[mythtv] RE:Re: [PATCH] Increase block size

Mark Frey markfrey at fastmail.fm
Tue Dec 23 14:01:44 EST 2003


Edward wrote:
>RE:Re: [PATCH] Increase block size [In reply to]
>> Scaling could
>> start lower than 128k but I don't think it's going to make much
difference
>> since everything up to 128k is all one go.
>
>Is this likely to cause additional latency during the pausing
>phase when channel changing takes place? It used to be pretty
>bad (up to 800ms), but at the moment it seems to be quite
>reasonable, about 50-150ms (thanks to whoever fixed this!).
>
>I'm not sure where the delays occur, but one area
>I haven't investigated is whether the frontend has to consume
>all buffers before playback is completely paused and channel change can
begin?
>
>Probably a load of rubbish, if so sorry for the wasted bandwidth..

Not a load of rubbish. The frontend does wait until the data retrieval
finishes when changing a channel I believe. Since the patch causes larger
blocks to be retrieved it will increase the latency some. How much is hard
to tell. It shouldn't be anywhere near 650ms unless you're using a very slow
network connection. IF the code is currently retrieving data the additional
latency would be somewhere near 25ms (50ms max) for bitrates < 17000 and
75ms (150ms max) for bitrates above that value. This is an average and
assumes a 10mbs connection. But again this would only be if the code is in
the actual retrieval section. The best way to know for sure is to test it
;-) I don't have a frontent != backend setup so there's no significant
additional latency for me, you could try and let us know the actual effect.

-Mark



More information about the mythtv-dev mailing list