[petsc-dev] Bug in MatMatMultTranspose for sequantial AIJ matrices from petsc-dev?

[Stefano Zampini]

Thank you. Using MatMatTransposeMult the code works.

You said this accident is historic, but I encountered no problems with
MatMatMultTranspose using previous PETSc versions (e.g 2.x, 3.0 and 3.1).
Weird, but maybe it's related to some properties of symmetric matrices with
which I used to work, and the fact the now I compiled petsc-dev with
errorchecking=1.

Again, thank you for the fast answer.
Stefano

2011/11/9 Dmitry Karpeev <karpeev at mcs.anl.gov>

> It seems to me the problem here is a historic accident:
> Originally MatMatMultTraspose did just what Stefano (and, a perhaps
> outdated, webpage) says: A^T*B.
> Then MatMatTransposeMult was introduced to do that, and
> MatMatMultTranspose was repurposed to do A*B^T.
> That's the state that Stefano's petsc-dev is probably in -- already
> inconsistent with that old webpage, but before
> the final renaming of the subroutines.
>
> Dmitry.
>
> On Wed, Nov 9, 2011 at 8:42 AM, Jed Brown <jedbrown at mcs.anl.gov> wrote:
>
>> On Wed, Nov 9, 2011 at 08:12, Stefano Zampini <stefano.zampini at gmail.com>wrote:
>>>
>>> I had some problems using MatMatMultTranspose for SeqAIJ matrices. It
>>> seems to me that MatMatMultTranspose(A,B,...,MAT_INITIAL_MATRIX,&C), which
>>> should perform C=A^T*B according to the web pages, doesn't its job.
>>
>>
>> petsc-dev has several new MatMatMult routines, so we had to rename some
>> of these functions for consistency. The function you name does not exist
>> any more, it is replaced by MatTransposeMatMult(), and there is a new
>> MatMatTransposeMult().
>>
>> http://lists.mcs.anl.gov/pipermail/petsc-dev/2011-November/006213.html
>>
>> Hong, please update src/docs/website/documentation/changes/dev.html when
>> you make API changes.
>>
>
>


-- 
Stefano
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20111109/0c74afd7/attachment.html>