[MPlayer-dev-eng] Audio synchronization

[Eric Anholt]

-nobps fixed the synchronization, thanks. (although, it seems to get a little 
out of sync for a bit after fast-forwarding, but then recovers.  odd.)

I've got the code to the point that it works on my Celerons with all the cpu 
opimizations enabled (3dnow, sse, mmx, etc).  Of course, most of them are 
disabled during runtime so I can't tell for sure if they still work.  I'm 
going to be installing FreeBSD on my P3 laptop and trying it out (giving up 
on RH7.1 after finding it had the broken gcc) to check if SSE is working.  
I'm sure it's not completely right, because I don't have an MMX2 test in the 
CPU detection iirc.  The file is at 
http://gladstone.uoregon.edu/~eanholt/mplayer/mplayer-step1.tar.gz for anyone 
interested.  The code also will not work on non-cpuid-capable machines right 
now, as I just don't know the assembly necessary to get the cpuid detection 
working right.

The goal of this is to be able to make a single package of mplayer that is 
fast on as many cpu combinations as possible.

On Friday 21 September 2001 07:22, you wrote:
> On Friday, 21. September 2001 10:46, Eric Anholt wrote:
> > I've recently discovered mplayer through the FreeBSD mailinglists, and
> > was trying it out on a divx movie I have.  Of course, the first thing
> > that bit me was cpu detection (It said my Celerons had 3dnow support. 
> > Errors followed). I've been working since then off of a July 8 pre5
> > tarball trying to add runtime cpu autodetection, and I've got cpuid
> > working correctly on my machine now (I'll try on a P3, too), and am
> > working autodetecting kernel support for MMX2/SSE.  It's using function
> > pointers and template headers rather than if statements, and I have quite
> > a few to go still, but the process is fairly mechanical.
>
> try if recent cvs still misdetects 3dnow on your celery.
>
> > However, when I was able to download CVS and work against that, the audio
> > on my movie is totally out of sync, late by a little over a second. 
> > Before I could start it up and it matched almost perfectly.  The patches
> > I have to the system are almost the same between the tarball and the CVS,
> > and I'm running the same commandline (tried a few different audio and
> > video options on both of them), but the newer CVS is just out of sync. 
> > Have other people seen this?
>
> Try playing it back with mplayer cvs and -nobps switch, in pre5 -nobps
> still was default but now -bps is default because it works better with vbr
> mp3 audio in divx which is used in many divx lately.

-- 
Eric Anholt
eanholt at gladstone.uoregon.edu