and an optimized memory access mode, and the calculation efficiency is remarkably improved.

The invention provides a method and system for dynamically exporting Excel based on asynchronous streaming block data, and belongs to the technical field of large data volume export. Comprising the following steps: receiving a user export request, and generating a unique request ID; an export task is generated based on the export request, a task ID is returned, and an asynchronous response is made; submitting the task to a thread pool queue, and obtaining block data from a database in batches through streaming query; wherein the core thread count and the maximum thread count of the thread pool are adjusted based on a dynamic resource allocation method, and load change is smoothed in combination with a sliding window algorithm; writing the block data into an Excel file batch by batch by adopting a streaming writing technology; and after the task is completed, pushing a notification to the user side. According to the method, the rapidity, the availability and the safety during data export can be ensured when large data volume export is performed.

The invention discloses a cloud hard disk data backup and recovery method and device, electronic equipment and a storage medium, and relates to the technical field of cloud computing. A cloud hard disk is segmented into a plurality of segmentation blocks and a multi-thread parallel processing mode is adopted, a backup and recovery task of a large-capacity cloud hard disk is decomposed into a plurality of sub-tasks to be processed at the same time, and the data backup and recovery efficiency is improved. The operation time is greatly shortened, the number of threads in the thread pool is determined according to the number of the segmentation blocks and the current available computing resources, the utilization rate of the computing resources can be improved, cloud hard disk segmentation strategy parameters can be flexibly configured according to service requirements and cloud hard disk characteristics, data protection requirements in different scenes are met, and the user experience is improved. Consistency verification is carried out on the transmitted data, so that the integrity and consistency of the data in the backup and recovery process are ensured, the data loss risk is reduced, and the purposes of improving the backup and recovery efficiency of the cloud hard disk and enhancing the system reliability and the computing resource utilization rate are achieved.

The invention discloses an inter-thread operand collection strategy and an application system, and relates to the technical field of cross-thread data sharing of a GPGPU (General Purpose Graphics Processing Unit). In order to solve the problems of read-write delay and cache occupation caused by synchronous operation of the GPGPU in a data sharing scene, the scheme adopted by the invention comprises a decoding unit, a transmitting unit and a write-back unit, and the three units cooperate with one another to form a complete cross-thread data read-write path. Wherein the decoding unit is responsible for decoding a self-defined instruction and remapping a thread index; the transmitting unit undertakes the tasks of instruction caching, data synchronization control and cross-thread data reading; and the write-back unit is responsible for writing an instruction execution result back to the storage space and updating the state of a related register. According to the method, the parallel performance of the GPGPU in a multi-thread parallel computing scene can be improved.

The invention relates to the technical field of data management, in particular to a real-time system synchronous data management system based on multi-thread processing, which comprises a thread data dependency sensing modeling module, a data processing module, a data processing module and a data processing module, and is characterized in that the thread data dependency sensing modeling module is used for constructing an implicit dependency graph among threads and generating a dependency node table and an access conflict weight matrix; the dynamic master-slave role switching module is used for outputting a main thread guide table and a synchronous window configuration table; according to the synchronous track reconstruction execution module, a master synchronous thread initiates a data consistency instruction according to the synchronous window configuration table, a slave thread executes data playback and state merging operation according to a track deviation correction strategy in a latest master thread guide table, and a unified synchronous data state diagram is output. According to the method, potential conflict paths can be dynamically found, the conflict identification precision and coverage capability in an asynchronous access mode are improved, and the method is particularly suitable for a real-time system which is high in concurrency, mixed in read-write and complex in access relation.

The invention relates to the related technical field of equipment life cycle management, in particular to an aviation equipment full life cycle management method and system based on a digital thread. Determining main data of a first digital thread and a first-layer associated test sequence and main data of a second digital thread and a second-layer associated test sequence, mapping the deviation between a physical equipment state and a virtual simulation state, and identifying a performance difference point under a cross-stage data association link; and setting a management strategy for collaborative optimization. The technical problem that the reliability of aeronautical equipment is limited due to the fact that a cross-stage performance difference point equipment performance fluctuation root positioning fuzzy cannot be recognized in real time is solved, first and second digital thread main data and an associated test sequence are constructed, the deviation between physical equipment and a virtual simulation state is accurately mapped, and the reliability of the aeronautical equipment is improved. And a management strategy is set in combination with dual-thread data and performance difference points, and full-life-cycle collaborative optimization is performed, so that the technical effect of equipment reliability is effectively improved.

The embodiment of the invention provides a read-write test method and device for a memory and a storage medium, and belongs to the technical field of memories. The method comprises the following steps: acquiring the number of data blocks of a target memory to be tested and the line number of each data block; determining the number of to-be-created test threads according to the number of the data blocks, the number of the rows and the read-write type; creating the number of test threads, and respectively determining a read-write initial position of each test thread from the row region of each data block; and synchronously calling each created test thread, and outputting a test result, so as to synchronously perform a read-write test from the corresponding read-write initial position through each test thread. The embodiment of the invention can improve the test efficiency.

The invention discloses a processor, a chip, equipment and an instruction execution method. The processor comprises a shared memory for storing first quantity information, second quantity information, third quantity information and fourth quantity information; the first number information is the total number of threads in the working group, the second number information is the number of threads waiting to be synchronized on the N instruction branches, the third number information is the number of threads not needing to be synchronized, and the fourth number information is the number of finished threads. And the synchronization instruction assembly line is used for calculating the sum of the thread numbers of the second, third and fourth number information, and when the sum of the thread numbers is equal to the total number of the threads, unfinished threads in the working group are triggered to execute an instruction behind the rendezvous point. According to the scheme, under the condition that the sum of the thread numbers is equal to the total thread number, the synchronization instruction assembly line can trigger the threads which are not finished in the working group to continue to execute subsequent instructions, and the problem of task deadlock caused by the fact that part of the threads cannot reach the synchronization point is avoided.

Disclosed in the present application are data processing methods, an electronic device, a storage medium and a computer program product. A method comprises: acquiring data to be processed, said data being data obtained after data sequence length padding is performed on original input data; on the basis of pre-recorded data sequence length information, selecting the original input data from the data to be processed, the data sequence length information being used for recording an initial sequence length corresponding to the original input data; and calling a plurality of threads on a preset type processor to perform target computing on the original input data, so as to obtain a target computing result, the thread number of the plurality of threads being determined by means of the data sequence length information. The present application solves the technical problems in the related art of high consumption of computing resources and low training efficiency caused by performing redundant attention computing on data matrices constructed by padding operations.

The invention provides a system operation lagging construction method and device, equipment, a medium and a program product. Relates to the technical field of system performance testing. The method comprises the steps that the CPU thread number and a watchdog threshold value of a current system are obtained; on the basis of the CPU thread number and the watchdog threshold value, the duration of the target thread executing busy waiting operation is calculated through a preset formula; initializing a spin lock of a kernel mode, and creating target threads of which the number is the same as that of the CPU threads, so that the target threads execute preset processing logic to construct system operation jamming; the preset processing logic comprises the following steps: periodically obtaining a spin lock; responding to the situation that the spin lock is not obtained, and continuing to execute the spin lock obtaining operation; and in response to obtaining the spin lock, executing a busy waiting operation. According to the method, CPU resources are continuously occupied through kernel spin lock competition, and a real system jamming scene is simulated. The busy waiting operation duration is dynamically calculated, it is ensured that the single-thread lock holding time is slightly lower than a watchdog threshold value, and crash of a system triggering unrecoverable is avoided.

The invention provides a concurrent scheduling and fault-tolerant optimization method and a concurrent scheduling and fault-tolerant optimization system for distributed jobs of a Spark application layer. The method comprises the following steps: acquiring a scheduling parameter, and analyzing the parameter to generate a configuration object; dividing data in a starting and ending date range according to business logic based on the configuration object, and generating a discrete data fragment list; based on the discrete data fragment list and the concurrent thread count, creating a thread pool, packaging each discrete data fragment into an independent task, and storing the independent task into a to-be-processed fragment queue; obtaining discrete data fragments through a thread pool working thread, calling SparkSession to execute service processing logic, and submitting a job to a cluster to obtain an execution result; performing judgment processing on the execution result, and recording a task processing state; based on all processing states, an execution report is generated. According to the method, the execution efficiency and the resource utilization rate of the Spark application are remarkably improved.

The invention discloses a storage resource allocation method and device, a computing device and a storage medium. The storage resource allocation method is used for allocation of shared registers in a processor and comprises the steps that a discontinuous register block list is added into an available register block list, the discontinuous register block list comprises a plurality of discontinuous register blocks with the interval step length being M, and M is a positive integer larger than 1; in response to an allocation request that the current instruction occupies K discontinuous registers, allocating a discontinuous register group to the current instruction, K being a positive integer greater than 1; and updating the state of the distributed discontinuous register block in the discontinuous register block list. According to the storage resource allocation method, the utilization rate of the register can be improved, so that the number of synchronously executed threads is increased, and the program execution parallelism is improved.

The embodiment of the invention discloses a data synchronization method and device, electronic equipment and a storage medium. The method comprises the steps that connection with a data source is established, metadata information of the data source is acquired and analyzed, and data source characteristics are obtained; determining a target index according to the real-time operation index of the data source and a local real-time allowable performance index, determining an adjustment strategy of the collection thread count according to the target index, and adjusting the current collection thread count according to the adjustment strategy; and determining the synchronization content of the to-be-synchronized data according to the data source characteristics and the database log, determining a parallel pipeline according to the synchronization content, and issuing the synchronization content to the parallel pipeline to synchronize the to-be-synchronized data. According to the method, the acquisition thread count is accurately adjusted, and the load balance of the two ends in the data transmission process is ensured while the data transmission load pressure of the data source and the local is relieved.

The invention discloses a streaming task operation method and device based on a distributed batch scene, and relates to the technical field of distribution.The method comprises the steps that after a task scheduling instruction distributed by a scheduler is received, the global streaming read thread count in a distributed cache is inquired, and when the task scheduling instruction does not exceed a preset thread count threshold value, the task scheduling instruction is executed; adding the job task into a streaming read task execution queue, scheduling a to-be-executed task from the streaming read task execution queue under the condition of monitoring that the target cache triggers a resource release event, confirming a target thread corresponding to the to-be-executed task, and reading task data from a database through a streaming read mode according to the to-be-executed task; after task processing is completed, thread resources corresponding to the target thread are released, and the thread occupancy state in the target cache is updated. According to the method and the device, the technical problems of long task processing time, excessive consumption of thread resources and influence on database use due to adoption of a streaming reading strategy in a distributed batch scene in related technologies are solved.

Systems and methods are described for allocating computing resources to load data into a database. In some aspects, a database service obtains a request to load a number of data chunks. Responsive to the request, the database service scales a number of worker threads to load the data chunks. The scaling may include selecting a maximum number of worker threads to load (e.g., concurrently) the data chunks based on a number of data chunks, and assigning individual worker threads to load individual data chunks. The scaling may additionally include tracking a number of available worker threads relative to the maximum number of worker threads to reassign a first worker thread to load another data chunk upon the completing loading a prior data chunk, or assign a new worker thread that has become available based on a change in the maximum number of worker threads.

The embodiment of the invention discloses a multi-dimensional tensor protocol calculation method and device, equipment, a storage medium and a product. The method comprises the following steps: determining m reduction dimensions and (n-m) non-reduction dimensions of an n-dimensional tensor; a protocol calculation instruction is generated, K thread blocks are started on the artificial intelligence chip after the protocol calculation instruction is executed, and the K thread blocks conduct protocol calculation on m protocol dimensions in a parallel mode; the number of threads contained in each thread block is associated with the number of elements contained in m protocol dimensions, K is associated with the number of elements contained in (n-m) non-protocol dimensions, and each thread in each thread block reads respective protocol data in a parallel mode; the protocol calculation instruction is sent to the artificial intelligence chip, and the artificial intelligence chip executes the protocol calculation instruction. According to the method, protocol calculation can be carried out on any dimension of the multi-dimensional tensor, the universality is improved, the calculation time is shortened, and the calculation efficiency is improved.

A system for scaling parallelism in sweeping stale data associated with multi-version data management is disclosed. The system is programmed to manage a targeted sweep of stale data stored on distributed storage associated with a set of computing nodes, where the stale data includes metadata for database transactions. The system is programmed to dynamically determine, on a computing node, how many threads to simultaneously perform deletion for the targeted sweep based on historical performance of the computing nodes, historical performance of the distributed storage associated with the computing nodes, or other factors. The system is programmed to then cause a certain number of threads to run on the computing node, where each thread identifies and deletes specific metadata for data transactions within the scope of the targeted sweep.

The invention relates to the technical field of chips, and provides a method and a device for optimizing memory stack conflicts during access to a local register. The method comprises the steps that M arithmetic logic groups are set, the maximum number of threads capable of being accessed by a register in parallel is N, the number of concurrent threads of each arithmetic logic group is increased to N from N / M, and the number of pipelines connected with each arithmetic logic group and an instruction module is increased to N; therefore, the total number of the threads concurrently accessing the registers in each clock period is increased to K, K = M * N, the number of the concurrent threads in each clock period is increased to K from original N, and the requirement that the number of the concurrent threads and the registers form a many-to-few mapping relation is met; the instruction module receives the access requests of the K threads and plans the issuing time of the access requests of the K threads, so that the number of the access requests issued in each clock period reaches the maximum issuing number. According to the method and the device, the access requests of the threads are issued in a smaller number of clock cycles as much as possible, and the instruction performance is improved.

The invention relates to the technical field of graphic rendering, and provides a task processing method and device, a storage medium, electronic equipment and a program product. The task processing method comprises the steps that a plurality of first calculation tasks are converted into at least one second calculation task, and the number of threads in the first calculation tasks is smaller than the number of threads in the second calculation tasks; and executing the at least one second calculation task in parallel through the plurality of processing units to obtain a task processing result. According to the method, a plurality of small-order first computing tasks with serial execution sequences are converted into at least one large-order second computing task, and the plurality of small-order first computing tasks are compactly arranged, so that the large-order second computing tasks can be simultaneously configured to a plurality of existing processing units to be processed; therefore, parallel processing of the multiple first computing tasks is achieved, multiple existing processing units are fully utilized, GPU computing resource waste is avoided, and task processing efficiency is improved.

The invention discloses a self-adaptive memory allocation optimization method and device, terminal equipment and a storage medium, and belongs to the field of memory management. The method comprises the steps of monitoring thread data and system state data in real time, extracting thread memory behavior characteristics and resource state characteristics when one variable quantity exceeds a safety threshold value, and performing weighting calculation according to a preset weight coefficient to obtain a target strategy score; when the strategy score difference value of the target strategy score and the current strategy score is smaller than a preset difference value threshold value, the current memory allocation strategy is kept; when the strategy score difference value is not smaller than a difference value threshold value, a target memory allocation strategy is matched in a preset strategy library according to the thread memory behavior and the resource state characteristics, a current memory allocation strategy is updated, memory allocation parameters are adjusted, and a current strategy score is updated; when the current memory allocation strategy and the current strategy score are compared for the first time, the current memory allocation strategy and the current strategy score are the initial memory allocation strategy and the corresponding initial score, and the problem that efficiency and compatibility cannot be considered at the same time in the prior art is solved.

According to the pressure testing method and system, the storage medium and the computer equipment provided by the invention, when the pressure testing server executes the pressure testing task, the control parameters input by the user and the thread states of all available threads of all the pressure testing agent machines connected with the server can be firstly obtained; then calculating a first thread count required by pressure measurement at the current moment according to the control parameter, and notifying at least one pressure measurement agent machine to start a thread corresponding to the first thread count to execute the pressure measurement work according to the first thread count and the thread state of each available thread; and obtaining an execution result returned after the execution of the at least one pressure measurement agent machine is completed, updating the thread state of the thread executing the pressure measurement work according to the execution result, and after the current moment is replaced with the next moment, continuing to execute the pressure measurement work until the pressure measurement task is completed. The process can realize the accurate control of the simulation flow of the pressure measurement agent machine, thereby achieving the effect of reproducing the flow at the real peak moment.

The invention discloses a quick temperature field simulation method based on a multi-thread architecture and a background calculation strategy, and relates to the field of information technologies and industrial processes. The method comprises the following steps: firstly, establishing data response logic of a main thread and a slave thread of a temperature field rapid simulation system based on a multi-thread architecture, blocking and releasing response actions of the threads by using a blocking queue, and improving the reliability of multi-thread data interaction; and secondly, establishing asynchronous data response logic of the temperature field rapid simulation system based on a background calculation strategy, calling CFD software in advance from a thread to simulate and calculate temperature field simulation data of a (k + 1) th period, and realizing uninterrupted asynchronous data playing and response of the temperature field rapid simulation system. According to the multi-thread asynchronous response method for the temperature field rapid simulation system, the problem that multi-thread data response of the temperature field rapid simulation system is disordered can be effectively avoided, meanwhile, the problem that calculation cannot keep up with data playing due to complex model calculation is solved, and the multi-thread asynchronous response method for the temperature field rapid simulation system with complex background calculation is achieved.

The thread allocation method, device, storage medium and computer device provided by the application, when the target task to be executed and the total number of threads to be allocated are determined, since the target task includes a plurality of concurrently executed subtasks, and the number of subtasks is not greater than the total number of threads, therefore, at the initial running of the system, the application can perform at least one thread allocation operation on each subtask according to a preset first allocation strategy until the number of allocated threads reaches the total number of threads; further, in order to avoid allocation errors caused by uneven data size, the application can set a round interval in advance, and after the thread allocation is completed, the current execution round of the target task is monitored, and when the current execution round reaches the preset round interval, at least one thread allocation operation is performed on each subtask according to a preset second allocation strategy until the number of allocated threads reaches the total number of threads, which can improve the resource utilization.

The invention provides a resource scheduling method and related device.The method comprises the steps that operation parameters of an instance are detected according to a preset detection rule, the operation efficiency of the instance is obtained, and the operation parameters comprise the number of threads and the number of execution tasks; under the condition that the running efficiency of the instance does not meet the preset running requirement, determining the instance type of the instance; and determining a corresponding preset resource scheduling rule according to the instance type, adjusting the thread count of the instance to a first thread count according to the preset resource scheduling rule, and adjusting the execution task number of the instance to a first task number. The task number and the thread number of the instance can be dynamically adjusted, so that the task execution efficiency of the instance is improved.

The embodiment of the application provides a kind of method and device for determining target thread, storage medium, electronic equipment, which comprises: performing classification operation to N test class files in test case set, obtain M test class file set;Determine the first thread interval required to run the test case included in each test class file set based on the number of test class files included in each test class file set;Based on the running time of test case and the first thread interval included in each test class file set, the predicted running time of each test class file set is calculated, and M predicted running times are obtained;According to target predicted running time and the running time of test case included in target test class file set, the target thread required to run the test case included in target test class file set is calculated. Through the present application, the problem that the number of threads cannot be reasonably determined in the related art can be solved, and the effects of reasonably starting threads and reducing resource consumption are achieved.

Code running on a processing unit can retain a hold on an active thread thereby preventing another thread running on the processing unit from executing. This can occur when the code requests more threads to be executed than the processing unit can give forward progress to, e.g., more threads than can be run simultaneously. Inserting the conventional yield instruction can lead to processing inefficiency such that the yield instruction can be executed more times than necessary thereby leading to an inefficiency in processing cycles of the processing unit performing the thread-switching routines. A yield occasionally instruction can be inserted into the code instead of the yield instruction. The yield occasionally can limit the times a yield instruction is executed thereby improving the efficiency of the processing cycles. The yield occasionally instruction can be used in non-terminating regions of the code, where the forward progress of other threads cannot be guaranteed.