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
|
<html>
<body BGCOLOR="FFFFFF">
<h1>Docs: Changes: 2.0.24</h1>
<p align="center"><font color="#FF0000" size="5"><a name="CHANGES">CHANGES</a> in PETSc 2.0.24 (see new <a href="2024.html#Features">features</a>
below)</font></p>
<p><font size="4"><strong><u>General:</u></strong></font> <ul>
<li>Changed PETSC_ARCH on Windows platforms from nt to win32, and from nt_gnu to
win32_gnu.</li>
<li>Moved all the compiler/optimization information for each PETSC_ARCH from the various <strong>base.g</strong>,
<strong>base.O_c++</strong> etc. files all into the single file <strong>base_variables</strong>.
Now configuration for your system should require only editing the files ${PETSC_ARCH}/base
and ${PETSC_ARCH}/<strong>base.site</strong></li>
<li>The PETSC_ARCH <strong>rs6000</strong> is now intended for IBM rs6000 workstations
running MPICH or LAM, while <strong>rs6000_sp</strong> is for IBM SPs.</li>
<li>Removed the gvec directory from the distribution. If you are using this please let us
and know and we can provide it or provide alternative code to do what you need.</li>
<li>Routines with the suffix "<strong>withcoloring</strong>" have been changed to
the suffix "<strong>color</strong>" to get around the 31 character limit in
names.</li>
<li>The source code directory structure has changed slightly: pc and ksp are now
subdirectories of sles, and ao and da are subdirectories of a new directory called dm
(data management).</li>
<li>The PETSc libraries are now: -lpetscts -lpetscsnes -lpetscsles -lpetscdm -lpetscmat
-lpetscvec -lpetsc. Each library is now only dependent on libraries to the right of it in
the list above. For example, if you are only using the linear solvers, you
only need to list "-lpetscsles -lpetscdm -lpetscmat -lpetscvec -lpetsc" and if
you are only using the vectors you only need to list "-lpetscvec -lpetsc".</li>
<li>For users who previously could not use the PETSc makefiles because the PETSc rules for
making .o files conflicted with the user's rules, we have introduce a new file that may be
included by the user, <strong>${PETSC_DIR}/bmake/${PETSC_ARCH}/base_variables</strong>,
which defines all the useful PETSc makefile variables, but not the rules. This should make
it easier for people who previously had trouble using the PETSc makefiles. We highly
recommend using the PETSc makefiles when possible because they are immediately portable
across all systems that support PETSc.</li>
</ul>
<p><font size="4"><strong><u>AO (Application Orderings):</u></strong></font></p>
<p><font size="4"><strong><u>TS (Timestepping Solvers):</u></strong></font> <ul>
<li>Removed <strong>TSSetRHSJacobianDefault</strong>() - you can now call <strong>TSSetRHSJacobian</strong>(...
<strong>TSDefaultJacobianColor</strong>,....) directly from Fortran</li>
<li>Changed <strong>TSDefaultJacobianSlow()</strong> to <strong>TSDefaultJacobian()</strong></li>
</ul>
<p><font size="4"><strong><u>SNES (Nonlinear Solvers): </u></strong></font><ul>
<li>S<strong>NESSetConvergenceHistory()</strong> takes an additional argument; see the
manual page for details.</li>
<li>Changed the interface for matrix-free methods in SNES so that is more flexible, in
particular:<br>
Changed<strong> SNESDefaultMatrixFreeMatCreate()</strong> to <strong>MatSNESFDMFCreate()</strong>.<br>
Added routines<br>
<strong> MatSNESFDMFSetHHistory(Mat,Scalar *,int);<br>
MatSNESFDMFResetHHistory(Mat,Scalar *,int);<br>
MatSNESFDMFSetParameters(Mat,double,double);<br>
MatSNESFDMFGetH(Mat,Scalar *);<br>
MatSNESFDMFKSPMonitor(KSP,int,double,void *);</strong> and the corresponding<br>
option <strong>-snes_mf_ksp_monitor</strong><br>
Added support for an alternative way to compute the differencing parameter in matrix-free
methods<br>
using the Walker-Pernice srategy, accessible via the option <strong>-snes_mf_type</strong>
wp</li>
<li>Changed the calling sequence of <strong>SNESSetLineSearch() </strong>so that it accepts
an optional user-defined<br>
context variable. Likewise, changed the calling sequences of the various line search
routines, e.g., <strong>SNESCubicLineSearch(), SNESQuadraticLineSearch()</strong>, etc.</li>
</ul>
<p><font size="4"><strong><u>SLES (Linear Solvers):</u></strong></font> <ul>
<li>See PC and KSP </li>
</ul>
<p><font size="4"><strong><u>KSP (Krylov Subspace Methods):</u></strong></font> <ul>
<li>Changed the calling sequence of <strong>KSPComputeEigenvalues(KSP,int,double*,double*,int
*neig);</strong><br>
to return as the final value the number of eigenvalues computed; see the manual page for
more details.</li>
<li>Changed GMRES to use the Gram-Schmidt method with one optional step of iterative<br>
refinement as the default orthogonalization procedure, instead of the modified
Gram-Schmidt method.<br>
This change was urged by Rich Lehoucq. It could cause very minor differences in
convergence<br>
rates for codes run before and after this change.</li>
<li>The option <strong>-ksp_bsmonitor</strong> is not longer needed; use -<strong>ksp_truemonitor</strong>
for the same affect.</li>
<li>Added an additional argument to <strong>KSPSetResidualHistory()</strong>; see the manual
page for details. This change was suggested by Maria Murillo.<br>
</li>
</ul>
<p><font size="4"><strong><u>PC (Preconditioners): </u></strong></font><ul>
<li>Removed support for block Gauss-Seidel from the distribution; <strong>-pc_type bgs</strong>
is no longer supported.</li>
<li>The Eisenstat preconditioner (SSOR variant) now uses diagonal scaling by default, and
the options<br>
<strong>PCEisenstatUseDiagonalScaling()</strong> and <strong>-pc_eisenstat_diagonal_scaling</strong>
have been<br>
replaced with <strong>PCEisenstatNoDiagonalScaling()</strong> and <strong>-pc_eisenstat_no_diagonal_scaling</strong></li>
</ul>
<p><font size="4"><strong><u>MAT (Matrices): </u></strong></font><ul>
<li>Replaced<strong> MatGetSubMatrixCall</strong>() with <strong>MatReuse().</strong></li>
<li>Changed the calling sequence of <strong>MatILUFactor()</strong> and <strong>MatILUFactorSymbolic()</strong>
to increase<br>
flexibility of adding new ILU options.</li>
<li>Added an additional argument to <strong>MatCopy()</strong>; <strong>SAME_NONZERO_PATTERN</strong>
or <strong>DIFFERENT_NONZERO_PATTERN.</strong></li>
<li>Added an additional argument to <strong>MatDuplicate()</strong>, <strong>MAT_COPY_VALUES</strong>
or <strong>MAT_DO_NOT_COPY_VALUES.</strong></li>
<li>Changed the flag <strong>MAT_NEW_NONZERO_LOCATION_ERROR</strong> to <strong>MAT_NEW_NONZERO_LOCATION_ERR</strong>;
also changed <strong>MAT_NEW_ALLOCATION_ERROR</strong> to <strong>MAT_NEW_ALLOCATION_ERR</strong></li>
<li>Changed <strong>MatOrder</strong>xxx to <strong>MatOrdering</strong>xxx; changed <strong>MatOrderingType</strong>
to char* from enum</li>
<li>Changed <strong>Mat</strong>xxx<strong>Reordering</strong>xxx to <strong>Mat</strong>xxx<strong>Ordering</strong>xxx
changed command line option <strong>-mat_order</strong> to -<strong>mat_ordering_type; </strong>changed
<strong>MatColoringType </strong>to char* from enum</li>
<li>Changed <strong>Partitioning</strong>xxx to <strong>MatPartitioning</strong>xxx</li>
</ul>
<p><font size="4"><strong><u>DA (Distributed Arrays): </u></strong></font><ul>
<li><strong>DAGetInfo()</strong> has a new final argument, the stencil type. See the manual
page; to <br>
change your current code to work, you only need add an extra <strong>PETSC_NULL</strong>
as a final argument.</li>
</ul>
<p><font size="4"><strong><u>VEC (Vectors):</u></strong></font> <ul>
<li>After a call to <strong>VecCreate()</strong> you MUST either call <strong>VecSetType()</strong>
or <strong>VecSetFromOptions()</strong> on the vector before using it for any purpose. The
<strong>VecCreate</strong>xxx() routines (e.g., <strong>VecCreateSeq(), VecCreateMPI()</strong>)
do not need to have calls to VecSetType() or VecSetFromOptions() after them.</li>
<li>All <strong>VecGetArray()</strong> calls MUST now be matched with a <strong>VecRestoreArray()
</strong>call; previous PETSc versions did not force compliance with this rule.</li>
<li>The command line names for changing the vector type in <strong>VecCreate(</strong>) are
now <strong>-vec_type mpi</strong> or <strong>-vec_type shared</strong> rather than <strong>-vec_mpi</strong>
or <strong>-vec_shared</strong></li>
<li>Changed the <strong>VecType</strong> variable type from an enum type to a string.</li>
<li>Changed the calling sequence for <strong>VecGetType()</strong>; see the manual page for
details.</li>
<li>Changed <strong>VecGhostGetLocalRepresentation()</strong> and <strong>VecGhostRestoreLocalRepresentation()</strong>,<br>
to <strong>VecGhostGetLocalForm()</strong> and <strong>VecGhostRestoreLocalForm()</strong>,
respectively; the previous names had too many characters for some machines.</li>
</ul>
<p><font size="4"><strong><u>IS (Index Sets): </u></strong></font></p>
<p><font size="4"><strong><u>Draw (Graphics):</u></strong></font> <ul>
<li>Replaced<strong> VIEWER_DRAWX_</strong>.... with <strong>VIEWER_DRAW_</strong>....</li>
<li>The arguments of <strong>ViewerDrawGetDraw</strong>() have been changed; see the manual
page for details.</li>
<li><strong>DrawTensorContour()</strong> has changed; its final argument is now an array
rather than a vector and it can only be used for a single processor. For DA vectors you
can now use <strong>VecView()</strong> to create the contour plot.<br>
</li>
</ul>
<p><font size="4"><strong><u>Viewers: </u></strong></font><ul>
<li>Changed the names associated with the Matlab viewer to Socket viewer; this reflects our
plan to support additional graphics engines with PETSc.</li>
<li>Changed <strong>ViewerFileOpenASCII()</strong> to <strong>ViewerASCIIOpen().</strong></li>
<li>Changed <strong>ViewerDrawOpenX(</strong>) to <strong>ViewerDrawOpen().</strong></li>
<li>Changed <strong>ViewerFileOpenBinary()</strong> to <strong>ViewerBinaryOpen().</strong></li>
<li>Changed <strong>ViewerType</strong> from an enum to a char *.</li>
</ul>
<p><font size="4"><strong><u>System:</u></strong></font> <ul>
<li>Changed the calling sequence of <strong>PetscGetHomeDirectory(char dir[],int
maxlen)</strong>.</li>
<li>Changed the calling sequence of <strong>PetscFixFilename(const char filein[],char
fileout[]).</strong></li>
<li>Changed the calling sequence of <strong>PetscGetDate().</strong></li>
</ul>
<p><font size="4"><strong><u>Error Handling:</u></strong></font></p>
<p><font size="4"><strong><u>Event Logging:</u></strong></font></p>
<p><font size="4"><strong><u>Fortran Interface:</u></strong></font> </p>
<p> </p>
<p align="center"><font color="#FF0000" size="5">New <a name="Features">Features</a> in
PETSc 2.0.24 (see <a href="2024.html#CHANGES">changes</a> above)</font></p>
<p><font size="4"><strong><u>General:</u></strong></font> <ul>
<li>Add the routine <strong>OptionsGetAll().</strong></li>
<li>Added support for Sun Solaris running on Intel processors; PETSC_ARCH solaris_x86.</li>
<li>Added option -<strong>optionsleft_off</strong> so BOPT=g version of code will not print
warnings about unused options.</li>
<li>The environment variable PETSC_ARCH is now required to use the NT distribution.</li>
<li>Added the routine <strong>OptionsGetLogical().</strong></li>
<li>We now include PETSC_ARCH of solaris_gnu, rs6000_gnu for those systems when using the
gnu compilers.</li>
<li>Thanks to the generosity of the FreeBSD organization, we are now supporting freebsd
again.</li>
<li>Added optional filename to <strong>-log_info</strong> <filename> option. PETSc
will open a seperate file, called filename.rank for each processor and print the output
there.</li>
<li>Added <strong>PetscTypeCompare()</strong>; see the manual page.</li>
</ul>
<p><font size="4"><strong><u>AO (Application Orderings):</u></strong></font> <ul>
<li><strong>AOPetscToApplicationIS()</strong> and <strong>AOApplicationToPetscIS()</strong>,
which used to crash<br>
when passed a stride index set, now handle them correctly.</li>
</ul>
<p><font size="4"><strong><u>TS (Timestepping Solvers):</u></strong></font> <ul>
<li>Added <strong>TSSetTypeFromOptions().</strong></li>
</ul>
<p><font size="4"><strong><u>SNES (Nonlinear Solvers): </u></strong></font><ul>
<li>Added <strong>-snes_vecmonitor</strong> and <strong>SNESVecViewMonitor()</strong> to
allow graphically monitoring the solution.</li>
<li>Added <strong>SNESSetTypeFromOptions().</strong></li>
<li>Added the routine <strong>SNESSetLineSearchCheck()</strong>, which sets a routine to
check the validity of new iterates computed by line search routines; see the manual page
for details. </li>
</ul>
<p><font size="4"><strong><u>SLES (Linear Solvers):</u></strong></font> <ul>
<li>See PC and KSP </li>
<li>Added <strong>SLESSetTypesFromOptions().</strong></li>
</ul>
<p><font size="4"><strong><u>KSP (Krylov Subspace Methods):</u></strong></font> <ul>
<li>Added the function <strong>KSPSetAvoidNorms()</strong> and corresponding option <strong>-ksp_avoid_norms</strong>.
For the methods CG, Richardson, Bi-CG-stab, CR, and CGS, this eliminates the computation
of the residual norm at each iteration (which is an expensive, collective operation). Then
the convergence criterion depends only on the maximum number of iterations the users sets
with <strong>KSPSetTolerances().</strong> This is useful when using the Krylov method as a
smoother with a fixed number of iterations (say, e.g., 3 iterations).</li>
<li>Added the KSP type of <strong>KSPBICG </strong>(contributed by Victor Eijkhout) and
added support to some of the preconditioners to apply the preconditioner transpose
(required by KSPBICG).</li>
<li>Added <strong>-ksp_vecmonitor</strong> and <strong>KSPVecViewMonitor()</strong> to allow
graphically monitoring the solution's convergence.</li>
<li>Added <strong>KSPSetTypeFromOptions().</strong></li>
</ul>
<p><font size="4"><strong><u>PC (Preconditioners): </u></strong></font><ul>
<li>Added <strong>PCLUSetReuseReordering()</strong> and <strong>PCLUSetReuseFill()</strong>
so that SLES can reuse the reordering and fill information for a new factorization of a
matrix with slightly different nonzero structure than the previous factorization. The
corresponding command line options are -<strong>pc_lu_reuse_reordering</strong> and <strong>-pc_lu_reuse_fill</strong>.
These two options only have an effect when the <strong>SLESSetOperators()</strong> option<br>
of <strong>DIFFERENT_NONZERO_PATTERN</strong> is used; otherwise the ordering and fill
information is always <br>
reused.</li>
<li>Added <strong>PCGetVector()</strong>, not expected to be used by application
programmers.</li>
<li>Added the option <strong>-pc_ilu_diagonal_fill</strong> and the corresponding
routine <strong>PCILUSetAllowDiagonalFill()</strong>; see the manual page for details;
helps ILU(k) on matrices with missing diagonal entries. Suggested by David Hysom.</li>
<li>Added <strong>PCSetTypeFromOptions().</strong></li>
</ul>
<p><font size="4"><strong><u>MAT (Matrices): </u></strong></font><ul>
<li>Added <strong>MatSeqAIJSetColumnIndices()</strong> and <strong>MatSeqBAIJSetColumnIndices()</strong>.</li>
<li>If the option <strong>MAT_IGNORE_OFF_PROC_ENTRIES</strong> is used, then the routines <strong>MatAssemblyBegin()</strong><br>
and <strong>MatAssemblyEnd()</strong> will not cause any parallel communication or
synchronization.</li>
<li>If the option <strong>MAT_NO_NEW_NONZERO_LOCATIONS</strong> is used, one less collective
call will be made in the <strong>MatAssemblyEnd()</strong> routines for MPIAIJ and
MPIBAIJ.</li>
<li>Fixed bugs in <strong>MatReorderForNonzeroDiagonal()</strong> and made it more robust. </li>
<li>Fixed bugs in factorization for AIJ matrices with variable size inodes.</li>
<li>Indices that are negative passed into <strong>MatSetValues...()</strong> are now ignored
(rather than generating an error). This allows easily keeping homogeneous Dirichlet
boundary nodes out of the vectors and matrices for unstructured grid problems; just
indicate those node numbers with negative numbers.</li>
<li>Added additional optimized code for BAIJ matrices with block sizes 2, 3, 5, 6, and 7.</li>
<li>Improved the performance of <strong>MatSetValues()</strong> and <strong>MatAssemblyXXX()
</strong>for parallel matrices a great deal.</li>
<li>Added the option <strong>-matstash_initial_size</strong> <stashsize> and <strong>-matstash_block_initial_size</strong>
<stashsize> and the routine <strong>MatSetStashInitialSize() </strong>to allow
further optimization of off-processor<strong> MatSetValues().</strong></li>
</ul>
<p><font size="4"><strong><u>DA (Distributed Arrays): </u></strong></font><ul>
<li><font size="3">Added graphics support to DA vectors in 1d and 2d, with contour plotters
for 2d.</font></li>
</ul>
<p><font size="4"><strong><u>VEC (Vectors):</u></strong></font> <ul>
<li>Extended <strong>VecCreate()</strong> by allowing registration of new vector types with <strong>VecRegister()</strong>
see the manual page for details.</li>
<li>Indices that are negative passed into <strong>VecSetValues...()</strong> are now ignored
(rather than generating an error). This allows easily keeping homogeneous Dirichlet
boundary nodes out of the vectors and matrices for unstructured grid problems; just
indicate those node numbers with negative numbers.</li>
<li>Added <strong>VecDotBegin()</strong>, <strong>VecDotEnd(),</strong> <strong>VecNormBegin()</strong>,
<strong>VecNormEnd(), </strong>which reduce communication overhead in parallel;<strong> </strong>see
the manual pages for details (suggested by Victor Eijkhout).</li>
<li>If the option <strong>VEC_IGNORE_OFF_PROC_ENTRIES</strong> is used, the <strong>VecAssemblyBegin()</strong>
and <strong>VecAssemblyEnd()</strong> routines will not cause any parallel communication
or synchronization (previously <strong>VecAssemblyBegin()</strong> still used a barrier).</li>
<li>Added <strong>VIEWER_FORMAT_ASCII_INDEX</strong> which causes the index for each vector
entry to be printed with the vector values.</li>
<li>Added three new vector scatter options; <strong>-vecscatter_merge</strong> and <strong>-vecscatter_ssend</strong>
and<br>
<strong>-vecscatter_sendfirst</strong>; these may improve performance a great deal when
using MPICH for MPI on ethernet or fast ethernet.</li>
<li>Added <strong>VecStrideNorm(), VecStrideScatter()</strong>, and <strong>VecStrideGather().</strong></li>
<li>Improved the performance of <strong>VecSetValues()</strong> and <strong>VecAssemblyXXX()
</strong>for parallel vectors.</li>
<li>Added the option <strong>-vecstash_initial_size</strong> <stashsize> and <strong>-vecstash_block_initial_size</strong>
<stashsize> and the routine <strong>VecSetStashInitialSize() </strong>to allow
further optimization of off-processor <strong>VecSetValues()</strong>.<br>
</li>
</ul>
<p><font size="4"><strong><u>IS (Index Sets): </u></strong></font><ul>
<li>Added <strong>ISStrideToGeneral()</strong>, which converts a stride type index set to a
general type.</li>
</ul>
<p><font size="4"><strong><u>Draw (Graphics):</u></strong></font></p>
<p><font size="4"><strong><u>Viewers: </u></strong></font><ul>
<li>Added <strong>ViewerASCIIPushTab()</strong>, <strong>ViewerASCIIPopTab()</strong> and <strong>ViewerASCIIPrintf()</strong>.
These allow multi-level display of solver information (tabing the text further for each
inner solver). Thanks to Victor Eijkhout and Matt Knepley for pushing this. </li>
</ul>
<p><font size="4"><strong><u>System:</u></strong></font></p>
<p><font size="4"><strong><u>Error Handling:</u></strong></font></p>
<p><font size="4"><strong><u>Event Logging:</u></strong></font></p>
<p><font size="4"><strong><u>Fortran Interface:</u></strong></font> <ul>
<li>Added Fortran 90 interface support for HP Convex and Solaris.machine.<br>
</li>
</ul>
</body>
</html>
|
