I read that NV_vdpau_interop exposes decoded textures to applications as individual fields, even for progressive content. As a result, when I want to get untouched 4:2:0 YCbCr content out of VDPAU, I have to map the two textures as fields and then weave them manually in my own code.
From the documentation at https://www.opengl.org/registry/specs/NV/vdpau_interop.txt it is suggested that this is because the nvidia VDPAU implementation splits textures into fields internally. Since interlaced content is basically dead in 2016, is there any particular reason this is still being done today or has the standard and implementation simply never been updated / reconsidered?
For reference, d3d11va (Windows), VAAPI EGL interop (Intel/Linux) and VideoToolbox (OS X) all expose progressive frames for progressive content, so I’m not entirely sure why VDPAU decided to go the other way here. Comments would be appreciated.