Alistair Mills writes:
> I would like to discuss the opinion expressed here by Richard vs my opinion!
I'm afraid that I can't understand what point you are making. The
central tenet of what I said was
>> Handling different ranks is *EXACTLY* one of the main purposes of
>> ELEMENTAL.
I don't see what the reply has to do with this. Note, by the way,
that I said nothing at all (nor do I plan to) about the performance
issues involved.
In particular, the point I was arguing with was
> think that trying to use both scalar and vectors with pure elemental
> functions is dangerous!
and the
> rather than a feature for dealing with multiple dimensions and ranks.
Perhaps I wasn't specific enough about which point of yours I was
disagreeing with. I was not disagreeing with your characterization of
elemental as invoking the function as many times as needed. Although
the implementation doesn't have to work exactly that way, that is one
implementation model, and one that is pretty easy to describe. My
disagreement was only with the statement that somehow elemental wasn't
suitable for handling multiple dimensions and ranks (including scalar
vs vector). I don't understand what would make you draw that
conclusion at all.
> As the pure elemental function has to work without changes to the standard
> calling sequences between routines on most platforms,
That is not true. Some of the limitations relating to elemental are
specifically crafted so that elemental procedures do *NOT* have to
have the same calling sequences. Thats why, for example, you can't
pass elemental procedures as actual arguments corresponding to
non-elemental dummies - that won't work if the calling sequences
are different.
There are multiple approaches to implementing elemental. Some of
these involve using the same calling sequence and some don't.
--
Richard Maine | Good judgment comes from experience;
[log in to unmask] | experience comes from bad judgment.
| -- Mark Twain
|