PCIe P2P-based data processing method and device, and medium

By using the PCIE P2P data transmission method in the computing architecture, direct data transmission is used to use the FPGA board card, GPU device, NVME SSD, and network card to solve the problem of the traditional computing architecture that is difficult to meet high data throughput and low latency. , Achieve more efficient data processing and transmission performance.

CN119336683AActive Publication Date: 2025-01-21INST OF AUTOMATION CHINESE ACAD OF SCI
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411909319.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-24
Publication Date
2025-01-21
Estimated Expiration
2044-12-24

AI Technical Summary

Technical Problem

Traditional computing architectures and data transmission methods are difficult to meet the needs of high data throughput and low latency, especially in scenarios such as scientific simulation, big data analysis and real-time signal processing, resulting in insufficient data processing speed and efficiency.

Method used

Using a data processing method based on PCIe P2P, direct data transmission is carried out through the PCIe bus between FPGA board, GPU device, NVMe SSD and network card, avoiding the CPU's participation in data transfer, and realizing FPGA board and GPU device, FPGA board and NVMe PCIe P2P data transmission between SSD, GPU device and NVMe SSD, GPU device and network card.

Benefits of technology

Improve the performance of data transmission, reduce data transportation between system memory, avoid delays and system loads caused by CPU participation in data transmission, and avoid packet loss problems caused by Ether network transmission.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119336683A_ABST
    Figure CN119336683A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a PCIe P2P-based data processing method and device and a medium, and the method comprises the steps: transmitting data to a GPU device through a PCIe bus between an FPGA board card and the corresponding GPU device according to a corresponding relation between the FPGA board card and the GPU device; calculating the data through the GPU equipment to obtain a corresponding data processing result; a data processing result is transmitted to a rear end through a PCIe bus between the GPU equipment and the network card; according to the method, the NVMe SSD is used for determining the type of equipment communicating with the NVMe SSD according to the data or the data processing result, and the data or the data processing result is stored in the NVMe SSD through the PCIe bus between the NVMe SSD and the type of the equipment, so that the problems of certain delay and system load caused by the fact that data generally needs to move back and forth between the equipment and a host memory in a traditional PCIe data transmission mode are solved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Heterogeneous intelligent calculation system

    CN111159093A

  • Method for managing and deploying bit stream by FPGA cluster

    CN113973049A

  • Server system and communication method thereof

    CN116501684A

  • Computing board card and storage device

    CN216352163U

  • Extended storage device based on pcie bus

    US20200364175A1