19results about How to "Reduce resource overhead" patented technology

The invention relates to a scheduling method which schedules deep learning jobs through a primitive-dual framework and a dynamic programming algorithm in a heterogeneous GPU cluster and can meet different deadline requirements of the jobs. According to the method, on-line analysis and historical data are combined in a heterogeneous GPU cluster environment, so that the overhead of obtaining the job throughput is reduced; according to the method, job scheduling and job placement are decoupled, the complexity of job scheduling is reduced, meanwhile, a simple and effective job placement scheme is designed, and the degree of resource fragmentation is effectively reduced; a scheduling target is modeled in a scheduling module, reward value functions are respectively designed for strict deadline operation and soft deadline operation, an optimization problem is constructed by taking maximization of an overall reward value as the scheduling target, and strict deadline requirements and soft deadline requirements of the operation are considered at the same time; and reconstructing the target optimization function into an integer linear programming problem, and solving by using a primitive-dual framework and a dynamic programming algorithm to ensure the theoretical effectiveness of the solution.

The invention discloses a virtual machine operation method, device and equipment, and the method comprises the steps: responding to at least one virtual machine in an operation termination state, if a virtual machine restart instruction is received, determining a first mirror image operation memory and a first mirror image backup memory which are pre-allocated for each virtual machine through a virtual machine controller; copying the virtual machine mirror images stored in the first mirror image backup memories to the first mirror image running memory through the virtual machine controller; and executing a virtual machine recovery instruction through the virtual machine controller, and triggering to jump to a kernel entry address of the first mirror running memory so as to recover running of the virtual machine. According to the scheme, restarting of the virtual machine does not need to depend on a privileged virtual machine, the virtual machine controller loads the virtual machine mirror image from the mirror image backup memory to the virtual machine running memory, and therefore independent restarting of the virtual machine is achieved.

This invention relates to a log access method, apparatus, and storage medium, belonging to the field of embedded software log management technology. Addressing the problems of existing embedded log access systems, such as complexity, high resource consumption, low access efficiency, and poor operability, this invention employs a fast access algorithm based on time-address mapping. It directly calculates the physical storage address of the log using time information, eliminating the need for a file system and mapping table, and requiring no additional data structures. This invention features a simple algorithm, high efficiency, and low resource consumption, enabling rapid log location and stable access. It is suitable for resource-constrained scenarios such as embedded devices, effectively balancing the reliability, timeliness, and operability of log access.

The application provides a PCIe link monitoring and self-repair system for an FPGA acceleration card, and the system comprises: a signal acquisition module having a main sampler anchored at a central sampling point and an offset sampler capable of biased sampling; a monitoring point calibration module for controlling the offset sampler to scan along a voltage axis and a time axis after the link is ready, determining an eye diagram boundary and calculating a monitoring sampling point; an online monitoring module for polling each monitoring point and interrupting when an error code exceeds a limit, and outputting a repair trigger signal; a repair decision module for determining a distortion type according to an abnormal point position and outputting a coding signal; and a parameter adjustment module for selecting a target equalizer according to the coding signal, and performing adaptive iterative adjustment with the monitoring result as feedback until the link returns to normal. The application realizes real-time monitoring, rapid diagnosis and closed-loop self-repair of the PCIe link signal under the premise of uninterrupted service, and significantly improves the stability and reliability of the system.

The application provides a unified systolic array computing unit for ANN and SNN, comprising a shift-add computing processing unit and a systolic array formed by arrayed deployment of the shift-add computing processing unit; an execution of artificial neural network operation or pulse neural network operation is switched through a configuration mode selection signal terminal; in the artificial neural network mode, multiplication operation is unfolded in the time dimension, and a shift register and an adder constitute a time division multiplexing operation sequence to complete multiplication and accumulation operation; in the pulse neural network mode, a weighting accumulation is directly completed through multiplexing of an addition link, and a neuron time step dependent chain is decoupled through a soft reset mechanism to realize cross-time step weight multiplexing; control signals and data are propagated in the same direction in the systolic array, which is used for realizing control logic array level multiplexing, and fixed data flow is adopted for the output, so that partial sums are preferentially resident in the array interior; the working frequency can be improved, the bandwidth and power consumption can be reduced, and the area efficiency and timing convergence can be improved.

The OTFS system channel estimation method based on hybrid multi-pilot frequency combination comprises the following steps: setting an OTFS transmitting frame which comprises a data symbol and a pilot frequency symbol which are simultaneously modulated to a DD domain; a DD domain sending signal is modulated by the OTFS and then reaches a receiving end through a channel, and a signal received by the receiving end is demodulated by the OTFS and then is converted into a DD domain receiving signal; the method comprises the following steps: receiving a signal based on a DD domain, and estimating time delay and integer Doppler of a channel by utilizing related characteristics of a pilot frequency sequence to obtain estimated values of the time delay and the integer Doppler; estimating decimal Doppler according to time invariance of a DD domain channel and adjustability of Doppler, and finally obtaining an estimated value of Doppler; performing LMMSE estimation on the attenuation coefficient according to the estimation values of the time delay and the Doppler to obtain an estimation value of the attenuation coefficient, and completing channel estimation of the OTFS system; according to the method, the pilot symbols and the data symbols are mixed, joint estimation is carried out, and accurate estimation of the OTFS decimal Doppler channel is achieved.

The application provides a time synchronization system and method based on a TDMA system wireless network virtual-real joint simulation, which comprises a physical network and a virtual network simulation system, time synchronization between multiple physical nodes of the physical network is completed through radio frequency signals, and wireless communication of the TDMA system is adopted; one of the physical nodes is a gateway node, the gateway node is provided with a first interface and a second interface, the first interface is used for generating a time slot pulse signal, and the second interface is used for generating a time slot numerical signal; the virtual network simulation system communicates with the physical network, and the virtual physical network system is provided with a third interface for receiving the time slot pulse signal and a fourth interface for receiving the time slot numerical signal; wherein the time slot pulse signal carries synchronization interrupt information, the time slot numerical signal carries time slot time information, and the virtual network simulation system realizes time synchronization with the physical network through the time slot pulse signal and the time slot numerical signal. The application has high time synchronization precision and low resource consumption.

The application discloses a virtual machine vulnerability detection method and device based on a large model and differential testing, and relates to the technical field of blockchain system security. The method comprises the following steps: S100, collecting multi-dimensional original data of a blockchain virtual machine, and constructing a standardized vulnerability knowledge base through natural language processing; S200, comparing bytecode streams before and after vulnerability repair, and calculating a bytecode semantic difference degree factor; S300, based on the standardized vulnerability knowledge base and the difference degree factor, constructing an engineering prompt word template by using a pre-trained large language model. Through high-targeted test case generation guided by the large model semantics, low-false alarm heterogeneous virtual machine differential consistency verification, and closed-loop self-optimization of the detection strategy driven by reinforcement learning, the pain points of traditional virtual machine vulnerability detection case generation, such as blindness, high false alarm rate and fixed detection capability, are solved, and the efficiency, accuracy and deep logic vulnerability detection capability of the blockchain virtual machine vulnerability detection are greatly improved.

The invention discloses a communication and navigation integrated implementation system and method of a low earth orbit satellite Internet of Things system. The system comprises a low earth orbit satellite module, a ground terminal receiving module, a positioning and parameter calculation module, an access control and compensation module and an uplink communication link establishment module. A positioning measurement signal is introduced into a satellite downlink NB-IoT (Narrow Band Internet of Things) or eMTC (Enhanced Machine Type Communication) signal, so that a ground terminal can simultaneously complete positioning data acquisition, optimal satellite selection and time and frequency compensation of an uplink signal, thereby effectively solving the problems of low access efficiency and difficulty in synchronization and compensation in the prior art. According to the invention, deep fusion of communication and navigation is realized, the access speed and reliability of the low-orbit satellite Internet of Things system are improved, and the method has good application prospect and popularization value.

The invention provides a PCIe link monitoring and self-repairing system for an FPGA accelerator card, and the system comprises a signal collection module which is provided with a main sampler anchored at a central sampling point and an offset sampler capable of performing offset sampling; the monitoring point calibration module is used for controlling an offset sampler to scan along a voltage axis and a time axis after the link is ready, determining an eye pattern boundary and calculating a monitoring sampling point; the on-line monitoring module is used for polling and sampling each monitoring point, interrupting when the error code exceeds the limit, and outputting a repair trigger signal; the restoration decision module is used for determining a distortion type according to the position of the abnormal point and outputting a coded signal; and the parameter adjusting module is used for selecting a target equalizer according to the coded signal, and carrying out adaptive iterative adjustment by taking the monitoring result as feedback until the link returns to normal. According to the PCIe link signal diagnosis system and method, real-time monitoring, rapid diagnosis and closed-loop self-repairing of PCIe link signals are achieved on the premise that services are not interrupted, and the stability and reliability of the system are remarkably improved.

The invention relates to a task processing method and device, equipment, a storage medium and a program product. The method comprises the following steps: in response to received task configuration information of a target periodic task, generating a current actual task corresponding to the target periodic task based on the task configuration information, sending the current actual task to a server, and in response to the current actual task being in a task end state, sending the current actual task to the server; and generating a current virtual task corresponding to the target periodic task on the basis of task interval time included in the task configuration information, generating a next actual task corresponding to the target periodic task on the basis of the task configuration information in response to triggering of a callback event corresponding to the current virtual task, and sending the next actual task to the server. By adopting the method, resource consumption can be reduced.

The application relates to the technical field of ECU management, and discloses an ECU boot dynamic self-recovery system, method and device based on an on-vehicle diagnosis communication protocol and physical sensing, which are characterized in that the technical scheme comprises a microcontroller, a nonvolatile memory, a memory protection unit and a power supply voltage detection unit; the microcontroller runs the boot guarantee module, the logic evolution module, the dynamic driving unit and the snapshot management and power-off solidification unit; the boot guarantee module is used for power-on priority startup, snapshot state reading, fault self-recovery and breakpoint continuation; the logic evolution module is used for executing the boot loader main body service; the dynamic driving unit is loaded into the running memory and is used for executing the erasing and writing operation on the updatable boot area and updating the snapshot information in the running memory in real time during the flashing process; the power supply voltage detection unit is used for triggering the power-off processing; and the snapshot management and power-off solidification unit is used for responding to the power-off signal and writing the snapshot information into the snapshot storage area.

A business processing system and method include: a business request center layer, a routing platform layer, and a business processing layer. The business request center layer includes at least one business request port corresponding to at least one business request center. Each business request port receives business requests from users within its own business request center and sends them to the routing platform layer. The routing platform layer includes at least one routing module corresponding to at least one business request center. Each routing module receives business requests sent from the business request ports within its corresponding business request center and routes them to the business processing layer according to routing information. The business processing layer includes at least one business processing module corresponding to at least one business processing center. Each business processing module receives business requests routed from any routing module and performs corresponding operations. This embodiment of the disclosure can reduce the number of servers deployed and save resource overhead.

Embodiments of this application provide a communication method, apparatus, and system, to reduce resource overheads for sending coordination information. The method includes: A first terminal apparatus determines coordination information, where the coordination information is used to assist a second terminal apparatus in determining a sidelink sending resource; and the first terminal apparatus sends the coordination information to the second terminal apparatus, where a sequence carrying the coordination information is mapped to a first time-frequency resource, the first time-frequency resource is a subset of a second time-frequency resource, and the second time-frequency resource and a physical sidelink feedback resource overlap in time domain, and are orthogonal in frequency domain.

Embodiments of the present application provide a data processing method and device, computer equipment and a readable storage medium. The method comprises: obtaining sample spliced media features and N-dimensional labeled interaction labels of a sample object for sample multimedia data; calling a shared network in an initial recognition model to extract shared media features associated with N media recognition tasks from the sample spliced media features; calling a dedicated network i in the initial recognition model to extract dedicated media features i from the sample spliced media features; calling a recognition network i in the initial recognition model to identify an i-dimensional predicted interaction label according to the shared media features and the dedicated media features i; and training the initial recognition model according to the N-dimensional labeled interaction labels and the N-dimensional predicted interaction labels to obtain a multi-task recognition model. The media recognition accuracy of the trained initial recognition model can be improved, and the push accuracy of the multimedia data can be improved.

A method, apparatus, readable storage medium, and terminal for exporting multi-source heterogeneous data are disclosed. The method includes: upon receiving a data export task, obtaining descriptive information of the data to be exported from the data export task. This descriptive information indicates the attribute information of the data to be exported and the various data sources from which it originates. The attribute information of the data to be exported is set based on preset metadata. A target execution engine is selected based on the various data sources from which the data to be exported originates and the current resource load of the execution engine. The data export task is converted into an executable task adapted to the target execution engine, and the executable task is distributed to the target execution engine. The target execution engine then performs data queries on the corresponding data sources based on the executable task. The export result is obtained based on the data query results and output. This solution enables the fusion and export of heterogeneous data from different data sources, improving the convenience of data export.

The invention discloses an active and passive cooperative network telemetering method based on sparsity prediction and path planning, and the method comprises the steps: designing an adaptive probability sampling function combining the queue depth and the number of INT labels in a passive sensing module of a data plane; a mark cooling mechanism aiming at a port is combined to reduce extra expenditure brought by a telemetering label, but monitoring of high-value flow state information is not lost; on a control plane, the invention provides an algorithm for converting the packet forwarding rate in an actual environment into a logarithmic form by using logarithmic normal distribution transformation and Holt double-exponential smoothing prediction, so as to carry out smoothing prediction on a future flow condition, thereby finding a sparse risk link of which the packet forwarding rate is smaller than a certain threshold value in advance; the discovery of the detection blind area is changed from passive to active; then, according to the predicted sparse link set and a prospective sparse link coverage and reconstruction (PSCR) mechanism, the topology of the whole network is partitioned, a series of loops are generated, an optimal loop is selected to serve as an active detection path, and a detection task is executed by the nearest host; therefore, the consumed bandwidth is minimized, and potential blind areas can be detected as much as possible; the method is suitable for a large-scale data center network, the blind area of network detection can be eliminated under extremely low overhead, and the visibility of the whole network reaches 99%.

The embodiment of the invention provides a self-adaptive scanning method and device, and relates to the technical field of network security. The method comprises the following steps: acquiring a scanning task for a target network; judging whether the type of the scanning task is port detection scanning or vulnerability vulnerability scanning according to the configuration information of the scanning task; determining a corresponding user mode protocol stack instance according to the scanning task type; and executing scanning based on the selected user mode protocol stack instance, and returning a scanning result. Therefore, according to the method, two types of tasks of a port detection scanning scene and a vulnerability vulnerability scanning scene can be accurately routed to the corresponding simple or complete user mode protocol stack instance for processing. In a port detection scanning scene, a high throughput rate and low resource overhead can be realized; in a vulnerability vulnerability scanning scene, while the interaction reliability of a complete protocol is ensured, the problems of kernel bottleneck and message back mismatch under the condition that multiple queues are parallel are avoided.

This invention provides a phase self-synchronization method, apparatus, and radar for a digital array radar. The phase self-synchronization method includes the following steps: Step 1. Periodically injecting a receive correction signal into the receiving channel to obtain the offset value of the receive phase difference between channels; Step 2. Extracting the phase of the radio frequency signal leaked from the transmit channel to the receive channel, and calculating the offset value of the closed-loop phase difference between the transmit and receive channels; Step 3. Decoupling the offset value of the receive phase difference between channels from the offset value of the closed-loop phase difference between channels to obtain the offset value of the transmit phase difference between channels; Step 4. Compensating for the phase offset of each receive and transmit channel based on the offset values ​​of the receive phase difference and the transmit phase difference between channels. This invention can compensate for the phase difference drift between the receive and transmit channels in a digital array radar in real time, improving the antenna beam performance under long-term operating conditions.