[Ffmpeg-devel] dirac codec

Rich Felker dalias
Wed Jun 22 16:49:02 CEST 2005


On Wed, Jun 22, 2005 at 11:03:59AM +0300, Tuukka Toivonen wrote:
> On Tue, 21 Jun 2005, Rich Felker wrote:
> 
> >Something that requires 10ghz to decode normal sized video is a joke
> >no matter how well it compresses...
> 
> No, that's called "under development". Unless you can point out
> some inherent, not-fixable, thing which make it slow.

It's usually impossible to get performance without building it into
the design at some level. Anyway my point was about the current
implementation. No one will bother with it when the implementation is
total crap and there's another free wavelet codec on the verge of
being totally usable.

> H.264 certainly isn't a joke even though JM requires much, much
> more powerful CPU than 10 GHz for realtime
> (at least for encoding--haven't tested decoding but it isnt fast either)

Encoding time is irrelevant.

> Besides, even 10 GHz requirements would be OK if it compresses
> superiorly. It just would need couple of years waiting...

No, speed will not keep increasing without bound, and making things
slower and slower without bound is not acceptable. At best it's
something like 10-20 years, not "a couple", and at worst it's never.
Remember you're not waiting for the first 10ghz system to hit the
stores. You're waiting until it's baseline, and that won't happen for
a LONG LONG time. As far as we're concerned (MPlayer devs), 300-600
mhz is baseline now.

Rich





More information about the ffmpeg-devel mailing list