On Mon, Mar 24, 2003 at 11:49:04PM +0200, Ville Syrj�l� wrote: > > I think -vop yuy2 is quite optimized. I'm not sure how accurate it's > though.
:-)
> Then again I'm not sure it alwasy has to do the conversion at all.
Right.
> Maybe some codecs will output YUY2 directly when -vop yuy2 is used. I
> really don't know enough about mplayer internals to give a real answer.
OK.
> Anyway performance is quite simple to test with -benchmark.
Right. I will give that a go.
> I origianlly just set the flag in preinit() or somewhere didn't I?
Yes.
> It got
> changed back somewhere.
Really? I thought I actually tested in the mplayer.c main loop to be
sure it was being affected. Let me go test again...
I must have either mis-tested or got the results wrong. I did put the
flag setting code back into pre-init but I do still get the mplayer.c
loop sleeping with:
if(time_frame>0.001 && !(vo_flags&256)){
fprintf(stderr, "mplayer sleeping\n");
...
every now and then.
> I think the right way is returning VFCAP_TIMER
> from query_format().
Indeed, that is how the dxr3 and it's ilk do it. But surely not
getting the semantics correct for setting the flag could not be
justification for removing it, so what other issues are there?
> Well it sort of achieves it with vsync, without threads and if the input
> framerate is the same as output framerate.
Why only sort of? The audio problem you mention below?
> But the real problem is still
> the audio.
What problem? Granted, my input was always NTSC, same as my output
(although my input now is quite frequently 23.98 fps as I am
inverse-telecining material where appropriate to give more of the
compression bandwidth to useful information rather than wasting it on
redundant fields).
I did try returning VFCAP_TIMER from query_format() and I did stop
mplayer.c from sleeping in it's main loop doing so. I played both
NTSC and 23.98 fps material (albeit briefly) and they seemed to stay
in sync, however I was getting:
ALSA lib pcm_hw.c:380:(snd_pcm_hw_start) SNDRV_PCM_IOCTL_START failed: File descriptor
in bad state
errors. Perhaps this is what you are referring to.
b.
--
Brian J. Murrell
pgp00000.pgp
Description: PGP signature
