[FFmpeg-devel] [PATCH 3/3] libutvideo: Make CODEC_CAP_DR1 capable

Reimar Döffinger Reimar.Doeffinger at gmx.de
Mon Nov 7 19:10:29 CET 2011


On Mon, Nov 07, 2011 at 10:45:30AM -0500, Derek Buitenhuis wrote:
> On 07/11/2011 3:20 AM, Reimar Döffinger wrote:
> >That is_not_  making things better, the point of
> >DR1 is to make things faster.
> >If does not make things_faster_  it's pointless, if
> >it even makes things slower it's just plain nonsense.
> 
> A few people were insistent I hop on the DR1
> bandwagon, so I guess this is a lesson in peer
> pressure. :/

Well, DR1 is something that really ought to be supported,
but for external libraries they need a sanely designed
API that allows supporting it in a way that makes sense.
Well, and admittedly for some codecs
it does not make sense because a memcpy would be faster
than changing decoding to support arbitrary-stride
output - though particularly in an embedded context
there's still the cache and power usage aspects of
copying a whole frame so that it would be good to at least
support slice rendering then.
Either way in this case we are limited to what the library
supports, and if it only supports a low-performance method
there's no point in pretending anything else.


More information about the ffmpeg-devel mailing list