Worthwhile to use PETSc to solve lots of small matrices?

[Barry Smith]

  Don't use PETSc


On Sun, 17 Jun 2007, Ben Tay wrote:

> Hi,
> 
> I am doing CFD and at each time step, I need to interpolate values at
> different points. I'm doing bi-quadratic interpolation and hence to get each
> interpolated value, I need to solve a 9x9 matrices. There's about 100+ pts
> and hence 100+ 9x9 matrices to solve per time step.
> 
> I'm wondering if it is worthwhile to use PETSc to solve them or am I better
> off using those direct guassian elimination solvers?
> 
> Thanks
>