Amesos Package Browser (Single Doxygen Collection)  Development
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Groups Pages
amesos_amd_2.c
Go to the documentation of this file.
1 /* ========================================================================= */
2 /* === AMD_2 =============================================================== */
3 /* ========================================================================= */
4 
5 /* ------------------------------------------------------------------------- */
6 /* AMD, Copyright (c) Timothy A. Davis, */
7 /* Patrick R. Amestoy, and Iain S. Duff. See ../README.txt for License. */
8 /* email: davis at cise.ufl.edu CISE Department, Univ. of Florida. */
9 /* web: http://www.cise.ufl.edu/research/sparse/amd */
10 /* ------------------------------------------------------------------------- */
11 
12 /* AMD_2: performs the AMD ordering on a symmetric sparse matrix A, followed
13  * by a postordering (via depth-first search) of the assembly tree using the
14  * AMD_postorder routine.
15  */
16 
17 #include "amesos_amd_internal.h"
18 
19 /* ========================================================================= */
20 /* === clear_flag ========================================================== */
21 /* ========================================================================= */
22 
23 static Int amesos_clear_flag (Int wflg, Int wbig, Int W [ ], Int n)
24 {
25  Int x ;
26  if (wflg < 2 || wflg >= wbig)
27  {
28  for (x = 0 ; x < n ; x++)
29  {
30  if (W [x] != 0) W [x] = 1 ;
31  }
32  wflg = 2 ;
33  }
34  /* at this point, W [0..n-1] < wflg holds */
35  return (wflg) ;
36 }
37 
38 
39 /* ========================================================================= */
40 /* === AMD_2 =============================================================== */
41 /* ========================================================================= */
42 
43 GLOBAL void AMD_2
44 (
45  Int n, /* A is n-by-n, where n > 0 */
46  Int Pe [ ], /* Pe [0..n-1]: index in Iw of row i on input */
47  Int Iw [ ], /* workspace of size iwlen. Iw [0..pfree-1]
48  * holds the matrix on input */
49  Int Len [ ], /* Len [0..n-1]: length for row/column i on input */
50  Int iwlen, /* length of Iw. iwlen >= pfree + n */
51  Int pfree, /* Iw [pfree ... iwlen-1] is empty on input */
52 
53  /* 7 size-n workspaces, not defined on input: */
54  Int Nv [ ], /* the size of each supernode on output */
55  Int Next [ ], /* the output inverse permutation */
56  Int Last [ ], /* the output permutation */
57  Int Head [ ],
58  Int Elen [ ], /* the size columns of L for each supernode */
59  Int Degree [ ],
60  Int W [ ],
61 
62  /* control parameters and output statistics */
63  double Control [ ], /* array of size AMD_CONTROL */
64  double Info [ ] /* array of size AMD_INFO */
65 )
66 {
67 
68 /*
69  * Given a representation of the nonzero pattern of a symmetric matrix, A,
70  * (excluding the diagonal) perform an approximate minimum (UMFPACK/MA38-style)
71  * degree ordering to compute a pivot order such that the introduction of
72  * nonzeros (fill-in) in the Cholesky factors A = LL' is kept low. At each
73  * step, the pivot selected is the one with the minimum UMFAPACK/MA38-style
74  * upper-bound on the external degree. This routine can optionally perform
75  * aggresive absorption (as done by MC47B in the Harwell Subroutine
76  * Library).
77  *
78  * The approximate degree algorithm implemented here is the symmetric analog of
79  * the degree update algorithm in MA38 and UMFPACK (the Unsymmetric-pattern
80  * MultiFrontal PACKage, both by Davis and Duff). The routine is based on the
81  * MA27 minimum degree ordering algorithm by Iain Duff and John Reid.
82  *
83  * This routine is a translation of the original AMDBAR and MC47B routines,
84  * in Fortran, with the following modifications:
85  *
86  * (1) dense rows/columns are removed prior to ordering the matrix, and placed
87  * last in the output order. The presence of a dense row/column can
88  * increase the ordering time by up to O(n^2), unless they are removed
89  * prior to ordering.
90  *
91  * (2) the minimum degree ordering is followed by a postordering (depth-first
92  * search) of the assembly tree. Note that mass elimination (discussed
93  * below) combined with the approximate degree update can lead to the mass
94  * elimination of nodes with lower exact degree than the current pivot
95  * element. No additional fill-in is caused in the representation of the
96  * Schur complement. The mass-eliminated nodes merge with the current
97  * pivot element. They are ordered prior to the current pivot element.
98  * Because they can have lower exact degree than the current element, the
99  * merger of two or more of these nodes in the current pivot element can
100  * lead to a single element that is not a "fundamental supernode". The
101  * diagonal block can have zeros in it. Thus, the assembly tree used here
102  * is not guaranteed to be the precise supernodal elemination tree (with
103  * "funadmental" supernodes), and the postordering performed by this
104  * routine is not guaranteed to be a precise postordering of the
105  * elimination tree.
106  *
107  * (3) input parameters are added, to control aggressive absorption and the
108  * detection of "dense" rows/columns of A.
109  *
110  * (4) additional statistical information is returned, such as the number of
111  * nonzeros in L, and the flop counts for subsequent LDL' and LU
112  * factorizations. These are slight upper bounds, because of the mass
113  * elimination issue discussed above.
114  *
115  * (5) additional routines are added to interface this routine to MATLAB
116  * to provide a simple C-callable user-interface, to check inputs for
117  * errors, compute the symmetry of the pattern of A and the number of
118  * nonzeros in each row/column of A+A', to compute the pattern of A+A',
119  * to perform the assembly tree postordering, and to provide debugging
120  * ouput. Many of these functions are also provided by the Fortran
121  * Harwell Subroutine Library routine MC47A.
122  *
123  * (6) both int and UF_long versions are provided. In the descriptions below
124  * and integer is and int or UF_long depending on which version is
125  * being used.
126 
127  **********************************************************************
128  ***** CAUTION: ARGUMENTS ARE NOT CHECKED FOR ERRORS ON INPUT. ******
129  **********************************************************************
130  ** If you want error checking, a more versatile input format, and a **
131  ** simpler user interface, use amd_order or amd_l_order instead. **
132  ** This routine is not meant to be user-callable. **
133  **********************************************************************
134 
135  * ----------------------------------------------------------------------------
136  * References:
137  * ----------------------------------------------------------------------------
138  *
139  * [1] Timothy A. Davis and Iain Duff, "An unsymmetric-pattern multifrontal
140  * method for sparse LU factorization", SIAM J. Matrix Analysis and
141  * Applications, vol. 18, no. 1, pp. 140-158. Discusses UMFPACK / MA38,
142  * which first introduced the approximate minimum degree used by this
143  * routine.
144  *
145  * [2] Patrick Amestoy, Timothy A. Davis, and Iain S. Duff, "An approximate
146  * minimum degree ordering algorithm," SIAM J. Matrix Analysis and
147  * Applications, vol. 17, no. 4, pp. 886-905, 1996. Discusses AMDBAR and
148  * MC47B, which are the Fortran versions of this routine.
149  *
150  * [3] Alan George and Joseph Liu, "The evolution of the minimum degree
151  * ordering algorithm," SIAM Review, vol. 31, no. 1, pp. 1-19, 1989.
152  * We list below the features mentioned in that paper that this code
153  * includes:
154  *
155  * mass elimination:
156  * Yes. MA27 relied on supervariable detection for mass elimination.
157  *
158  * indistinguishable nodes:
159  * Yes (we call these "supervariables"). This was also in the MA27
160  * code - although we modified the method of detecting them (the
161  * previous hash was the true degree, which we no longer keep track
162  * of). A supervariable is a set of rows with identical nonzero
163  * pattern. All variables in a supervariable are eliminated together.
164  * Each supervariable has as its numerical name that of one of its
165  * variables (its principal variable).
166  *
167  * quotient graph representation:
168  * Yes. We use the term "element" for the cliques formed during
169  * elimination. This was also in the MA27 code. The algorithm can
170  * operate in place, but it will work more efficiently if given some
171  * "elbow room."
172  *
173  * element absorption:
174  * Yes. This was also in the MA27 code.
175  *
176  * external degree:
177  * Yes. The MA27 code was based on the true degree.
178  *
179  * incomplete degree update and multiple elimination:
180  * No. This was not in MA27, either. Our method of degree update
181  * within MC47B is element-based, not variable-based. It is thus
182  * not well-suited for use with incomplete degree update or multiple
183  * elimination.
184  *
185  * Authors, and Copyright (C) 2004 by:
186  * Timothy A. Davis, Patrick Amestoy, Iain S. Duff, John K. Reid.
187  *
188  * Acknowledgements: This work (and the UMFPACK package) was supported by the
189  * National Science Foundation (ASC-9111263, DMS-9223088, and CCR-0203270).
190  * The UMFPACK/MA38 approximate degree update algorithm, the unsymmetric analog
191  * which forms the basis of AMD, was developed while Tim Davis was supported by
192  * CERFACS (Toulouse, France) in a post-doctoral position. This C version, and
193  * the etree postorder, were written while Tim Davis was on sabbatical at
194  * Stanford University and Lawrence Berkeley National Laboratory.
195 
196  * ----------------------------------------------------------------------------
197  * INPUT ARGUMENTS (unaltered):
198  * ----------------------------------------------------------------------------
199 
200  * n: The matrix order. Restriction: n >= 1.
201  *
202  * iwlen: The size of the Iw array. On input, the matrix is stored in
203  * Iw [0..pfree-1]. However, Iw [0..iwlen-1] should be slightly larger
204  * than what is required to hold the matrix, at least iwlen >= pfree + n.
205  * Otherwise, excessive compressions will take place. The recommended
206  * value of iwlen is 1.2 * pfree + n, which is the value used in the
207  * user-callable interface to this routine (amd_order.c). The algorithm
208  * will not run at all if iwlen < pfree. Restriction: iwlen >= pfree + n.
209  * Note that this is slightly more restrictive than the actual minimum
210  * (iwlen >= pfree), but AMD_2 will be very slow with no elbow room.
211  * Thus, this routine enforces a bare minimum elbow room of size n.
212  *
213  * pfree: On input the tail end of the array, Iw [pfree..iwlen-1], is empty,
214  * and the matrix is stored in Iw [0..pfree-1]. During execution,
215  * additional data is placed in Iw, and pfree is modified so that
216  * Iw [pfree..iwlen-1] is always the unused part of Iw.
217  *
218  * Control: A double array of size AMD_CONTROL containing input parameters
219  * that affect how the ordering is computed. If NULL, then default
220  * settings are used.
221  *
222  * Control [AMD_DENSE] is used to determine whether or not a given input
223  * row is "dense". A row is "dense" if the number of entries in the row
224  * exceeds Control [AMD_DENSE] times sqrt (n), except that rows with 16 or
225  * fewer entries are never considered "dense". To turn off the detection
226  * of dense rows, set Control [AMD_DENSE] to a negative number, or to a
227  * number larger than sqrt (n). The default value of Control [AMD_DENSE]
228  * is AMD_DEFAULT_DENSE, which is defined in amd.h as 10.
229  *
230  * Control [AMD_AGGRESSIVE] is used to determine whether or not aggressive
231  * absorption is to be performed. If nonzero, then aggressive absorption
232  * is performed (this is the default).
233 
234  * ----------------------------------------------------------------------------
235  * INPUT/OUPUT ARGUMENTS:
236  * ----------------------------------------------------------------------------
237  *
238  * Pe: An integer array of size n. On input, Pe [i] is the index in Iw of
239  * the start of row i. Pe [i] is ignored if row i has no off-diagonal
240  * entries. Thus Pe [i] must be in the range 0 to pfree-1 for non-empty
241  * rows.
242  *
243  * During execution, it is used for both supervariables and elements:
244  *
245  * Principal supervariable i: index into Iw of the description of
246  * supervariable i. A supervariable represents one or more rows of
247  * the matrix with identical nonzero pattern. In this case,
248  * Pe [i] >= 0.
249  *
250  * Non-principal supervariable i: if i has been absorbed into another
251  * supervariable j, then Pe [i] = FLIP (j), where FLIP (j) is defined
252  * as (-(j)-2). Row j has the same pattern as row i. Note that j
253  * might later be absorbed into another supervariable j2, in which
254  * case Pe [i] is still FLIP (j), and Pe [j] = FLIP (j2) which is
255  * < EMPTY, where EMPTY is defined as (-1) in amd_internal.h.
256  *
257  * Unabsorbed element e: the index into Iw of the description of element
258  * e, if e has not yet been absorbed by a subsequent element. Element
259  * e is created when the supervariable of the same name is selected as
260  * the pivot. In this case, Pe [i] >= 0.
261  *
262  * Absorbed element e: if element e is absorbed into element e2, then
263  * Pe [e] = FLIP (e2). This occurs when the pattern of e (which we
264  * refer to as Le) is found to be a subset of the pattern of e2 (that
265  * is, Le2). In this case, Pe [i] < EMPTY. If element e is "null"
266  * (it has no nonzeros outside its pivot block), then Pe [e] = EMPTY,
267  * and e is the root of an assembly subtree (or the whole tree if
268  * there is just one such root).
269  *
270  * Dense variable i: if i is "dense", then Pe [i] = EMPTY.
271  *
272  * On output, Pe holds the assembly tree/forest, which implicitly
273  * represents a pivot order with identical fill-in as the actual order
274  * (via a depth-first search of the tree), as follows. If Nv [i] > 0,
275  * then i represents a node in the assembly tree, and the parent of i is
276  * Pe [i], or EMPTY if i is a root. If Nv [i] = 0, then (i, Pe [i])
277  * represents an edge in a subtree, the root of which is a node in the
278  * assembly tree. Note that i refers to a row/column in the original
279  * matrix, not the permuted matrix.
280  *
281  * Info: A double array of size AMD_INFO. If present, (that is, not NULL),
282  * then statistics about the ordering are returned in the Info array.
283  * See amd.h for a description.
284 
285  * ----------------------------------------------------------------------------
286  * INPUT/MODIFIED (undefined on output):
287  * ----------------------------------------------------------------------------
288  *
289  * Len: An integer array of size n. On input, Len [i] holds the number of
290  * entries in row i of the matrix, excluding the diagonal. The contents
291  * of Len are undefined on output.
292  *
293  * Iw: An integer array of size iwlen. On input, Iw [0..pfree-1] holds the
294  * description of each row i in the matrix. The matrix must be symmetric,
295  * and both upper and lower triangular parts must be present. The
296  * diagonal must not be present. Row i is held as follows:
297  *
298  * Len [i]: the length of the row i data structure in the Iw array.
299  * Iw [Pe [i] ... Pe [i] + Len [i] - 1]:
300  * the list of column indices for nonzeros in row i (simple
301  * supervariables), excluding the diagonal. All supervariables
302  * start with one row/column each (supervariable i is just row i).
303  * If Len [i] is zero on input, then Pe [i] is ignored on input.
304  *
305  * Note that the rows need not be in any particular order, and there
306  * may be empty space between the rows.
307  *
308  * During execution, the supervariable i experiences fill-in. This is
309  * represented by placing in i a list of the elements that cause fill-in
310  * in supervariable i:
311  *
312  * Len [i]: the length of supervariable i in the Iw array.
313  * Iw [Pe [i] ... Pe [i] + Elen [i] - 1]:
314  * the list of elements that contain i. This list is kept short
315  * by removing absorbed elements.
316  * Iw [Pe [i] + Elen [i] ... Pe [i] + Len [i] - 1]:
317  * the list of supervariables in i. This list is kept short by
318  * removing nonprincipal variables, and any entry j that is also
319  * contained in at least one of the elements (j in Le) in the list
320  * for i (e in row i).
321  *
322  * When supervariable i is selected as pivot, we create an element e of
323  * the same name (e=i):
324  *
325  * Len [e]: the length of element e in the Iw array.
326  * Iw [Pe [e] ... Pe [e] + Len [e] - 1]:
327  * the list of supervariables in element e.
328  *
329  * An element represents the fill-in that occurs when supervariable i is
330  * selected as pivot (which represents the selection of row i and all
331  * non-principal variables whose principal variable is i). We use the
332  * term Le to denote the set of all supervariables in element e. Absorbed
333  * supervariables and elements are pruned from these lists when
334  * computationally convenient.
335  *
336  * CAUTION: THE INPUT MATRIX IS OVERWRITTEN DURING COMPUTATION.
337  * The contents of Iw are undefined on output.
338 
339  * ----------------------------------------------------------------------------
340  * OUTPUT (need not be set on input):
341  * ----------------------------------------------------------------------------
342  *
343  * Nv: An integer array of size n. During execution, ABS (Nv [i]) is equal to
344  * the number of rows that are represented by the principal supervariable
345  * i. If i is a nonprincipal or dense variable, then Nv [i] = 0.
346  * Initially, Nv [i] = 1 for all i. Nv [i] < 0 signifies that i is a
347  * principal variable in the pattern Lme of the current pivot element me.
348  * After element me is constructed, Nv [i] is set back to a positive
349  * value.
350  *
351  * On output, Nv [i] holds the number of pivots represented by super
352  * row/column i of the original matrix, or Nv [i] = 0 for non-principal
353  * rows/columns. Note that i refers to a row/column in the original
354  * matrix, not the permuted matrix.
355  *
356  * Elen: An integer array of size n. See the description of Iw above. At the
357  * start of execution, Elen [i] is set to zero for all rows i. During
358  * execution, Elen [i] is the number of elements in the list for
359  * supervariable i. When e becomes an element, Elen [e] = FLIP (esize) is
360  * set, where esize is the size of the element (the number of pivots, plus
361  * the number of nonpivotal entries). Thus Elen [e] < EMPTY.
362  * Elen (i) = EMPTY set when variable i becomes nonprincipal.
363  *
364  * For variables, Elen (i) >= EMPTY holds until just before the
365  * postordering and permutation vectors are computed. For elements,
366  * Elen [e] < EMPTY holds.
367  *
368  * On output, Elen [i] is the degree of the row/column in the Cholesky
369  * factorization of the permuted matrix, corresponding to the original row
370  * i, if i is a super row/column. It is equal to EMPTY if i is
371  * non-principal. Note that i refers to a row/column in the original
372  * matrix, not the permuted matrix.
373  *
374  * Note that the contents of Elen on output differ from the Fortran
375  * version (Elen holds the inverse permutation in the Fortran version,
376  * which is instead returned in the Next array in this C version,
377  * described below).
378  *
379  * Last: In a degree list, Last [i] is the supervariable preceding i, or EMPTY
380  * if i is the head of the list. In a hash bucket, Last [i] is the hash
381  * key for i.
382  *
383  * Last [Head [hash]] is also used as the head of a hash bucket if
384  * Head [hash] contains a degree list (see the description of Head,
385  * below).
386  *
387  * On output, Last [0..n-1] holds the permutation. That is, if
388  * i = Last [k], then row i is the kth pivot row (where k ranges from 0 to
389  * n-1). Row Last [k] of A is the kth row in the permuted matrix, PAP'.
390  *
391  * Next: Next [i] is the supervariable following i in a link list, or EMPTY if
392  * i is the last in the list. Used for two kinds of lists: degree lists
393  * and hash buckets (a supervariable can be in only one kind of list at a
394  * time).
395  *
396  * On output Next [0..n-1] holds the inverse permutation. That is, if
397  * k = Next [i], then row i is the kth pivot row. Row i of A appears as
398  * the (Next[i])-th row in the permuted matrix, PAP'.
399  *
400  * Note that the contents of Next on output differ from the Fortran
401  * version (Next is undefined on output in the Fortran version).
402 
403  * ----------------------------------------------------------------------------
404  * LOCAL WORKSPACE (not input or output - used only during execution):
405  * ----------------------------------------------------------------------------
406  *
407  * Degree: An integer array of size n. If i is a supervariable, then
408  * Degree [i] holds the current approximation of the external degree of
409  * row i (an upper bound). The external degree is the number of nonzeros
410  * in row i, minus ABS (Nv [i]), the diagonal part. The bound is equal to
411  * the exact external degree if Elen [i] is less than or equal to two.
412  *
413  * We also use the term "external degree" for elements e to refer to
414  * |Le \ Lme|. If e is an element, then Degree [e] is |Le|, which is the
415  * degree of the off-diagonal part of the element e (not including the
416  * diagonal part).
417  *
418  * Head: An integer array of size n. Head is used for degree lists.
419  * Head [deg] is the first supervariable in a degree list. All
420  * supervariables i in a degree list Head [deg] have the same approximate
421  * degree, namely, deg = Degree [i]. If the list Head [deg] is empty then
422  * Head [deg] = EMPTY.
423  *
424  * During supervariable detection Head [hash] also serves as a pointer to
425  * a hash bucket. If Head [hash] >= 0, there is a degree list of degree
426  * hash. The hash bucket head pointer is Last [Head [hash]]. If
427  * Head [hash] = EMPTY, then the degree list and hash bucket are both
428  * empty. If Head [hash] < EMPTY, then the degree list is empty, and
429  * FLIP (Head [hash]) is the head of the hash bucket. After supervariable
430  * detection is complete, all hash buckets are empty, and the
431  * (Last [Head [hash]] = EMPTY) condition is restored for the non-empty
432  * degree lists.
433  *
434  * W: An integer array of size n. The flag array W determines the status of
435  * elements and variables, and the external degree of elements.
436  *
437  * for elements:
438  * if W [e] = 0, then the element e is absorbed.
439  * if W [e] >= wflg, then W [e] - wflg is the size of the set
440  * |Le \ Lme|, in terms of nonzeros (the sum of ABS (Nv [i]) for
441  * each principal variable i that is both in the pattern of
442  * element e and NOT in the pattern of the current pivot element,
443  * me).
444  * if wflg > W [e] > 0, then e is not absorbed and has not yet been
445  * seen in the scan of the element lists in the computation of
446  * |Le\Lme| in Scan 1 below.
447  *
448  * for variables:
449  * during supervariable detection, if W [j] != wflg then j is
450  * not in the pattern of variable i.
451  *
452  * The W array is initialized by setting W [i] = 1 for all i, and by
453  * setting wflg = 2. It is reinitialized if wflg becomes too large (to
454  * ensure that wflg+n does not cause integer overflow).
455 
456  * ----------------------------------------------------------------------------
457  * LOCAL INTEGERS:
458  * ----------------------------------------------------------------------------
459  */
460 
461  Int deg, degme, dext, lemax, e, elenme, eln, i, ilast, inext, j,
462  jlast, jnext, k, knt1, knt2, knt3, lenj, ln, me, mindeg, nel, nleft,
463  nvi, nvj, nvpiv, slenme, wbig, we, wflg, wnvi, ok, ndense, ncmpa,
464  dense, aggressive ;
465 
466  unsigned Int hash ; /* unsigned, so that hash % n is well defined.*/
467 
468 /*
469  * deg: the degree of a variable or element
470  * degme: size, |Lme|, of the current element, me (= Degree [me])
471  * dext: external degree, |Le \ Lme|, of some element e
472  * lemax: largest |Le| seen so far (called dmax in Fortran version)
473  * e: an element
474  * elenme: the length, Elen [me], of element list of pivotal variable
475  * eln: the length, Elen [...], of an element list
476  * hash: the computed value of the hash function
477  * i: a supervariable
478  * ilast: the entry in a link list preceding i
479  * inext: the entry in a link list following i
480  * j: a supervariable
481  * jlast: the entry in a link list preceding j
482  * jnext: the entry in a link list, or path, following j
483  * k: the pivot order of an element or variable
484  * knt1: loop counter used during element construction
485  * knt2: loop counter used during element construction
486  * knt3: loop counter used during compression
487  * lenj: Len [j]
488  * ln: length of a supervariable list
489  * me: current supervariable being eliminated, and the current
490  * element created by eliminating that supervariable
491  * mindeg: current minimum degree
492  * nel: number of pivots selected so far
493  * nleft: n - nel, the number of nonpivotal rows/columns remaining
494  * nvi: the number of variables in a supervariable i (= Nv [i])
495  * nvj: the number of variables in a supervariable j (= Nv [j])
496  * nvpiv: number of pivots in current element
497  * slenme: number of variables in variable list of pivotal variable
498  * wbig: = INT_MAX - n for the int version, UF_long_max - n for the
499  * UF_long version. wflg is not allowed to be >= wbig.
500  * we: W [e]
501  * wflg: used for flagging the W array. See description of Iw.
502  * wnvi: wflg - Nv [i]
503  * x: either a supervariable or an element
504  *
505  * ok: true if supervariable j can be absorbed into i
506  * ndense: number of "dense" rows/columns
507  * dense: rows/columns with initial degree > dense are considered "dense"
508  * aggressive: true if aggressive absorption is being performed
509  * ncmpa: number of garbage collections
510 
511  * ----------------------------------------------------------------------------
512  * LOCAL DOUBLES, used for statistical output only (except for alpha):
513  * ----------------------------------------------------------------------------
514  */
515 
516  double f, r, ndiv, s, nms_lu, nms_ldl, dmax, alpha, lnz, lnzme ;
517 
518 /*
519  * f: nvpiv
520  * r: degme + nvpiv
521  * ndiv: number of divisions for LU or LDL' factorizations
522  * s: number of multiply-subtract pairs for LU factorization, for the
523  * current element me
524  * nms_lu number of multiply-subtract pairs for LU factorization
525  * nms_ldl number of multiply-subtract pairs for LDL' factorization
526  * dmax: the largest number of entries in any column of L, including the
527  * diagonal
528  * alpha: "dense" degree ratio
529  * lnz: the number of nonzeros in L (excluding the diagonal)
530  * lnzme: the number of nonzeros in L (excl. the diagonal) for the
531  * current element me
532 
533  * ----------------------------------------------------------------------------
534  * LOCAL "POINTERS" (indices into the Iw array)
535  * ----------------------------------------------------------------------------
536 */
537 
538  Int p, p1, p2, p3, p4, pdst, pend, pj, pme, pme1, pme2, pn, psrc ;
539 
540 /*
541  * Any parameter (Pe [...] or pfree) or local variable starting with "p" (for
542  * Pointer) is an index into Iw, and all indices into Iw use variables starting
543  * with "p." The only exception to this rule is the iwlen input argument.
544  *
545  * p: pointer into lots of things
546  * p1: Pe [i] for some variable i (start of element list)
547  * p2: Pe [i] + Elen [i] - 1 for some variable i
548  * p3: index of first supervariable in clean list
549  * p4:
550  * pdst: destination pointer, for compression
551  * pend: end of memory to compress
552  * pj: pointer into an element or variable
553  * pme: pointer into the current element (pme1...pme2)
554  * pme1: the current element, me, is stored in Iw [pme1...pme2]
555  * pme2: the end of the current element
556  * pn: pointer into a "clean" variable, also used to compress
557  * psrc: source pointer, for compression
558 */
559 
560 /* ========================================================================= */
561 /* INITIALIZATIONS */
562 /* ========================================================================= */
563 
564  /* Note that this restriction on iwlen is slightly more restrictive than
565  * what is actually required in AMD_2. AMD_2 can operate with no elbow
566  * room at all, but it will be slow. For better performance, at least
567  * size-n elbow room is enforced. */
568  ASSERT (iwlen >= pfree + n) ;
569  ASSERT (n > 0) ;
570 
571  /* initialize output statistics */
572  lnz = 0 ;
573  ndiv = 0 ;
574  nms_lu = 0 ;
575  nms_ldl = 0 ;
576  dmax = 1 ;
577  me = EMPTY ;
578 
579  mindeg = 0 ;
580  ncmpa = 0 ;
581  nel = 0 ;
582  lemax = 0 ;
583 
584  /* get control parameters */
585  if (Control != (double *) NULL)
586  {
587  alpha = Control [AMD_DENSE] ;
588  aggressive = (Control [AMD_AGGRESSIVE] != 0) ;
589  }
590  else
591  {
592  alpha = AMD_DEFAULT_DENSE ;
593  aggressive = AMD_DEFAULT_AGGRESSIVE ;
594  }
595  /* Note: if alpha is NaN, this is undefined: */
596  if (alpha < 0)
597  {
598  /* only remove completely dense rows/columns */
599  dense = n-2 ;
600  }
601  else
602  {
603  dense = (int) ( alpha * sqrt ((double) n) ) ;
604  }
605  dense = MAX (16, dense) ;
606  dense = MIN (n, dense) ;
607  AMD_DEBUG1 (("\n\nAMD (debug), alpha %g, aggr. "ID"\n",
608  alpha, aggressive)) ;
609 
610  for (i = 0 ; i < n ; i++)
611  {
612  Last [i] = EMPTY ;
613  Head [i] = EMPTY ;
614  Next [i] = EMPTY ;
615  /* if separate Hhead array is used for hash buckets: *
616  Hhead [i] = EMPTY ;
617  */
618  Nv [i] = 1 ;
619  W [i] = 1 ;
620  Elen [i] = 0 ;
621  Degree [i] = Len [i] ;
622  }
623 
624 #ifndef NDEBUG
625  AMD_DEBUG1 (("\n======Nel "ID" initial\n", nel)) ;
626  AMD_dump (n, Pe, Iw, Len, iwlen, pfree, Nv, Next, Last,
627  Head, Elen, Degree, W, -1) ;
628 #endif
629 
630  /* initialize wflg */
631  wbig = Int_MAX - n ;
632  wflg = amesos_clear_flag (0, wbig, W, n) ;
633 
634  /* --------------------------------------------------------------------- */
635  /* initialize degree lists and eliminate dense and empty rows */
636  /* --------------------------------------------------------------------- */
637 
638  ndense = 0 ;
639 
640  for (i = 0 ; i < n ; i++)
641  {
642  deg = Degree [i] ;
643  ASSERT (deg >= 0 && deg < n) ;
644  if (deg == 0)
645  {
646 
647  /* -------------------------------------------------------------
648  * we have a variable that can be eliminated at once because
649  * there is no off-diagonal non-zero in its row. Note that
650  * Nv [i] = 1 for an empty variable i. It is treated just
651  * the same as an eliminated element i.
652  * ------------------------------------------------------------- */
653 
654  Elen [i] = FLIP (1) ;
655  nel++ ;
656  Pe [i] = EMPTY ;
657  W [i] = 0 ;
658 
659  }
660  else if (deg > dense)
661  {
662 
663  /* -------------------------------------------------------------
664  * Dense variables are not treated as elements, but as unordered,
665  * non-principal variables that have no parent. They do not take
666  * part in the postorder, since Nv [i] = 0. Note that the Fortran
667  * version does not have this option.
668  * ------------------------------------------------------------- */
669 
670  AMD_DEBUG1 (("Dense node "ID" degree "ID"\n", i, deg)) ;
671  ndense++ ;
672  Nv [i] = 0 ; /* do not postorder this node */
673  Elen [i] = EMPTY ;
674  nel++ ;
675  Pe [i] = EMPTY ;
676 
677  }
678  else
679  {
680 
681  /* -------------------------------------------------------------
682  * place i in the degree list corresponding to its degree
683  * ------------------------------------------------------------- */
684 
685  inext = Head [deg] ;
686  ASSERT (inext >= EMPTY && inext < n) ;
687  if (inext != EMPTY) Last [inext] = i ;
688  Next [i] = inext ;
689  Head [deg] = i ;
690 
691  }
692  }
693 
694 /* ========================================================================= */
695 /* WHILE (selecting pivots) DO */
696 /* ========================================================================= */
697 
698  while (nel < n)
699  {
700 
701 #ifndef NDEBUG
702  AMD_DEBUG1 (("\n======Nel "ID"\n", nel)) ;
703  if (AMD_debug >= 2)
704  {
705  AMD_dump (n, Pe, Iw, Len, iwlen, pfree, Nv, Next,
706  Last, Head, Elen, Degree, W, nel) ;
707  }
708 #endif
709 
710 /* ========================================================================= */
711 /* GET PIVOT OF MINIMUM DEGREE */
712 /* ========================================================================= */
713 
714  /* ----------------------------------------------------------------- */
715  /* find next supervariable for elimination */
716  /* ----------------------------------------------------------------- */
717 
718  ASSERT (mindeg >= 0 && mindeg < n) ;
719  for (deg = mindeg ; deg < n ; deg++)
720  {
721  me = Head [deg] ;
722  if (me != EMPTY) break ;
723  }
724  mindeg = deg ;
725  ASSERT (me >= 0 && me < n) ;
726  AMD_DEBUG1 (("=================me: "ID"\n", me)) ;
727 
728  /* ----------------------------------------------------------------- */
729  /* remove chosen variable from link list */
730  /* ----------------------------------------------------------------- */
731 
732  inext = Next [me] ;
733  ASSERT (inext >= EMPTY && inext < n) ;
734  if (inext != EMPTY) Last [inext] = EMPTY ;
735  Head [deg] = inext ;
736 
737  /* ----------------------------------------------------------------- */
738  /* me represents the elimination of pivots nel to nel+Nv[me]-1. */
739  /* place me itself as the first in this set. */
740  /* ----------------------------------------------------------------- */
741 
742  elenme = Elen [me] ;
743  nvpiv = Nv [me] ;
744  ASSERT (nvpiv > 0) ;
745  nel += nvpiv ;
746 
747 /* ========================================================================= */
748 /* CONSTRUCT NEW ELEMENT */
749 /* ========================================================================= */
750 
751  /* -----------------------------------------------------------------
752  * At this point, me is the pivotal supervariable. It will be
753  * converted into the current element. Scan list of the pivotal
754  * supervariable, me, setting tree pointers and constructing new list
755  * of supervariables for the new element, me. p is a pointer to the
756  * current position in the old list.
757  * ----------------------------------------------------------------- */
758 
759  /* flag the variable "me" as being in Lme by negating Nv [me] */
760  Nv [me] = -nvpiv ;
761  degme = 0 ;
762  ASSERT (Pe [me] >= 0 && Pe [me] < iwlen) ;
763 
764  if (elenme == 0)
765  {
766 
767  /* ------------------------------------------------------------- */
768  /* construct the new element in place */
769  /* ------------------------------------------------------------- */
770 
771  pme1 = Pe [me] ;
772  pme2 = pme1 - 1 ;
773 
774  for (p = pme1 ; p <= pme1 + Len [me] - 1 ; p++)
775  {
776  i = Iw [p] ;
777  ASSERT (i >= 0 && i < n && Nv [i] >= 0) ;
778  nvi = Nv [i] ;
779  if (nvi > 0)
780  {
781 
782  /* ----------------------------------------------------- */
783  /* i is a principal variable not yet placed in Lme. */
784  /* store i in new list */
785  /* ----------------------------------------------------- */
786 
787  /* flag i as being in Lme by negating Nv [i] */
788  degme += nvi ;
789  Nv [i] = -nvi ;
790  Iw [++pme2] = i ;
791 
792  /* ----------------------------------------------------- */
793  /* remove variable i from degree list. */
794  /* ----------------------------------------------------- */
795 
796  ilast = Last [i] ;
797  inext = Next [i] ;
798  ASSERT (ilast >= EMPTY && ilast < n) ;
799  ASSERT (inext >= EMPTY && inext < n) ;
800  if (inext != EMPTY) Last [inext] = ilast ;
801  if (ilast != EMPTY)
802  {
803  Next [ilast] = inext ;
804  }
805  else
806  {
807  /* i is at the head of the degree list */
808  ASSERT (Degree [i] >= 0 && Degree [i] < n) ;
809  Head [Degree [i]] = inext ;
810  }
811  }
812  }
813  }
814  else
815  {
816 
817  /* ------------------------------------------------------------- */
818  /* construct the new element in empty space, Iw [pfree ...] */
819  /* ------------------------------------------------------------- */
820 
821  p = Pe [me] ;
822  pme1 = pfree ;
823  slenme = Len [me] - elenme ;
824 
825  for (knt1 = 1 ; knt1 <= elenme + 1 ; knt1++)
826  {
827 
828  if (knt1 > elenme)
829  {
830  /* search the supervariables in me. */
831  e = me ;
832  pj = p ;
833  ln = slenme ;
834  AMD_DEBUG2 (("Search sv: "ID" "ID" "ID"\n", me,pj,ln)) ;
835  }
836  else
837  {
838  /* search the elements in me. */
839  e = Iw [p++] ;
840  ASSERT (e >= 0 && e < n) ;
841  pj = Pe [e] ;
842  ln = Len [e] ;
843  AMD_DEBUG2 (("Search element e "ID" in me "ID"\n", e,me)) ;
844  ASSERT (Elen [e] < EMPTY && W [e] > 0 && pj >= 0) ;
845  }
846  ASSERT (ln >= 0 && (ln == 0 || (pj >= 0 && pj < iwlen))) ;
847 
848  /* ---------------------------------------------------------
849  * search for different supervariables and add them to the
850  * new list, compressing when necessary. this loop is
851  * executed once for each element in the list and once for
852  * all the supervariables in the list.
853  * --------------------------------------------------------- */
854 
855  for (knt2 = 1 ; knt2 <= ln ; knt2++)
856  {
857  i = Iw [pj++] ;
858  ASSERT (i >= 0 && i < n && (i == me || Elen [i] >= EMPTY));
859  nvi = Nv [i] ;
860  AMD_DEBUG2 ((": "ID" "ID" "ID" "ID"\n",
861  i, Elen [i], Nv [i], wflg)) ;
862 
863  if (nvi > 0)
864  {
865 
866  /* ------------------------------------------------- */
867  /* compress Iw, if necessary */
868  /* ------------------------------------------------- */
869 
870  if (pfree >= iwlen)
871  {
872 
873  AMD_DEBUG1 (("GARBAGE COLLECTION\n")) ;
874 
875  /* prepare for compressing Iw by adjusting pointers
876  * and lengths so that the lists being searched in
877  * the inner and outer loops contain only the
878  * remaining entries. */
879 
880  Pe [me] = p ;
881  Len [me] -= knt1 ;
882  /* check if nothing left of supervariable me */
883  if (Len [me] == 0) Pe [me] = EMPTY ;
884  Pe [e] = pj ;
885  Len [e] = ln - knt2 ;
886  /* nothing left of element e */
887  if (Len [e] == 0) Pe [e] = EMPTY ;
888 
889  ncmpa++ ; /* one more garbage collection */
890 
891  /* store first entry of each object in Pe */
892  /* FLIP the first entry in each object */
893  for (j = 0 ; j < n ; j++)
894  {
895  pn = Pe [j] ;
896  if (pn >= 0)
897  {
898  ASSERT (pn >= 0 && pn < iwlen) ;
899  Pe [j] = Iw [pn] ;
900  Iw [pn] = FLIP (j) ;
901  }
902  }
903 
904  /* psrc/pdst point to source/destination */
905  psrc = 0 ;
906  pdst = 0 ;
907  pend = pme1 - 1 ;
908 
909  while (psrc <= pend)
910  {
911  /* search for next FLIP'd entry */
912  j = FLIP (Iw [psrc++]) ;
913  if (j >= 0)
914  {
915  AMD_DEBUG2 (("Got object j: "ID"\n", j)) ;
916  Iw [pdst] = Pe [j] ;
917  Pe [j] = pdst++ ;
918  lenj = Len [j] ;
919  /* copy from source to destination */
920  for (knt3 = 0 ; knt3 <= lenj - 2 ; knt3++)
921  {
922  Iw [pdst++] = Iw [psrc++] ;
923  }
924  }
925  }
926 
927  /* move the new partially-constructed element */
928  p1 = pdst ;
929  for (psrc = pme1 ; psrc <= pfree-1 ; psrc++)
930  {
931  Iw [pdst++] = Iw [psrc] ;
932  }
933  pme1 = p1 ;
934  pfree = pdst ;
935  pj = Pe [e] ;
936  p = Pe [me] ;
937 
938  }
939 
940  /* ------------------------------------------------- */
941  /* i is a principal variable not yet placed in Lme */
942  /* store i in new list */
943  /* ------------------------------------------------- */
944 
945  /* flag i as being in Lme by negating Nv [i] */
946  degme += nvi ;
947  Nv [i] = -nvi ;
948  Iw [pfree++] = i ;
949  AMD_DEBUG2 ((" s: "ID" nv "ID"\n", i, Nv [i]));
950 
951  /* ------------------------------------------------- */
952  /* remove variable i from degree link list */
953  /* ------------------------------------------------- */
954 
955  ilast = Last [i] ;
956  inext = Next [i] ;
957  ASSERT (ilast >= EMPTY && ilast < n) ;
958  ASSERT (inext >= EMPTY && inext < n) ;
959  if (inext != EMPTY) Last [inext] = ilast ;
960  if (ilast != EMPTY)
961  {
962  Next [ilast] = inext ;
963  }
964  else
965  {
966  /* i is at the head of the degree list */
967  ASSERT (Degree [i] >= 0 && Degree [i] < n) ;
968  Head [Degree [i]] = inext ;
969  }
970  }
971  }
972 
973  if (e != me)
974  {
975  /* set tree pointer and flag to indicate element e is
976  * absorbed into new element me (the parent of e is me) */
977  AMD_DEBUG1 ((" Element "ID" => "ID"\n", e, me)) ;
978  Pe [e] = FLIP (me) ;
979  W [e] = 0 ;
980  }
981  }
982 
983  pme2 = pfree - 1 ;
984  }
985 
986  /* ----------------------------------------------------------------- */
987  /* me has now been converted into an element in Iw [pme1..pme2] */
988  /* ----------------------------------------------------------------- */
989 
990  /* degme holds the external degree of new element */
991  Degree [me] = degme ;
992  Pe [me] = pme1 ;
993  Len [me] = pme2 - pme1 + 1 ;
994  ASSERT (Pe [me] >= 0 && Pe [me] < iwlen) ;
995 
996  Elen [me] = FLIP (nvpiv + degme) ;
997  /* FLIP (Elen (me)) is now the degree of pivot (including
998  * diagonal part). */
999 
1000 #ifndef NDEBUG
1001  AMD_DEBUG2 (("New element structure: length= "ID"\n", pme2-pme1+1)) ;
1002  for (pme = pme1 ; pme <= pme2 ; pme++) AMD_DEBUG3 ((" "ID"", Iw[pme]));
1003  AMD_DEBUG3 (("\n")) ;
1004 #endif
1005 
1006  /* ----------------------------------------------------------------- */
1007  /* make sure that wflg is not too large. */
1008  /* ----------------------------------------------------------------- */
1009 
1010  /* With the current value of wflg, wflg+n must not cause integer
1011  * overflow */
1012 
1013  wflg = amesos_clear_flag (wflg, wbig, W, n) ;
1014 
1015 /* ========================================================================= */
1016 /* COMPUTE (W [e] - wflg) = |Le\Lme| FOR ALL ELEMENTS */
1017 /* ========================================================================= */
1018 
1019  /* -----------------------------------------------------------------
1020  * Scan 1: compute the external degrees of previous elements with
1021  * respect to the current element. That is:
1022  * (W [e] - wflg) = |Le \ Lme|
1023  * for each element e that appears in any supervariable in Lme. The
1024  * notation Le refers to the pattern (list of supervariables) of a
1025  * previous element e, where e is not yet absorbed, stored in
1026  * Iw [Pe [e] + 1 ... Pe [e] + Len [e]]. The notation Lme
1027  * refers to the pattern of the current element (stored in
1028  * Iw [pme1..pme2]). If aggressive absorption is enabled, and
1029  * (W [e] - wflg) becomes zero, then the element e will be absorbed
1030  * in Scan 2.
1031  * ----------------------------------------------------------------- */
1032 
1033  AMD_DEBUG2 (("me: ")) ;
1034  for (pme = pme1 ; pme <= pme2 ; pme++)
1035  {
1036  i = Iw [pme] ;
1037  ASSERT (i >= 0 && i < n) ;
1038  eln = Elen [i] ;
1039  AMD_DEBUG3 ((""ID" Elen "ID": \n", i, eln)) ;
1040  if (eln > 0)
1041  {
1042  /* note that Nv [i] has been negated to denote i in Lme: */
1043  nvi = -Nv [i] ;
1044  ASSERT (nvi > 0 && Pe [i] >= 0 && Pe [i] < iwlen) ;
1045  wnvi = wflg - nvi ;
1046  for (p = Pe [i] ; p <= Pe [i] + eln - 1 ; p++)
1047  {
1048  e = Iw [p] ;
1049  ASSERT (e >= 0 && e < n) ;
1050  we = W [e] ;
1051  AMD_DEBUG4 ((" e "ID" we "ID" ", e, we)) ;
1052  if (we >= wflg)
1053  {
1054  /* unabsorbed element e has been seen in this loop */
1055  AMD_DEBUG4 ((" unabsorbed, first time seen")) ;
1056  we -= nvi ;
1057  }
1058  else if (we != 0)
1059  {
1060  /* e is an unabsorbed element */
1061  /* this is the first we have seen e in all of Scan 1 */
1062  AMD_DEBUG4 ((" unabsorbed")) ;
1063  we = Degree [e] + wnvi ;
1064  }
1065  AMD_DEBUG4 (("\n")) ;
1066  W [e] = we ;
1067  }
1068  }
1069  }
1070  AMD_DEBUG2 (("\n")) ;
1071 
1072 /* ========================================================================= */
1073 /* DEGREE UPDATE AND ELEMENT ABSORPTION */
1074 /* ========================================================================= */
1075 
1076  /* -----------------------------------------------------------------
1077  * Scan 2: for each i in Lme, sum up the degree of Lme (which is
1078  * degme), plus the sum of the external degrees of each Le for the
1079  * elements e appearing within i, plus the supervariables in i.
1080  * Place i in hash list.
1081  * ----------------------------------------------------------------- */
1082 
1083  for (pme = pme1 ; pme <= pme2 ; pme++)
1084  {
1085  i = Iw [pme] ;
1086  ASSERT (i >= 0 && i < n && Nv [i] < 0 && Elen [i] >= 0) ;
1087  AMD_DEBUG2 (("Updating: i "ID" "ID" "ID"\n", i, Elen[i], Len [i]));
1088  p1 = Pe [i] ;
1089  p2 = p1 + Elen [i] - 1 ;
1090  pn = p1 ;
1091  hash = 0 ;
1092  deg = 0 ;
1093  ASSERT (p1 >= 0 && p1 < iwlen && p2 >= -1 && p2 < iwlen) ;
1094 
1095  /* ------------------------------------------------------------- */
1096  /* scan the element list associated with supervariable i */
1097  /* ------------------------------------------------------------- */
1098 
1099  /* UMFPACK/MA38-style approximate degree: */
1100  if (aggressive)
1101  {
1102  for (p = p1 ; p <= p2 ; p++)
1103  {
1104  e = Iw [p] ;
1105  ASSERT (e >= 0 && e < n) ;
1106  we = W [e] ;
1107  if (we != 0)
1108  {
1109  /* e is an unabsorbed element */
1110  /* dext = | Le \ Lme | */
1111  dext = we - wflg ;
1112  if (dext > 0)
1113  {
1114  deg += dext ;
1115  Iw [pn++] = e ;
1116  hash += e ;
1117  AMD_DEBUG4 ((" e: "ID" hash = "ID"\n",e,hash)) ;
1118  }
1119  else
1120  {
1121  /* external degree of e is zero, absorb e into me*/
1122  AMD_DEBUG1 ((" Element "ID" =>"ID" (aggressive)\n",
1123  e, me)) ;
1124  ASSERT (dext == 0) ;
1125  Pe [e] = FLIP (me) ;
1126  W [e] = 0 ;
1127  }
1128  }
1129  }
1130  }
1131  else
1132  {
1133  for (p = p1 ; p <= p2 ; p++)
1134  {
1135  e = Iw [p] ;
1136  ASSERT (e >= 0 && e < n) ;
1137  we = W [e] ;
1138  if (we != 0)
1139  {
1140  /* e is an unabsorbed element */
1141  dext = we - wflg ;
1142  ASSERT (dext >= 0) ;
1143  deg += dext ;
1144  Iw [pn++] = e ;
1145  hash += e ;
1146  AMD_DEBUG4 ((" e: "ID" hash = "ID"\n",e,hash)) ;
1147  }
1148  }
1149  }
1150 
1151  /* count the number of elements in i (including me): */
1152  Elen [i] = pn - p1 + 1 ;
1153 
1154  /* ------------------------------------------------------------- */
1155  /* scan the supervariables in the list associated with i */
1156  /* ------------------------------------------------------------- */
1157 
1158  /* The bulk of the AMD run time is typically spent in this loop,
1159  * particularly if the matrix has many dense rows that are not
1160  * removed prior to ordering. */
1161  p3 = pn ;
1162  p4 = p1 + Len [i] ;
1163  for (p = p2 + 1 ; p < p4 ; p++)
1164  {
1165  j = Iw [p] ;
1166  ASSERT (j >= 0 && j < n) ;
1167  nvj = Nv [j] ;
1168  if (nvj > 0)
1169  {
1170  /* j is unabsorbed, and not in Lme. */
1171  /* add to degree and add to new list */
1172  deg += nvj ;
1173  Iw [pn++] = j ;
1174  hash += j ;
1175  AMD_DEBUG4 ((" s: "ID" hash "ID" Nv[j]= "ID"\n",
1176  j, hash, nvj)) ;
1177  }
1178  }
1179 
1180  /* ------------------------------------------------------------- */
1181  /* update the degree and check for mass elimination */
1182  /* ------------------------------------------------------------- */
1183 
1184  /* with aggressive absorption, deg==0 is identical to the
1185  * Elen [i] == 1 && p3 == pn test, below. */
1186  ASSERT (IMPLIES (aggressive, (deg==0) == (Elen[i]==1 && p3==pn))) ;
1187 
1188  if (Elen [i] == 1 && p3 == pn)
1189  {
1190 
1191  /* --------------------------------------------------------- */
1192  /* mass elimination */
1193  /* --------------------------------------------------------- */
1194 
1195  /* There is nothing left of this node except for an edge to
1196  * the current pivot element. Elen [i] is 1, and there are
1197  * no variables adjacent to node i. Absorb i into the
1198  * current pivot element, me. Note that if there are two or
1199  * more mass eliminations, fillin due to mass elimination is
1200  * possible within the nvpiv-by-nvpiv pivot block. It is this
1201  * step that causes AMD's analysis to be an upper bound.
1202  *
1203  * The reason is that the selected pivot has a lower
1204  * approximate degree than the true degree of the two mass
1205  * eliminated nodes. There is no edge between the two mass
1206  * eliminated nodes. They are merged with the current pivot
1207  * anyway.
1208  *
1209  * No fillin occurs in the Schur complement, in any case,
1210  * and this effect does not decrease the quality of the
1211  * ordering itself, just the quality of the nonzero and
1212  * flop count analysis. It also means that the post-ordering
1213  * is not an exact elimination tree post-ordering. */
1214 
1215  AMD_DEBUG1 ((" MASS i "ID" => parent e "ID"\n", i, me)) ;
1216  Pe [i] = FLIP (me) ;
1217  nvi = -Nv [i] ;
1218  degme -= nvi ;
1219  nvpiv += nvi ;
1220  nel += nvi ;
1221  Nv [i] = 0 ;
1222  Elen [i] = EMPTY ;
1223 
1224  }
1225  else
1226  {
1227 
1228  /* --------------------------------------------------------- */
1229  /* update the upper-bound degree of i */
1230  /* --------------------------------------------------------- */
1231 
1232  /* the following degree does not yet include the size
1233  * of the current element, which is added later: */
1234 
1235  Degree [i] = MIN (Degree [i], deg) ;
1236 
1237  /* --------------------------------------------------------- */
1238  /* add me to the list for i */
1239  /* --------------------------------------------------------- */
1240 
1241  /* move first supervariable to end of list */
1242  Iw [pn] = Iw [p3] ;
1243  /* move first element to end of element part of list */
1244  Iw [p3] = Iw [p1] ;
1245  /* add new element, me, to front of list. */
1246  Iw [p1] = me ;
1247  /* store the new length of the list in Len [i] */
1248  Len [i] = pn - p1 + 1 ;
1249 
1250  /* --------------------------------------------------------- */
1251  /* place in hash bucket. Save hash key of i in Last [i]. */
1252  /* --------------------------------------------------------- */
1253 
1254  /* NOTE: this can fail if hash is negative, because the ANSI C
1255  * standard does not define a % b when a and/or b are negative.
1256  * That's why hash is defined as an unsigned Int, to avoid this
1257  * problem. */
1258  hash = hash % n ;
1259  ASSERT (((Int) hash) >= 0 && ((Int) hash) < n) ;
1260 
1261  /* if the Hhead array is not used: */
1262  j = Head [hash] ;
1263  if (j <= EMPTY)
1264  {
1265  /* degree list is empty, hash head is FLIP (j) */
1266  Next [i] = FLIP (j) ;
1267  Head [hash] = FLIP (i) ;
1268  }
1269  else
1270  {
1271  /* degree list is not empty, use Last [Head [hash]] as
1272  * hash head. */
1273  Next [i] = Last [j] ;
1274  Last [j] = i ;
1275  }
1276 
1277  /* if a separate Hhead array is used: *
1278  Next [i] = Hhead [hash] ;
1279  Hhead [hash] = i ;
1280  */
1281 
1282  Last [i] = hash ;
1283  }
1284  }
1285 
1286  Degree [me] = degme ;
1287 
1288  /* ----------------------------------------------------------------- */
1289  /* Clear the counter array, W [...], by incrementing wflg. */
1290  /* ----------------------------------------------------------------- */
1291 
1292  /* make sure that wflg+n does not cause integer overflow */
1293  lemax = MAX (lemax, degme) ;
1294  wflg += lemax ;
1295  wflg = amesos_clear_flag (wflg, wbig, W, n) ;
1296  /* at this point, W [0..n-1] < wflg holds */
1297 
1298 /* ========================================================================= */
1299 /* SUPERVARIABLE DETECTION */
1300 /* ========================================================================= */
1301 
1302  AMD_DEBUG1 (("Detecting supervariables:\n")) ;
1303  for (pme = pme1 ; pme <= pme2 ; pme++)
1304  {
1305  i = Iw [pme] ;
1306  ASSERT (i >= 0 && i < n) ;
1307  AMD_DEBUG2 (("Consider i "ID" nv "ID"\n", i, Nv [i])) ;
1308  if (Nv [i] < 0)
1309  {
1310  /* i is a principal variable in Lme */
1311 
1312  /* ---------------------------------------------------------
1313  * examine all hash buckets with 2 or more variables. We do
1314  * this by examing all unique hash keys for supervariables in
1315  * the pattern Lme of the current element, me
1316  * --------------------------------------------------------- */
1317 
1318  /* let i = head of hash bucket, and empty the hash bucket */
1319  ASSERT (Last [i] >= 0 && Last [i] < n) ;
1320  hash = Last [i] ;
1321 
1322  /* if Hhead array is not used: */
1323  j = Head [hash] ;
1324  if (j == EMPTY)
1325  {
1326  /* hash bucket and degree list are both empty */
1327  i = EMPTY ;
1328  }
1329  else if (j < EMPTY)
1330  {
1331  /* degree list is empty */
1332  i = FLIP (j) ;
1333  Head [hash] = EMPTY ;
1334  }
1335  else
1336  {
1337  /* degree list is not empty, restore Last [j] of head j */
1338  i = Last [j] ;
1339  Last [j] = EMPTY ;
1340  }
1341 
1342  /* if separate Hhead array is used: *
1343  i = Hhead [hash] ;
1344  Hhead [hash] = EMPTY ;
1345  */
1346 
1347  ASSERT (i >= EMPTY && i < n) ;
1348  AMD_DEBUG2 (("----i "ID" hash "ID"\n", i, hash)) ;
1349 
1350  while (i != EMPTY && Next [i] != EMPTY)
1351  {
1352 
1353  /* -----------------------------------------------------
1354  * this bucket has one or more variables following i.
1355  * scan all of them to see if i can absorb any entries
1356  * that follow i in hash bucket. Scatter i into w.
1357  * ----------------------------------------------------- */
1358 
1359  ln = Len [i] ;
1360  eln = Elen [i] ;
1361  ASSERT (ln >= 0 && eln >= 0) ;
1362  ASSERT (Pe [i] >= 0 && Pe [i] < iwlen) ;
1363  /* do not flag the first element in the list (me) */
1364  for (p = Pe [i] + 1 ; p <= Pe [i] + ln - 1 ; p++)
1365  {
1366  ASSERT (Iw [p] >= 0 && Iw [p] < n) ;
1367  W [Iw [p]] = wflg ;
1368  }
1369 
1370  /* ----------------------------------------------------- */
1371  /* scan every other entry j following i in bucket */
1372  /* ----------------------------------------------------- */
1373 
1374  jlast = i ;
1375  j = Next [i] ;
1376  ASSERT (j >= EMPTY && j < n) ;
1377 
1378  while (j != EMPTY)
1379  {
1380  /* ------------------------------------------------- */
1381  /* check if j and i have identical nonzero pattern */
1382  /* ------------------------------------------------- */
1383 
1384  AMD_DEBUG3 (("compare i "ID" and j "ID"\n", i,j)) ;
1385 
1386  /* check if i and j have the same Len and Elen */
1387  ASSERT (Len [j] >= 0 && Elen [j] >= 0) ;
1388  ASSERT (Pe [j] >= 0 && Pe [j] < iwlen) ;
1389  ok = (Len [j] == ln) && (Elen [j] == eln) ;
1390  /* skip the first element in the list (me) */
1391  for (p = Pe [j] + 1 ; ok && p <= Pe [j] + ln - 1 ; p++)
1392  {
1393  ASSERT (Iw [p] >= 0 && Iw [p] < n) ;
1394  if (W [Iw [p]] != wflg) ok = 0 ;
1395  }
1396  if (ok)
1397  {
1398  /* --------------------------------------------- */
1399  /* found it! j can be absorbed into i */
1400  /* --------------------------------------------- */
1401 
1402  AMD_DEBUG1 (("found it! j "ID" => i "ID"\n", j,i));
1403  Pe [j] = FLIP (i) ;
1404  /* both Nv [i] and Nv [j] are negated since they */
1405  /* are in Lme, and the absolute values of each */
1406  /* are the number of variables in i and j: */
1407  Nv [i] += Nv [j] ;
1408  Nv [j] = 0 ;
1409  Elen [j] = EMPTY ;
1410  /* delete j from hash bucket */
1411  ASSERT (j != Next [j]) ;
1412  j = Next [j] ;
1413  Next [jlast] = j ;
1414 
1415  }
1416  else
1417  {
1418  /* j cannot be absorbed into i */
1419  jlast = j ;
1420  ASSERT (j != Next [j]) ;
1421  j = Next [j] ;
1422  }
1423  ASSERT (j >= EMPTY && j < n) ;
1424  }
1425 
1426  /* -----------------------------------------------------
1427  * no more variables can be absorbed into i
1428  * go to next i in bucket and clear flag array
1429  * ----------------------------------------------------- */
1430 
1431  wflg++ ;
1432  i = Next [i] ;
1433  ASSERT (i >= EMPTY && i < n) ;
1434 
1435  }
1436  }
1437  }
1438  AMD_DEBUG2 (("detect done\n")) ;
1439 
1440 /* ========================================================================= */
1441 /* RESTORE DEGREE LISTS AND REMOVE NONPRINCIPAL SUPERVARIABLES FROM ELEMENT */
1442 /* ========================================================================= */
1443 
1444  p = pme1 ;
1445  nleft = n - nel ;
1446  for (pme = pme1 ; pme <= pme2 ; pme++)
1447  {
1448  i = Iw [pme] ;
1449  ASSERT (i >= 0 && i < n) ;
1450  nvi = -Nv [i] ;
1451  AMD_DEBUG3 (("Restore i "ID" "ID"\n", i, nvi)) ;
1452  if (nvi > 0)
1453  {
1454  /* i is a principal variable in Lme */
1455  /* restore Nv [i] to signify that i is principal */
1456  Nv [i] = nvi ;
1457 
1458  /* --------------------------------------------------------- */
1459  /* compute the external degree (add size of current element) */
1460  /* --------------------------------------------------------- */
1461 
1462  deg = Degree [i] + degme - nvi ;
1463  deg = MIN (deg, nleft - nvi) ;
1464  ASSERT (IMPLIES (aggressive, deg > 0) && deg >= 0 && deg < n) ;
1465 
1466  /* --------------------------------------------------------- */
1467  /* place the supervariable at the head of the degree list */
1468  /* --------------------------------------------------------- */
1469 
1470  inext = Head [deg] ;
1471  ASSERT (inext >= EMPTY && inext < n) ;
1472  if (inext != EMPTY) Last [inext] = i ;
1473  Next [i] = inext ;
1474  Last [i] = EMPTY ;
1475  Head [deg] = i ;
1476 
1477  /* --------------------------------------------------------- */
1478  /* save the new degree, and find the minimum degree */
1479  /* --------------------------------------------------------- */
1480 
1481  mindeg = MIN (mindeg, deg) ;
1482  Degree [i] = deg ;
1483 
1484  /* --------------------------------------------------------- */
1485  /* place the supervariable in the element pattern */
1486  /* --------------------------------------------------------- */
1487 
1488  Iw [p++] = i ;
1489 
1490  }
1491  }
1492  AMD_DEBUG2 (("restore done\n")) ;
1493 
1494 /* ========================================================================= */
1495 /* FINALIZE THE NEW ELEMENT */
1496 /* ========================================================================= */
1497 
1498  AMD_DEBUG2 (("ME = "ID" DONE\n", me)) ;
1499  Nv [me] = nvpiv ;
1500  /* save the length of the list for the new element me */
1501  Len [me] = p - pme1 ;
1502  if (Len [me] == 0)
1503  {
1504  /* there is nothing left of the current pivot element */
1505  /* it is a root of the assembly tree */
1506  Pe [me] = EMPTY ;
1507  W [me] = 0 ;
1508  }
1509  if (elenme != 0)
1510  {
1511  /* element was not constructed in place: deallocate part of */
1512  /* it since newly nonprincipal variables may have been removed */
1513  pfree = p ;
1514  }
1515 
1516  /* The new element has nvpiv pivots and the size of the contribution
1517  * block for a multifrontal method is degme-by-degme, not including
1518  * the "dense" rows/columns. If the "dense" rows/columns are included,
1519  * the frontal matrix is no larger than
1520  * (degme+ndense)-by-(degme+ndense).
1521  */
1522 
1523  if (Info != (double *) NULL)
1524  {
1525  f = nvpiv ;
1526  r = degme + ndense ;
1527  dmax = MAX (dmax, f + r) ;
1528 
1529  /* number of nonzeros in L (excluding the diagonal) */
1530  lnzme = f*r + (f-1)*f/2 ;
1531  lnz += lnzme ;
1532 
1533  /* number of divide operations for LDL' and for LU */
1534  ndiv += lnzme ;
1535 
1536  /* number of multiply-subtract pairs for LU */
1537  s = f*r*r + r*(f-1)*f + (f-1)*f*(2*f-1)/6 ;
1538  nms_lu += s ;
1539 
1540  /* number of multiply-subtract pairs for LDL' */
1541  nms_ldl += (s + lnzme)/2 ;
1542  }
1543 
1544 #ifndef NDEBUG
1545  AMD_DEBUG2 (("finalize done nel "ID" n "ID"\n ::::\n", nel, n)) ;
1546  for (pme = Pe [me] ; pme <= Pe [me] + Len [me] - 1 ; pme++)
1547  {
1548  AMD_DEBUG3 ((" "ID"", Iw [pme])) ;
1549  }
1550  AMD_DEBUG3 (("\n")) ;
1551 #endif
1552 
1553  }
1554 
1555 /* ========================================================================= */
1556 /* DONE SELECTING PIVOTS */
1557 /* ========================================================================= */
1558 
1559  if (Info != (double *) NULL)
1560  {
1561 
1562  /* count the work to factorize the ndense-by-ndense submatrix */
1563  f = ndense ;
1564  dmax = MAX (dmax, (double) ndense) ;
1565 
1566  /* number of nonzeros in L (excluding the diagonal) */
1567  lnzme = (f-1)*f/2 ;
1568  lnz += lnzme ;
1569 
1570  /* number of divide operations for LDL' and for LU */
1571  ndiv += lnzme ;
1572 
1573  /* number of multiply-subtract pairs for LU */
1574  s = (f-1)*f*(2*f-1)/6 ;
1575  nms_lu += s ;
1576 
1577  /* number of multiply-subtract pairs for LDL' */
1578  nms_ldl += (s + lnzme)/2 ;
1579 
1580  /* number of nz's in L (excl. diagonal) */
1581  Info [AMD_LNZ] = lnz ;
1582 
1583  /* number of divide ops for LU and LDL' */
1584  Info [AMD_NDIV] = ndiv ;
1585 
1586  /* number of multiply-subtract pairs for LDL' */
1587  Info [AMD_NMULTSUBS_LDL] = nms_ldl ;
1588 
1589  /* number of multiply-subtract pairs for LU */
1590  Info [AMD_NMULTSUBS_LU] = nms_lu ;
1591 
1592  /* number of "dense" rows/columns */
1593  Info [AMD_NDENSE] = ndense ;
1594 
1595  /* largest front is dmax-by-dmax */
1596  Info [AMD_DMAX] = dmax ;
1597 
1598  /* number of garbage collections in AMD */
1599  Info [AMD_NCMPA] = ncmpa ;
1600 
1601  /* successful ordering */
1602  Info [AMD_STATUS] = AMD_OK ;
1603  }
1604 
1605 /* ========================================================================= */
1606 /* POST-ORDERING */
1607 /* ========================================================================= */
1608 
1609 /* -------------------------------------------------------------------------
1610  * Variables at this point:
1611  *
1612  * Pe: holds the elimination tree. The parent of j is FLIP (Pe [j]),
1613  * or EMPTY if j is a root. The tree holds both elements and
1614  * non-principal (unordered) variables absorbed into them.
1615  * Dense variables are non-principal and unordered.
1616  *
1617  * Elen: holds the size of each element, including the diagonal part.
1618  * FLIP (Elen [e]) > 0 if e is an element. For unordered
1619  * variables i, Elen [i] is EMPTY.
1620  *
1621  * Nv: Nv [e] > 0 is the number of pivots represented by the element e.
1622  * For unordered variables i, Nv [i] is zero.
1623  *
1624  * Contents no longer needed:
1625  * W, Iw, Len, Degree, Head, Next, Last.
1626  *
1627  * The matrix itself has been destroyed.
1628  *
1629  * n: the size of the matrix.
1630  * No other scalars needed (pfree, iwlen, etc.)
1631  * ------------------------------------------------------------------------- */
1632 
1633  /* restore Pe */
1634  for (i = 0 ; i < n ; i++)
1635  {
1636  Pe [i] = FLIP (Pe [i]) ;
1637  }
1638 
1639  /* restore Elen, for output information, and for postordering */
1640  for (i = 0 ; i < n ; i++)
1641  {
1642  Elen [i] = FLIP (Elen [i]) ;
1643  }
1644 
1645 /* Now the parent of j is Pe [j], or EMPTY if j is a root. Elen [e] > 0
1646  * is the size of element e. Elen [i] is EMPTY for unordered variable i. */
1647 
1648 #ifndef NDEBUG
1649  AMD_DEBUG2 (("\nTree:\n")) ;
1650  for (i = 0 ; i < n ; i++)
1651  {
1652  AMD_DEBUG2 ((" "ID" parent: "ID" ", i, Pe [i])) ;
1653  ASSERT (Pe [i] >= EMPTY && Pe [i] < n) ;
1654  if (Nv [i] > 0)
1655  {
1656  /* this is an element */
1657  e = i ;
1658  AMD_DEBUG2 ((" element, size is "ID"\n", Elen [i])) ;
1659  ASSERT (Elen [e] > 0) ;
1660  }
1661  AMD_DEBUG2 (("\n")) ;
1662  }
1663  AMD_DEBUG2 (("\nelements:\n")) ;
1664  for (e = 0 ; e < n ; e++)
1665  {
1666  if (Nv [e] > 0)
1667  {
1668  AMD_DEBUG3 (("Element e= "ID" size "ID" nv "ID" \n", e,
1669  Elen [e], Nv [e])) ;
1670  }
1671  }
1672  AMD_DEBUG2 (("\nvariables:\n")) ;
1673  for (i = 0 ; i < n ; i++)
1674  {
1675  Int cnt ;
1676  if (Nv [i] == 0)
1677  {
1678  AMD_DEBUG3 (("i unordered: "ID"\n", i)) ;
1679  j = Pe [i] ;
1680  cnt = 0 ;
1681  AMD_DEBUG3 ((" j: "ID"\n", j)) ;
1682  if (j == EMPTY)
1683  {
1684  AMD_DEBUG3 ((" i is a dense variable\n")) ;
1685  }
1686  else
1687  {
1688  ASSERT (j >= 0 && j < n) ;
1689  while (Nv [j] == 0)
1690  {
1691  AMD_DEBUG3 ((" j : "ID"\n", j)) ;
1692  j = Pe [j] ;
1693  AMD_DEBUG3 ((" j:: "ID"\n", j)) ;
1694  cnt++ ;
1695  if (cnt > n) break ;
1696  }
1697  e = j ;
1698  AMD_DEBUG3 ((" got to e: "ID"\n", e)) ;
1699  }
1700  }
1701  }
1702 #endif
1703 
1704 /* ========================================================================= */
1705 /* compress the paths of the variables */
1706 /* ========================================================================= */
1707 
1708  for (i = 0 ; i < n ; i++)
1709  {
1710  if (Nv [i] == 0)
1711  {
1712 
1713  /* -------------------------------------------------------------
1714  * i is an un-ordered row. Traverse the tree from i until
1715  * reaching an element, e. The element, e, was the principal
1716  * supervariable of i and all nodes in the path from i to when e
1717  * was selected as pivot.
1718  * ------------------------------------------------------------- */
1719 
1720  AMD_DEBUG1 (("Path compression, i unordered: "ID"\n", i)) ;
1721  j = Pe [i] ;
1722  ASSERT (j >= EMPTY && j < n) ;
1723  AMD_DEBUG3 ((" j: "ID"\n", j)) ;
1724  if (j == EMPTY)
1725  {
1726  /* Skip a dense variable. It has no parent. */
1727  AMD_DEBUG3 ((" i is a dense variable\n")) ;
1728  continue ;
1729  }
1730 
1731  /* while (j is a variable) */
1732  while (Nv [j] == 0)
1733  {
1734  AMD_DEBUG3 ((" j : "ID"\n", j)) ;
1735  j = Pe [j] ;
1736  AMD_DEBUG3 ((" j:: "ID"\n", j)) ;
1737  ASSERT (j >= 0 && j < n) ;
1738  }
1739  /* got to an element e */
1740  e = j ;
1741  AMD_DEBUG3 (("got to e: "ID"\n", e)) ;
1742 
1743  /* -------------------------------------------------------------
1744  * traverse the path again from i to e, and compress the path
1745  * (all nodes point to e). Path compression allows this code to
1746  * compute in O(n) time.
1747  * ------------------------------------------------------------- */
1748 
1749  j = i ;
1750  /* while (j is a variable) */
1751  while (Nv [j] == 0)
1752  {
1753  jnext = Pe [j] ;
1754  AMD_DEBUG3 (("j "ID" jnext "ID"\n", j, jnext)) ;
1755  Pe [j] = e ;
1756  j = jnext ;
1757  ASSERT (j >= 0 && j < n) ;
1758  }
1759  }
1760  }
1761 
1762 /* ========================================================================= */
1763 /* postorder the assembly tree */
1764 /* ========================================================================= */
1765 
1766  AMD_postorder (n, Pe, Nv, Elen,
1767  W, /* output order */
1768  Head, Next, Last) ; /* workspace */
1769 
1770 /* ========================================================================= */
1771 /* compute output permutation and inverse permutation */
1772 /* ========================================================================= */
1773 
1774  /* W [e] = k means that element e is the kth element in the new
1775  * order. e is in the range 0 to n-1, and k is in the range 0 to
1776  * the number of elements. Use Head for inverse order. */
1777 
1778  for (k = 0 ; k < n ; k++)
1779  {
1780  Head [k] = EMPTY ;
1781  Next [k] = EMPTY ;
1782  }
1783  for (e = 0 ; e < n ; e++)
1784  {
1785  k = W [e] ;
1786  ASSERT ((k == EMPTY) == (Nv [e] == 0)) ;
1787  if (k != EMPTY)
1788  {
1789  ASSERT (k >= 0 && k < n) ;
1790  Head [k] = e ;
1791  }
1792  }
1793 
1794  /* construct output inverse permutation in Next,
1795  * and permutation in Last */
1796  nel = 0 ;
1797  for (k = 0 ; k < n ; k++)
1798  {
1799  e = Head [k] ;
1800  if (e == EMPTY) break ;
1801  ASSERT (e >= 0 && e < n && Nv [e] > 0) ;
1802  Next [e] = nel ;
1803  nel += Nv [e] ;
1804  }
1805  ASSERT (nel == n - ndense) ;
1806 
1807  /* order non-principal variables (dense, & those merged into supervar's) */
1808  for (i = 0 ; i < n ; i++)
1809  {
1810  if (Nv [i] == 0)
1811  {
1812  e = Pe [i] ;
1813  ASSERT (e >= EMPTY && e < n) ;
1814  if (e != EMPTY)
1815  {
1816  /* This is an unordered variable that was merged
1817  * into element e via supernode detection or mass
1818  * elimination of i when e became the pivot element.
1819  * Place i in order just before e. */
1820  ASSERT (Next [i] == EMPTY && Nv [e] > 0) ;
1821  Next [i] = Next [e] ;
1822  Next [e]++ ;
1823  }
1824  else
1825  {
1826  /* This is a dense unordered variable, with no parent.
1827  * Place it last in the output order. */
1828  Next [i] = nel++ ;
1829  }
1830  }
1831  }
1832  ASSERT (nel == n) ;
1833 
1834  AMD_DEBUG2 (("\n\nPerm:\n")) ;
1835  for (i = 0 ; i < n ; i++)
1836  {
1837  k = Next [i] ;
1838  ASSERT (k >= 0 && k < n) ;
1839  Last [k] = i ;
1840  AMD_DEBUG2 ((" perm ["ID"] = "ID"\n", k, i)) ;
1841  }
1842 }
#define AMD_NCMPA
Definition: amesos_amd.h:360
void f()
#define AMD_NMULTSUBS_LDL
Definition: amesos_amd.h:363
#define EMPTY
#define GLOBAL
#define AMD_DEBUG1(params)
#define AMD_DEBUG3(params)
#define Int
#define AMD_DEFAULT_AGGRESSIVE
Definition: amesos_amd.h:349
#define AMD_STATUS
Definition: amesos_amd.h:352
GLOBAL void AMD_2(Int n, Int Pe[], Int Iw[], Int Len[], Int iwlen, Int pfree, Int Nv[], Int Next[], Int Last[], Int Head[], Int Elen[], Int Degree[], Int W[], double Control[], double Info[])
Definition: amesos_amd_2.c:44
#define MAX(a, b)
#define AMD_NDENSE
Definition: amesos_amd.h:358
#define AMD_debug
#define NULL
#define AMD_dump
static Int amesos_clear_flag(Int wflg, Int wbig, Int W[], Int n)
Definition: amesos_amd_2.c:23
#define AMD_DENSE
Definition: amesos_amd.h:344
#define ASSERT(expression)
#define FLIP(i)
#define ID
#define AMD_AGGRESSIVE
Definition: amesos_amd.h:345
#define AMD_DEBUG4(params)
#define AMD_DMAX
Definition: amesos_amd.h:365
#define AMD_NMULTSUBS_LU
Definition: amesos_amd.h:364
#define AMD_DEBUG2(params)
#define Int_MAX
#define AMD_OK
Definition: amesos_amd.h:371
#define AMD_NDIV
Definition: amesos_amd.h:362
#define AMD_postorder
#define MIN(a, b)
#define AMD_DEFAULT_DENSE
Definition: amesos_amd.h:348
#define IMPLIES(p, q)
#define AMD_LNZ
Definition: amesos_amd.h:361