Skip to content

README.md

Latest commit

 

History

History
73 lines (53 loc) · 2.38 KB

README.md

File metadata and controls

73 lines (53 loc) · 2.38 KB
Raw

multi_thread_demo — Multi-Thread Measurement

Demonstrates XCP measurement and calibration across multiple concurrent threads. Shows thread-local event instances, shared calibration segments with safe atomic access, and experimental context/span instrumentation for duration measurement.

What it demonstrates

Feature How it is demonstrated
Thread-local event instances Each thread creates its own DaqCreateEvent instance
Thread-local measurement Local variables registered per-thread, measured independently
Shared calibration segment Single CalSeg accessed safely from all threads via lock/unlock
Consistent atomic update Multiple parameters updated atomically across all threads
Context and span API Experimental instrumentation to measure code block durations
Multi-thread queue contention Performance benchmark with configurable thread count and sleep time

Files

File Purpose
src/main.c Demo application — thread creation, per-thread events, shared calibration
CANape/ CANape project (A2L auto-upload, XCP UDP, port 5555)

Building

./build.sh examples
./build/multi_thread_demo

Or with CMake directly:

cmake -B build -S . -DXCPLITE_BUILD_EXAMPLES=ON -DCMAKE_BUILD_TYPE=Debug
cmake --build build --target multi_thread_demo
./build/multi_thread_demo

CANape

Open CANape/CANape.ini in CANape. The project is pre-configured for XCP on UDP, port 5555, with automatic A2L upload. If CANape cannot connect, verify the IP address in Device Configuration / Devices / XCP / Protocol / Transport Layer.

Performance

The lock-free queue in XCPlite is designed for minimal contention under multi-thread load. Example statistics with 8 threads at 50 µs sleep time (MacBook Pro M3):

Producer acquire lock time statistics:
  count=1404480  max_spins=4  max=67583ns  avg=76ns

Lock time histogram (1404480 events):
  Range                      Count        %  Bar
  --------------------  ----------  -------  ------------------------------
  0-40ns                    120470    8.58%  #####
  40-80ns                   627770   44.70%  ##############################
  80-120ns                  528258   37.61%  #########################
  120-160ns                  64571    4.60%  ###
  ...

See c_demo for single-thread minimum cycle time benchmarks.