Re: inlining + optimization = nuisance bugs

Luddy Harrison (luddy@concmp.com) wrote:
: I don't find any language in the document that expresses concern for
: the use of too *much* precision.


I believe it is implicit in the standard that the size and semantics
of single and double precision are fixed except for minor "allowed
deviations" in behavior. (E.g. single-rounding multiply-accumulate has
"special dispensation" from Kahan. Yet still, almost every platform
that has this feature also has a compiler switch to turn it off.)


Compilers which unpredicatably substitute extra precision into
programs are not doing programmers a favor. You can say things like
"Floating-point is imprecise anyway" and "Well if you'd just write
robust code, this wouldn't happen" but its just dancing around the
fact that the tools make it (very) difficult to express a precise
algorithm. (E.g. sometimes folks use IEEE floating-point to simulate
larger integer calculations. Such code may be dependent on IEEE 754,
but it should not be dependent on exactly what happens in the
compiler's optimizer.)


-Z-
--

Return to the comp.compilers page.
Search the comp.compilers archives again.