[mythtv] Profiling myth

[James Courtier-Dutton]

On 31 May 2010 14:43, Stuart Morgan <stuart at tase.co.uk> wrote:
> On Monday 31 May 2010 14:38:14 James Courtier-Dutton wrote:
>> Once I have this metric, I can start modifying code and be able to
>> tell if I am really improving it or not.
>
> Just remember, it's very easy to improve the channel change speed, it's just
> very hard to do it without breaking it or making it much worse for some
> people.
>

Maybe so, but having a proper metric so that users can report
improvements or not by just posting log files is more scientific that
what we currently have.
The improvements are going to be gradual, so if it does break it for
some people, we just reverse the change until we have a way of doing
it that does not break it.
I have extensive knowledge (over 10 years) of MPEG technologies and
media players.

The latency can of course be broken down into multiple areas. To name a few:
1) Tune to locked delay.
2) Time to first PAT and required PMT (if needed, this info may be cached)
3) Network latency passing the packets to the client machine.
4) Client playback latency.
5) Various buffers.

Now improving one area might adversely affect other areas, so one
needs the b-a metric first.

... or are you just telling me not to even bother?