In looking over the use cases I do not see anything in them that introduces anything that is beyond things that have to be done is stuff we are already doing for parallel mesh adaptation (including boundary layers). To the best of my understanding the questions we have have been discussing really do not influence on how to attack these problems. The issues we have been discussing continue to be of importance to how efficient implementations can be. Note - writing up the details of doing the use cases would be pretty extensive and time consuming. Mark [email protected] wrote:
Karen,
The compromise proposal will work fine with the FMDB.
One question is that the use case, efficiency, and the ease of handling of the entity ownership if we allow the user to specify the owner part of mesh entity.
It's not clear to me what's our scenario between #1 and #2.
#1 we assume that the owner part of the entity is automatically determined at time of entity creation by default and will allow some exceptional case where the user can specify the owner part for specific entities
#2 for every entity, the owner part shall be specified explicitly.
In case of the FMDB, the owner part is determined based on the partition model classification to which each entity belongs so the application has no right to change the entity ownership at the moment.
Seegyoung
Since we rescheduled the iMeshP conference call (now Thursday, May 27 at 8am PDT), I'll ask you to do some extra work. Please consider the attached use cases (drafted by Tim -- thanks, Tim!). For each case, consider whether the proposal works, particularly in the context of your implementation. We'll discuss the proposal and use cases on Thursday.
Thanks. Karen