Source code for dolfinx.la.petsc

# Copyright (C) 2025 Garth N. Wells, Jack S. Hale
#
# This file is part of DOLFINx (https://www.fenicsproject.org)
#
# SPDX-License-Identifier:    LGPL-3.0-or-later
"""Functions for working with PETSc linear algebra objects.

Note:
    Due to subtle issues in the interaction between petsc4py memory
    management and the Python garbage collector, it is recommended that
    the PETSc method ``destroy()`` is called on returned PETSc objects
    once the object is no longer required. Note that ``destroy()`` is
    collective over the object's MPI communicator.
"""

import functools
import itertools
import typing
from collections.abc import Iterable, Sequence

from petsc4py import PETSc

import numpy as np
import numpy.typing as npt

import dolfinx
from dolfinx.common import IndexMap
from dolfinx.la import Vector

assert dolfinx.has_petsc4py

__all__ = ["assign", "create_vector", "create_vector_wrap"]


def _ghost_update(x: PETSc.Vec, insert_mode: PETSc.InsertMode, scatter_mode: PETSc.ScatterMode):  # type: ignore[name-defined]
    """Helper function for ghost updating PETSc vectors"""
    if x.getType() == PETSc.Vec.Type.NEST:  # type: ignore[attr-defined]
        for x_sub in x.getNestSubVecs():
            x_sub.ghostUpdate(addv=insert_mode, mode=scatter_mode)
            x_sub.destroy()
    else:
        x.ghostUpdate(addv=insert_mode, mode=scatter_mode)


def _zero_vector(x: PETSc.Vec):  # type: ignore[name-defined]
    """Helper function for zeroing out PETSc vectors"""
    if x.getType() == PETSc.Vec.Type.NEST:  # type: ignore[attr-defined]
        for x_sub in x.getNestSubVecs():
            with x_sub.localForm() as x_sub_local:
                x_sub_local.set(0.0)
            x_sub.destroy()
    else:
        with x.localForm() as x_local:
            x_local.set(0.0)


[docs] def create_vector_wrap(x: Vector) -> PETSc.Vec: # type: ignore[name-defined] """Wrap a distributed DOLFINx vector as a PETSc vector. Args: x: The vector to wrap as a PETSc vector. Returns: A PETSc vector that shares data with ``x``. """ index_map = x.index_map ghosts = index_map.ghosts.astype(PETSc.IntType) # type: ignore[attr-defined] bs = x.block_size size = (index_map.size_local * bs, index_map.size_global * bs) return PETSc.Vec().createGhostWithArray( # type: ignore[attr-defined] ghosts, x.array, size=size, bsize=bs, comm=index_map.comm )
[docs] def create_vector( maps: typing.Sequence[tuple[IndexMap, int]], kind: str | None = None ) -> PETSc.Vec: # type: ignore[name-defined] """Create a PETSc vector from a sequence of maps and blocksizes. Three cases are supported: 1. If ``maps=[(im_0, bs_0), ..., (im_n, bs_n)]`` is a sequence of indexmaps and blocksizes and ``kind`` is ``None``or is ``PETSc.Vec.Type.MPI``, a ghosted PETSc vector whith block structure described by ``(im_i, bs_i)`` is created. The created vector ``b`` is initialized such that on each MPI process ``b = [b_0, b_1, ..., b_n, b_0g, b_1g, ..., b_ng]``, where ``b_i`` are the entries associated with the 'owned' degrees-of- freedom for ``(im_i, bs_i)`` and ``b_ig`` are the 'unowned' (ghost) entries. If more than one tuple is supplied, the returned vector has an attribute ``_blocks`` that holds the local offsets into ``b`` for the (i) owned and (ii) ghost entries for each ``V[i]``. It can be accessed by ``b.getAttr("_blocks")``. The offsets can be used to get views into ``b`` for blocks, e.g.:: >>> offsets0, offsets1, = b.getAttr("_blocks") >>> offsets0 (0, 12, 28) >>> offsets1 (28, 32, 35) >>> b0_owned = b.array[offsets0[0]:offsets0[1]] >>> b0_ghost = b.array[offsets1[0]:offsets1[1]] >>> b1_owned = b.array[offsets0[1]:offsets0[2]] >>> b1_ghost = b.array[offsets1[1]:offsets1[2]] 2. If ``V=[(im_0, bs_0), ..., (im_n, bs_n)]`` is a sequence of function space and ``kind`` is ``PETSc.Vec.Type.NEST``, a PETSc nested vector (a 'nest' of ghosted PETSc vectors) is created. Args: map: Sequence of tuples of ``IndexMap`` and the associated block size. kind: PETSc vector type (``VecType``) to create. Returns: A PETSc vector with the prescribed layout. The vector is not initialised to zero. """ if len(maps) == 1: # Single space case index_map, bs = maps[0] ghosts = index_map.ghosts.astype(PETSc.IntType) # type: ignore[attr-defined] size = (index_map.size_local * bs, index_map.size_global * bs) b = PETSc.Vec().createGhost(ghosts, size=size, bsize=bs, comm=index_map.comm) # type: ignore if kind == PETSc.Vec.Type.MPI: # type: ignore[attr-defined] _assign_block_data(maps, b) return b if kind is None or kind == PETSc.Vec.Type.MPI: # type: ignore[attr-defined] b = dolfinx.cpp.fem.petsc.create_vector_block(maps) _assign_block_data(maps, b) return b elif kind == PETSc.Vec.Type.NEST: # type: ignore[attr-defined] return dolfinx.cpp.fem.petsc.create_vector_nest(maps) else: raise NotImplementedError( "Vector type must be specified for blocked/nested assembly." f"Vector type '{kind}' not supported." "Did you mean 'nest' or 'mpi'?" )
[docs] @functools.singledispatch def assign( x0: npt.NDArray[np.inexact] | Sequence[npt.NDArray[np.inexact]], x1: PETSc.Vec, # type: ignore[name-defined] ): """Assign ``x0`` values to a PETSc vector ``x1``. Values in ``x0``, which is possibly a stacked collection of arrays, are assigned ``x1``. When ``x0`` holds a sequence of ``n``` arrays and ``x1`` has type ``NEST``, the assignment is:: [x0[0]] x1 = [x0[1]] [.....] [x0[n-1]] When ``x0`` holds a sequence of ``n`` arrays and ``x1`` **does not** have type ``NEST``, the assignment is:: [x0_owned[0]] x1 = [.....] [x0_owned[n-1]] [x0_ghost[0]] [.....] [x0_ghost[n-1]] Args: x0: An array or list of arrays that will be assigned to ``x1``. x1: Vector to assign values to. """ if x1.getType() == PETSc.Vec.Type().NEST: # type: ignore[attr-defined] x1_nest = x1.getNestSubVecs() for _x0, _x1 in zip(x0, x1_nest): with _x1.localForm() as x: x.array_w[:] = _x0 else: with x1.localForm() as _x: if isinstance(x0, Sequence): start = 0 for _x0 in x0: end = start + _x0.shape[0] _x.array_w[start:end] = _x0 start = end else: _x.array_w[:] = x0
@assign.register def _( x0: PETSc.Vec, # type: ignore[name-defined] x1: npt.NDArray[np.inexact] | Sequence[npt.NDArray[np.inexact]], ): """Assign PETSc vector ``x0`` values to (blocked) array(s) ``x1``. This function performs the reverse of the assignment performed by the version of :func:`.assign(x0: (npt.NDArray[np.inexact] | list[npt.NDArray[np.inexact]]), x1: PETSc.Vec)`. Args: x0: Vector that will have its values assigned to ``x1``. x1: An array or list of arrays to assign to. """ if x0.getType() == PETSc.Vec.Type().NEST: # type: ignore[attr-defined] x0_nest = x0.getNestSubVecs() for _x0, _x1 in zip(x0_nest, x1): with _x0.localForm() as x: _x1[:] = x.array_r[:] # type: ignore[index] else: with x0.localForm() as _x0: if isinstance(x1, Sequence): start = 0 for _x1 in x1: end = start + _x1.shape[0] _x1[:] = _x0.array_r[start:end] start = end else: x1[:] = _x0.array_r[:] def _assign_block_data(maps: Iterable[tuple[IndexMap, int]], vec: PETSc.Vec): # type: ignore[name-defined] """Assign block data to a PETSc vector. Args: maps: Iterable of tuples each containing an ``IndexMap`` and the associated block size ``bs``. vec: PETSc vector to assign block data to. """ # Early exit if the vector already has block data or is a nest vector if vec.getAttr("_blocks") is not None or vec.getType() == "nest": return maps = [(index_map, bs) for index_map, bs in maps] off_owned = tuple( itertools.accumulate(maps, lambda off, m: off + m[0].size_local * m[1], initial=0) ) off_ghost = tuple( itertools.accumulate( maps, lambda off, m: off + m[0].num_ghosts * m[1], initial=off_owned[-1] ) ) vec.setAttr("_blocks", (off_owned, off_ghost))