File: assemble_vector.py

package info (click to toggle)
dolfinx-mpc 0.9.3-1
  • links: PTS, VCS
  • area: main
  • in suites: sid
  • size: 1,188 kB
  • sloc: python: 7,263; cpp: 5,462; makefile: 69; sh: 4
file content (346 lines) | stat: -rw-r--r-- 13,254 bytes parent folder | download | duplicates (2)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
# Copyright (C) 2021 Jørgen S. Dokken
#
# This file is part of DOLFINX_MPC
#
# SPDX-License-Identifier:    MIT
from __future__ import annotations

from typing import Optional, Tuple

from petsc4py import PETSc as _PETSc

import cffi
import dolfinx
import dolfinx.cpp as _cpp
import dolfinx.fem as _fem
import dolfinx.la as _la
import dolfinx.log as _log
import numpy
import numpy.typing as npt
from dolfinx.common import Timer

import numba
from dolfinx_mpc.multipointconstraint import MultiPointConstraint

from .helpers import _forms, extract_slave_cells, pack_slave_facet_info
from .numba_setup import initialize_petsc

ffi, _ = initialize_petsc()


def assemble_vector(form: _forms, constraint: MultiPointConstraint, b: Optional[_PETSc.Vec] = None) -> _PETSc.Vec:  # type: ignore
    """
    Assemble a compiled DOLFINx form into vector b.

    Args:
        form: The complied linear form
        constraint: The multi point constraint
        b: PETSc vector to assemble into (optional)
    """

    _log.log(_log.LogLevel.INFO, "Assemble MPC vector")
    timer_vector = Timer("~MPC: Assemble vector (numba)")

    # Unpack Function space data
    V = form.function_spaces[0]
    x_dofs = V.mesh.geometry.dofmap
    x = V.mesh.geometry.x
    dofs = V.dofmap.map()
    block_size = V.dofmap.index_map_bs

    # Data from multipointconstraint
    coefficients = constraint.coefficients()[0]
    masters_adj = constraint.masters
    c_to_s_adj = constraint.cell_to_slaves
    cell_to_slave = c_to_s_adj.array
    c_to_s_off = c_to_s_adj.offsets
    is_slave = constraint.is_slave
    mpc_data = (
        masters_adj.array,
        coefficients,
        masters_adj.offsets,
        cell_to_slave,
        c_to_s_off,
        is_slave,
    )
    slave_cells = extract_slave_cells(c_to_s_off)

    # Get index map and ghost info
    if b is None:
        index_map = constraint.function_space.dofmap.index_map
        vector = _la.create_petsc_vector(index_map, block_size)
    else:
        vector = b

    # Pack constants and coefficients
    form_coeffs = _cpp.fem.pack_coefficients(form._cpp_object)
    form_consts = _cpp.fem.pack_constants(form._cpp_object)

    tdim = V.mesh.topology.dim
    num_dofs_per_element = V.dofmap.dof_layout.num_dofs

    # Assemble vector with all entries
    with vector.localForm() as b_local:
        _cpp.fem.assemble_vector(b_local.array_w, form._cpp_object, form_consts, form_coeffs)

    # Check if we need facet permutations
    # FIXME: access apply_dof_transformations here
    e0 = form.function_spaces[0].element
    needs_transformation_data = e0.needs_dof_transformations or form._cpp_object.needs_facet_permutations
    cell_perms = numpy.array([], dtype=numpy.uint32)
    if needs_transformation_data:
        V.mesh.topology.create_entity_permutations()
        cell_perms = V.mesh.topology.get_cell_permutation_info()
    if e0.needs_dof_transformations:
        raise NotImplementedError("Dof transformations not implemented")
    # Assemble over cells
    subdomain_ids = form._cpp_object.integral_ids(_fem.IntegralType.cell)
    num_cell_integrals = len(subdomain_ids)

    if _PETSc.ScalarType == numpy.float32:  # type: ignore
        nptype = "float32"
    elif _PETSc.ScalarType == numpy.float64:  # type: ignore
        nptype = "float64"
    elif _PETSc.ScalarType == numpy.complex64:  # type: ignore
        nptype = "complex64"
    elif _PETSc.ScalarType == numpy.complex128:  # type: ignore
        nptype = "complex128"
    else:
        raise RuntimeError(f"Unsupported scalar type {_PETSc.ScalarType}.")  # type: ignore
    ufcx_form = form.ufcx_form
    if num_cell_integrals > 0:
        V.mesh.topology.create_entity_permutations()

        # NOTE: This depends on enum ordering in ufcx.h
        cell_form_pos = ufcx_form.form_integral_offsets[0]
        for i, id in enumerate(subdomain_ids):
            cell_kernel = getattr(ufcx_form.form_integrals[cell_form_pos + i], f"tabulate_tensor_{nptype}")
            active_cells = form._cpp_object.domains(_fem.IntegralType.cell, id)
            coeffs_i = form_coeffs[(_fem.IntegralType.cell, id)]
            with vector.localForm() as b:
                assemble_cells(
                    numpy.asarray(b),
                    cell_kernel,
                    active_cells[numpy.isin(active_cells, slave_cells)],
                    (x_dofs, x),
                    coeffs_i,
                    form_consts,
                    cell_perms,
                    dofs,
                    block_size,
                    num_dofs_per_element,
                    mpc_data,
                )

    # Assemble exterior facet integrals
    subdomain_ids = form._cpp_object.integral_ids(_fem.IntegralType.exterior_facet)
    num_exterior_integrals = len(subdomain_ids)
    if num_exterior_integrals > 0:
        V.mesh.topology.create_entities(tdim - 1)
        V.mesh.topology.create_connectivity(tdim - 1, tdim)
        # Get facet permutations if required
        facet_perms = numpy.array([], dtype=numpy.uint8)
        if form._cpp_object.needs_facet_permutations:
            facet_perms = V.mesh.topology.get_facet_permutations()
        perm = (cell_perms, form._cpp_object.needs_facet_permutations, facet_perms)
        # NOTE: This depends on enum ordering in ufcx.h
        ext_facet_pos = ufcx_form.form_integral_offsets[1]
        for i, id in enumerate(subdomain_ids):
            facet_kernel = getattr(ufcx_form.form_integrals[ext_facet_pos + i], f"tabulate_tensor_{nptype}")
            coeffs_i = form_coeffs[(_fem.IntegralType.exterior_facet, id)]
            facets = form._cpp_object.domains(_fem.IntegralType.exterior_facet, id)
            facet_info = pack_slave_facet_info(facets, slave_cells)
            num_facets_per_cell = len(V.mesh.topology.connectivity(tdim, tdim - 1).links(0))
            with vector.localForm() as b:
                assemble_exterior_slave_facets(
                    numpy.asarray(b),
                    facet_kernel,
                    facet_info,
                    (x_dofs, x),
                    coeffs_i,
                    form_consts,
                    perm,
                    dofs,
                    block_size,
                    num_dofs_per_element,
                    mpc_data,
                    num_facets_per_cell,
                )
    timer_vector.stop()
    return vector


@numba.njit
def assemble_cells(
    b: npt.NDArray[_PETSc.ScalarType],  # type: ignore
    kernel: cffi.FFI.CData,
    active_cells: npt.NDArray[numpy.int32],
    mesh: Tuple[npt.NDArray[numpy.int32], npt.NDArray[dolfinx.default_real_type]],
    coeffs: npt.NDArray[_PETSc.ScalarType],  # type: ignore
    constants: npt.NDArray[_PETSc.ScalarType],  # type: ignore
    permutation_info: npt.NDArray[numpy.uint32],
    dofmap: npt.NDArray[numpy.int32],
    block_size: int,
    num_dofs_per_element: int,
    mpc: Tuple[  # type: ignore
        npt.NDArray[numpy.int32],
        npt.NDArray[_PETSc.ScalarType],
        npt.NDArray[numpy.int32],
        npt.NDArray[numpy.int32],
        npt.NDArray[numpy.int32],
        npt.NDArray[numpy.int32],
    ],
):
    """Assemble additional MPC contributions for cell integrals"""
    ffi_fb = ffi.from_buffer

    # Empty arrays mimicking Nullpointers
    facet_index = numpy.zeros(0, dtype=numpy.int32)
    facet_perm = numpy.zeros(0, dtype=numpy.uint8)

    # Unpack mesh data
    x_dofmap, x = mesh

    # NOTE: All cells are assumed to be of the same typecd
    geometry = numpy.zeros((x_dofmap.shape[1], 3), dtype=dolfinx.default_real_type)
    b_local = numpy.zeros(block_size * num_dofs_per_element, dtype=_PETSc.ScalarType)  # type: ignore

    for cell_index in active_cells:
        # Compute mesh geometry for cell
        geometry[:, :] = x[x_dofmap[cell_index]]

        # Assemble local element vector
        b_local.fill(0.0)
        kernel(
            ffi_fb(b_local),  # type: ignore
            ffi_fb(coeffs[cell_index, :]),  # type: ignore
            ffi_fb(constants),  # type: ignore
            ffi_fb(geometry),  # type: ignore
            ffi_fb(facet_index),  # type: ignore
            ffi_fb(facet_perm),  # type: ignore
        )
        # NOTE: Here we need to add the apply_dof_transformation function

        # Modify global vector and local cell contributions
        b_local_copy = b_local.copy()
        modify_mpc_contributions(b, cell_index, b_local, b_local_copy, mpc, dofmap, block_size, num_dofs_per_element)
        for j in range(num_dofs_per_element):
            for k in range(block_size):
                position = dofmap[cell_index, j] * block_size + k
                b[position] += b_local[j * block_size + k] - b_local_copy[j * block_size + k]


@numba.njit
def assemble_exterior_slave_facets(
    b: npt.NDArray[_PETSc.ScalarType],  # type: ignore
    kernel: cffi.FFI.CData,
    facet_info: npt.NDArray[numpy.int32],
    mesh: Tuple[npt.NDArray[numpy.int32], npt.NDArray[numpy.float64]],
    coeffs: npt.NDArray[_PETSc.ScalarType],  # type: ignore
    constants: npt.NDArray[_PETSc.ScalarType],  # type: ignore
    permutation_info: npt.NDArray[numpy.uint32],
    dofmap: npt.NDArray[numpy.int32],
    block_size: int,
    num_dofs_per_element: int,
    mpc: Tuple[  # type: ignore
        npt.NDArray[numpy.int32],
        npt.NDArray[_PETSc.ScalarType],
        npt.NDArray[numpy.int32],
        npt.NDArray[numpy.int32],
        npt.NDArray[numpy.int32],
        npt.NDArray[numpy.int32],
    ],
    num_facets_per_cell: int,
):
    """Assemble additional MPC contributions for facets"""
    ffi_fb = ffi.from_buffer

    # Unpack facet permutation info
    cell_perms, needs_facet_perm, facet_perms = permutation_info
    facet_index = numpy.zeros(1, dtype=numpy.int32)
    facet_perm = numpy.zeros(1, dtype=numpy.uint8)

    # Unpack mesh data
    x_dofmap, x = mesh

    geometry = numpy.zeros((x_dofmap.shape[1], 3), dtype=x.dtype)
    b_local = numpy.zeros(block_size * num_dofs_per_element, dtype=_PETSc.ScalarType)  # type: ignore
    for i in range(facet_info.shape[0]):
        # Extract cell index (local to process) and facet index (local to cell) for kernel
        cell_index, local_facet = facet_info[i]
        facet_index[0] = local_facet

        # Extract cell geometry
        geometry[:, :] = x[x_dofmap[cell_index]]

        # Compute local facet kernel
        if needs_facet_perm:
            facet_perm[0] = facet_perms[cell_index * num_facets_per_cell + local_facet]
        b_local.fill(0.0)
        kernel(
            ffi_fb(b_local),  # type: ignore
            ffi_fb(coeffs[cell_index, :]),  # type: ignore
            ffi_fb(constants),  # type: ignore
            ffi_fb(geometry),  # type: ignore
            ffi_fb(facet_index),  # type: ignore
            ffi_fb(facet_perm),  # type: ignore
        )
        # NOTE: Here we need to add the apply_dof_transformation

        # Modify local contributions and add global MPC contributions
        b_local_copy = b_local.copy()
        modify_mpc_contributions(b, cell_index, b_local, b_local_copy, mpc, dofmap, block_size, num_dofs_per_element)
        for j in range(num_dofs_per_element):
            for k in range(block_size):
                position = dofmap[cell_index, j] * block_size + k
                b[position] += b_local[j * block_size + k] - b_local_copy[j * block_size + k]


@numba.njit(cache=True)
def modify_mpc_contributions(
    b: npt.NDArray[_PETSc.ScalarType],  # type: ignore
    cell_index: int,  # type: ignore
    b_local: npt.NDArray[_PETSc.ScalarType],  # type: ignore
    b_copy: npt.NDArray[_PETSc.ScalarType],  # type: ignore
    mpc: Tuple[  # type: ignore
        npt.NDArray[numpy.int32],
        npt.NDArray[_PETSc.ScalarType],
        npt.NDArray[numpy.int32],
        npt.NDArray[numpy.int32],
        npt.NDArray[numpy.int32],
        npt.NDArray[numpy.int32],
    ],
    dofmap: npt.NDArray[numpy.int32],
    block_size: int,
    num_dofs_per_element: int,
):
    """
    Modify local entries of b_local with MPC info and add modified
    entries to global vector b.
    """

    # Unwrap MPC data
    masters, coefficients, offsets, cell_to_slave, cell_to_slave_offset, is_slave = mpc

    # Determine which slaves are in this cell,
    # and which global index they have in 1D arrays
    cell_slaves = cell_to_slave[cell_to_slave_offset[cell_index] : cell_to_slave_offset[cell_index + 1]]

    # Get local index of slaves in cell
    cell_blocks = dofmap[cell_index]
    local_index = numpy.empty(len(cell_slaves), dtype=numpy.int32)
    for i in range(num_dofs_per_element):
        for j in range(block_size):
            dof = cell_blocks[i] * block_size + j
            if is_slave[dof]:
                location = numpy.flatnonzero(cell_slaves == dof)[0]
                local_index[location] = i * block_size + j

    # Move contribution from each slave to the corresponding master dof
    # and zero out local b
    for local, slave in zip(local_index, cell_slaves):
        cell_masters = masters[offsets[slave] : offsets[slave + 1]]
        cell_coeffs = coefficients[offsets[slave] : offsets[slave + 1]]
        for m0, c0 in zip(cell_masters, cell_coeffs):
            b[m0] += c0 * b_copy[local]
            b_local[local] = 0