-
Notifications
You must be signed in to change notification settings - Fork 59
/
sparsepak.html
957 lines (862 loc) · 27.9 KB
/
sparsepak.html
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
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
<html>
<head>
<title>
SPARSEPAK - (Obsolete) Waterloo Sparse Matrix Package
</title>
</head>
<body bgcolor="#EEEEEE" link="#CC0000" alink="#FF3300" vlink="#000055">
<h1 align = "center">
SPARSEPAK <br> (Obsolete) Waterloo Sparse Matrix Package
</h1>
<hr>
<p>
<b>SPARSEPAK</b>
is a FORTRAN90 library which
solves large sparse systems of linear equations.
</p>
<p>
<b>SPARSEPAK</b> is an old version of the Waterloo Sparse Matrix Package.
<b>SPARSEPAK</b> can carry out direct solution of large sparse linear
systems. Only positive definite matrices should be used with this
program. Three different storage methods, and five solution
algorithms are supported.
</p>
<p>
The methods available are:
<ul>
<li>
<b>1WD</b>: One way dissection, partitioned tree storage.
</li>
<li>
<b>ND</b>: Nested dissection, compressed storage.
</li>
<li>
<b>QMD</b>: Quotient minimum degree, compressed storage.
</li>
<li>
<b>RCM</b>: Reverse Cuthill-McKee with envelope storage.
</li>
<li>
<b>RQT</b>: Refined quotient tree, partitioned tree storage.
</li>
</ul>
</p>
<p>
This version is not the most recent one. The most recent version
requires a license from the University of Waterloo, and includes
support for large sparse linear least squares problems and other features.
</p>
<p>
SPARSPAK offers five different methods for the direct solution of
a symmetric positive definite sparse linear system. Subroutines are
provided which try to reorder the matrix before factorization so that
as little storage as possible is required.
</p>
<p>
The five algorithms all follow the same steps:
<ol>
<li>
Determine the structure of the matrix.
</li>
<li>
Reorder rows and columns for efficiency.
</li>
<li>
Store numerical values in matrix and right hand side.
</li>
<li>
Factor the matrix.
</li>
<li>
Solve the system (may be repeated for multiple right hand sides).
</li>
<li>
Unpermute the solution.
</li>
</ol>
Each step requires calls to one or more subroutines in the group
associated with a particular method.
</p>
<p>
Methods available are:
<table>
<tr>
<th>Symbol</th><th>Description</th>
</tr>
<tr>
<td>1WD</td><td>One way dissection, partitioned tree storage.</td>
</tr>
<tr>
<td>ND</td><td>Nested dissection, compressed storage.</td>
</tr>
<tr>
<td>QMD</td><td>Quotient minimum degree, compressed storage.</td>
</tr>
<tr>
<td>RCM</td><td>Reverse Cuthill-McKee with envelope storage.</td>
</tr>
<tr>
<td>RQT</td><td>Refined quotient tree, partitioned tree storage.</td>
</tr>
</table>
</p>
<h2>
Data Structures
</h2>
<p>
The following is a list of the vectors required for the various methods,
including their type, size, and whether they are set by the user or not.
Unfortunately, for most of the arrays, although their values are
determined by the package, their minimum size is not known beforehand,
and is often hard to estimate. You must examine the reference to get
a better idea of how large to dimension arrays like XBLK and NZSUB,
for example.
</p>
<pre>
Values Used by:
Name Set by Size Type 1WD RCM RQT ND QMD
IADJ User varies I X X X X X
BNUM N I X X
DEG N I X
DIAG User N R X X X X X
ENV User varies R X X X
FATHER NBLKS I X X
FIRST N I X X X X
INVPRM N I X X X X X
IXLNZ N+1
LINK N I X X
LNZ R X X
LS N I X X X
MARKER N I X X X
MRGLNK N I X X
NBRHD I X
NODLVL N I X
NONZ R X X
NZSUB I X X
NZSUBS I X X
OVRLP X
PERM N I X X X X X
QLINK N I X
QSIZE N I X
RCHLNK N I X X
RCHSET N I X X
RHS User N R X X X X X
SEP I X
SUBG N I X X
TEMP N R X X X X
XADJ User N+1 I X X X X X
XBLK NBLKS+1 I X X
XENV User N+1 I X X X
XLNZ N I X X
XLS N+1 I X X X
XNONZ N+1 I X X
XNZSUB I X X
</pre>
<h2>
Adjacency Information
</h2>
<p>
You are responsible for defining the adjacency or nonzero structure of the
original matrix. This information is required by the routines to
efficiently reorder the system.
</p>
<p>
Suppose our linear system is a 6 by 6 matrix, where the entries
marked 'X' are nonzero:
<pre>
XX000X
XXXX00
0XX0X0
0X0X00
00X0XX
X000XX
</pre>
</p>
<p>
The adjacency information records, for each row I, which columns
(besides column I) contain a nonzero entry, and, for each column I,
which rows besides row I are nonzero.
</p>
<p>
In the example, the nonzero list for row I=1 is 2 and 6. This is because
A(1,2) and A(1,6) are nonzero. The whole list is
<table>
<tr>
<th>Row</th><th>Nonzero columns</th>
</tr>
<tr>
<td>1</td><td>2, 6</td>
</tr>
<tr>
<td>2</td><td>1, 3, 4</td>
</tr>
<tr>
<td>3</td><td>2, 5</td>
</tr>
<tr>
<td>4</td><td>2</td>
</tr>
<tr>
<td>5</td><td>3, 6</td>
</tr>
<tr>
<td>6</td><td>1, 5</td>
</tr>
</table>
Note that in the list of nonzeroes for row I, we never include I
itself. It is assumed that I is nonzero, or will become so during
the factorization.
</p>
<p>
The adjacency information consists of two vectors, IADJ and XADJ. The
vector IADJ contains the concatenated list of nonzeroes. The auxilliary
vector XADJ holds the location of the start of each row list in IADJ.
Thus, XADJ(1) should always be 1. XADJ(2) should be the location in
IADJ where the nonzero information for row 2 begins. Finally, XADJ(N+1)
points to the first unused entry in IADJ. For the example,
<table>
<tr>
<td>Index</td><td>1</td><td>2</td><td>3</td><td>4</td><td>5</td><td>6</td>
<td>7</td><td>8</td><td>9</td><td>10</td><td>11</td><td>12</td><td>13</td>
</tr>
<tr>
<td>IADJ</td><td>2</td><td>6</td><td>1</td><td>3</td><td>4</td>
<td>2</td><td>5</td><td>2</td><td>3</td><td>6</td><td>1</td><td>3</td>
</tr>
<tr>
<td>XADJ</td><td>1</td><td>3</td><td>6</td><td>8</td><td>9</td><td>11</td>
<td>13</td>
</tr>
</table>
</p>
<p>
You can see that to retrieve information about row 3, you would first
check XADJ(3) and XADJ(3+1), which have the values 6 and 8. This tells
us that the list for row 3 begins at IADJ(6) and stops at IADJ(8-1).
So the nonzeroes in row 3 occur in columns IADJ(6)=2 and IADJ(7)=5.
This seems a lot of work, but this construction can be automated, and
the storage savings can be enormous.
</p>
<h2>
Envelope Information
</h2>
<p>
Assume, as usual, that we are working with a symmetric matrix. The
lower row bandwidth of row I, denoted by BETA(I), is defined in the
obvious way to mean the distance between the diagonal entry in column I
and the first nonzero entry in row I.
</p>
<p>
The envelope of a matrix is then the collection of all subdiagonal
matrix positions (I,J) which lie within the lower row bandwidth of
their given row. This concept is sometimes also called profile
or skyline storage, particularly when columns are considered rather
than rows.
</p>
<p>
To store a matrix in envelope format, we require two objects,
an array ENV to hold the actual numbers, and a pointer array
XENV to tell us where each row starts and ends.
</p>
<p>
Let us return to our earlier example:
<pre>
XX000X
XXXX00
0XX0X0
0X0X00
00X0XX
X000XX
</pre>
</p>
<p>
We can analyze this matrix as follows:
<table>
<tr>
<th>Row</th><th>Bandwidth</th>
</tr>
<tr>
<td>1</td><td>0</td>
</tr>
<tr>
<td>2</td><td>1</td>
</tr>
<tr>
<td>3</td><td>1</td>
</tr>
<tr>
<td>4</td><td>2</td>
</tr>
<tr>
<td>5</td><td>2</td>
</tr>
<tr>
<td>6</td><td>5</td>
</tr>
</table>
</p>
<p>
This tells us that there are a total of 11 entries in the lower envelope.
XENV(I) tells us where the first entry of row I is stored, and XENV(I+1)-1
tells us where the last one is. For our example, then
<table>
<tr>
<th>Index</th><th>XENV</th>
</tr>
<tr>
<td>1</td><td>1</td>
</tr>
<tr>
<td>2</td><td>1</td>
</tr>
<tr>
<td>3</td><td>2</td>
</tr>
<tr>
<td>4</td><td>3</td>
</tr>
<tr>
<td>5</td><td>5</td>
</tr>
<tr>
<td>6</td><td>7</td>
</tr>
<tr>
<td>7</td><td>12</td>
</tr>
</table>
</p>
<p>
Note that the "extra" entry in XENV allows us to treat the last row in
the same way as the other ones. Now the ENV array will require 11
entries, and they can be stored and accessed using the XENV array.
</p>
<h2>
Setting Matrix Values
</h2>
<p>
Once you have defined the adjacency structure, you have set up empty slots
for your matrix. Into each position which you have set up, you can
store a numerical value A(I,J). You can use the user utility routines
ADDRCM, ADDRQT and ADDCOM to store values in A for you.
</p>
<p>
Alternatively, if you want to do this on your own, you must be
familiar with the storage method used. For example, in the envelope
storage scheme you must know, given I and J, in what entry of DIAG and
ENV you should store the value. Interested readers are referred to
the reference for more information on this.
</p>
<h2>
Example of 1WD Method
</h2>
<p>
<i>
Create the adjacency structure XADJ, IADJ, and reorder the variables
and equations.
</i>
<pre>
call gen1wd ( n, xadj, iadj, nblks, xblk, perm, xls, ls )
call perm_inverse ( n, perm, invprm )
call fnbenv ( xadj, iadj, perm, invprm, nblks, xblk, xenv, ienv,
marker, rchset, n )
</pre>
<i>
After zeroing out NONZ, DIAG and ENV, store the numerical values
of the matrix and the right hand side. You can use the user
routines as shown below.
</i>
<pre>
call addrqt ( isub, jsub, value, invprm, diag, xenv, env, xnonz, nonz,
nzsubs, n )
call addrhs ( invprm, isub, n, rhs, value )
</pre>
<i>
Factor and solve the system.
</i>
<pre>
call ts_factor ( nblks, xblk, father, diag, xenv, env, xnonz, nonz,
nzsubs, temp, first, n )
call ts_solve ( nblks, xblk, diag, xenv, env, xnonz, nonz, nzsubs,
rhs, temp, n )
</pre>
<i>
Unpermute the solution stored in RHS.
</i>
<pre>
call perm_rv ( n, rhs, perm )
</pre>
</p>
<h2>
Example of ND Method
</h2>
<p>
<i>
Create the adjacency structure XADJ, IADJ. Then reorder the variables and
equations.
</i>
<pre>
call gennd ( n, xadj, iadj, perm, xls, ls )
call perm_inverse ( n, perm, invprm )
call smb_factor ( n, xadj, iadj, perm, invprm, xlnz, nofnz, xnzsub,
nzsub, nofsub, rchlnk, mrglnk )
</pre>
<i>
Zero out the matrix and right hand side storage in DIAG, LNZ, and RHS,
then store numerical values. You can build up the values by calling
the following routines repeatedly.
</i>
<pre>
call addcom ( isub, jsub, value, invprm, diag, lnz, xlnz, nzsub,
xnzsub, n )
call addrhs ( invprm, isub, n, rhs, value )
</pre>
<i>
Factor and solve the system.
</i>
<pre>
call gs_factor ( n, xlnz, lnz, xnzsub, nzsub, diag, link, first )
call gs_solve ( n, xlnz, lnz, xnzsub, nzsub, diag, rhs )
</pre>
<i>
Unpermute the solution stored in RHS.
</i>
<pre>
call perm_rv ( n, rhs, perm )
</pre>
</p>
<h2>
Example of QMD Method
</h2>
<p>
<i>
Create the adjacency structure in XADJ, IADJ. Reorder the variables and
equations.
</i>
<pre>
call genqmd ( n, xadj, adjnc2, perm, invprm, marker, rchset,
nbrhd, qsize, qlink, nofsub )
call perm_inverse ( n, perm, invprm )
call smb_factor ( n, xadj, iadj, perm, invprm, xlnz, nofnz, xnzsub,
nzsub, nofsub, rchlnk, mrglnk )
</pre>
<i>
Zero out the matrix and right hand side storage in DIAG, LNZ and RHS.
Store numerical values, perhaps using the following routines repeatedly.
</i>
<pre>
call addcom ( isub, jsub, value, invprm, diag, lnz, xlnz, nzsub, xnzsub )
call addrhs ( invprm, isub, n, rhs, value )
</pre>
<i>
Factor and solve.
</i>
<pre>
call gs_factor ( n, xlnz, lnz, xnzsub, nzsub, diag, link, first )
call gs_solve ( n, xlnz, lnz, xnzsub, nzsub, diag, rhs )
</pre>
<i>
Unpermute the solution stored in RHS.
</i>
<pre>
call perm_rv ( n, rhs, perm )
</pre>
</p>
<h2>
Example of RCM Method
</h2>
<p>
<i>
Create the adjacency structure XADJ, IADJ. Then call the following
routines to reorder the system.
</i>
<pre>
call genrcm ( n, xadj, iadj, perm, xls )
call perm_inverse ( n, perm, invprm )
call fnenv ( n, xadj, iadj, perm, invprm, xenv, ienv, bandw )
</pre>
<i>
Now store actual matrix values in DIAG, ENV and RHS, calling the
following two routines as often as necessary. Be sure to first zero
out the storage for DIAG, ENV and RHS. Only call ADDRCM for A(I,J)
with I greater than or equal to J, since only the lower half of
the matrix is stored.
</i>
<pre>
call addrcm ( isub, jsub, value, invprm, diag, xenv, env, n )
call addrhs ( invprm, isub, n, rhs, value )
</pre>
<i>
Now factor and solve the system.
</i>
<pre>
call es_factor ( n, xenv, env, diag )
call el_solve ( n, xenv, env, diag, rhs )
call eu_solve ( n, xenv, env, diag, rhs )
</pre>
<i>
Unpermute the solution stored in RHS.
</i>
<pre>
call perm_rv ( n, rhs, perm )
</pre>
</p>
<h2>
Example of RQT Method
</h2>
<p>
<i>
Create the adjacency structure XADJ, IADJ. Reorder the variables and
equations.
</i>
<pre>
call genrqt ( n, xadj, iadj, nblks, xblk, perm, xls, ls, nodlvl )
call block_shuffle ( xadj, iadj, perm, nblks, xblk, xls, n )
call perm_inverse ( n, perm, invprm )
call fntadj ( xadj, iadj, perm, nblks, xblk, father, n )
call fntenv ( xadj, iadj, perm, invprm, nblks, xblk, xenv, ienv, n )
call fnofnz ( xadj, iadj, perm, invprm, nblks, xblk, xnonz, nzsubs,
nofnz, n )
</pre>
<i>
After zeroing out NONZ, DIAG and ENV, store the values of the matrix
and the right hand side. You can use the user utility routines as
shown below.
</i>
<pre>
call addrqt ( isub, jsub, value, invprm, diag, xenv, env, xnonz, nonz,
nzsubs, n )
call addrhs ( invprm, isub, n, rhs, value )
</pre>
<i>
Factor and solve the system.
</i>
<pre>
call ts_factor ( nblks, xblk, father, diag, xenv, env, xnonz, nonz,
nzsubs, temp, first, n )
call ts_solve ( nblks, xblk, diag, xenv, env, xnonz, nonz, nzsubs,
rhs, temp, n )
</pre>
<i>
Unpermute the solution stored in RHS.
</i>
<pre>
call perm_rv ( n, rhs, perm )
</pre>
</p>
<h3 align = "center">
Languages:
</h3>
<p>
<b>SPARSEPAK</b> is available in
<a href = "../../f77_src/sparsepak/sparsepak.html">a FORTRAN77 version</a> and
<a href = "../../f_src/sparsepak/sparsepak.html">a FORTRAN90 version</a>.
</p>
<h3 align = "center">
Related Data and Programs:
</h3>
<p>
<a href = "../../c_src/csparse/csparse.html">
CSPARSE</a>,
a C library which
carries out direct sparse matrix operations.
</p>
<p>
<a href = "../../f_src/dlap/dlap.html">
DLAP</a>,
a FORTRAN90 library which
solves sparse linear systems.
</p>
<p>
<a href = "../../f_src/hb_io/hb_io.html">
HB_IO</a>,
a FORTRAN90 library which
reads and writes sparse linear
systems stored in the Harwell-Boeing Sparse Matrix format.
</p>
<p>
<a href = "../../f77_src/hb_to_st/hb_to_st.html">
HB_TO_ST</a>,
a FORTRAN77 program which
converts the sparse matrix information stored in a Harwell-Boeing
file into a sparse triplet file.
</p>
<p>
<a href = "../../f_src/mgmres/mgmres.html">
MGMRES</a>,
a FORTRAN90 library which
applies the restarted GMRES
algorithm to solve a sparse linear system.
</p>
<p>
<a href = "../../f_src/mm_io/mm_io.html">
MM_IO</a>,
a FORTRAN90 library which
reads and writes sparse linear
systems stored in the Matrix Market format.
</p>
<p>
<a href = "../../f_src/rcm/rcm.html">
RCM</a>,
a FORTRAN90 library which
implements the reverse Cuthill-McKee
reordering.
</p>
<p>
<a href = "../../data/sparse_cc/sparse_cc.html">
SPARSE_CC</a>,
a data directory which
contains a description and examples of the CC format,
("compressed column") for storing a sparse matrix,
including a way to write the matrix as a set of three files.
</p>
<p>
<a href = "../../data/sparse_cr/sparse_cr.html">
SPARSE_CR</a>,
a data directory which
contains a description and examples of the CR format,
("compressed row") for storing a sparse matrix,
including a way to write the matrix as a set of three files.
</p>
<p>
<a href = "../../f_src/sparsekit/sparsekit.html">
SPARSEKIT</a>,
a FORTRAN90 library which
carries out various
operations on sparse matrices.
</p>
<h3 align = "center">
Reference:
</h3>
<p>
<ol>
<li>
Alan George, Joseph Liu,<br>
Computer Solution of Large Sparse Positive Definite Matrices,<br>
Prentice Hall, 1981,<br>
ISBN: 0131652745,<br>
LC: QA188.G46.
</li>
<li>
Norman Gibbs, William Poole, Paul Stockmeyer,<br>
An Algorithm for Reducing the Bandwidth
and Profile of a Sparse Matrix,<br>
SIAM Journal on Numerical Analysis,<br>
Volume 13, Number 2, April 1976, pages 236-250.
</li>
<li>
Norman Gibbs,<br>
Algorithm 509:
A Hybrid Profile Reduction Algorithm,<br>
ACM Transactions on Mathematical Software,<br>
Volume 2, Number 4, December 1976, pages 378-387.
</li>
<li>
Donald Kreher, Douglas Simpson,<br>
Combinatorial Algorithms,<br>
CRC Press, 1998,<br>
ISBN: 0-8493-3988-X,<br>
LC: QA164.K73.
</li>
</ol>
</p>
<h3 align = "center">
Source Code:
</h3>
<p>
<ul>
<li>
<a href = "sparsepak.f90">sparsepak.f90</a>, the source code;
</li>
<li>
<a href = "sparsepak.sh">sparsepak.sh</a>,
commands to compile the source code;
</li>
</ul>
</p>
<h3 align = "center">
Examples and Tests:
</h3>
<p>
<ul>
<li>
<a href = "sparsepak_prb.f90">sparsepak_prb.f90</a>, a sample problem;
</li>
<li>
<a href = "sparsepak_prb.sh">sparsepak_prb.sh</a>,
commands to compile, link and run the sample problem;
</li>
<li>
<a href = "sparsepak_prb_output.txt">sparsepak_prb_output.txt</a>,
output from the sample problem;
</li>
</ul>
</p>
<h3 align = "center">
List of Routines:
</h3>
<p>
<ul>
<li>
<b>ADDCOM</b> adds values to a matrix stored in compressed storage scheme.
</li>
<li>
<b>ADDRCM</b> adds values to a matrix stored in the RCM scheme.
</li>
<li>
<b>ADDRHS</b> adds a quantity to a specific entry of the right hand side.
</li>
<li>
<b>ADDRQT</b> adds values to a matrix stored in the implicit block storage scheme.
</li>
<li>
<b>ADJ_ENV_SIZE</b> computes the envelope size for an adjacency structure.
</li>
<li>
<b>ADJ_PRINT</b> prints the adjacency information stored in ADJ_ROW and ADJ.
</li>
<li>
<b>ADJ_SET</b> sets up the adjacency information ADJ_ROW and ADJ.
</li>
<li>
<b>ADJ_SHOW</b> displays a symbolic picture of a matrix.
</li>
<li>
<b>BLOCK_SHUFFLE</b> renumbers the nodes of each block to reduce its envelope.
</li>
<li>
<b>DEGREE</b> computes node degrees in a connected component, for the RCM method.
</li>
<li>
<b>EL_SOLVE</b> solves a lower triangular system stored in the envelope format.
</li>
<li>
<b>ES_FACTOR</b> factors a positive definite envelope matrix into L * L'.
</li>
<li>
<b>EU_SOLVE</b> solves an upper triangular system stored in the envelope format.
</li>
<li>
<b>FNBENV</b> finds the envelope of the diagonal blocks of a Cholesky factor.
</li>
<li>
<b>FNDSEP</b> finds a small separator for a connected component in a graph.
</li>
<li>
<b>FNENV</b> finds the envelope structure of a permuted matrix.
</li>
<li>
<b>FNLVLS</b> generates a rooted level structure, as part of the RQT method.
</li>
<li>
<b>FNOFNZ</b> finds columns of off-block-diagonal nonzeros.
</li>
<li>
<b>FNSPAN</b> finds the span of a level subgraph subset, as part of the RQT method.
</li>
<li>
<b>FNTADJ</b> determines the quotient tree adjacency structure for a graph.
</li>
<li>
<b>FNTENV</b> determines the envelope index vector.
</li>
<li>
<b>FN1WD</b> finds one-way dissectors of a connected component for the 1WD method.
</li>
<li>
<b>GENND</b> finds a nested dissection ordering for a general graph.
</li>
<li>
<b>GENQMD</b> implements the quotient minimum degree algorithm.
</li>
<li>
<b>GENRCM</b> finds the reverse Cuthill-Mckee ordering for a general graph.
</li>
<li>
<b>GENRQT</b> determines a partitioned ordering using refined quotient trees.
</li>
<li>
<b>GEN1WD</b> partitions a general graph, for the 1WD method.
</li>
<li>
<b>GS_FACTOR:</b> symmetric factorization for a general sparse system.
</li>
<li>
<b>GS_SOLVE</b> solves a factored general sparse system.
</li>
<li>
<b>I4_SWAP</b> switches two integer values.
</li>
<li>
<b>I4VEC_COPY</b> copies one integer vector into another.
</li>
<li>
<b>I4VEC_INDICATOR</b> sets an integer vector to the indicator vector.
</li>
<li>
<b>I4VEC_REVERSE</b> reverses the elements of an integer vector.
</li>
<li>
<b>I4VEC_SORT_INSERT_A</b> uses an ascending insertion sort on an integer vector.
</li>
<li>
<b>LEVEL_SET</b> generates the connected level structure rooted at a given node.
</li>
<li>
<b>PERM_INVERSE</b> produces the inverse of a given permutation.
</li>
<li>
<b>PERM_RV</b> undoes the permutation of the right hand side.
</li>
<li>
<b>QMDMRG</b> merges indistinguishable nodes for the QMD method.
</li>
<li>
<b>QMDQT</b> performs the quotient graph transformation.
</li>
<li>
<b>QMDRCH</b> determines the reachable set of a node, for the QMD method.
</li>
<li>
<b>QMDUPD</b> updates the node degrees, for the QMD method.
</li>
<li>
<b>RCM</b> renumbers a connected component by the reverse Cuthill McKee algorithm.
</li>
<li>
<b>RCM_SUB</b> finds the reverse Cuthill McKee ordering for a given subgraph.
</li>
<li>
<b>REACH</b> determines the reachable set of a node through a subset in a subgraph.
</li>
<li>
<b>ROOT_FIND</b> finds pseudo-peripheral nodes.
</li>
<li>
<b>RQTREE</b> finds a quotient tree ordering for a component, in the RQT method.
</li>
<li>
<b>SMB_FACTOR</b> performs symbolic factorization on a permuted linear system.
</li>
<li>
<b>TIMESTAMP</b> prints the current YMDHMS date as a time stamp.
</li>
<li>
<b>TS_FACTOR</b> performs the symmetric factorization of a tree-partitioned system.
</li>
<li>
<b>TS_SOLVE</b> solves a tree-partitioned factored system.
</li>
</ul>
</p>
<p>
You can go up one level to <a href = "../f_src.html">
the FORTRAN90 source codes</a>.
</p>
<hr>
<i>
Last revised on 24 February 2007.
</i>
<!-- John Burkardt -->
</body>
</html>