1 | /**CFile*********************************************************************** |
---|
2 | |
---|
3 | FileName [cuddRead.c] |
---|
4 | |
---|
5 | PackageName [cudd] |
---|
6 | |
---|
7 | Synopsis [Functions to read in a matrix] |
---|
8 | |
---|
9 | Description [External procedures included in this module: |
---|
10 | <ul> |
---|
11 | <li> Cudd_addRead() |
---|
12 | <li> Cudd_bddRead() |
---|
13 | </ul>] |
---|
14 | |
---|
15 | SeeAlso [cudd_addHarwell.c] |
---|
16 | |
---|
17 | Author [Fabio Somenzi] |
---|
18 | |
---|
19 | Copyright [Copyright (c) 1995-2004, Regents of the University of Colorado |
---|
20 | |
---|
21 | All rights reserved. |
---|
22 | |
---|
23 | Redistribution and use in source and binary forms, with or without |
---|
24 | modification, are permitted provided that the following conditions |
---|
25 | are met: |
---|
26 | |
---|
27 | Redistributions of source code must retain the above copyright |
---|
28 | notice, this list of conditions and the following disclaimer. |
---|
29 | |
---|
30 | Redistributions in binary form must reproduce the above copyright |
---|
31 | notice, this list of conditions and the following disclaimer in the |
---|
32 | documentation and/or other materials provided with the distribution. |
---|
33 | |
---|
34 | Neither the name of the University of Colorado nor the names of its |
---|
35 | contributors may be used to endorse or promote products derived from |
---|
36 | this software without specific prior written permission. |
---|
37 | |
---|
38 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
---|
39 | "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
---|
40 | LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS |
---|
41 | FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE |
---|
42 | COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, |
---|
43 | INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, |
---|
44 | BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; |
---|
45 | LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER |
---|
46 | CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
---|
47 | LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN |
---|
48 | ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
---|
49 | POSSIBILITY OF SUCH DAMAGE.] |
---|
50 | |
---|
51 | ******************************************************************************/ |
---|
52 | |
---|
53 | #include "util.h" |
---|
54 | #include "cuddInt.h" |
---|
55 | |
---|
56 | |
---|
57 | /*---------------------------------------------------------------------------*/ |
---|
58 | /* Constant declarations */ |
---|
59 | /*---------------------------------------------------------------------------*/ |
---|
60 | |
---|
61 | |
---|
62 | /*---------------------------------------------------------------------------*/ |
---|
63 | /* Stucture declarations */ |
---|
64 | /*---------------------------------------------------------------------------*/ |
---|
65 | |
---|
66 | |
---|
67 | /*---------------------------------------------------------------------------*/ |
---|
68 | /* Type declarations */ |
---|
69 | /*---------------------------------------------------------------------------*/ |
---|
70 | |
---|
71 | |
---|
72 | /*---------------------------------------------------------------------------*/ |
---|
73 | /* Variable declarations */ |
---|
74 | /*---------------------------------------------------------------------------*/ |
---|
75 | |
---|
76 | #ifndef lint |
---|
77 | static char rcsid[] DD_UNUSED = "$Id: cuddRead.c,v 1.6 2004/08/13 18:04:50 fabio Exp $"; |
---|
78 | #endif |
---|
79 | |
---|
80 | /*---------------------------------------------------------------------------*/ |
---|
81 | /* Macro declarations */ |
---|
82 | /*---------------------------------------------------------------------------*/ |
---|
83 | |
---|
84 | |
---|
85 | /**AutomaticStart*************************************************************/ |
---|
86 | |
---|
87 | /*---------------------------------------------------------------------------*/ |
---|
88 | /* Static function prototypes */ |
---|
89 | /*---------------------------------------------------------------------------*/ |
---|
90 | |
---|
91 | |
---|
92 | /**AutomaticEnd***************************************************************/ |
---|
93 | |
---|
94 | |
---|
95 | /*---------------------------------------------------------------------------*/ |
---|
96 | /* Definition of exported functions */ |
---|
97 | /*---------------------------------------------------------------------------*/ |
---|
98 | |
---|
99 | |
---|
100 | /**Function******************************************************************** |
---|
101 | |
---|
102 | Synopsis [Reads in a sparse matrix.] |
---|
103 | |
---|
104 | Description [Reads in a sparse matrix specified in a simple format. |
---|
105 | The first line of the input contains the numbers of rows and columns. |
---|
106 | The remaining lines contain the elements of the matrix, one per line. |
---|
107 | Given a background value |
---|
108 | (specified by the background field of the manager), only the values |
---|
109 | different from it are explicitly listed. Each foreground element is |
---|
110 | described by two integers, i.e., the row and column number, and a |
---|
111 | real number, i.e., the value.<p> |
---|
112 | Cudd_addRead produces an ADD that depends on two sets of variables: x |
---|
113 | and y. The x variables (x\[0\] ... x\[nx-1\]) encode the row index and |
---|
114 | the y variables (y\[0\] ... y\[ny-1\]) encode the column index. |
---|
115 | x\[0\] and y\[0\] are the most significant bits in the indices. |
---|
116 | The variables may already exist or may be created by the function. |
---|
117 | The index of x\[i\] is bx+i*sx, and the index of y\[i\] is by+i*sy.<p> |
---|
118 | On input, nx and ny hold the numbers |
---|
119 | of row and column variables already in existence. On output, they |
---|
120 | hold the numbers of row and column variables actually used by the |
---|
121 | matrix. When Cudd_addRead creates the variable arrays, |
---|
122 | the index of x\[i\] is bx+i*sx, and the index of y\[i\] is by+i*sy. |
---|
123 | When some variables already exist Cudd_addRead expects the indices |
---|
124 | of the existing x variables to be bx+i*sx, and the indices of the |
---|
125 | existing y variables to be by+i*sy.<p> |
---|
126 | m and n are set to the numbers of rows and columns of the |
---|
127 | matrix. Their values on input are immaterial. |
---|
128 | The ADD for the |
---|
129 | sparse matrix is returned in E, and its reference count is > 0. |
---|
130 | Cudd_addRead returns 1 in case of success; 0 otherwise.] |
---|
131 | |
---|
132 | SideEffects [nx and ny are set to the numbers of row and column |
---|
133 | variables. m and n are set to the numbers of rows and columns. x and y |
---|
134 | are possibly extended to represent the array of row and column |
---|
135 | variables. Similarly for xn and yn_, which hold on return from |
---|
136 | Cudd_addRead the complements of the row and column variables.] |
---|
137 | |
---|
138 | SeeAlso [Cudd_addHarwell Cudd_bddRead] |
---|
139 | |
---|
140 | ******************************************************************************/ |
---|
141 | int |
---|
142 | Cudd_addRead( |
---|
143 | FILE * fp /* input file pointer */, |
---|
144 | DdManager * dd /* DD manager */, |
---|
145 | DdNode ** E /* characteristic function of the graph */, |
---|
146 | DdNode *** x /* array of row variables */, |
---|
147 | DdNode *** y /* array of column variables */, |
---|
148 | DdNode *** xn /* array of complemented row variables */, |
---|
149 | DdNode *** yn_ /* array of complemented column variables */, |
---|
150 | int * nx /* number or row variables */, |
---|
151 | int * ny /* number or column variables */, |
---|
152 | int * m /* number of rows */, |
---|
153 | int * n /* number of columns */, |
---|
154 | int bx /* first index of row variables */, |
---|
155 | int sx /* step of row variables */, |
---|
156 | int by /* first index of column variables */, |
---|
157 | int sy /* step of column variables */) |
---|
158 | { |
---|
159 | DdNode *one, *zero; |
---|
160 | DdNode *w, *neW; |
---|
161 | DdNode *minterm1; |
---|
162 | int u, v, err, i, nv; |
---|
163 | int lnx, lny; |
---|
164 | CUDD_VALUE_TYPE val; |
---|
165 | DdNode **lx, **ly, **lxn, **lyn; |
---|
166 | |
---|
167 | one = DD_ONE(dd); |
---|
168 | zero = DD_ZERO(dd); |
---|
169 | |
---|
170 | err = fscanf(fp, "%d %d", &u, &v); |
---|
171 | if (err == EOF) { |
---|
172 | return(0); |
---|
173 | } else if (err != 2) { |
---|
174 | return(0); |
---|
175 | } |
---|
176 | |
---|
177 | *m = u; |
---|
178 | /* Compute the number of x variables. */ |
---|
179 | lx = *x; lxn = *xn; |
---|
180 | u--; /* row and column numbers start from 0 */ |
---|
181 | for (lnx=0; u > 0; lnx++) { |
---|
182 | u >>= 1; |
---|
183 | } |
---|
184 | /* Here we rely on the fact that REALLOC of a null pointer is |
---|
185 | ** translates to an ALLOC. |
---|
186 | */ |
---|
187 | if (lnx > *nx) { |
---|
188 | *x = lx = REALLOC(DdNode *, *x, lnx); |
---|
189 | if (lx == NULL) { |
---|
190 | dd->errorCode = CUDD_MEMORY_OUT; |
---|
191 | return(0); |
---|
192 | } |
---|
193 | *xn = lxn = REALLOC(DdNode *, *xn, lnx); |
---|
194 | if (lxn == NULL) { |
---|
195 | dd->errorCode = CUDD_MEMORY_OUT; |
---|
196 | return(0); |
---|
197 | } |
---|
198 | } |
---|
199 | |
---|
200 | *n = v; |
---|
201 | /* Compute the number of y variables. */ |
---|
202 | ly = *y; lyn = *yn_; |
---|
203 | v--; /* row and column numbers start from 0 */ |
---|
204 | for (lny=0; v > 0; lny++) { |
---|
205 | v >>= 1; |
---|
206 | } |
---|
207 | /* Here we rely on the fact that REALLOC of a null pointer is |
---|
208 | ** translates to an ALLOC. |
---|
209 | */ |
---|
210 | if (lny > *ny) { |
---|
211 | *y = ly = REALLOC(DdNode *, *y, lny); |
---|
212 | if (ly == NULL) { |
---|
213 | dd->errorCode = CUDD_MEMORY_OUT; |
---|
214 | return(0); |
---|
215 | } |
---|
216 | *yn_ = lyn = REALLOC(DdNode *, *yn_, lny); |
---|
217 | if (lyn == NULL) { |
---|
218 | dd->errorCode = CUDD_MEMORY_OUT; |
---|
219 | return(0); |
---|
220 | } |
---|
221 | } |
---|
222 | |
---|
223 | /* Create all new variables. */ |
---|
224 | for (i = *nx, nv = bx + (*nx) * sx; i < lnx; i++, nv += sx) { |
---|
225 | do { |
---|
226 | dd->reordered = 0; |
---|
227 | lx[i] = cuddUniqueInter(dd, nv, one, zero); |
---|
228 | } while (dd->reordered == 1); |
---|
229 | if (lx[i] == NULL) return(0); |
---|
230 | cuddRef(lx[i]); |
---|
231 | do { |
---|
232 | dd->reordered = 0; |
---|
233 | lxn[i] = cuddUniqueInter(dd, nv, zero, one); |
---|
234 | } while (dd->reordered == 1); |
---|
235 | if (lxn[i] == NULL) return(0); |
---|
236 | cuddRef(lxn[i]); |
---|
237 | } |
---|
238 | for (i = *ny, nv = by + (*ny) * sy; i < lny; i++, nv += sy) { |
---|
239 | do { |
---|
240 | dd->reordered = 0; |
---|
241 | ly[i] = cuddUniqueInter(dd, nv, one, zero); |
---|
242 | } while (dd->reordered == 1); |
---|
243 | if (ly[i] == NULL) return(0); |
---|
244 | cuddRef(ly[i]); |
---|
245 | do { |
---|
246 | dd->reordered = 0; |
---|
247 | lyn[i] = cuddUniqueInter(dd, nv, zero, one); |
---|
248 | } while (dd->reordered == 1); |
---|
249 | if (lyn[i] == NULL) return(0); |
---|
250 | cuddRef(lyn[i]); |
---|
251 | } |
---|
252 | *nx = lnx; |
---|
253 | *ny = lny; |
---|
254 | |
---|
255 | *E = dd->background; /* this call will never cause reordering */ |
---|
256 | cuddRef(*E); |
---|
257 | |
---|
258 | while (! feof(fp)) { |
---|
259 | err = fscanf(fp, "%d %d %lf", &u, &v, &val); |
---|
260 | if (err == EOF) { |
---|
261 | break; |
---|
262 | } else if (err != 3) { |
---|
263 | return(0); |
---|
264 | } else if (u >= *m || v >= *n || u < 0 || v < 0) { |
---|
265 | return(0); |
---|
266 | } |
---|
267 | |
---|
268 | minterm1 = one; cuddRef(minterm1); |
---|
269 | |
---|
270 | /* Build minterm1 corresponding to this arc */ |
---|
271 | for (i = lnx - 1; i>=0; i--) { |
---|
272 | if (u & 1) { |
---|
273 | w = Cudd_addApply(dd, Cudd_addTimes, minterm1, lx[i]); |
---|
274 | } else { |
---|
275 | w = Cudd_addApply(dd, Cudd_addTimes, minterm1, lxn[i]); |
---|
276 | } |
---|
277 | if (w == NULL) { |
---|
278 | Cudd_RecursiveDeref(dd, minterm1); |
---|
279 | return(0); |
---|
280 | } |
---|
281 | cuddRef(w); |
---|
282 | Cudd_RecursiveDeref(dd, minterm1); |
---|
283 | minterm1 = w; |
---|
284 | u >>= 1; |
---|
285 | } |
---|
286 | for (i = lny - 1; i>=0; i--) { |
---|
287 | if (v & 1) { |
---|
288 | w = Cudd_addApply(dd, Cudd_addTimes, minterm1, ly[i]); |
---|
289 | } else { |
---|
290 | w = Cudd_addApply(dd, Cudd_addTimes, minterm1, lyn[i]); |
---|
291 | } |
---|
292 | if (w == NULL) { |
---|
293 | Cudd_RecursiveDeref(dd, minterm1); |
---|
294 | return(0); |
---|
295 | } |
---|
296 | cuddRef(w); |
---|
297 | Cudd_RecursiveDeref(dd, minterm1); |
---|
298 | minterm1 = w; |
---|
299 | v >>= 1; |
---|
300 | } |
---|
301 | /* Create new constant node if necessary. |
---|
302 | ** This call will never cause reordering. |
---|
303 | */ |
---|
304 | neW = cuddUniqueConst(dd, val); |
---|
305 | if (neW == NULL) { |
---|
306 | Cudd_RecursiveDeref(dd, minterm1); |
---|
307 | return(0); |
---|
308 | } |
---|
309 | cuddRef(neW); |
---|
310 | |
---|
311 | w = Cudd_addIte(dd, minterm1, neW, *E); |
---|
312 | if (w == NULL) { |
---|
313 | Cudd_RecursiveDeref(dd, minterm1); |
---|
314 | Cudd_RecursiveDeref(dd, neW); |
---|
315 | return(0); |
---|
316 | } |
---|
317 | cuddRef(w); |
---|
318 | Cudd_RecursiveDeref(dd, minterm1); |
---|
319 | Cudd_RecursiveDeref(dd, neW); |
---|
320 | Cudd_RecursiveDeref(dd, *E); |
---|
321 | *E = w; |
---|
322 | } |
---|
323 | return(1); |
---|
324 | |
---|
325 | } /* end of Cudd_addRead */ |
---|
326 | |
---|
327 | |
---|
328 | /**Function******************************************************************** |
---|
329 | |
---|
330 | Synopsis [Reads in a graph (without labels) given as a list of arcs.] |
---|
331 | |
---|
332 | Description [Reads in a graph (without labels) given as an adjacency |
---|
333 | matrix. The first line of the input contains the numbers of rows and |
---|
334 | columns of the adjacency matrix. The remaining lines contain the arcs |
---|
335 | of the graph, one per line. Each arc is described by two integers, |
---|
336 | i.e., the row and column number, or the indices of the two endpoints. |
---|
337 | Cudd_bddRead produces a BDD that depends on two sets of variables: x |
---|
338 | and y. The x variables (x\[0\] ... x\[nx-1\]) encode |
---|
339 | the row index and the y variables (y\[0\] ... y\[ny-1\]) encode the |
---|
340 | column index. x\[0\] and y\[0\] are the most significant bits in the |
---|
341 | indices. |
---|
342 | The variables may already exist or may be created by the function. |
---|
343 | The index of x\[i\] is bx+i*sx, and the index of y\[i\] is by+i*sy.<p> |
---|
344 | On input, nx and ny hold the numbers of row and column variables already |
---|
345 | in existence. On output, they hold the numbers of row and column |
---|
346 | variables actually used by the matrix. When Cudd_bddRead creates the |
---|
347 | variable arrays, the index of x\[i\] is bx+i*sx, and the index of |
---|
348 | y\[i\] is by+i*sy. When some variables already exist, Cudd_bddRead |
---|
349 | expects the indices of the existing x variables to be bx+i*sx, and the |
---|
350 | indices of the existing y variables to be by+i*sy.<p> |
---|
351 | m and n are set to the numbers of rows and columns of the |
---|
352 | matrix. Their values on input are immaterial. The BDD for the graph |
---|
353 | is returned in E, and its reference count is > 0. Cudd_bddRead returns |
---|
354 | 1 in case of success; 0 otherwise.] |
---|
355 | |
---|
356 | SideEffects [nx and ny are set to the numbers of row and column |
---|
357 | variables. m and n are set to the numbers of rows and columns. x and y |
---|
358 | are possibly extended to represent the array of row and column |
---|
359 | variables.] |
---|
360 | |
---|
361 | SeeAlso [Cudd_addHarwell Cudd_addRead] |
---|
362 | |
---|
363 | ******************************************************************************/ |
---|
364 | int |
---|
365 | Cudd_bddRead( |
---|
366 | FILE * fp /* input file pointer */, |
---|
367 | DdManager * dd /* DD manager */, |
---|
368 | DdNode ** E /* characteristic function of the graph */, |
---|
369 | DdNode *** x /* array of row variables */, |
---|
370 | DdNode *** y /* array of column variables */, |
---|
371 | int * nx /* number or row variables */, |
---|
372 | int * ny /* number or column variables */, |
---|
373 | int * m /* number of rows */, |
---|
374 | int * n /* number of columns */, |
---|
375 | int bx /* first index of row variables */, |
---|
376 | int sx /* step of row variables */, |
---|
377 | int by /* first index of column variables */, |
---|
378 | int sy /* step of column variables */) |
---|
379 | { |
---|
380 | DdNode *one, *zero; |
---|
381 | DdNode *w; |
---|
382 | DdNode *minterm1; |
---|
383 | int u, v, err, i, nv; |
---|
384 | int lnx, lny; |
---|
385 | DdNode **lx, **ly; |
---|
386 | |
---|
387 | one = DD_ONE(dd); |
---|
388 | zero = Cudd_Not(one); |
---|
389 | |
---|
390 | err = fscanf(fp, "%d %d", &u, &v); |
---|
391 | if (err == EOF) { |
---|
392 | return(0); |
---|
393 | } else if (err != 2) { |
---|
394 | return(0); |
---|
395 | } |
---|
396 | |
---|
397 | *m = u; |
---|
398 | /* Compute the number of x variables. */ |
---|
399 | lx = *x; |
---|
400 | u--; /* row and column numbers start from 0 */ |
---|
401 | for (lnx=0; u > 0; lnx++) { |
---|
402 | u >>= 1; |
---|
403 | } |
---|
404 | if (lnx > *nx) { |
---|
405 | *x = lx = REALLOC(DdNode *, *x, lnx); |
---|
406 | if (lx == NULL) { |
---|
407 | dd->errorCode = CUDD_MEMORY_OUT; |
---|
408 | return(0); |
---|
409 | } |
---|
410 | } |
---|
411 | |
---|
412 | *n = v; |
---|
413 | /* Compute the number of y variables. */ |
---|
414 | ly = *y; |
---|
415 | v--; /* row and column numbers start from 0 */ |
---|
416 | for (lny=0; v > 0; lny++) { |
---|
417 | v >>= 1; |
---|
418 | } |
---|
419 | if (lny > *ny) { |
---|
420 | *y = ly = REALLOC(DdNode *, *y, lny); |
---|
421 | if (ly == NULL) { |
---|
422 | dd->errorCode = CUDD_MEMORY_OUT; |
---|
423 | return(0); |
---|
424 | } |
---|
425 | } |
---|
426 | |
---|
427 | /* Create all new variables. */ |
---|
428 | for (i = *nx, nv = bx + (*nx) * sx; i < lnx; i++, nv += sx) { |
---|
429 | do { |
---|
430 | dd->reordered = 0; |
---|
431 | lx[i] = cuddUniqueInter(dd, nv, one, zero); |
---|
432 | } while (dd->reordered == 1); |
---|
433 | if (lx[i] == NULL) return(0); |
---|
434 | cuddRef(lx[i]); |
---|
435 | } |
---|
436 | for (i = *ny, nv = by + (*ny) * sy; i < lny; i++, nv += sy) { |
---|
437 | do { |
---|
438 | dd->reordered = 0; |
---|
439 | ly[i] = cuddUniqueInter(dd, nv, one, zero); |
---|
440 | } while (dd->reordered == 1); |
---|
441 | if (ly[i] == NULL) return(0); |
---|
442 | cuddRef(ly[i]); |
---|
443 | } |
---|
444 | *nx = lnx; |
---|
445 | *ny = lny; |
---|
446 | |
---|
447 | *E = zero; /* this call will never cause reordering */ |
---|
448 | cuddRef(*E); |
---|
449 | |
---|
450 | while (! feof(fp)) { |
---|
451 | err = fscanf(fp, "%d %d", &u, &v); |
---|
452 | if (err == EOF) { |
---|
453 | break; |
---|
454 | } else if (err != 2) { |
---|
455 | return(0); |
---|
456 | } else if (u >= *m || v >= *n || u < 0 || v < 0) { |
---|
457 | return(0); |
---|
458 | } |
---|
459 | |
---|
460 | minterm1 = one; cuddRef(minterm1); |
---|
461 | |
---|
462 | /* Build minterm1 corresponding to this arc. */ |
---|
463 | for (i = lnx - 1; i>=0; i--) { |
---|
464 | if (u & 1) { |
---|
465 | w = Cudd_bddAnd(dd, minterm1, lx[i]); |
---|
466 | } else { |
---|
467 | w = Cudd_bddAnd(dd, minterm1, Cudd_Not(lx[i])); |
---|
468 | } |
---|
469 | if (w == NULL) { |
---|
470 | Cudd_RecursiveDeref(dd, minterm1); |
---|
471 | return(0); |
---|
472 | } |
---|
473 | cuddRef(w); |
---|
474 | Cudd_RecursiveDeref(dd,minterm1); |
---|
475 | minterm1 = w; |
---|
476 | u >>= 1; |
---|
477 | } |
---|
478 | for (i = lny - 1; i>=0; i--) { |
---|
479 | if (v & 1) { |
---|
480 | w = Cudd_bddAnd(dd, minterm1, ly[i]); |
---|
481 | } else { |
---|
482 | w = Cudd_bddAnd(dd, minterm1, Cudd_Not(ly[i])); |
---|
483 | } |
---|
484 | if (w == NULL) { |
---|
485 | Cudd_RecursiveDeref(dd, minterm1); |
---|
486 | return(0); |
---|
487 | } |
---|
488 | cuddRef(w); |
---|
489 | Cudd_RecursiveDeref(dd, minterm1); |
---|
490 | minterm1 = w; |
---|
491 | v >>= 1; |
---|
492 | } |
---|
493 | |
---|
494 | w = Cudd_bddAnd(dd, Cudd_Not(minterm1), Cudd_Not(*E)); |
---|
495 | if (w == NULL) { |
---|
496 | Cudd_RecursiveDeref(dd, minterm1); |
---|
497 | return(0); |
---|
498 | } |
---|
499 | w = Cudd_Not(w); |
---|
500 | cuddRef(w); |
---|
501 | Cudd_RecursiveDeref(dd, minterm1); |
---|
502 | Cudd_RecursiveDeref(dd, *E); |
---|
503 | *E = w; |
---|
504 | } |
---|
505 | return(1); |
---|
506 | |
---|
507 | } /* end of Cudd_bddRead */ |
---|
508 | |
---|
509 | /*---------------------------------------------------------------------------*/ |
---|
510 | /* Definition of internal functions */ |
---|
511 | /*---------------------------------------------------------------------------*/ |
---|
512 | |
---|
513 | |
---|
514 | /*---------------------------------------------------------------------------*/ |
---|
515 | /* Definition of static functions */ |
---|
516 | /*---------------------------------------------------------------------------*/ |
---|
517 | |
---|