[FFmpeg-devel] [ANNOUNCE] ffmpeg-mt merged

[Michael Niedermayer]

On Tue, Mar 22, 2011 at 07:37:10AM -0400, Ronald S. Bultje wrote:
> Hi,
> 
> On Tue, Mar 22, 2011 at 6:26 AM, Baptiste Coudurier
> <baptiste.coudurier at gmail.com> wrote:
> > On 3/22/11 2:41 AM, Michael Niedermayer wrote:
> >> On Tue, Mar 22, 2011 at 10:24:43AM +0100, Janne Grunau wrote:
> >>> On Mon, Mar 21, 2011 at 08:56:29PM +0100, Michael Niedermayer wrote:
> >>>>
> >>>> * merged ffmpeg-mt in
> >>>
> >>> thanks for ruining the history. see git log --reverse
> >>
> >> you talk to the wrong one here
> >> git log --reverse
> >> run in a new and unchanged ffmpeg-mt checkout looks identical
> >>
> >>
> >>>
> >>>> * fixed it to pass fate
> >>>
> >>> 1/2 of the h264 tests are broken and a couple of others. It might have
> >>> made sense to wait for Alexander and Ronald to fix things really.
> >>
> >> I just retested and all still pass, like they did yesterday when i tested
> >> before push,
> >> if you know of a specific bug that merging ffmpeg-mt introduced, tell me and
> >> ill fix it.
> >
> > fate-h264 passes here as well.
> 
> But does it use threading?
> 
> (Hint: I'm not that stupid.)

no, but you are smart enough to test with threading enabled on the
encoder side too,
This leads to encoding with multiple slices and causes alot more
differences than there are bugs.
Because you know, if you encode with different parameters you get a
different file, a file with 2 slices isnt the same as one with 1 slice

But dont take my word of it, see yourself:
Here is what your unchaneged fate does (paths shortened for readability)

with THREADS=1
ffmpeg -v 0 -threads 1 -y -flags +bitexact -dct fastint -idct simple -sws_flags +accurate_rnd+bitexact -benchmark -qscale 10 -f image2 -vcodec pgmyuv -i tests/vsynth1/%02d.pgm -f mpeg1video tests/data/vsynth1/mpeg1.mpg

db3127d1c005d47279ec3df25fc9091a0ba2be3a  tests/data/vsynth1/mpeg1.mpg


with THREADS=2
ffmpeg -v 0 -threads 2 -y -flags +bitexact -dct fastint -idct simple -sws_flags +accurate_rnd+bitexact -benchmark -qscale 10 -f image2 -vcodec pgmyuv -i tests/vsynth1/%02d.pgm -f mpeg1video tests/data/vsynth1/mpeg1.mpg

e5d0664d89420ccd16babffe34f5754b22db9676  tests/data/vsynth1/mpeg1.mpg


and now compare against THREADS=2 for decoding and THREADS=1 for encoding
ffmpeg -v 0 -threads 2 -y -flags +bitexact -dct fastint -idct simple -sws_flags +accurate_rnd+bitexact -benchmark -qscale 10 -f image2 -vcodec pgmyuv -i tests/vsynth1/%02d.pgm -f mpeg1video -threads 1 tests/data/vsynth1/mpeg1.mpg

db3127d1c005d47279ec3df25fc9091a0ba2be3a  tests/data/vsynth1/mpeg1.mpg


[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

It is dangerous to be right in matters on which the established authorities
are wrong. -- Voltaire
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20110324/7ef1bc42/attachment.asc>