Hardware accelerated encoding and/or decoding on Jetpack 4.6.2 (Multimedia version 32.7)

From what I managed to find and try out. It seems like there is no hardware acceleration for ffmpeg on Jetpack 4.6.2, only for 5.1 version. Am I wrong?

Also, I found some encoder/decoder samples in JetsonMultimediaAPI that use HW acceleration (NVDEC & NVENC). Does that mean no CUDA is used with that code samples? And if that is the case, would it be faster to allocate buffers on the GPU using CUDA and pass them to NVDEC/NVENC or is it the same as passing buffers from the CPU?

I’m currently on Jetpack 4.6.2 and I would like to use HW acceleration for ffmpeg since I already have encoder/decoder implemented (CPU code) using ffmpeg libs.

What would be an alternative if there is no HW acceleration for ffmpeg on Jetpack 4.6.2?

We have the implementation for hardware decoding. Please check developer guide

The implementation is based on JetsonMultimediaAPI. You may download the source code and take a look. FYI, here is a known issue of building application:
Jetson/L4T/r32.7.x patches - eLinux.org
[ffmpeg] libavcodec.so: undefined reference to v4l2_open


Thank you for your reply. Where exactly is the example code using decode functional flow described here under “Accelerated Decode with ffmpeg”. I can not find any example in doc folder that uses those functions.


I still cannot find a decoding example that uses those functions.

Please refer to this post:
Jetson TX2 and FFmpeg - Can't initialize nvrm channel - #5 by DaneLLL