Re: [Mlt-devel] Sluggish performance on MLT 0.7+

[Dan D. <da...@de...>]

On Sun, May 8, 2011 at 11:09 PM, Jonathan Thomas
<jon...@gm...> wrote:
> Thanks for looking into this!  I tested the newest GIT code, and did not see
> a big difference.  I'll continue to test this some more, and try and
> determine what is going on.  When seeking around a bit, it still seems to
> get stuck and starts moving 1 frame every few seconds.  I'll test some more
> tomorrow.
> Thanks!
> -Jonathan

I think I see it for a few seconds after I press play. Then, after
those few seconds it plays normal. Is that what you see?
I did an export to xml, play it with melt, and use the 5 & 6 keys to
pause and play, but it does not happen.

> On Sat, May 7, 2011 at 12:52 AM, Dan Dennedy <da...@de...> wrote:
>>
>> On Fri, May 6, 2011 at 3:38 PM, Dan Dennedy <da...@de...> wrote:
>> > On Fri, May 6, 2011 at 12:01 PM, Jonathan Thomas
>> > <jon...@gm...> wrote:
>> >> Hi Dan!
>> >> Just checking in with you about the de-interlace bug causing slow
>> >> performance.  Any success solving the issue yet?
>> >
>> > Yes, but it involves a bit a refactoring between the framework and the
>> > modules and needed more testing, but I had to prioritize some paid
>> > work the past couple of days. I do expect to check it in this weekend.
>>
>> The changes have been pushed to the git repo. Let me know if you think
>> it is fixed.
>>
>> >> Thanks!
>> >> -Jonathan
>> >>
>> >> On Tue, May 3, 2011 at 11:10 PM, Dan Dennedy <da...@de...> wrote:
>> >>>
>> >>> On Tue, May 3, 2011 at 9:05 PM, Jonathan Thomas
>> >>> <jon...@gm...> wrote:
>> >>> > Okay, good to know.  I will start updating my XML generation code to
>> >>> > use
>> >>> > INT
>> >>> > for frame numbers, instead of Floats.  If there is anything else I
>> >>> > need
>> >>> > to
>> >>> > do, to ensure which deinterlacer is used, etc... please let me know.
>> >>> >  Also,
>> >>> > if there is a way to reuse a producer in the XML, please let me know
>> >>> > the
>> >>> > syntax and I'll make that change as well.  Thanks for the help!
>> >>>
>> >>> heh, well I was going to say, don't reuse producers! There is a
>> >>> side-effect with the way the avformat producer handles audio. I will
>> >>> spare you the technical details for now.
>> >>>
>> >>> Also, the bug with deinterlace is in Mlt with respect to transitions,
>> >>> and it actually looks like a longstanding bug from the analysis so
>> >>> far. So, there is nothing you can do to workaround it.
>> >>>
>> >>> > On Tue, May 3, 2011 at 10:58 PM, Dan Dennedy <da...@de...>
>> >>> > wrote:
>> >>> >>
>> >>> >> On Tue, May 3, 2011 at 4:11 PM, Dan Dennedy <da...@de...>
>> >>> >> wrote:
>> >>> >> > On Tue, May 3, 2011 at 2:15 PM, Jonathan Thomas
>> >>> >> > <jon...@gm...> wrote:
>> >>> >> >> Greetings Dan!
>> >>> >> >> I have received many reports of sluggish playback using OpenShot
>> >>> >> >> with
>> >>> >> >> MLT
>> >>> >> >> 0.7.0 and 0.7.2 (including the GIT trunk).  I have tried to do a
>> >>> >> >> bit
>> >>> >> >> of
>> >>> >> >> research before reporting this bug, but it's a bit difficult to
>> >>> >> >> nail
>> >>> >> >> down.
>> >>> >> >>  It seems that when you take an HD video clip (I have seen
>> >>> >> >> various
>> >>> >> >> formats
>> >>> >> >> and codecs reported), and place it more than 1 time on the
>> >>> >> >> timeline,
>> >>> >> >> MLT
>> >>> >> >> starts to skip and stutter in an unusable way.  However, if the
>> >>> >> >> clip
>> >>> >> >> is
>> >>> >> >> only
>> >>> >> >> on the timeline 1 time, it plays back smoothly.  I'm not sure if
>> >>> >> >> these
>> >>> >> >> steps
>> >>> >> >
>> >>> >> > Are you creating distinct producers for each shot or reusing
>> >>> >> > producers?
>> >>> >>
>> >>> >> I checked, and no, you are not. While reviewing this, I remembered
>> >>> >> something I need to tell you. Stop using floating point for your
>> >>> >> time
>> >>> >> values. In a near future version, floating point will be
>> >>> >> interpreted
>> >>> >> as a clock value (.e.g. 3.0 = 3 seconds).
>> >>> >>
>> >>> >> I found the problem is due to the yadif deinterlacer being selected
>> >>> >> instead of onefield, which is supposed to be used for sdl by
>> >>> >> default.
>> >>> >> I am working on it.
>> >>> >>
>> >>> >> --
>> >>> >> +-DRD-+
>> >>> >
>> >>> >
>> >>>
>> >>>
>> >>>
>> >>> --
>> >>> +-DRD-+
>> >>
>> >>
>> >
>> >
>> >
>> > --
>> > +-DRD-+
>> >
>>
>>
>>
>> --
>> +-DRD-+
>
>



-- 
+-DRD-+