[petsc-users] Domain Decomposition Method for Parallel FEM Code

[Matthew Knepley]

On Wed, Feb 11, 2015 at 2:07 PM, Hanglin Ye <hanglinye at gmail.com> wrote:

> Hi,
> Thank you again for the reply. I am trying to run ex62. But there is an
> error :  "Mesh generation needs external package support. Please
> reconfigure with --download-triangle." I then configure again with
> --download triangle, and it is download and installed. But this error keeps
> showing up. Could you please let me know if I am missing anything?
>

Always send the entire error text. It will show your configure line, and
lots of other information.

  Thanks,

    Matt


> Thank you.
>
> On Wed, Feb 11, 2015 at 1:05 PM, Matthew Knepley <knepley at gmail.com>
> wrote:
>
>> On Wed, Feb 11, 2015 at 12:01 PM, Hanglin Ye <hanglinye at gmail.com> wrote:
>>
>>> Thank you very much for the reply.
>>>
>>> I am not very sure if structured/unstructured mesh means differently in
>>> PETSc, but my mesh are simply hexahedral mesh, which I assume is
>>> structured. Is there any difference when dealing with structured and
>>> unstructured mesh?
>>>
>>
>>
>>> And another question is: Can I say that I only need to provide the whole
>>> domain, and PETSc can take care of the decomposition, so that I do not need
>>> to use software such as Metis to partition the mesh in advance?
>>>
>>
>> Now your mesh sounds unstructured. SNES ex62 is an example of a finite
>> element code using DMPlex
>> which can have tetrahedral or hexahedral cells, and solves Stokes
>> equation.
>>
>> PETSc can handle partitioning and distribution for you.
>>
>>   Thanks,
>>
>>     Matt
>>
>>
>>> Thank you.
>>>
>>>
>>> On Feb 11, 2015, at 12:01, Matthew Knepley <knepley at gmail.com> wrote:
>>>
>>> On Wed, Feb 11, 2015 at 10:52 AM, Hanglin Ye <hanglinye at gmail.com>
>>> wrote:
>>>
>>>> Dear all,
>>>>
>>>> I am new to PETSc and I want to use it to parallel my current serial
>>>> FEM code. I want to use Domain Decomposition Method so that the whole FEM
>>>> domain is partitioned into sub-domains and computations are performed in
>>>> sub-domains then assemble together. Is there a way to use PETSc to realize
>>>> that ?
>>>>
>>>> I've been searching through tutorials but none seems to be clear about
>>>> this aspect. I mainly wish to know what solver of PETSc do I need.
>>>>
>>>
>>> Do you have a structured or unstructured mesh? In either case, you will
>>> use a DM to encapsulate your mesh, which will give you
>>> DMLocalToGlobal() and DMGlobalToLocal() to map between Vecs which are
>>> appropriate for the solver (global) and those with
>>> ghost regions which are appropriate for assembly (local). You can see an
>>> example in SNES ex5, ex12, and ex19.
>>>
>>>  Thanks,
>>>
>>>    Matt
>>>
>>>
>>>> Thank you very much.
>>>> --
>>>> Hanglin Ye
>>>> Ph.D.Student MANE, RPI
>>>>
>>>
>>>
>>>
>>> --
>>> What most experimenters take for granted before they begin their
>>> experiments is infinitely more interesting than any results to which their
>>> experiments lead.
>>> -- Norbert Wiener
>>>
>>>
>>
>>
>> --
>> What most experimenters take for granted before they begin their
>> experiments is infinitely more interesting than any results to which their
>> experiments lead.
>> -- Norbert Wiener
>>
>
>
>
> --
> Hanglin Ye
> Ph.D.Student MANE, RPI
>



-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20150211/48193b17/attachment.html>