Update Cygwin Ports URL.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16288 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Bring up to date the Cygwin docs.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16287 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Make h264 parser 50% faster.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16286 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

100l, I broke H.264 again, forgot one hunk.
Thanks to FATE for finding it.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16285 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Optimize 0 0 0-3 search, 45% faster on pentium dual.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16284 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

The diffutils package is also needed (for the 'cmp' command).
Spotted by Ramiro Polla.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16283 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Revert r16257:

Include "libavutil/common.h" where we use llrint() in case ffmpeg's
own llrint() is to be used.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16282 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Fix two identical warnings when compiling riff.c and nuv.c with icc:
warning #188: enumerated type mixed with another type

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16281 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Fix a warning when compiling with icc:
warning #188: enumerated type mixed with another type

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16280 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Simplify decode_cabac_mb_ref() a little bit, 2 cpu cycles faster on
pentium dual.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16279 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Indent

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16278 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

inline decode_cabac_mb_type for I & P frames, 9 cycles faster on pentium dual.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16277 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Negate 2 more variables, 1 cpu cycle faster on pentium dual.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16276 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Simplify if/else, no speed change

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16275 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Negate a few variables, this simplifies the code and makes it 5 cycles faster
on pentium dual.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16274 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Simplify ifs(), 8 cpu cycles faster on pentium dual

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16273 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Simplify if(), 3 cpu cycles faster in pentium dual.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16272 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Initialize s->opaque before calling url_resetbuf().
Fixes a regression test crash for icc 10.1 64 bit.

Patch by Aurelien.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16271 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Rename libavcodec/i386/ --> libavcodec/x86/.
It contains optimizations that are not specific to i386 and
libavutil uses this naming scheme already.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16270 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Offer an option (disabled by default) to enable the copy of the frames before
the first key frame in a stream.
Patch by Wolfram Gloger wmglo /!\ dent med uni (minus) muenchen de

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16269 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Decode TIFF image only after all tags have been decoded

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16268 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Some TIFFs declare -1 for the single strip height

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16267 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Add monochrome TIFF support

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16266 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Calculate line size variable correctly for lower bitdepths and use it for raw data copying

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16265 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Register TIFF tags for CCITT Group 3 and 4 compression options

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16264 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Remove a bunch of unused variables.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16263 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Use ARCH_X86_32 instead of !ARCH_X86_64, it is more straightforward.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16262 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Fix incorrectly constructed Dirac parse units that caused A/V sync loss.
Fixes issue 694.
patch by Anuradha Suraparaju, anuradha rd.bbc.co uk

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16261 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Do not use full include path for headers in the same directory.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16260 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Change mpeg audio parser so it only sets frame_size, channels and bit_rate
after it has a few valid frames.
Fixes issue762

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16259 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Silence the following icc warnings:
warning #1292: attribute "force_align_arg_pointer" ignored
warning #1292: attribute "alloc_size" ignored
warning #1292: attribute "cold" ignored

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16258 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Include "libavutil/common.h" where we use llrint() in case ffmpeg's
own llrint() is to be used.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16257 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Fix PTS returned by the demuxers for mpeg4 without a decoder.
More precissely fix has_b_frames.
Fixes issue757

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16256 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Move CODEC_FLAG_LOW_DELAY into ff_mpeg4_decode_picture_header().

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16255 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Fix detection of audio codec in K70707-ARIA229.flv.
Fixes issue760.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16254 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Remove useless code.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16253 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Only execute clear_blocks() when needed.
+0.3% speedup for both aladin & cathedral.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16252 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

AAC: Reindent after last commit

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16251 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

AAC: Fix AAC prediction when used in conjunction with the CPE

Patch by Alex Converse (alex converse gmail com)

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16250 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

AAC: Don't write an illegal sampling_index in a PCE to the MPEG4AudioConfig
member

Patch by Alex Converse (alex converse gmail com)

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16249 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Optimize get_dct8x8_allowed().
30 cpu cycles faster in pentium dual.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16248 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

RV30 decoder

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16247 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Use AV_GCC_VERSION_AT_LEAST() to simplify gcc version checks.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16246 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

RV30 loop filter

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16245 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Cosmetics by Kenan Gillet. Part 1 of 3 of his 'qcelp: silence handling'
changeset.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16244 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Put under #if LIBAVCODEC_VERSION_MAJOR < 53 the deprecated functions
av_set_string() and av_set_string2(), they will be dropped at the next
major bump.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16243 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Cosmetics, change variable name "o2" to "o" in ffserver_opt_default().

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16242 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Remove calls to deprecated av_set_string2() with calls to
av_set_string3().

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16241 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

H.264 loopfilter speed tweaks

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16240 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Port x264 deblocking code to libavcodec.  This includes SSE2 luma deblocking code and both MMXEXT and SSE2 luma intra deblocking code for H.264 decoding.  This assembly is available under --enable-gpl and speeds decoding of Cathedral by 7%.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16239 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Remove pointless malloc.h #include.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16238 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Document new idct requirements.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16237 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Fix decoding with the plain C idcts of
FRExt/HPCAMOLQ_BRCM_B
FRExt/HPCAQ2LQ_BRCM_B
FRExt/HPCVMOLQ_BRCM_B

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16236 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

a table for upcoming RV30 loop filtering

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16235 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Add automatic prefix handling to yasm functions.  Does nothing now, but will be useful for porting x264 asm in the future.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16234 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Replace /2 by >>1 in decode_cabac_mb_dqp()
3 cpu cycles speed up on pentium dual.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16233 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Simplify ctx update in decode_cabac_mb_dqp().
no speed change

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16232 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Simplify ctx calculation in decode_cabac_mb_dqp()
no speed change

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16231 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Clear FF_INPUT_BUFFER_PADDING_SIZE bytes at the end of NALs in rbsp_buffer.

Fixes valgrind uninitialized value warnings at the end of decoding H.264
frames.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16230 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Allocate and clear FF_INPUT_BUFFER_PADDING bytes at the end of mkv extradata.

Fixes valgrind invalid read warnings during H.264 extradata parsing.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16229 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Move filter_luma_intra into dsputil for later addition of asm.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16228 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Simplify chroma AC in CABAC residual decoding.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16227 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Optimize ctx calculation in decode_cabac_mb_mvd(), code by dark shikari.
The case for 16x16 blocks becomes 10 cpu cycles faster on pentium dual,
i could not find a speed difference in the case of subblocks though.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16226 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Remove unacceptable NULL pointer hack from mc code.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16225 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Check ref values in CABAC H.264 for validity.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16224 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Move idct_(dc)add closer to where it is needed.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16223 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

indent

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16222 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Reorder ifs in chroma hl_decode_mb to avoid a duplicate transform_bypass
check.
14 cpu cycles speedup

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16221 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

s/h->cbp_table[mb_xy]/h->cbp/

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16220 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Faster CAVLC decoding of trailing_ones. Based on a patch by dark shikari.
decode_residual is about 3.3% faster.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16219 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Replace i by trailing_ones, part of a patch by dark shikari.
No speed change meassureable with START/STOP_TIMER, but this is needed
for future optimizations.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16218 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Remove if() surrounding decode_cabac_mb_type() that can never be true.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16217 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Remove unreachable else clause, found by dark shikari.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16216 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Remove useless IS_8x8DCT check i forgot, spotted by dark shikari.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16215 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

10l, svq3 didnt set h->cbp, this broke decoding a little.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16214 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

cosmetics, remove useless parenthesis and whitespaces

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16213 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

set alac channels from extradata, fix alac mono in m4a

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16212 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

cosmetics: remove one unneeded space

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16211 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Do not calculate idct_dc_add/idct_add when the variables are unused.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16210 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Remove redundant nnz variable.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16209 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

indent

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16208 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Use the new idct functions (except chroma as it was slower in benchmarks)
cathedral +0.5% speed
aladin +0.6% speed [note aladin has been cat-ed 10 times to reduce the influence
of init time]
Speedup also verified via START/STOP_TIMER (difference was very significant
for the changed parts)

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16207 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

H.264 idct functions that include the chroma, inter luma and intra16 luma loops
thus avoiding the calling overhead.
New functions are not yet used.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16206 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Make asf_get_packet() and asf_parse_packet() public for use in RTSP-MS and
give them a ff_ prefix. See "[PATCH] asf.c: make functions of interest for
MS-RTSP public" thread on ML.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16205 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Add a ByteIOContext argument to ASF functions that will be shared with the
RTSP stack for RTSP-MS support. This way, they can read input from any given
input source. See discussion in "[PATCH] add gb argument to ASF functions of
interest to MS-RTSP" ML thread.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16204 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Skip non intra luma code when there is no coded luma.
0.7% speedup for the cathedral sample.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16203 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

1.6x faster pred*_vertical_add_c

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16202 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Skip chroma handling when there is no coded chroma.
0.5% overall speedup for the cathedral sample.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16201 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Fix weird indent.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16200 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Implement size autodetection if one of the width and height
AVFormatParameters for grab_read_header() is non-positive.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16199 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Reokace the width and height fields in VideoData with the struct
video_window video_win field.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16198 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Replace frame_rate and frame_rate_base fields in VideoData with an
AVRational time_base field.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16197 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Fix opt_default(), making it exit immediately in case of an invalid
argument.
See the thread: "[PATCH] Fix opt_default()".

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16196 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

add const qualifier to some pointers for input data

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16195 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

use precalculated mb_pos where available

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16194 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

ARM: make FASTDIV() an inline function

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16193 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Merge deblocking pattern with CBP for RV30/40 loop filtering

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16192 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Generate correct deblock pattern for RV30

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16191 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

Unfortunately RV30 luma thirdpel MC functions cannot be substituted with
applying two separate filtering passes, so implement them explicitly.

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16190 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b

4l: TIFF stores short strings inside tag, do not interpret it is as an offset

Fixes issue 753
-This line, and those below, will be ignored--

M    libavcodec/tiff.c

git-svn-id: file:///var/local/repositories/ffmpeg/trunk@16189 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b