No description
Find a file
Ganesh Ajjanagadde a5202bc968 swresample/resample: improve bessel function accuracy and speed
This improves accuracy for the bessel function at large arguments, and this in turn
should improve the quality of the Kaiser window. It also improves the
performance of the bessel function and hence build_filter by ~ 20%.
Details are given below.

Algorithm: taken from the Boost project, who have done a detailed
investigation of the accuracy of their method, as compared with e.g the
GNU Scientific Library (GSL):
http://www.boost.org/doc/libs/1_52_0/libs/math/doc/sf_and_dist/html/math_toolkit/special/bessel/mbessel.html.
Boost source code (also cited and licensed in the code):
https://searchcode.com/codesearch/view/14918379/.

Accuracy: sample values may be obtained as follows. i0 denotes the old bessel code,
i0_boost the approach here, and i0_real an arbitrary precision result (truncated) from Wolfram Alpha:
type "bessel i0(6.0)" to reproduce. These are evaluation points that occur for
the default kaiser_beta = 9.

Some illustrations:
bessel(8.0)
i0      (8.000000) = 427.564115721804739678191254
i0_boost(8.000000) = 427.564115721804796521610115
i0_real (8.000000) = 427.564115721804785177396791

bessel(6.0)
i0      (6.000000) = 67.234406976477956163762428
i0_boost(6.000000) = 67.234406976477970374617144
i0_real (6.000000) = 67.234406976477975326188025

Reason for accuracy: Main accuracy benefits come at larger bessel arguments, where the
Taylor-Maclaurin method is not that good: 23+ iterations
(at large arguments, since the series is about 0) can cause
significant floating point error accumulation.

Benchmarks: Obtained on x86-64, Haswell, GNU/Linux via a loop calling
build_filter 1000 times:
test: fate-swr-resample-dblp-44100-2626

new:
995894468 decicycles in build_filter(loop 1000),     256 runs,      0 skips
1029719302 decicycles in build_filter(loop 1000),     512 runs,      0 skips
984101131 decicycles in build_filter(loop 1000),    1024 runs,      0 skips

old:
1250020763 decicycles in build_filter(loop 1000),     256 runs,      0 skips
1246353282 decicycles in build_filter(loop 1000),     512 runs,      0 skips
1220017565 decicycles in build_filter(loop 1000),    1024 runs,      0 skips

A further ~ 5% may be squeezed by enabling -ftree-vectorize. However,
this is a separate issue from this patch.

Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
2015-11-08 21:18:16 -05:00
compat Merge commit '407ac22322' 2015-11-02 11:58:27 +01:00
doc swresample: allow double precision beta value for the Kaiser window 2015-11-08 21:11:07 -05:00
libavcodec avcodec/aacsbr: Use FLOAT_0 2015-11-08 11:05:48 +01:00
libavdevice avdevice/dshow_enummediatypes: check return of av_malloc 2015-11-06 09:36:57 -05:00
libavfilter avfilter: add anoisesrc 2015-11-08 18:50:26 +01:00
libavformat avformat/segafilm: Only add index entries when the stream exists 2015-11-07 14:09:17 +01:00
libavresample avresample/resample: remove unused variable 2015-09-19 23:40:34 +02:00
libavutil softfloat: handle INT_MIN correctly in av_int2sf 2015-11-08 21:06:40 +01:00
libpostproc Merge commit 'e88103a7f9' 2015-09-05 21:35:46 +02:00
libswresample swresample/resample: improve bessel function accuracy and speed 2015-11-08 21:18:16 -05:00
libswscale swscale: rename sws_rgb2rgb_init to ff_sws_rgb2rgb_init 2015-10-29 23:49:46 +01:00
presets presets: remove moldering iPod presets 2014-06-17 16:15:04 -08:00
tests Add pixblockdsp checkasm tests 2015-11-07 18:46:55 -08:00
tools Replace remaining occurances of av_free_packet with av_packet_unref 2015-10-27 14:35:30 +01:00
.gitattributes Treat all '*.pnm' files as non-text file 2014-11-28 17:52:43 -05:00
.gitignore gitignore: ignore object file temporaries 2015-10-10 20:30:41 +02:00
.travis.yml Merge commit '1e0b8bf0b3' 2015-09-12 13:31:22 +02:00
arch.mak use mmi instead of loongson3 as simd-optimization flag 2015-07-07 03:46:57 +02:00
Changelog avfilter: add anoisesrc 2015-11-08 18:50:26 +01:00
cmdutils.c all: use FFDIFFSIGN to resolve possible undefined behavior in comparators 2015-11-03 16:28:30 -05:00
cmdutils.h cmdutils: remove sws_opts usage, simplify code 2015-08-08 16:51:25 +02:00
cmdutils_common_opts.h opts: add list device sources/sinks options 2014-10-25 20:20:31 +02:00
cmdutils_opencl.c all: use FFDIFFSIGN to resolve possible undefined behavior in comparators 2015-11-03 16:28:30 -05:00
common.mak Merge commit 'c1aac39eac' 2015-10-14 13:54:50 +02:00
configure mmaldec: add vc1 decoding support 2015-11-07 19:24:49 +01:00
COPYING.GPLv2
COPYING.GPLv3
COPYING.LGPLv2.1 cosmetics: Delete empty lines at end of file. 2012-02-09 12:26:45 +01:00
COPYING.LGPLv3
CREDITS CREDITS: redirect to Git log, remove current outdated content 2013-01-31 18:02:52 +01:00
ffmpeg.c all: use FFDIFFSIGN to resolve possible undefined behavior in comparators 2015-11-03 16:28:30 -05:00
ffmpeg.h ffmpeg: add abort_on option to allow aborting on empty output 2015-10-22 19:03:01 +02:00
ffmpeg_dxva2.c ffmpeg_dxva2: call GetDesktopWindow() in place of GetShellWindow() 2015-06-03 16:25:08 +02:00
ffmpeg_filter.c ffmpeg_filter: check encoder before using it to set frame size. 2015-10-30 10:55:10 +01:00
ffmpeg_opt.c all: remove some casts of function pointer to void * 2015-10-24 15:14:59 -04:00
ffmpeg_qsv.c Merge commit 'fb472e1a11' 2015-10-22 16:18:02 +02:00
ffmpeg_vdpau.c ffmpeg_vdpau: Ignore decoder's max supported level 2015-08-16 15:02:33 -07:00
ffmpeg_videotoolbox.c ffmpeg/videotoolbox: protect UTGetOSTypeFromString on both VDA and VT 2015-10-15 10:22:31 +02:00
ffplay.c Replace remaining occurances of av_free_packet with av_packet_unref 2015-10-27 14:35:30 +01:00
ffprobe.c ffprobe: add support for printing packet strings metadata as packet tags 2015-10-30 20:47:43 +01:00
ffserver.c Replace remaining occurances of av_free_packet with av_packet_unref 2015-10-27 14:35:30 +01:00
ffserver_config.c ffserver_config: check for INT_MIN before doing FFABS 2015-10-13 19:40:09 -04:00
ffserver_config.h ffserver: Use singlejpeg muxer for jpeg 2015-06-08 03:36:22 +02:00
INSTALL.md INSTALL: add markdown syntax 2014-05-28 22:38:38 +02:00
library.mak build: add LDLIBFLAGS 2015-07-08 14:35:02 +02:00
LICENSE.md avfilter: add rubberband wrapper 2015-09-20 19:54:57 +02:00
MAINTAINERS lavfi: remove astreamsync. 2015-11-07 16:43:36 +01:00
Makefile Merge commit 'fb472e1a11' 2015-10-22 16:18:02 +02:00
README.md README: replace http with https 2015-10-06 13:27:29 +02:00
RELEASE RELEASE: update to 2.8.git 2015-09-09 23:53:15 -03:00
version.sh version.sh: add note that ffversion.h is auto-generated 2015-10-25 10:14:44 -04:00

FFmpeg README

FFmpeg is a collection of libraries and tools to process multimedia content such as audio, video, subtitles and related metadata.

Libraries

  • libavcodec provides implementation of a wider range of codecs.
  • libavformat implements streaming protocols, container formats and basic I/O access.
  • libavutil includes hashers, decompressors and miscellaneous utility functions.
  • libavfilter provides a mean to alter decoded Audio and Video through chain of filters.
  • libavdevice provides an abstraction to access capture and playback devices.
  • libswresample implements audio mixing and resampling routines.
  • libswscale implements color conversion and scaling routines.

Tools

  • ffmpeg is a command line toolbox to manipulate, convert and stream multimedia content.
  • ffplay is a minimalistic multimedia player.
  • ffprobe is a simple analysis tool to inspect multimedia content.
  • ffserver is a multimedia streaming server for live broadcasts.
  • Additional small tools such as aviocat, ismindex and qt-faststart.

Documentation

The offline documentation is available in the doc/ directory.

The online documentation is available in the main website and in the wiki.

Examples

Coding examples are available in the doc/examples directory.

License

FFmpeg codebase is mainly LGPL-licensed with optional components licensed under GPL. Please refer to the LICENSE file for detailed information.

Contributing

Patches should be submitted to the ffmpeg-devel mailing list using git format-patch or git send-email. Github pull requests should be avoided because they are not part of our review process. Few developers follow pull requests so they will likely be ignored.