24 results about "Asynchronous computation" patented technology

The invention provides a DCU-based high-performance sparse stiffness matrix vector multiplication method, which comprises the following steps of: according to a sparse stiffness matrix, dividing a non-zero element into a plurality of calculation unit blocks by rows, pre-loading non-zero element data to an L1 shared memory or a register file through an on-chip shared memory controller of the DCU, a high-bandwidth crossbar switch of the DCU is used for realizing data copying and transmission; starting multi-row fusion execution for a short row of which the row non-zero element is lower than a DCU single-instruction multi-data width threshold value; constructing a wavefront scheduler based on a DCU asynchronous computing engine: binding an independent instruction cache region for each wavefront, and loading a multiply-add operation instruction set in advance through a prefetch instruction queue; a calculation unit state register is established, and when a wavefront scheduler ready signal is triggered, a scalar unit of the DCU is activated to execute calculation; and realizing cross-thread block reduction by adopting a DCU atomic operation accelerator. According to the method, the calculation throughput and the memory bandwidth utilization rate of large-scale structural mechanics stiffness matrix vector multiplication are effectively improved.

The invention relates to a dynamic scheduling method and device for a power grid network frame topological structure based on graph calculation and a medium, and relates to the technical field of power grids, and the method comprises the steps: S1, collecting the multi-source data of a power grid, and constructing an initial power grid topological graph; s2, performing hierarchical processing on the power grid topological graph to obtain a plurality of sub-topological graph layers, establishing independent asynchronous calculation units, and monitoring state changes of power grid equipment in a jurisdiction area of the corresponding sub-topological graph layer by each calculation unit to generate update information; s3, predicting the change trend of the power grid topological structure in different time scales and space ranges based on the update information of each calculation unit, and generating a space-time prediction topological model; and S4, inputting the space-time prediction topology model into a decision generation module based on Monte Carlo tree search, generating a scheduling instruction, feeding back the scheduling instruction to the power grid topological graph, and performing verification and optimization until a target effect is met. Compared with the prior art, the method has the advantages of improving the corresponding scheduling speed and the like.

The invention discloses a dynamic building and scheduling system and method for a digital twin water conservancy scene. The system comprises a data processing module used for obtaining a standard geographic space data service and a standard three-dimensional model data service; the preliminary scene building module is used for obtaining a preliminary scene according to data of the standard geographic space data service in the scene template standard three-dimensional model data service; the scene rendering scheduling module is used for obtaining triangular mesh data from the vector data through an asynchronous calculation thread, pushing the obtained triangular mesh data to a rendering thread through a main thread in a framing manner to execute rendering operation, and obtaining a basic scene; and the scene editing and storage module is used for editing and structurally storing the basic scene. Through AI remote sensing identification, asynchronous rendering scheduling and structured storage, the problems of low customization development efficiency, difficulty in scene reuse and large-scale data rendering lagging in traditional digital twin water conservancy scene construction are solved, and the water conservancy management efficiency and the decision precision are improved.

The invention discloses an audio frequency spectrum fluidization interaction presentation method and system based on shader computing power, and belongs to the technical field of image processing, and the method comprises the steps: obtaining an audio stream and user interaction data extraction features in real time; generating a sound interaction total dynamic field in the GPU parallel computing shader through nonlinear coupling; inputting the field into a preset neural network model in a fragment shader for forward reasoning to complete fluid neural physical evolution; separating an audio semantic layer in the rendering pipeline and executing competitive visual attribute mapping to generate an image; and performing asynchronous scheduling on each processing pipeline through an asynchronous calculation engine. According to the method, full-pipeline GPU calculation and neural operator evolution are adopted, the contradiction between fluid simulation and real-time rendering can be effectively solved, frequent data transmission is abandoned, low-delay interaction feedback is achieved while the physical reality sense is guaranteed, and the immersion sense and expressive force of audio visualization are remarkably improved.

The application discloses a kind of based on Internet of Things distributed edge computing system and method, belong to edge computing and intelligent sensing technical field, by constructing node level three-domain coupling model, realize the collaborative modeling of original sampling sequence, clock signal and electromagnetic environment disturbance data;Further identify pseudo-synchronous interval and generate phase distortion indicator function, reconstruct the phase of data;On this basis, extract phase collapse entropy gradient parameter and cross-node phase hysteresis loop area parameter, construct topological mapping matrix to depict phase collapse propagation path;Task dynamic decomposition and path avoidance are realized in combination with the topological mapping matrix, and phase distortion compensation is introduced in asynchronous computing process;Finally, output high-confidence state evaluation results through reverse consistency check and fusion;The application can effectively inhibit the influence of phase collapse, improve the accuracy and stability of distributed edge computing.

This invention discloses a blockchain state management method and system based on delayed writes and a sliding window state tree. The method includes, in response to the execution of the current block, applying the generated state update to the sliding window state tree in memory and generating a local state commitment stored in the current block header; and, in response to the fulfillment of preset conditions, asynchronously calculating the global state root based on the accumulated state updates in the sliding window state tree and disclosing it in the header of subsequent blocks. This invention introduces a TCP-inspired congestion control strategy, adaptively adjusting the window span vector according to the number of active leaf nodes. By decoupling state root calculation from block verification, reducing disk I / O operations, and dynamically adjusting memory load, it significantly improves the transaction processing throughput, node verification efficiency, and attack resistance of the blockchain system, making it suitable for public blockchains, consortium blockchains, and permissioned blockchain environments.

The invention discloses a high-throughput split federated learning method based on an activated cache. The method comprises the following steps: carrying out system initialization and network structure variable definition; performing client parameter synchronization, local feature extraction and independent parameter updating; performing activation feature drift asynchronous calculation and maximum old-age window updating; performing double-branch independent training of the server side based on hybrid retrieval; global federal aggregation based on data volume weighting is carried out; and performing a multi-server integrated reasoning test of the global model. Through an innovative drift sensing activation reuse mechanism and a multi-GPU communication-free integrated training strategy, the throughput performance of the split federated learning system in heterogeneous equipment and a non-independent identically distributed data environment is remarkably improved.

The application discloses a multi-direction privacy calculation method based on a smart contract, comprising a login platform and a cloud data processing terminal, through cooperation between a confusion function module and a smart contract module, a synchronous and asynchronous multi-direction privacy calculation is established between participants in a confusion circuit mode, synchronous calculation and asynchronous calculation are used for trusted transactions without a third party, the transactions are traceable and irreversible, data security is facilitated when participants formulate a contract, data value can be explored at any time, through cooperation between the cloud data processing terminal, a data source gateway module and a data encryption module, participant privacy can be encrypted and protected and participant information can be stored through big data screening, synchronous and asynchronous multi-direction privacy calculation between participants is facilitated, and privacy calculation is facilitated according to online conditions of the participants.

The invention relates to a key technology in the field of computer image processing and data recovery, in particular to an asynchronous calculation method for a matrix singular value and a singular vector of the matrix singular value. The matrix singular value pure CPU asynchronous calculation method specifically comprises the following steps: uniformly partitioning an input matrix into a plurality of sub-matrixes according to rows, carrying out asynchronous parallel QR decomposition iterative calculation on each sub-matrix on a multi-core CPU platform, and extracting a column orthogonal basis and a row orthogonal basis of each sub-matrix; recombining the row orthogonal basis of each sub-matrix and then carrying out QR decomposition iteration again to generate an intermediate matrix of which the dimensionality is obviously reduced; performing approximate singular value decomposition on the intermediate matrix; and finally obtaining a plurality of first singular values and singular vectors of the original matrix through matrix multiplication recombination. According to the method, the parallel computing capacity of the multi-core CPU is fully utilized, GPU hardware support is not needed, and compared with a traditional SVD algorithm, the computing speed is obviously increased.

A system and method are provided for synchronizing interdependent asynchronous computing environments. A synchronization engine dynamically coordinates state updates across independent subsystems, ensuring consistency and reducing computational inefficiencies. A workflow management method assigns structured tasks to users, validating completion before triggering dependent state transitions. A synchronization method resolves conflicts, reallocates resources, and propagates updates to maintain system-wide consistency.

The application provides a double closed loop collaborative evolution control method, device and medium for full heterogeneous subjects, aiming at the calculation asynchrony and model synchronization error problem of full heterogeneous subjects in a dynamic environment, an asynchronous interaction "execution-evolution" double closed loop architecture is constructed: the subject side uses a time sequence belief network to execute local control flow and realizes asynchronous calculation and synchronization reporting of experience trajectory; the collaborative controller uses a global feature cross attention mechanism to execute strategy evolution flow and triggers a global model synchronization instruction when monitoring an environmental performance deficit. The core innovation is: the evolution timing is triggered accurately by actively determining the environmental drift, and a time and space semantic review mechanism is introduced to forcibly verify the semantic coherence of the parameters issued before the subject hot update and the current physical characteristics. The application solves the timing logic conflict and semantic mismatch risk in asynchronous evolution of large-scale heterogeneous systems, and realizes dynamic calculation synchronization and stable model synchronization among full heterogeneous subjects.

This application relates to the field of intelligent driving technology, specifically providing a proximal policy optimization training method, system, storage medium, and vehicle. It aims to solve the technical problem of how to construct a PPO training system based on asynchronous computation without changing the PPO optimization objective and convergence characteristics. To this end, this application includes: a stage load-aware scheduling module that uniformly manages data transmission and model parameter version synchronization between the training data generation module, the policy and value joint sampling module, the reward and advantage calculation module, and the policy and value network optimization module, based on a delay consistency alignment mechanism. Through the delay consistency alignment mechanism, this application achieves effective control over asynchronously sampled trajectory data, reward values, advantage values, and model parameter updates, ensuring that the parameter update process conforms to the on-policy constraints of the PPO algorithm. Simultaneously, it can improve the parallel processing capability of the training system and guarantee the stability and convergence of the PPO training process.

This invention discloses an audio spectrum fluidized interactive presentation method and system based on shader computing power, belonging to the field of image processing technology. It includes: real-time acquisition of audio streams and user interaction data to extract features; generation of the total dynamic field of audio interaction through nonlinear coupling in a GPU parallel computing shader; inputting this field into a preset neural network model for forward inference in a fragment shader to complete the fluid neurophysical evolution; separating the audio semantic layer in the rendering pipeline and performing competitive visual attribute mapping to generate images; and asynchronous scheduling of each processing pipeline through an asynchronous computing engine. This invention employs full-pipeline GPU computing and neural operator evolution, effectively resolving the contradiction between fluid simulation and real-time rendering, eliminating frequent data transmission, and achieving low-latency interactive feedback while ensuring physical realism, significantly improving the immersiveness and expressiveness of audio visualization.

The invention relates to the technical field of photovoltaic equipment, and discloses a full-size mechanical analysis method for a flexible offshore floating photovoltaic structure, which comprises the following steps: establishing a floating photovoltaic geometric model, carrying out multi-scale structural discretization based on stress characteristics, and calculating local adaptive time step lengths and global time step lengths of various components; for all the units of the component type i, the repeated displacement vectors, the internal force and the external force are solved respectively; synchronizing the particles in global time; calculating and integrating the sum of external force and internal force vectors of the particles; and judging whether the Z coordinate of the central point of the cable net structure meets an iteration stopping condition or not to obtain the geometric configuration of the structure under the steady-state effect. According to the scheme, multi-scale structural discretization is carried out based on the stress characteristics of all the components, a self-adaptive time step mechanism and an asynchronous calculation mechanism are introduced, and efficient structural full-size mechanical response calculation is achieved.

The present application relates to the technical field of photovoltaic equipment, and discloses a flexible offshore floating photovoltaic structure full-size mechanical analysis method, comprising: establishing a floating photovoltaic geometric model, and performing multi-scale structure discretization based on stress characteristics, calculating the local adaptive time step and the global time step of various components; for all component types i units, respectively repeating displacement vector, internal force and external force solution; synchronizing particles on the global time; calculating and integrating the sum of particle external force and internal force vectors; judging whether the Z coordinate of the center point of the cable net structure satisfies the stop iteration condition, and obtaining the geometric configuration of the structure under the action of the steady state. The scheme is based on the stress characteristics of each component to perform multi-scale structure discretization, introduces an adaptive time step mechanism and an asynchronous computing mechanism, and realizes efficient calculation of the full-size mechanical response of the structure.

The application discloses a real-time rendering optimization method based on ray tracing. The application combines the optimization of spatial data structure, ray beam tracing technology, adaptive sampling, denoising technology, light map precalculation, shader optimization, ray-scene intersection acceleration, light source clipping, importance sampling, hardware acceleration, software-hardware cooperation, machine learning prediction, hybrid rendering technology, asynchronous calculation and dynamic resolution adjustment and other technologies; thereby improving the calculation efficiency of ray tracing and reducing the demand for hardware performance, so that the real-time or near real-time rendering speed is realized without sacrificing the image quality.

The present invention belongs to the field of computer systems based on specific computing models, and in particular relates to an interpretation method and interpretation system for predicting oil recovery rate based on multimodal feature perturbation analysis. The present invention introduces a variety of feature perturbation types, such as noise, scaling, shifting, random deletion, random interpolation, and nonlinear transformation, to comprehensively explore the potential relationship between features that improve oil recovery rate, breaking through the current limitation of focusing on a single feature in feature importance, evaluating the effect of multiple combination features on the output of the prediction model, and providing solid data support for the optimization and understanding of the model; by introducing asynchronous computing methods and using the MapReduce model for parallel processing, the computing efficiency is significantly improved. The present invention aims to use feature perturbation technology to optimize the oil production prediction model, improve the accuracy, and improve the interpretability of the machine learning model.

The application relates to a GIS multitask asynchronous processing and management method, which comprises the following steps: parameter configuration and structuring are carried out on a task creation interface; a mutually independent MQTT network publishing channel and a signal slot communication channel are synchronously enabled to distribute tasks and trigger state updating; a task processing server end is internally provided with task processing logic, after receiving the tasks, the asynchronous calculation process is simulated through a timer, after the calculation is completed, the result is published through MQTT, and the user interface state indication is updated in real time. The application can solve the problem that interface response delay and processing reliability cannot be considered in the existing GIS multitask processing method, can guarantee instant response of the user interface, and can ensure task processing reliability.

The invention belongs to the technical field of multi-modal similarity control. The invention provides a multi-mode affinity quantitative modeling and asynchronous calculation method. According to the embodiment of the invention, firstly, a cross-modal affinity matrix for asynchronous multi-modal data processing is constructed, and searching is preferentially carried out in a surrounding neighbor state set by taking a current data fusion state as a guide; and the similarity between the modals is quantified and normalized by using a hierarchical analysis method so as to adapt to input of different types and quantities. And then introducing a conditional generative adversarial network of a class attention mechanism. The network efficiently predicts and complements missing or delayed input data by using a conditional generative adversarial mode in combination with a class attention mechanism. According to the method, challenges brought by asynchronous data are handled, and through structured modal affinity control, the efficiency and precision of data fusion are improved, the deduction delay is remarkably reduced, and the real-time response capability of the system is enhanced.

The application discloses a kind of digital twin water conservancy scene dynamic building and scheduling system and method, comprising: data processing module is used to obtain standard geospatial data service and standard three-dimensional model data service;Preliminary scene building module is used to obtain preliminary scene according to standard geospatial data service in the scene template standard three-dimensional model data service data;Scene rendering scheduling module is used to obtain triangular mesh data after vector data is obtained by asynchronous computing thread, the triangular mesh data obtained by main thread is framed and pushed to rendering thread to execute rendering operation, obtains basic scene;Scene editing and storage module is used to edit and structured storage to basic scene.The application is identified by AI remote sensing, asynchronous rendering scheduling and structured storage, solves the problems that traditional digital twin water conservancy scene construction exists low customization development efficiency, scene is difficult to reuse and large-scale data rendering lag, improves water conservancy management efficiency and decision-making accuracy.

The invention belongs to the field of integrated circuits, and particularly relates to a double-edge time-to-digital converter, a time domain storage circuit and a CIM chip. The TDC comprises a symbol judgment module and a numerical calculation module. The symbol judgment module is used for respectively quantizing values of symbol bits of a first calculation result corresponding to a rising edge and a second calculation result corresponding to a falling edge according to an input calculation signal A and a reference signal B; the numerical calculation module comprises a shaping unit, a first delay chain, a second delay chain, a quantization unit, a transcoding unit and a separate storage unit. The method comprises the following steps: shaping A and B to obtain an OR signal C and an AND signal D; obtaining thermometer codes of numerical bits of two calculation results according to C and D through gradient delay of the two delay chains and XOR operation of the quantization unit, and finally completing data transcoding and asynchronous calculation result extraction through a transcoding unit and a separate storage unit. According to the invention, the problems of large area and high power consumption of the existing circuit are solved.

The application discloses a load balancing method and device, and relates to the technical field of cloud computing. The method comprises the following steps: acquiring parameter information of a service end; determining real-time weights of each service end, and pre-grouping the service ends according to the real-time weights; when a calling request is received, determining the number of available addresses in each weight group by using a dynamic routing result and a pre-grouping result; generating a weight calculation result by using the real-time weights of the weight groups and the number of available addresses corresponding to the real-time weights; and determining a target calling address of the calling request according to the weight calculation result. In the application, the pre-grouping method is adopted, and a large amount of repeated calculation work in the weight calculation process is moved to background asynchronous calculation, so that the workload of traversing addresses to obtain parameters during runtime is removed, the application can be run with extremely low performance loss even when the cluster scale is particularly large, the weight calculation result is generated by combining the dynamic routing result and the pre-grouping result, and strong support is provided for preheat calculation in a large-scale deployment scenario.

The invention provides a public security-oriented Chinese text intelligent identification and classification method, relates to the technical field of network security text data natural language processing, and supports high-concurrency text stream processing by setting and introducing an asynchronous computing framework based on a GPU (Graphics Processing Unit); integrating a plurality of semantic perception models including context enhancement, entity recognition and event extraction for cooperative work; embedding a structured knowledge graph constructed by taking a case library of public security and the like and a judicial judgment abstract as a core into classification logic; a user-defined emotion dictionary is fused, and an emotion triggering map is constructed in combination with field hot events; and after the multi-level judgment is completed, a label cascade and level response strategy is triggered. The method can be used for text recognition in a professional field, and makes up for the defect that a general model lacks field support; moreover, the decision-making process has good interpretability, and strict compliance requirements can be met; moreover, the method can be effectively applied to various high-risk and real-time decision-making scenes.

The application provides a high-performance volume rendering method and system based on an autonomous controllable GPU pre-rendering. In the method, a curvature distribution feature of a voxel surface is extracted through dynamic topology analysis to construct pre-processing metadata, a reconfigurable computing array is driven to dynamically divide a hardware topology structure of a high-density light projection unit and a low-power interpolation unit according to the curvature feature, task allocation and collaborative execution of a heterogeneous computing unit are realized based on an asynchronous computing engine, finally, a hybrid weight algorithm optimized at a hardware level is used to fuse multi-scale sampling results in an autonomous controllable rendering pipeline, and a closed-loop feedback mechanism of a rendering result and the pre-processing metadata is established. The method innovatively deeply couples voxel geometry features, hardware architecture reconstruction and asynchronous computing scheduling, forms a technical closed loop of data feature-driven hardware resource configuration and rendering result-reversed optimization of pre-processing parameters. The technical scheme provided by the application realizes high-fidelity real-time volume rendering of medical images and scientific data.