[MPlayer-dev-eng] Delay at start of video playback

[Mike Lieman]

I have already posted a patch to this list to remove the fast-math library
which is causing the optimization issue in ffmpeg when they use a negative
infinity as a flag.  This resolves the issue.

I'll repost it here.  Can someone commit it to svn?


Index: configure
===================================================================
--- configure   (revision 38637)
+++ configure   (working copy)
@@ -2969,10 +2969,10 @@
   elif test "$cc_vendor" != "gnu" ; then
     CFLAGS="-O2 $_march $_mcpu $_pipe"
   else
-    CFLAGS="-O4 $_march $_mcpu $_pipe -ffast-math -fomit-frame-pointer"
+    CFLAGS="-O4 $_march $_mcpu $_pipe -fomit-frame-pointer"
     WARNFLAGS="-Wall -Wno-switch -Wno-parentheses -Wpointer-arith
-Wredundant-decls -Werror=format-security"
     WARN_CFLAGS="-Werror-implicit-function-declaration"
-    extra_ldflags="$extra_ldflags -ffast-math"
+    extra_ldflags="$extra_ldflags"
   fi

On Sun, Sep 1, 2024 at 4:23 PM The Wanderer <wanderer at fastmail.fm> wrote:

> With my just-built copy of MPlayer with current FFmpeg, when I play a
> file that has a video stream (all the ones I've tried so far, at least),
> it hangs for an extended period - with heavy CPU utilization - before it
> brings up the video window. So far as I've measured, the total delay
> appears to be on the order of 20 seconds. Once the video window appears,
> things seem to work normally; CPU utilization goes down, playback is
> normal, and I can seek around in the file without apparent issues.
>
> With one such file, I appended '-v -v -v -v -v -v' to the command line,
> and got in part the following output:
>
> ========8<========
> [mov,mp4,m4a,3gp,3g2,mj2 @ 0x55b746d80620]All info found
> mp_seek(0x55b77974d5b0, 0, 65536)
> mp_seek(0x55b77974d5b0, 0, 65536)
> mp_seek(0x55b77974d5b0, 0, 65536)
> [mov,mp4,m4a,3gp,3g2,mj2 @ 0x55b746d80620]stream 0: start_time:
>                         duration: 348.222875
> [mov,mp4,m4a,3gp,3g2,mj2 @ 0x55b746d80620]stream 1: start_time:
>                         duration: 348.29932
> [mov,mp4,m4a,3gp,3g2,mj2 @ 0x55b746d80620]format: start_time:
>                       duration: 348.29932 (estimate from stream)
> bitrate=398 kb/s
> [mov,mp4,m4a,3gp,3g2,mj2 @ 0x55b746d80620]After
> avformat_find_stream_info() pos: 117417 bytes read:134623 seeks:0 frames:12
> ==> Found video stream: 0
> ========8<========
>
> When the hang occurs, the final visible message is that third mp_seek.
> The following three messages appear one at a time, seconds apart, over
> the delay period; the ones after that appear to come at a more normal
> rate, without visible delays. Those three appear to come from
> ffmpeg/libavformat/demux.c:estimate_timings(); that suggests that the
> issue may be in that function, or possibly in a callsite for the same,
> but the recent changes to that file (per 'git log' in the internal
> FFmpeg checkout) do not seem to show anything that looks suspicious to
> me.
>
> I am not currently sure how to start investigating further, aside from
> possibly trying to bisect (which I could do, but don't have the time for
> this afternoon).
>
> This is fairly significantly impairing to convenient use of MPlayer, but
> does not block me from using it entirely.
>
> --
>    The Wanderer
>
> The reasonable man adapts himself to the world; the unreasonable one
> persists in trying to adapt the world to himself. Therefore all
> progress depends on the unreasonable man.         -- George Bernard Shaw
>
> _______________________________________________
> MPlayer-dev-eng mailing list
> MPlayer-dev-eng at mplayerhq.hu
> https://lists.mplayerhq.hu/mailman/listinfo/mplayer-dev-eng
>