[Ffmpeg-devel] [PATCH] Vorbis decoder

Rich Felker dalias
Thu May 12 20:57:31 CEST 2005


On Thu, May 12, 2005 at 10:26:36AM -0600, Mike Melanson wrote:
> Balatoni Denes wrote:
> >It is a somewhat twisted idea that a perceptual audio codec with 16 bits 
> >precission would need so many bits. In the case of Vorbis the spec. allows 
> >48 bits of precision, and in fact tremor uses an optimized floating point 
> >implementation (the idea: many numbers share the same exponent). In 
> >practice however (looking at the current encoder) 32 bits are more than 
> >enough for the decoder, but this could change. 
> 
> 	I fail to remember which step in the decoding recipe mentions it 
> 	(dot product?), but I know that the Vorbis decoder spec notes that the 
> product of one particular multiplication can overflow a 32-bit variable.

and it doesn't if you shift right first. there's no way that many bits
are relevant to the actual output. something that numerically unstable
will not be useful for compression.

rich





More information about the ffmpeg-devel mailing list