ffmpeg.git
7 years agoaacdec: fix typo in scalefactor clipping check
Justin Ruggles [Mon, 23 May 2011 19:56:52 +0000 (15:56 -0400)]
aacdec: fix typo in scalefactor clipping check

7 years agofate: fix fate-h264-conformance-frext-pph10i4-panasonic-a crcs.
Ronald S. Bultje [Mon, 23 May 2011 12:44:31 +0000 (08:44 -0400)]
fate: fix fate-h264-conformance-frext-pph10i4-panasonic-a crcs.

The sample on rsync was corrupt, this one is now bitexact w.r.t. JM.

7 years agofate: update 9/10bit refs.
Ronald S. Bultje [Mon, 23 May 2011 12:08:03 +0000 (08:08 -0400)]
fate: update 9/10bit refs.

7 years agoh264: Properly set coded_{width, height} when parsing H.264.
Felipe Contreras [Thu, 19 May 2011 21:39:12 +0000 (00:39 +0300)]
h264: Properly set coded_{width, height} when parsing H.264.

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Diego Biurrun <diego@biurrun.de>
7 years agox86 asm: Add SECTION_TEXT to dct32_sse.asm.
Dave Yeo [Sun, 22 May 2011 17:44:17 +0000 (10:44 -0700)]
x86 asm: Add SECTION_TEXT to dct32_sse.asm.

This fixes the following error on OS/2:
error: segment name `.text align=16' not recognized

Signed-off-by: Diego Biurrun <diego@biurrun.de>
7 years agoFix 9/10 bit in swscale.
Kieran Kunhya [Mon, 23 May 2011 02:02:12 +0000 (03:02 +0100)]
Fix 9/10 bit in swscale.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
7 years agoffmpeg: Don't trigger url_interrupt_cb on the first signal
Martin Storsjö [Sun, 22 May 2011 11:40:33 +0000 (14:40 +0300)]
ffmpeg: Don't trigger url_interrupt_cb on the first signal

Currently, the url_interrupt_cb callback will abort all IO
after the first received signal. This makes the output files
from e.g. the mov muxer to be unreadable if the transcode is
aborted with ctrl+c.

After this patch, the first signal cleanly breaks out of
the transcoding loop, but won't forcibly abort all IO.
After the second signal is received, the url_interrupt_cb
callback will abort all IO.

Signed-off-by: Martin Storsjö <martin@martin.st>
7 years agoavoptions: Check the return value from av_get_number
Martin Storsjö [Sun, 22 May 2011 18:34:49 +0000 (21:34 +0300)]
avoptions: Check the return value from av_get_number

This avoids doing a division by zero if the option wasn't found,
or wasn't an option of an appropriate type.

Signed-off-by: Martin Storsjö <martin@martin.st>
7 years agodct32_sse: eliminate some spills
Loren Merritt [Sat, 21 May 2011 21:36:23 +0000 (23:36 +0200)]
dct32_sse: eliminate some spills

125->104 cycles on penryn (x86_64 only)

7 years agoFix dct32() compilation with --disable-yasm
Vitor Sessak [Sun, 22 May 2011 10:04:33 +0000 (12:04 +0200)]
Fix dct32() compilation with --disable-yasm

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
7 years agoffmpeg: get rid of the -vglobal option.
Anton Khirnov [Sat, 21 May 2011 10:05:49 +0000 (12:05 +0200)]
ffmpeg: get rid of the -vglobal option.

It's badly documented and does the same thing as -flags global_header,
so it's redundant.

7 years agodct32: Add AVX implementation of 32-point DCT
Vitor Sessak [Sat, 14 May 2011 12:17:15 +0000 (14:17 +0200)]
dct32: Add AVX implementation of 32-point DCT

7 years agodct32: Change pass 6 permutation to allow for AVX implementation
Vitor Sessak [Sat, 14 May 2011 12:16:30 +0000 (14:16 +0200)]
dct32: Change pass 6 permutation to allow for AVX implementation

7 years agodct32: port SSE 32-point DCT to YASM
Vitor Sessak [Tue, 17 May 2011 16:26:01 +0000 (18:26 +0200)]
dct32: port SSE 32-point DCT to YASM

7 years agomultiple inclusion guard cleanup
Diego Biurrun [Tue, 17 May 2011 14:58:04 +0000 (16:58 +0200)]
multiple inclusion guard cleanup

Add missing multiple inclusion guards; clean up #endif comments;
add missing library prefixes; keep guard names consistent.

7 years agoavio: document buffer must created with av_malloc() and friends
Can Wu [Sat, 14 May 2011 09:34:28 +0000 (17:34 +0800)]
avio: document buffer must created with av_malloc() and friends

Else a later buffer resize in ffio_set_buf_size() will ABORT.

Signed-off-by: Anton Khirnov <anton@khirnov.net>
7 years agoavio: check AVIOContext malloc failure
Can Wu [Sat, 14 May 2011 09:27:31 +0000 (17:27 +0800)]
avio: check AVIOContext malloc failure

Signed-off-by: Anton Khirnov <anton@khirnov.net>
7 years agoswscale: point out an alternative to sws_getContext
Clément Bœsch [Thu, 5 May 2011 19:25:46 +0000 (21:25 +0200)]
swscale: point out an alternative to sws_getContext

Signed-off-by: Anton Khirnov <anton@khirnov.net>
7 years agosvq3: Do initialization after parsing the extradata
Ronald S. Bultje [Fri, 13 May 2011 07:24:31 +0000 (10:24 +0300)]
svq3: Do initialization after parsing the extradata

If done before, some parameters aren't known yet.

With svq3/rtp, initializing before some parameters are known
can lead to calling av_malloc(0), which on OS X currently returns
broken pointers.

7 years agoadd changelog entries for 0.7_beta2
Reinhard Tartler [Thu, 12 May 2011 16:18:54 +0000 (18:18 +0200)]
add changelog entries for 0.7_beta2

7 years agomp3lame: add #include required for AV_RB32 macro.
Ronald S. Bultje [Fri, 20 May 2011 18:56:44 +0000 (14:56 -0400)]
mp3lame: add #include required for AV_RB32 macro.

Fixes compilation with mp3lame enabled.

7 years agoconfigure: make executable again
Mans Rullgard [Fri, 20 May 2011 17:11:31 +0000 (18:11 +0100)]
configure: make executable again

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoLATM/AAC: Free previously initialized context on reinit.
Ronald S. Bultje [Fri, 20 May 2011 14:49:20 +0000 (10:49 -0400)]
LATM/AAC: Free previously initialized context on reinit.

Fixes memory leaks which are the result of overwriting already-initialized
MDCT contexts during context reinitialization, e.g. in valgrind
fate-aac-latm_000000001180bc60.

Signed-off-by: Diego Biurrun <diego@biurrun.de>
7 years agoconfigure: Do not unconditionally add -Wall to host CFLAGS.
Diego Biurrun [Mon, 16 May 2011 14:16:14 +0000 (16:16 +0200)]
configure: Do not unconditionally add -Wall to host CFLAGS.

Some compilers choke on -Wall, so only add the flag after checking it works.

7 years agoconfigure: Set OS/2 objformat to a.out.
Dave Yeo [Fri, 20 May 2011 15:26:52 +0000 (17:26 +0200)]
configure: Set OS/2 objformat to a.out.

Signed-off-by: Diego Biurrun <diego@biurrun.de>
7 years agoAdd support for a.out object format to assembler macros.
Dave Yeo [Fri, 20 May 2011 15:33:38 +0000 (17:33 +0200)]
Add support for a.out object format to assembler macros.

This format is still used by e.g. OS/2.

Signed-off-by: Diego Biurrun <diego@biurrun.de>
7 years agofate: disable threading for encoding
Mans Rullgard [Fri, 20 May 2011 11:57:12 +0000 (12:57 +0100)]
fate: disable threading for encoding

This explicitly disables threading for encoding as slices are otherwise
automatically activated.  This should be dropped once option resetting
between files is fully implemented.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agofate: add comment field
Mans Rullgard [Fri, 20 May 2011 11:47:42 +0000 (12:47 +0100)]
fate: add comment field

This adds a comment field to the report header, suitable for
extra information not covered by the automatic fields.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agofate: allow overriding default build and install dirs
Mans Rullgard [Fri, 20 May 2011 11:06:51 +0000 (12:06 +0100)]
fate: allow overriding default build and install dirs

This is useful e.g. for building in a different filesystem than
where the source is kept.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agompegtsenc: Add an AVClass pointer to the private data
Martin Storsjö [Fri, 20 May 2011 08:33:57 +0000 (11:33 +0300)]
mpegtsenc: Add an AVClass pointer to the private data

Since a private class is set for this muxer, the callers will
assume that the private data starts with an AVClass pointer.
If no such member exists, the first few bytes of the struct
will be overwritten, and the class pointer may be broken at
any later time.

Signed-off-by: Martin Storsjö <martin@martin.st>
7 years agompegaudio: clean up #includes
Mans Rullgard [Thu, 19 May 2011 12:27:24 +0000 (13:27 +0100)]
mpegaudio: clean up #includes

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agompegaudio: move all header parsing to mpegaudiodecheader.[ch]
Mans Rullgard [Thu, 19 May 2011 12:44:11 +0000 (13:44 +0100)]
mpegaudio: move all header parsing to mpegaudiodecheader.[ch]

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoqdm2: Use floating point synthesis filter.
Vitor Sessak [Thu, 19 May 2011 19:33:27 +0000 (21:33 +0200)]
qdm2: Use floating point synthesis filter.

This avoid needlessly convertion from floating point to fixed point and back.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
7 years agoh264: correct border check.
Ronald S. Bultje [Thu, 19 May 2011 20:47:59 +0000 (16:47 -0400)]
h264: correct border check.

When backing up the top-left border, check that the top-left
(rather than left) MB indeed does belong to our slice. If it
doesn't, backing up has no positive effect but may accidentally
interfere with other threads writing in the same space.

Fixes occasional one-off effects when enabling slice-MT.

7 years agoh264: fix loopfilter with threading at slice boundaries.
Ronald S. Bultje [Wed, 18 May 2011 12:10:49 +0000 (08:10 -0400)]
h264: fix loopfilter with threading at slice boundaries.

7 years agoFix ff_mpa_synth_filter_fixed() prototype
Mans Rullgard [Thu, 19 May 2011 19:46:24 +0000 (20:46 +0100)]
Fix ff_mpa_synth_filter_fixed() prototype

The prototype should use the same typedefs as the definition, or it
will fail where int32_t is not int (DOS apparently).

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoRename costablegen.c ---> cos_tablegen.c.
Diego Biurrun [Sun, 15 May 2011 22:53:07 +0000 (00:53 +0200)]
Rename costablegen.c ---> cos_tablegen.c.

This is consistent with how all other table generation programs are named.
Moreover this ensures that the cos table generation program is correctly
deleted when cleaning the tree.

7 years agoCollapse tableprint.c into tableprint.h.
Diego Biurrun [Thu, 19 May 2011 17:22:41 +0000 (19:22 +0200)]
Collapse tableprint.c into tableprint.h.

tableprint.c serves little purpose on its own and removing it allows building
the table generator programs with the normal HOSTPROGS Makefile rules.

7 years agoSimplify trig table rules
Mans Rullgard [Thu, 19 May 2011 17:34:40 +0000 (18:34 +0100)]
Simplify trig table rules

This collapses the make rules for the trig tables into a pattern
rule.  Based on a patch by Diego, modified to avoid using fragile
make constructs and allow future addition of fixed-point sin tables.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoRemove potentially unstable filenames from comments in generated files.
Diego Biurrun [Thu, 19 May 2011 16:12:17 +0000 (18:12 +0200)]
Remove potentially unstable filenames from comments in generated files.

7 years agoIgnore generated tables and generated table generator programs.
Diego Biurrun [Sun, 15 May 2011 22:59:50 +0000 (00:59 +0200)]
Ignore generated tables and generated table generator programs.

7 years agoSimplify CLEANFILES make variable by using wildcards.
Diego Biurrun [Sun, 15 May 2011 22:54:36 +0000 (00:54 +0200)]
Simplify CLEANFILES make variable by using wildcards.

Also ensures that generated file cos_fixed_tables.c is deleted on 'make clean'.

7 years agoRemove silly insults from avformat_version() Doxygen documentation.
Diego Biurrun [Thu, 19 May 2011 15:39:50 +0000 (17:39 +0200)]
Remove silly insults from avformat_version() Doxygen documentation.

7 years agompegaudiodsp: fix x86 and ppc makefiles
Mans Rullgard [Thu, 19 May 2011 15:26:39 +0000 (16:26 +0100)]
mpegaudiodsp: fix x86 and ppc makefiles

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoconfigure: Adjust AVX assembler check.
Diego Biurrun [Mon, 16 May 2011 12:23:59 +0000 (14:23 +0200)]
configure: Adjust AVX assembler check.

Older nasm versions have trouble assembling certain AVX instructions, but the
current AVX check did not detect this. Update the check to use an instruction
that triggers the nasm problem.

7 years agompegaudio: remove unused version of SAME_HEADER_MASK
Mans Rullgard [Thu, 19 May 2011 13:01:11 +0000 (14:01 +0100)]
mpegaudio: remove unused version of SAME_HEADER_MASK

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agompegaudio: remove useless #undef at end of file
Mans Rullgard [Thu, 19 May 2011 11:48:33 +0000 (12:48 +0100)]
mpegaudio: remove useless #undef at end of file

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoasfdec: add missing #include for av_bswap32()
Mans Rullgard [Thu, 19 May 2011 11:31:05 +0000 (12:31 +0100)]
asfdec: add missing #include for av_bswap32()

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agompegaudio: merge two #if CONFIG_FLOAT blocks
Mans Rullgard [Mon, 16 May 2011 17:59:25 +0000 (18:59 +0100)]
mpegaudio: merge two #if CONFIG_FLOAT blocks

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agompegaudio: move some struct definitions from mpegaudio.h
Mans Rullgard [Tue, 17 May 2011 13:22:25 +0000 (14:22 +0100)]
mpegaudio: move some struct definitions from mpegaudio.h

These structs are only used in mpegaudiodec.c, so move them there
and remove no longer needed #include lines from mpegaudio.h.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoMove some mpegaudio functions to new mpegaudiodsp subsystem
Mans Rullgard [Mon, 16 May 2011 15:52:01 +0000 (16:52 +0100)]
Move some mpegaudio functions to new mpegaudiodsp subsystem

This separation allows these functions to be used in a cleaner
fashion from other codecs (e.g. qdm2) and simplifies creating
optimised versions of them.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoClean up #includes in cmdutils.h.
Diego Biurrun [Sun, 15 May 2011 15:09:25 +0000 (17:09 +0200)]
Clean up #includes in cmdutils.h.

inttypes.h is not necessary, just stdint.h is enough.
Unconditionally #include avfilter.h in cmdutils.h.  It is an installed
header with no non-standard external dependencies, so it is safe.

7 years agog729: Merge g729.h into g729dec.c.
Diego Biurrun [Tue, 17 May 2011 13:29:58 +0000 (15:29 +0200)]
g729: Merge g729.h into g729dec.c.

The header contains just a single define that is only used in g729dec.c.

7 years ago10l: wrap float_interleave functions in HAVE_YASM.
Justin Ruggles [Thu, 19 May 2011 00:18:08 +0000 (20:18 -0400)]
10l: wrap float_interleave functions in HAVE_YASM.

fixes compilation with --disable-yasm

7 years agoAPIchanges: fill in date and commit for request_sample_fmt
Justin Ruggles [Wed, 18 May 2011 21:52:31 +0000 (17:52 -0400)]
APIchanges: fill in date and commit for request_sample_fmt

7 years agoAdd floating-point sample format support to the ac3, eac3, dca, aac, and vorbis
Justin Ruggles [Sat, 23 Apr 2011 01:30:19 +0000 (21:30 -0400)]
Add floating-point sample format support to the ac3, eac3, dca, aac, and vorbis
decoders.

Based on patches by clsid2 in ffdshow-tryout.

7 years agoAdd support for request_sample_format in ffmpeg and ffplay.
Justin Ruggles [Sun, 24 Apr 2011 20:48:42 +0000 (16:48 -0400)]
Add support for request_sample_format in ffmpeg and ffplay.

7 years agoAdd APIchanges entry for request_sample_fmt.
Justin Ruggles [Thu, 21 Apr 2011 14:48:05 +0000 (10:48 -0400)]
Add APIchanges entry for request_sample_fmt.

7 years agoAdd request_sample_fmt field to AVCodecContext.
Justin Ruggles [Tue, 19 Apr 2011 23:47:01 +0000 (19:47 -0400)]
Add request_sample_fmt field to AVCodecContext.

This will allow audio decoders to support output of different sample formats
as a runtime option.

7 years agoAdd float_interleave() to FmtConvertContext with x86-optimized versions.
Justin Ruggles [Sun, 24 Apr 2011 21:50:17 +0000 (17:50 -0400)]
Add float_interleave() to FmtConvertContext with x86-optimized versions.

Partially based on patches by clsid2 in ffdshow-tryout.
ff_float_interleave6() x86 improvements by Loren Merrit.

7 years agoRemove unused make variable SEEK_REFFILE
Mans Rullgard [Wed, 18 May 2011 16:13:15 +0000 (17:13 +0100)]
Remove unused make variable SEEK_REFFILE

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agofate: remove redundant aref and vref references
Mans Rullgard [Wed, 18 May 2011 16:23:47 +0000 (17:23 +0100)]
fate: remove redundant aref and vref references

The ref targets are included in the FATE_[AV]CODEC lists created
by configure so they do not need to be listed separately in the
makefile.  Filter them out when setting dependencies to avoid make
warnings about circular deps.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agofate: remove do_ffmpeg_nocheck function
Mans Rullgard [Wed, 18 May 2011 14:13:58 +0000 (15:13 +0100)]
fate: remove do_ffmpeg_nocheck function

This function is essentially an alias for run_ffmpeg and is only
used in one place.  This patch removes the function and replaces
the call with the equivalent (simpler) run_ffmpeg call.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agofate: do not collect -benchmark output
Mans Rullgard [Wed, 18 May 2011 14:05:26 +0000 (15:05 +0100)]
fate: do not collect -benchmark output

The old regtest scripts pass -benchmark and collect the utime values.
As these values are never used, this machinery can be removed.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agompegaudiodec: remove decode_end() function
Mans Rullgard [Mon, 16 May 2011 15:37:38 +0000 (16:37 +0100)]
mpegaudiodec: remove decode_end() function

This function is not needed since 721d6f2dc5 removed the DCT table
allocations for the configuration used here.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agofate: run aref and vref as regular tests
Mans Rullgard [Wed, 18 May 2011 12:05:39 +0000 (13:05 +0100)]
fate: run aref and vref as regular tests

These tests create reference files used for psnr calculation in
the other codec tests.  Treating them as (mostly) regular tests
simplifies the makefile and makes them visible in the fate reports.
The latter makes errors in these runs easier to identify.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agompegaudio: sanitise compute_antialias_* names
Mans Rullgard [Tue, 17 May 2011 19:39:09 +0000 (20:39 +0100)]
mpegaudio: sanitise compute_antialias_* names

This makes the compute_antialias functions use the same naming
convention as everything else.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agompeg12: add slice-threading checks to slice-threading initializers.
Ronald S. Bultje [Tue, 17 May 2011 14:26:29 +0000 (10:26 -0400)]
mpeg12: add slice-threading checks to slice-threading initializers.

Fixes "make THREADS=2 THREAD_TYPE=1 fate-mpeg2-field-enc".

7 years agoh264: copy pixel_shift between slice threading contexts.
Ronald S. Bultje [Tue, 17 May 2011 14:26:30 +0000 (10:26 -0400)]
h264: copy pixel_shift between slice threading contexts.

Fixes "make THREADS=2 THREAD_TYPE=2
fate-h264-conformance-frext-pph10i3_panasonic_a".

7 years agomdec: enable frame-level multithreading.
Alexander Strange [Tue, 10 May 2011 15:29:09 +0000 (11:29 -0400)]
mdec: enable frame-level multithreading.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
7 years agomdec.c: fix overread.
Ronald S. Bultje [Tue, 10 May 2011 15:29:08 +0000 (11:29 -0400)]
mdec.c: fix overread.

7 years agoFix compilation of iirfilter-test.
Justin Ruggles [Tue, 17 May 2011 22:06:51 +0000 (18:06 -0400)]
Fix compilation of iirfilter-test.

7 years agolibx264: handle closed GOP codec flag
Jindrich Makovicka [Tue, 10 May 2011 13:11:45 +0000 (15:11 +0200)]
libx264: handle closed GOP codec flag

Also update libx264 presets to keep closed gop as default.

Signed-off-by: Jindrich Makovicka <makovick@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
7 years agolavf: remove duplicate assignment in avformat_alloc_context.
Anton Khirnov [Fri, 29 Apr 2011 09:42:05 +0000 (11:42 +0200)]
lavf: remove duplicate assignment in avformat_alloc_context.

AVClass is already initialized in  avformat_get_context_defaults.

7 years agolavf: use designated initializers for AVClasses.
Anton Khirnov [Fri, 29 Apr 2011 09:30:02 +0000 (11:30 +0200)]
lavf: use designated initializers for AVClasses.

7 years agoflvdec: clenup debug code
Alex Converse [Tue, 17 May 2011 19:49:42 +0000 (12:49 -0700)]
flvdec: clenup debug code

7 years agoasfdec: fix possible overread on broken files.
Uoti Urpala [Thu, 12 May 2011 14:20:27 +0000 (10:20 -0400)]
asfdec: fix possible overread on broken files.

7 years agoasfdec: do not fall back to binary/generic search
Vladimir Pantelic [Tue, 17 May 2011 15:30:05 +0000 (17:30 +0200)]
asfdec: do not fall back to binary/generic search

asf_read_seek() inside the asf demuxer already does the
right thing, it tries the index and if that fails it uses
binary search. If binary search is called from outside of asfdec.c
it will fail because the asf code cannot clean up after itself.
Therefore introduce AVFMT_NOBINSEARCH that prevents the seek
code to fallback to binary search and AVFMT_NOGENSEARCH that
prevents the seek code to fallback to generic search.

7 years agoasfdec: reindent after previous commit c7bd5ed
Vladimir Pantelic [Thu, 12 May 2011 08:26:32 +0000 (10:26 +0200)]
asfdec: reindent after previous commit c7bd5ed

7 years agoasfdec: fallback to binary search internally
Vladimir Pantelic [Thu, 12 May 2011 08:25:54 +0000 (10:25 +0200)]
asfdec: fallback to binary search internally

lavf will do that anyway in case seek by index fails

7 years agompegaudio: add _fixed suffix to some names
Mans Rullgard [Tue, 17 May 2011 12:07:08 +0000 (13:07 +0100)]
mpegaudio: add _fixed suffix to some names

This adds a _fixed suffix to the fixed-point versions of things
with both float and fixed-point variants.  This makes it more
consistent with other dual-implementation things, e.g. fft.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoModify x86util.asm to ease transitioning to 10-bit H.264 assembly.
Daniel Kang [Mon, 16 May 2011 21:18:22 +0000 (17:18 -0400)]
Modify x86util.asm to ease transitioning to 10-bit H.264 assembly.

Arguments for variable size instructions are added to many macros, along
with other various changes. The x86util.asm code was ported from x264.

Signed-off-by: Diego Biurrun <diego@biurrun.de>
7 years agodct: build dct32 as separate object files
Mans Rullgard [Tue, 17 May 2011 10:48:28 +0000 (11:48 +0100)]
dct: build dct32 as separate object files

This builds the float and fixed-point versions of dct32 separately
instead of #including the file in dct.c and mpegaudiodec.c.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoqdm2: include correct header for rdft
Mans Rullgard [Tue, 17 May 2011 12:42:43 +0000 (13:42 +0100)]
qdm2: include correct header for rdft

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoffmpeg: fix -aspect cli option
Baptiste Coudurier [Fri, 1 Apr 2011 15:30:45 +0000 (17:30 +0200)]
ffmpeg: fix -aspect cli option

Redesign the way -aspect option is handled. This is done by making
ffmpeg read the sample aspect ratio set in the corresponding input
stream by default, and overriding it using the value specified by
-aspect.

If the output display aspect ratio is specified with -aspect, it is
set at the end of the filterchain, thus overriding the value set by
filters in the filterchain.

This implementation is more robust, since does not modify the
filterchain description (which was creating potential syntax errors).

(Cherry-pick abf8342aa94bdf06bb324f6723a6743dd628d5c6)

Another aspect ratio fix try. This leaves the setdar addition at the end
(preferred by people).

(Cherry-pick e7c7b0d000e81d24327602e04d8fed400dbb7193)

7 years agoRestructure video filter implementation in ffmpeg.c.
Michael Niedermayer [Mon, 14 Feb 2011 22:02:10 +0000 (23:02 +0100)]
Restructure video filter implementation in ffmpeg.c.

This fixes several bugs like multiple outputs and -aspect mixed with -vf

(cherry picked from commit 1762d9ced70ccc46c5d3e5d64e56a48d0fbbd4f7)
(cherry picked from commit 5c20c81bfa526b3a269db9c88b0c9007861f0917)
(cherry picked from commit a7844c580d83d8466c161a0e3979b3902d0d9100)

7 years agoffplay: remove audio_write_get_buf_size() forward declaration
Stefano Sabatini [Tue, 12 Apr 2011 10:06:49 +0000 (12:06 +0200)]
ffplay: remove audio_write_get_buf_size() forward declaration

Move up the definition of audio_write_get_buf_size(), so that it is
defined before it is used. Simplify.
(cherry picked from commit 8776f3d22e401e30d17856e341f6cabbbefa92f7)

7 years agolavfi: print key-frame and picture type information in ff_dlog_ref()
Stefano Sabatini [Sat, 23 Apr 2011 17:55:59 +0000 (19:55 +0200)]
lavfi: print key-frame and picture type information in ff_dlog_ref()

Signed-off-by: Stefano Sabatini <stefano.sabatini-lala@poste.it>
(cherry picked from commit f7bdffb09da597c5d6afff5359523370470ad072)

7 years agomathops: remove ancient confusing comment
Mans Rullgard [Mon, 16 May 2011 20:31:06 +0000 (21:31 +0100)]
mathops: remove ancient confusing comment

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agocws2fws: Improve error message wording.
Diego Biurrun [Sun, 15 May 2011 20:17:35 +0000 (22:17 +0200)]
cws2fws: Improve error message wording.

7 years agotools: Check the return value of write().
Diego Biurrun [Sun, 15 May 2011 16:34:11 +0000 (18:34 +0200)]
tools: Check the return value of write().

This fixes several warnings of the type:
warning: ignoring return value of ‘write’, declared with attribute warn_unused_result

7 years agompegaudio: move OUT_FMT macro to mpegaudiodec.c
Mans Rullgard [Mon, 16 May 2011 16:06:30 +0000 (17:06 +0100)]
mpegaudio: move OUT_FMT macro to mpegaudiodec.c

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agompegaudio: remove OUT_MIN/MAX macros
Mans Rullgard [Mon, 16 May 2011 16:13:23 +0000 (17:13 +0100)]
mpegaudio: remove OUT_MIN/MAX macros

These macros are no longer needed after the s32 output was removed.
Change the relevant code to use av_clip_int16() instead of using
explicit limits.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoAdd missing #includes to mp3_header_(de)compress bsf
Mans Rullgard [Mon, 16 May 2011 14:23:22 +0000 (15:23 +0100)]
Add missing #includes to mp3_header_(de)compress bsf

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agodct: fix indentation
Mans Rullgard [Mon, 16 May 2011 14:57:36 +0000 (15:57 +0100)]
dct: fix indentation

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agodct: bypass table allocation for DCT_II of size 32
Mans Rullgard [Mon, 16 May 2011 14:57:04 +0000 (15:57 +0100)]
dct: bypass table allocation for DCT_II of size 32

The size-32 DCT_II has a special implementation which doesn't use
the normal tables.  Skipping allocation of these in this case saves
some memory.

Signed-off-by: Mans Rullgard <mans@mansr.com>
7 years agoh264dsp_mmx: Add #ifdefs around some mmxext functions on x86_64.
Gil Pedersen [Mon, 16 May 2011 12:40:56 +0000 (14:40 +0200)]
h264dsp_mmx: Add #ifdefs around some mmxext functions on x86_64.

This fixes linking errors due to undefined symbols on x86_64 OS X.

Signed-off-by: Diego Biurrun <diego@biurrun.de>
7 years agoRemove unused header mpegaudio3.h.
Diego Biurrun [Mon, 16 May 2011 01:34:20 +0000 (03:34 +0200)]
Remove unused header mpegaudio3.h.

The header is a part of an MP3 encoder that never saw the light of day.

7 years agoaacdec: Use float instead of int16_t for ltp_state to avoid needless rounding.
Justin Ruggles [Sun, 15 May 2011 15:07:29 +0000 (11:07 -0400)]
aacdec: Use float instead of int16_t for ltp_state to avoid needless rounding.

7 years agoacelp: Remove unused gray_decode table.
Diego Biurrun [Fri, 29 Apr 2011 17:42:00 +0000 (19:42 +0200)]
acelp: Remove unused gray_decode table.