13results about How to "Avoid double loading" patented technology

The invention discloses a page picture resource progressive loading method, device and system and a storage medium, and relates to the technical field of data processing. The method comprises the steps that a timer event is set, and when a page is loaded, picture resources are searched in advance through the timer event; monitoring the page, and when the page changes, searching for picture resources needing to be asynchronously loaded; when the picture resources needing to be asynchronously loaded are found, the picture resources needing to be asynchronously loaded are replaced with thumbnails, and asynchronous loading is prepared; and monitoring the intersection condition of the picture resource and the visual area of the page, and when the picture resource is intersected with the visual area, asynchronously loading the original resource of the picture. By adopting the method provided by the embodiment of the invention, the data volume when the page is loaded for the first time is reduced, so that the loading speed is increased; repeated loading of the same picture resource is avoided, and network requests and bandwidth occupation are reduced; therefore, the best loading experience can be obtained no matter how the user interacts with the page.

The invention relates to the technical field of Internet of Vehicles target detection, in particular to an Internet of Vehicles multi-scene target detection system and method based on an RISC-V architecture, and the method comprises the steps: S1, obtaining image data, storing the image data in a DMA cache region preset on an RISC-V hardware calculation module, and setting the DMA cache region to be of a fixed size; s2, mapping the image data in the DMA cache region to an address space of a process on an RISC-V hardware calculation module by utilizing an mmap mechanism; s3, preprocessing the image data by using an RVV instruction; s4, after target detection model parameters are loaded for the first time, caching the preprocessed image data into a heterogeneous shared memory by using a cache consistency protocol supported by an RISC-V hardware calculation module, performing model reasoning, and obtaining perception information; s5, performing post-processing on the model reasoning result through a self-defined post-processing function, and obtaining display information; and the real-time frame rate of target detection, the overall performance of the system and the memory access efficiency are improved.

The invention belongs to the related technical field of integrated circuit design, and particularly relates to a method for loading data required by an accelerator operation processing unit of a convolutional neural network, which comprises the following steps of: controlling a cache sequence of weight data and feature map data through a state machine, when controlling the cache of the feature map data, firstly performing input feature map format conversion, the method comprises the following steps: performing sub-segment division on external channel-by-channel input feature map data with the bit width of A bit, storing the external channel-by-channel input feature map data with the bit width of A bit in a grouping dislocation storage mode, and converting the stored feature map data into N paths of parallel data with the bit width of B bit (B = A / N) in a special reading mode, the B * N data flow can complete feature map data loading of N storage units at the same time, the problem that the off-chip data access bit width is not matched with the data bit width needed in the operation processing unit is solved, and the bandwidth utilization rate of the convolutional neural network accelerator is improved.

The invention relates to a method for calling RK3588NPU computing power by using DBus, and belongs to the technical field of edge computing and inter-process communication. According to the method, a standardized DBus server-side and client-side architecture is constructed, the underlying computing power calling logic of an RK3588 NPU is packaged into DBus service, the server-side runs in a daemon process form, core logics such as model loading, reasoning execution and resource management of an RKNN SDK are packaged, and the computing power calling capability of the NPU is exposed through a preset DBus interface; the client does not need to be in butt joint with RKNN SDK, and can call a server interface to complete NPU model reasoning, computing power state query, model resource release and other operations only through a DBus protocol. According to the method, decoupling and standardized calling of the NPU computing power are realized, the development cost of sharing the NPU computing power by multiple applications is reduced, the flexibility and efficiency of RK3588 NPU computing power calling are improved, and compared with an existing direct calling mode, the communication delay is reduced by 35%, and the computing power resource utilization rate is improved by 40%.

This application discloses a storage system, a control method for the storage system, and an apparatus for the storage system, relating to the field of computer technology. The storage system includes a node scheduler and multiple processor nodes, each processor node including a processor core, a processor port, and candidate storage devices. The node scheduler is used to allocate a target processor node to a target request based on the received information of the target request and the target mapping relationship. The target processor core is used to determine the operation position of the target request based on the request information of the target request and to generate a target instruction based on the target request and the operation position. The target storage device is used to execute the target instruction and generate execution information of the target instruction. The target processor core is also used to update the metadata database based on the execution information and send the execution information to the sending interface. This application solves the technical problem of low response rate of the storage system to read and write requests, and achieves the technical effect of improving the response rate of the storage system to read and write requests.

The application provides a unified evaluation method and device for a multi-control paradigm video world model, wherein the method comprises the following steps: based on unified case metadata of a to-be-evaluated model, an evaluation request carrying a control condition is sent to a unified service interface, and the unified service interface distributes the evaluation request based on a deployment mode of the to-be-evaluated model; and an evaluation video generated by calling the to-be-evaluated model is received by a remote server or a local server. By using unified case metadata and a unified service interface, the evaluation request is distributed to the remote server or the local server according to the deployment mode of the to-be-evaluated model, a unified evaluation entry is provided for various video world models of a control interface, the local server is designed as an isolated resident inference service started by a sub-process, the independence of the evaluation environment of each open source model is ensured, the model weight is loaded once when the service is started, repeated loading of a huge model weight is avoided, and the expansibility and automation level of the evaluation system are improved.

This invention relates to the field of byte processing technology, providing a method and apparatus for accelerating byte encryption and decryption. The invention adds operation bits to bitwise operation instructions; uses corresponding operation bits to perform byte selection, endianness conversion, single-input multiple shift processing, and / or efficient table lookup operations. Specifically, byte selection avoids invalid bit calculations, improving instruction utilization. Endianness conversion ensures correct data order, minimizing duplicate data requests due to order errors, thus reducing the need for bitwise operations and improving GPU bandwidth utilization. Single-input multiple shift processing reduces the number of shift instructions, increasing throughput and avoiding repeated loading of the same data; efficient table lookup operations improve bitwise computation throughput, significantly reducing bandwidth pressure. This invention optimizes the granularity of bitwise operations, enabling encryption and decryption algorithms to efficiently utilize computing resources on the GPU while minimizing bandwidth consumption.

The invention discloses a model training method and system based on artificial intelligence, a medium and a product, and relates to the field of data processing systems or methods specially suitable for administrative, commercial, financial, management, supervision or prediction purposes. Tasks with the same preposed continuous logic units are aggregated into a practical training task group, the preposed units serve as common baseline workflows to be separated from the tasks, and remaining independent training logics of the tasks are converted into personalized derivative workflows. A first computing resource is allocated to perform single centralized execution on the common baseline workflow to generate a reference context slice. And distributing a second computing resource for each derivative workflow, distributing the reference context slice for environment reconstruction, and executing each personalized derivative workflow in parallel, thereby completing concurrent training of multiple tasks. According to the method, the computing power redundancy consumption caused by high-similarity training tasks is reduced, and the overall efficiency of task scheduling is improved.

The present invention discloses a method, system and server for loading configuration information, wherein the method applied to the main process includes: receiving request information about a target domain name sent by a working process, and querying the status identifier of the target domain name; if the status identifier represents a loaded state, feeding back the loading information of the target domain name to the working process, so that the working process provides an acceleration service based on the loading information; if the status identifier represents a loading state or an unloaded state, writing the process identifier carried in the request information into a request linked list of a shared memory, and loading the configuration information of the target domain name into the shared memory; after providing the loading result to the working process represented by the process identifier, clearing the request linked list. The technical solution provided by the present application can save the memory consumption of the system.

The present disclosure provides a loading method and device of an object resource, a storage medium and an electronic device. The method comprises: in response to a loading request associated with a target object resource, searching for a target resource identifier used for indicating the target object resource in a cache pool matched with a current page, wherein the cache pool comprises a plurality of resource identifiers, and the resource identifier is used for indicating an object resource loaded in a memory area; in a case where the search result indicates that the target resource identifier is not included in the cache pool, obtaining a target compressed resource matched with the target object resource; loading the target object resource obtained by decompressing the target compressed resource to the memory area; and adding the target resource identifier to a target cache location in the cache pool, wherein the target resource identifier at the target cache location is to be preferentially traversed in an identifier search operation. The present disclosure solves the technical problem of low efficiency of a related resource loading method.

This application relates to a document processing method, apparatus, computer equipment, and storage medium, applicable to Enterprise Resource Planning (ERP) systems, including but not limited to manufacturing management systems, supply chain systems, human resources systems, financial systems, tax systems, and auditing systems. The method includes: when a document is opened, loading and caching the document data, returning the main document pagination data from the document data, so that the main document pagination data is displayed on the main document interface after the document is opened; the main document interface has at least one corresponding sub-document interface, different sub-document interfaces are used to display different categories of data in the document data; when a target sub-document interface is triggered to display, retrieving the sub-document pagination data under the target category corresponding to the target sub-document interface from the cached document data, so that the sub-document pagination data under the target category is displayed on the target sub-document interface. This method can reduce computer overhead.

The invention relates to the technical field of neural networks, in particular to a pulse neural network accelerator based on quaternary pulses and a calculation method thereof, and the specific method comprises the steps: constructing a neuron model, presetting three membrane thresholds, carrying out quaternary pulse generation trigger judgment according to the real-time membrane potential of the neuron and the three membrane thresholds, and calculating the pulse neural network accelerator based on the quaternary pulses. Meanwhile, executing a zero-skipping strategy in a quaternary pulse generation process, and screening to obtain an effective quaternary pulse; after neurons in the neuron model generate and give out effective quaternary pulses, resetting membrane potentials of the neurons through a membrane potential resetting strategy; synchronously using the input data with the quaternary pulse to train the deep residual pulse neural network with the quaternary pulse, and generating a final output membrane potential and the quaternary pulse through an additional summator; according to the method, the processing time window of the SNN is remarkably shortened, the calculation efficiency and energy efficiency are improved, and the method is suitable for various neural network calculation tasks.