Detailed Description

Sparsity pattern for the PETSc (parallel) build.

CSR representation with global column indices sorted within each local row. Used for type-agnostic preallocation via MATPREALLOCATOR. The index type is PetscInt, because that is the only type PETSc accepts.

Definition at line 42 of file SparsityPattern.h.

#include <SparsityPattern.h>

Public Member Functions
PetscInt	nnzInRow (PetscInt const row) const
	Number of nonzeros in the given local row.
PetscInt	numberOfRows () const
	Number of local rows the pattern describes.

Public Attributes
std::vector< PetscInt >	row_ptr
	CSR row pointers (length n_local_rows + 1).
std::vector< PetscInt >	col_idx
	Global column indices, sorted within each local row.

Member Function Documentation

◆ nnzInRow()

PetscInt MathLib::PETScSparsityPattern::nnzInRow ( PetscInt const row ) const

inline

Number of nonzeros in the given local row.

Definition at line 50 of file SparsityPattern.h.

    {
        return row_ptr[row + 1] - row_ptr[row];
    }

References row_ptr.

Referenced by MathLib::PETScMatrix::preallocateFromSparsityPattern().

◆ numberOfRows()

PetscInt MathLib::PETScSparsityPattern::numberOfRows ( ) const

inline

Number of local rows the pattern describes.

Definition at line 56 of file SparsityPattern.h.

    {
        // A default-constructed pattern has an empty row_ptr and describes
        // zero rows; only a populated pattern carries the trailing sentinel.
        return row_ptr.empty() ? 0 : static_cast<PetscInt>(row_ptr.size()) - 1;
    }