platform/upstream/libav.git
13 years agoRV3/4: calculate B-frame motion weights once per frame
Kostya Shishkov [Wed, 10 Aug 2011 09:26:24 +0000 (11:26 +0200)]
RV3/4: calculate B-frame motion weights once per frame

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agoMove RV3/4-specific DSP functions into their own context
Kostya Shishkov [Tue, 9 Aug 2011 09:00:09 +0000 (11:00 +0200)]
Move RV3/4-specific DSP functions into their own context

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agomjpeg: propagate decode errors from ff_mjpeg_decode_sos and ff_mjpeg_decode_dqt
Dustin Brody [Thu, 11 Aug 2011 15:16:09 +0000 (11:16 -0400)]
mjpeg: propagate decode errors from ff_mjpeg_decode_sos and ff_mjpeg_decode_dqt

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agoh264: notice memory allocation failure
Dustin Brody [Thu, 11 Aug 2011 12:57:58 +0000 (08:57 -0400)]
h264: notice memory allocation failure

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agoac3enc: add macros for option names to make them more understandable.
Justin Ruggles [Tue, 9 Aug 2011 22:11:08 +0000 (18:11 -0400)]
ac3enc: add macros for option names to make them more understandable.

13 years agoac3enc: fix 'channel_coupling' option to actually allow 'auto'.
Justin Ruggles [Wed, 10 Aug 2011 14:21:10 +0000 (10:21 -0400)]
ac3enc: fix 'channel_coupling' option to actually allow 'auto'.

Use 'auto' as the default value.

13 years agocavs: fix some crashes with invalid bitstreams
Mans Rullgard [Wed, 10 Aug 2011 17:52:11 +0000 (18:52 +0100)]
cavs: fix some crashes with invalid bitstreams

This removes all valgrind-reported invalid writes with one
specific test file.

Fixes http://www.ocert.org/advisories/ocert-2011-002.html

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agoac3enc: clip coupling coordinates during calculation
Justin Ruggles [Sun, 7 Aug 2011 21:12:49 +0000 (17:12 -0400)]
ac3enc: clip coupling coordinates during calculation

This is faster than using the clip_coefficients() function. Also,
clipping the coordinates before determining reuse is better since it
will compare coordinates that are closer to their final encoded values.

13 years agoRemove incorrect info in documentation of AVCodecContext.bits_per_raw_sample.
Justin Ruggles [Wed, 10 Aug 2011 18:07:35 +0000 (14:07 -0400)]
Remove incorrect info in documentation of AVCodecContext.bits_per_raw_sample.

bits_per_raw_sample is used in video as well, where sample_fmt is not used.

13 years agolavc: fix parentheses placement in avcodec_open2().
Baptiste Coudurier [Tue, 9 Aug 2011 06:41:50 +0000 (23:41 -0700)]
lavc: fix parentheses placement in avcodec_open2().

Signed-off-by: Anton Khirnov <anton@khirnov.net>
13 years agolavc: make avcodec_init() static on next bump.
Anton Khirnov [Wed, 10 Aug 2011 07:50:00 +0000 (09:50 +0200)]
lavc: make avcodec_init() static on next bump.

It's called from avcodec_register() anyway, so there's no reason for it
to be public.

13 years agoac3enc: remove unneeded #include
Justin Ruggles [Sun, 7 Aug 2011 21:09:47 +0000 (17:09 -0400)]
ac3enc: remove unneeded #include

13 years agoac3enc: restructure coupling coordinate reuse calculation
Justin Ruggles [Fri, 5 Aug 2011 21:34:53 +0000 (17:34 -0400)]
ac3enc: restructure coupling coordinate reuse calculation

13 years agoac3enc: allow new coupling coordinates to be sent independently for each
Justin Ruggles [Fri, 5 Aug 2011 20:28:39 +0000 (16:28 -0400)]
ac3enc: allow new coupling coordinates to be sent independently for each
channel.

13 years agoac3enc: separate exponent bit counting from exponent grouping.
Justin Ruggles [Fri, 5 Aug 2011 20:00:18 +0000 (16:00 -0400)]
ac3enc: separate exponent bit counting from exponent grouping.

Move bit counting to the bit allocation function. Move exponent grouping to
after bit allocation. This will allow for adjustment of bandwidth parameters
during bit allocation without having to do exponent grouping multiple times.

13 years agoh264: propagate error return values for AV_LOG_ERROR-triggering events
Dustin Brody [Thu, 4 Aug 2011 21:47:16 +0000 (17:47 -0400)]
h264: propagate error return values for AV_LOG_ERROR-triggering events

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agoaac: Don't attempt to output configure an invalid channel configuration.
Alex Converse [Fri, 5 Aug 2011 19:40:51 +0000 (12:40 -0700)]
aac: Don't attempt to output configure an invalid channel configuration.

13 years agoRename ffserver to avserver.
Anton Khirnov [Tue, 26 Jul 2011 19:56:56 +0000 (21:56 +0200)]
Rename ffserver to avserver.

13 years agoRename ffprobe to avprobe.
Anton Khirnov [Tue, 26 Jul 2011 17:04:43 +0000 (19:04 +0200)]
Rename ffprobe to avprobe.

13 years agoRename ffplay to avplay.
Anton Khirnov [Tue, 26 Jul 2011 15:51:34 +0000 (17:51 +0200)]
Rename ffplay to avplay.

13 years agoPretty-print RV3/4 decoder source
Kostya Shishkov [Tue, 9 Aug 2011 09:19:55 +0000 (11:19 +0200)]
Pretty-print RV3/4 decoder source

Signed-off-by: Diego Biurrun <diego@biurrun.de>
13 years agoSaner RV1/2 initialisation and RV2 header parsing.
Kostya Shishkov [Thu, 4 Aug 2011 19:00:12 +0000 (21:00 +0200)]
Saner RV1/2 initialisation and RV2 header parsing.

Old version was based on lots of guesswork which had different hacks based on
32-bit version field instead of splitting it into major/minor/micro version.
RV2 picture header parser also had a few places where it deviated from binary
decoder.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agolibx264: do not set pic quality if no frame is output
Baptiste Coudurier [Sun, 30 Jan 2011 01:05:42 +0000 (17:05 -0800)]
libx264: do not set pic quality if no frame is output

Avoids uninitialized reads.

Signed-off-by: Anton Khirnov <anton@khirnov.net>
13 years agolibx264: add preset, tune, profile and fastfirstpass private options
Baptiste Coudurier [Sat, 16 Apr 2011 23:50:50 +0000 (16:50 -0700)]
libx264: add preset, tune, profile and fastfirstpass private options

Signed-off-by: Anton Khirnov <anton@khirnov.net>
13 years agolavc: fix misspelling in comment
Dustin Brody [Sat, 6 Aug 2011 00:38:46 +0000 (20:38 -0400)]
lavc: fix misspelling in comment

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agoaac: propagate error return values for AV_LOG_ERROR-triggering events
Dustin Brody [Fri, 5 Aug 2011 02:32:09 +0000 (22:32 -0400)]
aac: propagate error return values for AV_LOG_ERROR-triggering events

13 years agolsws: remove optimization debug logs in sws_init_context()
Stefano Sabatini [Thu, 4 Aug 2011 15:53:44 +0000 (17:53 +0200)]
lsws: remove optimization debug logs in sws_init_context()

The logged information is possibly false, and it tends to be outdated
after each change since the logging code needs to be manually updated.

Simplify and prevent confusing wrong debug messages.

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
13 years agolsws: use array for storing the supported in/out information
Stefano Sabatini [Wed, 3 Aug 2011 23:34:49 +0000 (01:34 +0200)]
lsws: use array for storing the supported in/out information

Also remove the unnecessary isSupportedIn/Out macros.

Make the code more compact/readable, and simplify the access to
lsws-specific pixel format information.

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
13 years agoapplehttp: fix variant discard logic
Luca Barbato [Wed, 3 Aug 2011 23:06:58 +0000 (01:06 +0200)]
applehttp: fix variant discard logic

The v->ctx is always not NULL now, check for streams presence to
mark the read_header state.

Fixes bug #25, possibly introduced by 603b8bc

13 years agoh263dec: Fix asserts broken by the elimination of FF_COMMON_FRAME.
Alex Converse [Thu, 4 Aug 2011 00:31:11 +0000 (17:31 -0700)]
h263dec: Fix asserts broken by the elimination of FF_COMMON_FRAME.

13 years agocosmetics: remove some stray comments from AVCodec declarations
Diego Biurrun [Wed, 3 Aug 2011 21:28:26 +0000 (23:28 +0200)]
cosmetics: remove some stray comments from AVCodec declarations

13 years agocosmetics: fix indentation/alignment in AVCodec declarations
Diego Biurrun [Wed, 3 Aug 2011 21:27:50 +0000 (23:27 +0200)]
cosmetics: fix indentation/alignment in AVCodec declarations

13 years agoarm: Avoid using the movw instruction needlessly
Måns Rullgård [Mon, 1 Aug 2011 17:21:04 +0000 (20:21 +0300)]
arm: Avoid using the movw instruction needlessly

This fixes building for ARM11 without Thumb2.

Signed-off-by: Martin Storsjö <martin@martin.st>
13 years agoMove an int64_t down in MpegEncContext
Martin Storsjö [Mon, 1 Aug 2011 14:18:13 +0000 (17:18 +0300)]
Move an int64_t down in MpegEncContext

This allows using the same arm assembler offsets for both EABI
and the mach-o ABI.

Signed-off-by: Martin Storsjö <martin@martin.st>
13 years agoRevert "swscale: use 15-bit intermediates for 9/10-bit scaling."
Ronald S. Bultje [Tue, 2 Aug 2011 19:27:43 +0000 (12:27 -0700)]
Revert "swscale: use 15-bit intermediates for 9/10-bit scaling."

This reverts commit ac0fb5934893be554a44d2a1eb7a3bc7bf39da4a. It
causes valgrind errors which I'll want to investigate before
resubmitting this.

13 years agoswscale: use 15-bit intermediates for 9/10-bit scaling.
Ronald S. Bultje [Tue, 2 Aug 2011 04:04:19 +0000 (21:04 -0700)]
swscale: use 15-bit intermediates for 9/10-bit scaling.

13 years agodct32: Add SSE2 ASM optimizations
Vitor Sessak [Sat, 30 Jul 2011 16:39:25 +0000 (18:39 +0200)]
dct32: Add SSE2 ASM optimizations

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agoCorrect chroma vector calculation for RealVideo 3.
Kostya Shishkov [Tue, 2 Aug 2011 11:45:28 +0000 (13:45 +0200)]
Correct chroma vector calculation for RealVideo 3.

Old version divided it wrong, which resulted in chroma drift (visible on FATE
sample too as dirty trails left by clouds).

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agolavf: Add an option to discard corrupted frames
Zohar Kelrich [Sun, 24 Jul 2011 14:28:33 +0000 (17:28 +0300)]
lavf: Add an option to discard corrupted frames

Signed-off-by: Zohar Kelrich <lumimies@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
13 years agompegts: Mark wrongly-sized packets as corrupted
Zohar Kelrich [Sun, 24 Jul 2011 08:13:51 +0000 (11:13 +0300)]
mpegts: Mark wrongly-sized packets as corrupted

Signed-off-by: Zohar Kelrich <lumimies@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
13 years agompegts: Move scan test to handle_packets
Zohar Kelrich [Sun, 24 Jul 2011 08:13:50 +0000 (11:13 +0300)]
mpegts: Move scan test to handle_packets

This fixes an issue where packets which start being read
while reading the header stick around after a seek.

Signed-off-by: Zohar Kelrich <lumimies@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
13 years agompegts: Mark corrupted packets
Zohar Kelrich [Sun, 24 Jul 2011 08:13:50 +0000 (11:13 +0300)]
mpegts: Mark corrupted packets

Signed-off-by: Zohar Kelrich <lumimies@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
13 years agompegts: Reset continuity counter on seek
Zohar Kelrich [Sun, 24 Jul 2011 08:13:50 +0000 (11:13 +0300)]
mpegts: Reset continuity counter on seek

Signed-off-by: Zohar Kelrich <lumimies@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
13 years agompegts: Fix for continuity counter
Zohar Kelrich [Sun, 24 Jul 2011 08:13:50 +0000 (11:13 +0300)]
mpegts: Fix for continuity counter

Make continuity counter respect discontinuity flag
and null packets. Unpack the adaptation_field_control field.

Signed-off-by: Zohar Kelrich <lumimies@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
13 years agompegts: Silence "can't seek" warning on unseekable
Zohar Kelrich [Sun, 24 Jul 2011 08:13:50 +0000 (11:13 +0300)]
mpegts: Silence "can't seek" warning on unseekable

Do not try to seek when we already know we are not allowed to.
Silences warning that always happens when streaming.

Signed-off-by: Zohar Kelrich <lumimies@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
13 years agoapichange: add an entry for AV_PKT_FLAG_CORRUPT
Luca Barbato [Mon, 1 Aug 2011 15:46:29 +0000 (17:46 +0200)]
apichange: add an entry for AV_PKT_FLAG_CORRUPT

13 years agoavpacket: signal possibly corrupted packets
Luca Barbato [Mon, 1 Aug 2011 15:41:44 +0000 (17:41 +0200)]
avpacket: signal possibly corrupted packets

Use AV_PKT_FLAG_CORRUPT flag to mark packets that might be incomplete
or have errors. Formats that have a mean to validate their packets
should use this flag to deliver such information instead of silently
dropping the packets.

13 years agompeg4videodec: remove dead code that would have detected erroneous encoding
Dustin Brody [Thu, 28 Jul 2011 00:44:58 +0000 (20:44 -0400)]
mpeg4videodec: remove dead code that would have detected erroneous encoding

Signed-off-by: Diego Biurrun <diego@biurrun.de>
13 years agoaac: Remove some suspicious illegal memcpy()s from LTP.
Alex Converse [Fri, 29 Jul 2011 22:49:11 +0000 (15:49 -0700)]
aac: Remove some suspicious illegal memcpy()s from LTP.

13 years agobink: Eliminate unnecessary shadow declaration.
Alex Converse [Fri, 29 Jul 2011 21:45:04 +0000 (14:45 -0700)]
bink: Eliminate unnecessary shadow declaration.

13 years agoRemove the ffserver test.
Anton Khirnov [Wed, 27 Jul 2011 18:29:32 +0000 (20:29 +0200)]
Remove the ffserver test.

It doesn't work and is therefore useless.

13 years agoRemove ffserver.h
Anton Khirnov [Tue, 26 Jul 2011 19:57:15 +0000 (21:57 +0200)]
Remove ffserver.h

It's unused an serves no purpose.

13 years agolavc: Deprecate unused FF_ER_VERY_AGGRESSIVE
Dustin Brody [Fri, 29 Jul 2011 07:25:34 +0000 (03:25 -0400)]
lavc: Deprecate unused FF_ER_VERY_AGGRESSIVE

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agox11grab: add show_region AVOption.
Yu-Jie Lin [Sat, 30 Jul 2011 11:13:43 +0000 (19:13 +0800)]
x11grab: add show_region AVOption.

Draw the current grabbing region for indication.

Signed-off-by: Yu-Jie Lin <livibetter@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
13 years agox11grab: add follow_mouse AVOption.
Yu-Jie Lin [Sat, 30 Jul 2011 10:46:36 +0000 (18:46 +0800)]
x11grab: add follow_mouse AVOption.

-follow_mouse centered|PIXELS
  move grabbing region to where mouse pointer at the center; or
  only move when pointer reaches within PIXELS to the edge.

Signed-off-by: Yu-Jie Lin <livibetter@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
13 years agoDo not convert RGB buffer at once when stride does not fit exact samples.
Kostya Shishkov [Fri, 29 Jul 2011 10:55:01 +0000 (12:55 +0200)]
Do not convert RGB buffer at once when stride does not fit exact samples.

When converting RGB format to RGB format with the same bits per sample,
unscaled path performs conversion on the whole buffer at once. For
non-multiple-of-16 BGR24 to RGB24 conversion it means that padding at the
end of line will be converted too. Since it may be of arbitrary length
(e.g. 8 bytes), operating on the whole buffer produces obviously wrong
results.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agobuild: make examples build normally
Luca Barbato [Sat, 30 Jul 2011 09:01:17 +0000 (11:01 +0200)]
build: make examples build normally

The link target was missing $(EXAMPLES)

13 years agomxfdec: Include FF_INPUT_BUFFER_PADDING_SIZE when allocating extradata.
Alex Converse [Fri, 29 Jul 2011 22:27:36 +0000 (15:27 -0700)]
mxfdec: Include FF_INPUT_BUFFER_PADDING_SIZE when allocating extradata.

This prevents out of bounds reads when extradata is being decoded.

13 years agoH.264: tweak some other x86 asm for Atom
Jason Garrett-Glaser [Thu, 28 Jul 2011 01:09:49 +0000 (18:09 -0700)]
H.264: tweak some other x86 asm for Atom

13 years agoprobe: Fix insane flow control.
Alex Converse [Fri, 29 Jul 2011 01:53:19 +0000 (18:53 -0700)]
probe: Fix insane flow control.

A loop control variable was shadowed inside the loop and never updated.

13 years agompegts: remove invalid error check
Mike Williams [Thu, 14 Jul 2011 20:49:21 +0000 (16:49 -0400)]
mpegts: remove invalid error check

mpegts_read_header is used by both mpegts and mpegtsraw, so this
erro check is no longer valid.

Signed-off-by: Mike Williams <mike@mikebwilliams.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
13 years agos302m: use nondeprecated audio sample format API
Dustin Brody [Fri, 29 Jul 2011 06:36:11 +0000 (02:36 -0400)]
s302m: use nondeprecated audio sample format API

Signed-off-by: Anton Khirnov <anton@khirnov.net>
13 years agolavc: use designated initialisers for all codecs.
Anton Khirnov [Sun, 17 Jul 2011 10:54:31 +0000 (12:54 +0200)]
lavc: use designated initialisers for all codecs.

It's more readable and less prone to breakage.

13 years agox86: cabac: add operand size suffixes missing from 6c32576
Mans Rullgard [Fri, 29 Jul 2011 00:55:58 +0000 (01:55 +0100)]
x86: cabac: add operand size suffixes missing from 6c32576

This fixes build with clang.

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agoh263dec: Propagate AV_LOG_ERRORs from slice decoding through frame decoding with...
Dustin Brody [Wed, 27 Jul 2011 23:09:45 +0000 (19:09 -0400)]
h263dec: Propagate AV_LOG_ERRORs from slice decoding through frame decoding with sufficient error recognition

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agox86: cabac: don't load/store context values in asm
Mans Rullgard [Thu, 28 Jul 2011 19:21:33 +0000 (20:21 +0100)]
x86: cabac: don't load/store context values in asm

Inspection of compiled code shows gcc handles these fine on its own.
Benchmarking also shows no measurable speed difference.

Removing the remaining cases in get_cabac_bypass_sign_x86() does
cause more substantial changes to the compiled code with uncertain
impact.

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agoH.264: optimize CABAC x86 asm for Atom
Jason Garrett-Glaser [Wed, 27 Jul 2011 02:08:05 +0000 (19:08 -0700)]
H.264: optimize CABAC x86 asm for Atom

13 years agovp3/theora: flush after seek.
Ronald S. Bultje [Tue, 26 Jul 2011 17:58:29 +0000 (10:58 -0700)]
vp3/theora: flush after seek.

13 years agodoc/fftools-common-opts: wording fixes missing from the previous commit.
Anton Khirnov [Thu, 28 Jul 2011 15:43:24 +0000 (17:43 +0200)]
doc/fftools-common-opts: wording fixes missing from the previous commit.

13 years agodoc: document using AVOptions in fftools.
Anton Khirnov [Sun, 17 Jul 2011 12:54:20 +0000 (14:54 +0200)]
doc: document using AVOptions in fftools.

13 years agocmdutils: add codec_opts parameter to setup_find_stream_info_opts()
Stefano Sabatini [Sun, 17 Jul 2011 14:19:28 +0000 (16:19 +0200)]
cmdutils: add codec_opts parameter to setup_find_stream_info_opts()

Avoid brittle and obfuscating reference to a global.

Signed-off-by: Anton Khirnov <anton@khirnov.net>
13 years agocmdutils: clarify documentation for filter_codec_opts()
Stefano Sabatini [Sat, 16 Jul 2011 23:20:50 +0000 (01:20 +0200)]
cmdutils: clarify documentation for filter_codec_opts()

Signed-off-by: Anton Khirnov <anton@khirnov.net>
13 years agocmdutils: clarify documentation for setup_find_stream_info_opts()
Stefano Sabatini [Sat, 16 Jul 2011 23:30:43 +0000 (01:30 +0200)]
cmdutils: clarify documentation for setup_find_stream_info_opts()

Signed-off-by: Anton Khirnov <anton@khirnov.net>
13 years agolavf: add forgotten attribute_deprecated to av_find_stream_info()
Anton Khirnov [Sun, 17 Jul 2011 11:59:04 +0000 (13:59 +0200)]
lavf: add forgotten attribute_deprecated to av_find_stream_info()

13 years agoppc: remove redundant setting of Altivec IDCT
Mans Rullgard [Mon, 25 Jul 2011 15:18:17 +0000 (16:18 +0100)]
ppc: remove redundant setting of Altivec IDCT

This is already set by dsputil_init_ppc() and is best done in only
one place.

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agompegvideo: initialise DSPContext in ff_dct_common_init()
Mans Rullgard [Mon, 25 Jul 2011 14:51:36 +0000 (15:51 +0100)]
mpegvideo: initialise DSPContext in ff_dct_common_init()

The functions and tables initialised in this function rely on an
initialised DSPContext.  Make sure they always have one.

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agocosmetics: reindent
Justin Ruggles [Mon, 25 Jul 2011 16:54:25 +0000 (12:54 -0400)]
cosmetics: reindent

13 years agoeac3enc: support writing of basic mixing and info metadata
Justin Ruggles [Sun, 29 May 2011 23:16:46 +0000 (19:16 -0400)]
eac3enc: support writing of basic mixing and info metadata

13 years agodnxhdenc: fix declarations in for loops
Mans Rullgard [Wed, 27 Jul 2011 15:14:33 +0000 (16:14 +0100)]
dnxhdenc: fix declarations in for loops

Apparently the gcc warning doesn't trigger on these.

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agodsputil: remove stale bink prototypes and comments
Mans Rullgard [Wed, 27 Jul 2011 14:49:03 +0000 (15:49 +0100)]
dsputil: remove stale bink prototypes and comments

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agodsputil: move a bink-only function to binkdsp
Mans Rullgard [Wed, 27 Jul 2011 14:47:02 +0000 (15:47 +0100)]
dsputil: move a bink-only function to binkdsp

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agodsputil: remove some unused functions
Mans Rullgard [Wed, 27 Jul 2011 13:47:56 +0000 (14:47 +0100)]
dsputil: remove some unused functions

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agobink: make IDCT take 32-bit input
Kostya Shishkov [Sat, 23 Jul 2011 13:46:35 +0000 (15:46 +0200)]
bink: make IDCT take 32-bit input

Since IDCT transforming 32-bit input to 8-bit output is unusual and unpractical
for most codecs, move Bink IDCT into separate context. Get rid of an additional
permutation table while at it since SIMD support for Bink IDCT is unlikely to
be implemented in foreseeable future.
Quantisation tables also have to change type to signed for proper
dequantisation of DCT coefficients.

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agoconfigure: spelling cosmetics
Diego Biurrun [Wed, 13 Jul 2011 16:12:43 +0000 (18:12 +0200)]
configure: spelling cosmetics

13 years agodctref: make sure function declarations match between .c and .h file
Diego Biurrun [Wed, 13 Jul 2011 22:00:43 +0000 (00:00 +0200)]
dctref: make sure function declarations match between .c and .h file

13 years agox86: fix build with gcc 4.7
Mans Rullgard [Tue, 26 Jul 2011 11:23:43 +0000 (12:23 +0100)]
x86: fix build with gcc 4.7

The upcoming gcc 4.7 has more advanced constant propagation
resulting some inline asm operands becoming constants and thus
emitted as literals, sometimes in contexts where this results
in invalid instructions.

This patch changes the constraints of the relevant operands
to "rm" thus forcing a valid type.  While obviously suboptimal,
this is what older gcc versions already did, and there is no
change to the code generated with these.

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agoape: adjust some printf format strings to correctly match argument types
Diego Biurrun [Wed, 1 Jun 2011 13:06:08 +0000 (15:06 +0200)]
ape: adjust some printf format strings to correctly match argument types

13 years agojpegdec: actually search for and parse RSTn
Michael Niedermayer [Thu, 9 Jun 2011 01:35:50 +0000 (03:35 +0200)]
jpegdec: actually search for and parse RSTn

Fixes decoding of MJPEG files produced by some UVC Logitec web cameras,
such as "Notebook Pro" and "HD C910".

References:
http://trac.videolan.org/vlc/ticket/4215
http://ffmpeg.org/trac/ffmpeg/ticket/267

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Reviewed-by: Kostya <kostya.shishkov@gmail.com>
(cherry picked from commit 7b8ed831eb8432d202dad16dedc1758b018bb1fa)

13 years agodnxhddec: avoid a branch in 10-bit decode_dct_block()
Mans Rullgard [Thu, 21 Jul 2011 23:13:27 +0000 (00:13 +0100)]
dnxhddec: avoid a branch in 10-bit decode_dct_block()

The minimum weight value is 32 so this test can be skipped for the
10-bit case.  Overall speedup 3-4%.

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agoH.264: Add optimizations to predict x86 assembly.
Daniel Kang [Fri, 22 Jul 2011 01:15:58 +0000 (21:15 -0400)]
H.264: Add optimizations to predict x86 assembly.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agoriff: Add mpgv MPEG-2 fourcc
Alex Converse [Fri, 22 Jul 2011 17:13:22 +0000 (10:13 -0700)]
riff: Add mpgv MPEG-2 fourcc

Supported by mplayer and seen in the wild.

13 years agoadd Flash Screen Video 2 decoder
Kostya Shishkov [Sat, 16 Jul 2011 10:02:55 +0000 (12:02 +0200)]
add Flash Screen Video 2 decoder

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
13 years agodnxhddec: optimise dnxhd_decode_dct_block()
Mans Rullgard [Thu, 21 Jul 2011 18:04:37 +0000 (19:04 +0100)]
dnxhddec: optimise dnxhd_decode_dct_block()

Template the function for 8/10-bit and use lowlevel bitstream
macros.

6% faster overall on i7 gcc 4.5.

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agortp: remove disabled code
Diego Biurrun [Sun, 17 Jul 2011 11:27:46 +0000 (13:27 +0200)]
rtp: remove disabled code

13 years agoeac3enc: use different numbers of blocks per frame to allow higher bitrates
Justin Ruggles [Thu, 14 Jul 2011 17:02:45 +0000 (13:02 -0400)]
eac3enc: use different numbers of blocks per frame to allow higher bitrates

13 years agodnxhd: add regression test for 10-bit
Mans Rullgard [Wed, 20 Jul 2011 23:55:24 +0000 (00:55 +0100)]
dnxhd: add regression test for 10-bit

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agodnxhd: 10-bit support
Joseph Artsimovich [Wed, 20 Jul 2011 17:58:27 +0000 (18:58 +0100)]
dnxhd: 10-bit support

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agodsputil: update per-arch init funcs for non-h264 high bit depth
Mans Rullgard [Thu, 21 Jul 2011 09:05:15 +0000 (10:05 +0100)]
dsputil: update per-arch init funcs for non-h264 high bit depth

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agodsputil: template get_pixels() for different bit depths
Mans Rullgard [Thu, 21 Jul 2011 01:31:57 +0000 (02:31 +0100)]
dsputil: template get_pixels() for different bit depths

Signed-off-by: Mans Rullgard <mans@mansr.com>
13 years agodsputil: create 16/32-bit dctcoef versions of some functions
Mans Rullgard [Thu, 21 Jul 2011 11:39:41 +0000 (12:39 +0100)]
dsputil: create 16/32-bit dctcoef versions of some functions

High bitdepth H.264 needs 32-bit transform coefficients, whereas
dnxhd does not.  This creates a conflict with the templated
functions operating on DCTELEM data.  This patch adds a field
allowing the caller to choose the element size in dsputil_init()
and adds the required functions.

Signed-off-by: Mans Rullgard <mans@mansr.com>