In the summer of 2006, Paul Jones began work on converting the ffmpeg SoundCodecWMA decoder to fixed point so that it could be used with Rockbox (FS#4984). Although he did not complete the project, it raised interest in eventually getting ffmpeg's decoder working in Rockbox. This Google Summer of Code 2007 project completed his task.
Codec is currently in SVN and functioning well above real time on all software targets. Results appear comparable to ffmpeg for nearly all files. Noise coding and line spectral pairs are implemented, and appear to function as well as ffmpeg. Seeking has been implemented and appears robust. Performance is now extremely good on ARM. Coldfire performance is also very good but could be further improved.
Both WMA Std. flavors are supported. No special WMA variants are supported (Pro, Voice, Lossless) due to lack of an open source decoder.
* Additional ASM optimization (requantization, etc)
* Review Tremolo and see if more ASM code can be used in WMA too
* Improve the makefile to allow compiling a stand alone codec for testing
Who's working on this
* Michael Giacomelli (saratoga)
with lots of help from:
* linuxstb (developed the ASF parser)
* marsdaddy (Paul Jones)