42results about How to "Improve Parallel Processing Efficiency" patented technology

A parallel data processing method based on multiple graphic processing units (GPUs) is provided, including: creating, in a central processing unit (CPU), a plurality of worker threads for controlling a plurality of worker groups respectively, the worker groups including one or more GPUs; binding each worker thread to a corresponding GPU; loading a plurality of batches of training data from a nonvolatile memory to GPU video memories in the plurality of worker groups; and controlling the plurality of GPUs to perform data processing in parallel through the worker threads. The method can enhance efficiency of multi-GPU parallel data processing. In addition, a parallel data processing apparatus is further provided.

This patent discloses a convolution neural network accelerator based on PSoC device, including an off-chip memory, a CPU, a feature map input memory, a feature Map Output Memory, a bias memory, a weight memory, a direct memory access to the same number of cells as neurons. The calculation unit comprises a first-in first-out queue, a state machine, a data selector, an average pooling module, a maximum pooling module, a multiplication and addition calculation module and an activation function module, wherein the calculations in the multiplication and addition calculation module are executed in parallel, and can be used for a convolution neural network system of various architectures. The invention fully utilizes the programmable part in the PSoC (Programmable System on Chip) device to realize the convolution neural network calculation part with large calculation amount and high parallelism, and utilizes the CPU to realize the serial algorithm and the state control.

The invention discloses a distributed parallel processing database system and a data processing method thereof. Reliable data support is provided for data recovery by formulating a data fragment distribution rule, a redundant backup rule and a redo log mechanism. And according to the data operation type, log recovery or inter-node consistency recovery is redone to perform data automatic recovery.Distributed deployment is supported, automatic data recovery is achieved, and a stable storage computing service is provided.

The invention discloses a pipe-type communication method for interprocess communication, comprising the following steps: receiving the data processed by the process of a server at the previous level; caching the data to a first caching pool; reading the data in the first caching pool, and caching the data to a first memory buffer zone so as to bring convenience for the process of the server to process; writing the data processed by the process of the server into a second memory buffer zone; caching the data in the second memory buffer zone to the second caching pool; and sending the data in the second caching pool to the process of the server at the next level to process, wherein the first caching pool and the second caching pool are storage spaces arranged on an external memory. The invention also discloses a pipe-type communication system for interprocess communication. According to the method and the system, data in the pipe-type communication mode in the streamline concurrent processing process are cached by the large-capacity storage space of the external memory, thereby improving the reliability and the efficiency for data transmission during pipe-type communication.

The invention discloses a maintenance method for a distributed dynamic double-layer forwarding table, and the method comprises the training, querying and aging process of a forwarding table, wherein the training process of he forwarding table comprises the steps: S1), inputting a real random number from the outside; S2), receiving a source address and a virtual local area network SA_vlan; S3), carrying out the Hash calculation of the real random number and the virtual local area network SA_vlan, and calculating a to-be-written address; S4), transmitting a signal indication of an item corresponding to the address downwards; S5), carrying out no writing of the item if a to-be-written item is set to be invalid, or else writing the virtual local area network SA_vlan into the item in an sel circuit; S6), carrying out the Hash calculation of a target address DA_vlan and the real random value, obtaining the address of the item, and reading a to-be-queried forwarding port. According to the invention, a plurality of small forwarding tables employ the same Hash calculation rule, and form different Hash mapping with different real random numbers. The item load rate which is far greater than the item load rate of one unified forwarding table is obtained, and the precious BRAM resources in an FPGA are greatly saved.

The invention provides a transparent encryption method and apparatus of a network application layer. A data encryption method of the network application layer in a data transmission process by cooperatively using a message obtaining module, a depth detection module, an encryption/decryption module and a data transfer module, and transparent encryption is achieved according to a unique processing mode in the method. An efficient and reliable data encryption and decryption method of the network application layer is provided, the encryption and decryption granularity is small, the data size is small, and the parallel processing efficiency is higher; and the encryption and decryption do not change the network message structure of the original data message, and the data transmission is not affected. The possibility of message errors is reduced, and the impact of encryption and decryption on the current network environment is reduced; and rules can be freely configured according to differentservice requirements, the application is flexible, and the transparent encryption method and apparatus are applicable to different network environments.

The invention discloses a payment method and device used for refueling. The method comprises the following steps of acquiring first refueling transaction data; detecting whether the first refueling transaction data is bound with an account; and if the first refueling transaction data is detected to be bound with the account, using integrals to deduct corresponding refueling cost. In the invention,a technical problem that payment is not convenient during a refueling process is solved; and the parallel payment and the integral accumulation of the refueling process can be conveniently completed.

The invention relates to the field of parallel computation application technology and the field of high-performance scientific computation, in particular to an octree parallel construction method for visual reconstruction of CT slice data with a TB-level data processing ability. According to the method, octree parallel construction is performed based on an MPI+OpenMP parallel programming model by use of the characteristics of mesh generation of original volume data, non-dependency of octree node data, etc. based on the scheme of "construction on demand-Branch on need Octrees, BONOs" according to actual three-dimensional size of the volume data; on the one hand, waste of computation resources and storage resources and I/O expenditure in the construction process are reduced; and on the other hand, rapid octree data structure construction of the TB-level CT slice data is realized by means of parallel computation, and the MPI+OpenMP parallel programming technology meets the requirement for rapid construction of an octree data structural body of the TB-level CT slice data under different resolution requirements. The method has a good parallel speedup ratio and good parallel efficiency.

The invention discloses a data area overlapped boundary data zero communication parallel computing method, which comprises the following steps of: partitioning mother data to be processed into a plurality of sub blocks of data, wherein each sub block of data stores boundary data of the adjacent sub block of data in a redundant way; and carrying out parallel processing to the sub blocks of data. The invention also discloses a data area overlapped boundary data zero communication parallel computing system, which comprises a data partitioning module which is used for carrying out redundant partitioning to the mother data, and a parallel processing unit which is used for carrying out parallel processing to the sub blocks of data, and also comprises a data partitioning module which is used for carrying out non-redundant partitioning to the mother data, a data exchange module which is used for exchanging the boundary data of the adjacent sub blocks of data and carrying out redundant storage to the boundary data, and a parallel processing unit which is used for carrying out parallel processing to the sub blocks of data. According to the method and the system, disclosed by the invention, the waiting time during the data transmission can be saved, and the efficiency of the parallel processing can be improved.

The invention belongs to the technical field of big data application, and discloses a big data application system and method based on the Internet of Things. The system comprises a data acquisition module, a central control module, an Internet of Things communication module, a data parallel processing module, a data mining module, a data statistics module, a data retrieval module, a data analysismodule, a big data storage module and a display module. According to the invention, the parallel processing efficiency of heterogeneous data is greatly improved through the data parallel processing module; meanwhile, the big data storage module is designed and used for the big data industry; the system can be used independently, can be used as a main means for data storage and collection, and canalso be used in cooperation with a data service, so that big data is applied in practical application, the better effect of the data is achieved, quick storage of mass data is achieved, and the data efficiency is improved.

The invention discloses a cache file delete influence verification method, a device and a mobile terminal; the method comprises the following steps: obtaining a to be verified program, and obtaining a click path set and a cache path set corresponding to the to be verified program, wherein the click path set comprises a plurality of click paths, the cache path set comprises a plurality of cache paths, and each click path is matched with one cache path; obtaining one to be verified cache path from the cache path set, and deleting a to be verified cache file of the to be verified cache path; carrying out simulation clicking for the to be verified program according to the click path corresponding to the to be verified cache path; obtaining delete influence of the to be verified cache file according to the running result of the to be verified program. The method only needs simple device to automatically detect whether the deleted cache file can affect the program or not, thus improving parallel processing rate, and reducing labor cost.

The invention provides a parallel mosaic method and system for massive remote sensing images based on Spark, which distributes the remote sensing image mosaic task to each worker node through the Master node in the Spark cluster as a controller of the entire Spark cluster. Through calling a custom RDD operation operator, the image mosaic is completed. The method rewrites RDD's compute and getPartitions methods in Spark to customize the RDD for remote sensing image processing, and takes three key steps of image mosaic, overlap area estimation, image registration and image fusion as the operation of custom RDD Sub-image parallel implementation of the mosaic. The mosaic method and system for massive remote sensing images based on Spark is able to effectively increase the image mosaic efficiency of a large number of images while the image mosaic effect is ensured.

The invention discloses a method and device for improving the parallel processing efficiency of a bare computer system, a medium and equipment, and belongs to the field of computer processing. The method comprises the following steps: transmitting current frame data containing N pieces of target information to a CPU (Central Processing Unit); processing first target information in the current frame data by utilizing the CPU; the hardware controller is used for sending the first target information, and meanwhile the CPU processes the second target information in the current frame data; and sending the ith target information of the current frame data by using the hardware controller, and judging whether the current frame data is processed or not, if the current frame data is judged not to be processed, continuing to process the (i + 1) th target information in the current frame data by the CPU, otherwise, starting to process the target information in the next frame data containing the M pieces of information by the CPU. According to the method and the device, the problems that CPU time is wasted when low-speed equipment sends data, and the overall task processing period is prolonged due to the fact that other work cannot be processed in parallel are solved.

The invention relates to an application program processing method, device and system and a computer readable storage medium, and relates to the technical field of computers. The method comprises the steps of determining correlation between to-be-processed data of an application program in response to operation starting of the application program submitted by a user; dividing the to-be-processed data according to the correlation; dividing the application program into a plurality of task fragments according to a division result; and allocating each task fragment to each corresponding actuator node in the actuator node cluster for parallel processing so as to obtain a processing result of each task fragment. According to the technical scheme, the parallel processing efficiency can be improved.