libmpeg2-devel — main mailing list for libmpeg2 development

Re: [mpeg2-dev] general doubt about PES packetization.

[Michael N. <mic...@gm...>]

On Sun, Aug 24, 2008 at 07:11:09PM -0700, Michel Lespinasse wrote:
> On Fri, Aug 22, 2008 at 11:51:15PM +0200, anand meher wrote:
> >                I want to clarify one point, will a PES packet carry only one
> > coded picture, say an I or a B or a P? or a particular PES packet can have
> > last slice of a picture ending and then followed immediately by the next
> > picture header??
> 
> The mpeg standard makes no guarantees about this - the PES payload can
> start and end at arbitrary byte aligned positions in the ES stream.
> A startcode in the ES stream could even be split accross two (or up to 4,
> but that'd be beyond perverse) PES packets.
> 
> In practice, it's very common that PES packets are closed after the last
> byte of the last slice of every picture in the ES stream, but you should
> not rely on this if you want to be a compliant decoder.
> 

> I believe mplayer violates this - it seems to mostly get away with it, though.

Iam pretty sure mplayers demuxer does not violate this. Besides mplayer
can use libavformat for demuxing (-demuxer lavf) which does also not depend
on 1 picture per PES either.

And if someone wants streams that dont have ES==PES, the mpeg-ps
generated by ffmpeg should do.

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

I know you won't believe me, but the highest form of Human Excellence is
to question oneself and others. -- Socrates

[mpeg2-dev] [PATCH] remove a set of braces for consistency

[Diego B. <di...@bi...>]

Here is a small patch to remove an extra set of braces from
libmpeg2/motion_comp.c in order to keep things consistent with the rest
of the file.

Diego

[mpeg2-dev] How to obtain motion vectors and DC Coefficients?

[Muhammad A. <M.A...@st...>]

Hi Everyone,

I'm working with compressed video and after looking around, it seems that libmpeg2 is a good tools to decode MPEG-1 or 2 video compression.

In my work, I need to obtain the DC coefficients and also the motion vectors from compressed domain (the idea is to obtain these values without decompressing the whole video).
Is it possible to use libmpeg2 library to obtain these values?
Can somebody help me figure out where to start? I'm really lost right now.

kind regards,
Muhammad AlBaqir

[mpeg2-dev] Custom padding at 32bbp?

[Matthias H. <li...@wi...>]

Hi, I'm having a hard time figuring out where I could set
a custom padding value other than zero for 32bbp frames
after the R, G and B byte.
I guess it's somewhere in rgb.c but due to this optimized
code and my lack of knowledge of MPEG internals it's
kinda hard to find a suitable location in the sourcecode...

Regards,
Matthias H

[mpeg2-dev] SEQUENCE_MODIFIED and changes in byterate

[Marian Ď. <md...@bt...>]

Hi all,

  I'd like to ask for comments about the following behaviour:

Some DVB-S transpoders are using VBR for all TV stations, and seem to
change maxBps in almost every sequence header (see below).

  With libmpeg2-0.5.1, this triggers SEQUENCE_MODIFIED state, which
invokes player routines for aspect ratio changes etc.

  libmpeg2-0.4.1 was deliberately ignoring changes in maxBps and 
gave SEQUENCE_REPEATED state in such case.

  Should the players now check for this situation, or would it make sense
to put back the previous behaviour - i.e. giving SEQUENCE_REPEATED 
from libmpeg2 even if maxBps was changed?


        Thanks & kind regards,

               M.

  

   ff1fe S- --- --- - SEQUENCE MPEG2 MP@ML 720x576 chroma 360x288 fps 25
maxBps 976100 vbv 229376 picture 720x576 display 720x576 pixel 64x45 guessed
118x81
  1b6266 S- --- --- - SEQUENCE_MODIFIED MPEG2 MP@ML 720x576 chroma 360x288
fps 25 maxBps 976150 vbv 229376 picture 720x576 display 720x576 pixel 64x45
guessed 118x81
  2c16ee S- --- --- - SEQUENCE_MODIFIED MPEG2 MP@ML 720x576 chroma 360x288
fps 25 maxBps 705250 vbv 229376 picture 720x576 display 720x576 pixel 64x45
guessed 118x81
  386cc6 S- --- --- - SEQUENCE_MODIFIED MPEG2 MP@ML 720x576 chroma 360x288
fps 25 maxBps 722800 vbv 229376 picture 720x576 display 720x576 pixel 64x45
guessed 118x81
  4f82ee S- --- --- - SEQUENCE_MODIFIED MPEG2 MP@ML 720x576 chroma 360x288
fps 25 maxBps 976150 vbv 229376 picture 720x576 display 720x576 pixel 64x45
guessed 118x81
  661466 S- --- --- - SEQUENCE_REPEATED MPEG2 MP@ML 720x576 chroma 360x288
fps 25 maxBps 976150 vbv 229376 picture 720x576 display 720x576 pixel 64x45
guessed 118x81
  7813b6 S- --- --- - SEQUENCE_MODIFIED MPEG2 MP@ML 720x576 chroma 360x288
fps 25 maxBps 976100 vbv 229376 picture 720x576 display 720x576 pixel 64x45
guessed 118x81

Re: [mpeg2-dev] general doubt about PES packetization.

[Michel L. <wa...@zo...>]

On Fri, Aug 22, 2008 at 11:51:15PM +0200, anand meher wrote:
>                I want to clarify one point, will a PES packet carry only one
> coded picture, say an I or a B or a P? or a particular PES packet can have
> last slice of a picture ending and then followed immediately by the next
> picture header??

The mpeg standard makes no guarantees about this - the PES payload can
start and end at arbitrary byte aligned positions in the ES stream.
A startcode in the ES stream could even be split accross two (or up to 4,
but that'd be beyond perverse) PES packets.

In practice, it's very common that PES packets are closed after the last
byte of the last slice of every picture in the ES stream, but you should
not rely on this if you want to be a compliant decoder.

I believe mplayer violates this - it seems to mostly get away with it, though.

-- 
Michel "Walken" Lespinasse
A program is never fully debugged until the last user dies.

[mpeg2-dev] general doubt about PES packetization.

[anand m. <kvm...@gm...>]

Hi Keith,
              I have a doubt with PES packetization. According to what i
learnt PES packets carry elementary stream data , either video or audio but
not both. I also read that if the PES packet carries video elementary stream
then it carries the contiguous bits from the elementary stream.

               I want to clarify one point, will a PES packet carry only one
coded picture, say an I or a B or a P? or a particular PES packet can have
last slice of a picture ending and then followed immediately by the next
picutre header?? I am working on a small project where
I wrote a small wrapper over the libmpeg2 mpeg2dec.c to *get the whole bytes
* of a coded picture(included with the PES headers and transport stream
headers),The video packets are from a transport stream. i wanted to clarify
the way i do it,  i do it as follows.....

i modify the prototype of decode_mpeg2 of mpeg2dec.c as follows *decode_mpeg2
(uint8_t *start , unit8_t *end, uint8_t *buffer) * where the buffer refers
to the 188 byte packet(video packet) of a transport stream which is passed
explicitly by me. The buffer should contain all the headers of the transport
stream and then followed by the PES headers.
Now in decide_mpeg2 function
 in case STATE_PICTURE : when the picture header and picture coding
extension are recognized, i assume that a new picture has been identified
and then i start pushing the buffers (the 188 byte packets )into a data
structure till i again hit STATE_SLICE state.

so when i encounter STATE_SLICE , my logic assumes that the data structure
has all the data of that particular picture along with the PES headers and
transport stream headers.

I know the logic will work only if a new picture definition will not start
in middle of a PES packet that is right after the last picture has been
identified. It would be really helpful if you can let me know if i am doing
something wrong. I can explain if some point was not clear.


thanks & regards
Anand Meher

Re: [mpeg2-dev] mpeg2dec fails when using the default XVideo output

[Lionel D. <lio...@ya...>]

Hello Göran,

> I'm having another problem.  When trying to run mpeg2dec 0.5.1 on my
> (64 bit) Fedora systems now I get the following X error:
> 
> [libmpeg2-0.5.1]$ src/mpeg2dec -o xv \
> /usr/src/redhat/BUILD/m2vmp2cut-0.65-dev/sg1.d/sg1.m2v
> libmpeg2-0.5.1 - by Michel Lespinasse <wa...@zo...> and Aaron
> Holtzman
> X Error of failed request:  BadMatch (invalid parameter attributes)
>   Major opcode of failed request:  141 (XVideo)
>   Minor opcode of failed request:  19 ()
>   Serial number of failed request:  24
>   Current serial number in output stream:  25
> 
> This happens without any "-o" option or with an explicit "-o xv" as
> above.  Using "-o x11" it works.  My X server does have the XVideo
> extension:
> 
> [~]$ xrdb -symbols | grep XV
> -DEXT_XVideo_MotionCompensation
> -DEXT_XVideo
mpeg2dec has always behaved the same way for me, on two systems with 
different NVidia cards, both running SimplyMEPIS 7.0 (X.org 7.1.1) and 
the proprietary NVidia driver.

A while ago, I searched for occurrences of
"Major opcode of failed request" "(XVideo)". There are several relevant 
matches in help sections of forums and distro/project bug databases. 
People often attribute these XV errors to driver bugs.
But the open NVidia driver doesn't have accelerated 3D support, and I 
don't have an Intel or ATI/AMD card on any other machine I own, so I 
gave up...

The problem happens no matter SHM initialization is activated or not in 
the X11/XV/XV2 video output.


Regards,
Lionel Debroux.

[mpeg2-dev] mpeg2dec fails when using the default XVideo output

[G. U. <go...@ud...>]

Göran Uddeborg writes:
> ... I'm using the svn
> version of libmpeg2.  But it crashes, and in debugging this we have
> found out that even mpeg2dec crashes immediately.  

The good news is that this doesn't happen any more in 0.5.1.

I'm having another problem.  When trying to run mpeg2dec 0.5.1 on my
(64 bit) Fedora systems now I get the following X error:

[libmpeg2-0.5.1]$ src/mpeg2dec -o xv /usr/src/redhat/BUILD/m2vmp2cut-0.65-dev/sg1.d/sg1.m2v
libmpeg2-0.5.1 - by Michel Lespinasse <wa...@zo...> and Aaron Holtzman
X Error of failed request:  BadMatch (invalid parameter attributes)
  Major opcode of failed request:  141 (XVideo)
  Minor opcode of failed request:  19 ()
  Serial number of failed request:  24
  Current serial number in output stream:  25

This happens without any "-o" option or with an explicit "-o xv" as
above.  Using "-o x11" it works.  My X server does have the XVideo
extension:

[~]$ xrdb -symbols | grep XV
-DEXT_XVideo_MotionCompensation
-DEXT_XVideo

[mpeg2-dev] [ANN] libmpeg2 0.5.1

[Sam H. <sa...@zo...>]

   Hello,

   It was bound to happen: last week's unexpected yet acclaimed libmpeg2
0.5.0 release came with its own deal of new bugs. Fortunately, nothing
grave really happened, and the shiny new 0.5.1 solves all the problems
we are aware of.

   From the NEWS file:

This bugfix release fixes issues with the installation of headers and
pkgconfig files.

   For more information, please visit the libmpeg2 website at the
following address: <http://libmpeg2.sf.net/>. The new package can be
found here: <http://libmpeg2.sf.net/files/libmpeg2-0.5.1.tar.gz>.

Have fun,
-- 
Sam.

[mpeg2-dev] No header files installed ?

[Petr U. <pet...@su...>]

Hi list,

when I do typical configure,make,make install with libmpeg2-0.5.0, everything 
compiles and installs OK, but no header files are installed (contrary to 
previous version, where mpeg2convert.h and mpeg2.h were installed).

Do I miss something or is this a bug in libmpeg2 build scripts?

TIA

-- 
Best regards / s pozdravem

Petr Uzel, Packages maintainer
---------------------------------------------------------------------
SUSE LINUX, s.r.o.                          e-mail: pet...@su...
Lihovarská 1060/12                          tel: +420 284 028 964
190 00 Prague 9                             fax: +420 284 028 951
Czech Republic                              http://www.suse.cz   

Re: [mpeg2-dev] [ANN] libmpeg2 0.5.0

[M. Ď. <md...@bt...>]

On Sat, 12 Jul 2008 18:25:35 +0200, Christophe Massiot wrote

> Probably you'll only need to take care of STATE_SEQUENCE_MODIFIED used
> in some rare streams.
> 

Well, it's no so rare these days - every TV station that switches between 4:3
and 16:9 aspect ratios will heavily utilize it. And, e.g. in VLC it doesn't work
yet:

#ifdef STATE_SEQUENCE_MODIFIED
/* FIXME - the above ifdef is always FALSE, even with libmpeg2-0.5.0
 * Need to improve the implementation to avoid invalid/unref pictures */


   With kind regards,

         M.

Re: [mpeg2-dev] [ANN] libmpeg2 0.5.0

[Christophe M. <ma...@vi...>]

On Sat, Jul 12, 2008, Diego Biurrun wrote:

> > There are too few differences between 0.4 and 0.5, and the API is
> > compatible, so it wouldn't make much sense to go on with the branch.
> 
> Oh, OK, I thought there had been API changes.  I will try to update
> MPlayer to the new release soon.

Probably you'll only need to take care of STATE_SEQUENCE_MODIFIED used
in some rare streams.

-- 
Christophe Massiot.

Re: [mpeg2-dev] [ANN] libmpeg2 0.5.0

[Diego B. <di...@bi...>]

On Sat, Jul 12, 2008 at 05:24:09PM +0200, Christophe Massiot wrote:
> On Sat, Jul 12, 2008, Diego Biurrun wrote:
> 
> > > We are glad to announce a new version of libmpeg2 and mpeg2dec.
> > 
> > I'm happy to hear this.  Will you also make another 0.4 branch release
> > or do you consider 0.4 to have reached end of life?
> 
> There are too few differences between 0.4 and 0.5, and the API is
> compatible, so it wouldn't make much sense to go on with the branch.

Oh, OK, I thought there had been API changes.  I will try to update
MPlayer to the new release soon.

Diego

Re: [mpeg2-dev] [ANN] libmpeg2 0.5.0

[Christophe M. <ma...@vi...>]

On Sat, Jul 12, 2008, Diego Biurrun wrote:

> > We are glad to announce a new version of libmpeg2 and mpeg2dec.
> 
> I'm happy to hear this.  Will you also make another 0.4 branch release
> or do you consider 0.4 to have reached end of life?

There are too few differences between 0.4 and 0.5, and the API is
compatible, so it wouldn't make much sense to go on with the branch.

-- 
Christophe Massiot.

Re: [mpeg2-dev] [ANN] libmpeg2 0.5.0

[Diego B. <di...@bi...>]

On Sat, Jul 12, 2008 at 01:59:24PM +0200, Christophe Massiot wrote:
> 
> We are glad to announce a new version of libmpeg2 and mpeg2dec.

I'm happy to hear this.  Will you also make another 0.4 branch release
or do you consider 0.4 to have reached end of life?

Diego

[mpeg2-dev] [ANN] libmpeg2 0.5.0

[Christophe M. <ma...@vi...>]

Hi folks,

We are glad to announce a new version of libmpeg2 and mpeg2dec. Sam 
and I would like to thank all contributors who have sent us patches 
with determination and patience, and who have truly made this release 
possible.

 From the NEWS file :

New optimizations have been provided for SSE2 and ARM instruction sets,
while a lot of warnings triggered by recent compiler changes have been
fixed.

A new function has been added to guess the aspect ratio of streams where
it is not properly coded. Also it is now possible to retrieve the value
of the MPEG-2 repeat_first_field flag, and to be notified when the
sequence header of a stream changes (eg. aspect ratio on-the-fly changes).

Enjoy!

-- 
Christophe Massiot.

Re: [mpeg2-dev] [PATCH] AltiVec vector declaration syntax

[Christophe M. <ma...@vi...>]

At 19:22 +0200 2/07/08, Diego Biurrun wrote:

>  > In the meantime, my proposed patch should be a rather reasonable
>>  workaround.
>
>.. ping ..

Pong ! Applied, thanks,
-- 
Christophe Massiot.

Re: [mpeg2-dev] release?

[Diego B. <di...@bi...>]

On Wed, May 14, 2008 at 02:22:47PM +0200, Christophe Massiot wrote:
> On Wed, Apr 30, 2008, Jean-Baptiste Kempf wrote:
> > On Wed, Apr 30, 2008 at 07:39:00PM +0200, Diego Biurrun wrote :
> > > So, all patches that were posted recently have been applied.  How about
> > > making a libmpeg2 release now?  A branch release would be very helpful
> > > for me since it I could easily import it into MPlayer :)
> > 
> > I am for a release too. :D
> 
> Me too. Walken, is there still work to be done in TRUNK ?

I have a patch pending that should be committed :)

So what's it going to be?  Can we have a release?

Diego

Re: [mpeg2-dev] [PATCH] AltiVec vector declaration syntax

[Diego B. <di...@bi...>]

On Fri, May 30, 2008 at 02:29:43PM +0200, Diego Biurrun wrote:
> Here is a patch to handle AltiVec vector declaration syntax in a
> slightly better way.  Yes, this reverts a patch I sent earlier :)
> 
> Apple's gcc accepts vector declarations with () instead of {} like the
> FSF gcc.  Newer variants of the Apple compiler understand {} as well and
> complain about () when compiling with -maltivec, which is the portable
> and thus preferred flag.  Thus {} should be used where possible.
> 
> AFAIK altivec.h is always necessary when compiling with -maltivec, thus
> checking for ALTIVEC_H is better than __APPLE_CC__.
> 
> A cleaner solution would be to have a small test program in configure
> like
> 
> #include <altivec.h>
> #define AVV(x...) {x}
> int main (void) { (vector int) AVV(1); return 0; }
> 
> and if it compiles successfully, set a HAVE_ALTIVEC_VECTOR_BRACES to 1
> in config.h.  I have done this for FFmpeg and MPlayer, but I have to
> admit I do not know how to hack this into your autoconf scripts.
> 
> In the meantime, my proposed patch should be a rather reasonable
> workaround.

.. ping ..

Diego

[mpeg2-dev] mpeg2dec crashes immediately (maybe only in 64 bit version)

[G. U. <go...@ud...>]

Is this list an appropriate place to report a bug?  I looked for some
kind of bugzilla, but I couldn't find that the project used any.

As part of trying to build Tomi Ollila's m2vmp2cut I'm using the svn
version of libmpeg2.  But it crashes, and in debugging this we have
found out that even mpeg2dec crashes immediately.  So there is
something that doesn't work in libmpeg2 itself.

As I mentioned in the subject, I'm doing this on 64 bit machines.  I'm
using Fedora, with the relevant packages updated to Fedora 9.  It does
not seem to matter what file I give mpeg2dec, it crashes all the
time.  The version of mpeg2dec that is build by Livna, 0.4.1, works
fine on the same machines.

The commands I did was

    svn checkout svn://svn.videolan.org/libmpeg2/trunk libmpeg2
    cd libmpeg2
    ./bootstrap
    ./configure
    make
    src/mpeg2dec .../sg1.m2v

mpeg2dec crashes immediately on startup.  If I run mpeg2dec in the
debugger, I get this backtrace:


#0  mpeg2_idct_copy_sse2 (block=0x12364ee, dest=0x7f73f2597c00 "", stride=720)
    at idct_mmx.c:1227
#1  0x0000000000415f68 in mpeg2_slice (decoder=0x12363c0, 
    code=<value optimized out>, buffer=<value optimized out>) at slice.c:954
#2  0x000000000040b7f8 in mpeg2_parse (mpeg2dec=0x12363c0) at decode.c:188
#3  0x0000000000401df4 in decode_mpeg2 (current=0x123ab90 "", 
    end=<value optimized out>) at mpeg2dec.c:279
#4  0x0000000000403886 in main (argc=<value optimized out>, 
    argv=<value optimized out>) at mpeg2dec.c:758

Any ideas?  I haven't tried to understand the code, but I'd be glad to
help debug things if I can.

[mpeg2-dev] [PATCH] AltiVec vector declaration syntax

[Diego B. <di...@bi...>]

Here is a patch to handle AltiVec vector declaration syntax in a
slightly better way.  Yes, this reverts a patch I sent earlier :)

Apple's gcc accepts vector declarations with () instead of {} like the
FSF gcc.  Newer variants of the Apple compiler understand {} as well and
complain about () when compiling with -maltivec, which is the portable
and thus preferred flag.  Thus {} should be used where possible.

AFAIK altivec.h is always necessary when compiling with -maltivec, thus
checking for ALTIVEC_H is better than __APPLE_CC__.

A cleaner solution would be to have a small test program in configure
like

#include <altivec.h>
#define AVV(x...) {x}
int main (void) { (vector int) AVV(1); return 0; }

and if it compiles successfully, set a HAVE_ALTIVEC_VECTOR_BRACES to 1
in config.h.  I have done this for FFmpeg and MPlayer, but I have to
admit I do not know how to hack this into your autoconf scripts.

In the meantime, my proposed patch should be a rather reasonable
workaround.

Diego

Re: [mpeg2-dev] Patch to fix issues with Visual Studio

[Lionel D. <lio...@ya...>]

--- John Adcock <jo...@ad...> wrote :
> --snip--
> > Thanks for the tips, I'll try again later.
> Here is a more complete patch.

Tested-by: Lionel Debroux <lio...@ya...>

The patch passes `make check` for me on SimplyMEPIS 7.0
(based on Debian stable, GCC 4.1.x) running on a Core 2 Duo;
several other videos look OK.


Lionel.

__________________________________________________
Do You Yahoo!?
En finir avec le spam? Yahoo! Mail vous offre la meilleure protection possible contre les messages non sollicités 
http://mail.yahoo.fr Yahoo! Mail 

Re: [mpeg2-dev] Patch to fix issues with Visual Studio

[John A. <jo...@ad...>]

--snip--
> Thanks for the tips, I'll try again later.

Here is a more complete patch.

Thanks

John

Re: [mpeg2-dev] Patch to fix issues with Visual Studio

[John A. <Jo...@ad...>]

--snip--
> Apart from that, I'd surround the GCC-specific declaration of
> DECLARE_ALIGNED by a #ifdef __GNUC__, like this:
--snip--
> (Note that I updated the comment as well)
Thanks for the tips, I'll try again later.
 
John