124results about How to "Reduced execution time" patented technology

The invention provides a natural language SQL (Structured Query Language) conversion method based on a data platform and a large language model, which relates to the technical field of data processing and comprises the following steps of: performing vectorization storage on metadata, blood relationship, indexes and historical query information, performing intention judgment and vector matching on a user query request, and determining a query condition based on a matching result. An association path is constructed according to data consanguinity, and an optimal SQL statement is generated by utilizing a historical query optimization table connection sequence and combining index matching. According to the method, intelligent conversion from the natural language to the efficient SQL can be realized, the query efficiency is improved, and the use threshold of a user is reduced.

The invention discloses a computing task scheduling method and device in a multivariate heterogeneous environment and a medium, and relates to the technical field of parallel computing. The method comprises the following steps: monitoring resource states of various computing resources in a multivariate heterogeneous environment in real time, and quantifying the resource states to obtain corresponding resource information; under the condition that the to-be-processed calculation task is generated in the scheduling task list, performing demand analysis on the to-be-processed calculation task to determine a task demand report; calculating the task demand report and the resource information by using a preset scheduling strategy generation model to generate an intelligent scheduling strategy; according to the intelligent scheduling strategy, various computing resources are called to execute the to-be-processed computing task; and continuously monitoring the scheduling task list and the resource state so as to update the intelligent scheduling strategy corresponding to the to-be-processed calculation task based on the scheduling strategy strategy optimization logic.

The invention discloses a multi-NDP (Near-Data Processing) system read-write acceleration method based on a key value separation technology, and the method comprises the steps: when data is written, splitting a data key value pair written into a batch processing region of a multi-NDP system, determining target equipment, then carrying out key processing and value processing, carrying out key storage on keys, and carrying out data storage on the target equipment; a value corresponding to the key is added to a value log file of the target equipment; during data query, hash operation is carried out on a query key, target device position information of a corresponding value is obtained, then the master-slave multi-level hot cache is queried, and query retrieval of the corresponding value is carried out according to the obtained value position index. According to the method, the key value separation technology is introduced into the multi-NDP system, when the Compaction is executed, the execution time of the Compaction is shortened, the influence of the Compaction on the writing operation of the user is reduced, and the writing performance is greatly improved.

The invention provides a parallel training method, device and equipment based on adaptive re-calculation and load division and a computer readable medium. The method comprises the following steps: firstly, converting a dynamic calculation graph in a deep learning framework into a static calculation graph data structure; then, the calculation sequence is divided, and calculation sub-graphs are generated; constructing a mixed integer programming model according to the calculation time of the operator, the transfer cost of the operator and the constraint condition; then, solving the scheduling optimization problem of the operator by using a solver to generate a video memory optimization strategy; aiming at the generated video memory optimization strategy, evaluating the execution time of each calculation sub-graph, and calculating to obtain a load balancing rate; and finally, according to a calculation result of the load balancing rate, determining a calculation sub-graph which is finally divided, and performing parallel training by utilizing the calculation sub-graph which is finally divided. Through the method provided by the invention, the final operator execution sequence can be found, so that the peak video memory consumption during the operation of the computational graph is within a given limit, and the training time of the neural network model is minimized.

The invention discloses an intelligent self-adaptive real-time task scheduling method and system based on multi-dimensional optimization. The method comprises the following steps: S1, inputting a task; s2, a DAG task is decomposed into independent sequential sub-tasks; s3, dynamically calculating the priorities of the tasks by adopting a multi-factor dynamic priority scheduling module, and updating the priorities of the tasks; s4, dynamically adjusting an acceleration factor by adopting a self-adaptive acceleration factor management module; s5, vertex correlation analysis is carried out, and a variance of processor load imbalance is calculated by adopting a load balanced distribution algorithm and is minimized; s6, in combination with static analysis and dynamic inspection, a hybrid schedulability judgment module is adopted to judge the tasks, and judgment results are fused; and S7, fusing decisions by adopting a decision fusion algorithm, obtaining a scheduling strategy through a scheduling strategy coordinator, executing the task, and monitoring the task in real time. Real-time monitoring and real-time optimization scheduling are realized, and the utilization rate of resources is improved.

The invention discloses a hierarchical collaborative deep reinforcement learning decision-making method for multi-agent path planning, and belongs to the technical field of multi-agent path planning. The method comprises the following steps: constructing a PPO-based hierarchical collaborative decision-making model through a bottom-layer task allocation decision-making module and an upper-layer behavior decision-making module taking PPO as a framework; taking environment RGB image data, a task request and observation information as input of a layered collaborative decision model based on PPO; wherein the observation information comprises bottom layer observation information and upper layer observation information; and the optimal control strategy of the multi-agent system is obtained by continuously training the PPO-based hierarchical collaborative decision model. The method has the advantages that task resource allocation is efficient, the system state is stable, the execution time is shortened, and the robustness of the system is enhanced.

The invention discloses a transfer learning-based cross-data center storage and calculation cooperative training acceleration method. The method comprises the following steps of S1, deploying an initial version of a deep learning model in a plurality of heterogeneous data centers; s2, performing initial training on the deep learning model in the first data center to generate a first model parameter set; s3, generating a local deep learning model parameter set; s4, establishing a dynamic weight adjustment mechanism of the heterogeneous data center; s5, distributing different computing resources to each heterogeneous data center according to the participation weight; s6, scheduling storage resources and computing resources in real time according to resource use conditions and task requirements of the heterogeneous data centers; s7, performing multi-task parallel processing; s8, monitoring the resource utilization rate of each heterogeneous data center through a dynamic resource scheduling strategy; and S9, generating a final global optimization model parameter set, and applying the global optimization model parameter set to subsequent task deployment of each heterogeneous data center. According to the method, the collaborative training efficiency of the cross-heterogeneous data center is optimized.

The invention discloses an underwater robot intelligent control method and system based on deep learning, and relates to the technical field of underwater obstacle avoidance of underwater robotics.The method comprises the steps that a deep learning calibration model is established through a deep learning algorithm, and the avoidance rate index BrsL in the underwater robot avoidance process is constructed; the ocean current speed and direction of the area where the underwater robot is located and the real-time vertical descending depth, the underwater temperature value and the submerging seawater pressure value of the propeller in the obstacle avoiding process are monitored and monitored, an ocean current influence index YL and a relative strength index Xd are calculated and evaluated, and the ocean current influence index YL and the relative strength index Xd are evaluated; if a multi-stage calibration control strategy of the first calibration instruction and the second calibration instruction is triggered, the power of the propeller and the posture of the underwater robot are adjusted, and timely adjustment is carried out according to the change of the vertical descending depth so as to adapt to the ocean current environment, and it is guaranteed that the underwater robot stably moves in the obstacle avoidance process.

The invention provides a synchronous zero voltage modulation system and method of a cascade system, a medium and equipment wherein the synchronous zero voltage modulation system of the cascade system comprises a power circuit module and a modulation system; the power circuit module comprises a front-stage four-switch buck-boost converter and a rear-stage LCC-LCC compensated bidirectional wireless charging system; the modulation system comprises a voltage compensation module, a current compensation module, a current comparison module, a synchronization module and a switch driving signal generation module, the output end of the voltage compensation module is connected with the input end of the synchronization module, and the output end of the current compensation module is connected with the input end of the synchronization module. The output end of the current comparison module is connected with the input end of the synchronization module, and the output end of the synchronization module is connected with the driving signal generation module. According to the invention, a plurality of time scales are adopted, the implementation condition of synchronous zero-voltage switch modulation is accurately calculated, the calculation speed is high, the memory occupation amount is small, and the zero-voltage switch synchronous modulation of the cascade system is realized.

The invention discloses a data transmission method and device for a GPGPU and a host and a medium, and the method comprises the steps: obtaining a first transmission attribute corresponding to DMA transmission at this time through a running program at a host side; if the number of channels is a specified number representing automatic setting, determining the number of actual channels used at this time according to the data length of the first transmission attribute and the task emergency degree of DMA transmission at this time; through a DMA-CTRL controller, corresponding channels are selected based on the actual channel number, and second transmission attributes corresponding to the channels are obtained; and controlling the DMA controller through the DMA-CTRL controller, executing the DMA transmission based on the second transmission attribute, and feeding back a transmission result to the DMA-CTRL controller through the DMA controller. The DMA-CTRL controller is added in the GPGPU, and the DMA controllers and the PCIe-D controllers are controlled to perform data transmission at the same time, so that the data transmission speed between the GPGPU and the host can be effectively increased, the overall task execution time is shortened, and the influence of the PCIe interface on the overall calculation performance is improved.

The invention discloses a heterogeneous computing architecture testing method of an ISP module, and belongs to the technical field of heterogeneous computing. According to the technical scheme, a high-level synthesis tool is used for generating and verifying an IP to be tested; a target algorithm IP application system is constructed by using a microprocessor IP hardcore and an AXI bus connector of a heterogeneous computing platform, a target algorithm IP core is operated by using software to carry out an application test, and actually measured data is stored in a nonvolatile storage medium; transplanting an operation system and an OpenCV library to the heterogeneous computing platform for the development board, and storing a target algorithm software operation result to a data file after a test program is operated; and evaluating the function correctness and the performance index of the target algorithm IP by using the actually measured data and the software operation result. The method has the advantages that effective testing and optimization of the heterogeneous computing architecture of the ISP module are achieved through the software and hardware co-simulation and IP core verification technology, middle-level optimization, a comprehensive testing framework and other methods, and powerful support is provided for technology development, application and popularization in related fields.

The invention relates to a parallel test system based on a large model, which is characterized by comprising a test case analysis system and a case dependency relationship model service, the case dependency relationship model service comprises an inter-case dependency relationship service, and the inter-case dependency relationship service comprises an inter-case dependency relationship service interface and a prediction model. According to the method, the dependency relationship among the use cases is accurately obtained through a large model technology, test use cases or processes without the dependency relationship can be better scheduled and executed in parallel based on the dependency relationship, test resources are utilized to the maximum extent, and repeated operation can be avoided when a plurality of use cases have the same preconditions.

The invention discloses a database query method, device and system and a medium, and the method comprises the steps: receiving a database query statement, analyzing the query statement to generate an execution plan of a tree structure, segmenting the execution plan, and obtaining a plurality of execution plan subtrees; matching the plurality of execution plan sub-trees with the intermediate result set in the cache, and determining whether intermediate result set sub-trees equal to the execution plan sub-trees exist in the intermediate result set or not; if an intermediate result set sub-tree equal to the execution plan sub-tree exists in the cache, replacing the successfully matched execution plan sub-tree with the equal intermediate result set sub-tree, and generating a replacement execution plan; and performing scheduling execution according to the replacement execution plan to obtain a corresponding database query result. The subtrees in the execution plan are matched and replaced through the intermediate result set in the cache, so that the intermediate result set in the cache can be directly used in subsequent execution operators, resource consumption and execution time are reduced, and database query performance is improved.

The invention relates to the technical field of computers, and discloses a distributed integration-based adaptive software and hardware collaborative scheduling system and method.Working data is acquired through a sensing acquisition module, and system states are distinguished and task requirements are predicted through a sensing processing strategy; the resource management module dynamically schedules task resources of software and hardware according to the system state and the task requirement and generates a scheduling result of the task resources, and through refined resource management, the scheduling system can dynamically allocate and schedule the resources according to the actual task requirement and the system state, the resource utilization rate is increased, and waste is reduced; the task scheduling module determines a task scheduling strategy according to the scheduling result of the task resources; the control execution module converts the task scheduling strategy into a control signal and drives the automation equipment to execute the task; and the feedback adjustment module obtains a task execution result and a system state of the automation equipment, generates a feedback signal through a signal generation strategy, and dynamically adjusts a task scheduling strategy.

The embodiment of the invention provides a task testing method, computer equipment and a storage medium, and belongs to the technical field of computers. The method comprises the steps that test task information is obtained, a test task set is generated, and the test task set comprises at least one test task; when it is detected that the test environment is released, allocating the test environment to a current test task to be executed in the test task set; and based on the task step-by-step test information corresponding to the current test task, performing step-by-step test on the current test task in the test environment to obtain an execution result of the current test task. According to the technical scheme, the execution efficiency of the test task is improved.

The invention belongs to the technical field of distributed account books, and particularly relates to a performance optimization method of a distributed account book system, which comprises the following steps: step 1, node architecture optimization: improving the processing capacity, memory and storage of nodes, reasonably configuring the load of the nodes, and avoiding overload of some nodes; and step 2, consensus algorithm improvement: the workload proof is converted to the right proof or other more efficient consensus mechanism, and the communication complexity and the calculation amount in the consensus process are reduced. According to the distributed account book system, the performance, the efficiency and the safety of the distributed account book system can be comprehensively improved by improving the node performance, improving the consensus algorithm, optimizing the transaction processing and the network layer, saving the storage space, improving the intelligent contract, simplifying the protocol layer and considering the safety and the performance balance at the same time, and the more efficient and reliable distributed account book system is provided for users.

The invention discloses a compiling method, device and equipment and a storage medium, and belongs to the field of compilers, and the method comprises the steps that an instruction sequence in a loop BB is circularly unfolded, and a triple BB and a triple DAG corresponding to the triple BB are generated; the triple BB comprises an instruction sequence circulating for three times; the triple DAG is used for indicating a data dependency relationship among instructions in the triple BB; window scheduling is carried out based on the triple BB, and a target scheduling window is determined; the window scheduling is used for dividing an instruction sequence in the circulating BB into two parts according to a scheduling window; the target scheduling window is a scheduling window of which the execution time meets a time condition, and the execution time is time required for executing an instruction in the window and calculated according to a data dependency relationship in the triple DAG; and according to a scheduling result corresponding to the target scheduling window, compiling to obtain a target code of the circulating BB in the target program. According to the method, the compiling and executing efficiency of the loop BB can be improved.

The invention relates to the technical field of RISC-V instruction set processors, in particular to an RISC-V-based AI processor design method for small convolution kernel acceleration, which comprises the following steps: determining the acceleration demand of a 3 * 3 convolution kernel in a convolutional neural network (CNN), selecting an open architecture RV32I of RISC-V as a basis, defining a core architecture of an AI processor, comprising a calculation unit, a register file and a control unit, the processing of the small convolution kernel is optimized; designing and realizing a vector arithmetic logic unit (VALU) and a convolution operation unit (CUU); the VALU has parallel computing capability and can execute addition, subtraction and multiplication vector operations to support instruction set extension; a convolution arithmetic unit CUU processes 3 * 3 convolution operations, including dedicated hardware logic to perform element-by-element multiplication and accumulation operations, thereby accelerating the convolution calculation process.

The application discloses a memory fault detection method and system based on a four-cell coupling model, and relates to the field of memory fault detection.In the method, four adjacent memory cells arranged in a square are taken as a basic test unit, four data backgrounds, namely, a full zero background, a horizontal alternating background, a chessboard background and a double-row alternating background, are generated by a preset or dynamic generation method, and an optimization strategy of performing read and write operations on only the cells needing to change states when the data background is switched is adopted, so that full coverage detection of various non-associated static fault types, including single-cell faults and various coupling faults such as adjacent double cells, diagonal double cells, right-angle type triple cells and square type four cells, can be realized.The complete test sequence complexity of the algorithm is low, the test time is reduced under the premise of ensuring high fault coverage, and the test efficiency of high-density memory is effectively improved.The algorithm has high flexibility and configurability.

A method for facilitating playing a legacy game is described. The method includes receiving user input during the playing of the legacy game, determining whether one or more code blocks for servicing the user input are cached, and accessing one or more instructions of legacy game code upon determining that the one or more code blocks are not cached. The method also includes compiling the one or more code blocks from the one or more instructions of the legacy game code, caching the one or more code blocks, and executing the one or more code blocks to display a virtual environment.

The embodiment of the invention discloses a cloud computing task scheduling method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a cloud scheduling data set which comprises a cloud computing task data set and a virtual resource data set; according to the cloud scheduling data set, a cloud computing task execution time model is constructed, and the cloud computing task execution time model is used for indicating the expected time of processing the cloud computing task by the virtual resources in the scheduling scheme; constructing a cloud computing task scheduling optimization model according to the cloud computing task execution time model; according to a Q learning algorithm and the cloud scheduling data set, a primary population of a bacterial foraging algorithm is determined, and the primary population represents an initial solution of the cloud computing task scheduling optimization model; the primary population is optimized according to a bacterial foraging algorithm, an optimal solution is obtained, and the optimal solution indicates an optimal scheduling scheme; and scheduling each cloud computing task according to the optimal solution of the cloud computing task scheduling optimization model. The resource utilization rate and the calculation efficiency can be improved.

The present invention discloses a device snapshot polling adaptive task scheduling method and device, the method comprising the following steps: obtaining real-time status information of the device by performing snapshot polling on the device, the real-time status information including the real-time load and remaining resources of the device, when there is a new task, calculating the priority of the new task according to the time dimension requirement and importance requirement of the new task to update the task queue and selecting the highest priority tasks in turn to allocate to the device that meets the resource demand and has the lowest load, monitoring the execution status of the task and the real-time status information of the device in real time, calculating the device resource utilization rate according to the real-time status information of the device, calculating the task execution efficiency according to the execution status of the task, evaluating the performance indicators of the previous task scheduling strategy through the device resource utilization rate and task execution efficiency and optimizing the task scheduling strategy, thereby improving the task execution efficiency and the device resource utilization rate.

The invention provides a test method and device and electronic device.The method comprises the steps that a target file under a target path is read, the target file is used for recording the execution process of at least one test case under a target test environment, and the execution process of the at least one test case is executed; the parameter value of the operation method involved in any case step in the test case is at least one of a set value, a value of a predefined variable corresponding to the target test environment and a value of a user-defined variable set by a value obtained by any case step before the case step; analyzing each test case recorded in the target file to generate test data corresponding to each test case; and according to the operation method in the test data corresponding to each test case and the parameter value of the operation method, driving the execution of each test case. Therefore, the use threshold can be reduced, the test case can better adapt to different test requirements and environment changes due to the fact that the user predefined variables and the user-defined variables are supported, and the test flexibility is improved.

This invention discloses a cloud platform computing power resource allocation system and method based on load balancing, relating to the field of computing power resource allocation technology. The application includes: constructing a computing power supply and demand matching platform, connecting with idle computing power resource providers, and receiving computing power requests from computing power resource lessors; based on the computing power requests from the lessors, determining the cluster time, and allocating idle computing power resources to the lessors in the cluster; the lessors setting up key communication, and after key verification, the system accessing the cloud platform of the allocated lessor, achieving intelligent decision-making and optimizing the scheduling of allocated idle computing power resources in the cloud platform. This invention can maximize overall resource utilization, reduce operating costs, minimize task execution latency, and improve user experience. Simultaneously, dynamic allocation based on computing power sharing and cluster processing according to different time nodes improves the overall utilization efficiency of computing power resources.

This utility model discloses a gripping and rotating coordinated apple-picking end effector, installed at the end of a picking robotic arm, comprising a cylindrical housing, a linear motor, a transmission mechanism, and a gripper assembly. The transmission mechanism converts the linear motion output by the linear motor into linear and rotational motions. The gripper assembly opens and rotates via an upward motion, and closes and rotates in the opposite direction via a downward motion to twist and break the fruit stem. The transmission mechanism uses a pin sleeve in cooperation with a central groove shaft, and achieves step-by-step movement through a spiral groove and a limiting groove structure. The gripper assembly completes clamping and twisting through a guide ridge and a hinge seat. This design simplifies the drive structure, reduces costs, improves reliability and picking efficiency, and is suitable for apple picking scenarios with short stems and dense branches.

The invention relates to a chained SVG control system and a configuration method thereof, and belongs to the technical field of power electronics. The method is used for configuring a chained SVG control system of which a controller adopts a heterogeneous dual-core processor architecture based on an FPGA (Field Programmable Gate Array), and comprises the following steps: a control algorithm in the chained SVG control system is realized by an FPGA core, and a control logic in the chained SVG control system is realized by another core; the control algorithm comprises a virtual synchronous control function, commutation chain voltage sharing, control instruction generation and harmonic suppression; the control logic comprises chain type SVG control system start-stop control, unlocking and locking sequence control logic, chain type SVG control system state conversion and main and standby system switching. The control system comprises the controller, the controller adopts a heterogeneous dual-core processor architecture based on the FPGA, and the controller is configured by adopting the method. According to the method, on the premise that control hardware configuration is not changed, the execution time of a control algorithm is greatly shortened, and the cost is low while the control capability of the controller is improved.

The embodiment of the present invention discloses a resource scheduling method and device for a job chain, which constructs a data set for representing the resource allocation of each job chain on a database cluster; establishes a cost function for each job chain to allocate database resources; establishes a fitness function based on the data set and the cost function; solves the minimum value of the fitness function through a particle swarm algorithm, and determines the resource allocation plan corresponding to each job chain according to the minimum value; and allocates database resources to each job chain according to the resource allocation plan. In the embodiment of the present invention, the optimal job chain resource allocation plan is solved through a particle swarm algorithm to avoid the waste of database resources caused by job chains competing for resources, which can greatly improve the utilization rate of database resources. Moreover, each job chain can be allocated appropriate database resources, thereby improving the overall execution efficiency of the job chain and reducing the overall execution time of the job chain.

The invention belongs to the technical field of artificial intelligence, and relates to a target detection method and device, computer equipment and a storage medium, and the method comprises the steps: obtaining a playback image set and a pre-trained target detection model which comprises a front layer, a playback layer and a full connection layer; during forward propagation, inputting a training playback image to a front layer, extracting image features, obtaining an activation value corresponding to the training playback image from playback storage, connecting the image features and the activation value in a playback layer, inputting a connection data structure to a full connection layer, and determining a category result of a target object; and during back propagation, calculating a gradient according to a category result, updating a weight parameter of the full connection layer, and obtaining a trained target detection model. In addition, the invention also relates to a block chain technology, and data such as a target detection model can be stored in a block chain. According to the method and the device, disastrous forgetting of the model can be overcome, and accurate and efficient object detection can be realized on the edge equipment with limited resources.

The invention relates to a domestic operating system kernel backup and recovery method based on a domestic computing platform, and belongs to the technical field of trusted computing. According to the method, the design of domestic BIOS firmware is optimized, a temporary operating system in a current general solution in the industry is removed, and functions such as kernel measurement, acquisition backup and kernel recovery after measurement failure of the operating system are all submitted to the BIOS firmware to be responsible for, so that unbinding with the operating system is realized. And after the BIOS firmware finishes the kernel recovery of the operating system, the operating system does not need to be restarted, so that the execution time of the kernel recovery process is greatly shortened.

The invention relates to the technical field of approximate network-on-chip, discloses a convolutional neural network acceleration-oriented high-real-time approximate network-on-chip, and provides an approximate compression and segmented recovery scheme based on data classification in combination with the data distribution condition of a neural network, so that the reasoning precision of the neural network is kept while the communication overhead is reduced; a quality control model is provided, and the model can quickly search threshold values of all layers of the neural network meeting quality loss constraints, so that the network is ensured to keep accuracy while performing approximate communication; a threshold value dynamic adjustment scheme based on congestion sensing is further provided, and the approximate threshold value of each layer is dynamically adjusted by sensing the congestion condition in the network, so that the real-time optimization of the data compression rate is realized, and the congestion problem in the communication network is effectively relieved; and in combination with activated data distribution characteristics in a network after approximate communication, an approximate calculation module in the calculation array is optimized, the overall efficiency of the calculation array is improved, and calculation acceleration is realized.