fill for matrix multiplication

Barry Smith bsmith at
Sun Feb 18 12:34:47 CST 2007

 Use 1.0 for dense matrices; it is ignored since dense matrices
are always dense.


On Sun, 18 Feb 2007, Manav Bhatia wrote:

> Hi,
>  I am performing a matrix multiplication of two dense matrices with both
> MatMatMult and MatMatMultTranspose.
>  What do I choose a fill factor as? According to the definition in the
> documentation: fiill = expected fill as ratio of nnz(C)/(nnz(A) + nnz(B)).
> So, If I have two full matrices A and B, I will get a full matrix as my
> result. Hence, the fill factor will be 0.5. However, both these methods give
> an error with a fill factor less than 1.0. Also, if I use PETSC_DFAULT as the
> fill argument, it agains results in an error since its value is -2.
> Kindly help me with your advice here.
> Thanks
> Manav

More information about the petsc-users mailing list