(j3.2006) Liaison to IFIP WG 2.5

[Van Snyder]

On Tue, 2007-08-21 at 10:43 -0500, Bill Long wrote:
> We already have a great facility for intervals build into Fortran - 
> modules.  I think there would need to be two preconditions satisfied 
> before it was interesting to go beyond that:

Adequate, maybe, but not great.

> 1) There would have to be significant customer demand (so far I've
> seen exactly zero such demand), and

The numerical mathematics community tends to use what they have in the
way of languages rather than to try to put pressure on compiler vendors
to make extensions just for them.  They know they're small potatoes
compared to, say, MacNeal-Schwendler (now MSC).  Engineers and
engineering companies tend to believe they know all they need to know
about numerical mathematics, so don't appreciate the benefits of
interval arithmetic.  In many applications, they're right.  But when
their Newton method doesn't converge (which it does globally in interval
arithmetic) they just accept crappy answers or use some hand-waving
work-around from Numerical Recipes.

> 2) Hardware support needs to be widespread to make something beyond a 
> module worthwhile from a performance point of view.  That means, at a 
> minimum, that IEEE 754r++ specify a standard.

This is exactly what IFIP WG 2.5 hopes for.  As I remarked, Kulisch and
others plan to complain to IEEE that 754r doesn't provide adequate
support for interval arithmetic.  As I also remarked, I told them they
were not likely to get their wish unless somebody showed up at standards
committee meetings to actually write the words.  Interval arithmetic was
dropped from Fortran 2003 because nobody wrote any serious specs papers,
let alone syntax or edits papers.  Wolfgang didn't have funding to come
to many meetings, Baker Kearfott stopped coming, Tony Warnock never
wrote anything.  For their own unstated reasons, Sun didn't write any
significant papers on interval arithmetic.  Since February 1997, I
remember exactly one half-page paper, which proposed syntax for
arithmetic operators with directed rounding.  Proposals on the table
before then were half-baked trial balloons, such as allowing to specify
the precedence of user-defined operators, which would in turn allow a
given operator symbol to have different precedence in different scoping
units if they were in fact a different operator with the same spelling.

-- 
Van Snyder                    |  What fraction of Americans believe 
Van.Snyder at jpl.nasa.gov       |  Wrestling is real and NASA is fake?
Any alleged opinions are my own and have not been approved or
disapproved by JPL, CalTech, NASA, the President, or anybody else.