Hi Paul,
Thanks for your information. The amount of relative change in terms of
||delta U|| is about 1.0e-6 for my case with only 8 nodes. That's why I
doubt I may not setup the solver correctly.
If an iterative solver, what's the tolerance?
Are you asking how I solved x in Ax=b?
I just assembled the system:
{system.matrix->add_matrix (Ke, dof_indices);
system.rhs->add_vector (Fe, dof_indices);}
then called system.solve, where system is a <TransientLinearImplicitSystem>.
I did use the tolerance for the loop of iteration solver. it was 1.0e-5 for
the delta U, and the n-iteration-number = 1000;
The above mainly shows the difference of the first
iteration/calculation(Ax=b) between one processor and two processors.
By the way, What is the typical parameter setup I should use for the
iterative solver?
On Thu, May 22, 2014 at 8:21 AM, Paul T. Bauman <[email protected]> wrote:
> The amount of change your showing (< 1.0e-10) is not unexpected. What
> solver are you using? If an iterative solver, what's the tolerance? I
> suspect you are within solver tolerance. You cannot control the order of
> operations in MPI-based operations so you'll always have some difference
> near machine precision when running in parallel.
>
> Also recommend you update to the latest version of libMesh. 0.8.0 is
> getting a little long in the tooth...
>
> HTH,
>
> Paul
>
>
> On Thu, May 22, 2014 at 12:48 AM, walter kou <[email protected]>wrote:
>
>> By the way, this case only contains 8 elements...
>>
>>
>> On Wed, May 21, 2014 at 10:59 PM, walter kou <[email protected]>
>> wrote:
>>
>> > Deal all,
>> >
>> > I did a case on solving a LinearImplicitSystem for several steps, and I
>> > found the result for 2 processor are different with that for 1
>> processor:
>> > /***************/
>> > 1) case with one processor: ./exe_code
>> > final residual: 1.50996e-10 Nonlinear convergence: ||delta U|| =
>> 0.175641
>> >
>> > 2) case with 2 processors: mpiexec -np 2 ./exe_code
>> > final residual: 1.55163e-10 Nonlinear convergence: ||delta U|| =
>> 0.175642
>> >
>> > /*************/
>> >
>> > Does that necessarily mean my code has a bug?
>> > By the way, my version is 0.8.0
>> >
>> > Thanks,
>> >
>> > Walter
>> >
>>
>> ------------------------------------------------------------------------------
>> "Accelerate Dev Cycles with Automated Cross-Browser Testing - For FREE
>> Instantly run your Selenium tests across 300+ browser/OS combos.
>> Get unparalleled scalability from the best Selenium testing platform
>> available
>> Simple to use. Nothing to install. Get started now for free."
>> http://p.sf.net/sfu/SauceLabs
>> _______________________________________________
>> Libmesh-users mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/libmesh-users
>>
>
>
------------------------------------------------------------------------------
"Accelerate Dev Cycles with Automated Cross-Browser Testing - For FREE
Instantly run your Selenium tests across 300+ browser/OS combos.
Get unparalleled scalability from the best Selenium testing platform available
Simple to use. Nothing to install. Get started now for free."
http://p.sf.net/sfu/SauceLabs
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users
