]> git.sesse.net Git - ffmpeg/log
ffmpeg
10 years agodxva2: Log errors verbosely
Sam Lantinga [Sun, 19 Jan 2014 14:42:36 +0000 (15:42 +0100)]
dxva2: Log errors verbosely

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
10 years agodxva2: Retry IDirectXVideoDecoder_BeginFrame()
Sam Lantinga [Thu, 9 Jan 2014 19:59:54 +0000 (19:59 +0000)]
dxva2: Retry IDirectXVideoDecoder_BeginFrame()

If the function returns E_PENDING retry for a fixed number of times.

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
10 years agohevc: Bound check cu_qp_delta
Luca Barbato [Sat, 11 Jan 2014 19:13:20 +0000 (20:13 +0100)]
hevc: Bound check cu_qp_delta

The T-REC-H.265-2013044 page 91 states it has to be in the range
[-(26 + s->sps->qp_bd_offset / 2), (25 + s->sps->qp_bd_offset / 2)].

10 years agohevc: Bound check slice_qp
Luca Barbato [Sat, 11 Jan 2014 19:13:20 +0000 (20:13 +0100)]
hevc: Bound check slice_qp

The T-REC-H.265-2013044 page 79 states it has to be in the range
[-s->sps->qp_bd_offset, 51].

Sample-Id: 00001386-google
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
10 years agohevc: Prevent some integer overflows
Luca Barbato [Sat, 11 Jan 2014 11:33:42 +0000 (12:33 +0100)]
hevc: Prevent some integer overflows

get_ue_golomb_long() returns an unsigned.

Sample-Id: 00001541-google
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
10 years agohevc: Remove useless clip
Guillaume Martres [Sat, 11 Jan 2014 21:46:25 +0000 (22:46 +0100)]
hevc: Remove useless clip

The src buffer should only contain values in the interval
[0, (1 << BIT_DEPTH) - 1].

Since shift = (BIT_DEPTH - 5), src[x] >> shift must be in
the interval [0, 31], so no clip is needed.

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
10 years agohevc: clip pixels when transquant bypass is used
Guillaume Martres [Sat, 11 Jan 2014 21:46:24 +0000 (22:46 +0100)]
hevc: clip pixels when transquant bypass is used

The reconstructed picture should always be clipped (see section 8.6.5),
previously we did not clip coding units where
cu_transquant_bypass_flag == 1.

Sample-Id: 00001325-google
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
10 years agohevc: Use uint64 to check for tile dimensions
Luca Barbato [Sat, 11 Jan 2014 02:20:45 +0000 (03:20 +0100)]
hevc: Use uint64 to check for tile dimensions

And use unsigned datatypes.

Otherwise it would overflow.

Sample-Id: 00001315-google
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
10 years agohevc: Fix modulo operations
Michael Niedermayer [Fri, 10 Jan 2014 20:32:05 +0000 (21:32 +0100)]
hevc: Fix modulo operations

Keep qp fields within the range.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
10 years agohevc: check that VPS referenced from SPS exists
Michael Niedermayer [Fri, 10 Jan 2014 20:32:04 +0000 (21:32 +0100)]
hevc: check that VPS referenced from SPS exists

This matches how its done for SPS/PPS.
Fixes null pointer dereference.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
10 years agovc1: Always reset numref when parsing a new frame header.
Michael Niedermayer [Sun, 19 Jan 2014 15:28:25 +0000 (15:28 +0000)]
vc1: Always reset numref when parsing a new frame header.

Fixes an issue where the B-frame coding mode switches from interlaced
fields to interlaced frames, causing incorrect decisions in the motion
compensation code and resulting in visual artifacts.

CC: libav-stable@libav.org
Signed-off-by: Tim Walker <tdskywalker@gmail.com>
10 years agortmpproto: Send a full, absolute timestamp if it isn't monotonically growing
Martin Storsjö [Fri, 17 Jan 2014 12:30:57 +0000 (14:30 +0200)]
rtmpproto: Send a full, absolute timestamp if it isn't monotonically growing

The normal differential timestamps can't handle negative
differences, thus send a full packet header with an absolute
timestamp in these cases.

Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agortmpproto: Avoid using uninitialized memory
Martin Storsjö [Fri, 17 Jan 2014 12:29:57 +0000 (14:29 +0200)]
rtmpproto: Avoid using uninitialized memory

If the url ends with .flv, we stripped it but didn't initialize
rt->playpath, doing av_strlcat on an uninitialized buffer.

Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agomov: do not set avg_frame_rate in the demuxer
Justin Ruggles [Thu, 16 Jan 2014 20:59:05 +0000 (15:59 -0500)]
mov: do not set avg_frame_rate in the demuxer

The track duration is often not reliable or is not the duration
represented by the number of frames. In those cases, avg_frame_rate
was reported incorrectly. Removing this code falls back to the
default calculation in avformat_find_stream_info().

This is a partial revert of commit c3aeaa540.

Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
10 years agolavc: deprecate CODEC_FLAG_EMU_EDGE and avcodec_get_edge_width().
Anton Khirnov [Thu, 2 Jan 2014 10:07:11 +0000 (11:07 +0100)]
lavc: deprecate CODEC_FLAG_EMU_EDGE and avcodec_get_edge_width().

10 years agompegvideo: allocate edges when encoding.
Anton Khirnov [Wed, 1 Jan 2014 19:04:53 +0000 (20:04 +0100)]
mpegvideo: allocate edges when encoding.

Allocating edges will be removed from generic code, so the encoder must
handle this explicitly.

10 years agolavc: allow the caller to override dimensions in ff_get_buffer()
Anton Khirnov [Wed, 1 Jan 2014 19:03:05 +0000 (20:03 +0100)]
lavc: allow the caller to override dimensions in ff_get_buffer()

This will be useful for allocating edges in the encoders.

10 years agoeac3: cosmetics, re-indent.
Tim Walker [Fri, 17 Jan 2014 15:08:18 +0000 (16:08 +0100)]
eac3: cosmetics, re-indent.

10 years ago(e)ac3: clip surround mix level indexes.
Tim Walker [Fri, 17 Jan 2014 15:06:34 +0000 (16:06 +0100)]
(e)ac3: clip surround mix level indexes.

Values 0-2 are reserved, 3 (-1.5dB) should be used instead.

10 years ago(e)ac3: parse and store the Lt/Rt and LFE mix levels.
Tim Walker [Fri, 17 Jan 2014 14:57:40 +0000 (15:57 +0100)]
(e)ac3: parse and store the Lt/Rt and LFE mix levels.

10 years agoac3: set default matrix encoding modes in parse_frame_header.
Tim Walker [Fri, 17 Jan 2014 14:18:45 +0000 (15:18 +0100)]
ac3: set default matrix encoding modes in parse_frame_header.

Deduplicates some code.

10 years agoh264: check that an IDR NAL only contains I slices
Anton Khirnov [Thu, 28 Nov 2013 09:54:35 +0000 (10:54 +0100)]
h264: check that an IDR NAL only contains I slices

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC:libav-stable@libav.org

10 years agoh264: reset num_reorder_frames if it is invalid
Anton Khirnov [Thu, 28 Nov 2013 09:54:35 +0000 (10:54 +0100)]
h264: reset num_reorder_frames if it is invalid

An invalid VUI is not considered a fatal error, so the SPS containing it
may still be used. Leaving an invalid value of num_reorder_frames there
can result in writing over the bounds of H264Context.delayed_pic.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC:libav-stable@libav.org

10 years agoh264: reset ref count if decoding the slice header fails
Anton Khirnov [Thu, 28 Nov 2013 09:54:35 +0000 (10:54 +0100)]
h264: reset ref count if decoding the slice header fails

Otherwise the ER code might try to use some already freed references.

Fixes possible access to freed memory.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC:libav-stable@libav.org

10 years agoh264: prevent two matching fields from being both a short/long ref combination
Anton Khirnov [Thu, 28 Nov 2013 09:54:35 +0000 (10:54 +0100)]
h264: prevent two matching fields from being both a short/long ref combination

Fixes possible access to freed memory.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC:libav-stable@libav.org

10 years agoh264: do not call field_end if we do not have a current picture
Anton Khirnov [Thu, 28 Nov 2013 09:54:35 +0000 (10:54 +0100)]
h264: do not call field_end if we do not have a current picture

Fixes invalid reads.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC:libav-stable@libav.org

10 years agoh264: limit allowed pred modes in ff_h264_check_intra_pred_mode() to 3
Anton Khirnov [Thu, 28 Nov 2013 09:54:35 +0000 (10:54 +0100)]
h264: limit allowed pred modes in ff_h264_check_intra_pred_mode() to 3

Higher modes are not allowed for 16x16/chroma, which is what this
function is used for. Otherwise this function would return 0 (vertical
prediction) for invalid higher modes, which could result in invalid
reads.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC:libav-stable@libav.org

10 years agoh264: reset first_field if frame_start() fails for missing refs
Anton Khirnov [Thu, 28 Nov 2013 09:54:35 +0000 (10:54 +0100)]
h264: reset first_field if frame_start() fails for missing refs

In this case we may not have a current frame, while first_field being
set implies we do.

Fixes invalid reads.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC:libav-stable@libav.org

10 years agoPrepare for 10_alpha2 Release
Reinhard Tartler [Sat, 18 Jan 2014 15:27:27 +0000 (10:27 -0500)]
Prepare for 10_alpha2 Release

10 years agoavisynth: fix setting packet properties
Michael Niedermayer [Mon, 13 Jan 2014 15:09:26 +0000 (10:09 -0500)]
avisynth: fix setting packet properties

Reviewed-by: Stephen Hutchinson <qyot27@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
10 years agobuffersrc: handle non-refcounted frames in av_buffersrc_add_frame() correctly
Anton Khirnov [Thu, 16 Jan 2014 10:06:02 +0000 (11:06 +0100)]
buffersrc: handle non-refcounted frames in av_buffersrc_add_frame() correctly

10 years agoyuv4mpeg: set average frame rate
Anton Khirnov [Wed, 15 Jan 2014 20:14:46 +0000 (21:14 +0100)]
yuv4mpeg: set average frame rate

10 years agoavconv: print a warning when falling back to default 25fps
Anton Khirnov [Wed, 15 Jan 2014 20:32:38 +0000 (21:32 +0100)]
avconv: print a warning when falling back to default 25fps

10 years agomp3enc: allow omitting the id3v2 header with -id3v2_version 0
Anton Khirnov [Tue, 14 Jan 2014 21:06:03 +0000 (22:06 +0100)]
mp3enc: allow omitting the id3v2 header with -id3v2_version 0

10 years agomp3enc: add an option for disabling the Xing frame.
Anton Khirnov [Tue, 14 Jan 2014 20:47:45 +0000 (21:47 +0100)]
mp3enc: add an option for disabling the Xing frame.

10 years agodoxy: Document better the available AVFrame flags
Luca Barbato [Thu, 16 Jan 2014 00:04:57 +0000 (01:04 +0100)]
doxy: Document better the available AVFrame flags

10 years agodoxy: Add AVFrame to the structured modules
Luca Barbato [Thu, 16 Jan 2014 00:04:35 +0000 (01:04 +0100)]
doxy: Add AVFrame to the structured modules

10 years agoaviocat: Add support for specifying the input duration
Martin Storsjö [Thu, 16 Jan 2014 08:01:03 +0000 (10:01 +0200)]
aviocat: Add support for specifying the input duration

This avoids the caller having to calculate the byte rate if wanting
to push a file in a rate resembling realtime.

Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agoaviocat: Check the argv array length before reading element i+1
Martin Storsjö [Thu, 16 Jan 2014 07:58:03 +0000 (09:58 +0200)]
aviocat: Check the argv array length before reading element i+1

Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agox86: dsputil: Simplify xvmc deprecation conditional
Diego Biurrun [Tue, 14 Jan 2014 17:24:03 +0000 (18:24 +0100)]
x86: dsputil: Simplify xvmc deprecation conditional

10 years agobuild: Add separate hidden config option for the intrax8 code
Diego Biurrun [Tue, 14 Jan 2014 17:24:02 +0000 (18:24 +0100)]
build: Add separate hidden config option for the intrax8 code

This allows for slightly cleaner dependency specification.

10 years agoaarch64: port neon clobber test from arm
Janne Grunau [Sat, 11 Jan 2014 16:21:19 +0000 (17:21 +0100)]
aarch64: port neon clobber test from arm

10 years agoaarch64: h264 (bi)weight NEON optimizations
Janne Grunau [Fri, 10 Jan 2014 13:07:24 +0000 (14:07 +0100)]
aarch64: h264 (bi)weight NEON optimizations

Ported from ARMv7 NEON.

10 years agoaarch64: h264 loop filter NEON optimizations
Janne Grunau [Fri, 20 Dec 2013 20:02:43 +0000 (21:02 +0100)]
aarch64: h264 loop filter NEON optimizations

Ported from ARMv7 NEON.

10 years agoaarch64: hpeldsp NEON optimizations
Janne Grunau [Fri, 20 Dec 2013 19:03:58 +0000 (20:03 +0100)]
aarch64: hpeldsp NEON optimizations

Ported from ARMv7 NEON.

10 years agoaarch64: h264 qpel NEON optimizations
Janne Grunau [Wed, 18 Dec 2013 14:56:50 +0000 (15:56 +0100)]
aarch64: h264 qpel NEON optimizations

Ported from ARMv7 NEON.

10 years agoaarch64: h264 idct NEON assembler optimizations
Janne Grunau [Thu, 12 Dec 2013 23:33:48 +0000 (00:33 +0100)]
aarch64: h264 idct NEON assembler optimizations

Ported from ARMv7 NEON.

10 years agoaarch64: h264 chroma motion compensation NEON optimizations
Janne Grunau [Tue, 10 Dec 2013 19:16:08 +0000 (20:16 +0100)]
aarch64: h264 chroma motion compensation NEON optimizations

Since RV40 and VC-1 use almost the same algorithm so optimizations for
those two decoders are easy to do and included.

10 years agoaarch64: add cpuflags support for NEON and VFP
Janne Grunau [Tue, 10 Dec 2013 19:13:32 +0000 (20:13 +0100)]
aarch64: add cpuflags support for NEON and VFP

NEON and VFP are currently mandatory for all ARMv8 profiles. Both are
handled as extensions as far as cpuflags are concerned. This is
consistent with handling x86_64 which always has SSE2, but still
handles it as an extension.

10 years agodct-test: test ff_simple_idct_neon only on ARM for now
Janne Grunau [Mon, 13 Jan 2014 20:32:19 +0000 (21:32 +0100)]
dct-test: test ff_simple_idct_neon only on ARM for now

Signed-off-by: Janne Grunau <janne-libav@jannau.net>
10 years agomov: Free an earlier allocated array if allocating a new one
Martin Storsjö [Mon, 13 Jan 2014 12:46:07 +0000 (14:46 +0200)]
mov: Free an earlier allocated array if allocating a new one

It could probably also be considered an error if the pointer isn't
null at this point, but then we might risk rejecting some
slightly broken files that we might have handled so far.

Sample-Id: 00000496-google
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agomov: Free intermediate arrays in the normal cleanup function
Martin Storsjö [Mon, 13 Jan 2014 12:43:23 +0000 (14:43 +0200)]
mov: Free intermediate arrays in the normal cleanup function

These arrays are normally freed at the end of mov_read_trak,
but make sure they're freed in case mov_read_trak returned
early (due to errors) or in case the atoms that allocate arrays
are encountered at some other point than within a trak (which
we don't have checks against).

Sample-Id: 00000496-google
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agoaarch64: bswap inline assembly
Janne Grunau [Mon, 25 Nov 2013 15:58:54 +0000 (16:58 +0100)]
aarch64: bswap inline assembly

Signed-off-by: Janne Grunau <janne-libav@jannau.net>
10 years agoaarch64: configure: set arch dependent features
Janne Grunau [Thu, 12 Dec 2013 16:45:32 +0000 (17:45 +0100)]
aarch64: configure: set arch dependent features

Stack is always 16 byte aligned and clz, 64bit operations and unaligned
memory access are fast in aarch64 mode on ARMv8.

Signed-off-by: Janne Grunau <janne-libav@jannau.net>
10 years agocabac: remove leftovers from the cabac encoder
Janne Grunau [Mon, 13 Jan 2014 20:23:13 +0000 (21:23 +0100)]
cabac: remove leftovers from the cabac encoder

The cabac encoder was only used by the removed cabac test.

Signed-off-by: Janne Grunau <janne-libav@jannau.net>
10 years agomxf: Fix potential leak in mxf_read_local_tags()
Tomas Härdin [Wed, 24 Oct 2012 14:54:25 +0000 (16:54 +0200)]
mxf: Fix potential leak in mxf_read_local_tags()

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
10 years agomxf: Correctly support files from Pinnacle Thunder
Tomas Härdin [Wed, 4 Apr 2012 12:24:49 +0000 (14:24 +0200)]
mxf: Correctly support files from Pinnacle Thunder

Such files have IndexTableSegments which when parsed cover EditUnit
ranges like this:

 [0,1)
 [249,250)
 [249,377)
 [0,249)

where each interval is

 [IndexStartPosition, IndexStartPosition + IndexDuration)

This would be reduced to a sparse index like:

 [0,1), [249,250)

instead of the full range:

 [0,249), [249,377)

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
10 years agomxf: Fix a possible leak of extradata
Michael Niedermayer [Tue, 11 Dec 2012 13:43:47 +0000 (14:43 +0100)]
mxf: Fix a possible leak of extradata

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
10 years agomxf: Use av_malloc_array
Luca Barbato [Tue, 7 Jan 2014 22:54:09 +0000 (23:54 +0100)]
mxf: Use av_malloc_array

10 years agomxf: Drop unnecessary checks
Luca Barbato [Tue, 7 Jan 2014 20:56:33 +0000 (21:56 +0100)]
mxf: Drop unnecessary checks

av_reallocp_array does the check already.

10 years agomxf: Fix off by one error in d10 aes3 decoding
Marton Balint [Wed, 18 Jul 2012 23:23:20 +0000 (01:23 +0200)]
mxf: Fix off by one error in d10 aes3 decoding

Without this fix the last sample was missing from the packet.

10 years agomxf: Do not use int to check the seek position
Luca Barbato [Tue, 7 Jan 2014 20:26:14 +0000 (21:26 +0100)]
mxf: Do not use int to check the seek position

Overly large files are to be expected.

Reported-by: Jean-Baptiste Kempf <jb@videolan.org>
10 years agomxf: Set AV_FIELD_PROGRESSIVE
Reimar Döffinger [Thu, 3 Jan 2013 20:42:35 +0000 (21:42 +0100)]
mxf: Set AV_FIELD_PROGRESSIVE

Needed for AVC-intra

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
10 years agovf_yadif: Relicense from GPL to LGPL
Robert Krüger [Sun, 5 Jan 2014 18:53:11 +0000 (19:53 +0100)]
vf_yadif: Relicense from GPL to LGPL

All copyright holders have agreed to the relicensing.

10 years agox86: Consistently use cpu flag detection macros in places that still miss it
Diego Biurrun [Mon, 13 Jan 2014 17:52:32 +0000 (18:52 +0100)]
x86: Consistently use cpu flag detection macros in places that still miss it

10 years agoarm: Use full filenames as multiple inclusion guards
Diego Biurrun [Mon, 13 Jan 2014 14:26:16 +0000 (15:26 +0100)]
arm: Use full filenames as multiple inclusion guards

10 years agolavf: make av_probe_input_buffer more robust
Anton Khirnov [Mon, 13 Jan 2014 12:47:07 +0000 (13:47 +0100)]
lavf: make av_probe_input_buffer more robust

Always use the actually read size as the offset instead of making
possibly invalid assumptions.

Addresses: CVE-2012-6618

10 years agolavf: use a fixed width type
Anton Khirnov [Mon, 13 Jan 2014 10:56:59 +0000 (11:56 +0100)]
lavf: use a fixed width type

It's shorter and more consistent with the rest of the code.

10 years agolavf: simplify handling of offset in av_probe_input_buuffer()
Anton Khirnov [Mon, 13 Jan 2014 10:55:18 +0000 (11:55 +0100)]
lavf: simplify handling of offset in av_probe_input_buuffer()

10 years agox86/arm: Add clobber tests to libavresample
Martin Storsjö [Sat, 11 Jan 2014 20:43:26 +0000 (22:43 +0200)]
x86/arm: Add clobber tests to libavresample

Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agovc1: Enable the interlaced B-frame codepath.
Tim Walker [Sat, 11 Jan 2014 22:58:16 +0000 (23:58 +0100)]
vc1: Enable the interlaced B-frame codepath.

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
10 years agovc1: Add avg_no_rnd_vc1_chroma_mc4_c()
Michael Niedermayer [Thu, 25 Apr 2013 19:31:18 +0000 (21:31 +0200)]
vc1: Add avg_no_rnd_vc1_chroma_mc4_c()

Needed for proper interlaced support.

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
10 years agovc1: Factorize out chroma MC
Luca Barbato [Sat, 11 Jan 2014 22:59:54 +0000 (23:59 +0100)]
vc1: Factorize out chroma MC

10 years agovc1dsp: K&R formatting cosmetics
Luca Barbato [Sat, 11 Jan 2014 23:59:24 +0000 (00:59 +0100)]
vc1dsp: K&R formatting cosmetics

Signed-off-by: Diego Biurrun <diego@biurrun.de>
10 years agoflv: Workaround for buggy Omnia A/XE encoder
Benjamin Larsson [Sun, 5 Jan 2014 13:42:14 +0000 (14:42 +0100)]
flv: Workaround for buggy Omnia A/XE encoder

The Omnia A/XE encoder writes the explicit extra data incorrectly
and wrongly disables parametric stereo. Truncating the extra data
by setting the size to 2 works around this. The AAC extra data
parser will then only parse the correct parts.

Bug-id: 599

10 years agoarm: Add an option for making sure NEON registers aren't clobbered
Martin Storsjö [Fri, 20 Dec 2013 11:51:50 +0000 (13:51 +0200)]
arm: Add an option for making sure NEON registers aren't clobbered

This is pretty much based on the same test for XMM registers.

Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agofate: Explicitly specify the rgb555le pixel format for the mss2 rgb555 tests
Martin Storsjö [Fri, 10 Jan 2014 14:05:10 +0000 (16:05 +0200)]
fate: Explicitly specify the rgb555le pixel format for the mss2 rgb555 tests

This fixes these fate tests on big endian.

Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agoavutil: do not use avcodec header in frame.h
Vittorio Giovara [Wed, 8 Jan 2014 00:20:18 +0000 (01:20 +0100)]
avutil: do not use avcodec header in frame.h

10 years agoavfilter: fix leaks on error in ff_filter_frame
Vittorio Giovara [Wed, 18 Dec 2013 10:45:01 +0000 (11:45 +0100)]
avfilter: fix leaks on error in ff_filter_frame

10 years agoavfilter: add needs_writable field to the internal AVFilterPad structure
Vittorio Giovara [Wed, 18 Dec 2013 13:21:25 +0000 (14:21 +0100)]
avfilter: add needs_writable field to the internal AVFilterPad structure

10 years agoavfilter: add documentation for needs_writable
Vittorio Giovara [Wed, 18 Dec 2013 10:40:38 +0000 (11:40 +0100)]
avfilter: add documentation for needs_writable

10 years agofate: add framepack filter test
Vittorio Giovara [Tue, 5 Nov 2013 15:00:49 +0000 (16:00 +0100)]
fate: add framepack filter test

10 years agolavfi: add framepack filter
Vittorio Giovara [Thu, 9 Jan 2014 18:50:19 +0000 (19:50 +0100)]
lavfi: add framepack filter

10 years agoflac muxer: add option to disable writing the global header
Justin Ruggles [Thu, 9 Jan 2014 17:11:26 +0000 (12:11 -0500)]
flac muxer: add option to disable writing the global header

The global header is not explicitly required for playback, and
omitting it allows for simple concatenation.

10 years agoATRAC3+ decoder
Maxim Polijakowski [Fri, 3 Jan 2014 22:23:11 +0000 (23:23 +0100)]
ATRAC3+ decoder

Cleanup by Diego Biurrun.

Signed-off-by: Kostya Shishkov <kostya.shishkov@gmail.com>
10 years agox86: hpeldsp: Add missing av_cold attribute to init function
Diego Biurrun [Thu, 9 Jan 2014 13:06:32 +0000 (14:06 +0100)]
x86: hpeldsp: Add missing av_cold attribute to init function

10 years agox86: avcodec: Add a bunch of missing #includes for av_cold
Diego Biurrun [Thu, 9 Jan 2014 13:06:31 +0000 (14:06 +0100)]
x86: avcodec: Add a bunch of missing #includes for av_cold

10 years agoac3tab.h: #include the correct headers
Diego Biurrun [Thu, 9 Jan 2014 13:06:30 +0000 (14:06 +0100)]
ac3tab.h: #include the correct headers

10 years agomjpegdec: apply flipping after decoding, not before
Anton Khirnov [Fri, 20 Dec 2013 09:44:57 +0000 (10:44 +0100)]
mjpegdec: apply flipping after decoding, not before

This is simpler and removes a silly restriction on edges being present.

10 years agohevc: fix decoding of one PU wide files
Guillaume Martres [Fri, 3 Jan 2014 08:10:38 +0000 (09:10 +0100)]
hevc: fix decoding of one PU wide files

For those the block size may be larger than the source linesize (if the
edges are not allocated).

Signed-off-by: Anton Khirnov <anton@khirnov.net>
10 years agoFATE: add MSS2 tests
Anton Khirnov [Tue, 17 Dec 2013 11:44:17 +0000 (12:44 +0100)]
FATE: add MSS2 tests

10 years agovp9: drop support for real (non-emulated) edges
Anton Khirnov [Fri, 20 Dec 2013 10:20:24 +0000 (11:20 +0100)]
vp9: drop support for real (non-emulated) edges

They are not measurably faster on x86, they might be somewhat faster on
other platforms due to missing emu edge SIMD, but the gain is not large
enough to justify the added complexity.

10 years agovp8: drop support for real (non-emulated) edges
Anton Khirnov [Fri, 20 Dec 2013 10:20:24 +0000 (11:20 +0100)]
vp8: drop support for real (non-emulated) edges

They are not measurably faster on x86, they might be somewhat faster on
other platforms due to missing emu edge SIMD, but the gain is not large
enough to justify the added complexity.

10 years agompegvideo: drop support for real (non-emulated) edges
Anton Khirnov [Fri, 20 Dec 2013 10:20:24 +0000 (11:20 +0100)]
mpegvideo: drop support for real (non-emulated) edges

Several decoders disable those anyway and they are not measurably faster
on x86. They might be somewhat faster on other platforms due to missing
emu edge SIMD, but the gain is not large enough (and those decoders
relevant enough) to justify the added complexity.

10 years agoarm: Add a missing # as prefix for an immediate constant
Martin Storsjö [Tue, 7 Jan 2014 12:01:24 +0000 (14:01 +0200)]
arm: Add a missing # as prefix for an immediate constant

Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agoarm: Allow overriding the alignment set in the function macro
Martin Storsjö [Tue, 7 Jan 2014 10:13:11 +0000 (12:13 +0200)]
arm: Allow overriding the alignment set in the function macro

The function macro always sets .align 2 before declaring the
function label (since 5c5e1ea3) and always sets the section to
.text (since 278caa6a).

The .align 5 before certain functions, added in fc252eba, were added
before .text and .align were added to the function macro and thus
became useless/unused when the function macro got them.

This restores the original intention, to align the loop entry
points.

Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agoarm: Remove a leftover define for the pld instruction
Martin Storsjö [Tue, 7 Jan 2014 10:05:56 +0000 (12:05 +0200)]
arm: Remove a leftover define for the pld instruction

This file no longer uses the pld instruction at all, all such uses
have been split into hpeldsp_arm.S.

Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agoarm: cosmetics: Reindent the h264dsp neon init function
Martin Storsjö [Tue, 7 Jan 2014 09:15:49 +0000 (11:15 +0200)]
arm: cosmetics: Reindent the h264dsp neon init function

Signed-off-by: Martin Storsjö <martin@martin.st>
10 years agonetwork.h: Add #endif comments to improve readability
Diego Biurrun [Mon, 6 Jan 2014 10:26:20 +0000 (11:26 +0100)]
network.h: Add #endif comments to improve readability

10 years agoRename CONFIG_FFT_FLOAT ---> FFT_FLOAT
Diego Biurrun [Mon, 6 Jan 2014 13:48:18 +0000 (14:48 +0100)]
Rename CONFIG_FFT_FLOAT ---> FFT_FLOAT

The define does not originate from configure, so it should not
have a name that is CONFIG_-prefixed.