Memory-Centric Communication Mechanism for Real- time Autonomous Navigation Applications
Wei Liu1, Yifan Gong1, Hao Wu1, Jidong Zhai2, Jiangming Jin1
1. TuSimple Inc. 2. Tsinghua University, BNRist
Wei Liu 1 , Yifan Gong 1 , Hao Wu 1 , Jidong Zhai 2 , Jiangming Jin 1 - - PowerPoint PPT Presentation
Memory-Centric Communication Mechanism for Real- time Autonomous Navigation Applications Wei Liu 1 , Yifan Gong 1 , Hao Wu 1 , Jidong Zhai 2 , Jiangming Jin 1 1. TuSimple Inc. 2. Tsinghua University, BNRist Agenda 1. Background &
Wei Liu1, Yifan Gong1, Hao Wu1, Jidong Zhai2, Jiangming Jin1
1. TuSimple Inc. 2. Tsinghua University, BNRist
Resource Isolation Modularity Multi-language Latency Memory Throghput advantage disadvantage braking distance
perception result Inter-Process Communication
푰
+ 풂 풍풄 풖풍 풂‷ 㰍䅁 ੪ ≈ ퟐ%
data movement
Memory Allocation Data Movement Virtual Layer based Architecture Pre-record Memory Allocation problem solution Z-framework
ROS2-like Architecture Z-framework
Z-framework Python Support Problem Socket-base IPC Problem
Z-framework Python Support Problem Socket-base IPC Problem
Process-1 Process-2 Process-3
Low Efficiency for Memory Allocation A Memory Cost Model Memory Pool Techniques page fault Virtual memory Physical memory page-fault
ObjectMap
Efficient Data Assigment Efficient Data Extraction Distributed Environment Supporting
Total IPC Costs of Different Sizes of Images Image IPC Latency from 256KB to 8MB Image IPC Latency from 256KB to 8MB Mainly Solved: memory allocation overhead Mainly Solved: data movement overhead
Total IPC Costs of Different Types of Messages
Performance Improvement with Virtual Layer based Architecture Virtual Layer Technique Pre-record Technique benefit by:
Workflow Latency Workflow Throughput
Real-world workflow test