After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 701111 - matroskademux: seeking in Sintel 4K png-in-matroska file is dogslow
matroskademux: seeking in Sintel 4K png-in-matroska file is dogslow
Status: RESOLVED OBSOLETE
Product: GStreamer
Classification: Platform
Component: gst-plugins-good
git master
Other Linux
: Normal normal
: git master
Assigned To: GStreamer Maintainers
GStreamer Maintainers
Depends on:
Blocks:
 
 
Reported: 2013-05-28 02:55 UTC by Jean-François Fortin Tam
Modified: 2018-05-07 14:30 UTC
See Also:
GNOME target: ---
GNOME version: ---


Attachments
do_seek.c test case (2.83 KB, text/x-csrc)
2017-04-25 09:10 UTC, Tim-Philipp Müller
Details

Description Jean-François Fortin Tam 2013-05-28 02:55:00 UTC
Since the 4K version of Sintel is huge and almost impossible to find nowadays, I made a shortened sample (with Avidemux, without re-encoding). Simply download http://jeff.ecchi.ca/public/sintel-4k-torture-sample.mkv and try seeking it in various places.

The typical seeking time with totem/pitivi on gst 1.x is 15 to 30 seconds.
VLC manages to have typical seeking times of 2-3 seconds (sometimes 4-5).
mplayer somehow manages to seek instantly (?!)

Core 2 Quad Q9300 CPU, no hardware decoding acceleration.
Comment 1 Tim-Philipp Müller 2013-05-28 20:06:00 UTC
Let's move it to the demuxer for now.
Comment 2 Tim-Philipp Müller 2017-04-25 09:10:32 UTC
Created attachment 350381 [details]
do_seek.c test case

Little test program.
Comment 3 Tim-Philipp Müller 2017-04-25 09:15:41 UTC
Intel(R) Core(TM) i7-4702HQ CPU @ 2.20GHz, decoding with gst-libav.

Key-unit seeks:
100 seeks, min: 95ms, max 332ms, avg 220ms

Accurate seeks:
100 seeks, min: 157ms, max 2912ms, avg 1039ms

The file is 23 seconds long and has 5 cue entries. The first and second index entry are 10 seconds apart.

I'm a bit confused because the bug title talks about this being a PNG file (and it's what I remember too), but the file I just downloaded from the link has in fact H.264.

Either way, it's not clear to me there is anything outrageous happening here any longer.

Can we close this, or was this about a different file after all?
Comment 4 Edward Hervey 2018-05-07 14:30:49 UTC
Closing.

Jeff, if you can reproduce the slowness with tim's example program , re-open aand provide the actual file.