Fixed point matrix representation considered harmful

[Joerg Sonnenberger]

On Thu, Mar 20, 2008 at 08:04:39PM +0200, Daniel Stone wrote:
> On Thu, Mar 20, 2008 at 06:48:23PM +0100, Joerg Sonnenberger wrote:
> > On Wed, Mar 19, 2008 at 09:55:34AM -0700, Keith Packard wrote:
> > > On Wed, 2008-03-19 at 16:59 +0100, Joerg Sonnenberger wrote:
> > > > OK, so you basically want to specify a 32bit (or whatever) IEEE format
> > > > without all the nasty parts? E.g. subnormals and exception values?
> > > 
> > > Exception values (NaN, +/-inf) should be invalid, but denorms shouldn't
> > > cause any issues for us.
> > 
> > Except that it complicates decoding for non-IEEE platforms. Think VAX.
> 
> Are we actually designing for VAX these days? There aren't many other
> non-IEEE platforms that I know of: even lame embedded/consumer chips are
> pretty much all IEEE now, TTBOMK.

Well, at least some of the RISC platforms have IEEE-like floating point,
but don't do the magic values. Think about Alpha. My point is to just
specify a simple binary format and leave out all the special magic
values.

Joerg