On Feb 25, 09 21:08:00 +0100, Christian König wrote:
6. Correct me if i'm wrong but AFAIK the video hardware scales the picture and makes colour space conversion, just to realize that we can throw away 25% of the picture because it just contains black borders which are outside of the screen.
When color conversion is done by pixel shader this is not true, because the pixels aren't even rendered. It's just the upload to video memory that can be optimized. And that can be done transparently in the driver.
Sorry for blaming the player to be braindead, it's a little bit over sized, i should have blamed the television station in the first place. :/
:-P
And i 100% akk with you that backpropagating if parts of the displaying area are visible is like killing a fly with an atomic bomb, but hey from what i know about MPEG2 i would put an cropping feature between the VLD and iDCT stage, so every stage after this one doesn't even sees the black borders (ok doing MC at the picture border must be quite tricky then).
Eventually you could even skip decoding the non-visible parts of the
MPEG2 stream. But things get even more nasty then, because the
non-visible parts could be needed for motion compensation...
CU
Matthias
--
Matthias Hopf