Use memcpy instead of std::copy when bridging images

[caguero]

🦟 Optimization

Summary

While testing ros <-> gz communication using the bridge I noticed that the bridge was talking quite a bit of time copying images from Gazebo to ROS. I found that the std::copy operation that we're doing is substantially slower than the memcpy alternative. I think that in principle this shouldn't happen but the numbers are quite clear. Perhaps std::copy is doing something that doesn't use cache effectively...

How to test it?

First, modify this code to see some stats:

auto beg = std::chrono::high_resolution_clock::now();
// Option 1: Using memcpy
// memcpy(ros_msg.data.data(), gz_msg.data().c_str(), gz_msg.data().size());

// Option 2: Using std::copy
auto count = ros_msg.step * ros_msg.height;
std::copy(
  gz_msg.data().begin(),
  gz_msg.data().begin() + count,
  ros_msg.data.begin());

auto end = std::chrono::high_resolution_clock::now();
auto duration = std::chrono::duration_cast<std::chrono::microseconds>(end - beg);
std::cerr << "Elapsed Time: " << duration.count() << " us" << std::endl; 

Recompile and launch one of our examples that publish 320x240 images:

ros2 launch ros_gz_sim_demos gpu_lidar_bridge.launch.py

The default code shows:

[parameter_bridge-2] Elapsed Time: 548 us

Enable the memcpy, comment the old code and relaunch the example again. The new code shows:

[parameter_bridge-2] Elapsed Time: 11 us

Checklist

• Signed all commits for DCO
• Added tests
• Updated documentation (as needed)
• Updated migration guide (as needed)
• Consider updating Python bindings (if the library has them)
• codecheck passed (See contributing)
• All tests passed (See test coverage)
• While waiting for a review on your PR, please help review another open pull request to support the maintainers

Note to maintainers: Remember to use Squash-Merge and edit the commit message to match the pull request summary while retaining Signed-off-by messages.

[traversaro]

Not blocking, but a comment on why std::copy may be slow: perhaps the type being copied is not trivially copiable? In theory this can be tested via a static_assert with std::is_trivially_copyable (that in theory may be a good thing to check even before using memcpy).

[azeey]

CI seems to be broken on the ros2 branch. I'll be taking a look tomorrow.

[ahcorde]

https://github.com/Mergifyio backport jazzy

[mergify]

backport jazzy

✅ Backports have been created

• #585 Use memcpy instead of std::copy when bridging images (backport #565) has been created for branch jazzy