VictorLee
VictorLee
@fhwedel-hoe First of all, when I say IPC, I mean inter-process-comm(socket, shm(minimum copy), zero-copy), intra-process-comm is not included. > ros2 already does support efficient IPC in some circumstances (e.g. when...
> > Oh, this is what you really need, an intra-process shared pointer. > > Indeed. If I there was an option of loaning a message once and then re-use...
> In my case , it costs 0.000736ms. But when I use the VoxelGrid in cuda-pcl, it's worse than pcl. @NJUSTzwh Hi Your time comsuming looks creazy, could you share...
> 请问你的测试机器是? the program was tested on jetson xavier.
@SteveMacenski Thanks for your attention! would you please read this https://github.com/ZhenshengLee/ros2_shm_msgs/issues/25 and give some comments on zero-copy of ros2 from the view of perception?
As the current v0.1 arch is about multi-process-node with StaticSingleThreadedExecutor by a node, I don't see it necessary to add the executor-level-cbg. But I would like to keep it open.