Hi Roy, The advise to use current local solution instead of solution helped in fixing the physics, however the problem still remains with the averaging routine. Will see if I can fix that. Thanks, Subramanya
> From: [email protected] > To: [email protected]; [email protected] > Date: Wed, 10 Jul 2013 12:17:05 -0400 > Subject: Re: [Libmesh-users] Weird trouble with running in parallel. > > Hi Roy, > I replaced solution with current local solution and it still gives me the > same error. The weird thing is, I have the exact same approach to average > stresses and strains to the nodes and that works without a hitch.. There are > a couple of seemingly minor differences. I will try and change the code for > this part so that it matches the other more closely. Thanks, Subramanya > > From: [email protected] > To: [email protected]; [email protected] > Date: Wed, 10 Jul 2013 12:05:32 -0400 > Subject: Re: [Libmesh-users] Weird trouble with running in parallel. > > Hi Roy, Thanks, I will change this and see if it helps. The problem with the > previous one was that the discontinuity was not across elements, but that my > solution was markedly discontinuous across the mesh partitions, not across > the element boundaries. I have attached pictures , this should make it clear > what I was referring to. this is for 2 processors, If I run this code on 4 > partitions I get four such striations on the chemical potential plot. The > white line is a void, where the material properties are nearly 0. Thanks, > Subramanya > > > > Date: Wed, 10 Jul 2013 10:51:46 -0500 > > From: [email protected] > > To: [email protected] > > CC: [email protected] > > Subject: Re: [Libmesh-users] Weird trouble with running in parallel. > > > > > > On Wed, 10 Jul 2013, subramanya sadasiva wrote: > > > > > I have some weird trouble with some parallel vectors. I am trying to > > > compute the average value of the gradient of a field at the nodes. > > > > This makes a little more sense. Your previous post just asked whether > > the gradient would be the same on both sides of a partition, to which > > the answer on C0 elements is "of course not, gradients are > > discontinuous between elements". But if you're projecting the > > gradients onto a continuous space then your only problem is making > > sure they get communicated correctly. > > > > > However, when I run this in parallel, I get the following error. > > > > > Vector contains [9,15)And empty ghost array. > > > > This is just diagnostics following the "no index _ in current vector" > > error message. Basically some processor is trying to either read or > > set a value that doesn't live on it and doesn't have a local copy. > > That "empty ghost array" bit means that *no* remote value has a local > > copy, which for your algorithm is guaranteeing you problems. You > > probably need to be dealing with current_local_solution instead of > > just solution on your ExplicitSystem; the local solution should have a > > ghost array sufficient for your stencil. > > --- > > Roy > > > ------------------------------------------------------------------------------ > See everything from the browser to the database with AppDynamics > Get end-to-end visibility with application monitoring from AppDynamics > Isolate bottlenecks and diagnose root cause in seconds. > Start your free trial of AppDynamics Pro today! > http://pubads.g.doubleclick.net/gampad/clk?id=48808831&iu=/4140/ostg.clktrk > _______________________________________________ > Libmesh-users mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/libmesh-users > > ------------------------------------------------------------------------------ > See everything from the browser to the database with AppDynamics > Get end-to-end visibility with application monitoring from AppDynamics > Isolate bottlenecks and diagnose root cause in seconds. > Start your free trial of AppDynamics Pro today! > http://pubads.g.doubleclick.net/gampad/clk?id=48808831&iu=/4140/ostg.clktrk > _______________________________________________ > Libmesh-users mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/libmesh-users ------------------------------------------------------------------------------ See everything from the browser to the database with AppDynamics Get end-to-end visibility with application monitoring from AppDynamics Isolate bottlenecks and diagnose root cause in seconds. Start your free trial of AppDynamics Pro today! http://pubads.g.doubleclick.net/gampad/clk?id=48808831&iu=/4140/ostg.clktrk _______________________________________________ Libmesh-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-users
