[mythtv] Suggestion for improving SLOW channel changing
danielk at cuymedia.net
Sat Sep 10 17:53:20 UTC 2005
On Sat, 2005-09-10 at 22:38 +1000, Lincoln Dale wrote:
> Isaac Richards wrote:
> > And yes, this is how you speed up channel changing, not by coming up with
> > crazily overcomplicated schemes.
> i ran 0.17 for a while with setsockopt TCP_NODELAY to disable nagling.
> i didn't verify whether it was a benefit but based on my perception of
> delay, it shaved 100msec of various things.
I think this would be safe for the RingBuffer data socket...
Just in case someone is interested in getting in there and speeding
things up the RingBuffer is a major bottleneck in LiveTV. I actually
made some small changes in this for the signal monitoring, basically
allowing partial packets 256K packets to be sent after a timeout for
remote frontends, but I ended up just padding the dummyrec output
with null packets to get things working properly.
We really need something like adaptive packet size and timeouts based
on current data availability from the recorder. Right now we use the
rate estimation (I think from libav) at startup which isn't entirely
accurate. But when I fidgeted with these values a few weeks ago it
just caused more problems, so I think the first thing to do is
document the current code RingBuffer code....
More information about the mythtv-dev