Re: [linux-audio-dev] Fixed vs. floating point

On Mon, 2005-10-17 at 17:34 +0200, Maarten de Boer wrote:
> Very interesting reads. Thanks!
>
>
> > Here's some papers specifically geared toward DSP processors that
> > support the use of fixed point:
> >
> > Superior Audio Requires Fixed-Point DSP
> > http://www.rane.com/note153.html

what a lame author! he writes near the end:

   "This is not to say that floating-point DSPs will never have their
day in achieving superior audio -- it's just not today. What will it
take? Here are some pretty nasty "ifs" necessary for floating-point to
overtake fixed-point: if it is a 56-bit floating-point processor (i.e.,
48-bit mantissa plus 8-bit exponent) or 32-bit with double-precision
(requiring a large accumulator), if the parts run at the same speed as
the equivalent fixed-point part, if they use the same power, and if they
cost the same, then the choice is made."

this was written in 2002, when any FP unit worth mentioning in this
context could operate in double or single precision mode. double
precision mode vastly exceeds the requirements he mentions here,
although i will concede that they may use more power (miniscule amounts
in the overall scheme of things). the fact that most audio apps today
use single precision FP (32 bit) has as much to do with performance and
laziness by authors as with anything else. the technology is there to do
80 bit FP if people want to use it, and some do (Ron Kuper at Cakewalk
just did a presentation at AES on Sonar's new 64 bit mode which uses
double precision FP IIUC).

--p

  
Received on Tue Oct 18 00:15:09 2005