73 results about "Memory sharing" patented technology

The invention relates to the technical field of memory expansion, and discloses a memory access system and method for a memory pool, the system is connected with PC I e equipment and CXL equipment, the PC I e equipment is used for sending an access request to a host so as to access the CXL memory pool through the host; the CXL equipment is used for sending a read-write command to the CXL switch so as to access the CXL memory pool through the CXL switch, the system comprises a host which is respectively connected with the PC I e equipment and the CXL switch, and the host is used for receiving an access request sent by the PC I e equipment, converting the access request into the read-write command and sending the read-write command to the CXL switch so as to access the CXL memory pool through the CXL switch, cXL equipment and PC I e equipment which are short in memory are expanded through the CXL memory pool, the memory access system of the memory pool can be compatible with the CXL equipment and the PC I e equipment at the same time, unified memory distribution and memory management are provided for the CXL equipment and the PC I e equipment, memory sharing between the CXL equipment and the PC I e equipment is achieved, memories of the CXL equipment and the PC I e equipment are expanded, and therefore the memory access efficiency and the memory utilization rate are improved.

The invention discloses a class preloading method and device for Java micro-service cold start acceleration, and belongs to the technical field of cloud computing and Java virtual machine optimization. The method comprises the steps of class dependency analysis based on a historical call chain, hierarchical class preloading execution and shared memory snapshot management. The device comprises a class analysis engine, a preloading decision maker and a memory snapshot management and JVM optimization adaptation module. Through intelligent analysis of class popularity, hierarchical preloading and combination of a cross-JVM instance memory sharing technology, the Java micro-service cold start time is reduced from average 4.2 s to 0.8 s, memory occupation is reduced by 45%, compatibility with a standard JVM and a mainstream framework is kept, and the method is suitable for financial transaction, real-time recommendation and other micro-service applications in a Serverless scene.

The invention discloses a production line layout optimization decision support system based on data driving. The system comprises a data acquisition module, a data preprocessing module, a feature analysis module, a layout optimization module and a man-machine interaction module. The data acquisition module establishes a bidirectional data channel with an MES system interface through an Internet of Things sensor, and is used for acquiring equipment state data and production order information in real time; the data preprocessing module is connected with the data acquisition module and is used for receiving the original data stream and executing dynamic cleaning and fusion processing; the feature analysis module interacts with the data preprocessing module through a memory shared database, receives the standardized production feature matrix and outputs a bottleneck analysis result; and the layout optimization module communicates with the feature analysis module through an OPC UA protocol to obtain bottleneck parameters as optimization constraint conditions. The problem that in the prior art, modeling depends on historical data, production fluctuation cannot be responded in real time, and then an optimization scheme lags behind actual requirements is solved.

Datacenter workloads demand flexible memory architectures spanning from rack-level to pod-scale deployments. Embodiments herein disclose systems enabling CXL memory semantics over physical layers based on IEEE 802.3 PMA, facilitating memory disaggregation across datacenter fabric infrastructures. The embodiments comprise processing cores with coherent interconnects, MMUs for address translation, and memory channels supporting substantial memory capacities. Resource Provisioning Units (RPUs) translate between CXL data, optionally encapsulated, transmitted via physical layers based on IEEE 802.3 PMA, and CXL requests, enabling external entities to access memory across different physical address spaces. This architecture provides memory pooling using datacenter network infrastructure, supporting intra-rack memory sharing, inter-pod memory access, distributed AI training across datacenter resources, and elastic memory provisioning for cloud-native applications, overcoming physical layer limitations of traditional CXL implementations while maintaining protocol coherency suitable for GenAL, LLM inference, and HPC workloads.

The present application provides a method, system, device, and medium for establishing shared memory between virtual machines, relating to the field of memory sharing technology. The method includes: upon successfully establishing a two-stage address table of an initiating client, sending a memory message to a receiving client and saving its own client physical address to a virtualization platform; in response to the message, initiating a first request to the virtualization platform; in response to the first request, controlling the receiving client to suspend operation and establish a second-stage address table for each physical core on the receiving client based on the client's physical address, if the second-stage address table of each physical core is stored in the same physical space; upon completion of the establishment, controlling the receiving client to resume operation and sending a completion message to the receiving client; in response to the completion message, invoking a corresponding mapping interface to complete the establishment of its own first-stage address table, thereby achieving shared memory establishment. The purpose is to enhance the flexibility of shared memory establishment.

If sufficient memory resources are allowed access, components on an IC chip can operate faster or provide higher performance relative to power consumption. However, if each component is provided with its own memory, the chip becomes expensive. In the described implementation, memory is shared among two or more components (110, 114, 116). For example, a processing component (116) may include computing circuitry (206) and memory (106) coupled thereto. A multi-component cache controller (114) is coupled to memory (106). Logic circuitry (202) is coupled to the cache controller (114) and memory (106). The logic circuitry (202) selectively divides memory (106) into multiple memory partitions (108). A first memory partition (108-1) may be allocated to computing circuitry (206) and provide storage for computing circuitry (206). A second memory partition (108-2) may be allocated to the cache controller (114) and provide storage for multiple components (110). The relative capacity of memory partitions is adjustable to accommodate fluctuating demand without having to dedicate separate memory to a component.

The invention provides a memory sharing method and equipment, in a computing cluster, when a first computing node needs to borrow a memory, a request containing a borrowing demand can be sent to a management node, the management node determines a borrowing strategy in a plurality of borrowing strategies according to the borrowing demand, and the borrowing strategy is used for borrowing the memory. The method comprises the steps that firstly, a borrowing strategy is established, information of a second computing node meeting the borrowing requirement is determined according to the borrowing strategy, the information of the second computing node is sent to the first computing node, then, the first computing node determines a third computing node providing borrowing memory for the first computing node according to the information of the second computing node, a borrowing request is sent to the third computing node, and finally, the third computing node provides borrowing memory for the first computing node. And the third computing node provides the memory for the first computing node, so that other nodes borrowed with the memory share the memory with the node needing to borrow the memory, the success rate of memory borrowing is improved, and the flexibility of memory sharing is improved.

The application discloses a multi-agent collaborative reasoning method and system based on memory sharing and conflict resolution, comprising: constructing a hybrid structured memory pool containing a vector database for semantic recall and a graph database for storing logical structure; deploying functionally heterogeneous multi-agent to normalize original multi-source data into observation nodes and write into the graph database; using the graph database trigger mechanism to listen to the writing event, and automatically triggering the reasoning to generate the hypothesis node and establish the deduction edge when matching the preset rule; detecting whether there is a logical conflict for the hypothesis node of the same subject, and triggering dynamic arbitration if there is; the arbitration mechanism calculates the weighted score based on the confidence, data source credibility and context consistency calculated based on KL divergence, and finally outputs the reasoning conclusion and evidence chain path. The application breaks the information island through the hybrid memory pool, realizes the automatic reasoning by using the trigger, and significantly improves the accuracy, logical consistency and explainability of reasoning in complex scenarios.

A high-capacity cache memory is implemented by multiple heterogenous DRAM dies, including a dedicated tag-storage DRAM die architected for low-latency tag-address retrieval and thus rapid hit / miss determination, and one or more capacity-optimized cache-line DRAM dies that render a net cache-line storage capacity orders of magnitude beyond that of state-of-the art SRAM cache implementations. The tag-storage die serves double-duty in some implementations, yielding rapid tag hit / miss determination for cache-line read / write requests while also serving as a high-capacity snoop-filter in a memory-sharing multiprocessor environment.

Provided in the present application are a virtual instance management method based on cloud computing technology, and a cloud management platform. The cloud management platform is used for managing a cloud data center, wherein the cloud data center comprises a plurality of servers that access a high-speed interconnect bus inside the cloud data center. The method comprises: receiving a virtual instance memory creation request that is input by a tenant; in response to the virtual instance memory creation request, selecting a first server from a plurality of servers and sending a first control command to the first server; receiving a first virtual instance memory binding request; and in response to the first virtual instance memory binding request, sending a second control command to the first server to instruct the first server to provide, by means of a high-speed interconnect bus, a virtual instance memory located on the first server to a first virtual instance located on a second server for use. By means of the technical solution, memory sharing across servers for virtual instances can be realized, thereby improving data communication efficiency and memory utilization efficiency.

Embodiments of the present disclosure disclose a method of sharing memory cross operating systems, apparatus, a circuit, a medium, and a device. The method of sharing memory cross operating systems includes: determining a first dynamic buffer area in a first operating system according to a memory sharing message; determining an address identifier corresponding to the first dynamic buffer area; and sending the address identifier to a second operating system, wherein the first operating system and the second operating system are in a same integrated circuit.

The invention relates to an edge system with a dual-processor architecture, which integrates a broadband processor and an enterprise network processor through coordinated function configuration. The service level of the packet flow is divided by the post-provisioning data and the telemetry data, and it is determined whether to transmit the packet by the broadband processor or to selectively perform a high-order function, such as deep packet detection, data encryption, or data compression, by the enterprise network processor. Embodiments include logical coupling of virtual area networks, a signalling mechanism of segment routing / micro identity identification, a transport mechanism of memory sharing and without duplication, and a hybrid coupling combining logical control and data transport entity structures.

The utility model discloses a storage server, relates to the computer field, said storage server includes at least one computing module, central backplane, sharing module, wherein each computing module includes mainboard, mainboard includes central processing unit and memory, sharing module includes the storage equipment that is used for providing the service for at least one computing module, the central processing unit includes the central processing unit and the memory. And each calculation module is connected with the sharing module through the middle backboard.

A multi-stream GPU memory multiplexing method for tensor structure equivalence, after the memory block of a computing stream is released, it is put into a memory sharing pool, the method comprises the following steps: extracting the structure information of a tensor, including dimension, data type and step layout, generating a unique structure identifier based on the structure information of the tensor; constructing and maintaining a mapping table, the mapping table stores a plurality of key-value pairs, the key is a unique structure identifier, and the value is a memory address, the memory block is located in the memory sharing pool; when allocating the memory of the tensor, in combination with the current execution context identifier, query whether there is a memory address in the mapping table that matches the unique structure identifier and is not occupied, if so, adjust the pointer to point to the memory address, and allocate the unoccupied memory to the tensor; if not, apply for a memory address and update the mapping table. The method can reuse a large number of non-overlapping tensor spaces without changing the original network structure and execution logic, and significantly reduces the peak memory occupancy.

The embodiment of the invention provides a memory sharing method and device, electronic equipment, a storage medium and a product, and relates to the technical field of memory management. Allocating a virtual memory area of a corresponding first process to a shared virtual memory space bound with the target shared file; in response to an access request of the first process for the target virtual memory area, determining a target shared virtual memory space corresponding to the target virtual memory area; under the condition that the access request is not sent by the first process for the target virtual memory area for the first time, if a page table entry corresponding to a corresponding target shared virtual memory space exists in a target host address space to which the target shared virtual memory space belongs, responding to the access request based on the page table entry; and otherwise, allocating a physical page to the target shared virtual memory space, and generating a corresponding page table entry in the target host address space. The memory consumption can be reduced.

The invention relates to an AI processor based on storage and calculation integration, three-dimensional integration and memory sharing and a control method of the AI processor. The calculation unit is coupled with the at least one storage unit through a three-dimensional integration technology, and the calculation unit comprises a neural network processing unit which is used for executing neural network correlation calculation; the interface device is arranged on the computing unit and is used for performing communication coupling on the AI processor and external main control equipment; the control module is arranged in the storage unit and used for switching the storage unit between a first working mode and a second working mode, and in the first working mode, the storage unit establishes communication coupling with external main control equipment through the interface device and executes storage operation in response to the read-write request; in the second working mode, the storage unit establishes communication coupling with the neural network processing unit, and responds to a calculation request of the neural network processing unit to assist in executing calculation operation; and the functions of high-performance AI acceleration and general storage are considered through working mode switching and time division multiplexing.

Techniques for managing the memory of computing devices that host multiple VMs by selecting optimally sized pages of virtual memory for programs running in the VMs. Virtual memory has traditionally been divided into pages of a single size (e.g., small, large, or huge) that are then offered to programs running in VMs. These page sizes have different benefits and drawbacks that must be considered when selecting the best page size for virtual memory. For example, small pages increase memory sharing between the VMs, but also result in an execution slowdown. Conversely, large pages improve execution performance by reducing latency in address translations, but limit memory sharing. This disclosure describes using small pages for sharable memory, and using large pages for private memory that cannot be shared. This provides the high performance needed for certain applications by using large pages, but still allows for increased memory sharing using small pages.

The invention discloses a transaction closed-loop processing method and system based on double-end collaborative hot update and containerization, and the method comprises the steps: carrying out the encryption storage of a script based on a management end, and pushing the encrypted script to a server; after receiving the script based on the server side, executing decryption and real-time compiling, generating an executable instance and storing the executable instance into a pre-compilation instance pool; the instance is automatically released by the JVM when the reference count returns to zero; the server side is deployed in a Kubernetes cluster in a containerization manner, and the capacity is dynamically expanded and shrunk; the method comprises the following steps of: performing closed-loop processing on a service request by the same server instance: completing multi-step operation through a local database transaction; realizing cross-method state transfer by using JVM memory sharing; the call link depth is constantly a preset value and is directly referenced and executed through a memory. According to the method, the management end pushes the script to the server end, the server end compiles in real time, and Kubernetes containerization capacity expansion and contraction are combined, so that service request closed-loop processing is realized, distributed transaction coordination is eliminated through local transaction and memory sharing, and transaction time consumption is reduced.

The invention discloses a telescope real-time control method and system based on a synchronous time service and memory sharing mechanism, the core of the method is to construct a hard real-time, high-synchronization and low-delay distributed control system for a telescope platform, and the method specifically comprises the following steps: constructing a real-time computing platform; constructing a high-precision time synchronization system; constructing a global shared memory network; defining and initializing a shared memory data area; the deterministic control logic based on the unified time reference is realized; and deploying a watchdog and a health monitoring mechanism. The method has a wide application prospect in the field of precise motion control of astronomical telescopes and the like. According to the method, the problems of poor real-time performance and low synchronization precision caused by network delay, clock asynchronization and process scheduling uncertainty of an existing telescope control system are solved.

The invention relates to a table connection method and device and computer equipment, and belongs to the technical field of data processing.The method comprises the steps that in response to a table connection processing request for a first data table and a second data table, a first occupied memory of the first data table and a second occupied memory of the second data table are determined; and determining the target local shared memory pool according to the minimum occupied memory between the first occupied memory and the second occupied memory. And sending the data table corresponding to the minimum occupied memory to the target local memory shared pool, so that each computing node calls the data table corresponding to the minimum occupied memory from the target local memory shared pool, and executes a table connection operation for the first data table and the second data table. On the basis of the table connection establishment method and device, the network bandwidth occupied by data table transmission during table connection establishment is effectively saved, and due to the fact that each computing node can acquire the data table corresponding to the minimum occupied memory in a local calling mode, the table connection establishment efficiency is effectively improved.

The invention relates to a memory sharing method and device, computer equipment, a storage medium and a program product. The method comprises the steps that in the process that a current thread in multiple threads processes a write request, a target data segment, to be written into a memory pool, of the write request is determined, and a first memory index structure and a second memory index structure of the threads are used for storing memory page indexes of memory pages distributed to the threads in the memory pool; acquiring a memory page index from a first memory index structure of the current thread, and applying for a memory page from an unallocated memory in a memory pool under the condition that acquisition fails; under the condition that the application in the memory pool fails, the memory page index is extracted from the second memory index structure of the current thread, the target data segment is written into a target memory page pointed by the extracted memory page index, and submission pause prompt information is returned to a submitting party of the write request. By adopting the method, the write request processing efficiency can be improved under the condition that memory resources are insufficient.

An accelerator memory sharing method, which is applied to the technical field of artificial intelligence. In the accelerator memory sharing method, the same accelerator is used to simultaneously execute an inference task and a training task, and a portion of a memory on the accelerator is allocated to the inference task, and the other portion of the memory is allocated to the training task. When an inference load of the accelerator changes, memory amounts on the accelerator that are allocated to the inference task and the training task change accordingly, such that the memory amount on the accelerator that is allocated to the inference task can adaptively meet the requirements of the inference load.

Modem datacenters require efficient mechanisms for memory resource sharing between accelerators and host processors to support AI / ML workloads, HPC applications, and distributed computing environments. Embodiments herein disclose systems incorporating RPUs that enable entities to access host memory through UALink-based interconnects. The processor utilizes a coherent interconnect coupling processing cores to memory controllers, with an MMU mapping virtual addresses to physical addresses within the processor's physical address space. The RPU performs hardware-accelerated physical address translations between UALink-associated addresses and the processor's physical address space, enabling entities to access memory via the UALink port, coherent interconnect, and memory controllers. Some embodiments support multiple RPUs with independent UALink ports serving entities with distinct physical address spaces, enabling dynamic resource allocation and memory pooling, which address memory disaggregation challenges for GenAI inference, LLM training, and next-generation datacenter architectures requiring flexible memory sharing across heterogeneous compute elements.

Modern datacenters require efficient mechanisms for memory resource sharing and utilization across distributed computing environments. Some of the disclosed embodiments introduce systems and methods incorporating a Resource Provisioning Unit (RPU) that performs host-to-host physical address translations, enabling external hosts to access memory resources utilizing CXL protocols. The system includes a processor coupled to DRAM, an MMU for virtual-to-physical address mapping, and a CXL device for host communication. The RPU enables hosts to access the DRAM utilizing messages conforming to CXL protocols, including CXL.mem with Host-managed Device Memory (HDM) regions and CXL.io with Transaction Layer Packets. Some embodiments support multiple hosts and CXL memory expanders utilizing additional CXL devices and root ports. The embodiments enable flexible memory architecture, improved resource utilization, and scalable memory sharing suitable for AI workloads, cloud computing, and next-generation datacenter deployments.

The application discloses a memory allocation method, a deployment method and related devices. The method comprises determining the execution order of each operator in the network model, the operator memory, the corresponding dependent operator and the local memory sharing operator; selecting a target operator according to the operator memory and the execution order; and allocating memory for the target operator and the input operator of the target operator according to the dependent operator of the local memory sharing operator of the target operator. The application obtains the dependent operator and the local memory sharing operator of each operator, allocates memory for the operator and the input operator of the operator based on the dependent operator of the local memory sharing operator, and controls the operator that can reuse the memory of the input operator to reuse the memory of the input operator, thereby reducing the required memory resources of the network model. Meanwhile, when the network model is used online, the operator with reused input operator memory is removed from the network model, so that the number of data carrying times in the execution process of the network model is reduced, and the inference speed of the network model is improved.

Embodiments of this application disclose a memory sharing method and an apparatus. The method in embodiments of this application includes: A host side establishes a first memory page table via a shared virtual address management unit. The first memory page table includes mapping from a virtual address on the host side to a physical address on the host side. A first device side establishes a second memory page table via a shared virtual address agent unit. The second memory page table includes mapping from a virtual address on the first device side to a physical address on the first device side. An agent system memory management unit SMMU on the first device side updates the second memory page table based on the first memory page table, or the host side updates the first memory page table based on the second memory page table.

The embodiment of the invention discloses a cross-operating-system memory sharing method, device and circuit, a medium and equipment, and the method comprises the steps: determining a first dynamic cache region in a first operating system according to a memory sharing message; determining an address identifier corresponding to the first dynamic cache region; sending the address identifier to a second operating system; the first operating system and the second operating system are located in the same integrated circuit. According to the embodiment of the invention, an address identifier of a unique identifier is determined for a first dynamic cache region determined in a first operating system and is used for pointing to the first dynamic cache region, and the address identifier is sent to a second operating system located in the same integrated circuit as the first operating system; the second operating system can directly access the first dynamic cache region through the address identifier, so that memory sharing between two different operating systems is realized, and the problem of memory sharing among multiple operating systems is solved.

The invention discloses a multi-device cooperation system for a conference system, and relates to the technical field of comprehensive communication, the multi-device cooperation system comprises an application layer, a core abstraction layer, a driving layer, a kernel layer and a hardware layer, the application layer operates conference application software, the core abstraction layer is a unified hardware abstraction interface layer, and provides a standardized device interface for the application layer; the driving layer and the kernel layer provide hardware driving, a dynamic priority allocation algorithm, an interrupt response and memory sharing mechanism and CPU / co-processing resource scheduling, and the hardware layer comprises a camera, a microphone, a sound box and an FPGA acceleration chip; the method has the advantages that multi-thread communication delay is effectively reduced, system fluency is guaranteed, multi-device cooperation efficiency of the conference machine in a real-time video cooperation scene is improved, high-resolution video streams and low-delay audio transmission are supported, and audio and picture synchronization errors are reduced.

The invention discloses a data redundancy storage control system, and relates to the technical field of electronic circuit control, the system realizes memory selection and memory sharing through a controller, and the data storage capacity is increased; key data are respectively stored in different memories, so that data redundancy control is realized, and the reliability of the data is improved; the processors report the occupation condition of the memories in real time through the I / O interfaces, the occupied memories are released after processing is completed, the capacity of rapidly obtaining multi-memory data is provided, and the task processing real-time performance is improved.

The invention relates to the technical field of intelligent cabins, and discloses a vehicle communication method and device, a vehicle, a storage medium and a program product.The intelligent cabin is divided into an instrument system, a remote communication system and a software application system which are independent, isolation of hardware resources is achieved, independent physical channels are distributed for all the systems, and the communication efficiency is improved. And the isolation and safety of the system are improved. And the instrument system decrypts the first encrypted data sent by the main controller to obtain remote communication data and / or software application data of other systems, so that information transmission between hardware isolation systems is realized, and system data leakage caused by memory sharing can be avoided. And the private communication protocol between the instrument system and the software application system is designed, and the first original data obtained through decryption is sent to the software application system, so that the communication delay between the systems is reduced, and the real-time performance of data display and instruction execution of the instrument system is improved.