10 years agoFix SPLATB_REG mess. Used to be a if/elseif/elseif/elseif spaghetti, so this
Ronald S. Bultje [Sat, 24 Jul 2010 19:33:05 +0000 (19:33 +0000)]
Fix SPLATB_REG mess. Used to be a if/elseif/elseif/elseif spaghetti, so this
splits it into small optimization-specific macros which are selected for each
DSP function. The advantage of this approach is that the sse4 functions now
use the ssse3 codepath also without needing an explicit sse4 codepath.

Originally committed as revision 24487 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoSimplify lavfi test script a little
Måns Rullgård [Sat, 24 Jul 2010 17:13:35 +0000 (17:13 +0000)]
Simplify lavfi test script a little

Originally committed as revision 24486 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoUnbreak parallel lavfi testing
Måns Rullgård [Sat, 24 Jul 2010 17:13:32 +0000 (17:13 +0000)]
Unbreak parallel lavfi testing

Originally committed as revision 24485 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoaacenc: TLS: Save maximum values for each swb in a table.
Alex Converse [Sat, 24 Jul 2010 17:11:51 +0000 (17:11 +0000)]
aacenc: TLS: Save maximum values for each swb in a table.

This gives an almost 20% speedup.

Originally committed as revision 24484 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years ago10l: Remove some commented out code that slipped in.
Alex Converse [Sat, 24 Jul 2010 17:01:50 +0000 (17:01 +0000)]
10l: Remove some commented out code that slipped in.

Originally committed as revision 24483 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoaacenc: TLS: Try to preserve some energy in each non-zero band.
Alex Converse [Sat, 24 Jul 2010 16:36:01 +0000 (16:36 +0000)]
aacenc: TLS: Try to preserve some energy in each non-zero band.

Reduce scalefactors in non-zero bands that underflow by twice as much as those
in bands that just fail to hit psy targets.

Originally committed as revision 24482 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoFix compilation, forgot to add const also to the
Reimar Döffinger [Sat, 24 Jul 2010 14:43:55 +0000 (14:43 +0000)]
Fix compilation, forgot to add const also to the
definition of ff_yuv2packedX_altivec

Originally committed as revision 31782 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

10 years agoUse "const" qualifier for pointers that point to input data of
Reimar Döffinger [Sat, 24 Jul 2010 13:59:49 +0000 (13:59 +0000)]
Use "const" qualifier for pointers that point to input data of
audio encoders.
This is purely for clarity/documentation purposes.

Originally committed as revision 24481 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoFix a "function declaration isn't a prototype" warning.
Reimar Döffinger [Sat, 24 Jul 2010 12:58:28 +0000 (12:58 +0000)]
Fix a "function declaration isn't a prototype" warning.

Originally committed as revision 24480 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoFix use of uninitialized data: A array element can not be
Reimar Döffinger [Sat, 24 Jul 2010 11:55:22 +0000 (11:55 +0000)]
Fix use of uninitialized data: A array element can not be
used to initialized another array element of the same array.

Originally committed as revision 24479 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoMore const-correctness fixes to avoid warnings.
Reimar Döffinger [Sat, 24 Jul 2010 11:37:58 +0000 (11:37 +0000)]
More const-correctness fixes to avoid warnings.

Originally committed as revision 31781 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

10 years agoAdd some "const" to avoid incompatible pointer type warnings
Reimar Döffinger [Sat, 24 Jul 2010 11:33:28 +0000 (11:33 +0000)]
Add some "const" to avoid incompatible pointer type warnings

Originally committed as revision 31780 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

10 years agoAdd missing comma, fix rgb2yuv_table[SWS_CS_SMPTE240M] coefficients.
Stefano Sabatini [Sat, 24 Jul 2010 11:19:39 +0000 (11:19 +0000)]
Add missing comma, fix rgb2yuv_table[SWS_CS_SMPTE240M] coefficients.

Originally committed as revision 31779 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

10 years agoAdd some comments to the rgb2yuv_table, clarifying the standard where
Stefano Sabatini [Sat, 24 Jul 2010 11:19:35 +0000 (11:19 +0000)]
Add some comments to the rgb2yuv_table, clarifying the standard where
are defined.

Originally committed as revision 31778 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

10 years agoNever shrink the ByteIOContext buffer in ff_rewind_with_probe_data
Martin Storsjö [Sat, 24 Jul 2010 09:04:44 +0000 (09:04 +0000)]
Never shrink the ByteIOContext buffer in ff_rewind_with_probe_data

If there is little unread data in the ByteIOContext buffer, this may lead
to reducing the size of the ByteIOContext buffer to little more the probe
data size. This can lead to suboptimal aviobuf behaviour, e.g. making some
demuxers fail to do short seeks backwards (if the input isn't seekable).

Originally committed as revision 24478 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agompegts: Warn if unable to seek back
Martin Storsjö [Sat, 24 Jul 2010 08:43:52 +0000 (08:43 +0000)]
mpegts: Warn if unable to seek back

Originally committed as revision 24477 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoaacenc: Prevent premature termination of the two loop search.
Alex Converse [Sat, 24 Jul 2010 04:23:26 +0000 (04:23 +0000)]
aacenc: Prevent premature termination of the two loop search.

Originally committed as revision 24476 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoaacdec: Only load and write each predictor variable once.
Alex Converse [Sat, 24 Jul 2010 02:57:08 +0000 (02:57 +0000)]
aacdec: Only load and write each predictor variable once.

This is slightly faster and opens the door for further optimization.

Originally committed as revision 24475 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoaacdec: 4% faster main profile decoding.
Alex Converse [Sat, 24 Jul 2010 02:41:47 +0000 (02:41 +0000)]
aacdec: 4% faster main profile decoding.

Originally committed as revision 24474 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoaacenc: Favor log2f() and sqrtf() over log2() and sqrt().
Alex Converse [Sat, 24 Jul 2010 02:10:59 +0000 (02:10 +0000)]
aacenc: Favor log2f() and sqrtf() over log2() and sqrt().

Originally committed as revision 24473 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoaacenc: Factorize some scalefactor utilities.
Alex Converse [Fri, 23 Jul 2010 22:37:42 +0000 (22:37 +0000)]
aacenc: Factorize some scalefactor utilities.

Originally committed as revision 24472 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoInline asm for VP56 arith coder
Eli Friedman [Fri, 23 Jul 2010 21:46:30 +0000 (21:46 +0000)]
Inline asm for VP56 arith coder

This is a lot more reliable to get cmov rather than trying to trick gcc into
generating it, useful since it's 2% faster overall.

Patch by Eli Friedman <eli.friedman at gmail>

Originally committed as revision 24471 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: Inline traversing vp8_small_mvtree
David Conrad [Fri, 23 Jul 2010 21:46:25 +0000 (21:46 +0000)]
VP8: Inline traversing vp8_small_mvtree

Much faster read_mv_component, slightly faster overall

Originally committed as revision 24470 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: Use vp56_rac_get_prob_branchy when the bit is only used by an if()
David Conrad [Fri, 23 Jul 2010 21:46:20 +0000 (21:46 +0000)]
VP8: Use vp56_rac_get_prob_branchy when the bit is only used by an if()

Originally committed as revision 24469 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoDecode DCT tokens by branching to a different code path for each branch
David Conrad [Fri, 23 Jul 2010 21:46:17 +0000 (21:46 +0000)]
Decode DCT tokens by branching to a different code path for each branch
on the huffman tree, instead of traversing the tree in a while loop.

Based on the similar optimization in libvpx's detokenize.c

10% faster at normal bitrates, and 30% faster for high-bitrate intra-only

Originally committed as revision 24468 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoMove renormalization of the VP56 arith decoder to before decoding a bit
David Conrad [Fri, 23 Jul 2010 21:46:14 +0000 (21:46 +0000)]
Move renormalization of the VP56 arith decoder to before decoding a bit

No difference at the moment, but allows a future branchy variant
of vp56_rac_get_prob to be significantly faster

Originally committed as revision 24467 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoSplit renorm of vp56 arith decoder to its own function
David Conrad [Fri, 23 Jul 2010 21:46:08 +0000 (21:46 +0000)]
Split renorm of vp56 arith decoder to its own function

Originally committed as revision 24466 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agovp56's arith decoder's code_word is only 16 bits, no need for unsigned long
David Conrad [Fri, 23 Jul 2010 21:46:01 +0000 (21:46 +0000)]
vp56's arith decoder's code_word is only 16 bits, no need for unsigned long

Originally committed as revision 24465 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoAdd myself to VP8 copyright and maintainers.
Jason Garrett-Glaser [Fri, 23 Jul 2010 21:42:35 +0000 (21:42 +0000)]
Add myself to VP8 copyright and maintainers.
Also add Ronald to maintainers.

Originally committed as revision 24464 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: always_inline some things to force gcc to do the right thing
Jason Garrett-Glaser [Fri, 23 Jul 2010 21:36:21 +0000 (21:36 +0000)]
VP8: always_inline some things to force gcc to do the right thing
Mostly seems to help in the MC code, which gets a hundred cycles faster.

Originally committed as revision 24463 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: use AV_RL24 instead of defining a new RL24.
Jason Garrett-Glaser [Fri, 23 Jul 2010 21:17:18 +0000 (21:17 +0000)]
VP8: use AV_RL24 instead of defining a new RL24.

Originally committed as revision 24462 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoswscale-test: scale from reference to source only once for each algorithm
Ramiro Polla [Fri, 23 Jul 2010 20:27:54 +0000 (20:27 +0000)]
swscale-test: scale from reference to source only once for each algorithm

Originally committed as revision 31777 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

10 years agoindent
Ramiro Polla [Fri, 23 Jul 2010 20:26:54 +0000 (20:26 +0000)]

Originally committed as revision 31776 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

10 years agoswscale-test: change order tests are run, maintaining the same algorithm for
Ramiro Polla [Fri, 23 Jul 2010 20:26:32 +0000 (20:26 +0000)]
swscale-test: change order tests are run, maintaining the same algorithm for
sequential geometries instead of running all algorithms sequentially for each

Originally committed as revision 31775 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

10 years agoVP8: Slightly faster MV selection
Jason Garrett-Glaser [Fri, 23 Jul 2010 19:06:22 +0000 (19:06 +0000)]
VP8: Slightly faster MV selection
Don't clamp best mv unless it's actually used.

Originally committed as revision 24461 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: use AV_ZERO32 instead of AV_WN32A where relevant
Jason Garrett-Glaser [Fri, 23 Jul 2010 10:42:19 +0000 (10:42 +0000)]
VP8: use AV_ZERO32 instead of AV_WN32A where relevant

Originally committed as revision 24460 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: eliminate redundant code in r24458
Jason Garrett-Glaser [Fri, 23 Jul 2010 10:34:21 +0000 (10:34 +0000)]
VP8: eliminate redundant code in r24458

Originally committed as revision 24459 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: shave a few clocks off check_intra_pred_mode
Jason Garrett-Glaser [Fri, 23 Jul 2010 10:24:38 +0000 (10:24 +0000)]
VP8: shave a few clocks off check_intra_pred_mode

Originally committed as revision 24458 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoAdd VP8 sign bias code regtest. Not tested by the official conformance suite.
Vitor Sessak [Fri, 23 Jul 2010 10:09:46 +0000 (10:09 +0000)]
Add VP8 sign bias code regtest. Not tested by the official conformance suite.

Originally committed as revision 24457 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: fix broken sign bias code in MV pred
Jason Garrett-Glaser [Fri, 23 Jul 2010 06:41:35 +0000 (06:41 +0000)]
VP8: fix broken sign bias code in MV pred
Apparently the official conformance test vectors don't test this feature,
even though libvpx uses it.

Originally committed as revision 24456 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: optimize DC-only chroma case in the same way as luma.
Jason Garrett-Glaser [Fri, 23 Jul 2010 06:02:52 +0000 (06:02 +0000)]
VP8: optimize DC-only chroma case in the same way as luma.
Add MMX idct_dc_add4uv function for this case.
~40% faster chroma idct.

Originally committed as revision 24455 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: Clean up some variable shadowing.
Jason Garrett-Glaser [Fri, 23 Jul 2010 03:44:37 +0000 (03:44 +0000)]
VP8: Clean up some variable shadowing.

Originally committed as revision 24454 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8 asm: cosmetics (spacing)
Jason Garrett-Glaser [Fri, 23 Jul 2010 03:02:56 +0000 (03:02 +0000)]
VP8 asm: cosmetics (spacing)

Originally committed as revision 24453 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: 30% faster idct_mb
Jason Garrett-Glaser [Fri, 23 Jul 2010 02:58:27 +0000 (02:58 +0000)]
VP8: 30% faster idct_mb
Take shortcuts based on statistically common situations.
Add 4-at-a-time idct_dc function (mmx and sse2) since rows of 4 DC-only DCT
blocks are common.
TODO: tie this more directly into the MB mode, since the DC-level transform is
only used for non-splitmv blocks?

Originally committed as revision 24452 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: smarter prefetching
Jason Garrett-Glaser [Fri, 23 Jul 2010 01:59:56 +0000 (01:59 +0000)]
VP8: smarter prefetching
Don't prefetch reference frames that were used less than 1/32th of the time so
far in the frame.
This helps speed up to ~2% on videos that, in many frames, make near-zero
(but not entirely zero) use of golden and/or alt-refs.
This is a very common property of videos encoded by libvpx.

Originally committed as revision 24451 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoIn h264 parser, return immediately if buf_size is 0, avoid printing
Baptiste Coudurier [Fri, 23 Jul 2010 00:34:09 +0000 (00:34 +0000)]
In h264 parser, return immediately if buf_size is 0, avoid printing
erroneous message for last frame.

Originally committed as revision 24450 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoIn mpegts muxer, fix rbsp trailing bits in AUD nal, fixes issue #2122
Baptiste Coudurier [Fri, 23 Jul 2010 00:27:17 +0000 (00:27 +0000)]
In mpegts muxer, fix rbsp trailing bits in AUD nal, fixes issue #2122

Originally committed as revision 24449 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: clear DCT blocks in iDCT instead of using clear_blocks.
Jason Garrett-Glaser [Fri, 23 Jul 2010 00:07:16 +0000 (00:07 +0000)]
VP8: clear DCT blocks in iDCT instead of using clear_blocks.
~0.3% faster overall.

Originally committed as revision 24448 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: avoid a memset for non-i4x4 blocks with no coefficients
Jason Garrett-Glaser [Fri, 23 Jul 2010 00:05:44 +0000 (00:05 +0000)]
VP8: avoid a memset for non-i4x4 blocks with no coefficients

Originally committed as revision 24447 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoGet rid of more unnecessary dereferences in VP8 deblocking
Jason Garrett-Glaser [Thu, 22 Jul 2010 23:11:40 +0000 (23:11 +0000)]
Get rid of more unnecessary dereferences in VP8 deblocking

Originally committed as revision 24446 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoShut up an uninitialized variable GCC warning in VP8.
Jason Garrett-Glaser [Thu, 22 Jul 2010 23:04:51 +0000 (23:04 +0000)]
Shut up an uninitialized variable GCC warning in VP8.

Originally committed as revision 24445 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoSmarter VP8 prefetching
Jason Garrett-Glaser [Thu, 22 Jul 2010 23:03:08 +0000 (23:03 +0000)]
Smarter VP8 prefetching
Prefetch all refs (including altref), but only if they've been used so far this
~2.5% faster overall.

TODO: Do something even smarter, like using how often each ref has been used
so far, so that a couple blocks of a rarely-used ref don't force us to prefetch

Originally committed as revision 24444 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoPrefer impersonal form over third person, for consistency with the
Stefano Sabatini [Thu, 22 Jul 2010 23:01:31 +0000 (23:01 +0000)]
Prefer impersonal form over third person, for consistency with the
rest of FFmpeg.

Originally committed as revision 31772 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

10 years agoFix stupid bug in VP8 prefetching code
Jason Garrett-Glaser [Thu, 22 Jul 2010 22:15:43 +0000 (22:15 +0000)]
Fix stupid bug in VP8 prefetching code

Originally committed as revision 24443 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoAdd an APIchanges entry after the M_PHI addition.
Stefano Sabatini [Thu, 22 Jul 2010 22:11:45 +0000 (22:11 +0000)]
Add an APIchanges entry after the M_PHI addition.

Originally committed as revision 24442 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoEliminate a LUT in escape decoding in VP8 decode_block_coeffs
Jason Garrett-Glaser [Thu, 22 Jul 2010 22:08:09 +0000 (22:08 +0000)]
Eliminate a LUT in escape decoding in VP8 decode_block_coeffs

Originally committed as revision 24441 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoRemove dependency on avcodec_get_chroma_sub_sample() and libavcodec,
Stefano Sabatini [Thu, 22 Jul 2010 22:06:46 +0000 (22:06 +0000)]
Remove dependency on avcodec_get_chroma_sub_sample() and libavcodec,
use the pixdesc API instead.

Originally committed as revision 24440 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoAdd the M_PHI constant, contains an approximation of the golden ratio
Stefano Sabatini [Thu, 22 Jul 2010 22:05:21 +0000 (22:05 +0000)]
Add the M_PHI constant, contains an approximation of the golden ratio
irrational number.

Originally committed as revision 24439 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoEliminate some repeated dereferences in VP8 inter_predict
Jason Garrett-Glaser [Thu, 22 Jul 2010 21:05:30 +0000 (21:05 +0000)]
Eliminate some repeated dereferences in VP8 inter_predict

Originally committed as revision 24438 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoUse pextrw for SSE4 mbedge filter result writing, speedup 5-10cycles on
Ronald S. Bultje [Thu, 22 Jul 2010 19:59:34 +0000 (19:59 +0000)]
Use pextrw for SSE4 mbedge filter result writing, speedup 5-10cycles on
CPUs supporting it.

Originally committed as revision 24437 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoswscale-test: use only 3 characters to print width/height since they're always
Ramiro Polla [Thu, 22 Jul 2010 19:05:10 +0000 (19:05 +0000)]
swscale-test: use only 3 characters to print width/height since they're always
smaller than 1000

Originally committed as revision 31771 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

10 years agoswscale-test: print test name before running test, so that it's available in
Ramiro Polla [Thu, 22 Jul 2010 19:02:49 +0000 (19:02 +0000)]
swscale-test: print test name before running test, so that it's available in
the output even if the test crashes.

Originally committed as revision 31770 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

10 years agoDo not automatically run configure
Måns Rullgård [Thu, 22 Jul 2010 17:47:56 +0000 (17:47 +0000)]
Do not automatically run configure

Making automatic reconfigure work properly in all cases proved more
difficult than it seemed, and people complained about it.  Instead,
print a warning when the configuration appears to be out of date.

Originally committed as revision 24436 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoswscale-test: allocate more memory to prevent scalers from writing out of bounds
Ramiro Polla [Thu, 22 Jul 2010 15:30:22 +0000 (15:30 +0000)]
swscale-test: allocate more memory to prevent scalers from writing out of bounds

Some converters (ie. unscaled rgb24 -> argb) may write some bytes out of
bounds. Ideally the converters should be fixed, but in the meantime we allocate
more memory to prevent heap corruption.

Originally committed as revision 31768 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

10 years agoFactorize indent definition in PRINT_LIB_INFO().
Stefano Sabatini [Thu, 22 Jul 2010 13:05:10 +0000 (13:05 +0000)]
Factorize indent definition in PRINT_LIB_INFO().

Originally committed as revision 24435 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoMerge maybe_print_config() and PRINT_LIB_CONFIG() in PRINT_LIB_INFO().
Stefano Sabatini [Thu, 22 Jul 2010 13:05:07 +0000 (13:05 +0000)]
Merge maybe_print_config() and PRINT_LIB_CONFIG() in PRINT_LIB_INFO().

Originally committed as revision 24434 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoRename:
Stefano Sabatini [Thu, 22 Jul 2010 13:05:04 +0000 (13:05 +0000)]
print_all_lib_versions() -> print_all_libs_info()

Originally committed as revision 24433 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoReindent.
Stefano Sabatini [Thu, 22 Jul 2010 13:05:01 +0000 (13:05 +0000)]

Originally committed as revision 24432 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoIncrease flexibility of PRINT_LIB_VERSION(), make it accept a flags
Stefano Sabatini [Thu, 22 Jul 2010 13:04:58 +0000 (13:04 +0000)]
Increase flexibility of PRINT_LIB_VERSION(), make it accept a flags
parameter. Allow a pending factorization.

Originally committed as revision 24431 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoMap settings for 2-pass libvpx encoding.
James Zern [Thu, 22 Jul 2010 12:35:32 +0000 (12:35 +0000)]
Map settings for 2-pass libvpx encoding.

Patch by James Zern, jzern at google

Originally committed as revision 24430 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoEliminate a pointless memset for intra blocks in P-frames in VP8
Jason Garrett-Glaser [Thu, 22 Jul 2010 12:15:29 +0000 (12:15 +0000)]
Eliminate a pointless memset for intra blocks in P-frames in VP8

Originally committed as revision 24429 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVP8: Don't store segment in macroblock struct anymore.
Jason Garrett-Glaser [Thu, 22 Jul 2010 11:55:55 +0000 (11:55 +0000)]
VP8: Don't store segment in macroblock struct anymore.
Not necessary with the previous patch.

Originally committed as revision 24427 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoConvert VP8 macroblock structures to a ring buffer.
Jason Garrett-Glaser [Thu, 22 Jul 2010 11:45:18 +0000 (11:45 +0000)]
Convert VP8 macroblock structures to a ring buffer.
Uses a slightly nonintuitive ring buffer size of (width+height*2) to simplify
addressing logic.
Also split out the segmentation map to a separate structure, necessary to
implement the ring buffer.

Originally committed as revision 24426 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoAdd APIchanges entry for the libavfilter media format generalization
Stefano Sabatini [Thu, 22 Jul 2010 11:24:31 +0000 (11:24 +0000)]
Add APIchanges entry for the libavfilter media format generalization
change of r24424.

Originally committed as revision 24425 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoGeneralize pixel format enum fields to int formats.
S.N. Hemanth Meenakshisundaram [Thu, 22 Jul 2010 11:12:47 +0000 (11:12 +0000)]
Generalize pixel format enum fields to int formats.

This is needed to make the libavfilter framework work with audio

In particular add a type field to AVFilterLink, change the field types:
enum PixelFormat format   -> int format   in AVFilterBuffer
enum PixelFormat *formats -> int *formats in AVFilterFormats
enum PixelFormat *format  -> int format   in AVFilterLink

and change the function signatures:
AVFilterFormats *avfilter_make_format_list(const enum PixelFormat *pix_fmts); ->
AVFilterFormats *avfilter_make_format_list(const int *fmts);

int avfilter_add_colorspace(AVFilterFormats **avff, enum PixelFormat pix_fmt); ->
int avfilter_add_format(AVFilterFormats **avff, int fmt);

AVFilterFormats *avfilter_all_colorspaces(void); ->
AVFilterFormats *avfilter_all_formats(enum AVMediaType type);

This change breaks libavfilter API/ABI.

Patch by S.N. Hemanth Meenakshisundaram |smeenaks|ucsd|edu|.

Originally committed as revision 24424 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoDocument create_filter().
Stefano Sabatini [Thu, 22 Jul 2010 09:56:09 +0000 (09:56 +0000)]
Document create_filter().

Originally committed as revision 24423 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoAdd simple doxy to link_filter().
Stefano Sabatini [Thu, 22 Jul 2010 09:56:06 +0000 (09:56 +0000)]
Add simple doxy to link_filter().

Originally committed as revision 24422 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoFix documentation for parse_link_name().
Stefano Sabatini [Thu, 22 Jul 2010 09:56:03 +0000 (09:56 +0000)]
Fix documentation for parse_link_name().

Originally committed as revision 24421 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoConsistently use Uppercase for the first character of the log
Stefano Sabatini [Thu, 22 Jul 2010 09:56:00 +0000 (09:56 +0000)]
Consistently use Uppercase for the first character of the log

Originally committed as revision 24420 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoMake link_filter() propagate the generated error code.
Stefano Sabatini [Thu, 22 Jul 2010 09:55:57 +0000 (09:55 +0000)]
Make link_filter() propagate the generated error code.

Originally committed as revision 24419 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoFix style: "if(" -> "if (".
Stefano Sabatini [Thu, 22 Jul 2010 09:49:15 +0000 (09:49 +0000)]
Fix style: "if(" -> "if (".

Originally committed as revision 24418 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoCalculate deblock strength per-MB instead of per-row
Jason Garrett-Glaser [Thu, 22 Jul 2010 07:24:22 +0000 (07:24 +0000)]
Calculate deblock strength per-MB instead of per-row
Gives better cache locality, since the VP8Macroblock structs are still in cache.
Inspired by the way x264 does it.

Originally committed as revision 24417 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoAvoid tracking i4x4 modes in P-frames in VP8
Jason Garrett-Glaser [Thu, 22 Jul 2010 07:04:45 +0000 (07:04 +0000)]
Avoid tracking i4x4 modes in P-frames in VP8
As in the previous commit, they aren't used for context selection, so it saves
memory this way.

Originally committed as revision 24416 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoAvoid useless fill_rectangle in P-frames in VP8
Jason Garrett-Glaser [Thu, 22 Jul 2010 06:39:54 +0000 (06:39 +0000)]
Avoid useless fill_rectangle in P-frames in VP8
In VP8, i4x4 only uses contexts based on neighbors in I-frames.

Originally committed as revision 24415 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoOptimize partition mv decoding in VP8
Jason Garrett-Glaser [Thu, 22 Jul 2010 06:29:26 +0000 (06:29 +0000)]
Optimize partition mv decoding in VP8

Originally committed as revision 24414 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoTake shortcuts for mv0 case in VP8 MC
Jason Garrett-Glaser [Thu, 22 Jul 2010 05:49:09 +0000 (05:49 +0000)]
Take shortcuts for mv0 case in VP8 MC
Avoid edge emulation -- it isn't needed if there isn't any subpel.

Originally committed as revision 24413 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoMuch faster VP8 mv and mode prediction
Jason Garrett-Glaser [Thu, 22 Jul 2010 04:26:41 +0000 (04:26 +0000)]
Much faster VP8 mv and mode prediction

Originally committed as revision 24412 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoConvert vp56_mv to 16-bit.
Jason Garrett-Glaser [Thu, 22 Jul 2010 03:33:29 +0000 (03:33 +0000)]
Convert vp56_mv to 16-bit.
Saves nothing except a bit of memory/cache now, but will allow future

Originally committed as revision 24411 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoAdd prefetching to VP8 decoder
Jason Garrett-Glaser [Thu, 22 Jul 2010 03:09:10 +0000 (03:09 +0000)]
Add prefetching to VP8 decoder
~5% faster overall, probably depends on CPU and resolution.

Originally committed as revision 24410 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoFix and enable horizontal >=SSE2 mbedge loopfilter.
Ronald S. Bultje [Thu, 22 Jul 2010 01:35:26 +0000 (01:35 +0000)]
Fix and enable horizontal >=SSE2 mbedge loopfilter.

Originally committed as revision 24409 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agorelicense h264 deblock sse2 to lgpl
Loren Merritt [Thu, 22 Jul 2010 00:39:49 +0000 (00:39 +0000)]
relicense h264 deblock sse2 to lgpl

Originally committed as revision 24408 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoAdd APIchanges entry after libavcore addition.
Stefano Sabatini [Wed, 21 Jul 2010 22:59:28 +0000 (22:59 +0000)]
Add APIchanges entry after libavcore addition.

Originally committed as revision 24407 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agosync yasm macros from x264
Loren Merritt [Wed, 21 Jul 2010 22:45:16 +0000 (22:45 +0000)]
sync yasm macros from x264

Originally committed as revision 24406 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoEliminate one instruction in VP8 dc_add_sse4
Jason Garrett-Glaser [Wed, 21 Jul 2010 22:41:37 +0000 (22:41 +0000)]
Eliminate one instruction in VP8 dc_add_sse4

Originally committed as revision 24405 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoAdd WMAPro regtests
Vitor Sessak [Wed, 21 Jul 2010 22:33:06 +0000 (22:33 +0000)]
Add WMAPro regtests

Originally committed as revision 24404 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoVarious VP8 x86 deblocking speedups
Jason Garrett-Glaser [Wed, 21 Jul 2010 22:11:03 +0000 (22:11 +0000)]
Various VP8 x86 deblocking speedups
SSSE3 versions, improve SSE2 versions a bit.
SSE2/SSSE3 mbedge h functions are currently broken, so explicitly disable them.

Originally committed as revision 24403 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agouse return value of ff_get_line()
Aurelien Jacobs [Wed, 21 Jul 2010 21:41:46 +0000 (21:41 +0000)]
use return value of ff_get_line()

Originally committed as revision 24402 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agodocument ff_get_line()
Aurelien Jacobs [Wed, 21 Jul 2010 21:40:50 +0000 (21:40 +0000)]
document ff_get_line()

Originally committed as revision 24401 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agoimprove ff_get_line to return line length
Aurelien Jacobs [Wed, 21 Jul 2010 21:40:10 +0000 (21:40 +0000)]
improve ff_get_line to return line length

Originally committed as revision 24400 to svn://svn.ffmpeg.org/ffmpeg/trunk

10 years agomove ff_get_line to aviobuf.c
Aurelien Jacobs [Wed, 21 Jul 2010 21:39:01 +0000 (21:39 +0000)]
move ff_get_line to aviobuf.c

Originally committed as revision 24399 to svn://svn.ffmpeg.org/ffmpeg/trunk