14results about How to "Improve throughput efficiency" patented technology

The invention provides a neural network processing unit scheduling system and method, relates to the field of artificial intelligence accelerators, and can set a task progress register in a neural network processing unit for recording the task processing progress of the neural network processing unit for processing and calculating sub-tasks and exposing the task processing progress to a software layer. A scheduler and a global shared task queue can be arranged in the runtime module. The scheduler is used for scheduling the calculation subtasks to the neural network processing units, screening the slow processing units according to the task processing progress of the neural network processing units, determining the remaining subtasks which are not completed by the slow processing units, taking the remaining subtasks as the calculation subtasks again, and sending the calculation subtasks to the neural network processing units; and re-issuing to the idle neural network processing unit through the global shared task queue. Therefore, the remaining sub-tasks which are not completed by the low-speed processing unit can be re-issued to the idle neural network processing unit, so that idle computing resources can be avoided, and the processing efficiency is improved.

The application discloses a kind of optical module detection equipment, optical module detection method and multiple parallel detection methods, comprising: tray conveying track;Empty tray backflow track;Code scanner;First jacking positioning mechanism and second jacking positioning mechanism, sequentially set on tray conveying track along the conveying direction;First jacking positioning mechanism is used to lift and keep tray in temporary storage standby position, and after jacking, the tray is separated from tray conveying track;Second jacking positioning mechanism is used to lift and keep tray in working position, and after jacking, the tray is separated from tray conveying track;Two test workstations, each test workstation includes from front to back sequentially arranged: fiber plug module, optical module product placement positioning assembly, test board assembly;Multi-axis manipulator;Control system;It realizes tray circulation zero waiting, test workstation continuous operation, equipment comprehensive utilization maximization, improves the overall throughput efficiency of single equipment and multiple parallel production line.

The invention discloses a nuclear medicine storage suspension type robot carrying system, which comprises a suspension type rail network, a carrying robot running along a rail, an automatic low-temperature storage cabinet, a workstation and a central control system, and is characterized in that the carrying robot comprises a sliding seat main body which is arranged on the suspension type rail network in a sliding manner; the rotary dispatching mechanism is arranged on the sliding seat main body; the plurality of independent and closed storage bins are carried by the rotary dispatching mechanism; the mechanical arm is fixedly mounted on the front side of the sliding seat main body; the self-adaptive clamping mechanism is mounted at the output end of the mechanical arm; and the rotary dispatching mechanism is configured to rotate and adjust any storage bin to a fixed operation station corresponding to the self-adaptive clamping mechanism.

The application provides a double-mode interconnected driving simulation device of a liquid-gas combined tensioner, relates to the field of ocean engineering equipment testing, and comprises a translation mechanism, a rotation adjusting mechanism, a vertical movement and inclination simulation mechanism, a riser, a support and a combined tensioner mechanism. The application adopts a positive configuration of active control simulation at the bottom and passive response verification at the top. The device is sequentially composed of a bottom drilling platform translation mechanism, a rotation adjusting mechanism, a middle drilling platform vertical movement and inclination simulation mechanism, a top ring direct-acting liquid-gas combined tensioner at the top of a riser assembly and the like from bottom to top, so that the translation mechanism can simulate the horizontal drift movement of the ocean platform, the rotation adjusting mechanism can reproduce the yaw movement posture of the platform, the vertical movement and inclination simulation mechanism can simulate the inclination posture of the offshore platform under different sea conditions, and the combined tensioner mechanism can simulate the mechanical characteristics of the real wellhead connection of the tensioner.

This invention discloses a hybrid gated loop unit circuit based on memristors. The invention includes a memristor circuit, a weight mapping circuit, a hybrid gate circuit, a candidate activation state circuit, and an output circuit. The signal output terminal of the memristor circuit is connected to the signal input terminal of the weight mapping circuit. Relying on an in-memory computing architecture, it can directly complete the entire computation process of the hybrid gated loop unit and its variant models at the hardware level, eliminating the need for frequent data transfer and interaction. This fundamentally reduces the amount of data transmission, effectively breaking through the bottleneck limitations of data transmission in traditional architectures. It integrates the update gate and reset gate in traditional gated loop units into a single hybrid gate structure. Through the gate value output by the hybrid gate, the retention ratio of historical hidden information can be flexibly and dynamically adjusted, while precisely controlling the degree of fusion between new and historical information, making the information processing of the entire unit more efficient and dynamically adaptable.

The invention relates to the technical field of network transmission optimization, and provides a data transmission method and device. The method comprises the steps of receiving credit authorization information sent by a receiving end, wherein the credit authorization information is used for authorizing a sending end to transmit to-be-transmitted data based on a credit authorization amount; determining a transmission performance index of each data transmission path; determining a credit allocation weight of each data transmission path according to the transmission performance index; determining the target credit of each data transmission path according to the credit authorization amount and the credit allocation weight of each data transmission path; and based on the target credit of each data transmission path, distributing the to-be-transmitted data to the corresponding data transmission path for transmission. According to the data transmission method provided by the invention, multi-path data transmission under active control is realized, and the robustness of data transmission and the resource utilization rate are improved.

The application provides a spatial optical circuit switching array based on a MEMS adjustable metasurface and a control method thereof, and relates to the technical field of optical communication.The array comprises a transmitting end array and a receiving end array;the transmitting end array and the receiving end array each comprise two layers of transmission metasurface layers arranged in parallel;each transmission metasurface layer comprises a plurality of planarly arranged metasurface units;the transmitting end array modulates the phase of the normally incident light of any input optical port to generate outgoing light with a deflection angle;the receiving end array applies a translation compensation opposite to the transmitting end array to the outgoing light to make the outgoing light recollimate and couple to a target output optical port.The control method realizes non-blocking spatial light path reconstruction of multiple light paths concurrently through master-slave cooperative control within a limited displacement stroke.Under the condition of multiple light paths concurrently, the application realizes non-blocking spatial light path reconstruction and dynamic switching within a limited displacement range, and effectively avoids light path crosstalk.

The invention discloses a warehouse in-out automatic processing system based on RFID and data visualization, and relates to the technical field of warehouse management. According to the warehouse in-out automatic processing system based on RFID and data visualization, label information of goods, carriers and personnel equipment in a warehouse is collected through the RFID real-time sensing unit, a global event flow is formed, and the dynamic environment modeling unit constructs a digital twin model in real time based on the event flow and records the article position and the operation state. The task analysis and scheduling trigger unit compares a model with an operation plan, generates a scheduling request, cooperates with the intelligent decision-making unit to optimize scheduling calculation and generates a task and path instruction, the visual guide unit converts the instruction into a visual data packet, the interaction terminal guides an operator to execute the task, and the real-time operation is ensured through the verification and updating unit. Through the RFID and data visualization technology, the warehouse state is monitored in real time, resources are intelligently scheduled and optimized, and the warehouse in-out management efficiency is improved.

The invention is suitable for the technical field of artificial intelligence, and relates to a storage controller, a storage system, computer equipment and a storage access method. The memory controller comprises a controller chip and a memory chip, the random access memory chip is bonded on the controller chip, the random access memory chip and the controller chip adopt the same address coding rule, and the random access memory chip comprises a first random access memory space used for storing a logic address mapping table; the second random access storage space is used for storing command queues corresponding to different storage channels, each storage channel is connected with at least one storage chip, and the command queues are used for caching operation commands issued by the controller chip to the storage chips; and the third random access storage space is used for storing read data returned by the storage chip in response to the operation command or storing write data required to be written by the storage chip in response to the operation command. According to the method, the strict requirements of a large-scale neural network model on high-bandwidth and low-delay storage access can be met.

The invention discloses a dynamic resource regulation and control system based on computing power reservation and container mapping, relates to the technical field of resource regulation and control, and is used for solving the problems of response delay of a transcoding task in a peak period and idle waste of computing power resources in an off-peak period. According to task attributes, video frame rates and categories, batch transcoding tasks are allocated to different computing power periods for execution, during a computing power reservation effective period, resource mapping and execution monitoring are carried out on the operation tasks, task completion rates and resource occupation data are collected in real time, dynamic reallocation of the task operation periods is realized through computing power adaptation scores, and real-time transcoding is realized. In the whole task execution process, the matching relation between the task and the computing power time period is continuously corrected, so that the resource utilization state and the task processing efficiency are maintained in a more stable balance interval, and the balance of the overall throughput performance of the cluster and the computing power distribution is kept in the multi-type task concurrent environment.

This application provides a data preheating and caching method, apparatus, device, and storage medium, relating to the field of big data processing technology. The method includes: before the release of target information, identifying first hot data related to the target information to be accessed, retrieving the first hot data from a database and writing it to multiple cache nodes; determining a target node among the multiple cache nodes according to preset rules, and writing second hot data cached on the target node to local storage; and responding to a received request to access the second hot data, reading the second hot data from local storage and returning it. This method utilizes multiple cache nodes for preheating and caching, thus ensuring high system availability and improving throughput efficiency. Furthermore, after preheating and caching, the cached data is stored in local storage and read from local storage, further improving data retrieval efficiency. Therefore, it can accelerate the speed at which users obtain data and improve user experience.

The invention relates to the technical field of route distribution, and discloses a block chain node route distribution method, device and equipment and a medium, and the method comprises the steps: obtaining the historical load data of block chain nodes, and calculating the expected processing delay, the expected concurrent request amount and the network topology distance of each block chain node based on the historical load data; performing performance quantitative analysis on the expected processing delay, the expected concurrent request quantity and the network topology distance to obtain a processing delay metric value, a concurrent request metric value and a network topology distance metric value; determining a corresponding scene weight coefficient according to the business application scene of the block chain node; calculating a routing distribution cost characteristic value according to the processing delay metric value, the concurrent request metric value, the network topology distance metric value and the scene weight coefficient; and performing a routing distribution decision on the block chain node according to the routing distribution cost feature value. According to the invention, the accuracy of node routing distribution can be improved.

The invention discloses a data processing method and device, a nonvolatile storage medium and electronic equipment. The method comprises the following steps: acquiring a to-be-transmitted file list, and detecting a storage performance index of a storage system for storing the to-be-transmitted file list and a network bandwidth of a transmission environment; according to the total data volume, the storage performance index and the network bandwidth of the to-be-transmitted file list, determining the number of target fragments used for performing fragment processing on the file; extracting a plurality of to-be-transmitted files with the data volume smaller than a first preset threshold value from the to-be-transmitted file list, and combining the plurality of to-be-transmitted files into a logic block with the preset data volume in the to-be-transmitted file list to obtain a target to-be-transmitted file list; and according to the target fragment number, performing fragment processing on the target to-be-transmitted file list to obtain a target fragment. The technical problems of low transmission efficiency and too long task time consumption caused by the fact that the fragmentation scale cannot be dynamically adjusted according to network bandwidth fluctuation and storage system performance in the prior art are solved.

This invention discloses a VOCs recovery and treatment device and method for a color-coated steel sheet production line, relating to the field of waste gas treatment technology in color-coated steel sheet production. The VOCs recovery and treatment device for the color-coated steel sheet production line includes a shell, within which an adsorption assembly is rotatably installed. A second partition is also provided within the shell, and a first partition is symmetrically installed within the shell. The first and second partitions divide the internal space of the shell into an adsorption zone, a desorption zone, and a cooling zone. Activated carbon plates periodically circulate within the adsorption, desorption, and cooling zones, achieving automated circulation of the entire adsorption, desorption, and cooling process. Compared to traditional fixed-bed systems that require shutdown for regeneration, this significantly improves system throughput efficiency. The arrangement of the first and second partitions, combined with the first and second arc-shaped plates, creates strict temperature zone separation, ensuring that ambient temperature adsorption in the adsorption zone, high-temperature desorption in the desorption zone, and cryogenic regeneration in the cooling zone do not interfere with each other, avoiding cross-heat loss and guaranteeing the processing efficiency of each stage.