Speaker
Anders Hellerup Madsen
Description
To gain cross-platform capabilities in our video processing pipelines, we at Veo Technologies decided to convert our custom GPU-dependent GStreamer elements from CUDA to Vulkan. Today, several of our production video processing pipelines run entirely on GstVulkan, processing hundreds of thousands of sports recordings every week, having performance parity with the CUDA pipelines.
- We currently run the same pipelines on Nvidia Tesla T4, Nvidia Jetson, MacOS, and other GPU vendors.
- We gained hands-on experience with GstVulkan in production.
- We converted many CUDA kernels to Vulkan and Slang.
- We learned a lot about Vulkan synchronisation…
- We solved a lot of challenges and upstreamed a few patches to GstVulkan along the way.
In this talk, we will share our experiences with using GstVulkan in production: which strategies we used, which challenges we ran into, and what we gained.
Speaker Bio
I am a Staff Engineer in the Video Team at Veo Technologies. I have been in the company for 6 years and worked primarily with video processing and GStreamer in that time.
| Duration of the talk |
|---|
