libmpeg2-devel — main mailing list for libmpeg2 development

[mpeg2-dev] Crashes in sse2_idct x86_64

[John S. <ste...@je...>]

Have there been any new developments on this problem?
http://sourceforge.net/mailarchive/message.php?msg_id=25140766

I am seeing similar (if not identical) behavior from libmpeg2 compiled with the mingw64 cross tools for fedora 
(available here http://build1.openftd.org/mingw-w64/)
After disabling gcc optimizations and enabling debug, gdb indicates the crash happens in sse2_idct at the first movdqa_m2r.

There are also reports of similar crashes in sse2_idct due to recent gcc 4.6 changes. 
http://comments.gmane.org/gmane.linux.mandrake.bugs/201430
The patches in this thread fix the problem on fedora 15 where this is reproducible, but not for mingw64 builds 
unfortunately.

Re: [mpeg2-dev] [PATCH] Replace int_fast types with their standar d POSIX counterparts.

[Rémi Denis-C. <re...@re...>]

On Tue, 17 May 2011 21:03:44 +0200, Diego Biurrun <di...@bi...>
wrote:
>> If they really are missing on a real-life system you can replace them
>> with autofoo AFAICT.
> 
> Just replacing them with their equivalents without "fast" in the name
> seems way easier. 

It is easier for you. But there must be a reason why the author used those
types. I guess they really are int_fast32_t is typically a 64-bits integer
on Alpha; that's for compilation. For development, I think telling that we
only care about the low order 32-bits is also valuable. So neither int32_t
nor int64_t fit here IMHO.

-- 
Rémi Denis-Courmont
http://www.remlab.net/

Re: [mpeg2-dev] [PATCH] Replace int_fast types with their standard POSIX counterparts.

[David I. L. <dl...@vt...>]

On Mon, May 16, 2011 at 7:54 PM, Diego Biurrun <di...@bi...> wrote:
> The int_fast types are less portable because they are a part of the
> standard that is actually implemented on fewer systems.
>

Which systems lack int_fast* support?

-dave

Re: [mpeg2-dev] [PATCH] Replace int_fast types with their standard?POSIX counterparts.

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

On Tue, May 17, 2011 at 06:05:48PM +0300, Rémi Denis-Courmont wrote:
> Le mardi 17 mai 2011 02:54:28 Diego Biurrun, vous avez écrit :
> > On Mon, May 16, 2011 at 10:53:56PM +0300, Rémi Denis-Courmont wrote:
> > > Le lundi 16 mai 2011 21:27:13 Diego Biurrun, vous avez écrit :
> > > > The _fast integer types provide no realworld benefits, but may
> > > > introduce portability issues and are just plain ugly.
> > > 
> > > int_fastXX_t are in ISO C just as intXX_t. So IMHO I don't see how they
> > > are less portable, less standard or less POSIX.
> > 
> > Maybe my commit message is a bit misleading - I intended for "standard
> > counterparts" to be read as "more common counterparts that als happen
> > to be (just as) standard", not as "counterparts that are POSIX standard,
> > unlike the originally used types".  I can adjust the log message.
> > 
> > The int_fast types are less portable because they are a part of the
> > standard that is actually implemented on fewer systems.
> 
> If they really are missing on a real-life system you can replace them with 
> autofoo AFAICT.

Just replacing them with their equivalents without "fast" in the name
seems way easier.  Anyway, is the patch accepted or rejected?

Diego

Re: [mpeg2-dev] [PATCH] Replace int_fast types with their standard POSIX counterparts.

[R. Denis-C. <re...@re...>]

Le mardi 17 mai 2011 02:54:28 Diego Biurrun, vous avez écrit :
> On Mon, May 16, 2011 at 10:53:56PM +0300, Rémi Denis-Courmont wrote:
> > Le lundi 16 mai 2011 21:27:13 Diego Biurrun, vous avez écrit :
> > > The _fast integer types provide no realworld benefits, but may
> > > introduce portability issues and are just plain ugly.
> > 
> > int_fastXX_t are in ISO C just as intXX_t. So IMHO I don't see how they
> > are less portable, less standard or less POSIX.
> 
> Maybe my commit message is a bit misleading - I intended for "standard
> counterparts" to be read as "more common counterparts that als happen
> to be (just as) standard", not as "counterparts that are POSIX standard,
> unlike the originally used types".  I can adjust the log message.
> 
> The int_fast types are less portable because they are a part of the
> standard that is actually implemented on fewer systems.

If they really are missing on a real-life system you can replace them with 
autofoo AFAICT.

-- 
Rémi Denis-Courmont
http://www.remlab.info/
http://fi.linkedin.com/in/remidenis

Re: [mpeg2-dev] [PATCH] Replace int_fast types with their standard POSIX counterparts.

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

On Mon, May 16, 2011 at 10:53:56PM +0300, Rémi Denis-Courmont wrote:
> Le lundi 16 mai 2011 21:27:13 Diego Biurrun, vous avez écrit :
> > The _fast integer types provide no realworld benefits, but may introduce
> > portability issues and are just plain ugly.
> 
> int_fastXX_t are in ISO C just as intXX_t. So IMHO I don't see how they are 
> less portable, less standard or less POSIX.

Maybe my commit message is a bit misleading - I intended for "standard
counterparts" to be read as "more common counterparts that als happen
to be (just as) standard", not as "counterparts that are POSIX standard,
unlike the originally used types".  I can adjust the log message.

The int_fast types are less portable because they are a part of the
standard that is actually implemented on fewer systems.

Diego

Re: [mpeg2-dev] [PATCH] Replace int_fast types with their standard POSIX counterparts.

[R. Denis-C. <re...@vi...>]

Le lundi 16 mai 2011 21:27:13 Diego Biurrun, vous avez écrit :
> The _fast integer types provide no realworld benefits, but may introduce
> portability issues and are just plain ugly.

int_fastXX_t are in ISO C just as intXX_t. So IMHO I don't see how they are 
less portable, less standard or less POSIX.

-- 
Rémi Denis-Courmont
http://git.remlab.net/cgi-bin/gitweb.cgi?p=vlc-courmisch.git;a=summary

[mpeg2-dev] [PATCH] Replace int_fast types with their standard POSIX counterparts.

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

The _fast integer types provide no realworld benefits, but may introduce
portability issues and are just plain ugly.
---
 libmpeg2/idct_alpha.c |   14 +++++++-------
 1 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/libmpeg2/idct_alpha.c b/libmpeg2/idct_alpha.c
index 8e94219..490b659 100644
--- a/libmpeg2/idct_alpha.c
+++ b/libmpeg2/idct_alpha.c
@@ -53,7 +53,7 @@ do {					\
 #else
 #define BUTTERFLY(t0,t1,W0,W1,d0,d1)	\
 do {					\
-    int_fast32_t tmp = W0 * (d0 + d1);	\
+    int32_t tmp = W0 * (d0 + d1);	\
     t0 = tmp + (W1 - W0) * d1;		\
     t1 = tmp - (W1 + W0) * d0;		\
 } while (0)
@@ -62,9 +62,9 @@ do {					\
 static inline void idct_row (int16_t * const block)
 {
     uint64_t l, r;
-    int_fast32_t d0, d1, d2, d3;
-    int_fast32_t a0, a1, a2, a3, b0, b1, b2, b3;
-    int_fast32_t t0, t1, t2, t3;
+    int32_t d0, d1, d2, d3;
+    int32_t a0, a1, a2, a3, b0, b1, b2, b3;
+    int32_t t0, t1, t2, t3;
 
     l = ldq (block);
     r = ldq (block + 4);
@@ -118,9 +118,9 @@ static inline void idct_row (int16_t * const block)
 
 static inline void idct_col (int16_t * const block)
 {
-    int_fast32_t d0, d1, d2, d3;
-    int_fast32_t a0, a1, a2, a3, b0, b1, b2, b3;
-    int_fast32_t t0, t1, t2, t3;
+    int32_t d0, d1, d2, d3;
+    int32_t a0, a1, a2, a3, b0, b1, b2, b3;
+    int32_t t0, t1, t2, t3;
 
     d0 = (block[8*0] << 11) + 65536;
     d1 = block[8*1];
-- 
1.5.6.5

[mpeg2-dev] How to debug libmpeg2?

[zhihang w. <zhi...@gm...>]

Hi, I use ./configure --enable-debug , make and gdb mpeg2dec
But I can't step into the libmpeg2.
How to debug libmpeg2?


Best Regards
zhihang wang

[mpeg2-dev] [PATCH] [RFC] workaround for crasher bug

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

I can reproduce a crash in libmpeg2 with MPlayer on the sample:

http://samples.mplayerhq.hu/MPEG2/libmpeg2-crash.vob

The crash happens both with the internal forked copy and with the
upstream version.

Here is a somewhat ugly patch that we used to apply to the forked
copy and that does fix the crash.  If you know a better solution,
we are all ears.

With the patch applies the sample gives ugly artifacting but does
not crash anymore.  To be precise when used with upstream libmpeg2
there is heavy artifacting, but not with MPlayer's forked copy.

Diego

Re: [mpeg2-dev] [PATCH] make headers compile standalone

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

On Thu, Jun 24, 2010, Diego Biurrun wrote:

> > Here is a version of the patch without the stdint.h #includes.
> > Please apply.

   Looks perfect. Thanks :-)

-- 
Sam.

Re: [mpeg2-dev] [PATCH] make headers compile standalone

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

On Thu, Jun 24, 2010 at 10:41:22AM +0200, Diego Biurrun wrote:
> On Sat, Jun 12, 2010 at 10:37:48AM +0200, Diego Biurrun wrote:
> > On Fri, Jun 11, 2010 at 11:56:09AM +0200, Sam Hocevar wrote:
> > > On Thu, Jun 10, 2010, Diego Biurrun wrote:
> > > > Here is a patch to make libmpeg2 headers compile standalone.
> > > > This allows #including them directly without worrying about
> > > > header inclusion order or having to #include system headers
> > > > before libmpeg2 headers.
> > > 
> > >    I am worried about the <stdint.h> inclusions. This will likely break
> > > compilation on Visual Studio until they have that header.
> > 
> > How backwards can they be?  Do they have inttypes.h?
> > 
> > In any case the rest of the patch should be worth applying nonetheless.
> 
> Here is a version of the patch without the stdint.h #includes.
> Please apply.

*Here* is a version of said patch - *sigh*..

Diego

Re: [mpeg2-dev] [PATCH] make headers compile standalone

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

On Sat, Jun 12, 2010 at 10:37:48AM +0200, Diego Biurrun wrote:
> On Fri, Jun 11, 2010 at 11:56:09AM +0200, Sam Hocevar wrote:
> > On Thu, Jun 10, 2010, Diego Biurrun wrote:
> > > Here is a patch to make libmpeg2 headers compile standalone.
> > > This allows #including them directly without worrying about
> > > header inclusion order or having to #include system headers
> > > before libmpeg2 headers.
> > 
> >    I am worried about the <stdint.h> inclusions. This will likely break
> > compilation on Visual Studio until they have that header.
> 
> How backwards can they be?  Do they have inttypes.h?
> 
> In any case the rest of the patch should be worth applying nonetheless.

Here is a version of the patch without the stdint.h #includes.
Please apply.

Diego

Re: [mpeg2-dev] Error Compiling mpeg2dec using sparc-elf-gcc

[R. Denis-C. <re...@re...>]

On Thursday 17 June 2010 06:29:34 ardimas andi, you wrote:
> can you guys help me? I'm so freakin' crazy to find this out.
> thx so much

You left out the object containing the SPARC VIS optimizations from the 
linking stage. As a consequence, linking fails.

-- 
Rémi Denis-Courmont

[mpeg2-dev] Error Compiling mpeg2dec using sparc-elf-gcc

[ardimas a. <ard...@ya...>]

hi guys, let me introduce my self because I am a new member. My name is Ardimas. Electrical Engineering Undergraduate Student in ITB, Indonesia.

I have a problem when I tried to compile mpeg2dec package using sparc (cross compiler of uP leon). 

sparc-elf-gcc mpeg2dec.c gettimeofday.c video_out.c video_out_null.c video_out_pgm.c  decode.c dump_state.c alloc.c header.c cpu_state.c cpu_accel.c slice.c idct.c motion_comp.c rgb.c rgb_vis.c -o mpeg2dec.exe

and the result is :
/cygdrive/c/DOCUME~1/KP/LOCALS~1/Temp/ccbhMdJk.o: In function `sigill_handler': cpu_accel.c:(.text+0x70): undefined reference to `sigprocmask' /cygdrive/c/DOCUME~1/KP/LOCALS~1/Temp/ccbhMdJk.o: In function `arch_accel': cpu_accel.c:(.text+0xd8): undefined reference to `sigprocmask' cpu_accel.c:(.text+0x164): undefined reference to `sigprocmask' /cygdrive/c/DOCUME~1/KP/LOCALS~1/Temp/ccMgiDgd.o: In function `mpeg2_mc_init': motion_comp.c:(.text+0x24): undefined reference to `mpeg2_mc_vis' motion_comp.c:(.text+0x28): undefined reference to `mpeg2_mc_vis'

can you guys help me? I'm so freakin' crazy to find this out.
thx so much



      

Re: [mpeg2-dev] [PATCH] make headers compile standalone

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

On Fri, Jun 11, 2010 at 11:56:09AM +0200, Sam Hocevar wrote:
> On Thu, Jun 10, 2010, Diego Biurrun wrote:
> > Here is a patch to make libmpeg2 headers compile standalone.
> > This allows #including them directly without worrying about
> > header inclusion order or having to #include system headers
> > before libmpeg2 headers.
> 
>    I am worried about the <stdint.h> inclusions. This will likely break
> compilation on Visual Studio until they have that header.

How backwards can they be?  Do they have inttypes.h?

In any case the rest of the patch should be worth applying nonetheless.

Diego

Re: [mpeg2-dev] [PATCH] make headers compile standalone

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

On Thu, Jun 10, 2010, Diego Biurrun wrote:
> Here is a patch to make libmpeg2 headers compile standalone.
> This allows #including them directly without worrying about
> header inclusion order or having to #include system headers
> before libmpeg2 headers.

   I am worried about the <stdint.h> inclusions. This will likely break
compilation on Visual Studio until they have that header.

-- 
Sam.

[mpeg2-dev] [PATCH] make headers compile standalone

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

Here is a patch to make libmpeg2 headers compile standalone.
This allows #including them directly without worrying about
header inclusion order or having to #include system headers
before libmpeg2 headers.

Diego

Re: [mpeg2-dev] Exception in mpeg2_idct_copy_sse2

[Matthew M. <mmc...@gm...>]

hmm. Not quite there.

The OSSBuild GStreamer package contains:

libmpeg2-0.dll
libmpeg2convert-0.dll
mpeg2.lib
mpeg2convert.lib

but unfortunately mpeg2convert.lib and mpeg2.lib are identical files. As a
result, I'm unable to link to libmpeg2convert.

Matt



On 5 May 2010 13:49, Matthew McGinity <mmc...@gm...> wrote:

> Hi David,
>
> Thanks. Those DLL's work and do indeed have SSE2 acceleration.
> (SSE2 produces approx 30% speed increase on my machine).
>
> But it is frustrating that I can't get my own build working in Visual
> Studio, as I would prefer to use a static library rather than a DLL.
>
> Thanks,
>
> Matt
>
>
>
> On 5 May 2010 02:46, Hoyt, David <ho...@ll...> wrote:
>
>>  You can get a copy built by the OSSBuild folks:
>>
>>
>>
>> http://code.google.com/p/ossbuild/
>>
>>
>>
>> Download and install the GPL version and then the GPL SDK and you’ll have
>> the headers and libs you need. I’m unsure if it’s built w/ SSE2 or not.
>>
>> My work’s filter likes to put asterisks (*) in URLs to prevent people from
>> clicking on just any link they find in an email, so if you see one, take it
>> out.
>>
>>
>>
>> *From:* Matthew McGinity [mailto:mmc...@gm...]
>> *Sent:* Tuesday, May 04, 2010 1:22 AM
>> *To:* Keith Winstein
>> *Cc:* lib...@li...
>> *Subject:* Re: [mpeg2-dev] Exception in mpeg2_idct_copy_sse2
>>
>>
>>
>> Hi,
>>
>> Can anyone confirm that they have SSE2 acceleration from a visual studio
>> 2005 build of libmpeg2?
>> How were those .obj files generated?
>>
>> Or perhaps someone could provide me with a release build of libmpeg2.lib
>> 0.5.1 ??
>>
>> Any ideas whatsoever greatly appreciated!
>>
>> Thanks
>>
>> Matt
>>
>> --
>> Matthew McGinity
>> iCinema Centre for Interactive Cinema Research
>> University of New South Wales, Sydney, Australia
>> http://*www.*icinema.unsw.edu.au
>> Tel: 0405 115 322
>>
>>  On 2 May 2010 17:46, Matthew McGinity <mmc...@gm...> wrote:
>>
>> Hi Keith,
>>
>> Just resending last mail without attachment so it's not bounced by the
>> mailing-list.
>>
>>
>> I've compiled both sample1 and mpeg2dec and they both suffer the same
>> problem.
>> I've compiled version 0.4.1 and that works fine - which is to be
>> expected as the SSE2 support was introduced in version .5
>>
>> > I didn't understand your remark about "I assume the source files for the
>> > .obj files in vc++ cannot be built with VC++". Are you saying you're not
>> > compiling libmpeg2 from the source code and just using somebody else's
>> > build?
>>
>> I'm building my own version. My project has these source files:
>>
>> libmpeg2\alloc.c
>> libmpeg2\decode.c
>> libmpeg2\header.c
>> libmpeg2\idct.c
>> libmpeg2\motion_comp.c
>> libmpeg2\motion_comp_vis.c
>> libmpeg2\slice.c
>> libmpeg2\convert\rgb.c
>> libmpeg2\convert\rgb_vis.c
>> libmpeg2\convert\uyvy.c
>>
>> and these object files:
>> vc++\cpu_accel.obj
>> vc++\cpu_state.obj
>> vc++\idct_mmx.obj
>> vc++\motion_comp_mmx.obj
>> vc++\rgb_mmx.obj
>>
>> I tried compiling these .obj files from their corresponding source .c
>> files but without success -  I don't think they're intended to be
>> built with vc++.
>>
>> The only changes I've made is removing #define restrict __restrict from
>> config.h
>>
>> I can easily email out my complete solution should anyone with windows
>> and Visual Studio 2005 be willing to take a look.
>> (I don't think I can send attachments to this forum)
>>
>> Cheers,
>> Matt
>>
>>
>> > On 2 May 2010 02:43, Keith Winstein <ke...@mi...> wrote:
>> >> Hi Matthew,
>> >>
>> >> We're here. Unfortunately I don't have Windows and would have a hard
>> time
>> >> trying to reproduce this problem for you. (It works fine for me with
>> gcc on
>> >> Linux; I recognize that's not very helpful to know.)
>> >>
>> >> Are you able to compile the "mpeg2dec" program that ships with
>> libmpeg2?
>> >> Does it also bomb out like this?
>> >>
>> >> I didn't understand your remark about "I assume the source files for
>> the
>> >> .obj files in vc++ cannot be built with VC++". Are you saying you're
>> not
>> >> compiling libmpeg2 from the source code and just using somebody else's
>> >> build? Sorry for the confusion.
>> >>
>> >> Regards,
>> >> Keith
>> >>
>> >> On Sat, 1 May 2010, Matthew McGinity wrote:
>> >>
>> >>> Hi,
>> >>>
>> >>> First question - is there anyone out there? Is this forum still alive?
>> >>>
>> >>> I have libmpeg2 version 5.0.1 compiling in VS2005.
>> >>> My test file, based on sample5.c, works perfectly, provided I don't
>> enable
>> >>> SSE2.
>> >>>
>> >>> That is, if I change mpeg2_init() to use
>> >>>  mpeg2_accel(0);
>> >>> or
>> >>>  mpeg2_accel(MPEG2_ACCEL_X86_MMX);
>> >>> or
>> >>>  mpeg2_accel(MPEG2_ACCEL_X86_MMXEXT);
>> >>> everything works fine.
>> >>>
>> >>> However, if I allow auto-detection
>> >>>  mpeg2_accel(MPEG2_ACCEL_DETECT);
>> >>> it chooses sse2 acceleration and I get an exception on the very first
>> >>> call to mpeg2_idct_copy in slice_intra_DCT in mpeg2_slice:
>> >>>
>> >>> "First-chance exception at 0x00445fa3 in icMPEG2Test.exe: 0xC0000005:
>> >>> Access violation reading location 0xffffffff."
>> >>>
>> >>> Stack location is:  _mpeg2_idct_copy_sse2()  + 0x13 bytes
>> >>>
>> >>> In my project, I'm linking to all the .obj files provided in directory
>> >>> \libmpeg2\vc++
>> >>>
>> >>> Not quite sure how to proceed. I assume the source files for the .obj
>> >>> files in vc++
>> >>> cannot be built with VC++. (I tried renaming __asm__ to __asm and
>> >>> __volatile__ to volatile, but it didn't get me far).
>> >>>
>> >>> Any advice appreciated,
>> >>>
>> >>> Matt
>> >>> - libmpeg2 v5.0.1, Windows XP, Intel Core2 Quad CPU Q9400, Visual
>> Studio
>> >>> 2005
>> >>>
>> >>>
>> >>>
>> >>> --
>> >>> Matthew McGinity
>> >>> iCinema Centre for Interactive Cinema Research
>> >>> University of New South Wales, Sydney, Australia
>> >>> http://*www.*icinema.unsw.edu.au
>> >>> Tel: 0405 115 322
>> >>>
>> >>>
>> >>>
>> ------------------------------------------------------------------------------
>> >>> _______________________________________________
>> >>> Libmpeg2-devel mailing list
>> >>> Lib...@li...
>> >>> https://*lists.sourceforge.net/lists/listinfo/libmpeg2-devel
>> >>>
>> >>
>> >
>>
>>
>>
>
>

Re: [mpeg2-dev] Exception in mpeg2_idct_copy_sse2

[Matthew M. <mmc...@gm...>]

Hi David,

Thanks. Those DLL's work and do indeed have SSE2 acceleration.
(SSE2 produces approx 30% speed increase on my machine).

But it is frustrating that I can't get my own build working in Visual
Studio, as I would prefer to use a static library rather than a DLL.

Thanks,

Matt


On 5 May 2010 02:46, Hoyt, David <ho...@ll...> wrote:

>  You can get a copy built by the OSSBuild folks:
>
>
>
> http://code.google.com/p/ossbuild/
>
>
>
> Download and install the GPL version and then the GPL SDK and you’ll have
> the headers and libs you need. I’m unsure if it’s built w/ SSE2 or not.
>
> My work’s filter likes to put asterisks (*) in URLs to prevent people from
> clicking on just any link they find in an email, so if you see one, take it
> out.
>
>
>
> *From:* Matthew McGinity [mailto:mmc...@gm...]
> *Sent:* Tuesday, May 04, 2010 1:22 AM
> *To:* Keith Winstein
> *Cc:* lib...@li...
> *Subject:* Re: [mpeg2-dev] Exception in mpeg2_idct_copy_sse2
>
>
>
> Hi,
>
> Can anyone confirm that they have SSE2 acceleration from a visual studio
> 2005 build of libmpeg2?
> How were those .obj files generated?
>
> Or perhaps someone could provide me with a release build of libmpeg2.lib
> 0.5.1 ??
>
> Any ideas whatsoever greatly appreciated!
>
> Thanks
>
> Matt
>
> --
> Matthew McGinity
> iCinema Centre for Interactive Cinema Research
> University of New South Wales, Sydney, Australia
> http://*www.*icinema.unsw.edu.au
> Tel: 0405 115 322
>
>  On 2 May 2010 17:46, Matthew McGinity <mmc...@gm...> wrote:
>
> Hi Keith,
>
> Just resending last mail without attachment so it's not bounced by the
> mailing-list.
>
>
> I've compiled both sample1 and mpeg2dec and they both suffer the same
> problem.
> I've compiled version 0.4.1 and that works fine - which is to be
> expected as the SSE2 support was introduced in version .5
>
> > I didn't understand your remark about "I assume the source files for the
> > .obj files in vc++ cannot be built with VC++". Are you saying you're not
> > compiling libmpeg2 from the source code and just using somebody else's
> > build?
>
> I'm building my own version. My project has these source files:
>
> libmpeg2\alloc.c
> libmpeg2\decode.c
> libmpeg2\header.c
> libmpeg2\idct.c
> libmpeg2\motion_comp.c
> libmpeg2\motion_comp_vis.c
> libmpeg2\slice.c
> libmpeg2\convert\rgb.c
> libmpeg2\convert\rgb_vis.c
> libmpeg2\convert\uyvy.c
>
> and these object files:
> vc++\cpu_accel.obj
> vc++\cpu_state.obj
> vc++\idct_mmx.obj
> vc++\motion_comp_mmx.obj
> vc++\rgb_mmx.obj
>
> I tried compiling these .obj files from their corresponding source .c
> files but without success -  I don't think they're intended to be
> built with vc++.
>
> The only changes I've made is removing #define restrict __restrict from
> config.h
>
> I can easily email out my complete solution should anyone with windows
> and Visual Studio 2005 be willing to take a look.
> (I don't think I can send attachments to this forum)
>
> Cheers,
> Matt
>
>
> > On 2 May 2010 02:43, Keith Winstein <ke...@mi...> wrote:
> >> Hi Matthew,
> >>
> >> We're here. Unfortunately I don't have Windows and would have a hard
> time
> >> trying to reproduce this problem for you. (It works fine for me with gcc
> on
> >> Linux; I recognize that's not very helpful to know.)
> >>
> >> Are you able to compile the "mpeg2dec" program that ships with libmpeg2?
> >> Does it also bomb out like this?
> >>
> >> I didn't understand your remark about "I assume the source files for the
> >> .obj files in vc++ cannot be built with VC++". Are you saying you're not
> >> compiling libmpeg2 from the source code and just using somebody else's
> >> build? Sorry for the confusion.
> >>
> >> Regards,
> >> Keith
> >>
> >> On Sat, 1 May 2010, Matthew McGinity wrote:
> >>
> >>> Hi,
> >>>
> >>> First question - is there anyone out there? Is this forum still alive?
> >>>
> >>> I have libmpeg2 version 5.0.1 compiling in VS2005.
> >>> My test file, based on sample5.c, works perfectly, provided I don't
> enable
> >>> SSE2.
> >>>
> >>> That is, if I change mpeg2_init() to use
> >>>  mpeg2_accel(0);
> >>> or
> >>>  mpeg2_accel(MPEG2_ACCEL_X86_MMX);
> >>> or
> >>>  mpeg2_accel(MPEG2_ACCEL_X86_MMXEXT);
> >>> everything works fine.
> >>>
> >>> However, if I allow auto-detection
> >>>  mpeg2_accel(MPEG2_ACCEL_DETECT);
> >>> it chooses sse2 acceleration and I get an exception on the very first
> >>> call to mpeg2_idct_copy in slice_intra_DCT in mpeg2_slice:
> >>>
> >>> "First-chance exception at 0x00445fa3 in icMPEG2Test.exe: 0xC0000005:
> >>> Access violation reading location 0xffffffff."
> >>>
> >>> Stack location is:  _mpeg2_idct_copy_sse2()  + 0x13 bytes
> >>>
> >>> In my project, I'm linking to all the .obj files provided in directory
> >>> \libmpeg2\vc++
> >>>
> >>> Not quite sure how to proceed. I assume the source files for the .obj
> >>> files in vc++
> >>> cannot be built with VC++. (I tried renaming __asm__ to __asm and
> >>> __volatile__ to volatile, but it didn't get me far).
> >>>
> >>> Any advice appreciated,
> >>>
> >>> Matt
> >>> - libmpeg2 v5.0.1, Windows XP, Intel Core2 Quad CPU Q9400, Visual
> Studio
> >>> 2005
> >>>
> >>>
> >>>
> >>> --
> >>> Matthew McGinity
> >>> iCinema Centre for Interactive Cinema Research
> >>> University of New South Wales, Sydney, Australia
> >>> http://*www.*icinema.unsw.edu.au
> >>> Tel: 0405 115 322
> >>>
> >>>
> >>>
> ------------------------------------------------------------------------------
> >>> _______________________________________________
> >>> Libmpeg2-devel mailing list
> >>> Lib...@li...
> >>> https://*lists.sourceforge.net/lists/listinfo/libmpeg2-devel
> >>>
> >>
> >
>
>
>

Re: [mpeg2-dev] Exception in mpeg2_idct_copy_sse2

[Hoyt, D. <ho...@ll...>]

You can get a copy built by the OSSBuild folks:

http://code.google.com/p/ossbuild/

Download and install the GPL version and then the GPL SDK and you'll have the headers and libs you need. I'm unsure if it's built w/ SSE2 or not.

My work's filter likes to put asterisks (*) in URLs to prevent people from clicking on just any link they find in an email, so if you see one, take it out.

From: Matthew McGinity [mailto:mmc...@gm...]
Sent: Tuesday, May 04, 2010 1:22 AM
To: Keith Winstein
Cc: lib...@li...
Subject: Re: [mpeg2-dev] Exception in mpeg2_idct_copy_sse2

Hi,

Can anyone confirm that they have SSE2 acceleration from a visual studio 2005 build of libmpeg2?
How were those .obj files generated?

Or perhaps someone could provide me with a release build of libmpeg2.lib 0.5.1 ??

Any ideas whatsoever greatly appreciated!

Thanks

Matt

--
Matthew McGinity
iCinema Centre for Interactive Cinema Research
University of New South Wales, Sydney, Australia
http://*www.*icinema.unsw.edu.au
Tel: 0405 115 322

On 2 May 2010 17:46, Matthew McGinity <mmc...@gm...<mailto:mmc...@gm...>> wrote:
Hi Keith,

Just resending last mail without attachment so it's not bounced by the
mailing-list.

I've compiled both sample1 and mpeg2dec and they both suffer the same problem.
I've compiled version 0.4.1 and that works fine - which is to be
expected as the SSE2 support was introduced in version .5

> I didn't understand your remark about "I assume the source files for the
> .obj files in vc++ cannot be built with VC++". Are you saying you're not
> compiling libmpeg2 from the source code and just using somebody else's
> build?

I'm building my own version. My project has these source files:

libmpeg2\alloc.c
libmpeg2\decode.c
libmpeg2\header.c
libmpeg2\idct.c
libmpeg2\motion_comp.c
libmpeg2\motion_comp_vis.c
libmpeg2\slice.c
libmpeg2\convert\rgb.c
libmpeg2\convert\rgb_vis.c
libmpeg2\convert\uyvy.c

and these object files:
vc++\cpu_accel.obj
vc++\cpu_state.obj
vc++\idct_mmx.obj
vc++\motion_comp_mmx.obj
vc++\rgb_mmx.obj

I tried compiling these .obj files from their corresponding source .c
files but without success -  I don't think they're intended to be
built with vc++.

The only changes I've made is removing #define restrict __restrict from config.h
I can easily email out my complete solution should anyone with windows
and Visual Studio 2005 be willing to take a look.
(I don't think I can send attachments to this forum)

Cheers,
Matt

> On 2 May 2010 02:43, Keith Winstein <ke...@mi...<mailto:ke...@mi...>> wrote:
>> Hi Matthew,
>>
>> We're here. Unfortunately I don't have Windows and would have a hard time
>> trying to reproduce this problem for you. (It works fine for me with gcc on
>> Linux; I recognize that's not very helpful to know.)
>>
>> Are you able to compile the "mpeg2dec" program that ships with libmpeg2?
>> Does it also bomb out like this?
>>
>> I didn't understand your remark about "I assume the source files for the
>> .obj files in vc++ cannot be built with VC++". Are you saying you're not
>> compiling libmpeg2 from the source code and just using somebody else's
>> build? Sorry for the confusion.
>>
>> Regards,
>> Keith
>>
>> On Sat, 1 May 2010, Matthew McGinity wrote:
>>
>>> Hi,
>>>
>>> First question - is there anyone out there? Is this forum still alive?
>>>
>>> I have libmpeg2 version 5.0.1 compiling in VS2005.
>>> My test file, based on sample5.c, works perfectly, provided I don't enable
>>> SSE2.
>>>
>>> That is, if I change mpeg2_init() to use
>>>  mpeg2_accel(0);
>>> or
>>>  mpeg2_accel(MPEG2_ACCEL_X86_MMX);
>>> or
>>>  mpeg2_accel(MPEG2_ACCEL_X86_MMXEXT);
>>> everything works fine.
>>>
>>> However, if I allow auto-detection
>>>  mpeg2_accel(MPEG2_ACCEL_DETECT);
>>> it chooses sse2 acceleration and I get an exception on the very first
>>> call to mpeg2_idct_copy in slice_intra_DCT in mpeg2_slice:
>>>
>>> "First-chance exception at 0x00445fa3 in icMPEG2Test.exe: 0xC0000005:
>>> Access violation reading location 0xffffffff."
>>>
>>> Stack location is:  _mpeg2_idct_copy_sse2()  + 0x13 bytes
>>>
>>> In my project, I'm linking to all the .obj files provided in directory
>>> \libmpeg2\vc++
>>>
>>> Not quite sure how to proceed. I assume the source files for the .obj
>>> files in vc++
>>> cannot be built with VC++. (I tried renaming __asm__ to __asm and
>>> __volatile__ to volatile, but it didn't get me far).
>>>
>>> Any advice appreciated,
>>>
>>> Matt
>>> - libmpeg2 v5.0.1, Windows XP, Intel Core2 Quad CPU Q9400, Visual Studio
>>> 2005
>>>
>>>
>>>
>>> --
>>> Matthew McGinity
>>> iCinema Centre for Interactive Cinema Research
>>> University of New South Wales, Sydney, Australia
>>> http://*www.*icinema.unsw.edu.au
>>> Tel: 0405 115 322
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> _______________________________________________
>>> Libmpeg2-devel mailing list
>>> Lib...@li...<mailto:Lib...@li...>
>>> https://*lists.sourceforge.net/lists/listinfo/libmpeg2-devel
>>>
>>
>

Re: [mpeg2-dev] Exception in mpeg2_idct_copy_sse2

[Matthew M. <mmc...@gm...>]

Hi,

Can anyone confirm that they have SSE2 acceleration from a visual studio
2005 build of libmpeg2?
How were those .obj files generated?

Or perhaps someone could provide me with a release build of libmpeg2.lib
0.5.1 ??

Any ideas whatsoever greatly appreciated!

Thanks

Matt

--
Matthew McGinity
iCinema Centre for Interactive Cinema Research
University of New South Wales, Sydney, Australia
http://www.icinema.unsw.edu.au
Tel: 0405 115 322


On 2 May 2010 17:46, Matthew McGinity <mmc...@gm...> wrote:

> Hi Keith,
>
> Just resending last mail without attachment so it's not bounced by the
> mailing-list.
>
> I've compiled both sample1 and mpeg2dec and they both suffer the same
> problem.
> I've compiled version 0.4.1 and that works fine - which is to be
> expected as the SSE2 support was introduced in version .5
>
> > I didn't understand your remark about "I assume the source files for the
> > .obj files in vc++ cannot be built with VC++". Are you saying you're not
> > compiling libmpeg2 from the source code and just using somebody else's
> > build?
>
> I'm building my own version. My project has these source files:
>
> libmpeg2\alloc.c
> libmpeg2\decode.c
> libmpeg2\header.c
> libmpeg2\idct.c
> libmpeg2\motion_comp.c
> libmpeg2\motion_comp_vis.c
> libmpeg2\slice.c
> libmpeg2\convert\rgb.c
> libmpeg2\convert\rgb_vis.c
> libmpeg2\convert\uyvy.c
>
> and these object files:
> vc++\cpu_accel.obj
> vc++\cpu_state.obj
> vc++\idct_mmx.obj
> vc++\motion_comp_mmx.obj
> vc++\rgb_mmx.obj
>
> I tried compiling these .obj files from their corresponding source .c
> files but without success -  I don't think they're intended to be
> built with vc++.
>
> The only changes I've made is removing #define restrict __restrict from
> config.h
>
> I can easily email out my complete solution should anyone with windows
> and Visual Studio 2005 be willing to take a look.
> (I don't think I can send attachments to this forum)
>
> Cheers,
> Matt
>
> > On 2 May 2010 02:43, Keith Winstein <ke...@mi...> wrote:
> >> Hi Matthew,
> >>
> >> We're here. Unfortunately I don't have Windows and would have a hard
> time
> >> trying to reproduce this problem for you. (It works fine for me with gcc
> on
> >> Linux; I recognize that's not very helpful to know.)
> >>
> >> Are you able to compile the "mpeg2dec" program that ships with libmpeg2?
> >> Does it also bomb out like this?
> >>
> >> I didn't understand your remark about "I assume the source files for the
> >> .obj files in vc++ cannot be built with VC++". Are you saying you're not
> >> compiling libmpeg2 from the source code and just using somebody else's
> >> build? Sorry for the confusion.
> >>
> >> Regards,
> >> Keith
> >>
> >> On Sat, 1 May 2010, Matthew McGinity wrote:
> >>
> >>> Hi,
> >>>
> >>> First question - is there anyone out there? Is this forum still alive?
> >>>
> >>> I have libmpeg2 version 5.0.1 compiling in VS2005.
> >>> My test file, based on sample5.c, works perfectly, provided I don't
> enable
> >>> SSE2.
> >>>
> >>> That is, if I change mpeg2_init() to use
> >>>  mpeg2_accel(0);
> >>> or
> >>>  mpeg2_accel(MPEG2_ACCEL_X86_MMX);
> >>> or
> >>>  mpeg2_accel(MPEG2_ACCEL_X86_MMXEXT);
> >>> everything works fine.
> >>>
> >>> However, if I allow auto-detection
> >>>  mpeg2_accel(MPEG2_ACCEL_DETECT);
> >>> it chooses sse2 acceleration and I get an exception on the very first
> >>> call to mpeg2_idct_copy in slice_intra_DCT in mpeg2_slice:
> >>>
> >>> "First-chance exception at 0x00445fa3 in icMPEG2Test.exe: 0xC0000005:
> >>> Access violation reading location 0xffffffff."
> >>>
> >>> Stack location is:  _mpeg2_idct_copy_sse2()  + 0x13 bytes
> >>>
> >>> In my project, I'm linking to all the .obj files provided in directory
> >>> \libmpeg2\vc++
> >>>
> >>> Not quite sure how to proceed. I assume the source files for the .obj
> >>> files in vc++
> >>> cannot be built with VC++. (I tried renaming __asm__ to __asm and
> >>> __volatile__ to volatile, but it didn't get me far).
> >>>
> >>> Any advice appreciated,
> >>>
> >>> Matt
> >>> - libmpeg2 v5.0.1, Windows XP, Intel Core2 Quad CPU Q9400, Visual
> Studio
> >>> 2005
> >>>
> >>>
> >>>
> >>> --
> >>> Matthew McGinity
> >>> iCinema Centre for Interactive Cinema Research
> >>> University of New South Wales, Sydney, Australia
> >>> http://www.icinema.unsw.edu.au
> >>> Tel: 0405 115 322
> >>>
> >>>
> >>>
> ------------------------------------------------------------------------------
> >>> _______________________________________________
> >>> Libmpeg2-devel mailing list
> >>> Lib...@li...
> >>> https://lists.sourceforge.net/lists/listinfo/libmpeg2-devel
> >>>
> >>
> >
>

Re: [mpeg2-dev] Exception in mpeg2_idct_copy_sse2

[Matthew M. <mmc...@gm...>]

Hi Keith,

Just resending last mail without attachment so it's not bounced by the
mailing-list.

I've compiled both sample1 and mpeg2dec and they both suffer the same problem.
I've compiled version 0.4.1 and that works fine - which is to be
expected as the SSE2 support was introduced in version .5

> I didn't understand your remark about "I assume the source files for the
> .obj files in vc++ cannot be built with VC++". Are you saying you're not
> compiling libmpeg2 from the source code and just using somebody else's
> build?

I'm building my own version. My project has these source files:

libmpeg2\alloc.c
libmpeg2\decode.c
libmpeg2\header.c
libmpeg2\idct.c
libmpeg2\motion_comp.c
libmpeg2\motion_comp_vis.c
libmpeg2\slice.c
libmpeg2\convert\rgb.c
libmpeg2\convert\rgb_vis.c
libmpeg2\convert\uyvy.c

and these object files:
vc++\cpu_accel.obj
vc++\cpu_state.obj
vc++\idct_mmx.obj
vc++\motion_comp_mmx.obj
vc++\rgb_mmx.obj

I tried compiling these .obj files from their corresponding source .c
files but without success -  I don't think they're intended to be
built with vc++.

The only changes I've made is removing #define restrict __restrict from config.h

I can easily email out my complete solution should anyone with windows
and Visual Studio 2005 be willing to take a look.
(I don't think I can send attachments to this forum)

Cheers,
Matt

> On 2 May 2010 02:43, Keith Winstein <ke...@mi...> wrote:
>> Hi Matthew,
>>
>> We're here. Unfortunately I don't have Windows and would have a hard time
>> trying to reproduce this problem for you. (It works fine for me with gcc on
>> Linux; I recognize that's not very helpful to know.)
>>
>> Are you able to compile the "mpeg2dec" program that ships with libmpeg2?
>> Does it also bomb out like this?
>>
>> I didn't understand your remark about "I assume the source files for the
>> .obj files in vc++ cannot be built with VC++". Are you saying you're not
>> compiling libmpeg2 from the source code and just using somebody else's
>> build? Sorry for the confusion.
>>
>> Regards,
>> Keith
>>
>> On Sat, 1 May 2010, Matthew McGinity wrote:
>>
>>> Hi,
>>>
>>> First question - is there anyone out there? Is this forum still alive?
>>>
>>> I have libmpeg2 version 5.0.1 compiling in VS2005.
>>> My test file, based on sample5.c, works perfectly, provided I don't enable
>>> SSE2.
>>>
>>> That is, if I change mpeg2_init() to use
>>>  mpeg2_accel(0);
>>> or
>>>  mpeg2_accel(MPEG2_ACCEL_X86_MMX);
>>> or
>>>  mpeg2_accel(MPEG2_ACCEL_X86_MMXEXT);
>>> everything works fine.
>>>
>>> However, if I allow auto-detection
>>>  mpeg2_accel(MPEG2_ACCEL_DETECT);
>>> it chooses sse2 acceleration and I get an exception on the very first
>>> call to mpeg2_idct_copy in slice_intra_DCT in mpeg2_slice:
>>>
>>> "First-chance exception at 0x00445fa3 in icMPEG2Test.exe: 0xC0000005:
>>> Access violation reading location 0xffffffff."
>>>
>>> Stack location is:  _mpeg2_idct_copy_sse2()  + 0x13 bytes
>>>
>>> In my project, I'm linking to all the .obj files provided in directory
>>> \libmpeg2\vc++
>>>
>>> Not quite sure how to proceed. I assume the source files for the .obj
>>> files in vc++
>>> cannot be built with VC++. (I tried renaming __asm__ to __asm and
>>> __volatile__ to volatile, but it didn't get me far).
>>>
>>> Any advice appreciated,
>>>
>>> Matt
>>> - libmpeg2 v5.0.1, Windows XP, Intel Core2 Quad CPU Q9400, Visual Studio
>>> 2005
>>>
>>>
>>>
>>> --
>>> Matthew McGinity
>>> iCinema Centre for Interactive Cinema Research
>>> University of New South Wales, Sydney, Australia
>>> http://www.icinema.unsw.edu.au
>>> Tel: 0405 115 322
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> _______________________________________________
>>> Libmpeg2-devel mailing list
>>> Lib...@li...
>>> https://lists.sourceforge.net/lists/listinfo/libmpeg2-devel
>>>
>>
>

Re: [mpeg2-dev] Exception in mpeg2_idct_copy_sse2

[Keith W. <ke...@MI...>]

Hi Matthew,

We're here. Unfortunately I don't have Windows and would have a hard time 
trying to reproduce this problem for you. (It works fine for me with gcc 
on Linux; I recognize that's not very helpful to know.)

Are you able to compile the "mpeg2dec" program that ships with libmpeg2? 
Does it also bomb out like this?

I didn't understand your remark about "I assume the source files for the 
.obj files in vc++ cannot be built with VC++". Are you saying you're not 
compiling libmpeg2 from the source code and just using somebody else's 
build? Sorry for the confusion.

Regards,
Keith

On Sat, 1 May 2010, Matthew McGinity wrote:

> Hi,
>
> First question - is there anyone out there? Is this forum still alive?
>
> I have libmpeg2 version 5.0.1 compiling in VS2005.
> My test file, based on sample5.c, works perfectly, provided I don't enable SSE2.
>
> That is, if I change mpeg2_init() to use
>   mpeg2_accel(0);
> or
>   mpeg2_accel(MPEG2_ACCEL_X86_MMX);
> or
>   mpeg2_accel(MPEG2_ACCEL_X86_MMXEXT);
> everything works fine.
>
> However, if I allow auto-detection
>   mpeg2_accel(MPEG2_ACCEL_DETECT);
> it chooses sse2 acceleration and I get an exception on the very first
> call to mpeg2_idct_copy in slice_intra_DCT in mpeg2_slice:
>
> "First-chance exception at 0x00445fa3 in icMPEG2Test.exe: 0xC0000005:
> Access violation reading location 0xffffffff."
>
> Stack location is:  _mpeg2_idct_copy_sse2()  + 0x13 bytes
>
> In my project, I'm linking to all the .obj files provided in directory
> \libmpeg2\vc++
>
> Not quite sure how to proceed. I assume the source files for the .obj
> files in vc++
> cannot be built with VC++. (I tried renaming __asm__ to __asm and
> __volatile__ to volatile, but it didn't get me far).
>
> Any advice appreciated,
>
> Matt
> - libmpeg2 v5.0.1, Windows XP, Intel Core2 Quad CPU Q9400, Visual Studio 2005
>
>
>
> --
> Matthew McGinity
> iCinema Centre for Interactive Cinema Research
> University of New South Wales, Sydney, Australia
> http://www.icinema.unsw.edu.au
> Tel: 0405 115 322
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Libmpeg2-devel mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libmpeg2-devel
>

[mpeg2-dev] Exception in mpeg2_idct_copy_sse2

[Matthew M. <mmc...@gm...>]

Hi,

First question - is there anyone out there? Is this forum still alive?

I have libmpeg2 version 5.0.1 compiling in VS2005.
My test file, based on sample5.c, works perfectly, provided I don't enable SSE2.

That is, if I change mpeg2_init() to use
   mpeg2_accel(0);
or
   mpeg2_accel(MPEG2_ACCEL_X86_MMX);
or
   mpeg2_accel(MPEG2_ACCEL_X86_MMXEXT);
everything works fine.

However, if I allow auto-detection
   mpeg2_accel(MPEG2_ACCEL_DETECT);
it chooses sse2 acceleration and I get an exception on the very first
call to mpeg2_idct_copy in slice_intra_DCT in mpeg2_slice:

"First-chance exception at 0x00445fa3 in icMPEG2Test.exe: 0xC0000005:
Access violation reading location 0xffffffff."

Stack location is:  _mpeg2_idct_copy_sse2()  + 0x13 bytes	

In my project, I'm linking to all the .obj files provided in directory
\libmpeg2\vc++

Not quite sure how to proceed. I assume the source files for the .obj
files in vc++
cannot be built with VC++. (I tried renaming __asm__ to __asm and
__volatile__ to volatile, but it didn't get me far).

Any advice appreciated,

Matt
- libmpeg2 v5.0.1, Windows XP, Intel Core2 Quad CPU Q9400, Visual Studio 2005



--
Matthew McGinity
iCinema Centre for Interactive Cinema Research
University of New South Wales, Sydney, Australia
http://www.icinema.unsw.edu.au
Tel: 0405 115 322