> On Jul 10, 2018, at 11:50 PM, Marius Buerkle <mbuer...@web.de> wrote:
> 
> Sorry, I have yet another question. While it works well for MPIAIJ, I end up 
> in a deadlock if I use MatCreateSubmatrix on a MPIDENSE matrix. Should 
> MatCreateSubmatrix work on dense matricies ?

yes

> On Mon, Jul 9, 2018 at 10:02 PM Marius Buerkle <mbuer...@web.de> wrote:
> MatGetSubmatrix is in the current PETSc released called MatCreateSubmatrix. 
> Is this correct ?
>  
>  
> Yep. I am old and cannot follow API changes anymore.
>  
>   Matt
>   
> On Mon, Jul 9, 2018 at 8:38 PM Marius Buerkle <mbuer...@web.de> wrote:
>  I see. What I want to do is to calculate the matrix product C=A*B' between 
> two sparse matrices A=(A11 0 , A21 0) and B=(B11 0 , B21 0)  where C will be 
> dense in the end but I just want to calculate some selected entries C_ij  of 
> C. At the moment I extract submatricies for the corresponding rows and 
> columns,
>  
> I think that is the right way, but you should only need MatGetSubmatrix for 
> that.
>  
>   Thanks,
>  
>      Matt
>  
> so I was wondering if there is a simpler or performancer-wise faster way. I 
> assume there is not such thing as a restricted MatMatMul which just 
> calculated the lets say predefined nonzero entries of C.
>  
> 
> These are "internal" routines that we rarely expect end users to use since 
> they are specific for particular matrix implementations. As such they are 
> also kind of strange to use from Fortran since none of the AIJ data 
> structures can be made visible to Fortran.
> 
> Could you explain why you want them from Fortran and maybe we'll have 
> alternative suggestions on how you can achieve the same effect.
> 
> Barry
> 
> 
> > On Jul 5, 2018, at 3:05 AM, Marius Buerkle <mbuer...@web.de> wrote:
> >
> > or MatMPIAIJGetLocalMatCondensed for that matter.
> >
> >
> >
> > Hi !
> >
> > Is MatMPIAIJGetSeqAIJ implemented for fortran?
> >
> > best,
> > Marius
>  
>  
>  
> --
> 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
>  
> https://www.cse.buffalo.edu/~knepley/
>  
>  
> --
> 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
>  
> https://www.cse.buffalo.edu/~knepley/

Reply via email to