I'm using decode from system memory buffers into system memory buffers, using hardware acceleration. Unfortunately, this gives me NV12 buffers that I then have to convert on the CPU to RGB32 for presentation, which is very slow. Is it possible to have the conversion be performed by the hardware, and what would be the simplest way of doing that?
I have checked the reference manual, and under Hardware Acceleration there is a section explaining that under D3D9 and D3D11, the decoder can have a "Decoder Render Target" of type RGB32, however I don't understand what that entails and how to set it up. Does that require using D3D surfaces as output? Where can you specify that you want RGB32 as output? Is there a way to still automatically convert back to system memory buffers?