(j3.2006) (SC22WG5.4096) [ukfortran] Standard intrinsics andcoarrays

[Bill Long]

Malcolm Cohen wrote:
> Bill Long wrote:
>> If an environment variable is never changed during execution of the program, 
>> then I don't see an issue with GET_ENVIRONMENT_VARIABLE either.
> 
> Why would we expect the same environment variables to exist for the whole 
> program?  

I would expect that the program launching mechanism would establish the 
same environment for each image.  However, as Nick's second version 
proposal suggests, this might not be true for some cases.


On a loosely-connected cluster, each image is going to have its own
> set of environment variables and some could well be different (like HOST for 
> example).
> 
> We probably ought to say that environment variables are per-image. 


I think this would be at least as bad - probably worse.  It would imply 
to the user that each image could independently set the value of an 
environment variable, which would be impossible on most systems.

The "processor-dependent" approach is better.  A user who wants portable 
code would call GET_ENVIRONMENT_VARIABLE only on image 1. It might be 
worth adding that suggestion to Nick's NOTE.

Cheers,
Bill




  The current
> text certainly implies that GET_ENVIRONMENT_VARIABLE will return the same value 
> on each image, and that is surely a limitation we don't want to apply.
> 
> Cheers,

-- 
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