(j3.2006) question about co-array definition and synchronization

[Jim Xia]

The first point to make is that the "new" terminology (by new I mean from 
Fairfax meeting) is that each of the X's on each image is a "co-array". So 

don't say "each image defines its local piece of co-array X". Instead, 
each 
defines the whole of a co-array.

Apologies if the reason behind the "new terminology" was made crystal 
clear to everyone else except me (since I didn't know what happened in 
Fairfax meeting).  I need some educations here because I find this 
statement more confusing
So don't say "each image defines its local piece of co-array X". Instead, 
each defines the whole of a co-array.

So what are the relationships among an image, its local portion of a 
co-array and the whole co-array?  Does the word co-array mean the whole 
array that is accessible by ALL images via co-indices, or it refers to a 
part of it that is affiliated with a particular image, or it refers to 
both?  In UPC the word "affinity" is used to describe the logical 
association between thread (image) and shared object (co-array I should 
say?)  Do we have a terminology to describe the association between 
co-array and image?

Thanks,

Jim Xia

XL Fortran Compiler Testing
IBM Toronto Lab at 8200 Warden Ave.
Phone (905) 413-3444  Tie-line 969-3444
D2/NAH/8200 /MKM



Aleksandar Donev <donev1 at llnl.gov> 
Sent by: j3-bounces at j3-fortran.org
06/20/2007 02:22 PM
Please respond to
fortran standards email list for J3 <j3 at j3-fortran.org>


To
fortran standards email list for J3 <j3 at j3-fortran.org>
cc

Subject
Re: (j3.2006) question about co-array definition and synchronization






On Wednesday 20 June 2007 11:11, Michael Ingrassia wrote:

> Is this draft-standard-conformable?
Yes, of course :-)

> If so, how does it not fall afoul of the rule [07-007r2:212:4-5]
The first point to make is that the "new" terminology (by new I mean from 
Fairfax meeting) is that each of the X's on each image is a "co-array". So 

don't say "each image defines its local piece of co-array X". Instead, 
each 
defines the whole of a co-array.

> Should the words say something like "the part of the co-array that is
> defined shall not be referenced, defined or become undefined in a 
segment
> on another image" ?  Or are they correct as written?
I already proposed (in private paper drafts) to change this to say 
"subobject 
of the co-array", but for different reasons. Specifically, we do want to 
allow image 1 to modify X[2](1) and image 2 to modify X[2](2) 
"concurrently". 
A more realistic example would be each image working on a column of an 
array. 
I think it is not clear from the way the words are written that this is 
OK.

Thanks,
Aleks

-- 
Aleksandar Donev, Ph.D.
Lawrence Postdoctoral Fellow @ LLNL
High Performance Computational Materials Science and Chemistry
E-mail: donev1 at llnl.gov
Phone: (925) 424-6816  Fax: (925) 423-0785
Address: P.O.Box 808, L-367, Livermore, CA 94551-9900
Web: http://cherrypit.princeton.edu/donev

_______________________________________________
J3 mailing list
J3 at j3-fortran.org
http://j3-fortran.org/mailman/listinfo/j3

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://j3-fortran.org/pipermail/j3/attachments/20070620/90925155/attachment.html