Hi,
A question to C. Coats first: Did you ever investigate the reason behind the
performance degradation. Usually it is in the way you wrote the interface
and the way you set up the variables you passed to it--it might be a copy
overhead, or something of the sort. In general it can often be "blamed" on
the programmer, not the standard...
I think the question is somewhat silly, as most of the things in the
standard were written in a certain way because there is an efficient way to
compile that kind of high level construct, in fact, in most cases, a more
efficient way than what can be expressed via F77. Whether vendors got the
correct points right is another issue. In my experience the only problem is
the fact F90 leaves a lot to the compiler, and compiler vendors all hide
internal things like they are some kind of precious secret (I think not), so
I have to use smart debuggers, memory profilers, and other tricks to see
that the compiler did something silly (or smart). More openness and better
documentation can do marvels.
So the way for the talk, W. Breinard, is, in
my opinion, to explain to "those" people what the Fortran committee has
done, why, what needs to be done, etc., rather then try to prove something
that is ill-posed to start with (you are comparing, as we say in my
language, frogs and grandmas (F77 and F90, a subset to a something that is a
lot more expressive)!), with numbers or otherwise.
A quick random thought...
Aleksandar
|