Abstract: An image compression sampling method and assembly are provided. The method includes: performing sparse representation on a target image by using an initial sparse matrix, quantifying an initial sparse representation result to obtain an optimized sparse representation result, and obtaining an optimized sparse matrix; constructing a product matrix by using the optimized sparse matrix and an initial measurement matrix, and adjusting absolute values of off-diagonal elements in the product matrix to be less than a correlation threshold; performing singular value decomposition on the product matrix to obtain a diagonal matrix and a left singular matrix, and updating the diagonal matrix according to a quantity of samplings of the initial measurement matrix; and optimizing the initial measurement matrix by using the left singular matrix and the updated diagonal matrix to obtain an optimized measurement matrix, and collecting image data by using the optimized sparse matrix and the optimized measurement matrix.

Abstract: Provided is a data processing method, which includes: in response to a logical volume receiving a write request, whether a logical address carried in the write request is occupied by a data unit in the logical volume is determined; if not, a data grain which is closest to the size of a data block and is greater than the size of the data block is determined; a new data unit is created in the logical volume by use of the logical address as an initial address and by use of the closest data grain as the length, and a logical address range occupied by the data block in the new data unit is recorded; the data block is written into an underlying storage and a written physical address is returned; and a mapping relationship between the initial address and the physical address is established and saved.

Abstract: This application discloses a high-reliability protection circuit and a power supply system.

Abstract: Provided are a chip and a pin line-out design method therefor, which are applied to a BGA packaged chip. The method includes: according to pin position information and pin definition information of a chip, determining a number of circuit board layers required for pin line-out of the chip (S1); allocating line-out layers to pins of the chip in their respective circuit boards (S2); and according to a pin density and transmission line width requirement of the chip, determining a specification of a via hole in each circuit board for leading the pin of the chip out to the corresponding line-out layer, to perform a corresponding line-out design on the basis of the via hole (S3). It may be seen that the described unified pin line-out design for the BGA packaged chip is more refined, and the quality of the line-out design of the pins of the chip is ensured.

Abstract: A method and device for model compression of a neural network. The method comprises: recording input and output parameters of each layer of network in a network structure; dividing the network structure into several small networks according to the input and output parameters; setting a pruning flag bit of a first convolutional layer in each small network to be zero to obtain a pruned small network; training each pruned small network to obtain a network weight and a weight mask; recording a pruned channel index number of each convolutional layer of a pruned small network with the weight mask of zero; and carrying out decomposition calculation on each pruned small network according to the pruned channel index number. According to the method, a calculation amount and the size of a model is reduced, and during network deployment, the model can be loaded with one click, thus reducing usage difficulty.

Abstract: The present disclosure discloses a method for determining a PCIe signal bandwidth, apparatus and device, and a computer readable storage medium. A target frequency point at which the energy spectrum density of a PCIe signal of a transmission link to be measured, after an output signal of said transmission link is subjected to inverse Fourier transform, reaches a preset proportion (determined according to a measurement error requirement of said transmission link) of that of an original signal is obtained by testing, and the target frequency point is taken as a bandwidth of the PCIe signal of said transmission link. Therefore, accurate high-speed signal bandwidth selection is realized.

Abstract: Provided is an image processing method, which includes: determining whether the number of written rows of source image data in a DDR is a preset number of rows or a stop write number of rows, the preset number of rows being an integer multiple of the number of rows of an image block determined according to a compression condition, and the image block being any image block in an image block matrix determined on the basis of an entire frame of image; if yes, reading the source image data of the current image block from the DDR on the basis of a first read address of the current image block, the number of rows, and the number of columns; and performing compression on the source image data of the current image block to obtain compressed data, and writing the compressed data to the DDR.

Abstract: Provided is a FPGA-based design method for equally dividing an interval, including the following steps: dividing the oscillation periods of a second pulse signal of a crystal oscillator clock of a FPGA board by the number of equally divided sampling pulses, and obtaining the remainder thereof; dividing the remainder by the number of the equally divided sampling pulses to serve as an error within each sampling interval; using a counter to count from the second pulse, and stopping the counting of the counter once whenever the error within the sampling interval, which is accumulated within the second pulse interval, is greater than or equal to the vibration period. Further provided is a FPGA-based design device for equally dividing an interval. The present application makes full use of the feature of interval equal division calculation, has high precision, and is easy to implement.

Abstract: A virtual machine management method and apparatus, a device, and a readable storage medium. The virtual machine management method is applied to a private cloud, and comprises: acquiring a SELinux label, the SELinux label comprising five elements: User, Role, Type, Sensitivity, and Category (S101); setting elements other than Category in the SELinux label as default values (S102); assigning different values to Category according to a preset Category variable combination so as to obtain multiple non-duplicate label groups (S103); and configuring each label group to a virtual machine in the private cloud, so that virtual machines configured with the same label group communicate with each other, and virtual machines configured with different label groups are isolated from each other (S104).

Abstract: Provided is a dynamic Huffman encoding method based on a sorting network. Compared with traditional dynamic Huffman coding solutions, the method implements sorting on the basis of the sorting network, therefore the sorting process is not only stable, but also may ensure a stable sorting result; and moreover, sorting steps and related operations are simpler, thereby greatly simplifying the sorting and iteration processes, and thus the sorting efficiency is higher. In addition, the sorting process better facilitates program implementation and transplantation, and implementation of hardware and software may achieve good effects. In addition, the present disclosure further provides a dynamic Huffman coding apparatus and device based on a sorting network, and a readable storage medium, and the technical effects thereof correspond to the technical effects of the above method.

Abstract: A power supply device with dual power source planes and a server. The device includes: a plurality of controllers connected in parallel; a first Power Supply Unit (PSU) power source group and a second PSU power source group each includes two PSU power sources connected in parallel, and the first PSU power source group and the second PSU power source group are respectively located on two independent power source planes; and a power source backboard, wherein the power source backboard includes a first copper skin layer and a second copper skin layer, which are not connected with each other, the first PSU power source group is connected to the input end of each controller by means of the first copper skin layer, and the second PSU power source group is connected to the input end of each controller by means of the second copper skin layer.

Abstract: A K8S-based service deployment method, apparatus, device, and medium, including: acquiring service data, the service data includes data corresponding to one service; deploying the service on a first target pod according to the service data; deploying a proxy corresponding to the service on the first target pod based on a sidecar manner, so as to use the proxy as a network proxy of the service; and acquiring a preset ConfigMap corresponding to the first target pod, whereby the proxy performs communication between different pods according to the preset ConfigMap, the preset ConfigMap includes routing information. As such, only one service is deployed on one pod, and a separate service does not need to be deployed to string together each independent service on one pod, thus reducing the workload in a service deployment process, and improving resource utilization.

Abstract: The present application discloses a pre-reading method and system of a kernel client, and a computer-readable storage medium. The method includes: receiving a reading request for a file and determining whether the reading of the file is continuous; if the reading of the file is discontinuous, generating a head node of a file inode, and constructing a linked list embedded in the head node; determining whether the file includes a reading rule for the file, and if the file includes the reading rule for the file, acquiring, based on the reading rule, the number of reading requests for the file and a reading offset corresponding to each request, generating a map route based on the number of reading requests and corresponding reading offsets, and storing the map route in the linked list; and executing pre-reading based on the linked list.

Abstract: A method, system and device for pipeline processing of instructions and a computer storage medium. The method comprises: acquiring a target instruction set (S101); acquiring a target prediction result, wherein the target prediction result is a result obtained by predicting a jump mode of the target instruction set (S102); performing pipeline processing on the target instruction set according to the target prediction result (S103); determining if a pipeline flushing request is received (S104); and if so, correspondingly saving the target instruction set and a corresponding pipeline processing result, so as to perform pipeline processing on the target instruction set again on the basis of the pipeline processing result (S105).

Abstract: A resource management method for a node in Kubernetes, a computer device, and a readable storage medium. The method includes following steps: initializing a component to start a main process; defining, by means of the main process, subprocesses for managing resources respectively; in response to a detection from the main process that a plurality of actual resources is present in a node among the resources, starting a subprocess corresponding to each of the actual resources, and starting gRPC service for each started subprocess; and registering the resources corresponding to a subprocesses where the gRPC service is successfully started to Kubernetes by using the main process enabling the subprocesses where the gRPC service is successfully started to manage the actual resources.

Abstract: The present application discloses a method for modifying an internal configuration of a virtual machine, a system and a device, wherein the method is applied to a virtual machine installed with a proxy service therein, and the proxy service is configured for, after the proxy service itself is started up, sending a datum request to a preset IP address via a virtual network card corresponding to the virtual machine. The method includes, when there is a target virtual network card sending a datum request to the preset IP address, according to a predetermined corresponding relation between virtual network cards and virtual machines, determining a target virtual machine corresponding to the target virtual network card; from a database, obtaining target configuration data corresponding to the target virtual machine.

Abstract: The present application discloses a method for processing a file read-write service. The method includes: in response to receiving a read-write service of a file, determining, based on a file serial number, whether a cache handle of the file is present in an index container; in response to the cache handle of the file not being present in the index container, opening, based on the read-write service, a corresponding handle of the file; encapsulating a flag and a pointer of the corresponding handle and the file serial number so as to obtain a cache handle of the file; adding the cache handle of the file into the index container and a first queue; processing the read-write service by using the corresponding handle of the file; and in response to completion of processing the read-write service, moving the cache handle of the file from the first queue to a second queue.

Abstract: A resource allocation method and a system after system restart and a related component. The method comprises: allocating, from a resource pool, a first part of resources to an initialization pre-application module; allocating, from the resource pool, a second part of resources to a cache module, such that the cache module restores cache data to be restored that is in an initialization stage; and repeating the following steps: determining whether current remaining resources in the resource pool can meet a preparation stage application requirement of the cache module or whether there is an IO restoration requirement; allocating resources to the cache module or the IO module, from the resource pool, according to the result of determination; and determining whether the preparation stage application requirement and the IO restoration requirement are completely met, and if so, jumping out of the loop.

Abstract: A measurement correction method and apparatus for a sensor, and a server power supply. The method is applied to a sensor comprising a shunt resistor and a differential amplifier, and comprises: performing compensation verification on the sensor in advance to obtain a compensation coefficient of the sensor (S1); obtaining a voltage signal output by the sensor (S2); and correcting the voltage signal according to an error value of the shunt resistor and the compensation coefficient to obtain a voltage correction signal for use in system optimization management (S3).

Abstract: A method, apparatus, and system for creating a training task on an AI training platform, and a computer-readable storage medium. The method includes: dividing nodes of the AI training platform into a plurality of virtual groups in advance, dividing a preset quota of disk space from each node to form a shared storage space of a virtual group, receiving training task configuration information inputted by a user, and determining task configuration conditions according to the training task configuration information; and determining whether there are first nodes satisfying the task configuration conditions among the nodes of the AI training platform, if so, selecting a target node from the first nodes according to a preset filtering method, creating a corresponding training task on the target node, and caching a training dataset obtained from a remote data center into an independent storage space of the target node, and recording a corresponding storage path.