13 results about "High performance parallel computing" patented technology

The invention discloses a resource optimization scheduling method for a high-performance parallel computing data center, and relates to the technical field of resource optimization scheduling, and the method comprises the steps: collecting multi-dimensional parameters of the high-performance parallel computing data center, and carrying out the preprocessing; calculating a comprehensive scheduling target index according to the preprocessed parameters; calculating a comprehensive resource scheduling index according to the comprehensive scheduling target index, setting the comprehensive resource scheduling index as a target function, defining a constraint condition, and constructing a multi-target optimization model; and solving the multi-objective optimization model according to a multi-objective optimization algorithm, generating a resource allocation scheme, and allocating tasks to corresponding resource nodes to complete resource optimization scheduling. The comprehensive resource scheduling index is constructed, and the optimal solution is efficiently searched in the complex high-dimensional solution space by using the multi-objective optimization algorithm, so that the comprehensiveness of resource evaluation, the construction efficiency of the scheduling index and the optimization quality and practicability of the resource allocation scheme are remarkably improved.

The invention discloses a CFD (Computational Fluid Dynamics) efficient GPU (Graphics Processing Unit) computing method without a synchronous ILU (Inter-Link Unit) precondition, which belongs to the field of high-performance parallel computing and comprises the following steps of: taking a block size as a block row group, taking the block row group as a parallel granularity, and initializing a dependency relationship for the block row group; determining the index of each thread processing task line, and finding the diagonal position of the line; a busy waiting mechanism is used for judging the change of the dependency relationship between the block row groups, and accumulation operation is carried out on each row; the dependency relationship in the block is solved, and the latest value is transmitted to the thread in the block; and completing calculation of all corresponding solution vectors x of the block row group, setting the dependency relationship of the block row group to be solved, and performing notification. According to the method, the solving efficiency is improved, the synchronization overhead is reduced, the data access performance is improved, and the method can meet the complex calculation requirement of a large-scale unstructured grid.

The invention relates to a high-precision sensing device capable of being used for GIS accidental partial discharge positioning and type diagnosis. The high-precision sensing device comprises an ultrahigh frequency sensor and a data acquisition system. The ultrahigh frequency sensor is responsible for capturing weak accidental partial discharge signals generated in the GIS equipment; the captured original signal is then sent to a filtering and amplifying module and a synchronization module for hardware preprocessing, and is digitalized through a high-speed sampling assembly; the digitized mass data are transmitted to the high-performance GPU parallel computing unit, and the data are efficiently analyzed by utilizing the strong parallel processing capability of the high-performance GPU parallel computing unit. According to the invention, the equipment sampling rate is improved to a high sampling rate of 5GHz, so that the partial discharge pulse sampling interval of the GIS partial discharge monitoring equipment reaches 0.2 ns, and the sampling points and the data volume are greatly increased; gPU parallel computing is combined and utilized to greatly improve the pulse processing capacity, so that the system can effectively find and capture intermittent and accidental partial discharge signals, and the detection rate of the intermittent and accidental signals is remarkably improved.

The application discloses a method and device for constructing a folded hypercube edge-disjoint Hamiltonian cycle, and relates to the technical field of interconnection network topology. The method comprises the following steps: recursively constructing a first Hamiltonian cycle on an n-dimensional folded hypercube; applying a mapping f to each vertex u in the first Hamiltonian cycle, and sequentially connecting the mapped vertices in the original order to obtain a second Hamiltonian cycle; and the first Hamiltonian cycle and the second Hamiltonian cycle do not have any common edge in the n-dimensional folded hypercube. The above method provides key technical support for the topology design and routing protocol of a high-performance parallel computing system.

This invention discloses a feature-preserving mesh processing method and system based on high-performance parallel computing. The input is an unlabeled point cloud dataset, used to train a point cloud feature generator and a prediction head MLP. α The input is a grid dataset with semantic segmentation labels. The grid surface is sampled to generate a sparse point cloud, and noise is added to the grid to represent the redundant structure of the grid geometry clipping task. Then, the trained prediction head MLP is removed. α Train prediction head MLPs separately β With MLP γ The user inputs the mesh to be processed, all vertices are converted into a sparse point cloud, and the point cloud feature generator and prediction head MLP are used. β MLP γ The invention generates semantic segmentation labels and redundant structure category labels for the sparse point cloud, respectively; finally, it performs a culling operation on the points marked as redundant structures to complete the mesh pruning. This invention can effectively understand the features in the mesh data, and the semantic segmentation results are more accurate. The noise addition process of this invention uses a CUDA parallel computing architecture to ensure the high efficiency of the entire computation process.

The invention discloses a high-performance time sequence parallel simulation system and method based on a time acceleration and synchronous control mechanism, and belongs to the technical field of high-performance parallel calculation. The system comprises a timing trigger module, a time mapping engine module, a task generation and distribution module, a parallel processing module, a synchronous blocking module, a state aggregation module and a visual output module. By establishing a double-layer time system of physical time takt-simulation time window, a time mapping engine dynamically calculates the number of virtual time steps needing to be propelled according to acceleration multiplied speed, and acceleration mapping of physical time and simulation time is achieved. The parallel processing module executes tasks in a thread pool with a fixed size, and the synchronous blocking module performs unified time sequence control on execution of each thread through a counting synchronizer, so that state consistency and calculation reproducibility under a high-speed operation condition are ensured. And the state aggregation module is used for integrating the synchronized local calculation results and updating the global system state. According to the method, high-speed simulation and high-fidelity playback of large-scale time-dependent data are realized on the premise of not damaging sequential logic, and the modeling verification efficiency and stability of a complex system can be remarkably improved.

The application discloses a high-performance parallel computing dynamic scheduling method and system based on eBPF, and the method comprises the following steps: S1, capturing an application layer function based on an eBPF program, and jumping to step S2 if a specified effective computing function is captured; the specified effective computing function refers to a computing function that calls a specified mathematical library to perform mathematical operation and starts parallel computing; S2, judging whether the CPU where the current computing task is located is idle and whether the current computing task is a heavy computing task or a light computing task, and if the CPU where the current computing task is located is idle or the current computing task is a light computing task, then the process of the current computing task is bound to the CPU where the current computing task is located, the scheduling strategy of the process of the current computing task is changed to a FIFO scheduling strategy, and the method is ended and exited. The application solves the problem of long computing time in high-performance computing.

The present invention discloses a CFD high - efficiency GPU computing method without a synchronous ILU pre - conditioner, belonging to the field of high - performance parallel computing, including the steps of: taking the rows of a block size as a block - row group, with the block - row group as the parallel granularity, and initializing the dependency relationship for the block - row group; determining the index of each thread to process the task row and finding the diagonal position of the row; using a busy - waiting mechanism to judge the change of the dependency relationship between block - row groups and performing an accumulation operation on each row; solving the dependency relationship within the block and passing the latest value to the threads within the block; completing the calculation of all corresponding solution vectors x of the block - row group, setting the dependency relationship of the block - row group as resolved, and giving a notification. The present invention improves the solution efficiency, reduces the synchronization overhead, enhances the data access performance, and can adapt to the complex computing requirements of large - scale unstructured grids.

The invention discloses a distributed uncertainty multidisciplinary optimization design method and system for a carrier rocket, and aims to process uncertain parameters, convert high-dimensional uncertain data into low-dimensional independent variables, reduce the calculation amount and improve the optimization design efficiency. When the data in the quantization parameter matrix is reconstructed, the optimization problem is solved based on the quantized parameters to obtain the probability density function, compared with a traditional uncertainty quantization means, the method can more accurately describe the probability distribution of the uncertain parameters, a reliable data basis is provided for subsequent optimization design, and the probability density function is obtained. The later calculation result is more accurate and does not need to depend on design experience. The constructed subject uncertainty and subject analysis agent model replaces high-cost high-fidelity subject analysis, and by constructing an offline agent model and combining with high-performance parallel calculation, frequent calling of a high-fidelity model in a UMDO solving process is avoided, and the calculation efficiency is remarkably improved.

The invention discloses an eBPF-based high-performance parallel computing dynamic scheduling method and system, and the method comprises the steps: S1, capturing an application layer function based on an eBPF program, and skipping to a step S2 if a specified effective computing function is captured; the specified effective calculation function refers to a calculation function for calling a specified math library to perform mathematical operation and starting parallel calculation; s2, judging whether the CPU where the current calculation task is located is idle or not and whether the current calculation task is a heavy calculation task or a light calculation task, if the CPU where the current calculation task is located is idle or the current calculation task is the light calculation task, binding the process of the current calculation task to the CPU where the current calculation task is located, and if not, binding the process of the current calculation task to the CPU where the current calculation task is located; and changing the scheduling strategy of the process of the calculation task into an FIFO scheduling strategy, ending and exiting. According to the method, the problem of overlong calculation time in high-performance calculation is solved.

The invention relates to a full-flow acquisition and intelligent analysis detection system based on hardware acceleration, and belongs to the technical field of network security. According to the full-flow acquisition and intelligent analysis detection system based on hardware acceleration, the processing efficiency and accuracy of network flow data can be remarkably improved. Through high-performance parallel computing capabilities of the high-speed NIC, the FPGA and the GPU, the system can realize zero packet loss and zero packet leakage in a high-bandwidth environment, high efficiency and stability of data acquisition and preprocessing are ensured, and data capture and preprocessing performance is improved. By rapidly screening and filtering irrelevant traffic, the system overhead is reduced, and high-quality data support is provided for subsequent intelligent traffic analysis.

The invention belongs to the field of high-performance parallel computing, particularly relates to an implementation method of a SpMV on a CPU-DCU heterogeneous platform based on a self-adaptive strategy, and aims to improve the SpMV operation efficiency. The method comprises the following steps: acquiring sparse matrix data at a CPU end, and converting the sparse matrix data into a CSR format to obtain a CSR format array; based on the CSR format array, performing adaptive decomposition on a sparse matrix represented by the CSR format array to obtain a dense block and a sparse block, and constructing an RDCSR format array based on the CSR format array; the RDCSR format array comprises a CSR format array, an array for recording a dense block row index and an array for recording a sparse block row index. And transmitting the RDCSR format array to the DCU, performing adaptive thread allocation on the dense blocks and the sparse blocks at the DCU end, and then executing SpMV calculation in parallel to obtain a merged calculation result. And the merged calculation result is transmitted from the DCU to the CPU, and the calculation performance of the SpMV is evaluated.

The invention provides a direct current optimal power flow problem rapid solving method and device based on a GPU, and belongs to the field of power system optimization and high-performance parallel computing. The method comprises the following steps: establishing a direct current optimal power flow model; re-expressing the direct current optimal power flow model as a compact form of a linear programming model, and then converting the direct current optimal power flow model into a corresponding dual form; and solving the dual form of the direct current optimal power flow model on the GPU in parallel by using an sGS-HPR algorithm combined with symmetric Gauss-Seidel decomposition and a Halpern Peaeman-Rachford algorithm based on a positive semidefinite near-end item, and obtaining an optimization result of the direct current optimal power flow. According to the method, the parallel computing capability of the GPU can be fully utilized, the power grid optimization problem of the large-scale security constraint is efficiently solved, and the method has a relatively high application value.