if running with multiple threads and CODEC_FLAGS2_FAST is set.
Thus, it may decode the slices in parallel to gain speed.
Patch by Andreas Öman: [andreas olebyn nu]
Originally committed as revision 10431 to svn://svn.ffmpeg.org/ffmpeg/trunk
Patch by Andreas Öman % andreas A olebyn P nu %
NB: depends on having a thread library activated at config time, and on
having a source encoded with multiple slices
Original threads:
date: May 18, 2007 11:00 PM
subject: [FFmpeg-devel] Parallelized h264 proof-of-concept
date: Jun 15, 2007 10:10 PM
subject: [FFmpeg-devel] [PATCH] h264 parallelized, (was: Parallelized h264 proof-of-concept)
date: Jun 25, 2007 7:02 PM
subject: Re: [FFmpeg-devel] [PATCH] h264 parallelized
Originally committed as revision 10407 to svn://svn.ffmpeg.org/ffmpeg/trunk
patch by Carl Eugen Hoyos: [cehoyos ag or at]
original thread:
[FFmpeg-devel] [PATCH] Don't let ctx->skip_frame>0 produce errors
date: 08/30/2007 01:30 PM
Originally committed as revision 10286 to svn://svn.ffmpeg.org/ffmpeg/trunk
you write them in the right order it comes out backwards.
This removes them from fill_rectangle().
patch by Alexander Strange %astrange A ithinksw P com%
Original thread:
Date: Aug 14, 2007 5:36 AM
Subject: [FFmpeg-devel] [PATCH] two small h264 optimizations
Originally committed as revision 10118 to svn://svn.ffmpeg.org/ffmpeg/trunk
returns 0. This leads to a 0.4% speed-up.
Patch by Alexander Strange astrange at_ ithinksw dot com
Original thread:
Date: Aug 11, 2007 11:44 PM
Subject: [FFmpeg-devel] [PATCH] h264: don't check decode_cabac_residual return
Originally committed as revision 10084 to svn://svn.ffmpeg.org/ffmpeg/trunk
Patch by Andreas Öman %andreas A olebyn P nu%
Original thread:
Date: Jul 7, 2007 1:23 AM
Subject: [FFmpeg-devel] Corrupted blocks and seeking issues in H264 disc sources
Originally committed as revision 9836 to svn://svn.ffmpeg.org/ffmpeg/trunk
since the MVs are in qpel res.
Patch by Andreas Öman % andreas A olebyn P nu %
Date: Jul 14, 2007 12:40 PM
Subject: [FFmpeg-devel] [PATCH] h264 mv visualization
Originally committed as revision 9688 to svn://svn.ffmpeg.org/ffmpeg/trunk
for Cr and Cb
Patch by Andreas Öman % andreas A olebyn P nu %
Original thread:
Date: Jun 26, 2007 8:48 PM
subject: [FFmpeg-devel] Color corruption and seeking errors with H264 disc sources
Originally committed as revision 9505 to svn://svn.ffmpeg.org/ffmpeg/trunk
this saves speed for the upcoming secondqp fix.
Patch by Andreas Öman % andreas A olebyn P nu %
Original thread:
Date: Jun 26, 2007 8:48 PM
subject: [FFmpeg-devel] Color corruption and seeking errors with H264 disc sources
Originally committed as revision 9498 to svn://svn.ffmpeg.org/ffmpeg/trunk
this saves speed for the upcoming secondqp fix.
Patch by Andreas Öman % andreas A olebyn P nu %
Original thread:
Date: Jun 26, 2007 8:48 PM
subject: [FFmpeg-devel] Color corruption and seeking errors with H264 disc sources
Originally committed as revision 9497 to svn://svn.ffmpeg.org/ffmpeg/trunk
at slice boundaries for deblocking-type 2 content.
This is needed for slice based threading only and doesn't do much
good or bad otherwise.
Patch by Andreas Oman %andreas A olebyn P nu%
Original thread:
date: Jun 18, 2007 1:21 PM
subject: Re: [FFmpeg-devel] [PATCH] h264 parallelized,
Originally committed as revision 9380 to svn://svn.ffmpeg.org/ffmpeg/trunk
the intra and inter -nal units are escaped
patch by Andreas Öman: \andreas olebyn nu/
original thread:
[FFmpeg-devel] [PATCH] h264: rbsp de-escape and data partitioning..
date: 06/20/2007 09:32 AM
Originally committed as revision 9374 to svn://svn.ffmpeg.org/ffmpeg/trunk
(done in order to implement slice-level parallel decoding)
Patch by Andreas Öman % andreas olebyn nu %
Original thread:
Date: Jun 15, 2007 10:10 PM
Subject: [FFmpeg-devel] [PATCH] h264 parallelized, (was: Parallelized h264 proof-of-concept)
Originally committed as revision 9371 to svn://svn.ffmpeg.org/ffmpeg/trunk
original thread:
Date: Jun 15, 2007 10:10 PM
Subject: [FFmpeg-devel] [PATCH] h264 parallelized, (was: Parallelized h264 proof-of-concept)
Originally committed as revision 9341 to svn://svn.ffmpeg.org/ffmpeg/trunk
patch by Andreas Öman: [andreas olebyn nu]
original thread:
subject: [FFmpeg-devel] [patch] h264: use 'simple' in border backup / xchg
date: 06/07/2007 03:24 PM
Originally committed as revision 9237 to svn://svn.ffmpeg.org/ffmpeg/trunk
patch by Carl Eugen Hoyos cehoyos chez ag or at
original thread: [FFmpeg-devel] [PATCH] attribute_unused -> av_unused
date: 05/29/2007 01:23 PM
Originally committed as revision 9155 to svn://svn.ffmpeg.org/ffmpeg/trunk
patch by Andreas Öman andreas ta olebyn tod nu
reference thread:
subject: [FFmpeg-devel] [PATCH] h264: allocate PPS and SPS dynamically
date: 05/28/2007 03:00 PM
Originally committed as revision 9148 to svn://svn.ffmpeg.org/ffmpeg/trunk
the H.264 encoder. These functions are: pred8x8_left_dc_c, pred8x8_top_dc_c,
pred16x16_left_dc_c and pred16x16_top_dc_c.
Originally committed as revision 9107 to svn://svn.ffmpeg.org/ffmpeg/trunk
attribute_unused and attribute_used respectively to ease compiling on non-gcc.
Originally committed as revision 9024 to svn://svn.ffmpeg.org/ffmpeg/trunk
Patch by Mean % fixounet A free P fr %
Original thread:
Date: Apr 29, 2007 2:00 PM
Subject: Re: [Ffmpeg-devel] [patch] h264.c, dont go beyond buffer in h264_decode_nal_unit
Originally committed as revision 8858 to svn://svn.ffmpeg.org/ffmpeg/trunk
This allows building shared libraries on AMD64 again.
based on a patch by Diego 'Flameeyes' Pettenò and suggestions by Michael
original thread:
Date: Wed, 18 Apr 2007 11:26:12 +0200
Subject: [Ffmpeg-devel] [PATCH] (try 2) Build shared libraries on AMD64 again
Originally committed as revision 8849 to svn://svn.ffmpeg.org/ffmpeg/trunk
Patch by Limin Wang % lance P lmwang A gmail P com %
Original thread:
date: 04/09/2007 03:54 PM
subject: [Ffmpeg-devel] [PATCH] fix segment fault in h264_parse if buf_size is zero
Originally committed as revision 8714 to svn://svn.ffmpeg.org/ffmpeg/trunk
calls decode_rbsp_trailing() and therefore bit_length might get negative.
Although the remaining code is able to handle a negative bit_length, avoid
the calculation at all by setting bit_length to 0 for dst_length == 0.
patch by Reinhard Nissl, rnissl gmx de
Originally committed as revision 8690 to svn://svn.ffmpeg.org/ffmpeg/trunk
Consider the following byte sequence
00 00 01 0a 00 00 00 01 09 ...
^ ^
A B
decode_nal() determines dst_length to be 1 (i. e. the byte between label
A and B above). However, this byte is a trailing zero byte as the spec
says the the current NAL unit is terminated by a byte sequence 00 00 00.
The current code used a loop to decrement dst_length accordingly. But the
loop doesn't start as the loop condition checks for dst_length > 1, which
should read dst_length > 0.
patch by Reinhard Nissl, rnissl gmx de
Originally committed as revision 8689 to svn://svn.ffmpeg.org/ffmpeg/trunk
i.e. the four bytes 00 00 01 0a.
When decode_nal() decodes the end of sequence NAL unit, it returns with
dst_length == 0. The original code leads to a return -1 which discards
the current properly decoded frame.
patch by Reinhard Nissl, rnissl gmx de
Originally committed as revision 8688 to svn://svn.ffmpeg.org/ffmpeg/trunk
interested in using a debugger to debug FFmpeg.
Original thread:
Subject: [PATCH] Fix compilation when using --disable-opts
Date: 2007-03-15 16:58:35 GMT
Originally committed as revision 8549 to svn://svn.ffmpeg.org/ffmpeg/trunk
144095->142319 dezicycles for hl_decode_mb() on duron
trailing whitespace removed by me
Originally committed as revision 8106 to svn://svn.ffmpeg.org/ffmpeg/trunk
remove silly ref_count<0 and ref_count==0 checks its impossible for this variable to have such a value
Originally committed as revision 7999 to svn://svn.ffmpeg.org/ffmpeg/trunk
Patch by Francois Oligny-Lemieux % eucloid A gmail P com %
Original thread:
Date: Feb 9, 2007 12:00 AM
Subject: [Ffmpeg-devel] h264.c patch, always decoding extradata when on non avc stream
Originally committed as revision 7904 to svn://svn.ffmpeg.org/ffmpeg/trunk
increase delayed_pic buffer size (one temporary is used and a terminating NULL is assumed by most code so it has to be 18 large)
Originally committed as revision 7663 to svn://svn.ffmpeg.org/ffmpeg/trunk
Patch by Frank %eucloid A gmail P com%
date: Jan 18, 2007 6:48 PM
subject: Re: [Ffmpeg-devel] h264, protection against corrupted data (second try patch)
AND
date: Jan 17, 2007 8:22 PM
subject: [Ffmpeg-devel] h264, protection against corrupted data
this also fixes a possible security issue (the sps and pps ids where not checked,
then used as index into an array of sps/pps structs which was then filled with data from the bitstream)
Originally committed as revision 7585 to svn://svn.ffmpeg.org/ffmpeg/trunk
patch by Alexander Chemeris % ipse P ffmpeg A gmail.com %
Original thread:
date: Dec 5, 2006 7:26 PM
subject: [Ffmpeg-devel] [PATCH] Fix crush when truncated slice passed to H.264 decoder
Originally committed as revision 7229 to svn://svn.ffmpeg.org/ffmpeg/trunk
x86 is 4% faster on P3
C sign stuff + x86 code for everything else is also faster then before (sorry forgot to test pure C)
... and if i replace the second occurance of the sign decoding in decode_residual by the asm too then everything gets slower iam starting to think that it might be best to write the whole function in asm, playing this avoid random deoptimizations game with gcc is not fun at all
Originally committed as revision 6732 to svn://svn.ffmpeg.org/ffmpeg/trunk
