19 results about "Zero-copy" patented technology

A computer-implemented system for routing artificial intelligence (AI) queries. The system utilizes a zero-copy data pipeline, which processes prompts in memory-mapped buffers to eliminate at least one memory copy operation, thereby reducing latency relative to conventional serialization pipelines. The system continuously verifies AI provider compliance by injecting synthetic prompts containing invisible, Ed25519-signed Unicode watermarks. Algorithmic bias is detected by generating counterfactual “digital twin” prompts and applying Fisher exact statistical testing.Routing decisions for multi-tier autonomous systems are governed by safety-level requirements (ASIL-D, ASIL-B, QM) and may be constrained by external routing directives received via a meta-identifier. A hash-chained manifest, cryptographically signed using Ed25519 and consumed by downstream gateways, is generated for each routing decision, with its Merkle root asynchronously anchored to a blockchain to create a tamper-evident audit trail for regulatory compliance.

The invention discloses a data stream processing method and device, electronic equipment and a storage medium, and relates to the technical field of distributed storage. According to the data stream processing method and device, the data stream transmission mode is pre-judged to be intra-node transmission or cross-node transmission based on a hardware topology mapping table, and read-write caches are dynamically allocated according to the pre-judged data stream transmission mode; data received by the front-end network card is directly written into a controller memory buffer area of the target equipment through point-to-point direct memory access for data streams transmitted in the nodes, and is read and forwarded from the target equipment by the rear-end network card; the sending cache and the receiving cache of the data stream transmitted across the nodes are locked to the non-uniform memory access node to which the target network card belongs, and the data are sent to the target node through the remote direct memory access register memory, so that the problem that the data stream processing is not optimized in combination with hardware topology characteristics can be solved; the technical effects of reducing data replication overhead, avoiding redundant access across hardware nodes, improving data stream transmission efficiency and realizing end-to-end zero replication transmission are achieved.

The invention provides an RDMA-based big data transmission and storage system and method and a semiconductor detection device.The big data transmission and storage system comprises a hardware layer and a software layer, the hardware layer comprises an NVMe SSD storage array, an NVMe RAID acceleration card based on a GPU architecture and an RDMA network card, and the software layer comprises an NVMe SSD storage array, an NVMe RAID acceleration card based on the GPU architecture and an RDMA network card; the software layer comprises a multi-thread transmission module, a self-adaptive memory pool module, a transmission control module, a data verification and retransmission module, a scheduling and optimizing module and a disk falling and storage management module. According to the method, multi-channel transmission parallelization is achieved through data direct connection, zero copy transmission, RAID acceleration and multi-thread parallel RDMA transmission in a hardware link, data is directly connected and transmitted from a memory to the NVMe SSD storage array, data receiving and disk falling thread decoupling directly access the NVMe SSD storage array through asynchronous writing, network overload or storage blockage is avoided, and the data transmission efficiency is improved. The data transmission efficiency and reliability are high, and popularization and application in scenes needing large-scale image transmission processing such as semiconductor detection are facilitated.

A multi-processor communication framework (MCF) based on Transmission Control Protocol / Internet Protocol (TCP / IP) architecture is defined that can be used for inter-processor communication between heterogenous processing nodes. The MCF includes an application layer, a transport layer, a network layer, and an inter-processor communication (IPC) layer for each processing node. The IPC layer can be used to facilitate data transfer between two processing nodes via a physical channel within an electronic device, or through shared memory. The MCF can support multicast, broadcast, unicast, and zero-copy features like the TCP / IP stack.

The invention relates to the technical field of embedded computing and system collaboration, in particular to a multi-operating system heterogeneous collaboration method based on a lightweight message bus and related equipment. According to the method, a cross-operating system data exchange basis is established by initializing a shared memory annular buffer area and setting access authority; a message bus application interface is used for completing multi-system service registration in the unified namespace, and a service identifier mapping table is generated; the sender operating system efficiently writes the message data into a target buffer area based on a zero copy mechanism and triggers a notification signal; the receiver operating system analyzes the message according to the read-write pointer and outputs a processing state; and finally, efficient communication and running state monitoring among multiple operating systems are realized through event synchronization and a watchdog mechanism. According to the method, efficient heterogeneous cooperative execution among multiple operating systems is realized in a light-weight, low-delay and low-power-consumption manner, and the cross-system data interaction efficiency and the operation stability are remarkably improved.

The network interface card (NIC) comprises a controller and a local memory, and the network interface card (NIC) comprises the controller and the local memory. The controller is used for receiving a data message. The data packet includes a data payload, a header, and metadata. The controller is further configured to select a 0 copy buffer from a 0 copy descriptor ring having a host physical address (HPA), and to select a 0 copy buffer from a 0 copy descriptor ring having a host physical address (HPA). The controller is further configured to write the data payload of the data packet to the 0 copy buffer. The controller is also configured to parse the data packet to identify a virtual destination. The controller is also to select a virtual buffer with a virtual address from a descriptor ring of the virtual destination. The controller is also configured to map the virtual address of the selected virtual buffer to the HPA of the 0 copy buffer.

The invention discloses an Android container application display method based on a cross-system zero-copy rendering link, and the method comprises the steps: taking a gap as a host and an Android as a container, building encrypted communication between the two parties through a memory coordination service and an agent daemon process, completing capability negotiation, and generating a cross-system capability matrix; creating a secure shared buffer area pool by the host according to the secure shared buffer area pool, and distributing a unique identifier and a DRM Prime handle for the buffer area; a container side application initiates a buffer area rendering request, a host matches and distributes a buffer area, and a handle is imported into the container side to bind a rendering context; after the application finishes rendering, synchronizing an object mark completion state through the DRM and notifying the host; and the host verifies the data and waits for a synchronization signal, and submits a buffer area handle to a display controller in a zero copy mode at the VSync opportunity to complete on-screen operation.

The invention relates to a distributed knowledge base file storage system based on a high-speed optical fiber network. A hardware architecture comprises storage nodes, an optical fiber switch and an application server. The system optimizes communication through a remote direct memory access technology, and a software architecture comprises a high-speed network driving layer, a joint file system daemon process and a user space file system framework. The united file system daemon process provides a unified virtual directory view, and supports standard file operation and data consistency. And the network optimization synchronization module optimizes data transmission by using a zero copy technology and asynchronous I / O. And the dynamic configuration management module supports dynamic node management and persistent configuration. And the control management interface module provides a UNIX Socket interface, supports JSON format instructions and realizes dynamic scheduling of storage resources. The user space file system framework is responsible for daemon process management, provides a unified storage view service and supports large model knowledge base application. The system provides an efficient and extensible storage solution for a large-scale knowledge base.

The application relates to the technical field of big data processing and distributed computing, in particular to a distributed big data batch processing optimization method based on SpringBatch, which comprises the following steps: multi-dimensional entropy analysis and fragmentation: performing random pre-sampling calculation key value information entropy, dynamically adjusting histogram bucket number, and establishing a dynamic fragmentation strategy based on a work-stealing mode; self-adaptive feedback control: collecting JVM heap memory usage, dynamically adjusting data grabbing size by using an incremental PID algorithm and a variable step limiter, and forming a self-adaptive back pressure mechanism; zero copy and reuse: performing zero copy conversion by using off-heap memory mapping, and reducing object creation frequency by using object pool reuse technology; asynchronous submission and fault tolerance: combining and submitting large transactions based on an asynchronous disk brushing strategy, and performing hierarchical retry operation; the application realizes statistical perception self-adaptive control of a batch processing process, and significantly improves system throughput and data consistency.

The invention discloses a high-real-time power equipment inter-board communication method and system based on a zero copy technology, and relates to the technical field of power system communication, and the method comprises the steps: defining a communication equipment logic instance through an enumeration type, and binding communication equipment through registering a callback function; the registered communication equipment is polled in the high-speed sampling interruption, and a message to be analyzed is received; distinguishing the message type through the type code in the field of the to-be-analyzed message; directly calling a registered callback function to process the data of the message to be analyzed based on the message type; wherein the callback function is only used for data copying. According to the invention, the technical problems of single communication equipment type, poor expansibility, high data transmission delay and insufficient system reliability in the prior art are solved.

Deep neural network checkpoint optimization system and method based on non-volatile memory are provided, where the client module and the server module register corresponding network structures in non-volatile memory and create data indexes and data communication protocols based on remote direct memory access (RDMA) before the start of training deep neural networks, and during the neural network training process, provide zero-copy, asynchronous, end-to-end neural network data persistence, which allows users to perform fine-grained checkpointing to ensure fault tolerance and data persistence without affecting the training speed.

The invention discloses a Java-based big data stream-oriented computing framework and an optimization method thereof. The optimization method comprises the following steps: pre-distributing a GC visible out-of-heap region outside a JVM (Java Virtual Machine), and registering the GC visible out-of-heap region to a G1 / Shendoah region table; automatically generating a data difference log and a difference abstract in TCP transceiving soft interrupt by utilizing eBPF, and locking an old page by combining hardware copy-on-write; calculating a bandwidth time delay product in real time according to a TCP congestion window, and directly driving operator parallelism degree expansion and contraction; in the recovery stage, the physical page frame number is remapped through user faultfd, and cross-Pod thermal migration is completed within 200 ms. A unified SQL interface is exposed to the outside by the framework, direct memory byte codes are automatically generated, a user can enjoy zero copy through zero perception, GC is free of pause, a back pressure instant feedback effect is achieved, throughput is improved by 30-50% under the same hardware, end-to-end delay is reduced by 20-40%, and global pause time is reduced to lt; and 5 seconds per hour.

The invention discloses a memory application method and system for reducing message copy in communication middleware, which comprises the following steps: acquiring an actual allocated memory based on a request memory space and a reserved space, storing user data in the request memory space to obtain a memory block, acquiring an actual memory block address based on the request memory space, and storing the actual memory block address in the reserved space; obtaining a user available address based on the actual memory block address and the reserved space, when header information needs to be added to forward a message, writing the header information into the reserved space to obtain a new message initial address, the length of the message being the length of the user data and the header information, and when the memory block needs to be called again, writing the header information into the reserved space to obtain a new message initial address; and releasing the memory block based on the actual memory block address. The present invention enables zero copy message forwarding. According to the method, the memory management concept of one-time distribution and zero-copy forwarding is put forward for the first time, and the memory performance bottleneck in a message forwarding scene is fundamentally solved through an innovative pre-distribution strategy and an address calculation mechanism.

The invention discloses an efficient asynchronous caching method and system for DAS high-frequency data, and the method comprises the steps: caching acoustic data into a data packet, writing a data pointer of the data packet which is fully written into a data block, and writing a pointer of the data block which is fully written into a data queue. By using the two-stage data structure of the data packet and the data block to store the data, the snapshot can be efficiently created, the lock-free, zero-copy and low-delay (1 millisecond delay) can be realized, the requirement of a continuous memory is greatly reduced, the method is suitable for high-frequency (10khz-100khz) acoustic data of the DAS and a high-efficiency data storage algorithm, and the resource overhead of a CPU (Central Processing Unit) is greatly reduced.

The invention provides a table item learning method and device based on DMA, electronic equipment and a readable medium, and belongs to the field of communication. The table item learning method comprises the following steps: a switching chip encodes learning information into a table writing command to form a learning message when learning is needed; the DMA is written into a memory and is set to be valid in a learning queue record address; the processor writes the address into a table writing queue after detection; the table writing queue executes hard table writing accordingly; and after completion, updating the learning queue into a completion state. According to the table item learning method and device based on DMA, the electronic equipment and the readable medium, the learning message is encoded into the table writing command capable of being directly executed on the switch chip side, the table writing command is written into the memory and the address is registered through the DMA module, a processor can trigger table writing only by transferring the memory address between the learning queue and the table writing queue, and the table writing efficiency is improved. Therefore, a zero-copy and zero-analysis rapid learning link is realized.

Systems, apparatus, articles of manufacture, and methods are disclosed for zero-copy tensor combination via memory management unit remapping. An example method includes identifying physical memory addresses of tensors of a machine learning model, the tensors stored in a key value cache, the tensors including a first tensor and a second tensor stored at respective physical memory addresses, creating a range of virtual addresses for a combined tensor, the combined tensor to represent a combination of the first tensor and the second tensor, and binding the physical memory addresses of the tensors to the range of virtual addresses using a memory management unit.

A computer-implemented method includes receiving a query specifying an operation to perform on a first table of a plurality of data blocks stored. Each data block in the first table includes a respective reference count indicating a number of tables referencing the data block. The method also includes determining that the operation specified by the query includes copying the plurality of data blocks in the first table into a second table and, in response, for each data block of the plurality of data blocks in the first table copied into the second table, incrementing, the respective reference count associated with the data block in the first table, appending, by the data processing hardware, into metadata of the second table, a reference of the corresponding data block copied into the second table.

Disclosed in the present application are a data processing method, a device, a storage medium and a computer program product. The method comprises: at a computing layer, a first device receiving a first processing request sent by a second device, wherein the first processing request is used to request to write first data into a storage layer of the first device, the first data is stored in a first buffer, and the first processing request comprises address information of the first buffer and feature information of the first data; and using the address information of the first buffer and the feature information of the first data to write the first data into the storage layer. In the technical solution provided in the present application, data write processing or data read processing is implemented between devices by means of transmitting address information of a data buffer, thereby achieving zero copy of data. In this way, the present application can reduce resource occupation, overheads and processing delay, thereby greatly improving device performance.

In some implementations, a memory system may store a dataset in a portion of a fabric-attached memory, wherein the dataset is stored in a format that enables zero-copy analysis of the dataset by multiple host devices associated with a distributed workflow. The memory system may establish a respective direct access connection to the portion of the fabric-attached memory with each host device of the multiple host devices associated with the distributed workflow. The memory system may permit each host device, of the multiple host devices, to access the dataset via the respective direct access connection and by using a zero-copy access technique to extract a batch of data objects from the dataset for performing a computation associated with the distributed workflow.