forked from qutip/qutip
/
sparse_utils.pyx
225 lines (194 loc) · 7.79 KB
/
sparse_utils.pyx
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
# This file is part of QuTiP: Quantum Toolbox in Python.
#
# Copyright (c) 2011 and later, Paul D. Nation and Robert J. Johansson.
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are
# met:
#
# 1. Redistributions of source code must retain the above copyright notice,
# this list of conditions and the following disclaimer.
#
# 2. Redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution.
#
# 3. Neither the name of the QuTiP: Quantum Toolbox in Python nor the names
# of its contributors may be used to endorse or promote products derived
# from this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
# PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
# HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
###############################################################################
import numpy as np
cimport numpy as np
cimport cython
cdef inline int int_max(int x, int y):
return x ^ ((x ^ y) & -(x < y))
include "parameters.pxi"
@cython.boundscheck(False)
@cython.wraparound(False)
def _sparse_bandwidth(
np.ndarray[ITYPE_t, ndim=1] idx,
np.ndarray[ITYPE_t, ndim=1] ptr,
int nrows):
"""
Calculates the max (mb), lower(lb), and upper(ub) bandwidths of a
csr_matrix.
"""
cdef int lb, ub, mb, ii, jj, ldist
lb = -nrows
ub = -nrows
mb = 0
for ii in range(nrows):
for jj in range(ptr[ii], ptr[ii + 1]):
ldist = ii - idx[jj]
lb = int_max(lb, ldist)
ub = int_max(ub, -ldist)
mb = int_max(mb, ub + lb + 1)
return mb, lb, ub
@cython.boundscheck(False)
@cython.wraparound(False)
def _sparse_profile(np.ndarray[ITYPE_t, ndim=1] idx,
np.ndarray[ITYPE_t, ndim=1] ptr,
int nrows):
cdef int ii, jj, temp, ldist=0
cdef LTYPE_t pro=0
for ii in range(nrows):
temp = 0
for jj in range(ptr[ii], ptr[ii + 1]):
ldist = idx[jj] - ii
temp = int_max(temp, ldist)
pro += temp
return pro
@cython.boundscheck(False)
@cython.wraparound(False)
def _sparse_permute(
np.ndarray[cython.numeric, ndim=1] data,
np.ndarray[ITYPE_t, ndim=1] idx,
np.ndarray[ITYPE_t, ndim=1] ptr,
int nrows,
int ncols,
np.ndarray[ITYPE_t, ndim=1] rperm,
np.ndarray[ITYPE_t, ndim=1] cperm,
int flag):
"""
Permutes the rows and columns of a sparse CSR or CSC matrix according to
the permutation arrays rperm and cperm, respectively.
Here, the permutation arrays specify the new order of the rows and columns.
i.e. [0,1,2,3,4] -> [3,0,4,1,2].
"""
cdef int ii, jj, kk, k0, nnz
cdef np.ndarray[cython.numeric] new_data = np.zeros_like(data)
cdef np.ndarray[ITYPE_t] new_idx = np.zeros_like(idx)
cdef np.ndarray[ITYPE_t] new_ptr = np.zeros_like(ptr)
cdef np.ndarray[ITYPE_t] perm_r
cdef np.ndarray[ITYPE_t] perm_c
cdef np.ndarray[ITYPE_t] inds
if flag == 0: # CSR matrix
if len(rperm):
inds = np.argsort(rperm).astype(ITYPE)
perm_r = np.arange(len(rperm), dtype=ITYPE)[inds]
for jj in range(nrows):
ii = perm_r[jj]
new_ptr[ii + 1] = ptr[jj + 1] - ptr[jj]
for jj in range(nrows):
new_ptr[jj + 1] = new_ptr[jj+1] + new_ptr[jj]
for jj in range(nrows):
k0 = new_ptr[perm_r[jj]]
for kk in range(ptr[jj], ptr[jj + 1]):
new_idx[k0] = idx[kk]
new_data[k0] = data[kk]
k0 = k0 + 1
if len(cperm):
inds = np.argsort(cperm).astype(ITYPE)
perm_c = np.arange(len(cperm), dtype=ITYPE)[inds]
nnz = new_ptr[len(new_ptr) - 1]
for jj in range(nnz):
new_idx[jj] = perm_c[new_idx[jj]]
elif flag == 1: # CSC matrix
if len(cperm):
inds = np.argsort(cperm).astype(ITYPE)
perm_c = np.arange(len(cperm), dtype=ITYPE)[inds]
for jj in range(ncols):
ii = perm_c[jj]
new_ptr[ii + 1] = ptr[jj + 1] - ptr[jj]
for jj in range(ncols):
new_ptr[jj + 1] = new_ptr[jj + 1] + new_ptr[jj]
for jj in range(ncols):
k0 = new_ptr[perm_c[jj]]
for kk in range(ptr[jj], ptr[jj + 1]):
new_idx[k0] = idx[kk]
new_data[k0] = data[kk]
k0 = k0 + 1
if len(rperm):
inds = np.argsort(rperm).astype(ITYPE)
perm_r = np.arange(len(rperm), dtype=ITYPE)[inds]
nnz = new_ptr[len(new_ptr) - 1]
for jj in range(nnz):
new_idx[jj] = perm_r[new_idx[jj]]
return new_data, new_idx, new_ptr
@cython.boundscheck(False)
@cython.wraparound(False)
def _sparse_reverse_permute(
np.ndarray[cython.numeric, ndim=1] data,
np.ndarray[ITYPE_t, ndim=1] idx,
np.ndarray[ITYPE_t, ndim=1] ptr,
int nrows,
int ncols,
np.ndarray[ITYPE_t, ndim=1] rperm,
np.ndarray[ITYPE_t, ndim=1] cperm,
int flag):
"""
Reverse permutes the rows and columns of a sparse CSR or CSC matrix
according to the original permutation arrays rperm and cperm, respectively.
"""
cdef int ii, jj, kk, k0, nnz
cdef np.ndarray[cython.numeric, ndim=1] new_data = np.zeros_like(data)
cdef np.ndarray[ITYPE_t, ndim=1] new_idx = np.zeros_like(idx)
cdef np.ndarray[ITYPE_t, ndim=1] new_ptr = np.zeros_like(ptr)
if flag == 0: # CSR matrix
if len(rperm):
for jj in range(nrows):
ii = rperm[jj]
new_ptr[ii + 1] = ptr[jj + 1] - ptr[jj]
for jj in range(nrows):
new_ptr[jj + 1] = new_ptr[jj + 1] + new_ptr[jj]
for jj in range(nrows):
k0 = new_ptr[rperm[jj]]
for kk in range(ptr[jj], ptr[jj + 1]):
new_idx[k0] = idx[kk]
new_data[k0] = data[kk]
k0 = k0 + 1
if len(cperm):
nnz = new_ptr[len(new_ptr) - 1]
for jj in range(nnz):
new_idx[jj] = cperm[new_idx[jj]]
if flag == 1: # CSC matrix
if len(cperm):
for jj in range(ncols):
ii = cperm[jj]
new_ptr[ii + 1] = ptr[jj + 1] - ptr[jj]
for jj in range(ncols):
new_ptr[jj + 1] = new_ptr[jj + 1] + new_ptr[jj]
for jj in range(ncols):
k0 = new_ptr[cperm[jj]]
for kk in range(ptr[jj], ptr[jj + 1]):
new_idx[k0] = idx[kk]
new_data[k0] = data[kk]
k0 = k0 + 1
if len(rperm) != 0:
nnz = new_ptr[len(new_ptr) - 1]
for jj in range(nnz):
new_idx[jj] = rperm[new_idx[jj]]
return new_data, new_idx, new_ptr