Open
Description
Hello there,
Looking at how transcoding is being called up:
(a). Adding a GPU-based encoder, specifically NVENC, configurable with presets and the global GPU selection option, and;
(b). Most importantly, a GPU-based filter, such as scale_cuda, scale_npp, and yadif_cuda (targeting NVIDIA-based systems) would greatly eliminate the single largest performance bottleneck. CPU-bound scaling and deinterlacing is expensive, consuming precious CPU cycles that eventually lower transcode density per target host.
A good primer on what's possible with this addition can be found here.
Metadata
Metadata
Assignees
Labels
No labels