Re: Clause 10, Expression Evaluation

[Thorsten Siebenborn <7_born@xxxxxx>]

Nick Maclaren schrieb:
> Thorsten Siebenborn <7_born@xxxxxx> wrote:> 
> It isn't useful to rely on such a property in a standard like IEEE
> 754R.

I never claimed otherwise.

> Yes, there are such languages.  However, the vast majority of
> relevant languages are not like that, and most of the most important
> are so fundamentally different that such a property cannot be added
> to them.  I need just include C, C++ and Fortran to justify the
> latter.

Correct. And I never claimed otherwise.

> Furthermore, there are no parallel languages based on Hoare's
> seminal structure that have such a property and, as I have said
> several times, there cannot be any general ones with it.  It would
> be possible to have a restricted parallel language (perhaps even
> one based on BSP), but no more and not OpenMP or POSIX threads.
>
> Claiming that an IEEE 754R specification that makes sense in the
> context of Forth, Java and BSP but not C, C++, Fortran, OpenMP and
> POSIX threads is useful....words are failing me.

But that isn't that what I actually claimed.
------------------------ CITE from me ----------------
You are right that there are languages where strict order rules
don't exist and that these languages are still rampant in numerical
computation. You are also right that for these languages debugging
in debug mode may be effectively futile because you don't know how
the optimizing compiler mangles your code (I have one friend who
run into this problem and wasn't able to solve it). So you are right in
criticizing the proposal to define strict expression evaluation
if there are languages who violate this rule by default.
------------------------------------------------------------

So you are claiming something which I already stated.

I have attacked the image that it is impossible to test a Volvo
for testing a Porsche. I say: Translated in computer languages there
*are* Volvos which can be tested to verify that the Porsche
will function. And I added that these specifics Volvos are able
to use tools which are extraordinarily useful.

Malcolm started his claim with ">>Furthermore<<,
it is not useful". So the inherent inability of FORTRAN
to come up with a strict evaluation order is apparently not the
reason he came up with this image.

In fact, I don't know what exactly Malcolm attacked because
his verdict is so general. I want to know what exactly is
futile/senseless/whatever. It could be insisting on strict
evaluation order, languages who support them, whatever.

Regards,
Thorsten

P.S. : It nerves that my reply is per default sent privately
instead using the stds list.