Nvmsgbroker has delay in sending message

Hi everyone. I ran deepstream-test4 on AGX Xavier. So I used 4 videos as input and performed inference on them using resnet10. And the protocol I used is AMQP.

The message I got from the consumer side at the end of video is frame 3698( it’s also written in ds.log). But the last frame has been processed is frame 4198.

So is nvmsgbroker not fast enough to dispatch message? How can I solve this problem? Because it’s only a 3 min video, the delay is around 500 frames.

Thanks in advance.

• Hardware Platform (Jetson / GPU):AGX Xavier
• DeepStream Version: 5.0
• JetPack Version (valid for Jetson only):4.4 developer version
• TensorRT Version:7.0.1

sources/gst-plugins/gst-nvmsgbroker/gstnvmsgbroker.c::gst_nvmsgbroker_do_work

self->nvds_msgapi_do_work (self->connHandle);
// wait 10ms.
g_usleep (10 * 1000);

you could reduce the sleep time for higher throughput.

Thank you. This solved the problem. Does this waiting time have minimum limit?

It does not have limitation.