Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
Newer
Older
100644 247 lines (166 sloc) 8.602 kB
9ac1651 Private branch to revise build system.
d-orban authored
1 Fixed to locate numpy header files correctly
2
3 Automate installation of numpy (no flags to change in setup.py)
4
5 Removed some Numeric imports
6
7 Moved version number from __init__.py to __version__.py
8
9 1.0
10 --------------------------------------------------------------------------------
11 Moved modules inside "pysparse" name space.
12
13 Updated to use the new numpy, but still compatible with the old Numeric.
14
15
16 0.34.032
17 ---------------------------------------------------------------------------------
18 Fixed a memory leak in ll_mat.keys(). (Reported by c chui)
19
20 Improved the performance of ll_mat.scale(), ll_mat.values() and ll_mat.items().
21
22 0.34.031
23 ---------------------------------------------------------------------------------
24 Modules using the pysparse C API can now define NO_IMPORT_SPMATRIX
25 together with SPMATRIX_UNIQUE_SYMBOL to avoid multiple definition of
26 the SpMatrix_API pointer. This fix was suggested by Erica Madeleine Ligner.
27
28 0.34.030
29 ---------------------------------------------------------------------------------
30 Applied patches 1089844, 1089837, 1089831 (supplied by Daniel Wheeler
31 & Jonathan Guyer): Added three new ll_mat methods:
32 update_add_at(b,id1,id2), put(b,id1,id2), take(b,id1,id2), fixed OS X
33 compatibility issues, Numpy compatibility fixes.
34
35 Applied some additional fixes for 64-bit machines.
36
37 PySparse now works with Numeric 23.7 (also on 64-bit machines).
38
39 [1111727] Replaced #include <malloc.h> by #include <stdlib.h> to fix
40 compatibility issue on OS X (reported by Jonathan Guyer).
41
42 Removed reference to libm.a which broke compilation on Win32 (reported
43 by Mark Dausch).
44
45 0.33.029
46 ---------------------------------------------------------------------------------
47 added new umfpack module. (by Nathaniel C. Domingo)
48
49 added Umfpack library source files to distribution (by Nathaniel C. Domingo)
50
51 added new module sparray, implementing a multidimensional sparse array
52 (by Pascal Barbedor)
53
54 fixed bug in SSOR preconditioner (precon module) for omega != 1.0 and
55 steps > 1
56
57 0.32.028
58 ---------------------------------------------------------------------------------
59 added Xavier Vasseur's gmres solver to the itsolvers module. Thanks Xavier!
60
61 0.31.027
62 ---------------------------------------------------------------------------------
63 fixed: call to SpMatrix_MATVEC in Src/pcg.c
64
65 0.31.026
66 ---------------------------------------------------------------------------------
67 PySparse is placed under a BSD style license
68
69 0.30.025
70 ---------------------------------------------------------------------------------
71 fixed bug in "generalize" member function of ll_mat type. Reported by
72 Xavier Vasseur.
73
74 the result tuple returned by jdsym.jdsym now includes an additional
75 value representing the avg. number of inner iterations. This change
76 may be incompatble with existing code calling this module function.
77
78 0.29.024
79 ---------------------------------------------------------------------------------
80
81 PySparse can be build using MinGW on Win32 platforms. Check the
82 instructions in INSTALL.
83
84 0.29.023
85 ---------------------------------------------------------------------------------
86
87 fixed memory leaks in SpMatrix_LLMatSetItem and
88 SpMatrix_LLMatUpdateItemAdd that occur if PyMem_Resize would fail.
89
90 More detailed installation instructions for Win32 platforms in INSTALL
91
92 0.29.022
93 ---------------------------------------------------------------------------------
94 new method for ll_mat objects: compress
95
96 nicer output for ll_mat objects representing small matrices
97
98 optimised matrix-matrix-multiplication routine included (unsym-unsym case)
99
100 new version of bicgstab solver (uses less memory)
101
102 0.28.021
103 ---------------------------------------------------------------------------------
104 fixed error handling in ll_mat.generalize
105
106 fixed name mismatched in solver kernels in itsolvers
107
108 0.28.020
109 ---------------------------------------------------------------------------------
110 new method itsolvers.bicgstab (implemented by Oliver Broeker)
111
112 all solvers in the itsolvers module now check for exceptions after
113 each precon or matvec call.
114
115 solver kernels in itsolvers now have names that conflict less likely
116 with other modules
117
118 0.28.019
119 ---------------------------------------------------------------------------------
120 uncommented call to PrintStat in superlu/dgssv.c and superlu/dgssvx.c
121 to prevent unwanted output in ML package
122
123 fixed memory leak in sss_mat.c: diag array was not freed
124
125 new method for ll_mat objects: generalize
126
127 0.27.018
128 ---------------------------------------------------------------------------------
129 updated test modules to reflect earlier bug fix in poisson.poisson2d()
130
131 new default precision in ll_mat.export_mtx(): 16 instead of 6
132
133 fixed bug in Src/jdsymmodule.c: solution now copied correctly if not
134 all requested eigenvectors converged.
135
136 0.27.017
137 ---------------------------------------------------------------------------------
138 fixed bug in SRC/cgs.c: now returns correct iteration count, if initial guess
139 satisfies stopping criterion
140
141 fixed bug in Src/superlumodule.c solve transposed system, since matrix was
142 passed row-wise
143
144 new method for ll_mat matrices: norm(p)
145
146 0.26.016
147 ---------------------------------------------------------------------------------
148 changed interface of SpMatrix_Matvec(), now supports non-square matrices
149
150 0.26.015
151 ---------------------------------------------------------------------------------
152 reimplemented SpMatrix_LLMatBuildColIndex and SpMatrix_LLMatDestroyColIndex
153
154 0.26.014
155 ---------------------------------------------------------------------------------
156 new module itsolvers_util: encapsulation objects for all iterative solvers
157
158 0.25.013
159 ---------------------------------------------------------------------------------
160 fixed bug in poisson.poisson2d_sym_blk(): diagonal entries are now 4, not 2.
161
162 changed superlu_superlu_timer.c to support Win32
163
164 0.25.012
165 ---------------------------------------------------------------------------------
166 test module for superlu
167
168 poisson module for generating 1D and 2D Poisson matrices
169
170 0.25.011
171 ---------------------------------------------------------------------------------
172 incorporated a new version of SuperLU (date 2002/09/23)
173
174 changed postorder tree traversal in sp_coletree.c to be more stack friendly
175 (max stack size on Linux seems to be 6 MBytes)
176
177 removed stdout in get_perm_c.c
178
179 0.25.008
180 ---------------------------------------------------------------------------------
181 fixed bug in matvec method of csr_mat, occuring if stride of vector y is not one
182
183 ll_mat.shift now supports shift of a general matrix by a symmetric matrix
184
185 new matvec and matvec:transp object methods for ll_mat type
186
187 fixed bug when assigning a symmetric matrix to a block of a general matrix
188
189 delete_cols method added for ll_mat objects
190
191 0.24.007
192 ---------------------------------------------------------------------------------
193 fixed bug occurring when the spmatrix C API is used in an extension
194 consisting of more than one file. Now SPMATRIX_UNIQUE_SYMBOL has
195 to be defined. This broke the itsolvers.pcg function.
196
197 0.24.006
198 ---------------------------------------------------------------------------------
199 Bug fixed in spmatrix.h: LLMatType now exported properly
200
201 Genereation of boiler plate code for C API is automated now
202
203 More functions exported in C API
204
205 spmatrix module no longer uses extern symbols
206
207 delete_rows and delete_rowcols methods added for ll_mat objects
208
209 matvec and matvec_transp methods of csr_mat and sss_mat objects now
210 accept strided arrays as input
211
212 0.23.005
213 ---------------------------------------------------------------------------------
214 Header files are now installed to python directory tree. External
215 packages may now import pysparse.
216
217 0.23.004
218 ---------------------------------------------------------------------------------
219 new module Lib/pysparse_version.py defining pysparse version number
220
221 0.23.003
222 ---------------------------------------------------------------------------------
223 Updated README
224
225 Removed some illegal characters from superlu/superlu_timer.c
226
227 Removed SuperLU output from superlu/get_perm_c.c stating the matrix
228 reordering used.
229
230
231 0.23.002
232 ---------------------------------------------------------------------------------
233 SuperLU is now incorporated into pysparse. No need to build the
234 library separately from now on.
235
236 Now compiles under Win32. Tested under WinXP with MS Visual Studio 6.0
237 and the Inter Math Kernel Library for LAPACK and BLAS
238
239 Removed some unused local variables
240
241 Changed interface of LLMat_parse_index. Can now be safely used in sss_mat.c.
242
243
244 0.22.001
245 ---------------------------------------------------------------------
246 initial release
Something went wrong with that request. Please try again.