[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Cfp-interest] C binding for IEEE 754-2008


Thanks for the comments. Please see responses below.


On Jul 18, 2012, at 1:00 PM, Michel Hack wrote:

Dan Zuras wrote:
I see almost no mention of either the newly defined
tower of precisions or the elements of Clause 9 other
than their mention WRT rounding modes.  Although it
mentions that these will appear in parts 3 through 5.

Well, at least they *are* mentioned.

It would indeed be nice if a language like C could provide
for a class of precisions not just for FP, but also for
integer variables, to allow a compiler to synthesize the
appropriate platform-specific arbitrary-precision idioms
instead of each programmer having to invent their own,
less portable, implementations.

Part 3 of the CFP TS will specify the 754 precisions for FP. I think integer 
types are out of scope for our study group. What do you think is needed beyond 
<stdint.h> and <inttypes.h>?

I like the way sensible support for NaNs is finally being
introduced -- in particular the payload functions, which
may finally allow reasonably-portable code to be written,
and enable some of the decoration encodings we have been
discussing in stds-1788 for compressed intervals.

Good to hear.

One thing I missed was recommendations (if not mandates)
for payload preservation across casts.  Perhaps that would
be addressed in Part 3, which may address cross-radix casts.

These recommendations can be considered part of 754 operation semantics which 
are adopted by reference by the TS. Maybe an explicit mention would be helpful.

I also like the way conversions to and from strings have
been described -- and am glad that the rules were interpreted
the same way I did when I brought the AIX conversion routines
into 754-2008 conformance.

Very good.

---Sent: 2012-07-18 20:06:31 UTC

Cfp-interest mailing list

754 | revision | FAQ | references | list archive