From: Kirk, B. (JSC-EG311) <ben...@na...> - 2011-01-20 19:44:28
|
Anyone out there running with subdomain restricted variables in parallel? I'm pretty sure I've found a bug in the DOF distribution code but it seems like it should be hurting everyone using this feature... Specifically, in dof_map.C, we loop over variables, then elements. If the variable is active on the subdomain we then loop over the nodes of that element and assign DOF indices *if we own the node*. Unfortunately, this misses a corner case in which the processor which actually owns the node does not own an element which is active on the relevant subdomain. -Ben |
From: Roy S. <roy...@ic...> - 2011-01-20 20:06:37
|
On Thu, 20 Jan 2011, Kirk, Benjamin (JSC-EG311) wrote: > Anyone out there running with subdomain restricted variables in parallel? > I'm pretty sure I've found a bug in the DOF distribution code but it seems > like it should be hurting everyone using this feature... > > Specifically, in dof_map.C, we loop over variables, then elements. If the > variable is active on the subdomain we then loop over the nodes of that > element and assign DOF indices *if we own the node*. Unfortunately, this > misses a corner case in which the processor which actually owns the node > does not own an element which is active on the relevant subdomain. I wouldn't be too surprised if people have been using that feature combination but missing that corner case - it sounds an awful lot like a few old parallel adaptivity bugs that most apps managed to avoid simply by not having just the wrong combination of partitioning and hanging nodes. --- Roy |
From: Derek G. <fri...@gm...> - 2011-01-20 20:09:32
|
It's funny... we do have a code here that is using subdomain restricted variables daily.... and they've never been able to get it to work properly in parallel. They've always thought it was poor preconditioning in parallel (it's a very stiff system requiring some significant preconditioning).... but it could be that they are running into this bug! For quite a while they've just been working in serial.... Have you fixed the bug... or is there a straightforward way? I'm going to be using this capability extensively soon (like tomorrow)... so it would be good to know how it stands... Derek On Jan 20, 2011, at 12:44 PM, Kirk, Benjamin (JSC-EG311) wrote: > Anyone out there running with subdomain restricted variables in parallel? > I'm pretty sure I've found a bug in the DOF distribution code but it seems > like it should be hurting everyone using this feature... > > Specifically, in dof_map.C, we loop over variables, then elements. If the > variable is active on the subdomain we then loop over the nodes of that > element and assign DOF indices *if we own the node*. Unfortunately, this > misses a corner case in which the processor which actually owns the node > does not own an element which is active on the relevant subdomain. > > -Ben > > > ------------------------------------------------------------------------------ > Protect Your Site and Customers from Malware Attacks > Learn about various malware tactics and how to avoid them. Understand > malware threats, the impact they can have on your business, and how you > can protect your company and customers by using code signing. > http://p.sf.net/sfu/oracle-sfdevnl > _______________________________________________ > Libmesh-devel mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libmesh-devel |
From: Kirk, B. (JSC-EG311) <ben...@na...> - 2011-01-20 20:32:26
|
> Have you fixed the bug... or is there a straightforward way? I'm going to be > using this capability extensively soon (like tomorrow)... so it would be good > to know how it stands... I just checked in a change which should fix the problem... I think that should work for you now. -Ben |
From: Derek G. <fri...@gm...> - 2011-01-20 20:35:44
|
On Jan 20, 2011, at 1:32 PM, Kirk, Benjamin (JSC-EG311) wrote: > I just checked in a change which should fix the problem... I think that > should work for you now. Thanks! Derek |