88 results about "Execution time" patented technology

The invention relates to a method for filling a filling installation, comprising a step of receiving tasks to be performed and a set of resources to be used, a step of selecting available resources corresponding to the resources to be used, and of selecting the tasks to be performed and at least one production order such that the set of tasks to be performed of said at least one production order is selected, a step of defining a complete execution sequence of the selected tasks to be performed such that the time to perform each selected task to be performed is less than or equal to the available time of the selected available resources, and a step of filling at least a portion of the plurality of available containers according to the complete execution sequence. Abstract figure: Fig. 1

Various example embodiments relate to methods, devices, and / or systems for executing a plurality of tasks in a computer operating environment. The method comprises monitoring a remaining execution time of a current task executing on the processing circuitry, and performing, by the processing circuitry, pre-emption of the current task based on the remaining execution time of the current task and a desired execution threshold time, the desired execution threshold time being one of a desired value based on one or more configuration parameter of the current task or a dynamic value determined based on information related to ongoing activities in the computer operating environment.

The present application provides a power control device that can suppress the power consumption reduction effect obtained by leveling execution from being weakened by suppressing the power consumption amount after leveling execution. The power control device includes a communication unit and a control unit. The control unit stores information of a leveling instruction and information of a recovery mitigation request. The information of the leveling instruction includes information of an execution time and information of a first allowable average power in the execution time. The information of the recovery mitigation request includes information of a recovery period and information of a second allowable average power in the recovery period. The control unit controls the driving of a power load device in such a manner that the average consumption power of a facility in the execution time becomes lower than the first allowable average power, and controls the driving of the power load device in such a manner that the average consumption power of the facility in the recovery period becomes lower than the second allowable average power. The second allowable average power is greater than the first allowable average power.

A processor includes a time counter and issuing instructions and executing instructions at a future time, which is based on the time counter. The execution times are based on fixed latency times of instructions, with the exception of the load instruction, which is based on the data cache hit latency time. A data cache miss causes the load instruction to fetch data from the level 2 cache, wherein a time tracker unit adjusts the level 2 cache latency time based on a counter.

This application relates to the field of computer technology and discloses a data processing method, system, medium, and program product. In this method, the data processing system first allocates corresponding computing resources for a first task. Then, it determines the transmission time required for the task data of the first task to be transmitted to the data processing system. During the transmission of the task data of the first task, a second task is executed based on the computing nodes allocated to the first task; the execution time of the second task is less than the transmission time. Corresponding to the completion of the transmission of the task data of the first task and the completion of the execution of the second task, the first task is executed based on the computing nodes allocated to the first task. Thus, even if the transmission of the task data of the first task takes a significant amount of time, the computing nodes allocated to the first task can execute other tasks with shorter execution times during the transmission of the task data of the first task, thereby reducing the waste of computing resources and improving resource utilization.

This invention discloses an AI inference resource decoupling scheduling method and system based on a two-dimensional orthogonal state space. User AI service rights are decomposed into storage rights dimension σ and computing power rights dimension φ, constructing an orthogonal state space S=(σ,φ), where σ and φ change independently and have no causal dependency. Four service modes are derived from the state space: full service, read-only memory, temporary session, and guided recovery. State transitions in the computing power rights dimension are triggered by critical conditions of discrete quota counting: the cost r for a single token generation is defined as the lower limit of physical resources required for a single forward propagation of the model decoder. When the remaining quota Q(t) is less than r, the physical system cannot complete the next calculation, and the state will inevitably transition and terminate the computation flow. State switching is achieved through metadata-level atomic operations, and the execution time is independent of the user data scale. This invention clarifies the inevitability of the technical effect through logical deduction, is applicable to software or hardware-software co-implementation, and solves the problems of resource deadlock, switching latency, and service interruption under subscription systems.

This invention discloses a multi-dimensional resource scheduling method for professional equipment simulation training, specifically relating to the field of professional equipment simulation training management and control. It addresses issues such as inconsistent resource states, execution time conflicts, and on-site occupancy insertions leading to plan mismatches that are difficult to review and correct during concurrent training tasks. The method unifies and merges scheduling console operation logs, node heartbeats, and task registration forms to construct a schedulable resource base map. It then generates a task package matching table and training execution sequence by combining the resource demand and execution dependencies of injected task packages. During execution, it continuously verifies and records abnormal occupancy. Finally, it corrects the matching table and scheduling rules by comparing the baseline image and reset image with the standard playback results, forming an iteratively updatable next round of training execution sequence. Simultaneously, it accumulates samples to support subsequent stable scheduling decisions.

Systems, apparatuses, and methods for updating and optimizing task scheduling policies are disclosed. A new policy is obtained and updated at runtime by a client based on a server analyzing a wide spectrum of telemetry data on a relatively long time scale. Instead of only looking at the telemetry data from the client's execution of tasks for the previous frame, the server analyzes the execution times of tasks for multiple previous frames so as to determine a more optimal policy for subsequent frames. This mechanism enables making a more informed task scheduling policy decision as well as customizing the policy per application, game, and user without requiring a driver update. Also, this mechanism facilitates improved load balancing across the various processing engines, each of which has their own task queues. The improved load balancing is achieved by analyzing the telemetry data including resource utilization statistics for the different processing engines.

The application discloses an edge-cloud collaborative computing power scheduling architecture and a low-power consumption management method, and relates to the technical field of edge computing.The application collects multi-dimensional resource indexes of distributed edge computing nodes, calculates resource availability scores through standardized processing and weighted comprehensive score calculation; extracts static code features and dynamic running features of tasks, adopts a gradient boosting decision tree algorithm to predict execution time, CPU occupancy, memory requirement and network transmission data volume of the tasks, and performs node performance correction; adopts an improved non-dominated sorting genetic algorithm to solve a multi-objective optimization problem, and generates a task offloading decision; the application solves problems such as difficulty in resource quantification under an edge computing environment, lack of accurate prediction in task scheduling, poor single-objective optimization effect, and unintelligent power consumption control, realizes efficient task scheduling and low-power consumption operation, and has good scalability and adaptability.

The present disclosure relates to a method and device for performance evaluation of a deep learning model based on a GPGPU architecture, an apparatus and a medium. The method comprises: obtaining a computation graph of a deep learning model; converting the computation graph into an intermediate representation, wherein the intermediate representation comprises operator parameter information of the operators and a dependency relationship between the computation nodes; obtaining hardware configuration parameters of a GPGPU used for deploying the deep learning model; decomposing each operator corresponding to a computation node into an instruction-level operator sequence according to the hardware configuration parameters of the GPGPU; for each computation node, obtaining a performance evaluation formula for each operator item in the instruction-level operator sequence corresponding to the computation node; determining a node performance indicator of the computation node according to the performance evaluation formula corresponding to each operator item; and determining an execution time and a performance bottleneck of the deep learning model based on the dependency relationship between the computation nodes and the node performance indicator of each computation node.

A system and method for power and latency reduction when processing thread execution code in a multithreaded architecture is disclosed. The method includes receiving a plurality of threads, each thread including a plurality of instructions for execution on one of a plurality of cores of processing circuitry; detecting a plurality of subsequent independent instructions in a first thread of the plurality of threads; and inserting an instruction hint into an instruction, the instruction hint, when executed, configuring an instruction scheduler of the processing circuitry to serially execute the plurality of subsequent independent instructions.

This invention belongs to the field of cloud computing, specifically relating to a cost-based workflow scheduling method in an IaaS environment. The method includes: obtaining the execution time of each task in the workflow on different virtual machines, and the data communication time between different virtual machines; constructing a dynamic budget allocation model based on user-defined total cost constraints; merging the workflow into single parent-child task pairs to obtain a new directed acyclic graph (DAG); calculating the priority score of each task in the new DAG and sorting them to obtain a sorted list; calculating the cost constraint of each task sequentially according to the sorted list, and obtaining the currently allowed budget for scheduling the task based on the dynamic budget allocation model; calculating the earliest completion time of the task before and after copying the critical parent task on different virtual machines, and selecting the virtual machine with the smallest earliest completion time to place the task; this invention effectively reduces the workflow completion time.

This invention relates to the field of hydrogen electrolysis technology, and more particularly to an automatic control method for an electrolysis hydrogen production system. The method includes: constructing a time sequence table of experimental operating parameters in the host computer based on periodic bidirectional communication between a host computer and a programmable logic controller (PLC); dividing the experimental process into multiple time-sequential operating stages; and configuring corresponding operating parameters and durations for each operating stage. The experimental execution progress is uniformly managed by initializing the experimental state machine and setting an execution pointer. During the experimental operation, the host computer periodically sends out operating parameters and collects operating status parameters, updates the execution pointer based on the collected results, and performs anomaly detection. When an anomaly occurs, the PLC is controlled to stop current input. Simultaneously, the host computer periodically records the execution pointer to form breakpoint information, enabling the recovery of operating stages and remaining execution time after an experimental interruption, and the continuation of the experimental process.

This invention provides a method and apparatus for multi-task scheduling in federated learning. The method involves acquiring a target federated learning task and its current events, wherein the target federated learning task includes multiple events, which are divided into stages based on their execution time and required resources. An event priority ratio is determined based on the attribute information of the target federated learning task and the type of the current event. A cumulative priority is determined based on the event priority ratio and the time the current event entered the scheduling queue. A predefined adaptive resource allocation algorithm is used to allocate computing nodes to the current event, and the current event is executed based on the computing nodes and according to the cumulative priority. This invention ensures fairness in task execution, reduces the overall task execution time of the system, and improves resource utilization.

A system for accelerator functions as a service is disclosed. The system may receive a job that includes a computer (CPU) portion and an accelerator portion. When the accelerator portion is performed, an execution time associated with a time for the accelerator to return results is determined. Resources or a portion thereof allocated to the job, or the CPU portion are freed or reallocated to another job at least during the execution time. The job is queued to receive resources when the results are received from the accelerator.

This application discloses a tensor program performance prediction method and system based on LLVM, relating to the field of computer system performance modeling and compiler optimization technology. This invention determines the instruction type of each instruction by semantic recognition of the tensor intermediate representation and transforms the instructions into task nodes with semantic tags. By calculating the symbolic execution count and corresponding execution cost of each instruction type tag within each basic block, the program load is quantified without executing the code. Furthermore, the basic blocks are scheduled and arranged under the physical resource constraints of the target hardware, achieving deep coupling between program logic semantics and hardware physical constraints. Finally, the total predicted execution time of the entire tensor program is obtained based on the predicted execution time of each basic block and the number of times each basic block appears. This enables accurate evaluation of the execution time of a high-performance tensor kernel without running on the target hardware platform.

This invention discloses a method, apparatus, medium, and product for a multi-engine asynchronous parallel computing architecture. The method includes: acquiring interval marker pairs, which indicate the start and end of a target program segment in a preset program; in response to the interval marker pairs, determining the boundary notification corresponding to the target program from among multiple notifications sent to the control core when each asynchronous engine related to the target program segment performs an operation; and determining the actual execution time of the target program segment based on the sending times of all boundary notifications sent by all asynchronous engines related to the target program segment.

Techniques and systems are disclosed for evaluating and optimizing logical plans using machine learning. In an aspect, a logical plan is received and submitted to a machine learning model, which predicts whether applying a rule to the logical plan provides a benefit, such as reduced execution time or resource use. Based on the prediction, the rule is applied or marked for application, or not applied or not marked for application. Training operations for the machine learning model include calculating and comparing execution costs of applying or not applying a rule, generating a numerical representation of the logical plan, and assigning labels based on cost differences. A machine learning model trained with the representations and labels predicts rule benefits for logical plans, including both new logical plans and those used during training.

The application discloses a kernel parameter optimization method, system and device based on an XGBoost model, and relates to the field of computer architecture and high-performance computing optimization. The matrix geometric features of a target sparse matrix are acquired, the hardware alignment features are calculated according to the basic execution unit size of a heterogeneous computing platform, and the matrix feature vector is constructed. The matrix feature vector is combined and spliced with each kernel parameter in a kernel candidate pool to generate an inference matrix. The inference matrix is input into a pre-trained XGBoost model to obtain the predicted execution time consumption of each kernel parameter combination. The optimal kernel parameter combination is determined according to the predicted execution time consumption. By introducing the hardware alignment features and the logarithmic time consumption label transformation, the application realizes high-precision prediction of the performance of sparse matrices across orders of magnitude, and reduces the online optimization overhead to the level of microsecond inference delay.

The embodiment of the application provides a data management method and device, electronic equipment and storage medium, the method obtains monitoring information of data by responding to the execution time of the set task, and judges whether the data meets the preset operation condition according to the monitoring information, and when the data meets the operation condition, the target operation associated with the operation condition is used to process the data, based on the above-mentioned mode, for each data to be managed, the target operation associated with the operation condition can be automatically processed when the corresponding set operation condition is met, effectively reducing data accumulation, thereby improving the query efficiency of the data, at the same time, the processing result of the data is monitored and recorded, the life cycle management of the data is realized, thereby facilitating the management personnel to more clearly and directly understand the running state of each data based on the record period, and improving the data management efficiency.

Provided in the present disclosure are a method and apparatus for data processing, and a program, a medium and a device. The method comprises: determining dependency relationships among a plurality of tasks in a workflow to be processed and respective execution times of the plurality of tasks; on the basis of the dependency relationships and the respective execution times of the plurality of tasks, determining a scheduling and orchestration policy for the plurality of tasks, wherein the scheduling and orchestration policy indicates a plurality of first-level hierarchies and at least one second-level hierarchy under each first-level hierarchy, each second-level hierarchy comprises at least one task group, each task group comprises at least one task of the plurality of tasks, the plurality of first-level hierarchies are configured to be executed in series, and each task group in each second-level hierarchy is configured to be executed in series or in parallel; and on the basis of the scheduling and orchestration policy, scheduling the plurality of tasks in the workflow for execution. In this way, the time consumption of workflow processing can be reduced, and the processing efficiency can be improved.

This application relates to a script orchestration method, apparatus, computer device, storage medium, and computer program product, belonging to the field of artificial intelligence technology. The method includes: acquiring a script execution plan for a test task and preset adjustment information for the script execution plan; when the planned total execution time corresponding to the script execution plan is not equal to the actual total execution time, adjusting the total number of execution days and / or daily execution time of the script execution plan according to the preset adjustment information to obtain a script orchestration plan; the actual total execution time is the sum of the actual execution times of each script in the full set of scripts in the script execution plan; iteratively executing all scripts according to the script orchestration plan to obtain the average execution time of each script in the full set of scripts; updating the script orchestration plan according to the average execution time of each script to obtain an updated script orchestration plan. Using this method can improve the utilization rate of automated testing resources.

The application discloses a method and system for interference joint control oriented to continuous batch processing reasoning. The method determines current running state data in a scheduling period; calculates interference degree and execution time consumption of the scheduling period according to the current running state data; based on the current running state data, the interference degree and the execution time consumption, effect evaluation and energy consumption estimation are performed on a preset frequency candidate set and a pre-filled shaping quota candidate set to screen out a feasible scheme list; a target control scheme is determined to generate a control instruction; and the frequency and the pre-filled limit are controlled according to the control instruction, and the safety margin is updated according to the actual running result, so that the updated safety margin is used for screening of the feasible scheme list next time, interference-driven closed-loop joint control is realized, the hardware energy consumption is reduced, and the service reliability and the running efficiency of the system under high-concurrency mixed load are improved while the large model reasoning response speed and the output stability are ensured.

The application relates to a task scheduling method, device and computer equipment. The method comprises the following steps: obtaining task information of each to-be-scheduled task and resource states of each computing node in a computing cluster; wherein the task information of each to-be-scheduled task at least comprises data volume and data distribution information of processing data required by the to-be-scheduled task; determining execution time lengths of each computing node for executing each to-be-scheduled task according to the task information of each to-be-scheduled task and the resource states of each computing node; determining load imbalance degrees corresponding to each candidate scheduling strategy according to the execution time lengths of each computing node for executing each to-be-scheduled task; the candidate scheduling strategy comprises node allocation of each to-be-scheduled task; selecting a target scheduling strategy from each candidate scheduling strategy according to the load imbalance degrees corresponding to each candidate scheduling strategy; and scheduling each to-be-scheduled task based on the target scheduling strategy. The method can improve resource utilization and task processing efficiency.

The invention relates to a method for filling a filling installation, comprising a step of receiving tasks to be performed and a set of resources to be used, a step of calculating the adjusted available time, a step of selecting available resources corresponding to the resources to be used, and of selecting the tasks to be performed and at least one production order such that the set of tasks to be performed of said at least one production order is selected, a step of defining a complete execution order of the selected tasks to be performed such that the time to perform each selected task to be performed is less than or equal to the available time of the selected available resources, and a step of filling at least a portion of the plurality of available containers according to the complete execution order. Abstract figure: Fig. 1

The application provides a database connection operation optimization method, a storage medium and a computer device. The method comprises the following steps: starting optimization of a connection operation by an optimizer; judging whether two side sub-tables participating in the connection operation satisfy a preset optimization condition; if yes, creating an alternative plan according to a preset connection order and a connection mode; and adding the alternative plan into an execution plan alternative queue, and selecting an actual execution plan from the execution plan alternative queue by the optimizer. According to the scheme, the execution efficiency of the connection operation is improved by targeted optimization, the execution complexity of the connection query is reduced, and the execution time of the connection operation is saved.

This application discloses a method, apparatus, and electronic device for dynamically adjusting a task queue. The method includes: generating tasks based on task configuration information and storing the tasks in a pre-execution task queue; storing tasks in the pre-execution task queue that meet the execution conditions into a task execution queue; determining a prediction coefficient for the task based on the task's actual execution time in the previous run and its historical average execution time, and determining the estimated execution time of the task based on the historical average execution time and the prediction coefficient; determining the number of executable tasks based on the server's real-time resource load, the task's historical average resource load, and the number of tasks currently being executed by the server, and selecting tasks to be executed from the task execution queue; if, based on the historical average resource load and real-time resource load of the tasks to be executed, it is determined that the resources do not meet the concurrency requirements, then the tasks to be executed are scheduled according to the estimated execution time. The above technical solution improves task execution efficiency.

The application discloses an instruction processing device, method, processor and computer equipment, and belongs to the technical field of computers. The instruction processing device comprises an execution unit and a processing unit. The execution unit is used for executing a first instruction to obtain an execution result of the first instruction, and the first instruction is any instruction in candidate instructions. The processing unit is used for discarding the execution result of the first instruction in the case that the execution result of the first instruction meets a discard condition. The discard condition is used for indicating that other instructions in the candidate instructions except the first instruction are still normally executed after the execution result of the first instruction is discarded. Based on the above device, method, processor and computer equipment, the execution times of write operations can be reduced by discarding the execution result of the first instruction, so that the power consumption of the processor can be saved, and the performance of the processor can be improved.

This invention discloses a large-model-driven method for the automatic generation and optimization of hybrid workflows, comprising: receiving user task requirements and constructing an initial workflow; establishing a space representing the structural and computational characteristics of the workflow, and iteratively optimizing the workflow within this space; during the optimization process, analyzing historical feedback using a large model and generating improvement strategies, thereby generating candidate workflows that integrate semantic reasoning units and deterministic computation units; performing execution tests on the candidate workflows, and filtering and updating them through multi-dimensional performance evaluation until a heterogeneous workflow meeting preset conditions is formed. This invention also discloses a system for implementing the above optimization method, which has broad application value. This invention can automatically construct heterogeneous workflows for complex tasks, significantly improving task processing accuracy while greatly reducing inference costs and execution time, achieving optimal resource utilization.