[mythtv] HD3000 buffer overrun solution - patches for hdtvrecorder.cpp/.h against mythtv 0.16

[John Patrick Poet]

I have not actually posted my patch anywhere yet.  I am still trying to
improve it.

I have not looked at your patch yet, but I will try it tonight.  Mine also
adds a ringbuffer thread to keep data flowing from the HD-3000 device
driver.

In my testing, recording from a single card is easy.  Recording from two
cards is still doable.  Recording three shows simultaneously is a serious
challenge.

My goal is to be able to record three shows simultaneously for at least an
hour.  I have achieved this while testing during the day.  I need to test it
during prime-time, since I think the data-rate may be higher then.

I also need to test if I can record two programs for 2 hours, while
recording a third for only an hour.  I am hoping that if one of the shows
drops off after an hour, the other two will be able to continue without any
overrurns.

My patch is actually against Daniel's *new* hdtvrecorder.  He is doing some
cool stuff there, which will eventually even allow us to see subtitles while
watching a show.  The downside, is that Daniel's hdtvrecorder takes more
processing time than the stock CVS hdtvrecorder -- this results in even more
overruns (without the ringbuffer).

I have a 3.0GHz HT P4 with 1 gig of RAM.  To get three programs to make it
an hour without an overrun, I am currently using a ringbuffer size of 96 meg.

I look forward to looking at your patch, since it may give me ideas on how
to make mine work better -- or visa-versa.

If you want to look at what I have done, I will send it to you tonight.

John

--

On Mon, 6 Dec 2004, Eric Anderson wrote:

>
> Hi Daniel -
>
> Sure. If someone can send me a pointer to John's patch (I just joined
> the mailing
> list today, should I search the archive web site for 'buffer overrun'
> or ??-- I
>   wasn't able to find stuff in the past), I'd be happy to review it.
>
> I actually bought two HD3000 cards last month, but based on the initial
> performance I was initially seeing, I didn't think I'd be able to run
> two at
> once. I can certainly give that a try, though.
>
> I'll also do some investigation as to what buffer sizes work.
>
> Note that I haven't seen any time-dependent overruns so far with this
> patch. I'll watch out for that, though.
>
> Regards.
>
> -Eric
>
> On Dec 6, 2004, at 9:41 AM, Daniel Thor Kristjansson wrote:
>
> >
> > Eric, first thanks for contributing to the HDTV code, we need more
> > contributors. John Poet has a similar patch he's been testing with
> > three
> > HD3000 cards. Can you compare notes with him to see which patch to
> > apply, and what buffer sizes to use?
> >
> > We held back on submitting that patch in part because we hoped the
> > newest HD3000 driver might eliminate the problem, but it didn't so I
> > think we'll go ahead and submit it. But we also held back on submitting
> > that patch because he was still seeing buffer overruns starting about
> > 90
> > minutes into recording when using multiple HD3000 cards. I'd like to
> > know if your patch eliminates this problem as well.
> >
> > -- Daniel
> > _______________________________________________
> > mythtv-dev mailing list
> > mythtv-dev at mythtv.org
> > http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-dev
>
>