PeerStreamer

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

Originally committed as revision 16217 to svn://svn.ffmpeg.org/ffmpeg/trunk

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

Originally committed as revision 16215 to svn://svn.ffmpeg.org/ffmpeg/trunk

Remove redundant nnz variable.

Originally committed as revision 16209 to svn://svn.ffmpeg.org/ffmpeg/trunk

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...

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

Originally committed as revision 16201 to svn://svn.ffmpeg.org/ffmpeg/trunk

Split filter_mb_dir() out of filter_mb().
1% overall decoding speed up for cathedral-beta2-400extra-crop-avc.mp4
no speed change for Aladin.mpg
Benchmarks done on Pentium dual

Originally committed as revision 16182 to svn://svn.ffmpeg.org/ffmpeg/trunk

Factorize 3 multiplications out, code becomes 3 cpu cycles faster.
(not significant as thats just per MB)

Originally committed as revision 16174 to svn://svn.ffmpeg.org/ffmpeg/trunk

Do not recalculate constant part of is_complex for every MB.

Originally committed as revision 16172 to svn://svn.ffmpeg.org/ffmpeg/trunk

Enable filter_mb_fast if CODEC_FLAG2_FAST is set.

Originally committed as revision 16141 to svn://svn.ffmpeg.org/ffmpeg/trunk

Check that cpb_count is within the allowed range.

Originally committed as revision 16010 to svn://svn.ffmpeg.org/ffmpeg/trunk

Allocate parameter sets sanely instead of using alloc_parameter_set().
Avoids ending up with half initialized parameter sets.
Fixes issue282.

Originally committed as revision 15971 to svn://svn.ffmpeg.org/ffmpeg/trunk

Move encoder-specific quantize_c and related tables to the H.264 encoder.

Originally committed as revision 15937 to svn://svn.ffmpeg.org/ffmpeg/trunk

Fix H.264 decoding.
(Problem analyzed by Uoti Urpala)

Originally committed as revision 15816 to svn://svn.ffmpeg.org/ffmpeg/trunk

h264: Implement decoding of picture timing SEI message.

Now correct values are propagated to interlaced_frame, top_field_first
and repeat_pict in AVFrame structure.

patch by ffdshow tryouts

Originally committed as revision 15773 to svn://svn.ffmpeg.org/ffmpeg/trunk

uses FF_ARRAY_ELEMS() where appropriate

Originally committed as revision 15662 to svn://svn.ffmpeg.org/ffmpeg/trunk

Rename copy_picture to ff_copy_picture.

Originally committed as revision 15508 to svn://svn.ffmpeg.org/ffmpeg/trunk

Avoid out of picture accesses when CODEC_FLAG_EMU_EDGE is set.
Fixes issue427.

Originally committed as revision 15418 to svn://svn.ffmpeg.org/ffmpeg/trunk

Initialize next/last_picture* as the generic code does not always
set them to sane values.

Originally committed as revision 15412 to svn://svn.ffmpeg.org/ffmpeg/trunk

Fix CAVLC+8x8+MBAFF loop filter.
Fixes FREXT02_JVC_C.

Originally committed as revision 15410 to svn://svn.ffmpeg.org/ffmpeg/trunk

A few empty lines for readability.

Originally committed as revision 15385 to svn://svn.ffmpeg.org/ffmpeg/trunk

Fix ref_shift so that it is correct for more/all? MBAFF/PAFF mixes.
Fixes at least MR9_BT_B.

Originally committed as revision 15380 to svn://svn.ffmpeg.org/ffmpeg/trunk

Correct comment in the direct mode code.
(note, yes this is unrelated to the previous simplification, the
code always behaved like it is documented now.)

Originally committed as revision 15378 to svn://svn.ffmpeg.org/ffmpeg/trunk

Make MBAFF temporal direct mode closer to the spec.
Fixes at least:
CAMANL2_TOSHIBA_B
CVMANL2_TOSHIBA_B
camp_mot_mbaff0_full

Originally committed as revision 15375 to svn://svn.ffmpeg.org/ffmpeg/trunk

Set poc for field references in MBAFF.

Originally committed as revision 15373 to svn://svn.ffmpeg.org/ffmpeg/trunk

Remove check and error message for "MBAFF + !direct_8x8_inference is not implemented"
because this mode does not exist, H.264-2007 says "When frame_mbs_only_flag is
equal to 0, direct_8x8_inference_flag shall be equal to 1."

Originally committed as revision 15369 to svn://svn.ffmpeg.org/ffmpeg/trunk

Ensure that extradata is not parsed twice.
Fixes FRExt1_Panasonic.

Originally committed as revision 15361 to svn://svn.ffmpeg.org/ffmpeg/trunk

Rename error_resilience to error_recognition.

Originally committed as revision 15270 to svn://svn.ffmpeg.org/ffmpeg/trunk

Try to support invalid h264 that does not mark IDR with POC=0.
Fixes issue576.

Originally committed as revision 15153 to svn://svn.ffmpeg.org/ffmpeg/trunk

The deblock filter does not need all the intra availability stuff so do
not recalculate it.

Originally committed as revision 14689 to svn://svn.ffmpeg.org/ffmpeg/trunk

Support MBAFF + constrained intra prediction.
(no i would not have tried to implement this had i known what mess it is)
fixes at least:
CAMACI3_Sony_C

Originally committed as revision 14687 to svn://svn.ffmpeg.org/ffmpeg/trunk

Remove a few more lines of unneeded code in fill_caches() for MBAFF.
Also document why the remaining related part is apparently needed.

Originally committed as revision 14682 to svn://svn.ffmpeg.org/ffmpeg/trunk

Do not execute the part of fill_caches() for the loopfilter that overwrites
reference indexes and motion vectors.
Fixes at least:
CVMAQP2_Sony_G

Originally committed as revision 14678 to svn://svn.ffmpeg.org/ffmpeg/trunk

Remove odd pairwise deblock filtering applied to MBAFF MBs.

Originally committed as revision 14673 to svn://svn.ffmpeg.org/ffmpeg/trunk

indent

Originally committed as revision 14622 to svn://svn.ffmpeg.org/ffmpeg/trunk

simplify y_shift/ref_shift code

Originally committed as revision 14620 to svn://svn.ffmpeg.org/ffmpeg/trunk

Factorize some code between temporal and spatial direct mode.

Originally committed as revision 14618 to svn://svn.ffmpeg.org/ffmpeg/trunk

Use frame_num and reference instead of poc for matching frames for direct
mode MBs, this seems to work better with field/frame mixes. POC of both
can be the same and can be different that makes its use tricky.

Originally committed as revision 14545 to svn://svn.ffmpeg.org/ffmpeg/trunk

Port col_parity code from spatial mode to temporal mode,
I also love it if people duplicate code and fix one but not the other.

Originally committed as revision 14543 to svn://svn.ffmpeg.org/ffmpeg/trunk

Remove redundant code.

Originally committed as revision 14541 to svn://svn.ffmpeg.org/ffmpeg/trunk

Fix static leaks in h264.c.

Patch by Art Clarke aclarke A vlideshow D com

Originally committed as revision 14538 to svn://svn.ffmpeg.org/ffmpeg/trunk

Fix dist_scale_factor for long term reference pictures.

Originally committed as revision 14536 to svn://svn.ffmpeg.org/ffmpeg/trunk

Remove useless mb_field_decoding_flag setting code that was at the wrong spot.

Originally committed as revision 14519 to svn://svn.ffmpeg.org/ffmpeg/trunk

Picture.ref_count/ref_poc have to be stored per field (actually also per
slice but thats a seperate bug)
Fixes at least:
CABREF3_Sand_D.264
camp_mot_fld0_full.26l
CVFI2_Sony_H.jsv
CVNLFI2_Sony_H.jsv

Originally committed as revision 14511 to svn://svn.ffmpeg.org/ffmpeg/trunk

Use #define instead of a constant. Patch by Paul Kendall.

Originally committed as revision 14463 to svn://svn.ffmpeg.org/ffmpeg/trunk

Initialize outputed_poc to INT_MIN at decoder initialization.
Fixes issue 560

Originally committed as revision 14428 to svn://svn.ffmpeg.org/ffmpeg/trunk

And.... Fix another spec reference to be more specific and correct.

Originally committed as revision 14423 to svn://svn.ffmpeg.org/ffmpeg/trunk

Simple refactoring to resolve a FIXME. Patch by Paul Kendall.

Originally committed as revision 14419 to svn://svn.ffmpeg.org/ffmpeg/trunk

chroma_format_idc=0 aka grayscale support.
Can be disabled by removing #define ALLOW_NOCHROMA in case the extra if()
slow the code down measurably.
Fixes at least
FRExt/HPCAMOLQ_BRCM_B.264
FRExt/HPCVMOLQ_BRCM_B.264

Originally committed as revision 14407 to svn://svn.ffmpeg.org/ffmpeg/trunk

Remove another 2 incorrect checks.
These would ignore fields of different parity.

Originally committed as revision 14404 to svn://svn.ffmpeg.org/ffmpeg/trunk

Remove mysterious ref->data⁰ check.
What the standard calls non-existent is not related to the
value of the data⁰ pointer.

Originally committed as revision 14402 to svn://svn.ffmpeg.org/ffmpeg/trunk