Ben Avison [Thu, 10 Jul 2014 23:12:31 +0000 (00:12 +0100)]
armv6: Accelerate ff_imdct_half for general case (mdct_bits != 6)
The previous implementation targeted DTS Coherent Acoustics, which only
requires mdct_bits == 6. This relatively small size lent itself to
unrolling the loops a small number of times, and encoding offsets
calculated at assembly time within the load/store instructions of each
iteration.
In the more general case (codecs such as AAC and AC3) much larger arrays
are used - mdct_bits == [8, 9, 11]. The old method does not scale for
these cases, so more integer registers are used with non-unrolled versions
of the loops (and with some stack spillage). The postrotation filter loop
is still unrolled by a factor of 2 to permit the double-buffering of some
VFP registers to facilitate overlap of neighbouring iterations.
I benchmarked the result by measuring the number of gperftools samples
that hit anywhere in the AAC decoder (starting from aac_decode_frame())
or specifically in ff_imdct_half_c / ff_imdct_half_vfp, for the same
example AAC stream:
Before After
Mean StdDev Mean StdDev Confidence Change
aac_decode_frame 2368.1 35.8 2117.2 35.3 100.0% +11.8%
ff_imdct_half_* 457.5 22.4 251.2 16.2 100.0% +82.1%
avcodec/me_cmp: restore author attribution and copyrights
These where removed by libav in
See: git show -C 2d60444331fca1910510038dd3817bea885c2367
diff --git a/libavcodec/dsputil.c b/libavcodec/me_cmp.c
similarity index 98%
rename from libavcodec/dsputil.c
rename to libavcodec/me_cmp.c
index ba71a99..9fcc937 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/me_cmp.c
@@ -1,8 +1,4 @@
/*
- * DSP utils
- * Copyright (c) 2000, 2001 Fabrice Bellard
- * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
- *
* This file is part of Libav.
*
* Libav is free software; you can redistribute it and/or
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Nicolas George [Tue, 15 Jul 2014 17:26:52 +0000 (19:26 +0200)]
lavfi: check refcount before merging.
When merging the formats around the automatically inserted
convert filters, the refcount of the format lists can not be 0.
Coverity does not detect it, and suspects a memory leak,
because if refcount is 0 the newly allocated lists are not
stored anywhere. That gives CIDs 1224282, 1224283 and 1224284.
Lists with refcount 0 are used in can_merge_formats(), so the
asserts can not be moved inside the merge functions.
Nicolas George [Tue, 15 Jul 2014 14:04:49 +0000 (16:04 +0200)]
lavd/x11grab: check 32-bits color masks.
The X11 servers by VNC, at 32-bits depths, has the following masks:
R:0x000007ff G:0x003ff800 B:0xffc00000
This is not compatible with AV_PIX_FMT_0RGB32, and the result
is success with completely wrong colors.
lavf/segment: set segment end time when the first packet arrives
Avoid negative durations in case there is a single packet in the current
segment, since in that case the end time is still set to the previous
segment end time.
Simon Thelen [Wed, 9 Jul 2014 19:40:43 +0000 (21:40 +0200)]
libavformat/segment: change segment_list_size behavior to match hls_list_size behavior.
Make the segment muxer keep segment_list_size segments instead of
segment_list_size + 1 segments. This patch also changes the
documentation for segment_list_size to reduce possible confusion over
how many segments are kept.
this allows the segment list to
be limited to containing only one segment which used to be impossible
because a segment_list_size of 0 kept all the segments and a
segment_list_size of 1 kept 2 segments.
Signed-off-by: Simon Thelen <ffmpeg-dev@c-14.de> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Ben Avison [Thu, 10 Jul 2014 23:14:31 +0000 (00:14 +0100)]
armv6: Accelerate butterflies_float
I benchmarked the result by measuring the number of gperftools samples that
hit anywhere in the AAC decoder (starting from aac_decode_frame()) or
specifically in butterflies_float_c() / ff_butterflies_float_vfp() for the
same sample AAC stream:
Before After
Mean StdDev Mean StdDev Confidence Change
Audio decode 1542.8 43.7 1470.5 41.5 100.0% +4.9%
butterflies_float 130.0 11.9 70.2 12.1 100.0% +85.2%
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Ben Avison [Thu, 10 Jul 2014 23:14:30 +0000 (00:14 +0100)]
armv6: Accelerate vector_fmul_window
I benchmarked the result by measuring the number of gperftools samples that
hit anywhere in the AAC decoder (starting from aac_decode_frame()) or
specifically in vector_fmul_window_c() / ff_vector_fmul_window_vfp() for the
same sample AAC stream:
Before After
Mean StdDev Mean StdDev Confidence Change
Audio decode 1598.2 47.4 1529.2 25.4 100.0% +4.5%
vector_fmul_window 244.0 22.1 188.9 22.3 100.0% +29.2%
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
It works by exporting necessary fields as metadata tags in matroskadec
and use those values to write the appropriate XML fields as per the WebM
DASH Specification [1]. Some ideas are adopted from webm-tools project
[2].
Blackframe video filter now sets metadata accordingly.
the libavfilter/vf_blackframe.c filter now not only logs detected
values, but also sets frame metadata. Currently, only `pblack` value is set
but `SET_META` macro has been introduced to ease development in the future.
Signed-off-by: Stepan Bujnak <stepan.bujnak@gmail.com> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
MSVC does not allow passing file pointers between libs
This API can thus not work with MSVC and as it was very recently added
and its it was in no release its removial should not cause any problems
A better API will be implemented, but its not finished yet, this revert is
to avoid potentially blocking the release
Found-by: Hendrik Leppkes <h.leppkes@gmail.com> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
hevc/rext: basic infrastructure for supporting range extension
- support for 4:2:2 and 4:4:4 up to 12 bits
- add a new profile for range extension
(cherry picked from commit d3c067fa65bbc871758d28aa07f54123430ca346)
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
avformat: add av_stream_get_parser() to access avformat AVParser
The AVStream.parser field is considered private and its location cannot be
preserved while preserving also ABI compatibility to libav, as libav added fields
before it.
Some tools like ffmpeg.c access this field though
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Ben Avison [Thu, 10 Jul 2014 23:14:28 +0000 (00:14 +0100)]
armv6: Accelerate ff_imdct_half for general case (mdct_bits != 6)
The previous implementation targeted DTS Coherent Acoustics, which only
requires mdct_bits == 6. This relatively small size lent itself to
unrolling the loops a small number of times, and encoding offsets
calculated at assembly time within the load/store instructions of each
iteration.
In the more general case (codecs such as AAC and AC3) much larger arrays
are used - mdct_bits == [8, 9, 11]. The old method does not scale for
these cases, so more integer registers are used with non-unrolled versions
of the loops (and with some stack spillage). The postrotation filter loop
is still unrolled by a factor of 2 to permit the double-buffering of some
VFP registers to facilitate overlap of neighbouring iterations.
I benchmarked the result by measuring the number of gperftools samples
that hit anywhere in the AAC decoder (starting from aac_decode_frame())
or specifically in ff_imdct_half_c / ff_imdct_half_vfp, for the same
example AAC stream:
Before After
Mean StdDev Mean StdDev Confidence Change
aac_decode_frame 2368.1 35.8 2117.2 35.3 100.0% +11.8%
ff_imdct_half_* 457.5 22.4 251.2 16.2 100.0% +82.1%
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>