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
|
# Copyright (C) 2023 Chris N. Richardson
#
# This file is part of DOLFINx (https://www.fenicsproject.org)
#
# SPDX-License-Identifier: LGPL-3.0-or-later
"""Unit tests for MatrixCSR"""
from mpi4py import MPI
import numpy as np
import pytest
import ufl
from dolfinx import cpp as _cpp
from dolfinx import fem
from dolfinx.common import IndexMap
from dolfinx.cpp.la import BlockMode, SparsityPattern
from dolfinx.la import matrix_csr
from dolfinx.mesh import GhostMode, create_unit_square
def create_test_sparsity(n, bs):
im = IndexMap(MPI.COMM_WORLD, n)
sp = SparsityPattern(MPI.COMM_WORLD, [im, im], [bs, bs])
if bs == 1:
for i in range(2):
for j in range(2):
sp.insert(2 + i, 4 + j)
elif bs == 2:
sp.insert(1, 2)
sp.finalize()
return sp
@pytest.mark.parametrize("dtype", [np.float32, np.float64, np.complex64, np.complex128])
def test_add(dtype):
# Regular CSR Matrix 6x6 with bs=1
sp = create_test_sparsity(6, 1)
mat1 = matrix_csr(sp, dtype=dtype)
# Insert a block using plain indices
mat1.add([1.0, 2.0, 3.0, 4.0], [2, 3], [4, 5], 1)
# Insert to same block using bs=2 data
mat1.add([10.0, 20.0, 30.0, 40.0], [1], [2], 2)
A1 = mat1.to_dense()
# Block CSR Matrix 3x3 with bs=2
sp = create_test_sparsity(3, 2)
mat2 = matrix_csr(sp, dtype=dtype)
# Insert a block using bs=1 data
mat2.add([10.0, 20.0, 30.0, 40.0], [2, 3], [4, 5], 1)
# Insert a block using bs=2 data
mat2.add([1.0, 2.0, 3.0, 4.0], [1], [2], 2)
A2 = mat2.to_dense()
assert np.allclose(A1, A2)
# Block CSR Matrix 3x3 with bs=2, expanded (should be same as A1)
mat3 = matrix_csr(sp, BlockMode.expanded, dtype=dtype)
# Insert a block using bs=1 data
mat3.add([10.0, 2.0, 30.0, 4.0], [2, 3], [4, 5], 1)
# Insert a block using bs=2 data
mat3.add([1.0, 20.0, 3.0, 40.0], [1], [2], 2)
A3 = mat3.to_dense()
assert np.allclose(A1, A3)
mat3.set_value(0.0)
assert mat3.squared_norm() == 0.0 # /NOSONAR
@pytest.mark.parametrize("dtype", [np.float32, np.float64, np.complex64, np.complex128])
def test_set(dtype):
mpi_size = MPI.COMM_WORLD.size
# Regular CSR Matrix 6x6 with bs=1
sp = create_test_sparsity(6, 1)
mat1 = matrix_csr(sp, dtype=dtype)
# Set a block with bs=1
mat1.set([2.0, 3.0, 4.0, 5.0], [2, 3], [4, 5], 1)
n1 = mat1.squared_norm()
assert n1 == 54.0 * mpi_size # /NOSONAR
# Set same block with bs=2
mat1.set([2.0, 3.0, 4.0, 5.0], [1], [2], 2)
n2 = mat1.squared_norm()
assert n1 == n2
@pytest.mark.parametrize("dtype", [np.float32, np.float64, np.complex64, np.complex128])
def test_set_blocked(dtype):
mpi_size = MPI.COMM_WORLD.size
# Blocked CSR Matrix 3x3 with bs=2
sp = create_test_sparsity(3, 2)
mat1 = matrix_csr(sp, dtype=dtype)
# Set a block with bs=1
mat1.set([2.0, 3.0, 4.0, 5.0], [2, 3], [4, 5], 1)
n1 = mat1.squared_norm()
assert n1 == 54.0 * mpi_size # /NOSONAR
@pytest.mark.parametrize("dtype", [np.float32, np.float64, np.complex64, np.complex128])
def test_distributed_csr(dtype):
size = MPI.COMM_WORLD.size
rank = MPI.COMM_WORLD.rank
if size == 1:
return
# global size N
N = 36
nghost = 3
nbr = (rank + 1) % size
n = int(N / size)
ghosts = np.array(range(n * nbr, n * nbr + nghost), dtype=np.int64)
owner = np.ones_like(ghosts, dtype=np.int32) * nbr
im = IndexMap(MPI.COMM_WORLD, n, ghosts, owner)
sp = SparsityPattern(MPI.COMM_WORLD, [im, im], [1, 1])
for i in range(n):
for j in range(n + nghost):
sp.insert(i, j)
for i in range(n, n + nghost):
for j in range(n, n + nghost):
sp.insert(i, j)
sp.finalize()
mat = matrix_csr(sp, dtype=dtype)
irow = np.array(range(n), dtype=np.int32)
icol = np.array(range(n + nghost), dtype=np.int32)
data = np.ones(len(irow) * len(icol), dtype=dtype) * 2.0
mat.add(data, irow, icol, 1)
irow = np.array(range(n, n + nghost), dtype=np.int32)
icol = np.array(range(n, n + nghost), dtype=np.int32)
data = np.ones(len(irow) * len(icol), dtype=dtype)
mat.add(data, irow, icol, 1)
pre_final_sum = mat.data.sum()
mat.scatter_reverse()
assert np.isclose(mat.data.sum(), pre_final_sum)
@pytest.mark.parametrize(
"dtype",
[
np.float32,
np.float64,
pytest.param(np.complex64, marks=pytest.mark.xfail_win32_complex),
pytest.param(np.complex128, marks=pytest.mark.xfail_win32_complex),
],
)
def test_set_block_matrix(dtype):
mesh_dtype = np.real(dtype(0)).dtype
ghost_mode = GhostMode.shared_facet
mesh = create_unit_square(MPI.COMM_WORLD, 2, 4, ghost_mode=ghost_mode, dtype=mesh_dtype)
V = fem.functionspace(mesh, ("Lagrange", 1, (2,)))
u, v = ufl.TrialFunction(V), ufl.TestFunction(V)
a = fem.form(ufl.inner(u, v) * ufl.dx, dtype=dtype)
A = fem.create_matrix(a)
As = A.to_scipy(ghosted=True)
assert As.blocksize == (2, 2)
@pytest.mark.parametrize(
"dtype",
[
np.float32,
np.float64,
pytest.param(np.complex64, marks=pytest.mark.xfail_win32_complex),
pytest.param(np.complex128, marks=pytest.mark.xfail_win32_complex),
],
)
def test_set_diagonal_distributed(dtype):
mesh_dtype = np.real(dtype(0)).dtype
ghost_mode = GhostMode.shared_facet
mesh = create_unit_square(MPI.COMM_WORLD, 5, 5, ghost_mode=ghost_mode, dtype=mesh_dtype)
V = fem.functionspace(mesh, ("Lagrange", 1))
tdim = mesh.topology.dim
cellmap = mesh.topology.index_map(tdim)
num_cells = cellmap.size_local + cellmap.num_ghosts
# Integration domain includes ghost cells
cells_domains = [(1, np.arange(0, num_cells))]
dx = ufl.Measure("dx", subdomain_data=cells_domains, domain=mesh)
u, v = ufl.TrialFunction(V), ufl.TestFunction(V)
a = fem.form(ufl.inner(u, v) * dx(1), dtype=dtype)
# get index map from function space
index_map = V.dofmap.index_map
num_dofs = index_map.size_local + index_map.num_ghosts
# list of dofs including ghost dofs
dofs = np.arange(0, num_dofs, dtype=np.int32)
# create matrix
A = fem.create_matrix(a)
As = A.to_scipy(ghosted=True)
# set diagonal values
value = dtype(1.0)
_cpp.fem.insert_diagonal(A._cpp_object, dofs, value)
# check diagonal values: they should be 1.0, including ghost dofs
diag = As.diagonal()
reference = np.full_like(diag, value, dtype=dtype)
assert np.allclose(diag, reference)
# Update matrix: this will remove ghost rows and diagonal values of
# ghost rows will be added to diagonal of corresponding process
A.scatter_reverse()
diag = As.diagonal()
nlocal = index_map.size_local
assert (diag[nlocal:] == dtype(0.0)).all()
shared_dofs = index_map.index_to_dest_ranks()
for dof in range(nlocal):
owners = shared_dofs.links(dof)
assert diag[dof] == len(owners) + 1
# create matrix
A = fem.create_matrix(a)
As = A.to_scipy(ghosted=True)
# set diagonal values using dirichlet bc: this will set diagonal values of
# owned rows only
bc = fem.dirichletbc(dtype(0.0), dofs, V)
_cpp.fem.insert_diagonal(A._cpp_object, a.function_spaces[0], [bc._cpp_object], value)
# check diagonal values: they should be 1.0, except ghost dofs
diag = As.diagonal()
reference = np.full_like(diag, value, dtype=dtype)
assert np.allclose(diag[:nlocal], reference[:nlocal])
assert np.allclose(diag[nlocal:], np.zeros_like(diag[nlocal:]))
# Update matrix:
# this will zero ghost rows and diagonal values are already zero.
A.scatter_reverse()
assert (As.diagonal()[nlocal:] == dtype(0.0)).all()
assert (As.diagonal()[:nlocal] == dtype(1.0)).all()
@pytest.mark.parametrize("dtype", [np.float32, np.float64, np.complex64, np.complex128])
def test_bad_entry(dtype):
sp = create_test_sparsity(6, 1)
mat1 = matrix_csr(sp, dtype=dtype)
# Set block in bs=1 matrix (tests insert_blocked_csr)
with pytest.raises(RuntimeError):
mat1.set([1.0, 2.0, 3.0, 4.0], [0], [0], 2)
# Set an single entry in bs=1 matrix (tests insert_csr)
with pytest.raises(RuntimeError):
mat1.add([1.0], [0], [0], 1)
sp = create_test_sparsity(3, 2)
mat2 = matrix_csr(sp, BlockMode.compact, dtype=dtype)
# set unblocked in bs=2 matrix (tests insert_nonblocked_csr)
with pytest.raises(RuntimeError):
mat2.add([2.0, 3.0, 4.0, 5.0], [0, 1], [0, 1], 1)
|
