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
347
|
Sept 12, 2017: version 3.2.0
* replaced UFget with ssget
* added comment to cs_lu.c
May 4, 2016: version 3.1.9
* minor changes to Makefile
Apr 1, 2016: version 3.1.8
* licensing simplified (no other change); refer to CSparse/Doc/License.txt
Feb 17, 2016: version 3.1.8beta (not released publically)
* minor update to cs_sprealloc, for handling sparse empty matrices.
Feb 1, 2016: version 3.1.7
* update to Makefiles
Jan 30, 2016: version 3.1.6
* correction to Makefiles
* correction to mexFunctions required for MATLAB R2015b,
for sparse empty matrices and NULL mxGetPr
Jan 1, 2016: version 3.1.5
* Makefile now creates both static (*.a) and shared (*.so / *.dylib)
libraries. No change to C code except version number in cs.h.
Oct 10, 2014: version 3.1.4
modified cs_make.m. No change to C code except version number.
Mar 26, 2014: version 3.1.3
* minor update to UFget
Apr 16, 2013: verison 3.1.2
* removed a useless line from cs_sqr.c (no integer overflow can
occur in S->lnz, S->unz at that point)
Jun 20, 2012: verison 3.1.1
* minor update for Windows (removed filesep)
Jun 1, 2012: version 3.1.0
* minor changes to enable the creation of CXSparse 3.1.0
Dec 7, 2011: version 3.0.2
* fixed the Makefile to better align with CFLAGS and other standards
* minor fix to MATLAB cs_install.m
Jan 18, 2010: version 3.0.0
* changed "int" to "csi", which is normally "ptrdiff_t". This way,
CSparse can be used on a 64-bit platform (and in 64-bit MATLAB
in particular), without the need to use CXSparse. If you wish
to use "int", simply edit CSparse/Include/cs.h and change the
definition of csi, or compile with -Dcsi=int.
Jan 25, 2011: version 2.2.5
* minor fix to "make install"
* minor change to cs_util.c, typecast return (cs_free (...))
* minor fixes to UFget: help file, UFsettings.txt
Nov 30, 2009: version 2.2.4
* added "make install" and "make uninstall"
* minor change to cs_make.m
Jan 20, 2009, v2.2.3
* all tabs expanded to 8 spaces (except for Makefiles)
* example corrected in cs_gaxpy.m
* minor change to cs_sparse to allow for i,j,x to be either
row or column vectors (or any mixture)
Sept 8, 2008, v2.2.2
* minor change to cs_make.m, to change "/" to filesep ("/" on Unix, "\"
on Windows), to handle limitations in some Windows compilers
Nov 1, 2007, v2.2.1
* very minor change to Include/cs.h: Changed name of 2nd argument of
cs_permute, to match the code. This has no affect on the code itself,
since the type ("int *") is unchanged. It's just a documentation issue.
* minor lint cleanup in mexFunctions
Mar 31, 2007, v2.2.0
* few changes to primary Source/ files. Changes mostly affect MATLAB
interface.
* Source/cs_house.c: correction to comment
* Souce/cs_updown.c: whitespace changed to reflect change in CXSparse,
no impact at all on CSparse itself.
* Doc/, Lib/ and Include/ directories created.
* Source/cs.h moved to Include/cs.h, version number changed to 2.2.0.
* modification to Makefiles, cs_make.m
* correction to help comments in cs_dmperm.m, cs_qr.m, cs_scc2.m, cs_scc.m
* if complex matrix passed to CSparse in MATLAB, error message
suggests using CXSparse instead
* minor performance fix to cs_sparse_mex.c
* cs_randperm added to MATLAB/Makefile; already appeared in cs_make.m
* minor improvements to MATLAB/ demos.
Mar 1, 2007, v2.1.0
* Source/cs_add.c: added test for matrix dimensions
* Source/cs_multiply.c: added test for matrix dimensions
* correction to MATLAB demo3 (no error in C code version of the demo)
* minor corrections to MATLAB m-file help comments.
Dec 12, 2006, v2.0.7
* minor MATLAB cleanup
Dec 6, 2006, v2.0.6
* Update to UFget. Now relies on the MATLAB urlwrite function instead of
my own Java code.
Nov 2006, v2.0.5
* Added UFgrep to UFget toolbox.
* No changes to C Source code, except for version number and date.
* Added two test matrices: ibm32a and ibm32b. ibm32a is the Harwell/
Boeing matrix ibm32, but with the last column removed. ibm32b
is the transpose of ibm32a. With optimization enabled (-O),
2 lines in cs_maxtrans.c are not tested; these matrices correct
that problem.
* Fixed UFget. Earlier version could not download matrices larger than
32MB.
* Modified UFget/UFweb, to reflect changes in the UF Sparse Matrix
Collection.
* Added ccspy.m and cs_scc2.m MATLAB functions
* Added examples to help info in each *.m MATLAB file
* modified cs_dmspy to speed up the plotting of large matrices with many
singletons
* minor change to cspy: now draws a box around the matrix.
* minor changes to MATLAB demos and tests.
Oct 13, 2006, v2.0.4
* minor modification to cs_updown.c. "n" was incorrectly declared "double".
It should be "int". This was safe, just a little confusing (n was only
used in an argument to cs_malloc, and is thus typecast).
Sept 28, 2006, v2.0.3
* minor modifications to MATLAB interface, to allow CSparse to be used
in MATLAB 6.5.
* added examples to m-files, other minor m-file cleanup.
* bug fix to cspy, to handle NaN's properly.
Aug 23, 2006: v2.0.2
* change to cs_updown mexFunction, to handle multiple rank updates
* symbolic links removed from Tcov/ directory in the distribution.
They are now created by Tcov/Makefile as needed. This makes the
zip files cleaner. Tcov/*test.c test files renamed.
July 19, 2006:
* minor fix to cs_must_compile.m and cs_make.m, to allow CSparse to be
compiled in MATLAB 6.5 with cs_make.
* minor fix to cspy for complex matrices (imaginary part was ignored).
* no change to version number or date, since there are no changes that
affect the appearance of CSparse in the book ("Direct Methods for
Sparse Linear Systems", SIAM, 2006).
June 24, 2006:
* minor typos in comments corrected. No change in code itself,
and no change in version number or date.
May 27, 2006, v2.0.1: (this version is printed in the book)
* minor bug fix. cs_util.c modified, so that cs_sprealloc (T,0) works
properly for a triplet matrix T (setting T->nzmax equal to T->nz).
The line in v2.0.0:
nzmax = (nzmax <= 0) ? (A->p [A->n]) : nzmax ;
changes to the following in v2.0.1:
if (nzmax <= 0) nzmax = (CS_CSC (A)) ? (A->p [A->n]) : A->nz ;
* minor typographical changes arising from the editting of the book.
Apr 12, 2006, v2.0.0:
* random permutation option added to cs_maxtrans and cs_dmperm, to help
avoid rare cases where the O(|A|n) time complexity is reached in
practice (GHS_indef/boyd2 in the UF sparse matrix collection, for
example). New cs_randperm function added.
Apr 10, 2006:
* stylistic changes for the book (except for the bug fix):
* "int order" parameter of cs_amd, cs_lusol, cs_cholsol, cs_qrsol, cs_sqr,
cs_schol changed. Now 0 means no ordering, 1 is A+A', 2 is S*S', and
3 is A*A'. In v1.2 and earlier, "order" took on a value ranging from
-1 to 2. "int n" parameter rearranged for cs_ipvec, cs_pvec, cs_post
(int n moved to the end). cs_triplet renamed cs_compress.
To ensure that these changes are propagated into user code that calls
CSparse, the "order" parameter has been placed as the first parameter
in all these routines. Your compiler will complain (gcc will halt) if
you upgrade from v1.2 to v2.0 without changing your code. This is much
better than a silent error in which you get the wrong ordering by
mistake (with a huge penalty in run-time performance and no compiler
warnings).
New syntax (v2.0 and later):
----------------------------
order = 0: natural ordering
order = 1: amd (A+A')
order = 2: amd (S'*S), where S=A except dense rows dropped
order = 3: amd (A'*A)
int cs_cholsol (int order, const cs *A, double *b) ;
int cs_lusol (int order, const cs *A, double *b, double tol) ;
int cs_qrsol (int order, const cs *A, double *b) ;
int *cs_amd (int order, const cs *A) ;
css *cs_schol (int order, const cs *A) ;
css *cs_sqr (int order, const cs *A, int qr) ;
int *cs_post (const int *parent, int n) ;
int cs_ipvec (const int *p, const double *b, double *x, int n) ;
int cs_pvec (const int *p, const double *b, double *x, int n) ;
cs *cs_compress (const cs *T) ;
Old syntax (v1.2 and earlier):
------------------------------
order = -1: natural ordering
order = 0: amd (A+A')
order = 1: amd (S'*S), where S=A except dense rows dropped
order = 2: amd (A'*A)
int cs_cholsol (const cs *A, double *b, int order) ;
int cs_lusol (const cs *A, double *b, int order, double tol) ;
int cs_qrsol (const cs *A, double *b, int order) ;
int *cs_amd (const cs *A, int order) ;
css *cs_schol (const cs *A, int order) ;
css *cs_sqr (const cs *A, int order, int qr) ;
int *cs_post (int n, const int *parent) ;
int cs_ipvec (int n, const int *p, const double *b, double *x) ;
int cs_pvec (int n, const int *p, const double *b, double *x) ;
cs *cs_triplet (const cs *T) ;
* CS_OVERFLOW macro removed (removed from cs_malloc.c; not needed).
* S->leftmost added to css (it was tacked onto S->pinv before).
* P,Q,R,S components of csd struct changed to p,q,r,s.
* Pinv and Q components of css struct changed to pinv and q.
* CS_CSC and CS_TRIPLET macros added, to clarify which CSparse functions
accept cs matrices in compressed column form, triplet form, or both.
* check for negative row/column indices added to cs_entry.
* cs_ereach and cs_leaf functions added.
* call to cs_sprealloc added to cs_fkeep.
* bug fixes in cs_counts and cs_amd (memory leak under rare out-of-memory
conditions).
Mar 15, 2006:
* cs_scc modified so that the row and columns of each component are put in
their natural order. cs_dmperm modified so that columns of each block
(instead of rows) are placed in their natural order.
* cs_splsolve renamed cs_spsolve, generalized to handle both Lx=b and
Ux=b, and non-unit diagonal, and placed in its own file (cs_spsolve.c;
it was a static function in cs_lu.c). cs_lsolve_mex.c and
cs_splsolve_mex.c merged (the latter was removed). cs_spsolve.c file
added
* cs_dmspy changed, so that block borders line up better with the matrix.
Mar 6, 2006:
* Makefile modified so that the Tcov tests (which may not be portable)
are not compiled and run by the default "make" in the CSparse directory.
To compile everything, including the Tcov tests, use "make all".
Trivial change to cs.h.
Feb 27, 2006:
* cs_reach, cs_dfs, cs_splsolve, cs_lu, and cs_scc changed to remove O(n)
initialized workspace.
* cs_reach and cs_splsolve now user-callable (were static in cs_lu.c).
Feb 20, 2006:
* various changes to simplify the construction of CXSparse
Feb 7, 2006:
* changed prototypes, adding "const" where appropriate.
|
