fixed 2 orders of magnitude bug in blockswitch
small DRM fix in huffman coding
made DRM to compile, not sure if it really works
added faacgui.exe to the build, does compile but doesn't really work.
small cleanup in bitstream.h
set default mode to ABR@128
added block switching threshold limit
use new quantize/code functions
renamed fields in coderinfo
initial windows grouping support
removed old huffman code
fixed compiler warnings
cleaner and more compatible SSE code
clang compatibility fix
MINGW DLL compatibility
added x86_64 to SSE check
use SSE only on i686 hosts
removed obsolete project files
Merge pull request #5 from vividos/master
moved faacgui project file to project/msvc folder and updated it to Visual Studio 2017
moved faac project file to project/msvc folder and updated it to Visual Studio 2017
moved Visual Studio project files to folder project/msvc
Merge pull request #4 from vividos/master
ignore some Visual Studio folders
updated Visual Studio project files to VS 2017 and reorganized output folder
simplified bandwidth setting; modified ABR code
version -> 1.29.7.6 (source package fix)
add kiss_fft include files to autotools
clang compatibility fix
autotools packaging fix
autotools support update
added -no-undefined option (MSYS2 project)
avoid clang error (initializer element is not a compile-time constant)
better framebuffer size check
frame buffer overrun check
Merge branch 'master' of ssh://git.code.sf.net/p/faac/faac
increased frame buffer sixe to 8k for all stream types and allow higher ADTS bitrates
increased frame buffer sixe to 8k for all stream types and allow higher ADTS bitrates
Revert "library interface version++"
library interface version++
MP4 max frame size calculation moved to library
added scalefactor output counter, it will be useful for window grouping;
coder data struct cleaned up; fields renamed
LTP and Main object types removed
mp4write: esds, stsc, stco atoms changed for better compatibility
raised max framesize and max quality (5000) for mp4 files
quantizer: calculate scalefactor first then band multiplier
cutoff frequency tuned to fully utilize top band; removed mid frequency boost (it turned out useless)
print faac version header before help
print help options in help
removed --help-long option, it was just an alias for -H
fixed help: --long-help was renamed to --help-long
bugfix: integer calculation should be double; fix quality multiplier accordingly
quantizer: use standard "MAGIC_NUMBER" for integer conversion
updated help, manpage and changelog
The strncmp() declaration needs to conform to the standard to fix build failure with musl libc.
check if output file exists; --overwrite option added
frontend: help/usage reorganized
block switch: produce a little more long frames
block switching parametrized: more long blocks at lower bitrates, more short blocks at higher bitrates
give more quality to lower frequencies, less quality to highs
quantizer: return back to using fixed bias with RMS energy matching
max quality setting raised to 2000
a better way to balance band volume
default bandwidth setting simplified; highened cutoff
quantizer: use unscaled data for most calculations