Glenn Fowler wrote:
> On Tue, 03 Mar 2009 16:40:43 +0100 Roland Mainz wrote:
> > Glenn Fowler wrote:
> > > the second field contains letters that correspond to the
> > > optional compile-time SHOPT_* features (when ksh was built)
> > >
> > >         A       SHOPT_AUDIT
> > >         B       SHOPT_BASH
> 
> > Is it possible reserve "D" (and "typeset -D") for "IEEE 754-2008-style
> > decimal floating point", please (this differs from normal IEEE754
> > floating-point math that it uses bash10 instead of base2 and is needed
>                                    ^^^^^^
> (maybe you need a break from shell hacking?:)

No, I am serious in this case. IEEE 754-2008 supports two _different_ 
floating-point formats:
a) The "traditional" base2 floating-point format we all know from IEEE
754-1985 and ISO C99 and ISO C++ etc.
b) The "new" base10 (="decimal") floating-point format (which comes with
new headers, library functions (see decNumber below) and C datatypes)

These two are different entities, use different datatypes and can be
used in the same application (the difference are in binary
representation (decimal floating-point uses densely packed
floating-point), rounding and text<--->binary conversions and other
little details).

> > for stuff like financial applications (somewhere I've queued a longer
> > email abóut this in my drafts folder)) ?
> 
> what are the coding/runtime mechanisms for IEEE 754-2008 decimal floating 
> point

1. See
http://en.wikipedia.org/wiki/Decimal_floating_point#IEEE_754-2008_encoding
2. Newer compilers like gcc >= 4.2 support "Decimal Floating Point"
directly (e.g. it is expected that in the future all newer compilers
will support decimal floating-point - the transition may need another
four or five years for all platforms (IBM is likely going to lead in
this area since their high-end Power hardware supports decimal floating
point in hardware already)).
3. IBM provides the ANSI-C "decNumber" library (see
http://www.alphaworks.ibm.com/tech/decnumber) under a MIT license for
compilers/platforms which do not support decimal-floating-point yet

> recall that ast does its own string<=>fp conversions
> these work for IEEE 754
> but no provisions have been made for IEEE 754-2008
> (it may be that no provisions are necessary)

No provisions are neccesary for the traditional IEEE 754-2008 base2
floating-point math ([a] above) ... but base10 math ([b] above) is
something completely new.

> in particular, can a single process switch between IEEE 754 and IEEE 754-2008?

Yes, see above. Both [a] and [b] are seperate things.

----

Bye,
Roland

-- 
  __ .  . __
 (o.\ \/ /.o) [email protected]
  \__\/\/__/  MPEG specialist, C&&JAVA&&Sun&&Unix programmer
  /O /==\ O\  TEL +49 641 3992797
 (;O/ \/ \O;)
_______________________________________________
ast-users mailing list
[email protected]
https://mailman.research.att.com/mailman/listinfo/ast-users

Reply via email to