#include "petscmat.h" PetscErrorCode MatSetValuesBlockedStencil(Mat mat,PetscInt m,const MatStencil idxm[],PetscInt n,const MatStencil idxn[],const PetscScalar v[],InsertMode addv)Not Collective
mat | - the matrix | |
m | - number of rows being entered | |
idxm | - grid coordinates for matrix rows being entered | |
n | - number of columns being entered | |
idxn | - grid coordinates for matrix columns being entered | |
v | - a logically two-dimensional array of values | |
addv | - either ADD_VALUES or INSERT_VALUES, where ADD_VALUES adds values to any existing entries, and INSERT_VALUES replaces existing entries with new values |
Calls to MatSetValuesBlockedStencil() with the INSERT_VALUES and ADD_VALUES options cannot be mixed without intervening calls to the assembly routines.
The grid coordinates are across the entire grid, not just the local portion
MatSetValuesBlockedStencil() uses 0-based row and column numbers in Fortran as well as in C.
For setting/accessing vector values via array coordinates you can use the DMDAVecGetArray() routine
In order to use this routine you must either obtain the matrix with DMCreateMatrix() or call MatSetBlockSize(), MatSetLocalToGlobalMapping() and MatSetStencil() first.
The columns and rows in the stencil passed in MUST be contained within the ghost region of the given process as set with DMDACreateXXX() or MatSetStencil(). For example, if you create a DMDA with an overlap of one grid level and on a particular process its first local nonghost x logical coordinate is 6 (so its first ghost x logical coordinate is 5) the first i index you can use in your column and row indices in MatSetStencil() is 5.
In Fortran idxm and idxn should be declared as
MatStencil idxm(4,m),idxn(4,n)and the values inserted using
idxm(MatStencil_i,1) = i
idxm(MatStencil_j,1) = j
idxm(MatStencil_k,1) = ketc
Negative indices may be passed in idxm and idxn, these rows and columns are simply ignored. This allows easily inserting element stiffness matrices with homogeneous Dirchlet boundary conditions that you don't want represented in the matrix.
Inspired by the structured grid interface to the HYPRE package (https://computation.llnl.gov/projects/hypre-scalable-linear-solvers-multigrid-methods)