http://www.mltframework.org/bin/view/MLT/Questions#Does_MLT_take_advantage_of_multi
Instead of real_time=0, try real_time=-1. On your 8-core machine, you might try
real_time=-8. The negative number indicates that frame dropping is disabled.
The real_time property changes the MLT consumer behavior. You might also want
to try tweaking the libx264 threading behavior. You do that by setting:
x264opts="threads=1"
~Brian
From: Javier Matas <javier.ma...@yump.com>
To: Dan Dennedy <d...@dennedy.org>
Cc: mlt-devel@lists.sourceforge.net
Sent: Tuesday, April 28, 2015 4:47 AM
Subject: Re: [Mlt-devel] Repeated frames with melt
Another question, is there any option in melt config to execute this command in
just one core?
The commands of threads and slices seem to do nothing for us. I've proven this
command on melt 0.8.0 and 0.9.0 and got no difference (8 frames repeated in a
8-core machine):
melt -provider in.mov -consumer avformat:out.mp4 vcodec=libx264 an=1
preset=ultrafast vb=2M qmin=10 qmax=22 me_method=hex g=120 threads=0 slices=0
cpu-used=1 real_time=0
Thank you.
2015-04-28 10:53 GMT+02:00 Javier Matas <javier.ma...@yump.com>:
Hello, problem solved!!
We executed that melt command in a machine with only 1 core and it worked.
We've realized that if we execute it on a machine with 2 cores, 2 frames were
repeated. In a machine with 4 cores...4 frames were repeated, ....
I don't know if it's a bug of melt multithreading, but it only works for us in
that case. Machines with one core are the key.
Thank you.
2015-04-27 21:25 GMT+02:00 Dan Dennedy <d...@dennedy.org>:
On Mon, Apr 27, 2015 at 10:34 AM Javier Matas <javier.ma...@yump.com> wrote:
Hello,
We are using melt in order to build videos from .mov fragments, and when we
process the fragments, we realize that the output videos have around 10
repeated frames at the beginning.
The melt command we use is:
melt -provider in.mov -consumer avformat:out.mp4 vcodec=libx264 an=1
preset=ultrafast vb=2M qmin=10 qmax=22 me_method=hex g=120 threads=4 slices=4
And its console output is:
libdc1394 error: Failed to initialize libdc1394No LADSPA plugins were found!
Check your LADSPA_PATH environment variable.[mov,mp4,m4a,3gp,3g2,mj2 @
0x7f14c88d16a0] max_analyze_duration reached[mov,mp4,m4a,3gp,3g2,mj2 @
0x7f14c8988600] max_analyze_duration reached[mov,mp4,m4a,3gp,3g2,mj2 @
0x7f14c8da2900] max_analyze_duration reached[mov,mp4,m4a,3gp,3g2,mj2 @
0x7f14c88e6820] max_analyze_duration reached[mov,mp4,m4a,3gp,3g2,mj2 @
0x7f14c88e0cc0] max_analyze_duration reached[mov,mp4,m4a,3gp,3g2,mj2 @
0x7f14c899e240] max_analyze_duration reached[mov,mp4,m4a,3gp,3g2,mj2 @
0x7f14c88c1ea0] max_analyze_duration reached[mov,mp4,m4a,3gp,3g2,mj2 @
0x7f14c88c9fc0] max_analyze_duration reached[mov,mp4,m4a,3gp,3g2,mj2 @
0x7f14c8da0fe0] max_analyze_duration reached+-----+ +-----+ +-----+ +-----+
+-----+ +-----+ +-----+ +-----+ +-----+|1=-10| |2= -5| |3= -2| |4= -1| |5= 0|
|6= 1| |7= 2| |8= 5| |9= 10|+-----+ +-----+ +-----+ +-----+ +-----+ +-----+
+-----+ +-----+
+-----++---------------------------------------------------------------------+|
H = back 1 minute, L = forward 1 minute ||
h = previous frame, l = next frame || g =
start of clip, j = next clip, k = previous clip || 0 =
restart, q = quit, space = play
|+---------------------------------------------------------------------+[mov,mp4,m4a,3gp,3g2,mj2
@ 0x7f14c94cd880] max_analyze_duration reached[mov,mp4,m4a,3gp,3g2,mj2 @
0x7f14c94d31e0] max_analyze_duration reached[mov,mp4,m4a,3gp,3g2,mj2 @
0x7f14c94ce960] max_analyze_duration reached[libx264 @ 0x7f14c98b8a20]
lookaheadless mb-tree requires intra refresh or infinite keyintCurrent
Position: 488
Tho OSs we use are debian 7 and ubuntu 14.04, and we've realized that the
number of repeated frames depends only on the machine in which we're executing
it (from 5 to 12). This causes annoying stops in the videos.
Our theory is that the architecture of the machine has a lot to do with these
problems, but trying to do the fine tuning with threads, slices and other
parameters like thread_type or real_time is proving useless.
We usually run this command from a Docker with debian 7 and melt 0.8.0, but as
I said before, the OS and the melt version don't matter because in each machine
we always have the same number of repeated frames, and this number is only
different among different host machines.
Thank you very much.
I do not understand how or why different hosts running the same code will
produce different numbers of repeated frames for the same input. However,
Debian 7 and Ubuntu 14.04 probably have different versions of
libavformat/libavcodec, and melt v0.8.0 is quite old.
I recommend that you try one of these clips in Shotcut on your desktop. You can
see if reproduces by stepping frame-by-frame in the player from the beginning
of the clip, by encoding the clip as H.264 within Shotcut, or by using the melt
that comes with Shotcut. (If using Linux, you must use the wrapper script at
Shotcut.app/melt instead of running Shotcut.app/bin/melt directly.)
------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Mlt-devel mailing list
Mlt-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlt-devel
------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Mlt-devel mailing list
Mlt-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlt-devel