I am new to FLAC, from previous discussions i could make out that code is already available. Is this code fixed point? it is optimized for which processor?

If you are asking if it will encode integer audio data then yes.

http://flac.sourceforge.net/features.html



You can write an integer decoder if you want, if the reference decoder is not already written that way.

There are integer only flac decoders if thats what you mean. The rockbox flac decoder is a modified version of ffmpeg's flac decoder, and is extremely fast on ARM and Coldfire.

libFLAC only uses floating point in parts of the encoder. (decoding is integer only.) there is a fixed point flavor you can turn on with #ifdefs but it's not very good.

The same applies for FFmpeg. The FLAC decoder is integer only, and the encoder uses floating-point. The exception is for encoding with compression levels 0 to 2, which only use integers.

Could you tell me more about rockbox flac decoder? How it is different from "Original one" available from http://sourceforge.net/project/showfiles.p...ackage_id=12677

Do you have have comparison figures of rockbox and Original decoder on ARM?

Thanks Josh Coalson.
Is it possible for you to also provide me an idea on Approx. no. of
Multiplications,
Adds/subs,
ORs/Shifts,
Movs etc
involved in decoding of per sample of audio frame by library libFlac, .. this can give idea on performance on any platform.
I guess i have not asked smthing unanswerable

Its about 3-4x faster as I recall on ARM7TDMI. I think libflac simply isn't well optimized.

The version of libflac Rockbox used was 1.1.2, which predates the bitbuffer optimizations in 1.1.4. Don't know how much that affects things though.