(j3.2006) Alternatives for SYMMETRIC (Was: a question on cobounds)

[Tom Clune]

Something seems terribly wrong if the only definition of "symmetric"  
that anyone in a technical discipline has seen is that of symmetric  
matrices.    The terms "symmetric"  and "symmetry" have simple well- 
established definitions that cover a variety of relationships among  
entities, and should not be restricted by one special case, however  
common in practice.   (Are symmetric matrices even mentioned in the  
standard?)   The symmetry in question here centers on what remains  
constant under "translations of image" for co-arrays.   Seems a  
perfect place for the term "symmetric", IMO.   I am open to a better/ 
comparable term here, but the alternatives that have been mentioned in  
this thread are so far off-the-mark that I am not optimistic.

Cheers,

- Tom



On Jul 28, 2009, at 8:08 PM, Craig Dedo wrote:

>> -----Original Message-----
>> From: j3-bounces at j3-fortran.org [mailto:j3-bounces at j3-fortran.org] On
>> Behalf Of Bill Long
>> Sent: Tuesday, July 28, 2009 13:00
>> To: Loren P Meissner
>> Cc: 'fortran standards email list for J3'
>> Subject: Re: (j3.2006) a question on cobounds
>>
>> Loren P Meissner wrote:
>>> Can somebody come up with a better term than SYMMETRIC for this
>> concept?
>>> (Unless the term is already imbedded in the standard.) Do you mean
>>> something like "the array size is invariant across images"?
>>
>> The term "symmetric" appears only once in the standard, in  
>> reference to
>> files and units.
>>
>>> I always thought a "symmetric array" is supposed to be one where  
>>> A(I,
>> J)
>>> and A(J, I) have the same VALUE for all I and J.
>> An obvious potential source of confusion, especially if the reader  
>> has
>> a
>> background in mathematics.
>>
>> This terminology is
>>> obviously not intended here, so it needs to at least be explained
>> away
>>> if people continue to use it.
>
> 	FWIW, I had the same confusion that Loren did.
>
>> The use of "symmetric" is historical, dating back to the SHMEM  
>> library,
>> where memory was allocated on a "symmetric heap", with the property
>> that
>> the allocated space had the same base address on every processor.   
>> The
>> symmetry is in the location of the data, not its values. Coarrays
>> evolved out of this model and the terminology has stuck.  The  
>> standard
>> does not mandate that coarrays be stored at "symmetric" memory
>> addresses, though there are obvious advantages for a distributed  
>> memory
>> implementation to do that.
>>
>> Cheers,
>> Bill
>
> 	Thanks for the explanation of the origin of this use of "SYMMETRIC",
> Bill.  Unfortunately, like a lot of terms that software developers  
> use, this
> use of the word in this context is confusing and misleading.
>
> 	I firmly believe in straightforward terminology.  We should
> diligently search for a better term for this concept.  Here are a few
> suggestions.  Perhaps someone else will come up with a better term.
>
> 	COMPATIBLE - the coarrays are compatible with each other on all
> images.
>
> 	CONSISTENT - the coarrays are consistent with each other on all
> images.
>
> Anyone have any better ideas?
>
>>>
>>> LOREN
>>>
>>> -----Original Message-----
>>> From: j3-bounces at j3-fortran.org [mailto:j3-bounces at j3-fortran.org]  
>>> On
>>> Behalf Of Bill Long
>>> Sent: Tuesday, July 28, 2009 9:23 AM
>>> To: fortran standards email list for J3
>>> Subject: Re: (j3.2006) a question on cobounds
>>> [...]
>>>
>>> 2) Having the cobounds the same on all images enhances the concept
>> that
>>> THE ARRAYS ARE SYMMETRIC ACROSS IMAGES.
>>>
>>
>> --
>> 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
>>
>
> Sincerely,
> Craig T. Dedo
> 17130 W. Burleigh Place
> P. O. Box 423                  Mobile Phone:  (414) 412-5869
> Brookfield, WI   53008-0423    E-mail:  <craig at ctdedo.com>
> USA
> Linked-In:  http://www.linkedin.com/in/craigdedo
>
> _______________________________________________
> J3 mailing list
> J3 at j3-fortran.org
> http://j3-fortran.org/mailman/listinfo/j3

Thomas Clune					
Acting Head of the Software Integration and Visualization Office
NASA GSFC (610.3)
301-286-4635 (W)	240-266-0400 (F)
<Thomas.L.Clune at nasa.gov>