(j3.2006) 08-267

Jim Xia jimxia
Tue Sep 23 16:18:40 EDT 2008


j3-bounces at j3-fortran.org wrote on 09/22/2008 02:44:06 PM:

> I understand the reasoning behind 08-267, but I'm not sure I understand
> all the ramifications.
> 
> If one can't reference a coindexed object that is polymorphic, or that
> has an allocatable ultimate component that is polymorphic or has a
> length parameter, why allow such objects to have codimensions in the
> first place?


Back to 2007-08 London meeting, I've submitted a report (Canadian National 
report, if I remember correctly) saying that coarrays with OO and PDT is 
potentially a dangerous combination that may cause grief in implementation 
and applications, and we should be looking at option to forbid it.  But 
back then without any concrete evidence, we simply convinced ourselves 
that this was only fear, not reality.


> 
> Shouldn't the restriction be on declaring such objects, not on
> referencing them?
> 
> If we can't declare or reference such objects, it seems that coarrays
> are essentially built on Fortran 95, not Fortran 2003.  Is this the
> intent?


Coarrays are for performance, and polymorphism is, in general, hurting 
performance (although I'm not against the OO features in F2003).  Don't 
know anything about PDT in real applications -- does anyone know?  From 
this angle, why would anyone use coarrays with OO polymorphism?  Maybe 
Cray can provide good reasons.

Cheers,

Jim Xia

RL Fortran Compiler Test
IBM Toronto Lab at 8200 Warden Ave, Markham, On, L6G 1C7
Phone (905) 413-3444  Tie-line 313-3444
email: jimxia at ca.ibm.com
D2/YF7/8200 /MKM
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://j3-fortran.org/pipermail/j3/attachments/20080923/3ccba7dd/attachment.html 



More information about the J3 mailing list