rockbox/apps/codecs/libwmapro
Nils Wallménius e436483b66 libmwapro: use codeclib asm macros for XNPROD31, speeds up decoding of a 55kbps wma pro file by 40% on h300.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@27612 a1c6a512-1295-4272-9138-f99709370657
2010-07-29 11:01:30 +00:00
..
libwmapro.make libwmapro: build with -O2 for coldfire, 5% speedup. 2010-07-27 06:48:15 +00:00
mdct_tables.c Submit part of FS#11498. Major speedup for WMA Professional on ARM and Coldfire CPUs. Introduce asm routines for multiplications, move arrays with major impact on decoding speed to IRAM. Speeds up decoding by 25% on PP5022 and 34% on mcf5249. 2010-07-26 21:43:07 +00:00
mdct_tables.h Partial conversion of the wmapro decoder to fixed point arithmetic. Currently inverse quantization & rescaling, 2010-07-05 22:33:37 +00:00
quant.h Increase the size of quant table in libwmapro to accomodate more values for exp. Fixes wma pro failure on test files in download.rockbox.org/test_files. 2010-07-17 09:13:55 +00:00
README.rockbox libwmapro no longer needs libm in the sim; update the makefile and README.rockbox. 2010-07-17 08:19:05 +00:00
SOURCES More cleaning for libwmapro; use codeclib's bitstream functions and drop those from the libwmapro. 2010-07-17 07:53:14 +00:00
wma.c Correct the scaling of the coeffs to 17bits. 2010-07-28 18:39:10 +00:00
wma.h libwmapro : remove dead/unneeded code from wma.[ch] 2010-07-17 08:09:17 +00:00
wmapro_math.h libwmapro: coldfire asm for vector_fixmul_scalar, 1% speedup 2010-07-28 23:25:45 +00:00
wmapro_mdct.c libmwapro: use codeclib asm macros for XNPROD31, speeds up decoding of a 55kbps wma pro file by 40% on h300. 2010-07-29 11:01:30 +00:00
wmapro_mdct.h Partial conversion of the wmapro decoder to fixed point arithmetic. Currently inverse quantization & rescaling, 2010-07-05 22:33:37 +00:00
wmaprodata.h Remove the floating point code from wmaprodec.c and change variable names accordingly. 2010-07-14 17:55:17 +00:00
wmaprodec.c Refinement of initial lowmem handling for libwmapro (r27593). Set maximum supported channels to 2 instead of 8 for low memory targets. This solution saves more RAM than the former solution and does avoid possible NULLPTR access. As add-on remove tabs. 2010-07-28 19:01:45 +00:00
wmaprodec.h Modify the wma pro decoder to produce non-interleaved samples, and work directly on the buffers in WMAProDecCtx instead to avoid the redundant copying of the output data. ~10% speedup (-2MHz) on pp502x and ~1.5% speedup (-3.8Mhz) 2010-07-26 22:03:20 +00:00

Library: libwmapro
Imported: 2010-04-30 by Mohamed Tarek

This set of files form the files needed from ffmpeg's libavcodec and libavutil
to build a standalone wma professional decoder.

LICENSING INFORMATION

ffmpeg is licensed under the Lesser GNU General Public License and the file
wmaprodec.c is copyright (c) 2007 Baptiste Coudurier, Benjamin Larsson, Ulion
and copyright (c) 2008 - 2009 Sascha Sommer, Benjamin Larsson.

IMPORT DETAILS

Based on ffmpeg svn r22886 dated 15 April 2010.

The decoder now (17 July 2010) uses fixed point arithmetic everywhere. It still
doesn't work on 3 targets, namely clip, c200v2 and m200v4 due to their small 
codec buffer compared to wma pro's needs. 

A comparison against ffmpeg floating point decoder gives :
0.2~0.3%  max relative error
0.005~0.01% average relative error