Using VPI in GStreamer

[Honey_Patouceul]

This is an old post in early VPI versions, but you may have a look to:

VPI in a GStreamer pipeline

You may have to register clients and manage ressources for each. So far I assume the same thread is running calls to gpu_process for each running instance, so checking there can help. I’ve quickly tried, the following is far from clean code but it seems working as a starting point just running 2 pipelines. Not tried what if dynamically creating/destroying pipelines, nor evaluated the overhead and limitations. You may improve and share.

rotation_filter.cu

/* * Copyright (c) 2016, NVIDIA CORPOR…