126results about How to "Improve parallel efficiency" patented technology

The invention relates to a method for image mosaic without control points, which comprises the following steps: (1) acquiring an image sequence; (2) extracting a characteristic point set of each image in the image sequence; (3) searching homonymy point pairs of characteristic points between two adjacent images; (4) calculating a Homograph transformation relation between the adjacent images by using an RANSAC error-tolerant algorithm; and (5) using a continual multiplication formula and fusion technology to obtain mosaic results. Aiming at each characteristic point, the method can automatically collect a color channel with highest characteristic significance degree, so as to not only improve amount of the characteristic points greatly, but also improve significance and accuracy of the characteristic points. The method can be used as a correction method for human-computer interactive air photos to correct micro errors accumulated during mosaic. The method needs no control points, and has the advantages of high accuracy, low application cost and high efficiency.

Packet flows are processed, e.g. to perform an intrusion detection function in a communication network, by means of a multiprocessor system including a plurality of processing units. The packets are distributed for processing among the processing units via a distribution function. Such a distribution function is selectively allotted to one of the processing units of the plurality. A preferred embodiment of the arrangement involves using a single Symmetric Multi-Processor machine with a single network port to Gigabit / sec link. The corresponding system architecture does not require any intermediate device, or any external load balancing mechanism. All the processing work is performed on a single system, which is able to dynamically balance the traffic load among the several independent CPUs. By resorting to a specific scheduling arrangement, such a system is able to effectively distribute the computations required to perform both the loadbalancing and the detection operations.

The invention discloses a heterogeneous multi-core infrared image processing system and method. The method mainly includes: 1, utilizing programmable logic of a Zynq to realize a driving schedule of a detector to complete collection and caching of an infrared image; 2, utilizing a CPU0 (central processing unit 0) of the Zynq to operate a Linux operating system to write a Qt-based display and control interface; 3, utilizing a CPU1 of the Zynq to respond to a serial port command in real time, and setting a system parameter and a video driving chip online; 4, utilizing the programmable logic to realize algorithmic hardware acceleration and a driving schedule of a video chip. The heterogeneous multi-core infrared image processing system and method has the advantages that a structure of ARM+FPGA in an all programmable platform Zynq chip is utilized to develop the heterogeneous multi-core infrared image processing system, design difficulty of hardware is remarkably reduced, a system bandwidth chokepoint brought by interconnection of conventional framework chip levels is removed, simplicity and flexibility in design and integration of a user IP (internet protocol) are realized, and high universality is realized.

The invention discloses a parallel computation method for Newton power flow of a large-scale electric power system, which belongs to the technical field of safe analysis and computation of electric power systems. A modified equation set of the Newton power flow is resolved by using a computer according to a parallel cluster through a program in combination with parallelization technology for updating a Jacobian matrix by a generalized minimal residual method combined with a quasi-diagonal parallel preconditioner so as to realize parallelization of a power flow resolving process of the large-scale electric power system. The method does not need network partitioning in any form and has simple computation and high generality; a forward / backward sweeping dependency relationship does not exist inside an algorithm, and parallel efficiency and computation speed are high; and computation can be performed by the common parallel computer and the parallel cluster and the method is convenient to popularize and apply, and the like. The method can be applied to the parallel power flow computation of the electric power system and is particularly suitable for the parallel computation of the Newton power flow of a large-scale complex electric power system.

The invention discloses a parallel method for simulating and optimizing the computer architecture scheme based on a cluster system, and aims to provide a parallel method for simulating and optimizing the design scheme of the computer architecture. The technical scheme is that a parallel computer system which consists of a main control node and simulation nodes and is provided with a remote command execution environment is firstly built, and a global configuration program, a simulated configuration file generating program, a task dispatching program and a result analyzing program are arranged on the main control node, wherein, the global configuration program is used for arranging global configuration; the simulated configuration file generating program is used for generating all simulated configuration files; the task dispatching program distributes simulation evaluation tasks to each node, controls each simulation node and performs simulation evaluation; and the result analyzing program searches simulation result files sent from the simulation nodes for statistics, screens out optimal configuration parameter values, and outputs a report. By adopting the invention, the time for evaluation and optimization can be reduced, and the selection accuracy is improved.

The invention relates to an NAND flash memory single-channel synchronous controller with a dynamic instruction scheduling function. The controller comprises five parts including a user interface layer, a dynamic instruction scheduling layer, a control layer, a physical layer and a data caching module. Complicated operation of an NAND flash memory chip is converted into a simple instruction of a client, the instruction execution efficiency is improved to the hilt, operation parallelism of different substrates is developed, the throughput rate of data read and write is increased, and for the user, operation of basic storage arrays is controlled and optimized transparently; data are read and written synchronously, eight substrates in four flash memory chips are controlled in a single channel, the dynamic instruction scheduling function and an assembly line technology are adopted, and the read-write throughput rate and extendibility of a flash memory are effectively improved; and besides, four instructions including reading, writing, erasing and moving are supported, an ECC (error correction code) module is provided, and various operation requirements of the user layer are met.

The invention relates to a parallel programming method oriented to data intensive application based on multiple data architecture centers. The method comprises the following steps of: constructing a main node of system architecture, constructing a sub node of the system architecture, performing loading, performing execution and the like. The parallel programming method has the advantages that a technicist in the field of a large scale of data intensive scientific data does not need to know well a parallel calculation mode based on multiple data centers and does not need to have a MapReduce and multi-point interface (MPI) parallel programming technology relevant to high-performance calculation; a plurality of distributed clusters are simply configured, and a MapReduce calculation task is loaded to the distributed clusters; the hardware and software collocation of the existing cluster system is not required to be changed, and the architecture can quickly parallel the data intensive application based on the MapReduce programming model on multiple data centers; and therefore, relatively high parallelization efficiency is achieved, and the processing capacity of the large-scale distributed data intensive scientific data can be greatly improved.

The invention discloses a video decoding macro-block-grade parallel scheduling method for perceiving calculation complexity. The method comprises two critical technologies: the first one involves establishing a macro-block decoding complexity prediction linear model according to entropy decoding and macro-block information after reordering such as the number of non-zero coefficients, macro-block interframe predictive coding types, motion vectors and the like, performing complexity analysis on each module, and fully utilizing known macro-block information so as to improve the parallel efficiency; and the second one involves combining macro-block decoding complexity with calculation parallel under the condition that macro-block decoding dependence is satisfied, performing packet parallel execution on macro-blocks according to an ordering result, dynamically determining the packet size according to the calculation capability of a GPU, and dynamically determining the packet number according to the number of macro-blocks which are currently parallel so that the emission frequency of core functions is also controlled while full utilization of the GPU is guaranteed and high-efficiency parallel is realized. Besides, parallel cooperative operation of a CPU and the GPU is realized by use of a buffer area mode, resources are fully utilized, and idle waiting is reduced.

The invention relates to a radar signal processing system. A FPGA preprocessing module is used for preprocessing radar information. A DSP processing array is used for processing the preprocessed radar information. A multimode loading module is used for loading different preprocessing programs to the FPGA preprocessing module according to different instructions. A power consumption management module is used for managing the power consumption of the system. A state monitoring module is used for transmitting the state information of the system to the master control unit of the radar system and receiving the feedback information of the master control unit. The system uses FPGA + DSP hardware processing architecture, gives scope to the characteristics of the FPGA and the DSP, completes real-time operation of a complex radar information processing algorithm, can achieve the maximum cost-effectiveness of intensive information processing, and solve the memory bottlenecks of the DSP and FPGA by using DDR3 and high-speed SRAM technology.

The invention discloses a moving object detecting method based on a graphics processing unit (GPU). The moving object detecting method comprises the steps of loading video training data to a GPU video memory, memorizing the video training data in a local Z-shaped block mode, extracting multi-feature data of the video training data, adopting a self-adaption weight model to integrate the multi-feature data so as to build a codebook model based on multiple features, loading video testing data to the GPU video memory, memorizing the video testing data in a local Z-shaped block mode, utilizing the codebook model based on multiple features to perform moving object detection to the video testing data, utilizing a moving object detection result to update the codebook model based on multiple features and memorizing the moving object detection result on the GPU side. The moving object detecting method has the advantages of being high in applicability, strong in expandability, high in efficiency and low in cost and enables moving object detection to meet the requirements for accuracy and timeliness.

The invention discloses a large-scale transport vehicle parallel operation transporting system and a method. The parallel operation method includes that four large-scale semi-trailers are combined to form a transporting unit; when the transporting unit performs cargo-loading transportation, the four large-scale semi-trailers of the transporting unit perform cargo-loading transportation respectively; and when the transporting unit performs empty load transportation, by means of a rollover system, two large-scale semi-trailers of the transporting unit are loaded on the rest two large-scale semi-trailers so that a parallel operation transporting method that the two large-scale semi-trailers transport the rest two large-scale semi-trailers is formed. Under the condition that the transporting unit integrated by the four semi-trailers is not ultra-high, ultra-long and ultra-wide, the transporting unit is capable of performing reliable transportation according to requirements of road transportations and substantially reducing transportation costs.

The invention relates to the key technology in the data processing of the oil and earthquake exploration which provides the earthquake data of the subtly geological structure description. The technology includes: (1) the cannon path collecting wave field extrapolated figure in the mix field; (2) the master-slave dynamic load balance parallel algorithm of the cannon path collecting wave field extrapolation; (3) the breakpoint recovery method of the parallel algorithm. The invention has improved the figure precision of the wave equation pre-fold depth excursion to proper for the land data processing; the parallel algorithm can use the free source and increase the work computing efficient. The breakpoint recovery method can not waste the pre-processing work, so it can shorten the processing period.

An analog front-end circuit comprising: a first current mode differential charge amplifier for a first sensing line, comprising: a unity gain buffer with current replication circuit, wherein the first sensing line current being fed to the unity gain buffer with current replication circuit to generate a pull up and a pull down output currents; and a current mirroring and amplification circuit, wherein the pull up and pull down output currents being fed to the current mirroring and amplification circuit to generate a positively scaled and a negatively scaled output currents; a current combining circuit for combining the positively scaled output generated for the first sensing line and a negatively scaled output current generated for a second sensing line to generate a differential current; and a differential current-to-voltage converter for converting the differential current to an output voltage usable by touch controller application.

The invention relates to a data-intensive remote sensing image processing-oriented generic programmed framework programming method. The technical scheme comprises the following steps of 1, according to characteristics of large mass, complex data structure and the like of sensing image data, building a distributed remote sensing image model; 2, on the basis of remote data parallel processing algorithm analysis, modeling and abstracting a public computing mode which repeatedly appears, and building remote a sensing data processing algorithm-oriented parallel algorithm; and 3, packaging the parallel algorithm through a template mechanism to realize a generic programmed algorithm model. The method has the beneficial effects that a parallel programming mode is provided by remote sensing application algorithm, so that the parallel programming difficulty of a large amount of sensing data processing applications can be effectively reduced, the defect that the requirement on the programming ability of a domain expert in the traditional remote data processing is high is overcome, and wide application of parallel computing technology to large area even global massive remote data processing of global change, dynamic disaster monitoring and the like is greatly promoted.

The invention relates to a small-size integrated parallel rotation ultrasonic motor and belongs to the technical field of ultrasonic motors. The ultrasonic motor comprises a stator (1) and a mover (2) used as a bearing, the stator (1) is formed by parallelly connecting N (N>=3) composite vibrators by means of flexible support components, and the composite vibrators are symmetrically placed around the mover along the circumferential direction; each of the composite vibrators is formed by mutually intersecting two Langevin vibrators in a certain angle mode; simultaneously, heads of the two Langevin vibrators are connected into a whole to form drive feet (13) of the composite vibrators, and the Langevin vibrators are located on the same plane; and each of the Langevin vibrators is composed of an elastomer (11) and a piezoelectric ceramic piece (12), a positive pole of the piezoelectric ceramic piece faces outward, and a negative pole of the piezoelectric ceramic piece points to a base body surface of the elastomer (11). The small-size integrated parallel rotation ultrasonic motor is high in energy density, compact in structure and still capable of providing large output torque under the condition that the motor is longitudinally thin.

The invention relates to a distributed parallel spatial viewshed analysis method. A module is subjected to parallel dynamic division by using digital elevation model data, parallelization and data scheduling of a serial algorithm are quickly completed, and irregular data partitioning and a redundancy mechanism of a sight line region are applied, so that it is ensured that a slave node can access to the digital elevation model data corresponding to a target grid point region in a distributed calculation environment, wherein an irregular data parallel policy has important guidance meaning for viewshed algorithms of different realization modes; and in combination with a buffer region mechanism, correct operations of different dependence degree algorithms are effectively ensured and resource utilization efficiency of the slave node is improved, so that the method has wide industrial application prospects.

The invention relates to a parallel optimization method for a convolutional neural network, and the method comprises the steps: carrying out the convolution kernel operation of the convolutional neural network through employing a winograd algorithm f (2x2, 3x3), so as to reduce the time complexity and reduce the number of times of multiplication; for the for circulation structure part without circulation data dependence, using an OpenMP for opening up a plurality of threads to carry out operation; carrying out vectorization processing on the data operation part with the same mode, so that one-time instruction multiple operation can be realized. The convolutional neural network program improved by adopting the method provided by the invention can greatly improve the parallel efficiency, reduce the operation complexity, fundamentally reduce the operation expenditure, and reduce the operation time of the program.

The present invention is to clarify causes of hampering the performance by quantitatively associating a value of a parallel efficiency with factors of hampering the improvement of the performance of a parallel computer system, Processing time α(p, n) for a portion to be sequentially processed, processing time β(p, n) / p for a portion to be parallel processed and processing time σ(p, n) caused by an overhead for the parallel processing at the time of the execution of a parallel processing program are measured. A parallelized rate Rpara(p, n), a sequential calculation time ratio Rα(p, n) and a parallel overhead ratio Rσ(p, n) are calculated by using the obtained processing time σ(p, n) for the portion to be sequentially processed, β(p, n) for the portion to be parallel processed and σ(p, n) caused by the overhead for the parallel processing. A parallel efficiency Epara(p, n) is calculated in accordance with an expression 1 / Rpara(p, n)×(1−Rα(p, n)−Rσ(p, n)).

The invention relates to a variable-depth cable ghost wave suppression method based on wave equation boundary value inversion. The method includes the following steps: 1) extracting a common-shot-point gather from seismic data collected by a variable-depth cable; 2) performing one-dimensional Fourier transform on the seismic data of a certain shot of the common-shot-point gather to obtain a frequency spectrum containing ghost wave seismic data; 3) calculating frequency domain seismic data U<AB> containing ghost waves corresponding to a certain frequency omega in the frequency spectrum; 4) calculating G1 and G2 according to detection point depth H<0><(2)>; 5) calculating frequency domain seismic data F<CD> after ghost wave suppression which are corresponding to the frequency omega according to U<AB>, G1 and G2; 6) performing steps 3) to 5) on all frequencies omega on the frequency spectrum in the common-shot-point gather to obtain frequency domain seismic data after ghost wave suppression; 7) performing steps 2) to 6) on other shots in the common-shot-point gather to obtain frequency domain seismic data after ghost wave suppression; and 8) performing inverse Fourier transform on frequency domain seismic data F<CD> after ghost wave suppression to obtain a common-shot-point gather after ghost wave suppression. The variable-depth cable ghost wave suppression method based on wave equation boundary value inversion can be used in the processing process of seismic data collected by a maritime variable-depth cable.

The invention discloses a multi-grid Chebyshev parallel spectral element method. The method includes the steps that a, a calculation area is divided into large and regular units; b, main grids and auxiliary grids are defined in the units, and the main grids are Chebyshev collocation points; c, Wave fields in the units are subjected to Chebyshev truncation expansion by the utilization of the main grids, and medium parameters and external forces in the units are subjected to truncation expansion by the utilization of the auxiliary grids; d, wave field, medium parameter and external force truncation expansion formulas are substituted into a wave equation to obtain unit mass matrixes, unit stiffness matrixes and unit external force vectors; e, on the basis of the unit mass matrixes, the unit stiffness matrixes and the unit external force vectors, an elastic wave expansion is solved on the main grids.

The invention relates to a circuit simulation method for realizing parallel computation through time domain division, belonging to the technical field of integrated circuit design. The method comprises the following steps of: giving an input file for circuit simulation, wherein the file is used for describing a topological structure of a circuit required to be simulated, an external exciting signal and a simulation parameter; averagely segmenting simulation time into a plurality of slices according to the actual condition of calculating resources; reserving time for one part to be overlapped between two adjacent time slices; independently and transiently analyzing each time slice; simultaneously carrying out different time slices on different calculation units by using different threads or progresses; and after finishing transient analysis of all time slices, carrying out waveform splicing on the result obtained by the transient analysis of all time slices to obtain the waveform of the voltage at each node of the circuit in the whole simulation time range and further processing to obtain other required information. According to the invention, the design efficiency of the circuit can be improved and the time of circuit products for appearing on the market is shortened.

The invention discloses a multiple parallel method based on a characteristic line method. After radial MOC inhomogeneous calculation is carried out on each layer of a core of a reactor, the layering homogenization cross section of a lattice cell is obtained in a homogenization mode, three-dimensional transportation calculation is carried out, and then an axial leakage item for radial MOC inhomogeneous calculation is updated until iterative convergence is achieved. By means of the mode of analyzing the effects of the parallel efficiency of space, the parallel efficiency of angles and the parallel efficiency of characteristic lines on the total parallel efficiency at the same time, a multiple parallel optimization scheme is determined.

The invention discloses a method for obtaining power system parallelization track sensitivity. Track sensitivity has important application value in the fields of power system dynamic performance assessment, stable verification and control and the like, but in large-scale power system analysis, the technical difficulty that calculation time is excessively long exists and accordingly track sensitivity cannot be effectively applied in an existing power system based on real-time control. To solve the preceding problem, the novel method for obtaining power system track sensitivity more quickly by means of a parallel computing technique is put forward. By digging and making use of the Newton iteration matrix structure property in forward sensitivity analysis, a special dishonest Newton iteration strategy is designed, linear equation systems in Newton iteration can be decoupled and deployed on different computing processing units to be calculated, parallelization of track sensitivity obtaining is achieved, computational efficiency is greatly improved and the method can be used for the fields of power system stable analysis and control and the like.

The invention discloses a parallel scanning method used for nuclear reactor shielding calculation. The method comprises the following steps: carrying out splitting approximation on refined nets for computational geometry, and then dividing the refined nets into a plurality of areas, wherein one CPU is responsible for each area, and meanwhile each area for which one CPU is responsible is divided into a plurality of parallel units; and after the calculation of the parallel units on one CPU is finished, computing information can be transferred to adjacent parallel units, so that efficient parallel shielding calculation is realized, and a powerful safeguard is provided for the safety and economy of nuclear power plant design.

The invention provides a material irradiation damage spatial resolution random cluster dynamics parallel simulation method, which can realize large-scale SRSCD simulation. The method comprises the following steps: allocating simulation volumes of spatially resolved random cluster dynamics to different processes, establishing a three-dimensional Cartesian topological structure of the processes, anddividing a region of the processes into a plurality of sectors; determining a communication data structure; traversing each sector in sequence according to a sector number sequence; calculating a time threshold of circulation in each sector, and entering internal circulation; and randomly selecting a reaction in the current sector, updating the defect according to the selected reaction, filling the communication data list and updating the related reaction, communicating with the neighbor process when the evolution time of the internal loop is greater than the time threshold of the internal loop, synchronizing the corresponding defect information and updating the related reaction. The invention relates to the technical field of nuclear material irradiation damage simulation and the field of parallel computing.

The invention discloses a moving particle method parallel computing equivalent particle load balancing acceleration method, which comprises the following steps of: 1, generating a corresponding background grid according to initial distribution of particles in a computing domain; 2, performing calculation region division according to the number of nodes, and determining a division mode according tothe number of particles and a calculation region; 3, dividing a particle computational domain by adopting an equivalent particle equilibrium distribution method; 4, adopting a memory sharing type OpenMP parallel model to realize parallel computing in the computing nodes; 5, parallel computing between nodes is realized by adopting a message passing type MPI parallel model; 6, achieving the load balance of each node by adopting equivalent particle balance distribution, so that the parallel efficiency is greatly improved. A Lagrange method-based moving particle method can be accelerated in parallel on a large scale, a more balanced division mode is developed according to the calculation amount difference of different types of particles, the use efficiency of the calculation nodes is improved, and the parallel mode is suitable for current mainstream cluster type calculation hardware.

The invention discloses a parallel symbol execution system based on multi-Agent distributed scheduling. The system comprises a distributed scheduling and monitoring unit, an execution track generating unit, a symbol execution unit and a constraint solving unit. The distributed scheduling and monitoring unit is used for achieving pipeline parallel scheduling on the execution track generating unit, the symbol execution unit and the constraint solving unit, the parallel operating degree of the system is increased, and loads are balanced. The execution track generating unit is used for actually executing specific input drive measured programs, detecting whether abnormalities are generated or not, recording a program execution instruction sequence and generating a program execution track in an intermediary language expression form through analysis and transformation. The symbol execution unit is used for conducting symbolic simulation execution on the program execution track, collecting constraint conditions of route branches and generating a corresponding constraint expression. The constraint solving unit is used for conducting depth-first negation solving on the constraint expression, generating a new input sample, marking the negated expression and avoiding repeated route analysis. A realized task scheduling policy can realize pipeline parallelism, and the problem of load imbalance caused by the fact that only program execution sub-trees serve as scheduling loads is avoided.

The invention discloses a method for carrying out parallel simulation by repeatedly switching a plurality of operation modes of a simulator, and aims at reducing the loss of simulation precision and the simulation time. The technical scheme of the invention is as follows: a host computer software and hardware platform of an operation simulator is constructed, a server node is provided with a control software of a server-side, and a simulation node is provided with a control software of a client; the control software of the server-side carries out global configuration and initialization on thenodes participating in the parallel simulation; the server node and the simulation node are performed with communication initialization; the control software of the client operates the simulator on the simulation node; the simulation nodes parallelly simulate the same Benchmark program in the three modes, namely functional simulation, detailed pre-heating and detailed simulation; and the control software of the client monitors the changes of a local performance index file and a status file, acquires the operation information and sends the information to the server nodes, and the server node receives the information and correspondingly processes to obtain a simulation result. The invention has obvious accelerated effect, high parallel efficiency, and small precision loss.

The invention relates to a power grid reliability real time parallel assessment method base on Monte-Carlo simulation task dynamic distribution. Firstly, data are input, based on the De Moivre-Laplace central-limit theorem, a sampling frequency corresponding to LOLP index variance coefficient convergence of the system is estimated for conservative projection. Then, the task distribution is carried out according to the estimated value of the sampling frequency. After each processor receives a simulation task, N / p system states are extracted through parallel random number generators for carrying out fault result analysis to fault states of the same class so as to form reliable indexes in the N / p system states. Simulation results are sent to a major processor and tasks are finished, and the major processor immediately receives the simulation results sent by the processors after completing task%. The beneficial effects are that the method is high in simulation precision, the inter-processor communication frequency (or time) is also reduced while the convergence precision is ensured, and the Monte-Carlo simulation efficiency is further improved.

The invention discloses a Markov Monte Carlo algorithm acceleration method and system based on an FPGA (field-programmable gate array) heterogeneous platform. The method comprises: allowing a host to finish initialization, and transmitting generated random number data to a DDR (double data rate) memory of an FPGA terminal through a PCI-e (peripheral component interconnect express) interface, wherein the initialization includes constructing a probability space and generating random numbers; allowing the FPGA terminal to read the random number data in the DDR memory, and using a constructed Markov chain to perform concurrent pipeline operation on the random number data, and acquiring and transmitting corresponding result data to the host. The random number data in the DDR memory are read via the FPGA terminal, and the constructed Markov chain is used to perform concurrent pipeline operation on the random number data; it is possible to use the constructed Markov chain in an FPGA board to finish a core computing process of the Markov Monte Carlo method in a concurrent and pipeline manner, the cost is reduced, and concurrency efficiency is improved.