Re: Function PRESENT and logical combinations
From: Ron Shepard (ron-shepard_at_NOSPAM.comcast.net)
Date: Tue, 15 Jun 2004 10:20:20 -0500
In article <2j7n9bFt4slrU1@uni-berlin.de>,
Jan Vorbruggen <firstname.lastname@example.org> wrote:
> > I will state as a point of absolute truth and without qualifiacation
> > that the standard explicit says the value of a function reference
> > is determined by its execution.
> I haven't heard even (cough, cough) Richard Maine claim that this isn't
> so 8-). But the Talmudic arguments surely start with clarifying what the
> words "execution" and "determined" mean, and perhaps even what the word
> "its" means 8-).
There does seem to be some ambiguity in the above wording (which
comes from the standard). Some people interpret the above to mean
that the value of a function reference is determined *ONLY* by its
execution (e.g. by executing the compiled code). Others think it
means that any mechanism that returns the same value as its
execution would have returned is legal -- i.e., the execution is the
ultimate authority on the value, but any mechanism is allowed to
return that value (within the vagaries of floating point precision
and dependence on external devices such as timers). The latter
mechanism includes such things as inlining and other optimizations
(e.g. replacing fortran statements with hardware execution) that the
former would not allow.
I think I understand the controversy, but I see benefits to both
interpretations and I don't have an opinion on which would be best
in the long run.
$.02 -Ron Shepard