[petsc-users] grid sequencing & DMComposite

[Matthew Knepley]

On Fri, Sep 11, 2015 at 8:37 AM, Gideon Simpson <gideon.simpson at gmail.com>
wrote:

> There doesn’t seem to be a global analog of DMCompositeGetLocalVectors.
> Do I need to do manual indexing?
>

Its simple, lets just write it:


http://www.mcs.anl.gov/petsc/petsc-current/src/dm/impls/composite/pack.c.html#DMCompositeGetLocalVectors

and from looking at the code here


http://www.mcs.anl.gov/petsc/petsc-current/src/dm/impls/composite/pack.c.html#DMCompositeScatter

All we have to do is write one function which gives back a bunch of Global
vectors pointing at the same memory.
We use PlaceArray() on the Get call, and ResetArray() on the Restore call.

I can't believe this isn't in there already, but I think O(0) people use
DMComposite right now.

   Matt


> -gideon
>
> On Sep 11, 2015, at 9:36 AM, Matthew Knepley <knepley at gmail.com> wrote:
>
> On Fri, Sep 11, 2015 at 8:15 AM, Gideon Simpson <gideon.simpson at gmail.com>
> wrote:
>
>> Ok, so here’s some behavior I don’t understand.  I am working with a
>> DMComposite structure, and I do the following,
>>
>>     SNESComputeFunction(snes,U_refine,r);
>>     SNESGetDM(snes, &dm_refine);
>>
>>     DMCompositeGetLocalVectors(dm_refine, &rp, &rQ);
>>     DMCompositeScatter(dm_refine, r, rp, rQ);
>>
>>     VecNorm(r, NORM_2, &r_norm);
>>     VecNorm(rp, NORM_2, &rp_norm);
>>     VecNorm(rQ, NORM_2, &rQ_norm);
>>
>>     VecGetSize(r,&r_size);
>>     VecGetSize(rp, &rp_size);
>>     VecGetSize(rQ, &rQ_size);
>>
>>     DMCompositeRestoreLocalVectors(dm_refine, &rp, &rQ);
>>     VecDestroy(&r);
>>
>>     PetscPrintf(PETSC_COMM_WORLD," ||r|| = %g, %i
>> entries\n",r_norm,r_size);
>>     PetscPrintf(PETSC_COMM_WORLD," ||rp|| = %g, %i
>> entries\n",rp_norm,rp_size);
>>     PetscPrintf(PETSC_COMM_WORLD," ||rQ|| = %g, %i
>> entries\n",rQ_norm,rQ_size);
>>
>> and my output is:
>>
>>  ||r|| = 225.31, 7999 entries
>>  ||rp|| = 140.021, 3 entries
>>  ||rQ|| = 176.56, 8004 entries
>>
>>
>> The arithmetic, is off, no?
>>
>
> I think you want Global vectors for this, not Local vectors.
>
>   Matt
>
>
>> -gideon
>>
>> On Sep 10, 2015, at 5:28 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:
>>
>>
>>  SNESGetDM()  this will return not the original DM you set but the
>> refined one.
>>
>> On Sep 10, 2015, at 3:47 PM, Gideon Simpson <gideon.simpson at gmail.com>
>> wrote:
>>
>> I’m using a DMCompsosite along with grid sequencing and here’s what I’m
>> trying to accomplish.  After running the SNES solve, I’d like to evaluate
>> the residual on the refined grid, on each piece of the DMComposite.  How do
>> I get a DM for the refined grid which I can then use with
>> DMCompositeGetLocalVectors in order to get each piece of the problem?  Or
>> is there a better way?
>>
>> -gideon
>>
>>
>>
>>
>
>
> --
> 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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20150911/322a7298/attachment.html>