[petsc-dev] DMGetMatrix --> DMGetMatrices?

[Dmitry Karpeev]

In a somewhat related matter, it appears that I cannot duplicate a
preallocated MATXXXAIJ until it has been assembled:
if my DM implementation keeps a preallocated MATSEQAIJ, which it wants to
duplicate on every call to DMGetMatrix,
it would have to put in fake entries before any duplication is possible.

Dmitry.

On Thu, Feb 9, 2012 at 11:51 PM, Dmitry Karpeev <karpeev at mcs.anl.gov> wrote:

> Good point.  It is still strange, however, that SNES insists on using the
> same Mat obtained from DMGetMatrix for J and P.
> Dmitry.
> On Thu, Feb 9, 2012 at 11:46 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:
>
>> On Thu, Feb 9, 2012 at 23:41, Dmitry Karpeev <karpeev at mcs.anl.gov> wrote:
>>
>>> Currently DMGetMatrix(DM dm, MatType mattype , Mat *J)  is supposed to
>>> produce the system Jacobian, according to the docs.
>>> SNES will use it, but, since DMGetMatrix returns a single matrix, it
>>> can't use a different preconditioning matrix.
>>> Why not change this to DMGetMatrices(DM dm, MatType jtype, Mat *J,
>>> MatType  ptype, Mat *P)?
>>> I'm not exactly sure what the best way of handling MatType should be
>>> (same MatType for J and P, separate types, how dm->mattype
>>> affects the result), but I think it can be sorted out, whereas the
>>> absence of P is a clear limitation.
>>>
>>
>> Umm, this doesn't put in values, it just creates a matrix of the right
>> size. You can make two if you want two different matrices.
>>
>> If you are talking about DMDAFormJacobianLocal(), that does actually
>> restrict what you can do, and you have to hack it a bit to use two
>> different matrices (except by MFFD because the library specifically
>> supports that).
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20120209/090cd272/attachment.html>