(j3.2006) 07-276r1

[Bill Long]

Michael Ingrassia wrote:
>> There 
>> is absolutely no reason to forbid, for example, C_LOC(co_array), and call C 
>> to do something with co_array, is there?
>>     
>
>
> Would this preclude being able in some future standard to interoperate
> as a full coarray (with all its coindexed subobjects) 
> in a C compiler that understood coarrays ?
>
>   
No.  In a future standard we can relax old restrictions to allow for 
added capability.  Existing programs written to the old standard are 
still conforming - they just do not use the new capability.  This a 
general concept underlying the standard, independent of any specific 
feature.  As an example, we used to have a restriction that the variable 
in an assignment statement could not be a function reference.  Now we're 
loosening that restriction to allow references to a limited class of 
functions. 

If there is a demand for binding to UPC in the future, I think our 
current infrastructure is sufficiently general.  For example we could 
add a bind(upc) clause if we did not want to overload the current bind(c).

Cheers,
Bill

-- 
Bill Long                                   longb at cray.com
Fortran Technical Support    &              voice: 651-605-9024
Bioinformatics Software Development         fax:   651-605-9142
Cray Inc., 1340 Mendota Heights Rd., Mendota Heights, MN, 55120