* commit '9de9b828ef005dec37052548c195a6b4f18fc701':
lavc: don't overwrite display dimensions with coded dimensions.
lavc: extend / update the [coded_]{width,height} doxy
Conflicts:
libavcodec/avcodec.h
libavcodec/utils.c
The change to the w/h handling is not merged as it breaks lowres
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit '8a776ad90e00ab2b98e8683ac6182d641a383c3a':
h261: Move shared data tables from a header to a proper C file
Conflicts:
libavcodec/Makefile
libavcodec/h261data.c
libavcodec/h261dec.c
libavcodec/h261enc.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit '85deb51a01f1ecc5ac5faa52ad8ea141c384e23a':
h264: Only initialize dsputil if error resilience is enabled
Conflicts:
libavcodec/h264.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit 'e8cafd2773bc56455c8816593cbd9368f2d69a80':
h264: Clear the mb members via memset instead of using dsputil
Conflicts:
libavcodec/h264.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit '6d25c9db11e87ec58652de6588448225137b0c57':
dsputil: Make square put/avg_pixels functions local to h264qpel
Conflicts:
libavcodec/dsputil.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
It is only used for error resilience. This allows building the
h264 decoder without dsputil, if error resilience is disabled.
Signed-off-by: Martin Storsjö <martin@martin.st>
The non-intra-pcm branch in hl_decode_mb (simple, 8bpp) goes from 700
to 672 cycles, and the complete loop of decode_mb_cabac and hl_decode_mb
(in the decode_slice loop) goes from 1759 to 1733 cycles on the clip
tested (cathedral), i.e. almost 30 cycles per mb faster.
Signed-off-by: Martin Storsjö <martin@martin.st>
Put a copy of the 8bit functions only in dsputil, where they are
used for some other things (e.g. mpeg4qpel, mspel, cavsqpel).
Signed-off-by: Martin Storsjö <martin@martin.st>
The EC code does not support fields currently thus it makes no
sense to wait for these cases (which also the check doesnt handle
correctly)
Fixes Ticket 2454
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
The pointers that get assigned ff_cropTbl were made const in
9e0f14f1, but other variables that transitively are assigned
based on these variables were missed.
Signed-off-by: Martin Storsjö <martin@martin.st>
* qatar/master:
lavc: Move ff_cropTbl and ff_zigzag_direct from dsputil to mathtables
Conflicts:
libavcodec/mathtables.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit '610b18e2e3d8ef5eca3e78f33a0625689b8d2bb9':
x86: qpel: Move fullpel and l2 functions to a separate file
bfin: Make vp3 functions static
Conflicts:
libavcodec/bfin/vp3_bfin.c
libavcodec/x86/Makefile
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit 'aa8d89536d35af0a0c8d8bac2b452ffe7b82cae5':
bfin: Don't use the vp3 idct functions if bitexact behaviour is expected
Conflicts:
libavcodec/bfin/vp3_bfin.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
These are widely used throughout libavcodec, nothing dsputil-specific.
Change ff_cropTbl to a statically initialized table, to avoid
initializing it with a function call.
Signed-off-by: Martin Storsjö <martin@martin.st>
This way, they can be shared between mpeg4qpel and h264qpel without
requiring either one to be compiled unconditionally.
Signed-off-by: Martin Storsjö <martin@martin.st>
This makes the vp3 decoder less dependent on dsputil, and will aid
in making it (eventually) dsputil-independent.
Signed-off-by: Martin Storsjö <martin@martin.st>
In the non-bitexact mode, vp3 currently decodes to the same
frame crcs as before 28f9ab702 (and the output visually looks
correct).
Signed-off-by: Martin Storsjö <martin@martin.st>
From 312 to 89/68 (sse/sse2) cycles on Arrandale and Win64.
Sandybridge: 68/47 cycles.
Having a loop counter is a 7 cycle gain.
Unrolling is another 7 cycle gain.
Working in reverse scan is another 6 cycles.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Text subtitles packets are not 0-terminated (and if they are,
it is handled by the recoding process since 0 is a valid
Unicode code point). The terminating 0 would overwrite the
last payload octet.
OTOH, packets must be 0-padded.
Fix a problem reported in trac ticket #2431.
This patch can be controversial, by assuming floats are IEEE-754 and
particular behaviour of the FPU will get in the way.
Timing on Arrandale and Win32 (thus, x87 FPU is used in the reference).
sbr_qmf_pre_shuffle_c: 115 to 76
sbr_neg_odd_64_c: 84 to 55
sbr_qmf_post_shuffle_c: 112 to 83
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Timing on Arrandale:
C SSE
Win32: 57 44
Win64: 47 38
Unrolling and not storing mask both save some cycles.
Signed-off-by: Diego Biurrun <diego@biurrun.de>
