[vdr] Re: VDR and 2.6.0-test1?

[Michael Hunold <hunold@convergence.de>]

Hello Klaus,

> > I'm using both branches. I prefer DVB because tuning is faster.
There is a bit of FUD going on with "dvb-kernel".

The "problem" is within the saa7146 i2c layer. The "DVB" driver is 
monolithic and has a i2c routine of it's own.

The saa7146 core that's used in "dvb-kernel" drives the av7110 cards, 
budget cards, 5 different analog tv cards. The i2c routine is shared, 
but is not optimized as the DVB driver. If you change something on the 
one end, you break the other.

The transfer speed can be tuned for every different driver, that's not 
the problem. The problem is, that the routine has to handle two differnt 
scenarious: bulk transfer of firmware data (ie. higher transfer rate in 
favour of system responsibility by using busy waiting) against tuning 
(ie. small data transfers, not timing critical).

But: I just did a quick test of DVB CVS against dvb-kernel CVS. *I* did 
not notice any significant differences with my DVB-C with ves1820. 
Tuning from 3sat to ZDF here in Berlin takes the same time, same for 
other channel transitions. IMHO tuning is not time critical -- it's more 
interesting to have a good signal and a good tuner, so you'll get a fast 
lock.

Other cards with other tuners may vary, of course.

> Well, that's a reason for me, too, then, to stay with DVB.
> I hope this is only a momentary thing - or will dvb-kernel
> be tuning slower than DVB in the long run (i.e.: is this
> a principle problem)?
No. If someone comes up with a all-in-one solution that works with all 
cards, we'll be happy. 8-)

> Klaus
CU
Michael.



--
Info:
To unsubscribe send a mail to ecartis@linuxtv.org with "unsubscribe vdr" as subject.