152 results about "CUDA" patented technology

The invention provides a method and system for achieving CUDA call tracking based on an eBPF, and relates to the field of GPU call tracking of the eBPF. The method comprises the following steps: dynamically mounting an eBPF program at an API function entry and an API function exit of a user mode CUDA runtime library through an uprobe technology of the eBPF; the eBPF program captures metadata of the CUDA API calling event in a kernel mode and packages the metadata into a predefined structural body; asynchronously transmitting the structural body data to a user mode through an annular buffer area; and calling a standard function library by a user mode program to load an eBPF program, reading data in the annular buffer area in real time, and completing analysis and visual display of a CUDA calling event. According to the method, the calling parameter, the return value and the nanosecond timestamp are effectively captured, and the observability and the debugging efficiency of the GPU application program are improved.

The invention provides a CUDA code generation and optimization method and system for multi-dimensional feature search and enhancement, and the method comprises the steps: generating an initial CUDA code through a large language model based on task description and GPU hardware parameters; executing triple verification and a feedback mechanism on the initial CUDA code; according to feedback information of triple verification, optimizing CUDA codes through multi-dimensional feature search; based on error information, adjusting large language model input to regenerate codes; dynamically selecting an optimization strategy based on a performance index, wherein the optimization strategy comprises a thread block size and a memory access mode; and iteratively executing the steps until the CUDA code which passes triple verification and meets the target performance is generated. According to the method, the code features of the CUDA code are analyzed, and a triple verification and feedback mechanism including compilation feasibility, logic correctness and performance during execution is formed, so that an automatic evaluation process of the CUDA code is realized, and the problems that the CUDA code optimization process is difficult to be automated and an optimization target is lacked in the CUDA code optimization process are solved.

The invention relates to a world model construction method, in particular to a high-fidelity lightweight world model construction method for an end-to-end automatic driving test, which is used for constructing a high-fidelity world model and performing knowledge distillation on the world model to solve the problems of huge parameters, low reasoning efficiency and the like of the world model. Model parameters are reduced on the basis of reserving the world model generation capability, and the reasoning efficiency is improved; a CUDA operator is developed in a user-defined mode for the bottleneck part of world model calculation, video memory allocation is optimized, and the reasoning efficiency of a high-fidelity world model is improved based on a single-device multi-thread scheduling and multi-device cooperative calculation method. According to the method, the high-fidelity lightweight world model for the end-to-end automatic driving test can be constructed, the problems that an existing world model is low in multi-modal information alignment precision, poor in cross-view and cross-frame consistency, low in reasoning efficiency and the like are effectively solved, the confidence coefficient of the end-to-end automatic driving system test process is improved, and the test efficiency of the end-to-end automatic driving system is improved. And the testing efficiency of the end-to-end automatic driving system is greatly improved.

The invention discloses a multi-view collaborative 3D Gaussian splash optimization method and system, and the method comprises the steps: constructing a multi-level heterogeneous video memory pool, and dynamically dividing a video memory in 3D Gaussian reconstruction into a view exclusive memory block and a global shared memory pool; a mixed rendering-gradient pipeline is designed, and hardware-level pipeline parallelism in 3D Gaussian reconstruction is realized through a double-buffer asynchronous switching mechanism based on a CUDA Warp-level parallel primitive fusion forward rendering and back propagation thread group; performing multi-view gradient joint optimization, screening an effective gradient path in 3D Gaussian reconstruction based on the visibility mask matrix, and performing projection error weighted fusion on a multi-view gradient tensor; and implementing a multi-modal densification decision, generating a 3D Gaussian candidate splitting position in 3D Gaussian reconstruction through Monte Carlo sampling, calculating a joint optimization objective function by combining a multi-view projection residual error and a gradient contribution factor, and finally realizing 3D Gaussian reconstruction. According to the invention, high-precision and low-delay large-scale scene real-time rendering and training can be realized.

The invention provides a CUDA kernel generation method and device based on reinforcement learning, and a CUDA kernel generation method and device based on reinforcement learning, which utilize a CUDA kernel generation and optimization model of reinforcement learning training and adopt a multi-round reinforcement learning training scheme. According to the method, the continuous optimization and iteration process of code trigger generation, execution and feedback followed by an engineer during development of the CUDA kernel is simulated, and performance feedback in a real execution environment can be fused into a model training process, so that the model autonomously explores an optimization path under a reinforcement learning mechanism, and automatic and intelligent generation of the high-performance CUDA kernel is realized.

The invention discloses a strong real-time signal processing method based on multi-stream processing, and the method comprises the steps: carrying out the remainder of a preset stream number of received coherent processing interval data through a CPU according to a CPI number, obtaining a stream number, distributing the coherent processing interval data to each stream distributed calculation program corresponding to the stream number, each stream distributed computing program comprises a CUDA stream and a CPU processing thread which are executed in sequence; the GPU and the CPU execute each stream distributed calculation program to overlap the calculation time and the data copying time of the CUDA stream and the CPU processing thread to obtain each trace point condensation data with different completion time; the CPU obtains the corresponding report to be sent according to the trace point condensation data, and orderly sends the report to be sent to the target program based on a report waiting synchronization mechanism, and the application can complete front-end downloading data receiving, detection processing and report sending in strong real time, thereby improving the development efficiency, shortening the product development period, and reducing the project cost.

The invention discloses an infrared bidirectional thermal effect simulation method based on radiation and GPU acceleration, and belongs to the field of thermal radiation simulation and parallel computing. Dispersing the complex geometry into patch units, constructing an inter-patch radiation energy balance equation based on a radiance theory, and iteratively solving the final temperature of the patches through a radiance method; a three-level parallel strategy of a task level, a data level and an instruction level is designed, data intensive tasks such as shape factor calculation and radiance iteration are migrated to a GPU to be executed, data storage is optimized in combination with a structure array (SoA) layout and a sparse matrix compression technology, and efficient data sharing of a CPU end and a GPU end is achieved through a CUDA zero copy technology. According to the method, the dependence of a traditional method on regular grids is broken through, the calculation efficiency and precision of million-level surface patch heat radiation transmission in a complex scene are remarkably improved, and an efficient tool is provided for heat radiation indirect transmission calculation and a global illumination model in the field of three-dimensional scene infrared simulation.

The invention discloses a lightweight CNN (Convolutional Neural Network) image classification method and a lightweight CNN image classification device for assisting agricultural rapid detection. The classification method comprises the following steps: step 1, designing a crop disease and insect pest detection network; 2, designing a patch merging module; 3, designing a convolution block module; 4, preparing an experimental data set; 5, testing the performance of the model; step 6, model performance index determination; and 7, comparing the model parameter quantity with the calculated quantity. The lightweight CNN image classification device is composed of an edge calculation host, a high-precision image acquisition system and a high-speed communication module, and a complete edge calculation solution is formed. The edge computing host adopts NVIDIA JetsonAGX Orin as a core computing unit, integrates a 12-core ARM Cortex-A78AE CPU (central processing unit) and an Ampere architecture GPU (graphics processing unit) with 2048 CUDA (compute unified device architecture) cores, is equipped with a 64GB LPDDR5 memory, and has the beneficial effects that efficient detection and accurate recognition are realized; edge calculation is adaptive, low in consumption and high in efficiency; a data fusion and model optimization mechanism; and light weight and hardware adaptive design are realized, so that good practicability and adaptability are realized.

The invention provides an AI multi-agent and digital twinborn fusion production scheduling process visualization method, medium and system, and belongs to the technical field of AI multi-agent production scheduling. Large-scale parallel computing is achieved by constructing a GPU three-layer CUDA processing architecture, a first layer data preprocessing grid executes data cleaning in parallel, a second layer data preprocessing grid executes data cleaning in parallel, and a third layer data preprocessing grid executes data cleaning in parallel; the second layer of negotiation analysis grid operates an agent interaction recognition model based on Transform to perform parallel mode recognition, the third layer of visual calculation grid operates an LSTM-CNN fused production scheduling process mapping model to generate visual data, and calculation resource allocation is dynamically adjusted through a multi-head attention mechanism. The calculation performance is optimized by adopting a pipeline parallel and data parallel strategy, and the memory access delay is hidden by combining an asynchronous data transmission and calculation overlapping technology, so that the technical problem that the real-time parallel processing of the multi-agent high-frequency negotiation data cannot be realized is solved.

Apparatuses, systems, and techniques to perform multi-threaded memory allocation in parallel by one or more software programs being performed on a parallel processing unit (PPU), such as a graphics processing unit (GPU), or any other processing unit capable of supporting multi-threaded software execution. In at least one embodiment, one or more software programs expressed in part by code using an application programming interface for parallel computing, such as CUDA, perform allocation, search, and deallocation of memory efficiently and in parallel on a GPU.

The invention provides a space-based full-link photoelectronic imaging simulation method based on CUDA (Compute Unified Device Architecture), and solves the problems of complicated calculation, insufficient simulation precision and high hardware requirements in the prior art. The method comprises the following steps: dividing the surface of a target three-dimensional model into triangular surface elements and calculating target intrinsic radiation scattering data of each surface element; generating background intrinsic radiation scattering data in real time through CUDA (Compute Unified Device Architecture) parallel calculation according to the space-time spectrum parameters and the earth surface / cloud layer / limb basic model library; calculating characterization radiation before entrance pupil based on target and background intrinsic data; the entrance pupil radiation is converted into a detector voltage signal through CUDA parallel optimization, and a gray infrared simulation image is generated. According to the method, the calculation efficiency is remarkably improved, and space-based full-link photoelectric imaging simulation with high authenticity and real-time performance is realized.

Apparatuses, systems, and techniques to execute CUDA programs. In at least one embodiment, an application programming interface is performed to indicate one or more attributes of one or more groups of blocks of one or more threads.

The application provides a space-time big data distributed privacy computing system and equipment based on homomorphic encryption, and relates to the technical field of data processing. It comprises a space privacy computing operator, which is used for the computing characteristics of vector data and raster data, and realizes homomorphic encryption computing of vector data and raster data by using TFHE homomorphic encryption algorithm and CKKS homomorphic encryption algorithm respectively; a distributed space privacy computing framework is used, each node of the distributed space privacy computing framework is respectively provided with a space privacy computing operator, and the distributed space privacy computing framework realizes multi-node task parallelization and safe cooperation based on a key management mechanism under a distributed environment and a remote Boolean decryption protocol based on inadvertent transmission; each node in the distributed space privacy computing framework is provided with a GPU acceleration architecture, the GPU acceleration architecture is respectively provided with a corresponding CUDA heterogeneous computing optimization scheme for the Boolean logic operation of vector calculation and the matrix operation characteristics of raster calculation, and the processing efficiency of a single node is improved.

The application relates to a world model construction method, in particular to an end-to-end automatic driving test high-fidelity lightweight world model construction method, which constructs a high-fidelity world model, solves the problems of large world model parameters and low inference efficiency, performs knowledge distillation on the world model, reduces model parameters on the basis of retaining world model generation capacity, and improves inference efficiency; a CUDA operator is self-defined for a world model calculation bottleneck part, memory allocation is optimized, and a single-device multi-thread scheduling and multi-device cooperative calculation method are used to improve the inference efficiency of the high-fidelity world model. The application can construct an end-to-end automatic driving test high-fidelity lightweight world model, effectively solve the problems of low multi-modal information alignment accuracy, poor cross-view and cross-frame consistency, and low inference efficiency of the existing world model, improve the confidence of an end-to-end automatic driving system test process, and greatly accelerate the test efficiency of the end-to-end automatic driving system.

Embodiments of the present specification provide a data processing method and system, the method comprising: a reasoning framework, in response to a cold start instruction sent by a cloud platform, cold starting on a target graphics processor, sending an opening hijacking command to a hijacking module, and during the process of loading first model weights of a first reasoning model, initiating a video memory application to the target graphics processor; the hijacking module, in response to the opening hijacking command, hijacking the video memory application, and redirecting the video memory application to a first virtual address in a fixed virtual address space; the reasoning framework, in a case where it is determined that the first model weights are completed loading, sending an ending hijacking command to the hijacking module, and based on the first model weights in the first virtual address, constructing a first reasoning execution graph to execute a first reasoning task using the first reasoning execution graph. By hijacking the CUDA video memory allocation and redirecting it to a pre-reserved fixed address, the address of the model weights is ensured to be persistent and stable, so that the CUDA Graph continues to be effective in multiple loadings and hot switching.

The invention provides a method and a system for quickly encoding an image in a 3D (three-dimensional) engine. The method for quickly encoding the image in the 3D engine comprises the following steps of: rendering the contents of a plurality of Cameras to the same RenderTexture, and distinguishing the positions of different Cameras; obtaining a bottom layer graphics library pointer of a RenderTexture object through a RenderTexture.GetNativeTexturePtr method provided by Unity, and obtaining a bottom layer graphics library pointer of the RenderTexture object according to the bottom layer graphics library pointer of the RenderTexture object and the bottom layer graphics library pointer of the RenderTexture object and the bottom layer graphics library pointer of the RenderTexture object and the bottom layer graphics library pointer of the RenderTexture object and the bottom layer graphics library pointer of the RenderTexture object. A graphic pointer is obtained through RenderTexture, a resource format is converted and mapped into a CUDA memory, a GPU is used for coding into a preset format, and a video memory is reused to reduce resource consumption. The system comprises modules corresponding to the steps of the method.

The invention discloses a GPU (Graphics Processing Unit) parallel thread block adjustment and kernel function scheduling optimization method and system. The method comprises the following steps: acquiring GPU equipment attributes and kernel function execution attributes; determining a feasible thread block size solution space of each kernel function based on GPU equipment attributes and kernel function execution attributes, and selecting a thread block size which enables the load balance degree evaluation model value to be maximum as optimal execution configuration of the kernel function; on the basis of the determined optimal thread block size of each kernel function, dividing an execution grid of each kernel function into a plurality of kernel function slices, and establishing a concurrent execution relationship between the kernel functions; a CUDA graph is constructed in a mixed mode according to the established concurrent execution relation between the kernel functions and data dependence; and by taking an actual model of particle simulation software as a test object, respectively comparing GPU occupancy rates before and after thread block adaptive adjustment, multi-stream concurrent scheduling and hybrid CUDA graph scheduling optimization, kernel function execution time and overall simulation efficiency, and outputting an optimized particle simulation result.

The present application belongs to the technical field of speech processing, and particularly relates to a double-talking detection and acoustic echo cancellation cooperative processing method based on double CUDA streams. The method comprises the following steps: creating a first CUDA stream and a second CUDA stream on the same graphics processing unit; establishing a shared memory control package for cross-stream control between the first CUDA stream and the second CUDA stream; organizing a microphone near-end signal and a far-end reference signal into continuous frames with a preset frame length and frame shift; in the second CUDA stream, recording a synchronization event after writing sub-band control parameters and frame-level mode markers into the shared memory control package; in the first CUDA stream, performing partition blocking frequency domain sub-band adaptive filtering to update filter coefficients and generate echo suppression output. The present application realizes a highly cooperative, low-delay, high-robust real-time processing system for double-talking detection and echo cancellation, and has operation efficiency, control accuracy and engineering realizability, and is suitable for voice communication, remote conference and intelligent voice interaction systems.

The application belongs to the technical field of video fusion, and particularly relates to a multi-path video real-time fusion method for industrial intelligentization. The method comprises the steps of device installation, mask making, first frame image processing and real-time video fusion. The first frame image processing is to obtain the first frame image of each camera corresponding to the video, process the first frame image of the video, and obtain a to-be-stitched image; the starting position of the to-be-stitched image in the final fusion image and the size of the final fusion image are calculated; the left and right black matrixes to be added for each image are obtained according to the starting position and the size of the final fusion image; the mask template and the black matrixes are uploaded from the CPU end to the GPU end; in the real-time video fusion step, the boundary of each sequence image at the same time is added and fused in turn based on CUDA acceleration, so as to generate a wide-view seamless fusion panoramic image.

The application relates to the technical field of electromagnetic simulation, and provides a time-domain electromagnetic field simulation calculation method based on GPU parallel acceleration and a corresponding product.The method comprises the following steps: setting simulation parameters; precalculating physical coordinates of all grid points, generating an observation point coordinate array and a source point coordinate array, and transmitting the arrays to GPU device memory; initializing a GPU solver, allocating GPU memory resources, and configuring a CUDA thread block and a grid structure, so that the number of threads matches the number of observation points; for each time step corresponding to a time step, source field data of the current time step is asynchronously updated into a ring buffer; when the simulation time exceeds the time required for electromagnetic waves to propagate from a source area to an observation area, a GPU kernel function without atomic operation is started, the electromagnetic field values of each observation point are calculated in parallel based on the precalculated coordinate array and the source field data in the ring buffer; and the calculation results are transmitted from the GPU device back to the host end for storage and post-processing verification.

The invention relates to a video coding method and device based on heterogeneous cooperative work and electronic equipment. The method comprises the following steps: acquiring video data to be processed, inputting a video segment to which a previous frame in the video data belongs to a complexity detection model, and predicting the complexity of a next frame in the video data; the central processing unit extracts a next frame and a reference frame from the video data and sends the next frame and the reference frame to the graphics processor; wherein each of the next frame and the reference frame comprises a plurality of coding tree units; each coding tree unit comprises a plurality of subunits; the graphics processor performs matching search on each subunit contained in the next frame in parallel in the reference frame based on the CUDA core to obtain a motion vector corresponding to each subunit; and according to the complexity and the motion vector, resource scheduling is carried out on a central processor and a graphics processor based on a load balancing principle, and a compressed bit stream is generated. In this way, the graphics processor utilizes the CUDA core to execute matching search on the multiple subunits in parallel, and the coding efficiency is improved.

The application provides an image processing method, device and electronic equipment, the method comprising: dividing a to-be-processed image into a plurality of sub-blocks; processing the plurality of sub-blocks using a plurality of CUDA (Compute Unified Device Architecture) streams, each CUDA stream in the plurality of CUDA streams comprising: a calculation operation and a memory operation, and a sub-block in the plurality of sub-blocks being processed asynchronously between the calculation operation in a first CUDA stream and the memory operation in a second CUDA stream; and splicing and fusing the processing results of the plurality of sub-blocks to obtain a processing result of the to-be-processed image. In the implementation process of the above scheme, the multi-core resources of the GPU are fully utilized through the asynchronous processing between the calculation operation and the memory operation in different CUDA streams, the idle time of the processing unit is reduced, the resource utilization rate of the CUDA stream is maximized, and therefore the overall processing efficiency of the high-resolution image is improved.

This invention relates to the field of autonomous driving technology, providing a CUDA-based collision detection method, apparatus, electronic device, and storage medium. The CUDA-based collision detection method includes: in response to a planned candidate trajectory and collected obstacle information, abstracting the candidate trajectory into a sequence of bounding boxes and the obstacle information into a set of obstacle points; processing the bounding box sequence and the obstacle point set in parallel using CUDA to determine matching pairs with potential collision risks, where each matching pair consists of a bounding box and an obstacle point; and in response to the obtained matching pairs, detecting the collision risk of each matching pair in parallel using CUDA to obtain the collision detection result of the candidate trajectory. This invention utilizes the parallel computing power of CUDA to greatly improve detection efficiency, and further improves the speed and accuracy of collision detection by first coarsely screening matching pairs with potential collision risks and then finely detecting the collision risks of each matching pair, achieving efficient and accurate collision detection of candidate trajectories.

According to the large rotary equipment assembling unbalance optimization method based on GPU parallel acceleration and the video memory control technology, inherent attributes of a rotor are stored in a CUDA constant memory, a dynamic combination batch scheduling mechanism is adopted, and GPU computing resources can be fully utilized in a high-dimensional optimization space of million-level phase combination; and while the calculation accuracy is ensured, the use pressure of a video memory is remarkably reduced, kernel operation interruption and system crash are effectively avoided, and stable execution of a large-scale parallel calculation task is ensured.

The application provides a large model training acceleration method and system based on CUDA shared memory and a PCIe expansion card, hardware configuration information is acquired to identify the PCIe expansion card and build an expansion storage space, and a unified management heterogeneous memory space is formed in combination with GPU shared memory; in the training process, according to a preset initialization strategy, model data blocks are loaded to different levels of the heterogeneous memory, data access frequency is monitored in real time, data migration instructions are dynamically generated based on multiple frequency thresholds, and intelligent scheduling of the data blocks among the levels of storage is realized; a transparent address remapping mechanism is used to enable CUDA kernel functions to seamlessly access the heterogeneous memory space; periodic data snapshots and persistent storage are simultaneously supported, and quick recovery after interruption of training is ensured; the application breaks through the GPU display memory capacity limit, improves the large model training efficiency, reduces the dependence on multi-card hardware and cost, and guarantees data security and training reliability.

The invention discloses a GPU resource dynamic allocation method, system and device and a storage medium, and relates to the technical field of GPU resource management, and the method can accurately grasp the use condition of the current resource through collecting the resource state information and the load condition in real time, and provides an accurate basis for resource allocation. And on the basis of a CUDA API detection mechanism, the actual demand of a task on GPU resources can be accurately judged, and invalid locking and waste of the resources are avoided. And resource allocation is carried out according to the preset dynamic allocation strategy, so that the rationality and high efficiency of resource allocation are ensured. And the GPU resources are released in time when the preset release condition is met, so that the utilization rate and the circulation efficiency of the resources are further improved. In this way, the utilization rate of GPU resources can be remarkably increased, and the problems of GPU resource recovery and re-circulation can be effectively solved.

Apparatuses, systems, and techniques to execute CUDA programs. In at least one embodiment, an application programming interface is performed to indicate two or more blocks of threads to be scheduled in parallel.

According to the hybrid expert model multi-lexical element prediction method and device based on dense-sparse parallel calculation, the medium, the program product and the terminal provided by the invention, parallel execution is realized in a Tensor Core and a CUDA Core through a dense-sparse collaborative scheduling mechanism, the problem of low hardware utilization rate caused by expert activation fragmentation when MTP and MoE are combined is solved, and the prediction efficiency of the hybrid expert model multi-lexical element prediction method and device based on dense-sparse parallel calculation is improved. The hardware efficiency and throughput performance of MTP and MoE joint reasoning are obviously improved; according to the method, the structure of the MoE model does not need to be modified or additionally trained, high-performance reasoning can be realized only through optimization in the operation process, and the method is suitable for a production scene in which the model cannot be modified; according to the dynamic pruning strategy based on the activation frequency and the gating weight, redundant calculation is effectively eliminated, the model reasoning speed is remarkably increased while the precision of the output result is guaranteed, and the stability and reliability of the model generation quality are ensured.

The application relates to a CUDA-based rule fast searching method, device and computer equipment, wherein the CUDA-based rule fast searching method comprises the following steps: obtaining a constant string of each rule, and compiling the constant string through a string matching algorithm to obtain a first Boolean array of single non-repeated characters in the constant string; further, obtaining a to-be-matched string of to-be-matched data, and creating a second Boolean array based on the to-be-matched string; and based on CUDA, performing multi-thread matching processing on the first Boolean array and the second Boolean array; when the to-be-matched string contains all the constant strings, the rule and the data are matched successfully. Through the application, the problem that data cannot be quickly matched with rules is solved, and the efficiency of rule searching is improved.

This application provides a GPU-based parallel spectral delay correction method and apparatus, applicable to the field of artificial intelligence technology. The method includes: acquiring multiple parallel units; determining the current parallel unit in chronological order, wherein all time sub-intervals within the current parallel unit simultaneously undergo multiple iterative correction scans through their respective CUDA blocks until a stopping iteration condition is met; after completing multiple iterative correction scans, serially compensating the time sub-intervals within the parallel unit to obtain the accurate solution for the current parallel unit; and sending the accurate solution of the last time sub-interval in the current parallel unit to the next parallel unit. The correction scan process includes: combining the solution calculations of the computing nodes in the current parallel unit into a batch; calling a batch solver to perform batch calculations to obtain an initial solution for each computing node; and performing parallel batch refinement based on the initial solution through the thread group corresponding to each computing node. This application improves GPU utilization.