linalg.topetsc#

Functions#

`get_index_shift_per_block_per_process`(V)
`get_npts_local`(V)	Compute the local number of nodes per dimension owned by the actual process.
`get_npts_per_block`(V)	Compute the number of nodes per block, process and dimension.
`mat_topetsc`(mat)	Convert operator from Psydac format to a PETSc.Mat object.
`petsc_local_to_psydac`(V, petsc_index)	Convert the PETSc local index (starting from 0 in each process) to a Psydac local index (natural multi-index, as grid coordinates).
`psydac_to_petsc_global`(V, block_indices, ...)	Convert the Psydac local index (natural multi-index, as grid coordinates) to a PETSc global index.
`toIJVrowmap`(mat_block, bd, bc, I, J, V, ...)
`vec_topetsc`(vec)	Convert vector from Psydac format to a PETSc.Vec object.

Details#

petsc_local_to_psydac(V: VectorSpace, petsc_index: int)[source]#

Convert the PETSc local index (starting from 0 in each process) to a Psydac local index (natural multi-index, as grid coordinates).

Parameters:

VVectorSpace: The vector space to which the Psydac vector belongs. This defines the number of blocks, the size of each block, and how each block is distributed across MPI processes.
petsc_indexint: The local PETSc index. The 0 index is only owned by every process.

Returns:

block: tuple: The block where the Psydac multi-index belongs to.
psydac_indextuple: The Psydac local multi-index. This index is local the block.

psydac_to_petsc_global(V: VectorSpace, block_indices, ndarray_indices) → int[source]#

Convert the Psydac local index (natural multi-index, as grid coordinates) to a PETSc global index. Performs a search to find the process owning the multi-index.

Parameters:

VVectorSpace: The vector space to which the Psydac vector belongs. This defines the number of blocks, the size of each block, and how each block is distributed across MPI processes.
block_indicestuple[int]: The indices which identify the block in a (possibly nested) block vector. In the case of a StencilVector this is an empty tuple.
ndarray_indicestuple[int]: The multi-index which identifies an element in the _data array, excluding the ghost regions.

Returns:

petsc_indexint: The global PETSc index. The 0 index is only owned by the first process.

get_npts_local(V: VectorSpace) → list[source]#

Compute the local number of nodes per dimension owned by the actual process. This is a local variable, its value will be different for each process.

Returns:

list: Local number of nodes per dimension owned by the actual process. In case of a StencilVectorSpace the list contains a single list with length equal the number of dimensions in the domain. In case of a BlockVectorSpace the list has length equal the number of blocks.

get_npts_per_block(V: VectorSpace) → list[source]#

Compute the number of nodes per block, process and dimension. This is a global variable, its value is the same for all processes.

Returns:

list: Number of nodes per block, process and dimension.

vec_topetsc(vec)[source]#

Convert vector from Psydac format to a PETSc.Vec object.

Parameters:

vecpsydac.linalg.stencil.StencilVector | psydac.linalg.block.BlockVector: Psydac StencilVector or BlockVector. In the case of a BlockVector, only the case where the blocks are StencilVector is implemented.

Returns:

gvecPETSc.Vec: PETSc vector

mat_topetsc(mat)[source]#

Convert operator from Psydac format to a PETSc.Mat object.

Parameters:

matpsydac.linalg.stencil.StencilMatrix | psydac.linalg.block.BlockLinearOperator: Psydac operator. In the case of a BlockLinearOperator, only the case where the blocks are StencilMatrix is implemented.

Returns:

gmatPETSc.Mat: PETSc Matrix