Jetson VC-2 Video Encoding

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

I have a Jetson TX2 working great doing low latency video encoding using GStreamer with the nvv4l2h264enc encoder. I’d like to encode to VC-2 though with the same low latency but it doesn’t seem there’s a GStreamer encoder available for this.

There is the avenc_vc2 encoder but I don’t think it utilizes hardware acceleration on the Jetson because when running gst-inspect-1.0 on the nvv or omx encoders I see in Jetson examples they always show this under the sink capabilities: video/x-raw(memory:NVMM)

Where the avenc_vc2 and other encoders only show “video/x-raw”. I’m under the impression the NVMM is a DMA peripheral that allows data to be transferred between Jetson hardware components. So it would make sense the hardware encoders supported it and the ones without NVIDIA hardware acceleration do not.

So, two part question:

  1. Is my understanding that GStreamer encoders with NVMM listed under sink capabilities means they support Jetson hardware encoding and those without NVMM listed do not support hardware encoding?
  2. Does anyone know of a GStreamer encoder available to do VC-2 encoding?
3

Hi,
In TX2, there is no hardware VC2 encoder. All supported hardware encoders are listed in
document. Please take a look.

For using software encoder on Jetson platforms, please execute sudo nvpmodel -m 0 and sudo jetson_clocks. The steps enable CPU cores at max clock.

1 Like
4

Okay, thank you @DaneLLL

5

@DaneLLL - Hi, follow up question: Theoretically, could someone who was an adept software/firmware engineer, yet not an NVIDIA employee, modify the existing GStreamer VC2 software encoder and enhance it to use the HWENC?

I suppose what I’m asking exactly is if writing code that utilizes the HWENC by a third party (me and/or my organization) is allowed or is this proprietary and/or undocumented and intended to be done only by NVIDIA?

6

Hi,
Since gstreamer plugins are open source, a possible solution is to implement VC2 encoding through CUDA so that the encoding task can be moved to GPU.

7

Thank you @DaneLLL

9

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.