itaps-parallel Notes and assignments

[Tim Tautges]

As far as I can tell, this all depends on what your definition of "in" 
is, whether that includes just the objects in the original partition or 
whether it can include also entities adjacent to those in the original 
partition.  Surprisingly, though, it seems to me that either of those 
yield the same result.  In fact, I think this means that definition X+Y 
proposed by Onkar is correct.

- tim

Onkar Sahni wrote:
> Karen,
> 
> I would like to point out one premises for earlier definition of
> 'neighbors of a part', which included "objects" in its definition:
> 
> --------------------------------------------
> Objects exist only on one single part (lets avoid any ghosting).
> 
> Early definition X reads like does any object (on current part) have
> common adj. entity of type A with objects (on other parts) (for example,
> region1/object1 in part 1 and region2/object2 in part 2 have a common adj.
> face (type A=FACE)).
> 
> Early definition Y reads like does any entity of type A have objects adj.
> to it that lie on separate parts (for example, face (type A=FACE) has adj.
> regions/objects in parts 1 and 2).
> --------------------------------------------
> 
> If one does not consider "objects" and/or "rules" (basically fact that
> "objects" exist only on one single part) into these definitions and just
> considers arbitrary entity-types A and B then both definition can be one,
> see below.
> 
> Definition X + Y: Two (or more) parts are neighbors, if any entity of type
> A has at least one adjacent entity of type B in each of the parts (type A
> and type B are inputs).
> 
> For example,
> **1) type A = REGION, type B = FACE (does any region has an adj. face that
> exists on parts 1 and 2).
> 
> **2) type A = FACE, type B = REGION (does any face has adj. regions in
> parts 1 and 2).
> 
> As of now I think we need to put some restriction on one of entity-type
> and if one does that then again there are two ways to see 'neighbors of a
> part' (like def. X and def. Y).
> 
> Please let me know if you have questions.
> 
> - Onkar
> 
> 
>> OK, we're in the home stretch in preparation for bootcamp.
>>
>> Our next phone conf is scheduled 2/28 at 1pm PST.  The biggest issues
>> still
>> in question are (1) copies and communication for them, (2) communication
>> of
>> tag info, and (3) what should be included in prefix_UpdatePartitionPar.
>>
>> We probably have only two phone confs left before bootcamp, so please
>> schedule 1.5 hours for the call on 2/28.  I'd like to make significant
>> progress on (1) and (2).
>>
>> Assignments:
>>
>> Carl & Onkar:  Revise the functions for copies and non-collective
>> communication.
>>
>> Tim:  Submit functions for exchanging tag data.
>>
>> All:
>> -  Send a list of data you'd like to have pre-computed in
>> prefix_UpdatePartitionPar (allowing other functions to work without
>> communication).
>>
>> -  Vote for your favorite definition of part neighbor (or submit your
>> objections to both along with a definition that you like better):
>>
>>    Definition X:  Two (or more) parts are neighbors with respect to entity
>> type A if at least one entity in each of the parts has a common lower- or
>> higher-dimensional adjacent entity of specified entity type A.
>>
>>    Definition Y:  Two (or more) parts are neighbors with respect to entity
>> type A if any entity of specified entity type A has at least one adjacent
>> higher- or lower-dimensional entity in each of the parts.
>>
>> -  Read the attached updated DraftInterface.h, paying attention to items
>> marked TODO.
>>
>>
> 
> 
> 

-- 
================================================================
"You will keep in perfect peace him whose mind is
   steadfast, because he trusts in you."               Isaiah 26:3

              Tim Tautges            Argonne National Laboratory
          (tautges at mcs.anl.gov)      (telecommuting from UW-Madison)
          phone: (608) 263-8485      1500 Engineering Dr.
            fax: (608) 263-4499      Madison, WI 53706