platform/upstream/libav.git
14 years agoDon't use tcp_fd if we're not using TCP-based connections (e.g. when
Jeremy Morton [Tue, 26 Jan 2010 15:51:54 +0000 (15:51 +0000)]
Don't use tcp_fd if we're not using TCP-based connections (e.g. when
reading direct SDP files to set up UDP-based RTP-streams). Fixes
issue 1713. Patch by Jeremy Morton <ffmpeg game-point net>.

Originally committed as revision 21461 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoremove unneeded check
Michael Niedermayer [Tue, 26 Jan 2010 15:34:21 +0000 (15:34 +0000)]
remove unneeded check

Originally committed as revision 21460 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoUse left_mb_xy from fill_caches instead of recalculating it.
Michael Niedermayer [Tue, 26 Jan 2010 14:57:53 +0000 (14:57 +0000)]
Use left_mb_xy from fill_caches instead of recalculating it.

Originally committed as revision 21459 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoSimplify left_xy content for the loop filter, this also makes it closer to
Michael Niedermayer [Tue, 26 Jan 2010 14:55:19 +0000 (14:55 +0000)]
Simplify left_xy content for the loop filter, this also makes it closer to
what is needed and its faster too.

Originally committed as revision 21458 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoSimplify loop filter a little by using top/left_type.
Michael Niedermayer [Tue, 26 Jan 2010 13:39:26 +0000 (13:39 +0000)]
Simplify loop filter a little by using top/left_type.

Originally committed as revision 21457 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoSet top & left types for deblock in fill_caches().
Michael Niedermayer [Tue, 26 Jan 2010 13:38:18 +0000 (13:38 +0000)]
Set top & left types for deblock in fill_caches().

Originally committed as revision 21456 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years ago100l typo fix, mixed up +-1 forcing the loop filter skip to never skip.
Michael Niedermayer [Tue, 26 Jan 2010 13:30:33 +0000 (13:30 +0000)]
100l typo fix, mixed up +-1 forcing the loop filter skip to never skip.

Originally committed as revision 21455 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agocosmetic
Michael Niedermayer [Tue, 26 Jan 2010 13:28:55 +0000 (13:28 +0000)]
cosmetic

Originally committed as revision 21454 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoFix qp_thres loop filter check for MBAFF.
Michael Niedermayer [Tue, 26 Jan 2010 13:27:22 +0000 (13:27 +0000)]
Fix qp_thres loop filter check for MBAFF.

Originally committed as revision 21453 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoOptimize mb neighbor initialization for MBAFF in fill_caches().
Michael Niedermayer [Tue, 26 Jan 2010 10:35:36 +0000 (10:35 +0000)]
Optimize mb neighbor initialization for MBAFF in fill_caches().
~10 cpu cycles speedup.

Originally committed as revision 21452 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoMetasoft MJPEG codec has Cb and Cr swapped, fixes issue 1611.
Carl Eugen Hoyos [Tue, 26 Jan 2010 10:30:24 +0000 (10:30 +0000)]
Metasoft MJPEG codec has Cb and Cr swapped, fixes issue 1611.

Originally committed as revision 21451 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoUse isAnyRGB() where possible
Michael Niedermayer [Tue, 26 Jan 2010 08:03:28 +0000 (08:03 +0000)]
Use isAnyRGB() where possible

Originally committed as revision 30435 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

14 years agoAdd 3 more RGB makros that allow specifying RGB in bytes and any rgb/bgr.
Michael Niedermayer [Tue, 26 Jan 2010 08:01:50 +0000 (08:01 +0000)]
Add 3 more RGB makros that allow specifying RGB in bytes and any rgb/bgr.

Originally committed as revision 30434 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

14 years agoRename isRGB() and isBGR() as their meaning is confusing.
Michael Niedermayer [Tue, 26 Jan 2010 07:40:14 +0000 (07:40 +0000)]
Rename isRGB() and isBGR() as their meaning is confusing.

Originally committed as revision 30433 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

14 years agoMerge two "if (flags & SWS_PRINT_INFO) { ... }" blocks, slightly
Stefano Sabatini [Mon, 25 Jan 2010 22:46:09 +0000 (22:46 +0000)]
Merge two "if (flags & SWS_PRINT_INFO) { ... }" blocks, slightly
simplify code.

Originally committed as revision 30432 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

14 years agoFix heap overflow due to lack of nb_components check.
Michael Niedermayer [Mon, 25 Jan 2010 13:26:10 +0000 (13:26 +0000)]
Fix heap overflow due to lack of nb_components check.

Originally committed as revision 21450 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoRelease buffer when decoding is finished.
Carl Eugen Hoyos [Mon, 25 Jan 2010 10:46:32 +0000 (10:46 +0000)]
Release buffer when decoding is finished.

Originally committed as revision 21449 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoPrint config of each lib if different from that of ffmpeg/ffplay/ffserver
Måns Rullgård [Mon, 25 Jan 2010 03:48:15 +0000 (03:48 +0000)]
Print config of each lib if different from that of ffmpeg/ffplay/ffserver

Originally committed as revision 21448 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoOptimize decode_cabac_field_decoding_flag().
Michael Niedermayer [Mon, 25 Jan 2010 02:44:34 +0000 (02:44 +0000)]
Optimize decode_cabac_field_decoding_flag().
~4 cpu cycles faster

Originally committed as revision 21447 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoconfigure: fix --cpu=nocona
Måns Rullgård [Mon, 25 Jan 2010 02:03:10 +0000 (02:03 +0000)]
configure: fix --cpu=nocona

Use x86_64_suggest instead of _select as otherwise fast_cmov being
disabled (as it should be for nocona) also disables x86_64 breaking
the build.

Originally committed as revision 21446 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years ago100l, correct wrong H.264+adaptive MBAFF decoding
Alexander Strange [Mon, 25 Jan 2010 01:56:13 +0000 (01:56 +0000)]
100l, correct wrong H.264+adaptive MBAFF decoding

Fixes several FATE tests.

Originally committed as revision 21445 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoCheck if dxva2api.h is recent enough to compile vc1_dxva2_hwaccel.
Ramiro Polla [Mon, 25 Jan 2010 01:34:48 +0000 (01:34 +0000)]
Check if dxva2api.h is recent enough to compile vc1_dxva2_hwaccel.

Originally committed as revision 21444 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoLoad the proper getaddrinfo functions from ws2_32.dll, if they're present.
Martin Storsjö [Mon, 25 Jan 2010 01:28:01 +0000 (01:28 +0000)]
Load the proper getaddrinfo functions from ws2_32.dll, if they're present.

Patch by Martin Storsjö <martin at martin dot st>.

Originally committed as revision 21443 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoCheck for getaddrinfo in the winsock headers.
Martin Storsjö [Mon, 25 Jan 2010 01:26:54 +0000 (01:26 +0000)]
Check for getaddrinfo in the winsock headers.

This makes us link directly to getaddrinfo if it is available, making
binaries incompatible with anything prior to windows XP. If compatibility
is desired, compile with --target-os=mingw32-prexp.

Patch by Martin Storsjö <martin at martin dot st>.

Originally committed as revision 21442 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoOnly define _WIN32_WINNT=0x0501 if $target_os = "mingw32".
Martin Storsjö [Mon, 25 Jan 2010 01:25:59 +0000 (01:25 +0000)]
Only define _WIN32_WINNT=0x0501 if $target_os = "mingw32".
If compatibility with pre-XP versions is desired, use
--target-os=mingw32-prexp
Patch by Martin Storsjö <martin at martin dot st>.

Originally committed as revision 21441 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoH.264: Use 64-/128-bit write-combining macros for copies
Alexander Strange [Mon, 25 Jan 2010 00:30:44 +0000 (00:30 +0000)]
H.264: Use 64-/128-bit write-combining macros for copies

2-3% faster decode on x86-32 core2.

Originally committed as revision 21440 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoOnly check frame size if the header is valid.
Justin Ruggles [Sun, 24 Jan 2010 23:55:33 +0000 (23:55 +0000)]
Only check frame size if the header is valid.

Originally committed as revision 21439 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoFix decoding ATI VCR2 codec, sample in
Carl Eugen Hoyos [Sun, 24 Jan 2010 22:47:50 +0000 (22:47 +0000)]
Fix decoding ATI VCR2 codec, sample in
samples/V-codecs/ATI-VCR12/ATI/VCR2test.avi
Software decoding was broken since r8865, when two new context variables
were added, but not initialized in vcr2_init_sequence().
Another such context variable was added in r18381.

Originally committed as revision 21438 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoDoxument url_open().
Stefano Sabatini [Sun, 24 Jan 2010 22:07:32 +0000 (22:07 +0000)]
Doxument url_open().

Originally committed as revision 21437 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoForce codec_tag I420 for Quicktime Planar Y'CbCr 8-bit 4:2:0 raw codec,
Carl Eugen Hoyos [Sun, 24 Jan 2010 21:37:35 +0000 (21:37 +0000)]
Force codec_tag I420 for Quicktime Planar Y'CbCr 8-bit 4:2:0 raw codec,
because YV12 triggers an UV swap in rawdec.c, fixes issue 1687.

Originally committed as revision 21436 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoFix mis-computation of the needsDither variable erroneously introduced
Stefano Sabatini [Sun, 24 Jan 2010 21:28:17 +0000 (21:28 +0000)]
Fix mis-computation of the needsDither variable erroneously introduced
in r30419, which was causing a swscale-example regression.

Also increase my liter count by 20.0 units.

Originally committed as revision 30431 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

14 years agoBranchless setting of MB_TYPE_8x8DCT.
Michael Niedermayer [Sun, 24 Jan 2010 20:54:09 +0000 (20:54 +0000)]
Branchless setting of MB_TYPE_8x8DCT.
Not benchmarked as i failed to find a sample that uses this one. But it should be faster.

Originally committed as revision 21435 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoRemove cruft.
Michael Niedermayer [Sun, 24 Jan 2010 20:52:49 +0000 (20:52 +0000)]
Remove cruft.

Originally committed as revision 21434 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoMark index as unsigend so gcc knows it doesnt need a sign extension on x86_64.
Michael Niedermayer [Sun, 24 Jan 2010 19:19:09 +0000 (19:19 +0000)]
Mark index as unsigend so gcc knows it doesnt need a sign extension on x86_64.
Any tips on how i can convince gcc that it doesnt need a
mov     %eax, %eax
in every get_bits() ?

Originally committed as revision 21433 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoa[b-1] -> (a-1)[b].
Michael Niedermayer [Sun, 24 Jan 2010 18:42:22 +0000 (18:42 +0000)]
a[b-1] -> (a-1)[b].
Helps gcc not to add seperate -1 instructions.

Originally committed as revision 21432 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoOptimize suffix_length computation, 1 cpu cycle speedup.
Michael Niedermayer [Sun, 24 Jan 2010 18:23:46 +0000 (18:23 +0000)]
Optimize suffix_length computation, 1 cpu cycle speedup.

Originally committed as revision 21431 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoSimplify suffix_length computation, same speed.
Michael Niedermayer [Sun, 24 Jan 2010 18:18:08 +0000 (18:18 +0000)]
Simplify suffix_length computation, same speed.

Originally committed as revision 21430 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoFavor the term "URL" over "filename", as the API deals with URLs
Stefano Sabatini [Sun, 24 Jan 2010 18:18:03 +0000 (18:18 +0000)]
Favor the term "URL" over "filename", as the API deals with URLs
rather than with simple filenames.

Originally committed as revision 21429 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoOptimize level_code computation, 6cpu cycles speedup.
Michael Niedermayer [Sun, 24 Jan 2010 18:17:01 +0000 (18:17 +0000)]
Optimize level_code computation, 6cpu cycles speedup.

Originally committed as revision 21428 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoMake url_read_complete handle EAGAIN more intelligently.
Reimar Döffinger [Sun, 24 Jan 2010 18:09:46 +0000 (18:09 +0000)]
Make url_read_complete handle EAGAIN more intelligently.
Only retry 2 - 5 times in quick succession and afterwards sleep a bit
to avoid creating high CPU load without any progress.

Originally committed as revision 21427 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoFix crash in MLP decoder due to integer overflow.
Reimar Döffinger [Sun, 24 Jan 2010 18:07:29 +0000 (18:07 +0000)]
Fix crash in MLP decoder due to integer overflow.
Probably only DoS, init_get_bits sets buffer to NULL, thus causing a
NULL-dereference directly after.

Originally committed as revision 21426 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years ago1 cpu cycle faster suffix_length calculation.
Michael Niedermayer [Sun, 24 Jan 2010 18:05:02 +0000 (18:05 +0000)]
1 cpu cycle faster suffix_length calculation.

Originally committed as revision 21425 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoMerge two "if (flags & SWS_PRINT_INFO) { ... }" blocks.
Stefano Sabatini [Sun, 24 Jan 2010 17:44:43 +0000 (17:44 +0000)]
Merge two "if (flags & SWS_PRINT_INFO) { ... }" blocks.

Originally committed as revision 30425 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

14 years agoVC-1/WMV3 DXVA2 implementation
Laurent Aimar [Sun, 24 Jan 2010 17:27:00 +0000 (17:27 +0000)]
VC-1/WMV3 DXVA2 implementation

 It allows VLD VC-1/WMV3 decoding using DXVA2 (GPU assisted decoding API under
VISTA and Windows 7).
 It is implemented by using AVHWAccel API.

Originally committed as revision 21424 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoRemove is_complex from loop_filter() its useless there in its current form.
Michael Niedermayer [Sun, 24 Jan 2010 17:02:44 +0000 (17:02 +0000)]
Remove is_complex from loop_filter() its useless there in its current form.

Originally committed as revision 21423 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoAdded a few missing consts in dxva_h264.
Laurent Aimar [Sun, 24 Jan 2010 16:40:58 +0000 (16:40 +0000)]
Added a few missing consts in dxva_h264.

Originally committed as revision 21422 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoAdded a missing const to ff_h264_get_slice_type().
Laurent Aimar [Sun, 24 Jan 2010 16:37:12 +0000 (16:37 +0000)]
Added a missing const to ff_h264_get_slice_type().

Originally committed as revision 21421 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoDoxument url_close().
Stefano Sabatini [Sun, 24 Jan 2010 16:33:56 +0000 (16:33 +0000)]
Doxument url_close().

Originally committed as revision 21420 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoAdded missing const to get_bits_count().
Laurent Aimar [Sun, 24 Jan 2010 16:22:09 +0000 (16:22 +0000)]
Added missing const to get_bits_count().

Originally committed as revision 21419 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoPrefer cbp over cbp_table.
Michael Niedermayer [Sun, 24 Jan 2010 13:43:26 +0000 (13:43 +0000)]
Prefer cbp over cbp_table.

Originally committed as revision 21418 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoDoxument url_open_protocol().
Stefano Sabatini [Sun, 24 Jan 2010 13:31:25 +0000 (13:31 +0000)]
Doxument url_open_protocol().

Originally committed as revision 21417 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoRemove all uses of slice_type* from the loop filter, also remove its
Michael Niedermayer [Sun, 24 Jan 2010 13:20:17 +0000 (13:20 +0000)]
Remove all uses of slice_type* from the loop filter, also remove its
initialization befre the loop filter.

Originally committed as revision 21416 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoDoxument url_read().
Stefano Sabatini [Sun, 24 Jan 2010 13:07:54 +0000 (13:07 +0000)]
Doxument url_read().

Originally committed as revision 21415 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoUse av_get_bits_per_pixel() for computing the bits per pixel of the
Stefano Sabatini [Sun, 24 Jan 2010 12:55:05 +0000 (12:55 +0000)]
Use av_get_bits_per_pixel() for computing the bits per pixel of the
source and destination format, cache those values in the newly added
SwsContext:srcFormatBpp and SwsContext:dstFormatBpp fields, and remove
the fmt_depth() function.

Originally committed as revision 30419 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

14 years agoIndent.
Ramiro Polla [Sun, 24 Jan 2010 03:02:40 +0000 (03:02 +0000)]
Indent.

Originally committed as revision 30413 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

14 years agoDon't redefine M_PI. It is already defined from libavutil/mathematics.h.
Ramiro Polla [Sun, 24 Jan 2010 02:16:26 +0000 (02:16 +0000)]
Don't redefine M_PI. It is already defined from libavutil/mathematics.h.

Originally committed as revision 30412 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

14 years agoSplit swscale.c into scaler code (swscale.c) and utility code (utils.c).
Ramiro Polla [Sun, 24 Jan 2010 02:08:22 +0000 (02:08 +0000)]
Split swscale.c into scaler code (swscale.c) and utility code (utils.c).

Originally committed as revision 30411 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

14 years agoRemove unneeded reset of non_zero_count_cache for deblock.
Michael Niedermayer [Sat, 23 Jan 2010 22:11:46 +0000 (22:11 +0000)]
Remove unneeded reset of non_zero_count_cache for deblock.

Originally committed as revision 21414 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoRemove useless things from the deblock side of fill_caches().
Michael Niedermayer [Sat, 23 Jan 2010 21:57:36 +0000 (21:57 +0000)]
Remove useless things from the deblock side of fill_caches().

Originally committed as revision 21413 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agomake mv_cache init 64bit where possible.
Michael Niedermayer [Sat, 23 Jan 2010 21:45:12 +0000 (21:45 +0000)]
make mv_cache init 64bit where possible.

Originally committed as revision 21412 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agovfwcap: Include windows.h before vfw.h since the latter requires defines
kemuri [Sat, 23 Jan 2010 20:58:29 +0000 (20:58 +0000)]
vfwcap: Include windows.h before vfw.h since the latter requires defines
from the former.
Patch by kemuri <kemuri9 at gmail dot com>

Originally committed as revision 21411 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoReplace every usage of -lvfw32 with what is particularly necessary for
kemuri [Sat, 23 Jan 2010 20:42:00 +0000 (20:42 +0000)]
Replace every usage of -lvfw32 with what is particularly necessary for
that case:
Avisynth -> -lavifil32
VFW Cap -> -lavicap32
Patch by kemuri <kemuri9 at gmail dot com>

Originally committed as revision 21410 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoUse $vfwcap_indev_extralibs variable instead of duplicating "-lvfw32" in
Ramiro Polla [Sat, 23 Jan 2010 20:38:17 +0000 (20:38 +0000)]
Use $vfwcap_indev_extralibs variable instead of duplicating "-lvfw32" in
capCreateCaptureWindow() check.
Based on patch by kemuri <kemuri9 at gmail dot com>

Originally committed as revision 21409 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoIndent.
Ramiro Polla [Sat, 23 Jan 2010 20:18:38 +0000 (20:18 +0000)]
Indent.

Originally committed as revision 21408 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoMoved reusable functions from dxva2_h264.c to dxva2.c
Laurent Aimar [Sat, 23 Jan 2010 18:17:57 +0000 (18:17 +0000)]
Moved reusable functions from dxva2_h264.c to dxva2.c

Originally committed as revision 21407 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoMove +52 from the loop filter to the alpha/beta offsets in the context.
Michael Niedermayer [Sat, 23 Jan 2010 18:05:30 +0000 (18:05 +0000)]
Move +52 from the loop filter to the alpha/beta offsets in the context.
This should fix a segfault, also it might be faster on systems where the
+52 wasnt free.

Originally committed as revision 21406 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoRelease buffer when the codec is closed.
Reimar Döffinger [Sat, 23 Jan 2010 18:04:09 +0000 (18:04 +0000)]
Release buffer when the codec is closed.

Originally committed as revision 21405 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoUse qstride = 0 and allocate only a single line for qscale_table since
Reimar Döffinger [Sat, 23 Jan 2010 18:01:38 +0000 (18:01 +0000)]
Use qstride = 0 and allocate only a single line for qscale_table since
for this codecs all quants are the same for a single frame.

Originally committed as revision 21404 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoAdded ff_dxva2_ prefix to get_surface(_index) and commit_buffer functions.
Laurent Aimar [Sat, 23 Jan 2010 18:01:28 +0000 (18:01 +0000)]
Added ff_dxva2_ prefix to get_surface(_index) and commit_buffer functions.

Originally committed as revision 21403 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoReindent the content of one if(){} in ff_dxva2_common_end_frame.
Laurent Aimar [Sat, 23 Jan 2010 18:00:51 +0000 (18:00 +0000)]
Reindent the content of one if(){} in ff_dxva2_common_end_frame.

Originally committed as revision 21402 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoMoved reusable code from dxva2_h264.c:end_frame to ff_dxva2_common_end_frame.
Laurent Aimar [Sat, 23 Jan 2010 17:59:08 +0000 (17:59 +0000)]
Moved reusable code from dxva2_h264.c:end_frame to ff_dxva2_common_end_frame.

Originally committed as revision 21401 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoSet edges based on cbp and mv partitioning, not just skiped MBs.
Michael Niedermayer [Sat, 23 Jan 2010 15:28:34 +0000 (15:28 +0000)]
Set edges based on cbp and mv partitioning, not just skiped MBs.
This is faster for videos that have lots of MBs that fall in this category.

Originally committed as revision 21400 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoUse AVPALETTE_COUNT instead of hard-coding 256.
Reimar Döffinger [Sat, 23 Jan 2010 15:22:48 +0000 (15:22 +0000)]
Use AVPALETTE_COUNT instead of hard-coding 256.
This is consistent with other codecs and will also avoid a crash on the
memcpy to data[1] if AVPALETTE_SIZE ever increases.

Originally committed as revision 21399 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoRemove redundant initialization of the palette, it is part of the context
Reimar Döffinger [Sat, 23 Jan 2010 15:19:34 +0000 (15:19 +0000)]
Remove redundant initialization of the palette, it is part of the context
and that one is always 0-initialized already.

Originally committed as revision 21398 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoOptimize filter_mb_mbaff_edge*()
Michael Niedermayer [Sat, 23 Jan 2010 14:50:56 +0000 (14:50 +0000)]
Optimize filter_mb_mbaff_edge*()

Originally committed as revision 21397 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoOptmize 8x8dct check used to skip some borders in the loop filter.
Michael Niedermayer [Sat, 23 Jan 2010 13:54:02 +0000 (13:54 +0000)]
Optmize 8x8dct check used to skip some borders in the loop filter.
4 cpu cycles faster.

Originally committed as revision 21396 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoUse av_freep intead of av_free to free pointers in vp56 context.
Reimar Döffinger [Sat, 23 Jan 2010 13:49:09 +0000 (13:49 +0000)]
Use av_freep intead of av_free to free pointers in vp56 context.

Originally committed as revision 21395 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoZero the frame data on allocation for VB codec, e.g. the FATE sample seems to
Reimar Döffinger [Sat, 23 Jan 2010 10:25:22 +0000 (10:25 +0000)]
Zero the frame data on allocation for VB codec, e.g. the FATE sample seems to
rely on this.

Originally committed as revision 21394 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoMake url_read_complete retry on EAGAIN and return how much data it read
Reimar Döffinger [Sat, 23 Jan 2010 10:23:47 +0000 (10:23 +0000)]
Make url_read_complete retry on EAGAIN and return how much data it read
if it reached EOF, making it useful in more cases.

Originally committed as revision 21393 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoPrint a warning when DATA chunk is encountered in the middle of chunk.
Ronald S. Bultje [Sat, 23 Jan 2010 09:42:52 +0000 (09:42 +0000)]
Print a warning when DATA chunk is encountered in the middle of chunk.
From multirate RM patch by Ronald S. Bultje

Originally committed as revision 21392 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoExport wma_critical_freqs as ff_wma_critical_freqs
Daniel Verkamp [Sat, 23 Jan 2010 08:39:50 +0000 (08:39 +0000)]
Export wma_critical_freqs as ff_wma_critical_freqs

Originally committed as revision 21391 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoFix heap overflow with -async.
Michael Niedermayer [Sat, 23 Jan 2010 01:25:19 +0000 (01:25 +0000)]
Fix heap overflow with -async.
Fixes issue1666

Originally committed as revision 21390 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoThe SSSE3 version of dct_quantize in mpegvideo_mmx_template.c needs
Reimar Döffinger [Fri, 22 Jan 2010 23:40:18 +0000 (23:40 +0000)]
The SSSE3 version of dct_quantize in mpegvideo_mmx_template.c needs
inv_zigzag_direct16 16-byte aligned, so mark it appropriately.
Fixes encoder crashes e.g. with MPlayer's -vf lavc.

Originally committed as revision 21389 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoRemove conditional M_PI definition, it is already defined in
Stefano Sabatini [Fri, 22 Jan 2010 23:16:23 +0000 (23:16 +0000)]
Remove conditional M_PI definition, it is already defined in
libavutil/mathematics.h indirectly included by libavformat/avformat.h.

Originally committed as revision 21388 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoReplace deprecated guess_format() with av_guess_format().
Martin Storsjö [Fri, 22 Jan 2010 23:09:08 +0000 (23:09 +0000)]
Replace deprecated guess_format() with av_guess_format().

Patch by Martin Storsjö reverse(<ts.nitram@nitram>).

Originally committed as revision 21387 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoImplement an sse version of scalarproduct_float().
Alex Converse [Fri, 22 Jan 2010 23:07:58 +0000 (23:07 +0000)]
Implement an sse version of scalarproduct_float().

Originally committed as revision 21386 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoCheck for chunk boundaries when decoding VB codec data
Kostya Shishkov [Fri, 22 Jan 2010 19:07:44 +0000 (19:07 +0000)]
Check for chunk boundaries when decoding VB codec data

Originally committed as revision 21385 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoZero palette in case not all entries are initialized later
Kostya Shishkov [Fri, 22 Jan 2010 18:28:41 +0000 (18:28 +0000)]
Zero palette in case not all entries are initialized later

Originally committed as revision 21384 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoMake style a little bit more K&R'ish.
Ronald S. Bultje [Fri, 22 Jan 2010 16:12:55 +0000 (16:12 +0000)]
Make style a little bit more K&R'ish.

Originally committed as revision 21383 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoRename a function which is no longer ipv6-specific. Patch by Martin Storsjö
Martin Storsjö [Fri, 22 Jan 2010 16:10:08 +0000 (16:10 +0000)]
Rename a function which is no longer ipv6-specific. Patch by Martin Storsjö
<$firstname () $firstname st>.

Originally committed as revision 21382 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoUse the control URI from the SDP (if present) rather than the input filename,
Alan Steremberg [Fri, 22 Jan 2010 16:04:15 +0000 (16:04 +0000)]
Use the control URI from the SDP (if present) rather than the input filename,
if present. This fixes playback of a number of MS-RTSP streams, mostly these
for which playback contains a session key in the URI. Fixes issue 1697.
Patch by Alan Steremberg <$firstname dot $lastname () gmail com>.

Originally committed as revision 21381 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoMove array specifiers outside DECLARE_ALIGNED() invocations
Måns Rullgård [Fri, 22 Jan 2010 03:26:30 +0000 (03:26 +0000)]
Move array specifiers outside DECLARE_ALIGNED() invocations

Originally committed as revision 30385 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

14 years agoSuppress TMS470 warnings about alignment directives
Måns Rullgård [Fri, 22 Jan 2010 03:25:27 +0000 (03:25 +0000)]
Suppress TMS470 warnings about alignment directives

Originally committed as revision 21380 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoDECLARE_ALIGNED macro for TMS470 compiler
Måns Rullgård [Fri, 22 Jan 2010 03:25:25 +0000 (03:25 +0000)]
DECLARE_ALIGNED macro for TMS470 compiler

This compiler supports gcc-style alignment attributes for struct,
but not for global variables.  For the latter, alignment can be
specified with a pragma, which does not work for struct members or
local variables.  By using both pragma and attribute, one or the
other will always take effect.

Unfortunately, no means exists for aligning stack variables.

Originally committed as revision 21379 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoAdd an AV_PRAGMA() macro for constructing _Pragma() directives
Måns Rullgård [Fri, 22 Jan 2010 03:25:21 +0000 (03:25 +0000)]
Add an AV_PRAGMA() macro for constructing _Pragma() directives

The seemingly equivalent _Pragma(AV_STRINGIFY(x)) is not accepted
by some compilers.

Originally committed as revision 21378 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoMove array specifiers outside DECLARE_ALIGNED() invocations
Måns Rullgård [Fri, 22 Jan 2010 03:25:11 +0000 (03:25 +0000)]
Move array specifiers outside DECLARE_ALIGNED() invocations

Originally committed as revision 21377 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoconfigure: fix --cpu=host
Måns Rullgård [Fri, 22 Jan 2010 03:14:19 +0000 (03:14 +0000)]
configure: fix --cpu=host

The awk command used inadvertently relied on non-standard features.

Originally committed as revision 21376 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoGcc idiocy fixes related to filter_mb_edge*.
Michael Niedermayer [Fri, 22 Jan 2010 01:59:17 +0000 (01:59 +0000)]
Gcc idiocy fixes related to filter_mb_edge*.
Change order of operands as gcc uses a hardcoded register per operand it seems
even for static functions
thus reducing unneeded moved (now functions try to pass the same argument in
the same spot).
Change signed int to unsigned int for array indexes as signed requires signed
extension while unsigned is free.
move the +52 up and merge it where it will end as a lea instruction, gcc always
splits the 52 out there turning the free +52 into an expensive one otherwise.
The changed code becomes a little faster.

Originally committed as revision 21375 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoUpdate regression tests after r21372
David Conrad [Thu, 21 Jan 2010 23:35:26 +0000 (23:35 +0000)]
Update regression tests after r21372

Originally committed as revision 21374 to svn://svn.ffmpeg.org/ffmpeg/trunk

14 years agoMove dquant check into qscale overflow check.
Michael Niedermayer [Thu, 21 Jan 2010 21:01:26 +0000 (21:01 +0000)]
Move dquant check into qscale overflow check.
This should be faster (couldnt meassue a difference), and its less picky
on slightly out of spec dquant.

Originally committed as revision 21373 to svn://svn.ffmpeg.org/ffmpeg/trunk