rtesbench: A Multi-core Benchmark Framework for Real-Time Embedded Systems
summary
rtesbench is a Python-driven framework that generates C workloads, runs them via platform scripts, captures serial output and plots histograms to measure contention and scalability across three target systems: Nuttx on a 6-core Sony Express, FreeRTOS on XCore-AI (16 logical cores on 2 physical), and TOPAS FMP on a 4-core Stylink. The prototype uses six API access patterns, finds FreeRTOS shows no change when scaling from 7→14 logical cores, shows fine-grained locks speed up parallel access but can regress at max stress, and spots a memory-related core activation failure on Nuttx.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:07Why multi-core benchmarking matters
- 3:36Framework design
- 5:25Framework users
- 6:50Benchmark workflow
- 7:53Prototype implementation
- 9:16Workload evaluation
- 10:20Scalability evaluation
- 11:03Target systems
- 12:07Baseline results
- 12:48Stress workload findings
- 14:14Lock design benefits
- 15:16Scalability insights
- 15:59Hardware/OS flaws
- 16:20Conclusion
related talks
35mMulti-core Application Development with Zephyr RTOS
26mTuning Real-Time Linux for ARM SoCs- 26mTuning Real-Time Linux for ARM SoCs
- 35mMulti-core Application Development with Zephyr RTOS
38mRTOS vs Bare metal Dawn of Constrained Device
16mAutomated Certification and Benchmarking for RISC-V Architectures
24mEmbench 0.5: A Free Benchmark Suite for IoT
12mDemo: RISC-V 64 Bit Debug and Trace