mschickler's profile - activity

Request for Cyclone DDS loaned message working example I've tried Galactic, Rolling, and building cyclone DDS and iceory

I am using ROS 2 galactic and running with my environment set for FastRTPS middleware. I am using the loaned message buffer API and the XML configuration for shared memory described in the Fast-DDS documentation (snippet):

<transport_descriptors>
    <!-- Create a descriptor for the new transport -->
    <transport_descriptor>
        <transport_id>shm_transport</transport_id>
        <type>SHM</type>
    </transport_descriptor>
</transport_descriptors>

<participant profile_name="SHMParticipant" is_default_profile="true">
    <rtps>
        <!-- Link the Transport Layer to the Participant -->
        <userTransports>
            <transport_id>shm_transport</transport_id>
        </userTransports>
    </rtps>
</participant>

When I run my publish code, I get the following output:

Currently used middleware can't loan messages. Local allocator will be used.

My testing shows that latency is increasing as data size increases so it appears that it really isn't using zero copy.

Code can be found at: https://github.com/mschickler/simple_...

It was announced in ROS discourse that the latest Fast-DDS supported zero copy / shared memory. Am I doing something wrong? Does the rmw layer not yet support the new functionality? Any ideas?

Thanks folks. Appreciate the help!

Do you know how I would go about configuring to use Cyclone DDS with iceoryx integration? When I try this test with def

Galactic/Fast-DDS middleware can't loan messages I am using ROS 2 galactic and running with my environment set for FastR

Galactic/Fast-DDS middleware can't loan messages I am using ROS 2 galactic and running with my environment set for FastR

Galactic/Fast-DDS middleware can't loan messages I am using ROS 2 galactic and running with my environment set for FastR

Thanks! It worked. I am a little surprised that it still looks like there is about 100-200 microseconds of latency whi

I've written a program to measure the latency of communication between separate ROS 2 foxy publisher and a subscriber nodes running in a single process. I'm dynamically allocating memory for the message using make_unique() and passing to publish(). My understanding was that this would result in zero-copy communication to the subscriber. However, the latency seems high and the memory address of the underlying message at the subscriber is different than the address of the originally allocated message. A similar test using shared pointer publish with nodelets in ROS 1 shows the memory address is unchanged at the receiver (i.e. zero-copy occurred).

My code is at: https://github.com/mschickler/simple_...

Any help or insight appreciated.

ROS 2 zero copy intraprocess communication I've written a program to measure the latency of communication between separa

The .to_msg() is critical!

This is critical!

The problem is that your unique_ptr never goes out of scope because it is global. Put "client.reset(nullptr);" before r

I think this is actually supposed to be "--cmake-args -DCMAKE_VEBOSE_MAKEFILE=TRUE"

In ROS1, it was possible to make a (synchronous) service call from a callback.

In ROS2, in a derived node class, if I do the following:

1. create a service client using "create_client"
2. within a callback function (e.g. a timer callback) call async_send_request() on the client and then:
3. call get() on the future that is returned

It will block forever. Instead, I must register a callback when I call async_send_request() in order to handle the response.

An async design does have its advantages, so I am wondering, is it the intent of ROS2 to force asynchronous service message handling in a node or is there a way to do a sync call that I haven't uncovered yet?

[ROS2] How to implement a sync service client in a node? In ROS1, it was possible to make a (synchronous) service call f

You can use the following to set the buffering policy to line buffering: stdbuf -o L ros2 launch pkg pkg.launch.py I w