Port-driven low-latency routing method and system based on PCIe network architecture
By assigning unique codes to switch ports in the PCIe network architecture and embedding the target path in the data packets, the problems of path opacity and imprecise QoS scheduling in the PCIe interconnect architecture are solved, achieving low-latency and efficient data packet forwarding, which is suitable for high-performance computing and real-time communication.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI XINLIJI SEMICON CO LTD
- Filing Date
- 2025-11-25
- Publication Date
- 2026-06-16
AI Technical Summary
The existing PCIe interconnect architecture struggles to achieve explicit path-level control and efficient QoS scheduling in high-density, multi-host collaborative scenarios, resulting in high packet forwarding complexity, opaque paths, and imprecise traffic scheduling, which fails to meet the needs of high-performance computing and real-time communication.
A port-driven low-latency routing method is adopted. By assigning a unique port code to each switch port, a global port mapping table is built, and the target path in the custom prefix is embedded in the data packet. The switch forwards the data packet directly according to the port code, avoiding the need to look up the BDF routing table.
It significantly reduces packet forwarding latency and complexity, improves data throughput efficiency, simplifies chip logic structure, and achieves deterministic and stable link latency, making it suitable for high-performance computing and real-time communication.
Smart Images

Figure CN122226682A_ABST