[Ffmpeg-devel] Fixed vs. Floating Point AAC

[Michel Lespinasse]

On Wed, Mar 08, 2006 at 05:32:53PM -0800, Mike Melanson wrote:
> 	Okay, so we need to start tracking these requests:
> 
> * pure float implementation
> * fast, low-precision, strictly 32-bit integer implementation
> * slower, higher-precision 32-bit integer implementation
> * pure 64-bit integer implementation
> 
> Anyone else have a request while we're in the planning phases? :)

Double precision float implementation (not to be used in production,
but nice to have at the development stage to have a reference to compare
the others precision against).

Floats are definitely easiest to start with. In my experience, they are
also fastest on about every <10 years old desktop hardware - though they
are very slow on some embedded chips that do not have hardware support.

If someone is looking for a benchmark, liba52 CVS supports both floats
and ints as a configure option - with the same code base and algorithms
otherwise. You can even switch between the 3 above mentionned integer
support modes by just changing a few ifdefs at the end of a52_internal.h

Hope this helps, though I'd be surprised if integer supporters changed
their mind about this. In my opinion, they've been ignoring the evidence
for 10 years already.

-- 
Michel "Walken" Lespinasse
"Bill Gates is a monocle and a Persian cat away from being the villain
in a James Bond movie." -- Dennis Miller