> The question underlying these questions is "If I use assumed-shape dummy
> arguments, can I count on performance (nearly) as good as for assumed-size
> dummy arguments in the case that the corresponding actual argument has unit
> stride in its first dimension?"
Care to write the answer such that non-computer scientists understand
it- once
you know it?
This is the relevant excerpt from the NAG f95 man page:
> -Oassumed=shape
> Optimises assumed-shape array dummy arguments
> according to the value of shape, which must be
> one of
>
> always_contig
> Optimised for contiguous actual argu-
> ments. If the actual argument is not
> contiguous a runtime error will occur
> (the compiler is not standard-conforming
> under this option).
>
> contig Optimised for contiguous actual argu-
> ments; if the actual argument is not con-
> tiguous (i.e. it is an array section) a
> contiguous local copy is made. This may
> speed up array section accessing if a
> sufficiently large number of array ele-
> ment or array operations is performed
> (i.e. if the cost of making the local
> copy is less than the overhead of discon-
> tiguous array accesses), but usually
> makes such accesses slower. Note that
> this option does not affect dummy argu-
> ments with the TARGET attribute; these
> are always accessed via the dope vector.
>
> section
> Optimised for low-moderate accesses to
> array section (discontiguous) actual
> arguments. This is the default.
>
> Note that CHARACTER arrays are not affected by
> these options.
Klaus
--
[log in to unmask] - www.ramstock.de
Am Krag 12g, 55286 Wörrstadt
phone +49 6732 963434 - fax +49 6732 963435
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|