On Wed, 12 Dec 2012, Ataollah Mesgarnejad wrote:

> On Dec 12, 2012, at 3:41 PM, Roy Stogner <[email protected]> wrote:
>
>> Yes - that would only have worked if you'd already manually
>> initialized every system and variable to match the solution file and
>> then init()ed it.

> I did this but still had the problem.

Define "the problem"?  The new vector had a size mismatch?  Wasn't
ghosted?  Was ghosted but didn't give you correct results in parallel?

This is the workaround I tried in a test app code myself; it worked
fine there.

>> If that's not possible then maybe you really will need to manually
>> upgrade the read() results to a GHOSTED vector.  That would mean
>> NumericVector::build(), init() of that vector with the send_list()
>> from your DofMap and the n,n_local sizes from the vector you're
>> replacing, and finally old_vec.swap(new_vec).

> Is there any place in /src or /examples that you do this so I can
> follow the steps?

Not that I know of, and I certainly hope not.  If we'd ever
encountered this problem before I'd like to believe we'd have found
time to think and talk about and fix it by now, not just put in a
workaround.

Although, now that we *are* thinking and talking about it, there may be
a simpler workaround if you change library code.  You could try
changing "this->add_vector(vec_name)" in system_io.C to
"this->add_vector(vec_name,true,GHOSTED)".  That's too inefficient to
put in the library for everyone, but it might be the fastest way for
you to get up and running.
---
Roy

------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users

Reply via email to