On Sat, Apr 30, 2016 at 10:02 AM Harald Albrecht <harald.albre...@gmx.net>
wrote:
> Tried with your options as suggested. While this gives an overall much
> better CPU usage (42% instead of 18%) during playback, the steep slowdown
> is still present. During the slowdown CPU usage drops significantly down to
> 16%. After the critical section, speed is back to normal with CPU usage
> again of 42%.
>
> On another note, I got a double free or corruption when trying to seek
> forward using "l".
>
>
Using a positive value > 1 for real_time is not very stable for me either.
In Shotcut, I only permit frame-threaded preview when frame-dropping is
disabled (negative real_time value).
As for the reported problem, I ran variations of the following command
line, adjusting from 4 to 7 video tracks above the black track:
melt -profile atsc_1080p_2398 color:black -track 00081.MTS -track 00090.MTS
-track 00091.MTS -track 00092.MTS -track 00093.MTS -track 00094.MTS
-transition affine geometry=0/0:33%x33% a_track=0 b_track=1 -transition
affine geometry=33%/0:33%x33% a_track=0 b_track=2 -transition affine
geometry=66%/0:33%x33% a_track=0 b_track=3 -transition affine
geometry=0/33%:33%x33% a_track=0 b_track=4 -transition affine
geometry=33%/33%:33%x33% a_track=0 b_track=5 -transition affine
geometry=66%/33%:33%x33% a_track=0 b_track=6 -consumer sdl real_time=-3
I did not see any significant drop in frame rate from one more track to the
next when going from 4 to 5, 5 to 6, or 6 to 7. It was a fairly
linear/incremental drop in speed as expected.
> Best regards,
> Harald
>
>
> Am 30.04.2016 um 18:47 schrieb Dan Dennedy:
>
> You must set real_time on the consumer to enable frame threading. Try
> running melt ... -consumer sdl real_time=3
>
> On Sat, Apr 30, 2016, 9:02 AM Harald Albrecht <harald.albre...@gmx.net>
> wrote:
>
>> Dan,
>>
>> thank you very much for your quick reply!
>>
>>
>> affine is rather slow, but it does support frame-threading if you enable
>> that.
>>
>> I've noticed that, but for me it's still good enough; it's an old habit
>> to use it instead of where a composite would be sufficient too. But this is
>> an optimization I can stick to in all the projects to come.
>>
>> My problem here is different as there is a steep degradation exactly at a
>> certain number of clips composited simultaneously.
>>
>> First, see if playing the project with melt outside of Kdenlive exhibits
>> the same problem.
>>
>> Playing the Kdenlive project with melt shows exactly the same slowdown.
>>
>> Next, set environment variable MLT_AVFORMAT_PRODUCER_CACHE=10 to see if
>> that helps with either/both melt or kdenlive.
>>
>> I first did an "export MLT_AVFORMAT_PRODUCER_CACHE=10", then run melt
>> again with my Kdenlive project: same slowdown (meltdown?) when reaching the
>> "critical" section with a certain number of clips simultaneously composed.
>>
>> CPU load still is only at max of 18% for my core i7, before and during
>> this critical section.
>>
>> Any ideas?
>>
>
>
------------------------------------------------------------------------------
Find and fix application performance issues faster with Applications Manager
Applications Manager provides deep performance insights into multiple tiers of
your business applications. It resolves application problems quickly and
reduces your MTTR. Get your free trial!
https://ad.doubleclick.net/ddm/clk/302982198;130105516;z
_______________________________________________
Mlt-devel mailing list
Mlt-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlt-devel
