Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.117
1.117 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.116 2004/12/01 12:36:17 noro Exp $ */
1.2 noro 2:
1.94 noro 3: #include "nd.h"
1.63 noro 4:
1.61 noro 5: int (*ndl_compare_function)(UINT *a1,UINT *a2);
1.94 noro 6: int nd_dcomp;
7: NM _nm_free_list;
8: ND _nd_free_list;
9: ND_pairs _ndp_free_list;
1.32 noro 10:
1.117 ! noro 11: static int nd_nalg;
1.103 noro 12: #if 0
1.74 noro 13: static int ndv_alloc;
1.103 noro 14: #endif
1.87 noro 15: #if 1
1.69 noro 16: static int nd_f4_nsp=0x7fffffff;
1.87 noro 17: #else
18: static int nd_f4_nsp=50;
19: #endif
1.42 noro 20: static double nd_scale=2;
1.61 noro 21: static UINT **nd_bound;
1.42 noro 22: static struct order_spec *nd_ord;
23: static EPOS nd_epos;
1.43 noro 24: static BlockMask nd_blockmask;
1.42 noro 25: static int nd_nvar;
26: static int nd_isrlex;
27: static int nd_epw,nd_bpe,nd_wpd,nd_exporigin;
1.61 noro 28: static UINT nd_mask[32];
29: static UINT nd_mask0,nd_mask1;
1.42 noro 30:
1.20 noro 31: static NDV *nd_ps;
1.53 noro 32: static NDV *nd_ps_trace;
1.42 noro 33: static RHist *nd_psh;
34: static int nd_psn,nd_pslen;
35: static RHist *nd_red;
1.96 noro 36: static int *nd_work_vector;
37: static int **nd_matrix;
38: static int nd_matrix_len;
1.97 noro 39: static struct weight_or_block *nd_worb;
40: static int nd_worb_len;
1.42 noro 41: static int nd_found,nd_create,nd_notfirst;
42: static int nmv_adv;
1.77 noro 43: static int nd_demand;
1.1 noro 44:
1.114 noro 45: UINT *nd_det_compute_bound(NDV **dm,int n,int j);
46: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d);
1.117 ! noro 47: ND nd_pseudo_monic(int m,ND p);
1.114 noro 48:
1.1 noro 49: void nd_free_private_storage()
50: {
51: _nm_free_list = 0;
1.5 noro 52: _ndp_free_list = 0;
1.71 noro 53: #if 0
1.1 noro 54: GC_gcollect();
1.71 noro 55: #endif
1.1 noro 56: }
57:
58: void _NM_alloc()
59: {
60: NM p;
61: int i;
62:
1.11 noro 63: for ( i = 0; i < 1024; i++ ) {
1.61 noro 64: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 65: p->next = _nm_free_list; _nm_free_list = p;
66: }
67: }
68:
69: void _ND_alloc()
70: {
71: ND p;
72: int i;
73:
74: for ( i = 0; i < 1024; i++ ) {
75: p = (ND)GC_malloc(sizeof(struct oND));
76: p->body = (NM)_nd_free_list; _nd_free_list = p;
77: }
78: }
79:
80: void _NDP_alloc()
81: {
82: ND_pairs p;
83: int i;
84:
1.11 noro 85: for ( i = 0; i < 1024; i++ ) {
1.1 noro 86: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
1.61 noro 87: +(nd_wpd-1)*sizeof(UINT));
1.1 noro 88: p->next = _ndp_free_list; _ndp_free_list = p;
89: }
90: }
91:
1.30 noro 92: INLINE int nd_length(ND p)
1.1 noro 93: {
94: NM m;
95: int i;
96:
97: if ( !p )
98: return 0;
99: else {
100: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
101: return i;
102: }
103: }
104:
1.61 noro 105: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 106: {
1.61 noro 107: UINT u1,u2;
1.1 noro 108: int i,j;
109:
1.34 noro 110: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 111: #if USE_UNROLL
1.1 noro 112: switch ( nd_bpe ) {
1.62 noro 113: case 3:
114: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
115: u1 = d1[i]; u2 = d2[i];
116: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
117: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
118: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
119: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
120: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
121: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
122: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
123: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
124: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
125: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
126: }
127: return 1;
128: break;
1.1 noro 129: case 4:
1.41 noro 130: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 131: u1 = d1[i]; u2 = d2[i];
132: if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0;
1.62 noro 133: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
134: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
135: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
136: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
137: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
138: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
139: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
1.1 noro 140: }
141: return 1;
142: break;
143: case 6:
1.41 noro 144: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 145: u1 = d1[i]; u2 = d2[i];
146: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
1.62 noro 147: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
148: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
149: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
150: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
1.1 noro 151: }
152: return 1;
153: break;
154: case 8:
1.41 noro 155: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 156: u1 = d1[i]; u2 = d2[i];
157: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
1.62 noro 158: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
159: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
160: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
1.1 noro 161: }
162: return 1;
163: break;
164: case 16:
1.41 noro 165: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 166: u1 = d1[i]; u2 = d2[i];
167: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
1.62 noro 168: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
1.1 noro 169: }
170: return 1;
171: break;
172: case 32:
1.41 noro 173: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 174: if ( d1[i] < d2[i] ) return 0;
175: return 1;
176: break;
177: default:
1.41 noro 178: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 179: u1 = d1[i]; u2 = d2[i];
180: for ( j = 0; j < nd_epw; j++ )
181: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
182: }
183: return 1;
184: }
1.65 noro 185: #else
186: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
187: u1 = d1[i]; u2 = d2[i];
188: for ( j = 0; j < nd_epw; j++ )
189: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
190: }
191: return 1;
192: #endif
1.1 noro 193: }
194:
1.61 noro 195: /*
196: * If the current order is a block order,
197: * then the last block is length 1 and contains
198: * the homo variable. Otherwise, the original
199: * order is either 0 or 2.
200: */
201:
202: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 203: {
1.61 noro 204: int w,i,e,n,omask0;
205:
206: omask0 = (1<<obpe)-1;
207: n = nd_nvar-1;
208: ndl_zero(r);
209: for ( i = 0; i < n; i++ ) {
210: e = GET_EXP_OLD(d,i);
211: PUT_EXP(r,i,e);
212: }
213: w = TD(d);
214: PUT_EXP(r,nd_nvar-1,weight-w);
215: TD(r) = weight;
216: if ( nd_blockmask ) ndl_weight_mask(r);
217: }
218:
219: void ndl_dehomogenize(UINT *d)
220: {
221: UINT mask;
222: UINT h;
1.31 noro 223: int i,bits;
1.23 noro 224:
1.44 noro 225: if ( nd_blockmask ) {
226: h = GET_EXP(d,nd_nvar-1);
1.45 noro 227: XOR_EXP(d,nd_nvar-1,h);
1.44 noro 228: TD(d) -= h;
229: d[nd_exporigin-1] -= h;
230: } else {
231: if ( nd_isrlex ) {
232: if ( nd_bpe == 32 ) {
233: h = d[nd_exporigin];
234: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
235: d[i-1] = d[i];
236: d[i-1] = 0;
237: TD(d) -= h;
238: } else {
239: bits = nd_epw*nd_bpe;
240: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
241: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
242: for ( i = nd_exporigin; i < nd_wpd; i++ )
243: d[i] = ((d[i]<<nd_bpe)&mask)
244: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
245: TD(d) -= h;
246: }
1.45 noro 247: } else {
248: h = GET_EXP(d,nd_nvar-1);
249: XOR_EXP(d,nd_nvar-1,h);
250: TD(d) -= h;
251: }
1.44 noro 252: }
1.23 noro 253: }
254:
1.61 noro 255: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 256: {
1.61 noro 257: UINT t1,t2,u,u1,u2;
1.43 noro 258: int i,j,l;
1.1 noro 259:
1.65 noro 260: #if USE_UNROLL
1.1 noro 261: switch ( nd_bpe ) {
1.62 noro 262: case 3:
263: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
264: u1 = d1[i]; u2 = d2[i];
265: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
266: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
267: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
268: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
269: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
270: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
271: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
272: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
273: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
274: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
275: d[i] = u;
276: }
277: break;
1.1 noro 278: case 4:
1.41 noro 279: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 280: u1 = d1[i]; u2 = d2[i];
281: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
1.62 noro 282: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
283: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
284: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
285: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
286: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
287: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
288: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
1.1 noro 289: d[i] = u;
290: }
291: break;
292: case 6:
1.41 noro 293: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 294: u1 = d1[i]; u2 = d2[i];
295: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
1.62 noro 296: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
297: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
298: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
299: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
1.1 noro 300: d[i] = u;
301: }
302: break;
303: case 8:
1.41 noro 304: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 305: u1 = d1[i]; u2 = d2[i];
306: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
1.62 noro 307: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
308: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
309: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
1.1 noro 310: d[i] = u;
311: }
312: break;
313: case 16:
1.41 noro 314: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 315: u1 = d1[i]; u2 = d2[i];
316: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
1.62 noro 317: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
1.1 noro 318: d[i] = u;
319: }
320: break;
321: case 32:
1.41 noro 322: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 323: u1 = d1[i]; u2 = d2[i];
324: d[i] = u1>u2?u1:u2;
325: }
326: break;
327: default:
1.41 noro 328: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 329: u1 = d1[i]; u2 = d2[i];
330: for ( j = 0, u = 0; j < nd_epw; j++ ) {
331: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
332: }
333: d[i] = u;
334: }
335: break;
336: }
1.65 noro 337: #else
338: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
339: u1 = d1[i]; u2 = d2[i];
340: for ( j = 0, u = 0; j < nd_epw; j++ ) {
341: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
342: }
343: d[i] = u;
344: }
345: #endif
1.39 noro 346: TD(d) = ndl_weight(d);
1.61 noro 347: if ( nd_blockmask ) ndl_weight_mask(d);
1.57 noro 348: }
349:
1.61 noro 350: int ndl_weight(UINT *d)
1.1 noro 351: {
1.61 noro 352: UINT t,u;
1.1 noro 353: int i,j;
354:
1.60 noro 355: if ( current_dl_weight_vector )
356: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
357: u = GET_EXP(d,i);
358: t += MUL_WEIGHT(u,i);
359: }
360: else
361: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
362: u = d[i];
363: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
364: t += (u&nd_mask0);
365: }
1.1 noro 366: return t;
367: }
368:
1.61 noro 369: void ndl_weight_mask(UINT *d)
1.43 noro 370: {
1.61 noro 371: UINT t,u;
372: UINT *mask;
373: int i,j,k,l;
1.43 noro 374:
1.61 noro 375: l = nd_blockmask->n;
376: for ( k = 0; k < l; k++ ) {
377: mask = nd_blockmask->mask[k];
378: if ( current_dl_weight_vector )
379: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
380: u = GET_EXP_MASK(d,i,mask);
381: t += MUL_WEIGHT(u,i);
382: }
383: else
384: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
385: u = d[i]&mask[i];
386: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
387: t += (u&nd_mask0);
388: }
389: d[k+1] = t;
390: }
1.43 noro 391: }
392:
1.61 noro 393: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 394: {
395: int i;
396:
1.41 noro 397: d1 += nd_exporigin;
398: d2 += nd_exporigin;
399: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
1.1 noro 400: if ( *d1 > *d2 )
1.32 noro 401: return nd_isrlex ? -1 : 1;
1.1 noro 402: else if ( *d1 < *d2 )
1.32 noro 403: return nd_isrlex ? 1 : -1;
1.1 noro 404: return 0;
405: }
406:
1.61 noro 407: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 408: {
409: int i,l,j,ord_o,ord_l;
410: struct order_pair *op;
1.61 noro 411: UINT t1,t2,m;
412: UINT *mask;
1.43 noro 413:
414: l = nd_blockmask->n;
415: op = nd_blockmask->order_pair;
416: for ( j = 0; j < l; j++ ) {
417: mask = nd_blockmask->mask[j];
418: ord_o = op[j].order;
419: if ( ord_o < 2 )
1.44 noro 420: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
421: else if ( t1 < t2 ) return -1;
1.43 noro 422: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.44 noro 423: m = mask[i];
424: t1 = d1[i]&m;
425: t2 = d2[i]&m;
1.43 noro 426: if ( t1 > t2 )
427: return !ord_o ? -1 : 1;
428: else if ( t1 < t2 )
429: return !ord_o ? 1 : -1;
430: }
431: }
432: return 0;
433: }
434:
1.96 noro 435: int ndl_matrix_compare(UINT *d1,UINT *d2)
436: {
437: int i,j,s;
438: int *v;
439:
440: for ( j = 0; j < nd_nvar; j++ )
441: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
442: for ( i = 0; i < nd_matrix_len; i++ ) {
443: v = nd_matrix[i];
444: for ( j = 0, s = 0; j < nd_nvar; j++ )
445: s += v[j]*nd_work_vector[j];
446: if ( s > 0 ) return 1;
447: else if ( s < 0 ) return -1;
448: }
449: return 0;
450: }
451:
1.97 noro 452: int ndl_composite_compare(UINT *d1,UINT *d2)
453: {
454: int i,j,s,start,end,len,o;
455: int *v;
456: struct sparse_weight *sw;
457:
458: for ( j = 0; j < nd_nvar; j++ )
459: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
460: for ( i = 0; i < nd_worb_len; i++ ) {
461: len = nd_worb[i].length;
462: switch ( nd_worb[i].type ) {
463: case IS_DENSE_WEIGHT:
464: v = nd_worb[i].body.dense_weight;
465: for ( j = 0, s = 0; j < len; j++ )
466: s += v[j]*nd_work_vector[j];
467: if ( s > 0 ) return 1;
468: else if ( s < 0 ) return -1;
469: break;
470: case IS_SPARSE_WEIGHT:
471: sw = nd_worb[i].body.sparse_weight;
472: for ( j = 0, s = 0; j < len; j++ )
473: s += sw[j].value*nd_work_vector[sw[j].pos];
474: if ( s > 0 ) return 1;
475: else if ( s < 0 ) return -1;
476: break;
477: case IS_BLOCK:
478: o = nd_worb[i].body.block.order;
479: start = nd_worb[i].body.block.start;
480: switch ( o ) {
481: case 0:
482: end = start+len;
483: for ( j = start, s = 0; j < end; j++ )
484: s += MUL_WEIGHT(nd_work_vector[j],j);
485: if ( s > 0 ) return 1;
486: else if ( s < 0 ) return -1;
487: for ( j = end-1; j >= start; j-- )
488: if ( nd_work_vector[j] < 0 ) return 1;
489: else if ( nd_work_vector[j] > 0 ) return -1;
490: break;
491: case 1:
492: end = start+len;
493: for ( j = start, s = 0; j < end; j++ )
494: s += MUL_WEIGHT(nd_work_vector[j],j);
495: if ( s > 0 ) return 1;
496: else if ( s < 0 ) return -1;
497: for ( j = start; j < end; j++ )
498: if ( nd_work_vector[j] > 0 ) return 1;
499: else if ( nd_work_vector[j] < 0 ) return -1;
500: break;
501: case 2:
502: for ( j = start; j < end; j++ )
503: if ( nd_work_vector[j] > 0 ) return 1;
504: else if ( nd_work_vector[j] < 0 ) return -1;
505: break;
506: }
507: break;
508: }
509: }
510: return 0;
511: }
512:
1.58 noro 513: /* TDH -> WW -> TD-> RL */
514:
1.61 noro 515: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 516: {
517: int i,m,e1,e2;
518:
519: if ( TD(d1) > TD(d2) ) return 1;
520: else if ( TD(d1) < TD(d2) ) return -1;
521: m = nd_nvar>>1;
522: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
523: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
524: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
525: }
526: if ( e1 > e2 ) return 1;
527: else if ( e1 < e2 ) return -1;
528: return ndl_lex_compare(d1,d2);
529: }
530:
1.61 noro 531: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 532: {
533: int i;
534:
1.81 noro 535: switch ( nd_wpd ) {
536: case 2:
537: if ( TD(d2) != TD(d1) ) return 0;
538: if ( d2[1] != d1[1] ) return 0;
539: return 1;
540: break;
541: case 3:
542: if ( TD(d2) != TD(d1) ) return 0;
543: if ( d2[1] != d1[1] ) return 0;
544: if ( d2[2] != d1[2] ) return 0;
545: return 1;
546: break;
547: default:
548: for ( i = 0; i < nd_wpd; i++ )
549: if ( *d1++ != *d2++ ) return 0;
550: return 1;
551: break;
552: }
1.1 noro 553: }
554:
1.61 noro 555: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 556: {
557: int i;
558:
559: switch ( nd_wpd ) {
1.41 noro 560: case 2:
1.34 noro 561: TD(d2) = TD(d1);
562: d2[1] = d1[1];
1.6 noro 563: break;
1.41 noro 564: case 3:
1.34 noro 565: TD(d2) = TD(d1);
1.6 noro 566: d2[1] = d1[1];
1.34 noro 567: d2[2] = d1[2];
1.6 noro 568: break;
569: default:
1.41 noro 570: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 571: d2[i] = d1[i];
572: break;
573: }
574: }
575:
1.61 noro 576: INLINE void ndl_zero(UINT *d)
577: {
578: int i;
579: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
580: }
581:
582: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 583: {
584: int i;
585:
1.43 noro 586: #if 1
1.6 noro 587: switch ( nd_wpd ) {
1.41 noro 588: case 2:
589: TD(d) = TD(d1)+TD(d2);
1.34 noro 590: d[1] = d1[1]+d2[1];
1.6 noro 591: break;
1.41 noro 592: case 3:
593: TD(d) = TD(d1)+TD(d2);
1.6 noro 594: d[1] = d1[1]+d2[1];
1.34 noro 595: d[2] = d1[2]+d2[2];
1.6 noro 596: break;
597: default:
1.43 noro 598: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 599: break;
600: }
1.43 noro 601: #else
602: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
603: #endif
1.6 noro 604: }
605:
1.55 noro 606: /* d1 += d2 */
1.61 noro 607: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 608: {
609: int i;
610:
611: #if 1
612: switch ( nd_wpd ) {
613: case 2:
614: TD(d1) += TD(d2);
615: d1[1] += d2[1];
616: break;
617: case 3:
618: TD(d1) += TD(d2);
619: d1[1] += d2[1];
620: d1[2] += d2[2];
621: break;
622: default:
623: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
624: break;
625: }
626: #else
627: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
628: #endif
629: }
630:
1.61 noro 631: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 632: {
633: int i;
634:
1.43 noro 635: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 636: }
637:
1.61 noro 638: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 639: {
1.61 noro 640: UINT t1,t2,u,u1,u2;
1.1 noro 641: int i,j;
642:
1.65 noro 643: #if USE_UNROLL
1.1 noro 644: switch ( nd_bpe ) {
1.62 noro 645: case 3:
646: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
647: u1 = d1[i]; u2 = d2[i];
648: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
649: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
650: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
651: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
652: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
653: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
654: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
655: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
656: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
657: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
658: }
659: return 1;
660: break;
1.1 noro 661: case 4:
1.41 noro 662: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 663: u1 = d1[i]; u2 = d2[i];
664: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 665: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
666: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
667: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
668: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
669: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
670: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
671: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 672: }
673: return 1;
674: break;
675: case 6:
1.41 noro 676: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 677: u1 = d1[i]; u2 = d2[i];
678: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 679: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
680: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
681: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
682: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 683: }
684: return 1;
685: break;
686: case 8:
1.41 noro 687: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 688: u1 = d1[i]; u2 = d2[i];
689: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 690: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
691: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
692: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 693: }
694: return 1;
695: break;
696: case 16:
1.41 noro 697: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 698: u1 = d1[i]; u2 = d2[i];
699: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 700: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 701: }
702: return 1;
703: break;
704: case 32:
1.41 noro 705: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 706: if ( d1[i] && d2[i] ) return 0;
707: return 1;
708: break;
709: default:
1.41 noro 710: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 711: u1 = d1[i]; u2 = d2[i];
712: for ( j = 0; j < nd_epw; j++ ) {
713: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
714: u1 >>= nd_bpe; u2 >>= nd_bpe;
715: }
716: }
717: return 1;
718: break;
719: }
1.65 noro 720: #else
721: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
722: u1 = d1[i]; u2 = d2[i];
723: for ( j = 0; j < nd_epw; j++ ) {
724: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
725: u1 >>= nd_bpe; u2 >>= nd_bpe;
726: }
727: }
728: return 1;
729: #endif
1.1 noro 730: }
731:
1.114 noro 732: int ndl_check_bound(UINT *d1,UINT *d2)
1.1 noro 733: {
1.61 noro 734: UINT u2;
1.5 noro 735: int i,j,ind,k;
1.1 noro 736:
1.5 noro 737: ind = 0;
1.65 noro 738: #if USE_UNROLL
1.5 noro 739: switch ( nd_bpe ) {
1.62 noro 740: case 3:
741: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
742: u2 = d2[i];
743: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
744: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
745: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
746: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
747: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
748: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
749: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
750: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
751: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
752: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
753: }
754: return 0;
755: break;
1.5 noro 756: case 4:
1.41 noro 757: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 758: u2 = d2[i];
759: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
760: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
761: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
762: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
763: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
764: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
765: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
766: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
767: }
768: return 0;
769: break;
770: case 6:
1.41 noro 771: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 772: u2 = d2[i];
773: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
774: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
775: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
776: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
777: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
778: }
779: return 0;
780: break;
781: case 8:
1.41 noro 782: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 783: u2 = d2[i];
784: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
785: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
786: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
787: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
788: }
789: return 0;
790: break;
791: case 16:
1.41 noro 792: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 793: u2 = d2[i];
794: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
795: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
796: }
797: return 0;
798: break;
799: case 32:
1.41 noro 800: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 801: if ( d1[i]+d2[i]<d1[i] ) return 1;
802: return 0;
803: break;
804: default:
1.41 noro 805: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 806: u2 = d2[i];
807: k = (nd_epw-1)*nd_bpe;
808: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
809: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
810: }
811: return 0;
812: break;
813: }
1.65 noro 814: #else
815: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
816: u2 = d2[i];
817: k = (nd_epw-1)*nd_bpe;
818: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
819: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
820: }
821: return 0;
822: #endif
1.1 noro 823: }
824:
1.114 noro 825: int ndl_check_bound2(int index,UINT *d2)
826: {
827: return ndl_check_bound(nd_bound[index],d2);
828: }
829:
1.61 noro 830: INLINE int ndl_hash_value(UINT *d)
1.1 noro 831: {
832: int i;
833: int r;
834:
1.34 noro 835: r = 0;
1.41 noro 836: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 837: r = ((r<<16)+d[i])%REDTAB_LEN;
838: return r;
839: }
840:
1.63 noro 841: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 842: {
1.13 noro 843: RHist r;
1.6 noro 844: int d,k,i;
1.1 noro 845:
1.63 noro 846: d = ndl_hash_value(dg);
1.13 noro 847: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 848: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 849: if ( k > 0 ) nd_notfirst++;
850: nd_found++;
1.13 noro 851: return r->index;
1.1 noro 852: }
853: }
1.13 noro 854: if ( Reverse )
855: for ( i = nd_psn-1; i >= 0; i-- ) {
856: r = nd_psh[i];
1.34 noro 857: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 858: nd_create++;
1.34 noro 859: nd_append_red(dg,i);
1.13 noro 860: return i;
861: }
862: }
863: else
864: for ( i = 0; i < nd_psn; i++ ) {
865: r = nd_psh[i];
1.34 noro 866: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 867: nd_create++;
1.34 noro 868: nd_append_red(dg,i);
1.13 noro 869: return i;
870: }
1.1 noro 871: }
1.6 noro 872: return -1;
1.1 noro 873: }
874:
1.63 noro 875: ND nd_merge(ND p1,ND p2)
876: {
877: int n,c;
878: int t,can,td1,td2;
879: ND r;
880: NM m1,m2,mr0,mr,s;
881:
882: if ( !p1 ) return p2;
883: else if ( !p2 ) return p1;
884: else {
885: can = 0;
886: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
887: c = DL_COMPARE(DL(m1),DL(m2));
888: switch ( c ) {
889: case 0:
890: s = m1; m1 = NEXT(m1);
891: can++; NEXTNM2(mr0,mr,s);
892: s = m2; m2 = NEXT(m2); FREENM(s);
893: break;
894: case 1:
895: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
896: break;
897: case -1:
898: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
899: break;
900: }
901: }
902: if ( !mr0 )
903: if ( m1 ) mr0 = m1;
904: else if ( m2 ) mr0 = m2;
905: else return 0;
906: else if ( m1 ) NEXT(mr) = m1;
907: else if ( m2 ) NEXT(mr) = m2;
908: else NEXT(mr) = 0;
909: BDY(p1) = mr0;
910: SG(p1) = MAX(SG(p1),SG(p2));
911: LEN(p1) = LEN(p1)+LEN(p2)-can;
912: FREEND(p2);
913: return p1;
914: }
915: }
916:
1.31 noro 917: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 918: {
919: int n,c;
1.34 noro 920: int t,can,td1,td2;
1.1 noro 921: ND r;
922: NM m1,m2,mr0,mr,s;
923:
1.34 noro 924: if ( !p1 ) return p2;
925: else if ( !p2 ) return p1;
1.71 noro 926: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1.113 noro 927: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 928: else {
1.30 noro 929: can = 0;
1.1 noro 930: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 931: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 932: switch ( c ) {
933: case 0:
1.19 noro 934: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 935: if ( t < 0 ) t += mod;
1.1 noro 936: s = m1; m1 = NEXT(m1);
937: if ( t ) {
1.34 noro 938: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.17 noro 939: } else {
1.34 noro 940: can += 2; FREENM(s);
1.17 noro 941: }
942: s = m2; m2 = NEXT(m2); FREENM(s);
1.95 noro 943: break;
944: case 1:
945: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
946: break;
947: case -1:
948: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
949: break;
950: }
951: }
952: if ( !mr0 )
953: if ( m1 ) mr0 = m1;
954: else if ( m2 ) mr0 = m2;
955: else return 0;
956: else if ( m1 ) NEXT(mr) = m1;
957: else if ( m2 ) NEXT(mr) = m2;
958: else NEXT(mr) = 0;
959: BDY(p1) = mr0;
960: SG(p1) = MAX(SG(p1),SG(p2));
961: LEN(p1) = LEN(p1)+LEN(p2)-can;
962: FREEND(p2);
963: return p1;
964: }
965: }
966:
967: /* XXX on opteron, the inlined manipulation of destructive additon of
968: * two NM seems to make gcc optimizer get confused, so the part is
969: * done in a function.
970: */
971:
1.113 noro 972: int nm_destructive_add_q(NM *m1,NM *m2,NM *mr0,NM *mr)
1.95 noro 973: {
974: NM s;
1.113 noro 975: Q t;
1.95 noro 976: int can;
977:
1.113 noro 978: addq(CQ(*m1),CQ(*m2),&t);
1.95 noro 979: s = *m1; *m1 = NEXT(*m1);
980: if ( t ) {
1.113 noro 981: can = 1; NEXTNM2(*mr0,*mr,s); CQ(*mr) = (t);
1.95 noro 982: } else {
983: can = 2; FREENM(s);
984: }
985: s = *m2; *m2 = NEXT(*m2); FREENM(s);
986: return can;
987: }
988:
1.113 noro 989: ND nd_add_q(ND p1,ND p2)
1.95 noro 990: {
991: int n,c,can;
992: ND r;
993: NM m1,m2,mr0,mr,s;
1.113 noro 994: Q t;
1.95 noro 995:
996: if ( !p1 ) return p2;
997: else if ( !p2 ) return p1;
998: else {
999: can = 0;
1000: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1001: c = DL_COMPARE(DL(m1),DL(m2));
1002: switch ( c ) {
1003: case 0:
1004: #if defined(__x86_64__)
1.113 noro 1005: can += nm_destructive_add_q(&m1,&m2,&mr0,&mr);
1.95 noro 1006: #else
1.113 noro 1007: addq(CQ(m1),CQ(m2),&t);
1.95 noro 1008: s = m1; m1 = NEXT(m1);
1009: if ( t ) {
1.113 noro 1010: can++; NEXTNM2(mr0,mr,s); CQ(mr) = (t);
1.95 noro 1011: } else {
1012: can += 2; FREENM(s);
1013: }
1014: s = m2; m2 = NEXT(m2); FREENM(s);
1015: #endif
1.17 noro 1016: break;
1017: case 1:
1018: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1019: break;
1020: case -1:
1021: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1022: break;
1023: }
1024: }
1025: if ( !mr0 )
1.34 noro 1026: if ( m1 ) mr0 = m1;
1027: else if ( m2 ) mr0 = m2;
1028: else return 0;
1029: else if ( m1 ) NEXT(mr) = m1;
1030: else if ( m2 ) NEXT(mr) = m2;
1031: else NEXT(mr) = 0;
1.17 noro 1032: BDY(p1) = mr0;
1033: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1034: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 1035: FREEND(p2);
1036: return p1;
1037: }
1038: }
1039:
1.71 noro 1040: ND nd_add_sf(ND p1,ND p2)
1041: {
1042: int n,c,can;
1043: ND r;
1044: NM m1,m2,mr0,mr,s;
1045: int t;
1046:
1047: if ( !p1 ) return p2;
1048: else if ( !p2 ) return p1;
1049: else {
1050: can = 0;
1051: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1052: c = DL_COMPARE(DL(m1),DL(m2));
1053: switch ( c ) {
1054: case 0:
1055: t = _addsf(CM(m1),CM(m2));
1056: s = m1; m1 = NEXT(m1);
1057: if ( t ) {
1058: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1059: } else {
1060: can += 2; FREENM(s);
1061: }
1062: s = m2; m2 = NEXT(m2); FREENM(s);
1063: break;
1064: case 1:
1065: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1066: break;
1067: case -1:
1068: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1069: break;
1070: }
1071: }
1072: if ( !mr0 )
1073: if ( m1 ) mr0 = m1;
1074: else if ( m2 ) mr0 = m2;
1075: else return 0;
1076: else if ( m1 ) NEXT(mr) = m1;
1077: else if ( m2 ) NEXT(mr) = m2;
1078: else NEXT(mr) = 0;
1079: BDY(p1) = mr0;
1080: SG(p1) = MAX(SG(p1),SG(p2));
1081: LEN(p1) = LEN(p1)+LEN(p2)-can;
1082: FREEND(p2);
1083: return p1;
1084: }
1085: }
1086:
1.1 noro 1087: /* ret=1 : success, ret=0 : overflow */
1.113 noro 1088: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 1089: {
1.11 noro 1090: ND d;
1.1 noro 1091: NM m,mrd,tail;
1.7 noro 1092: NM mul;
1.10 noro 1093: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1094: int c,c1,c2,dummy;
1.17 noro 1095: RHist h;
1.11 noro 1096: NDV p,red;
1.113 noro 1097: Q cg,cred,gcd,tq,qq;
1.21 noro 1098: double hmag;
1.1 noro 1099:
1.113 noro 1100: if ( dn ) {
1101: if ( mod )
1102: dn->m = 1;
1103: else
1104: dn->z = ONE;
1105: }
1.1 noro 1106: if ( !g ) {
1107: *rp = 0;
1108: return 1;
1109: }
1.113 noro 1110: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1111:
1.14 noro 1112: sugar0 = sugar = SG(g);
1.1 noro 1113: n = NV(g);
1.61 noro 1114: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 1115: for ( d = 0; g; ) {
1.63 noro 1116: index = ndl_find_reducer(HDL(g));
1.6 noro 1117: if ( index >= 0 ) {
1.17 noro 1118: h = nd_psh[index];
1119: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1120: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1121: nd_free(g); nd_free(d);
1122: return 0;
1123: }
1.77 noro 1124: if ( nd_demand )
1125: p = ndv_load(index);
1126: else
1127: p = ps[index];
1.71 noro 1128: if ( mod == -1 )
1129: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1130: else if ( mod ) {
1.19 noro 1131: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1132: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1.16 noro 1133: } else {
1.113 noro 1134: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1135: chsgnq(cg,&CQ(mul));
1136: nd_mul_c_q(d,cred); nd_mul_c_q(g,cred);
1.69 noro 1137: if ( dn ) {
1.113 noro 1138: mulq(dn->z,cred,&tq); dn->z = tq;
1.69 noro 1139: }
1.16 noro 1140: }
1.55 noro 1141: g = nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.34 noro 1142: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.113 noro 1143: if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) {
1144: tq = HCQ(g);
1.21 noro 1145: nd_removecont2(d,g);
1.69 noro 1146: if ( dn ) {
1.113 noro 1147: divq(tq,HCQ(g),&qq); divq(dn->z,qq,&tq); dn->z = tq;
1.69 noro 1148: }
1.113 noro 1149: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1150: }
1.1 noro 1151: } else if ( !full ) {
1152: *rp = g;
1153: return 1;
1154: } else {
1155: m = BDY(g);
1156: if ( NEXT(m) ) {
1.34 noro 1157: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1158: } else {
1159: FREEND(g); g = 0;
1160: }
1161: if ( d ) {
1.34 noro 1162: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1163: } else {
1.34 noro 1164: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1165: }
1166: }
1167: }
1.34 noro 1168: if ( d ) SG(d) = sugar;
1.1 noro 1169: *rp = d;
1170: return 1;
1171: }
1.28 noro 1172:
1.53 noro 1173: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1174: {
1175: int hindex,index;
1176: NDV p;
1177: ND u,d,red;
1178: NODE l;
1.31 noro 1179: NM mul,m,mrd,tail;
1.25 noro 1180: int sugar,psugar,n,h_reducible;
1181: PGeoBucket bucket;
1182: int c,c1,c2;
1.113 noro 1183: Q cg,cred,gcd,zzz;
1.25 noro 1184: RHist h;
1.28 noro 1185: double hmag,gmag;
1.77 noro 1186: int count = 0;
1187: int hcount = 0;
1.25 noro 1188:
1189: if ( !g ) {
1190: *rp = 0;
1191: return 1;
1192: }
1193: sugar = SG(g);
1194: n = NV(g);
1.113 noro 1195: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1196: bucket = create_pbucket();
1.31 noro 1197: add_pbucket(mod,bucket,g);
1.25 noro 1198: d = 0;
1.61 noro 1199: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1200: while ( 1 ) {
1.113 noro 1201: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1202: if ( hindex < 0 ) {
1.77 noro 1203: if ( DP_Print > 3 ) printf("(%d %d)",count,hcount);
1.34 noro 1204: if ( d ) SG(d) = sugar;
1.25 noro 1205: *rp = d;
1206: return 1;
1207: }
1208: g = bucket->body[hindex];
1.63 noro 1209: index = ndl_find_reducer(HDL(g));
1.25 noro 1210: if ( index >= 0 ) {
1.77 noro 1211: count++;
1212: if ( !d ) hcount++;
1.25 noro 1213: h = nd_psh[index];
1214: ndl_sub(HDL(g),DL(h),DL(mul));
1215: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1216: nd_free(d);
1.25 noro 1217: free_pbucket(bucket);
1218: *rp = 0;
1219: return 0;
1220: }
1.53 noro 1221: p = ps[index];
1.71 noro 1222: if ( mod == -1 )
1223: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1224: else if ( mod ) {
1.25 noro 1225: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1226: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1227: } else {
1.113 noro 1228: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1229: chsgnq(cg,&CQ(mul));
1230: nd_mul_c_q(d,cred);
1231: mulq_pbucket(bucket,cred);
1.26 noro 1232: g = bucket->body[hindex];
1.113 noro 1233: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1234: }
1.55 noro 1235: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1236: bucket->body[hindex] = nd_remove_head(g);
1237: red = nd_remove_head(red);
1.31 noro 1238: add_pbucket(mod,bucket,red);
1.34 noro 1239: psugar = SG(p)+TD(DL(mul));
1240: sugar = MAX(sugar,psugar);
1.28 noro 1241: if ( !mod && hmag && (gmag > hmag) ) {
1242: g = normalize_pbucket(mod,bucket);
1243: if ( !g ) {
1.34 noro 1244: if ( d ) SG(d) = sugar;
1.28 noro 1245: *rp = d;
1246: return 1;
1247: }
1248: nd_removecont2(d,g);
1.113 noro 1249: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1250: add_pbucket(mod,bucket,g);
1.28 noro 1251: }
1.25 noro 1252: } else if ( !full ) {
1253: g = normalize_pbucket(mod,bucket);
1.34 noro 1254: if ( g ) SG(g) = sugar;
1.25 noro 1255: *rp = g;
1256: return 1;
1257: } else {
1258: m = BDY(g);
1259: if ( NEXT(m) ) {
1.34 noro 1260: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1261: } else {
1262: FREEND(g); g = 0;
1263: }
1264: bucket->body[hindex] = g;
1265: NEXT(m) = 0;
1266: if ( d ) {
1.34 noro 1267: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1268: } else {
1.34 noro 1269: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1270: }
1271: }
1272: }
1273: }
1.27 noro 1274:
1.61 noro 1275: /* input : list of NDV, cand : list of NDV */
1.28 noro 1276:
1.61 noro 1277: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1278: {
1279: int n,i,stat;
1280: ND nf,d;
1.61 noro 1281: NDV r;
1.45 noro 1282: NODE t,s;
1.113 noro 1283: union oNDC dn;
1.45 noro 1284:
1.61 noro 1285: ndv_setup(0,0,cand);
1.31 noro 1286: n = length(cand);
1.28 noro 1287:
1288: /* membercheck : list is a subset of Id(cand) ? */
1289: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1290: again:
1.61 noro 1291: if ( nd_bpe > obpe )
1292: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1293: else
1294: r = (NDV)BDY(t);
1295: d = ndvtond(0,r);
1.69 noro 1296: stat = nd_nf(0,d,nd_ps,0,0,&nf);
1.45 noro 1297: if ( !stat ) {
1.103 noro 1298: nd_reconstruct(0,0);
1.45 noro 1299: goto again;
1300: } else if ( nf ) return 0;
1.71 noro 1301: if ( DP_Print ) { printf("."); fflush(stdout); }
1.28 noro 1302: }
1.71 noro 1303: if ( DP_Print ) { printf("\n"); }
1.28 noro 1304: /* gbcheck : cand is a GB of Id(cand) ? */
1.77 noro 1305: if ( !nd_gb(0,0,1) ) return 0;
1.28 noro 1306: /* XXX */
1.23 noro 1307: return 1;
1308: }
1.1 noro 1309:
1310: ND nd_remove_head(ND p)
1311: {
1312: NM m;
1313:
1314: m = BDY(p);
1315: if ( !NEXT(m) ) {
1.34 noro 1316: FREEND(p); p = 0;
1.31 noro 1317: } else {
1.34 noro 1318: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1319: }
1.1 noro 1320: FREENM(m);
1321: return p;
1322: }
1323:
1.69 noro 1324: ND nd_separate_head(ND p,ND *head)
1325: {
1326: NM m,m0;
1327: ND r;
1328:
1329: m = BDY(p);
1330: if ( !NEXT(m) ) {
1331: *head = p; p = 0;
1332: } else {
1333: m0 = m;
1334: BDY(p) = NEXT(m); LEN(p)--;
1335: NEXT(m0) = 0;
1336: MKND(NV(p),m0,1,r);
1337: *head = r;
1338: }
1339: return p;
1340: }
1341:
1.1 noro 1342: PGeoBucket create_pbucket()
1343: {
1344: PGeoBucket g;
1345:
1346: g = CALLOC(1,sizeof(struct oPGeoBucket));
1347: g->m = -1;
1348: return g;
1349: }
1350:
1.25 noro 1351: void free_pbucket(PGeoBucket b) {
1352: int i;
1353:
1.26 noro 1354: for ( i = 0; i <= b->m; i++ )
1.25 noro 1355: if ( b->body[i] ) {
1356: nd_free(b->body[i]);
1357: b->body[i] = 0;
1358: }
1359: GC_free(b);
1360: }
1361:
1.63 noro 1362: void add_pbucket_symbolic(PGeoBucket g,ND d)
1363: {
1364: int l,i,k,m;
1365:
1366: if ( !d )
1367: return;
1368: l = LEN(d);
1369: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1370: /* 2^(k-1) < l <= 2^k (=m) */
1371: d = nd_merge(g->body[k],d);
1372: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1373: g->body[k] = 0;
1374: d = nd_merge(g->body[k+1],d);
1375: }
1376: g->body[k] = d;
1377: g->m = MAX(g->m,k);
1378: }
1379:
1.31 noro 1380: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1381: {
1.31 noro 1382: int l,i,k,m;
1.1 noro 1383:
1.31 noro 1384: if ( !d )
1385: return;
1386: l = LEN(d);
1.29 noro 1387: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1388: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1389: d = nd_add(mod,g->body[k],d);
1390: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1391: g->body[k] = 0;
1.31 noro 1392: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1393: }
1394: g->body[k] = d;
1395: g->m = MAX(g->m,k);
1396: }
1397:
1.113 noro 1398: void mulq_pbucket(PGeoBucket g,Q c)
1.26 noro 1399: {
1400: int k;
1401:
1402: for ( k = 0; k <= g->m; k++ )
1.113 noro 1403: nd_mul_c_q(g->body[k],c);
1.26 noro 1404: }
1405:
1.63 noro 1406: NM remove_head_pbucket_symbolic(PGeoBucket g)
1407: {
1408: int j,i,k,c;
1409: NM head;
1410:
1411: k = g->m;
1412: j = -1;
1413: for ( i = 0; i <= k; i++ ) {
1414: if ( !g->body[i] ) continue;
1415: if ( j < 0 ) j = i;
1416: else {
1417: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1418: if ( c > 0 )
1419: j = i;
1420: else if ( c == 0 )
1421: g->body[i] = nd_remove_head(g->body[i]);
1422: }
1423: }
1424: if ( j < 0 ) return 0;
1425: else {
1426: head = BDY(g->body[j]);
1427: if ( !NEXT(head) ) {
1428: FREEND(g->body[j]);
1429: g->body[j] = 0;
1430: } else {
1431: BDY(g->body[j]) = NEXT(head);
1432: LEN(g->body[j])--;
1433: }
1434: return head;
1435: }
1436: }
1437:
1.19 noro 1438: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1439: {
1440: int j,i,c,k,nv,sum;
1.61 noro 1441: UINT *di,*dj;
1.1 noro 1442: ND gi,gj;
1443:
1444: k = g->m;
1445: while ( 1 ) {
1446: j = -1;
1447: for ( i = 0; i <= k; i++ ) {
1448: if ( !(gi = g->body[i]) )
1449: continue;
1450: if ( j < 0 ) {
1451: j = i;
1452: gj = g->body[j];
1453: dj = HDL(gj);
1.14 noro 1454: sum = HCM(gj);
1.1 noro 1455: } else {
1.34 noro 1456: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1457: if ( c > 0 ) {
1.34 noro 1458: if ( sum ) HCM(gj) = sum;
1459: else g->body[j] = nd_remove_head(gj);
1.1 noro 1460: j = i;
1461: gj = g->body[j];
1462: dj = HDL(gj);
1.14 noro 1463: sum = HCM(gj);
1.1 noro 1464: } else if ( c == 0 ) {
1.71 noro 1465: if ( mod == -1 )
1466: sum = _addsf(sum,HCM(gi));
1467: else {
1468: sum = sum+HCM(gi)-mod;
1469: if ( sum < 0 ) sum += mod;
1470: }
1.1 noro 1471: g->body[i] = nd_remove_head(gi);
1472: }
1473: }
1474: }
1.34 noro 1475: if ( j < 0 ) return -1;
1.1 noro 1476: else if ( sum ) {
1.14 noro 1477: HCM(gj) = sum;
1.26 noro 1478: return j;
1.31 noro 1479: } else
1.26 noro 1480: g->body[j] = nd_remove_head(gj);
1481: }
1482: }
1483:
1.113 noro 1484: int head_pbucket_q(PGeoBucket g)
1.26 noro 1485: {
1486: int j,i,c,k,nv;
1.113 noro 1487: Q sum,t;
1.26 noro 1488: ND gi,gj;
1489:
1490: k = g->m;
1491: while ( 1 ) {
1492: j = -1;
1493: for ( i = 0; i <= k; i++ ) {
1.34 noro 1494: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1495: if ( j < 0 ) {
1496: j = i;
1497: gj = g->body[j];
1.113 noro 1498: sum = HCQ(gj);
1.26 noro 1499: } else {
1500: nv = NV(gi);
1.34 noro 1501: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1502: if ( c > 0 ) {
1.113 noro 1503: if ( sum ) HCQ(gj) = sum;
1.34 noro 1504: else g->body[j] = nd_remove_head(gj);
1.26 noro 1505: j = i;
1506: gj = g->body[j];
1.113 noro 1507: sum = HCQ(gj);
1.26 noro 1508: } else if ( c == 0 ) {
1.113 noro 1509: addq(sum,HCQ(gi),&t);
1510: sum = t;
1.26 noro 1511: g->body[i] = nd_remove_head(gi);
1512: }
1513: }
1514: }
1.34 noro 1515: if ( j < 0 ) return -1;
1.26 noro 1516: else if ( sum ) {
1.113 noro 1517: HCQ(gj) = sum;
1.1 noro 1518: return j;
1.31 noro 1519: } else
1.1 noro 1520: g->body[j] = nd_remove_head(gj);
1521: }
1522: }
1523:
1.25 noro 1524: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1525: {
1.31 noro 1526: int i;
1.1 noro 1527: ND r,t;
1528:
1529: r = 0;
1.28 noro 1530: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1531: r = nd_add(mod,r,g->body[i]);
1.28 noro 1532: g->body[i] = 0;
1533: }
1534: g->m = -1;
1.1 noro 1535: return r;
1536: }
1537:
1.92 noro 1538: void do_diagonalize(int sugar,int m)
1539: {
1540: int i,nh,stat;
1541: NODE r,g,t;
1542: ND h,nf,s,head;
1543: NDV nfv;
1.113 noro 1544: Q q,num,den;
1545: union oNDC dn;
1.92 noro 1546:
1547: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1548: if ( nd_demand )
1549: nfv = ndv_load(i);
1550: else
1551: nfv = nd_ps[i];
1552: s = ndvtond(m,nfv);
1553: s = nd_separate_head(s,&head);
1554: nd_nf(m,s,nd_ps,1,&dn,&nf);
1555: if ( !m ) {
1.113 noro 1556: NTOQ(NM(dn.z),SGN(dn.z),num);
1557: mulq(HCQ(head),num,&q); HCQ(head) = q;
1558: if ( DN(dn.z) ) {
1559: NTOQ(DN(dn.z),1,den);
1560: nd_mul_c_q(nf,den);
1561: }
1.92 noro 1562: }
1563: nf = nd_add(m,head,nf);
1564: ndv_free(nfv);
1565: nd_removecont(m,nf);
1566: nfv = ndtondv(m,nf);
1567: nd_free(nf);
1568: nd_bound[i] = ndv_compute_bound(nfv);
1569: if ( nd_demand ) {
1570: ndv_save(nfv,i);
1571: ndv_free(nfv);
1572: } else
1573: nd_ps[i] = nfv;
1574: }
1575: }
1576:
1.27 noro 1577: /* return value = 0 => input is not a GB */
1578:
1.77 noro 1579: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1580: {
1581: int i,nh,sugar,stat;
1.23 noro 1582: NODE r,g,t;
1.1 noro 1583: ND_pairs d;
1584: ND_pairs l;
1.117 ! noro 1585: ND h,nf,s,head,nf1;
1.63 noro 1586: NDV nfv;
1.113 noro 1587: Q q,num,den;
1588: union oNDC dn;
1.1 noro 1589:
1.23 noro 1590: g = 0; d = 0;
1591: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1592: d = update_pairs(d,g,i);
1593: g = update_base(g,i);
1594: }
1595: sugar = 0;
1596: while ( d ) {
1597: again:
1598: l = nd_minp(d,&d);
1.14 noro 1599: if ( SG(l) != sugar ) {
1.92 noro 1600: if ( ishomo ) do_diagonalize(sugar,m);
1601:
1.14 noro 1602: sugar = SG(l);
1.71 noro 1603: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1604: }
1.53 noro 1605: stat = nd_sp(m,0,l,&h);
1.1 noro 1606: if ( !stat ) {
1607: NEXT(l) = d; d = l;
1.103 noro 1608: d = nd_reconstruct(0,d);
1.1 noro 1609: goto again;
1610: }
1.41 noro 1611: #if USE_GEOBUCKET
1.69 noro 1612: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1613: #else
1.69 noro 1614: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1615: #endif
1.1 noro 1616: if ( !stat ) {
1617: NEXT(l) = d; d = l;
1.103 noro 1618: d = nd_reconstruct(0,d);
1.1 noro 1619: goto again;
1620: } else if ( nf ) {
1.27 noro 1621: if ( checkonly ) return 0;
1.71 noro 1622: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1623: nd_removecont(m,nf);
1.117 ! noro 1624: if ( nd_nalg ) {
! 1625: nf1 = nd_pseudo_monic(m,nf); nd_free(nf);
! 1626: stat = nd_nf(m,nf1,nd_ps,1,0,&nf);
! 1627: if ( stat ) {
! 1628: NEXT(l) = d; d = l;
! 1629: d = nd_reconstruct(0,d);
! 1630: goto again;
! 1631: }
! 1632: }
1.63 noro 1633: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1634: nh = ndv_newps(m,nfv,0);
1.1 noro 1635: d = update_pairs(d,g,nh);
1636: g = update_base(g,nh);
1637: FREENDP(l);
1638: } else {
1.71 noro 1639: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1640: FREENDP(l);
1641: }
1642: }
1.77 noro 1643: if ( nd_demand )
1644: for ( t = g; t; t = NEXT(t) )
1645: BDY(t) = (pointer)ndv_load((int)BDY(t));
1646: else
1647: for ( t = g; t; t = NEXT(t) )
1648: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1649: return g;
1650: }
1651:
1.92 noro 1652: void do_diagonalize_trace(int sugar,int m)
1.91 noro 1653: {
1654: int i,nh,stat;
1655: NODE r,g,t;
1656: ND h,nf,nfq,s,head;
1657: NDV nfv,nfqv;
1.113 noro 1658: Q q,den,num;
1659: union oNDC dn;
1.91 noro 1660:
1661: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1662: /* for nd_ps */
1663: s = ndvtond(m,nd_ps[i]);
1664: s = nd_separate_head(s,&head);
1665: nd_nf_pbucket(m,s,nd_ps,1,&nf);
1666: nf = nd_add(m,head,nf);
1667: ndv_free(nd_ps[i]);
1668: nd_ps[i] = ndtondv(m,nf);
1669: nd_free(nf);
1670:
1671: /* for nd_ps_trace */
1672: if ( nd_demand )
1673: nfv = ndv_load(i);
1674: else
1675: nfv = nd_ps_trace[i];
1676: s = ndvtond(0,nfv);
1677: s = nd_separate_head(s,&head);
1678: nd_nf(0,s,nd_ps_trace,1,&dn,&nf);
1.113 noro 1679: NTOQ(NM(dn.z),SGN(dn.z),num);
1680: mulq(HCQ(head),num,&q); HCQ(head) = q;
1681: if ( DN(dn.z) ) {
1682: NTOQ(DN(dn.z),1,den);
1683: nd_mul_c_q(nf,den);
1684: }
1.91 noro 1685: nf = nd_add(0,head,nf);
1686: ndv_free(nfv);
1687: nd_removecont(0,nf);
1688: nfv = ndtondv(0,nf);
1689: nd_free(nf);
1690: nd_bound[i] = ndv_compute_bound(nfv);
1691: if ( nd_demand ) {
1692: ndv_save(nfv,i);
1693: ndv_free(nfv);
1694: } else
1695: nd_ps_trace[i] = nfv;
1696: }
1697: }
1698:
1.82 noro 1699: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 1700: {
1701: int i,nh,sugar,stat;
1.23 noro 1702: NODE r,g,t;
1.20 noro 1703: ND_pairs d;
1704: ND_pairs l;
1.82 noro 1705: ND h,nf,nfq,s,head;
1.63 noro 1706: NDV nfv,nfqv;
1.113 noro 1707: Q q,den,num;
1708: union oNDC dn;
1.20 noro 1709:
1.23 noro 1710: g = 0; d = 0;
1711: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1712: d = update_pairs(d,g,i);
1713: g = update_base(g,i);
1714: }
1715: sugar = 0;
1716: while ( d ) {
1717: again:
1718: l = nd_minp(d,&d);
1719: if ( SG(l) != sugar ) {
1.93 noro 1720: if ( ishomo ) do_diagonalize_trace(sugar,m);
1.20 noro 1721: sugar = SG(l);
1.71 noro 1722: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1723: }
1.53 noro 1724: stat = nd_sp(m,0,l,&h);
1.20 noro 1725: if ( !stat ) {
1726: NEXT(l) = d; d = l;
1.103 noro 1727: d = nd_reconstruct(1,d);
1.20 noro 1728: goto again;
1729: }
1.41 noro 1730: #if USE_GEOBUCKET
1.53 noro 1731: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1732: #else
1.69 noro 1733: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1734: #endif
1.20 noro 1735: if ( !stat ) {
1736: NEXT(l) = d; d = l;
1.103 noro 1737: d = nd_reconstruct(1,d);
1.20 noro 1738: goto again;
1739: } else if ( nf ) {
1.77 noro 1740: if ( nd_demand ) {
1741: nfqv = ndv_load(nd_psn);
1742: nfq = ndvtond(0,nfqv);
1743: } else
1744: nfq = 0;
1745: if ( !nfq ) {
1746: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1747: NEXT(l) = d; d = l;
1.103 noro 1748: d = nd_reconstruct(1,d);
1.77 noro 1749: goto again;
1750: }
1.70 noro 1751: }
1.20 noro 1752: if ( nfq ) {
1.70 noro 1753: /* m|HC(nfq) => failure */
1.113 noro 1754: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1.63 noro 1755:
1.71 noro 1756: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1757: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1758: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1.77 noro 1759: nh = ndv_newps(0,nfv,nfqv);
1.20 noro 1760: d = update_pairs(d,g,nh);
1761: g = update_base(g,nh);
1762: } else {
1.71 noro 1763: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1764: }
1765: } else {
1.71 noro 1766: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1767: }
1768: FREENDP(l);
1769: }
1.77 noro 1770: if ( nd_demand )
1771: for ( t = g; t; t = NEXT(t) )
1772: BDY(t) = (pointer)ndv_load((int)BDY(t));
1773: else
1774: for ( t = g; t; t = NEXT(t) )
1775: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.20 noro 1776: return g;
1777: }
1778:
1.23 noro 1779: int ndv_compare(NDV *p1,NDV *p2)
1780: {
1.34 noro 1781: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1782: }
1783:
1784: int ndv_compare_rev(NDV *p1,NDV *p2)
1785: {
1.34 noro 1786: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1787: }
1788:
1.61 noro 1789: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1790: {
1.69 noro 1791: int i,n,stat;
1792: ND nf,g,head;
1.23 noro 1793: NODE t,a0,a;
1.113 noro 1794: union oNDC dn;
1.69 noro 1795: NDV *w;
1.113 noro 1796: Q q,num,den;
1.23 noro 1797:
1.69 noro 1798: n = length(f);
1799: #if 0
1800: w = (NDV *)ALLOCA(n*sizeof(NDV));
1801: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1802: qsort(w,n,sizeof(NDV),
1803: (int (*)(const void *,const void *))ndv_compare);
1804: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1805: #endif
1806: ndv_setup(m,0,f);
1807: for ( i = 0; i < n; ) {
1808: g = ndvtond(m,nd_ps[i]);
1809: g = nd_separate_head(g,&head);
1810: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1811: if ( !stat )
1.103 noro 1812: nd_reconstruct(0,0);
1.69 noro 1813: else {
1.71 noro 1814: if ( DP_Print ) { printf("."); fflush(stdout); }
1.85 noro 1815: if ( !m ) {
1.113 noro 1816: NTOQ(NM(dn.z),SGN(dn.z),num);
1817: mulq(HCQ(head),num,&q); HCQ(head) = q;
1818: if ( DN(dn.z) ) {
1819: NTOQ(DN(dn.z),1,den);
1820: nd_mul_c_q(nf,den);
1821: }
1.85 noro 1822: }
1.69 noro 1823: nf = nd_add(m,head,nf);
1824: ndv_free(nd_ps[i]);
1.24 noro 1825: nd_removecont(m,nf);
1.69 noro 1826: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1827: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1828: i++;
1.23 noro 1829: }
1830: }
1.71 noro 1831: if ( DP_Print ) { printf("\n"); }
1.23 noro 1832: for ( a0 = 0, i = 0; i < n; i++ ) {
1833: NEXTNODE(a0,a);
1.69 noro 1834: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1835: }
1836: NEXT(a) = 0;
1837: return a0;
1838: }
1839:
1.1 noro 1840: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1841: {
1842: ND_pairs d1,nd,cur,head,prev,remove;
1843:
1844: if ( !g ) return d;
1845: d = crit_B(d,t);
1846: d1 = nd_newpairs(g,t);
1847: d1 = crit_M(d1);
1848: d1 = crit_F(d1);
1.55 noro 1849: if ( do_weyl )
1850: head = d1;
1851: else {
1852: prev = 0; cur = head = d1;
1853: while ( cur ) {
1854: if ( crit_2( cur->i1,cur->i2 ) ) {
1855: remove = cur;
1856: if ( !prev ) head = cur = NEXT(cur);
1857: else cur = NEXT(prev) = NEXT(cur);
1858: FREENDP(remove);
1859: } else {
1860: prev = cur; cur = NEXT(cur);
1861: }
1.1 noro 1862: }
1863: }
1864: if ( !d )
1865: return head;
1866: else {
1867: nd = d;
1.34 noro 1868: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1869: NEXT(nd) = head;
1870: return d;
1871: }
1872: }
1873:
1874: ND_pairs nd_newpairs( NODE g, int t )
1875: {
1876: NODE h;
1.61 noro 1877: UINT *dl;
1.34 noro 1878: int ts,s;
1.1 noro 1879: ND_pairs r,r0;
1880:
1.20 noro 1881: dl = DL(nd_psh[t]);
1.34 noro 1882: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1883: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1884: NEXTND_pairs(r0,r);
1885: r->i1 = (int)BDY(h);
1886: r->i2 = t;
1.20 noro 1887: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1888: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1889: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1890: }
1891: NEXT(r) = 0;
1892: return r0;
1893: }
1894:
1895: ND_pairs crit_B( ND_pairs d, int s )
1896: {
1897: ND_pairs cur,head,prev,remove;
1.61 noro 1898: UINT *t,*tl,*lcm;
1.1 noro 1899: int td,tdl;
1900:
1901: if ( !d ) return 0;
1.20 noro 1902: t = DL(nd_psh[s]);
1.1 noro 1903: prev = 0;
1904: head = cur = d;
1.61 noro 1905: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1906: while ( cur ) {
1907: tl = cur->lcm;
1908: if ( ndl_reducible(tl,t)
1.20 noro 1909: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1910: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1911: remove = cur;
1912: if ( !prev ) {
1913: head = cur = NEXT(cur);
1914: } else {
1915: cur = NEXT(prev) = NEXT(cur);
1916: }
1917: FREENDP(remove);
1918: } else {
1.34 noro 1919: prev = cur; cur = NEXT(cur);
1.1 noro 1920: }
1921: }
1922: return head;
1923: }
1924:
1925: ND_pairs crit_M( ND_pairs d1 )
1926: {
1927: ND_pairs e,d2,d3,dd,p;
1.61 noro 1928: UINT *id,*jd;
1.1 noro 1929:
1930: for ( dd = 0, e = d1; e; e = d3 ) {
1931: if ( !(d2 = NEXT(e)) ) {
1932: NEXT(e) = dd;
1933: return e;
1934: }
1.34 noro 1935: id = LCM(e);
1.1 noro 1936: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 1937: p = NEXT(d2);
1938: jd = LCM(d2);
1939: if ( ndl_equal(jd,id) )
1940: ;
1941: else if ( TD(jd) > TD(id) )
1.1 noro 1942: if ( ndl_reducible(jd,id) ) continue;
1943: else ;
1.34 noro 1944: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 1945: NEXT(d2) = d3;
1946: d3 = d2;
1947: }
1948: NEXT(e) = dd;
1949: dd = e;
1950: continue;
1951: /**/
1952: delit: NEXT(d2) = d3;
1953: d3 = d2;
1954: for ( ; p; p = d2 ) {
1955: d2 = NEXT(p);
1956: NEXT(p) = d3;
1957: d3 = p;
1958: }
1959: FREENDP(e);
1960: }
1961: return dd;
1962: }
1963:
1964: ND_pairs crit_F( ND_pairs d1 )
1965: {
1966: ND_pairs rest, head,remove;
1967: ND_pairs last, p, r, w;
1968: int s;
1969:
1970: for ( head = last = 0, p = d1; NEXT(p); ) {
1971: r = w = equivalent_pairs(p,&rest);
1.14 noro 1972: s = SG(r);
1.1 noro 1973: w = NEXT(w);
1974: while ( w ) {
1975: if ( crit_2(w->i1,w->i2) ) {
1976: r = w;
1977: w = NEXT(w);
1978: while ( w ) {
1979: remove = w;
1980: w = NEXT(w);
1981: FREENDP(remove);
1982: }
1983: break;
1.14 noro 1984: } else if ( SG(w) < s ) {
1.1 noro 1985: FREENDP(r);
1986: r = w;
1.14 noro 1987: s = SG(r);
1.1 noro 1988: w = NEXT(w);
1989: } else {
1990: remove = w;
1991: w = NEXT(w);
1992: FREENDP(remove);
1993: }
1994: }
1995: if ( last ) NEXT(last) = r;
1996: else head = r;
1997: NEXT(last = r) = 0;
1998: p = rest;
1999: if ( !p ) return head;
2000: }
2001: if ( !last ) return p;
2002: NEXT(last) = p;
2003: return head;
2004: }
2005:
2006: int crit_2( int dp1, int dp2 )
2007: {
1.20 noro 2008: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2009: }
2010:
1.40 noro 2011: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2012: {
2013: ND_pairs w,p,r,s;
1.61 noro 2014: UINT *d;
1.1 noro 2015:
2016: w = d1;
1.34 noro 2017: d = LCM(w);
1.1 noro 2018: s = NEXT(w);
2019: NEXT(w) = 0;
2020: for ( r = 0; s; s = p ) {
2021: p = NEXT(s);
1.34 noro 2022: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2023: NEXT(s) = w; w = s;
1.1 noro 2024: } else {
1.39 noro 2025: NEXT(s) = r; r = s;
1.1 noro 2026: }
2027: }
2028: *prest = r;
2029: return w;
2030: }
2031:
2032: NODE update_base(NODE nd,int ndp)
2033: {
1.61 noro 2034: UINT *dl, *dln;
1.1 noro 2035: NODE last, p, head;
2036:
1.20 noro 2037: dl = DL(nd_psh[ndp]);
1.1 noro 2038: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2039: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2040: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2041: p = NEXT(p);
2042: if ( last ) NEXT(last) = p;
2043: } else {
2044: if ( !last ) head = p;
2045: p = NEXT(last = p);
2046: }
2047: }
2048: head = append_one(head,ndp);
2049: return head;
2050: }
2051:
2052: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2053: {
2054: ND_pairs m,ml,p,l;
1.61 noro 2055: UINT *lcm;
1.33 noro 2056: int s,td,len,tlen,c,c1;
1.1 noro 2057:
2058: if ( !(p = NEXT(m = d)) ) {
2059: *prest = p;
2060: NEXT(m) = 0;
2061: return m;
2062: }
1.14 noro 2063: s = SG(m);
1.77 noro 2064: if ( !NoSugar ) {
2065: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2066: if ( (SG(p) < s)
2067: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
2068: ml = l; m = p; s = SG(m);
2069: }
2070: } else {
2071: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2072: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
2073: ml = l; m = p; s = SG(m);
2074: }
2075: }
1.1 noro 2076: if ( !ml ) *prest = NEXT(m);
2077: else {
2078: NEXT(ml) = NEXT(m);
2079: *prest = d;
2080: }
2081: NEXT(m) = 0;
2082: return m;
2083: }
2084:
1.63 noro 2085: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2086: {
1.69 noro 2087: int msugar,i;
1.63 noro 2088: ND_pairs t,dm0,dm,dr0,dr;
2089:
2090: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2091: if ( SG(t) < msugar ) msugar = SG(t);
2092: dm0 = 0; dr0 = 0;
1.69 noro 2093: for ( i = 0, t = d; t; t = NEXT(t) )
2094: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2095: if ( dm0 ) NEXT(dm) = t;
2096: else dm0 = t;
2097: dm = t;
1.69 noro 2098: i++;
1.63 noro 2099: } else {
2100: if ( dr0 ) NEXT(dr) = t;
2101: else dr0 = t;
2102: dr = t;
2103: }
2104: NEXT(dm) = 0;
2105: if ( dr0 ) NEXT(dr) = 0;
2106: *prest = dr0;
2107: return dm0;
2108: }
2109:
1.77 noro 2110: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2111: {
1.3 noro 2112: int len;
1.13 noro 2113: RHist r;
1.20 noro 2114: NDV b;
1.3 noro 2115:
1.1 noro 2116: if ( nd_psn == nd_pslen ) {
2117: nd_pslen *= 2;
1.11 noro 2118: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2119: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2120: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2121: nd_bound = (UINT **)
2122: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2123: }
1.39 noro 2124: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2125: nd_ps[nd_psn] = a;
1.39 noro 2126: if ( aq ) {
1.63 noro 2127: nd_ps_trace[nd_psn] = aq;
2128: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2129: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2130: } else {
1.63 noro 2131: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2132: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2133: }
1.77 noro 2134: if ( nd_demand ) {
2135: if ( aq ) {
2136: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2137: nd_ps_trace[nd_psn] = 0;
2138: } else {
2139: ndv_save(nd_ps[nd_psn],nd_psn);
2140: nd_ps[nd_psn] = 0;
2141: }
2142: }
1.1 noro 2143: return nd_psn++;
2144: }
2145:
1.61 noro 2146: void ndv_setup(int mod,int trace,NODE f)
1.1 noro 2147: {
1.5 noro 2148: int i,j,td,len,max;
1.1 noro 2149: NODE s,s0,f0;
1.61 noro 2150: UINT *d;
1.13 noro 2151: RHist r;
1.69 noro 2152: NDV *w;
1.61 noro 2153: NDV a,am;
1.11 noro 2154:
2155: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2156:
1.69 noro 2157: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2158: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2159: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
2160: qsort(w,nd_psn,sizeof(NDV),
2161: (int (*)(const void *,const void *))ndv_compare);
2162: nd_pslen = 2*nd_psn;
1.11 noro 2163: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2164: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2165: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2166: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2167:
1.11 noro 2168: if ( !nd_red )
1.13 noro 2169: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2170: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2171: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2172: if ( trace ) {
1.69 noro 2173: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2174: ndv_removecont(0,a);
2175: am = nd_ps[i] = ndv_dup(mod,a);
2176: ndv_mod(mod,am);
2177: ndv_removecont(mod,am);
2178: } else {
1.69 noro 2179: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2180: ndv_removecont(mod,a);
1.39 noro 2181: }
1.61 noro 2182: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2183: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2184: nd_psh[i] = r;
1.77 noro 2185: if ( nd_demand ) {
2186: if ( trace ) {
2187: ndv_save(nd_ps_trace[i],i);
2188: nd_ps_trace[i] = 0;
2189: } else {
2190: ndv_save(nd_ps[i],i);
2191: nd_ps[i] = 0;
2192: }
2193: }
1.20 noro 2194: }
2195: }
2196:
1.63 noro 2197: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2198: {
1.61 noro 2199: VL tv,fv,vv,vc;
1.1 noro 2200: NODE fd,fd0,r,r0,t,x,s,xx;
1.61 noro 2201: int e,max,nvar;
1.69 noro 2202: NDV b;
1.77 noro 2203: int ishomo;
1.1 noro 2204:
1.78 noro 2205: if ( !m && Demand ) nd_demand = 1;
2206: else nd_demand = 0;
2207:
1.103 noro 2208: #if 0
1.74 noro 2209: ndv_alloc = 0;
1.103 noro 2210: #endif
1.1 noro 2211: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2212: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2213: switch ( ord->id ) {
2214: case 1:
2215: if ( ord->nv != nvar )
2216: error("nd_{gr,f4} : invalid order specification");
2217: break;
2218: default:
2219: break;
2220: }
1.32 noro 2221: nd_init_ord(ord);
1.61 noro 2222: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2223: for ( tv = vv; tv; tv = NEXT(tv) ) {
2224: e = getdeg(tv->v,(P)BDY(t));
2225: max = MAX(e,max);
2226: }
2227: nd_setup_parameters(nvar,max);
1.77 noro 2228: ishomo = 1;
1.1 noro 2229: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2230: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.77 noro 2231: if ( ishomo )
2232: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2233: if ( m ) ndv_mod(m,b);
1.61 noro 2234: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2235: }
2236: if ( fd0 ) NEXT(fd) = 0;
1.61 noro 2237: ndv_setup(m,0,fd0);
1.77 noro 2238: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.82 noro 2239: nd_demand = 0;
1.61 noro 2240: x = ndv_reducebase(x);
2241: x = ndv_reduceall(m,x);
1.23 noro 2242: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2243: NEXTNODE(r0,r);
1.61 noro 2244: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2245: }
2246: if ( r0 ) NEXT(r) = 0;
2247: MKLIST(*rp,r0);
1.103 noro 2248: #if 0
1.74 noro 2249: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 2250: #endif
1.20 noro 2251: }
2252:
1.52 noro 2253: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2254: {
1.89 noro 2255: struct order_spec *ord1;
1.61 noro 2256: VL tv,fv,vv,vc;
1.27 noro 2257: NODE fd,fd0,in0,in,r,r0,t,s,cand;
1.61 noro 2258: int m,nocheck,nvar,mindex,e,max;
2259: NDV c;
2260: NMV a;
1.27 noro 2261: P p;
1.61 noro 2262: EPOS oepos;
1.82 noro 2263: int obpe,oadv,wmax,i,len,cbpe,ishomo;
1.20 noro 2264:
2265: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2266: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2267: switch ( ord->id ) {
2268: case 1:
2269: if ( ord->nv != nvar )
2270: error("nd_gr_trace : invalid order specification");
2271: break;
2272: default:
2273: break;
2274: }
1.52 noro 2275: nocheck = 0;
2276: mindex = 0;
2277:
1.78 noro 2278: if ( Demand ) nd_demand = 1;
2279: else nd_demand = 0;
2280:
1.52 noro 2281: /* setup modulus */
2282: if ( trace < 0 ) {
2283: trace = -trace;
2284: nocheck = 1;
2285: }
2286: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2287: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2288: for ( tv = vv; tv; tv = NEXT(tv) ) {
2289: e = getdeg(tv->v,(P)BDY(t));
2290: max = MAX(e,max);
1.23 noro 2291: }
1.61 noro 2292: nd_init_ord(ord);
2293: nd_setup_parameters(nvar,max);
2294: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 noro 2295: ishomo = 1;
1.61 noro 2296: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2297: c = ptondv(CO,vv,(P)BDY(t));
1.82 noro 2298: if ( ishomo )
2299: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2300: if ( c ) {
2301: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2302: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2303: }
1.61 noro 2304: }
2305: if ( in0 ) NEXT(in) = 0;
2306: if ( fd0 ) NEXT(fd) = 0;
1.82 noro 2307: if ( !ishomo && homo ) {
1.61 noro 2308: for ( t = in0, wmax = 0; t; t = NEXT(t) ) {
2309: c = (NDV)BDY(t); len = LEN(c);
2310: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2311: wmax = MAX(TD(DL(a)),wmax);
2312: }
2313: homogenize_order(ord,nvar,&ord1);
1.89 noro 2314: nd_init_ord(ord1);
1.61 noro 2315: nd_setup_parameters(nvar+1,wmax);
2316: for ( t = fd0; t; t = NEXT(t) )
2317: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2318: }
1.52 noro 2319: while ( 1 ) {
1.77 noro 2320: if ( Demand )
2321: nd_demand = 1;
1.61 noro 2322: ndv_setup(m,1,fd0);
1.82 noro 2323: cand = nd_gb_trace(m,ishomo || homo);
1.52 noro 2324: if ( !cand ) {
2325: /* failure */
1.61 noro 2326: if ( trace > 1 ) { *rp = 0; return; }
2327: else m = get_lprime(++mindex);
1.52 noro 2328: continue;
2329: }
1.82 noro 2330: if ( !ishomo && homo ) {
1.27 noro 2331: /* dehomogenization */
1.61 noro 2332: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2333: nd_init_ord(ord);
1.61 noro 2334: nd_setup_parameters(nvar,0);
1.27 noro 2335: }
1.77 noro 2336: nd_demand = 0;
1.61 noro 2337: cand = ndv_reducebase(cand);
2338: cand = ndv_reduceall(0,cand);
1.78 noro 2339: cbpe = nd_bpe;
1.61 noro 2340: if ( nocheck )
2341: break;
2342: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2343: /* success */
2344: break;
2345: else if ( trace > 1 ) {
2346: /* failure */
2347: *rp = 0; return;
1.61 noro 2348: } else {
1.52 noro 2349: /* try the next modulus */
2350: m = get_lprime(++mindex);
1.61 noro 2351: /* reset the parameters */
1.82 noro 2352: if ( !ishomo && homo ) {
1.89 noro 2353: nd_init_ord(ord1);
1.61 noro 2354: nd_setup_parameters(nvar+1,wmax);
2355: } else {
2356: nd_init_ord(ord);
2357: nd_setup_parameters(nvar,max);
2358: }
2359: }
1.52 noro 2360: }
1.27 noro 2361: /* dp->p */
1.61 noro 2362: nd_bpe = cbpe;
1.73 noro 2363: nd_setup_parameters(nd_nvar,0);
1.61 noro 2364: for ( r = cand; r; r = NEXT(r) ) BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.27 noro 2365: MKLIST(*rp,cand);
1.1 noro 2366: }
2367:
1.61 noro 2368: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2369: {
1.61 noro 2370: UINT *d;
1.57 noro 2371: int i,j,l,s,ord_l;
1.43 noro 2372: struct order_pair *op;
1.1 noro 2373:
2374: d = dl->d;
1.41 noro 2375: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2376: if ( nd_blockmask ) {
2377: l = nd_blockmask->n;
2378: op = nd_blockmask->order_pair;
2379: for ( j = 0, s = 0; j < l; j++ ) {
2380: ord_l = op[j].length;
1.57 noro 2381: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2382: }
2383: TD(r) = ndl_weight(r);
1.61 noro 2384: ndl_weight_mask(r);
1.43 noro 2385: } else {
1.56 noro 2386: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2387: TD(r) = ndl_weight(r);
2388: }
1.1 noro 2389: }
2390:
1.61 noro 2391: DL ndltodl(int n,UINT *ndl)
1.1 noro 2392: {
2393: DL dl;
2394: int *d;
1.57 noro 2395: int i,j,l,s,ord_l;
1.43 noro 2396: struct order_pair *op;
1.1 noro 2397:
2398: NEWDL(dl,n);
1.34 noro 2399: dl->td = TD(ndl);
1.1 noro 2400: d = dl->d;
1.43 noro 2401: if ( nd_blockmask ) {
2402: l = nd_blockmask->n;
2403: op = nd_blockmask->order_pair;
2404: for ( j = 0, s = 0; j < l; j++ ) {
2405: ord_l = op[j].length;
1.57 noro 2406: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2407: }
2408: } else {
1.56 noro 2409: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2410: }
1.1 noro 2411: return dl;
2412: }
2413:
1.61 noro 2414: void ndl_print(UINT *dl)
1.1 noro 2415: {
2416: int n;
1.57 noro 2417: int i,j,l,ord_l,s,s0;
1.43 noro 2418: struct order_pair *op;
1.1 noro 2419:
2420: n = nd_nvar;
2421: printf("<<");
1.43 noro 2422: if ( nd_blockmask ) {
2423: l = nd_blockmask->n;
2424: op = nd_blockmask->order_pair;
2425: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2426: ord_l = op[j].length;
1.57 noro 2427: for ( i = 0; i < ord_l; i++, s++ )
2428: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2429: }
2430: } else {
1.56 noro 2431: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2432: }
1.1 noro 2433: printf(">>");
2434: }
2435:
2436: void nd_print(ND p)
2437: {
2438: NM m;
2439:
2440: if ( !p )
2441: printf("0\n");
2442: else {
2443: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2444: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2445: else printf("+%d*",CM(m));
1.14 noro 2446: ndl_print(DL(m));
1.1 noro 2447: }
2448: printf("\n");
2449: }
2450: }
2451:
1.113 noro 2452: void nd_print_q(ND p)
1.16 noro 2453: {
2454: NM m;
2455:
2456: if ( !p )
2457: printf("0\n");
2458: else {
2459: for ( m = BDY(p); m; m = NEXT(m) ) {
2460: printf("+");
1.113 noro 2461: printexpr(CO,(Obj)CQ(m));
1.16 noro 2462: printf("*");
2463: ndl_print(DL(m));
2464: }
2465: printf("\n");
2466: }
2467: }
2468:
1.1 noro 2469: void ndp_print(ND_pairs d)
2470: {
2471: ND_pairs t;
2472:
1.34 noro 2473: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2474: printf("\n");
2475: }
2476:
1.20 noro 2477: void nd_removecont(int mod,ND p)
1.16 noro 2478: {
2479: int i,n;
1.113 noro 2480: Q *w;
2481: Q dvr,t;
1.16 noro 2482: NM m;
1.21 noro 2483: struct oVECT v;
1.113 noro 2484: N q,r;
1.16 noro 2485:
1.71 noro 2486: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2487: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2488: else {
2489: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
1.113 noro 2490: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2491: v.len = n;
2492: v.body = (pointer *)w;
1.113 noro 2493: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2494: removecont_array(w,n);
1.113 noro 2495: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2496: }
2497: }
2498:
1.21 noro 2499: void nd_removecont2(ND p1,ND p2)
2500: {
2501: int i,n1,n2,n;
1.113 noro 2502: Q *w;
2503: Q dvr,t;
1.21 noro 2504: NM m;
2505: struct oVECT v;
1.113 noro 2506: N q,r;
1.21 noro 2507:
2508: if ( !p1 ) {
2509: nd_removecont(0,p2); return;
2510: } else if ( !p2 ) {
2511: nd_removecont(0,p1); return;
2512: }
2513: n1 = nd_length(p1);
2514: n2 = nd_length(p2);
2515: n = n1+n2;
1.113 noro 2516: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2517: v.len = n;
2518: v.body = (pointer *)w;
1.113 noro 2519: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2520: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2521: removecont_array(w,n);
1.113 noro 2522: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2523: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.21 noro 2524: }
2525:
1.20 noro 2526: void ndv_removecont(int mod,NDV p)
1.16 noro 2527: {
2528: int i,len;
1.113 noro 2529: Q *w;
2530: Q dvr,t;
1.16 noro 2531: NMV m;
2532:
1.71 noro 2533: if ( mod == -1 )
2534: ndv_mul_c(mod,p,_invsf(HCM(p)));
2535: else if ( mod )
1.20 noro 2536: ndv_mul_c(mod,p,invm(HCM(p),mod));
2537: else {
2538: len = p->len;
1.113 noro 2539: w = (Q *)ALLOCA(len*sizeof(Q));
2540: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
2541: sortbynm(w,len);
2542: qltozl(w,len,&dvr);
2543: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2544: divq(CQ(m),dvr,&t); CQ(m) = t;
2545: }
1.16 noro 2546: }
1.21 noro 2547: }
2548:
1.61 noro 2549: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2550: {
2551: int len,i,max;
2552: NMV m,mr0,mr,t;
2553:
2554: len = p->len;
2555: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2556: max = MAX(max,TD(DL(m)));
2557: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2558: m = (NMV)((char *)mr0+(len-1)*oadv);
2559: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2560: t = (NMV)ALLOCA(nmv_adv);
2561: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2562: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
1.113 noro 2563: CQ(mr) = CQ(m);
1.61 noro 2564: ndl_copy(DL(t),DL(mr));
2565: }
2566: NV(p)++;
2567: BDY(p) = mr0;
2568: }
2569:
1.45 noro 2570: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2571: {
1.45 noro 2572: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.113 noro 2573: Q *w;
2574: Q dvr,t;
1.23 noro 2575: NMV m,r;
2576:
2577: len = p->len;
2578: newnvar = nd_nvar-1;
1.48 noro 2579: newexporigin = nd_get_exporigin(ord);
1.45 noro 2580: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2581: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2582: ndl_dehomogenize(DL(m));
1.23 noro 2583: if ( newwpd != nd_wpd ) {
1.90 noro 2584: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
1.61 noro 2585: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.113 noro 2586: CQ(r) = CQ(m);
1.45 noro 2587: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2588: adj = nd_exporigin-newexporigin;
2589: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2590: }
2591: }
2592: NV(p)--;
2593: }
2594:
1.113 noro 2595: void removecont_array(Q *c,int n)
1.21 noro 2596: {
2597: struct oVECT v;
1.113 noro 2598: Q d0,d1,a,u,u1,gcd;
1.110 noro 2599: int i,j;
1.113 noro 2600: N qn,rn,gn;
2601: Q *q,*r;
1.21 noro 2602:
1.113 noro 2603: q = (Q *)ALLOCA(n*sizeof(Q));
2604: r = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2605: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
1.113 noro 2606: igcdv_estimate(&v,&d0);
2607: for ( i = 0; i < n; i++ ) {
2608: divn(NM(c[i]),NM(d0),&qn,&rn);
2609: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2610: NTOQ(rn,SGN(c[i]),r[i]);
2611: }
1.34 noro 2612: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2613: if ( i < n ) {
2614: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
1.113 noro 2615: igcdv(&v,&d1);
2616: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2617: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
1.21 noro 2618: for ( i = 0; i < n; i++ ) {
1.113 noro 2619: mulq(a,q[i],&u);
2620: if ( r[i] ) {
2621: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2622: addq(u,u1,&q[i]);
2623: } else
2624: q[i] = u;
1.21 noro 2625: }
2626: }
1.34 noro 2627: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2628: }
2629:
1.19 noro 2630: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2631: {
2632: NM m;
2633: int c,c1;
2634:
1.34 noro 2635: if ( !p ) return;
1.115 noro 2636: if ( mul == 1 ) return;
1.71 noro 2637: if ( mod == -1 )
2638: for ( m = BDY(p); m; m = NEXT(m) )
2639: CM(m) = _mulsf(CM(m),mul);
2640: else
2641: for ( m = BDY(p); m; m = NEXT(m) ) {
2642: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2643: }
1.1 noro 2644: }
2645:
1.113 noro 2646: void nd_mul_c_q(ND p,Q mul)
1.16 noro 2647: {
2648: NM m;
1.113 noro 2649: Q c;
1.16 noro 2650:
1.113 noro 2651: if ( !p ) return;
1.115 noro 2652: if ( UNIQ(mul) ) return;
1.16 noro 2653: for ( m = BDY(p); m; m = NEXT(m) ) {
1.113 noro 2654: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 2655: }
2656: }
2657:
1.61 noro 2658: void nd_mul_c_p(VL vl,ND p,P mul)
2659: {
2660: NM m;
2661: P c;
2662:
2663: if ( !p ) return;
2664: for ( m = BDY(p); m; m = NEXT(m) ) {
2665: mulp(vl,CP(m),mul,&c); CP(m) = c;
2666: }
2667: }
2668:
1.1 noro 2669: void nd_free(ND p)
2670: {
2671: NM t,s;
2672:
1.34 noro 2673: if ( !p ) return;
1.1 noro 2674: t = BDY(p);
2675: while ( t ) {
2676: s = NEXT(t);
2677: FREENM(t);
2678: t = s;
2679: }
2680: FREEND(p);
2681: }
2682:
1.23 noro 2683: void ndv_free(NDV p)
2684: {
2685: GC_free(BDY(p));
2686: }
2687:
1.61 noro 2688: void nd_append_red(UINT *d,int i)
1.1 noro 2689: {
1.13 noro 2690: RHist m,m0;
1.1 noro 2691: int h;
2692:
1.13 noro 2693: NEWRHist(m);
1.34 noro 2694: h = ndl_hash_value(d);
1.13 noro 2695: m->index = i;
1.14 noro 2696: ndl_copy(d,DL(m));
1.1 noro 2697: NEXT(m) = nd_red[h];
2698: nd_red[h] = m;
2699: }
2700:
1.61 noro 2701: UINT *ndv_compute_bound(NDV p)
1.1 noro 2702: {
1.61 noro 2703: UINT *d1,*d2,*t;
2704: UINT u;
1.57 noro 2705: int i,j,k,l,len,ind;
1.45 noro 2706: NMV m;
1.1 noro 2707:
2708: if ( !p )
2709: return 0;
1.61 noro 2710: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2711: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2712: len = LEN(p);
2713: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2714: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2715: ndl_lcm(DL(m),d1,d2);
1.1 noro 2716: t = d1; d1 = d2; d2 = t;
2717: }
1.12 noro 2718: l = nd_nvar+31;
1.61 noro 2719: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 2720: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2721: u = d1[i];
2722: k = (nd_epw-1)*nd_bpe;
2723: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2724: t[ind] = (u>>k)&nd_mask0;
2725: }
2726: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 2727: return t;
2728: }
2729:
1.99 noro 2730: UINT *nd_compute_bound(ND p)
2731: {
2732: UINT *d1,*d2,*t;
2733: UINT u;
2734: int i,j,k,l,len,ind;
2735: NM m;
2736:
2737: if ( !p )
2738: return 0;
2739: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2740: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2741: len = LEN(p);
2742: m = BDY(p); ndl_copy(DL(m),d1); m = NEXT(m);
2743: for ( m = NEXT(m); m; m = NEXT(m) ) {
2744: ndl_lcm(DL(m),d1,d2);
2745: t = d1; d1 = d2; d2 = t;
2746: }
2747: l = nd_nvar+31;
2748: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
2749: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2750: u = d1[i];
2751: k = (nd_epw-1)*nd_bpe;
2752: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2753: t[ind] = (u>>k)&nd_mask0;
2754: }
2755: for ( ; ind < l; ind++ ) t[ind] = 0;
2756: return t;
2757: }
2758:
1.48 noro 2759: int nd_get_exporigin(struct order_spec *ord)
2760: {
1.51 noro 2761: switch ( ord->id ) {
1.96 noro 2762: case 0: case 2:
1.48 noro 2763: return 1;
1.41 noro 2764: case 1:
2765: /* block order */
1.43 noro 2766: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 2767: return ord->ord.block.length+1;
1.96 noro 2768: case 3:
2769: error("nd_get_exporigin : composite order is not supported yet.");
1.41 noro 2770: }
1.48 noro 2771: }
2772:
1.61 noro 2773: void nd_setup_parameters(int nvar,int max) {
1.79 noro 2774: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 2775: struct order_pair *op;
1.48 noro 2776:
1.73 noro 2777: nd_nvar = nvar;
1.79 noro 2778: if ( max ) {
1.83 noro 2779: /* XXX */
2780: if ( do_weyl ) nd_bpe = 32;
2781: else if ( max < 2 ) nd_bpe = 1;
1.79 noro 2782: else if ( max < 4 ) nd_bpe = 2;
2783: else if ( max < 8 ) nd_bpe = 3;
2784: else if ( max < 16 ) nd_bpe = 4;
2785: else if ( max < 32 ) nd_bpe = 5;
2786: else if ( max < 64 ) nd_bpe = 6;
2787: else if ( max < 256 ) nd_bpe = 8;
2788: else if ( max < 1024 ) nd_bpe = 10;
2789: else if ( max < 65536 ) nd_bpe = 16;
2790: else nd_bpe = 32;
2791: }
1.61 noro 2792: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 2793: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
2794: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 2795: wpd = nd_exporigin+elen;
2796: if ( wpd != nd_wpd ) {
2797: nd_free_private_storage();
2798: nd_wpd = wpd;
2799: }
1.1 noro 2800: if ( nd_bpe < 32 ) {
2801: nd_mask0 = (1<<nd_bpe)-1;
2802: } else {
2803: nd_mask0 = 0xffffffff;
2804: }
2805: bzero(nd_mask,sizeof(nd_mask));
2806: nd_mask1 = 0;
2807: for ( i = 0; i < nd_epw; i++ ) {
2808: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
2809: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
2810: }
1.90 noro 2811: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
1.57 noro 2812: nd_epos = nd_create_epos(nd_ord);
1.43 noro 2813: nd_blockmask = nd_create_blockmask(nd_ord);
1.96 noro 2814: nd_work_vector = (int *)REALLOC(nd_work_vector,nd_nvar*sizeof(int));
1.1 noro 2815: }
2816:
1.103 noro 2817: ND_pairs nd_reconstruct(int trace,ND_pairs d)
1.1 noro 2818: {
1.37 noro 2819: int i,obpe,oadv,h;
1.71 noro 2820: static NM prev_nm_free_list;
2821: static ND_pairs prev_ndp_free_list;
1.13 noro 2822: RHist mr0,mr;
2823: RHist r;
1.37 noro 2824: RHist *old_red;
1.71 noro 2825: ND_pairs s0,s,t;
1.43 noro 2826: EPOS oepos;
1.15 noro 2827:
1.1 noro 2828: obpe = nd_bpe;
1.11 noro 2829: oadv = nmv_adv;
1.43 noro 2830: oepos = nd_epos;
1.63 noro 2831: if ( obpe < 2 ) nd_bpe = 2;
2832: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 2833: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 2834: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 2835: else if ( obpe < 6 ) nd_bpe = 6;
2836: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 2837: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 2838: else if ( obpe < 16 ) nd_bpe = 16;
2839: else if ( obpe < 32 ) nd_bpe = 32;
2840: else error("nd_reconstruct : exponent too large");
1.5 noro 2841:
1.73 noro 2842: nd_setup_parameters(nd_nvar,0);
1.1 noro 2843: prev_nm_free_list = _nm_free_list;
2844: prev_ndp_free_list = _ndp_free_list;
2845: _nm_free_list = 0;
2846: _ndp_free_list = 0;
1.53 noro 2847: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
2848: if ( trace )
2849: for ( i = nd_psn-1; i >= 0; i-- )
2850: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 2851: s0 = 0;
2852: for ( t = d; t; t = NEXT(t) ) {
2853: NEXTND_pairs(s0,s);
2854: s->i1 = t->i1;
2855: s->i2 = t->i2;
1.14 noro 2856: SG(s) = SG(t);
1.61 noro 2857: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 2858: }
1.37 noro 2859:
2860: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 2861: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 2862: old_red[i] = nd_red[i];
2863: nd_red[i] = 0;
2864: }
2865: for ( i = 0; i < REDTAB_LEN; i++ )
2866: for ( r = old_red[i]; r; r = NEXT(r) ) {
2867: NEWRHist(mr);
1.13 noro 2868: mr->index = r->index;
1.20 noro 2869: SG(mr) = SG(r);
1.61 noro 2870: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 2871: h = ndl_hash_value(DL(mr));
2872: NEXT(mr) = nd_red[h];
2873: nd_red[h] = mr;
1.6 noro 2874: }
1.37 noro 2875: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
2876: old_red = 0;
1.11 noro 2877: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 2878: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 2879: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 2880: nd_psh[i] = r;
1.11 noro 2881: }
1.1 noro 2882: if ( s0 ) NEXT(s) = 0;
2883: prev_nm_free_list = 0;
2884: prev_ndp_free_list = 0;
1.71 noro 2885: #if 0
1.1 noro 2886: GC_gcollect();
1.71 noro 2887: #endif
1.1 noro 2888: return s0;
2889: }
2890:
1.61 noro 2891: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 2892: {
1.57 noro 2893: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 2894: struct order_pair *op;
1.1 noro 2895:
2896: n = nd_nvar;
1.61 noro 2897: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 2898: omask0 = (1<<obpe)-1;
1.34 noro 2899: TD(r) = TD(d);
1.41 noro 2900: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2901: if ( nd_blockmask ) {
2902: l = nd_blockmask->n;
2903: op = nd_blockmask->order_pair;
2904: for ( i = 1; i < nd_exporigin; i++ )
2905: r[i] = d[i];
2906: for ( j = 0, s = 0; j < l; j++ ) {
2907: ord_l = op[j].length;
1.57 noro 2908: for ( i = 0; i < ord_l; i++, s++ ) {
2909: ei = GET_EXP_OLD(d,s);
2910: PUT_EXP(r,s,ei);
2911: }
1.1 noro 2912: }
1.43 noro 2913: } else {
1.56 noro 2914: for ( i = 0; i < n; i++ ) {
2915: ei = GET_EXP_OLD(d,i);
2916: PUT_EXP(r,i,ei);
2917: }
1.1 noro 2918: }
2919: }
1.3 noro 2920:
1.6 noro 2921: ND nd_copy(ND p)
2922: {
2923: NM m,mr,mr0;
1.41 noro 2924: int c,n;
1.6 noro 2925: ND r;
2926:
2927: if ( !p )
2928: return 0;
2929: else {
2930: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
2931: NEXTNM(mr0,mr);
1.14 noro 2932: CM(mr) = CM(m);
2933: ndl_copy(DL(m),DL(mr));
1.6 noro 2934: }
2935: NEXT(mr) = 0;
1.31 noro 2936: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 2937: SG(r) = SG(p);
1.6 noro 2938: return r;
2939: }
2940: }
2941:
1.53 noro 2942: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 2943: {
2944: NM m;
2945: NDV p1,p2;
2946: ND t1,t2;
1.61 noro 2947: UINT *lcm;
1.31 noro 2948: int td;
1.11 noro 2949:
1.77 noro 2950: if ( !mod && nd_demand ) {
2951: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 2952: } else {
1.77 noro 2953: if ( trace ) {
2954: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
2955: } else {
2956: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
2957: }
1.20 noro 2958: }
1.34 noro 2959: lcm = LCM(p);
1.11 noro 2960: NEWNM(m);
1.113 noro 2961: CQ(m) = HCQ(p2);
1.34 noro 2962: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 2963: if ( ndl_check_bound2(p->i1,DL(m)) )
2964: return 0;
1.55 noro 2965: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 2966: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
2967: else if ( mod ) CM(m) = mod-HCM(p1);
1.113 noro 2968: else chsgnq(HCQ(p1),&CQ(m));
1.34 noro 2969: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 2970: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 2971: nd_free(t1);
2972: return 0;
2973: }
1.55 noro 2974: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 2975: *rp = nd_add(mod,t1,t2);
1.11 noro 2976: FREENM(m);
2977: return 1;
2978: }
2979:
1.19 noro 2980: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 2981: {
2982: NMV m;
2983: int c,c1,len,i;
2984:
1.34 noro 2985: if ( !p ) return;
1.14 noro 2986: len = LEN(p);
1.71 noro 2987: if ( mod == -1 )
2988: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
2989: CM(m) = _mulsf(CM(m),mul);
2990: else
2991: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
2992: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2993: }
1.11 noro 2994: }
2995:
1.113 noro 2996: void ndv_mul_c_q(NDV p,Q mul)
1.16 noro 2997: {
2998: NMV m;
1.113 noro 2999: Q c;
1.16 noro 3000: int len,i;
3001:
1.34 noro 3002: if ( !p ) return;
1.16 noro 3003: len = LEN(p);
3004: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 3005: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 3006: }
3007: }
3008:
1.55 noro 3009: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3010: int n2,i,j,l,n,tlen;
1.61 noro 3011: UINT *d0;
1.55 noro 3012: NM *tab,*psum;
3013: ND s,r;
3014: NM t;
3015: NMV m1;
3016:
3017: if ( !p ) return 0;
3018: n = NV(p); n2 = n>>1;
3019: d0 = DL(m0);
3020: l = LEN(p);
3021: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3022: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3023: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3024: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3025: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3026: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3027: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3028: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3029: for ( j = 0; j < tlen; j++ ) {
3030: if ( tab[j] ) {
3031: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3032: }
3033: }
3034: }
3035: for ( i = tlen-1, r = 0; i >= 0; i-- )
3036: if ( psum[i] ) {
3037: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3038: MKND(n,psum[i],j,s);
3039: r = nd_add(mod,r,s);
3040: }
1.56 noro 3041: if ( r ) SG(r) = SG(p)+TD(d0);
3042: return r;
1.55 noro 3043: }
3044:
1.56 noro 3045: /* product of monomials */
3046: /* XXX block order is not handled correctly */
3047:
1.55 noro 3048: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3049: {
1.56 noro 3050: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3051: UINT *d0,*d1,*d,*dt,*ctab;
1.113 noro 3052: Q *ctab_q;
3053: Q q,q1;
1.61 noro 3054: UINT c0,c1,c;
1.55 noro 3055: NM *p;
3056: NM m,t;
3057:
3058: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3059: if ( !m0 || !m1 ) return;
3060: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3061: NEWNM(m); d = DL(m);
1.56 noro 3062: if ( mod ) {
3063: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3064: } else
1.113 noro 3065: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3066: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3067: homo = n&1 ? 1 : 0;
3068: if ( homo ) {
3069: /* offset of h-degree */
3070: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3071: PUT_EXP(DL(m),n-1,h);
3072: TD(DL(m)) = h;
1.61 noro 3073: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3074: }
3075: tab[0] = m;
3076: NEWNM(m); d = DL(m);
1.57 noro 3077: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3078: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3079: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3080: /* xi^a*(Di^k*xi^l)*Di^b */
3081: a += l; b += k;
1.56 noro 3082: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3083: if ( !k || !l ) {
3084: for ( j = 0; j < curlen; j++ )
1.56 noro 3085: if ( t = tab[j] ) {
3086: dt = DL(t);
3087: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3088: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3089: }
3090: curlen *= k+1;
3091: continue;
3092: }
3093: min = MIN(k,l);
1.56 noro 3094: if ( mod ) {
1.61 noro 3095: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3096: mkwcm(k,l,mod,ctab);
3097: } else {
1.113 noro 3098: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3099: mkwc(k,l,ctab_q);
1.56 noro 3100: }
1.57 noro 3101: for ( j = min; j >= 0; j-- ) {
1.56 noro 3102: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3103: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3104: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3105: if ( homo ) {
3106: TD(d) = s;
1.56 noro 3107: PUT_EXP(d,n-1,s-h);
1.55 noro 3108: } else TD(d) = h;
1.61 noro 3109: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3110: if ( mod ) c = ctab[j];
1.113 noro 3111: else q = ctab_q[j];
1.57 noro 3112: p = tab+curlen*j;
3113: if ( j == 0 ) {
3114: for ( u = 0; u < curlen; u++, p++ ) {
3115: if ( tab[u] ) {
3116: ndl_addto(DL(tab[u]),d);
3117: if ( mod ) {
3118: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3119: } else {
1.113 noro 3120: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
1.57 noro 3121: }
3122: }
1.56 noro 3123: }
1.57 noro 3124: } else {
3125: for ( u = 0; u < curlen; u++, p++ ) {
3126: if ( tab[u] ) {
3127: NEWNM(t);
3128: ndl_add(DL(tab[u]),d,DL(t));
3129: if ( mod ) {
3130: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3131: } else
1.113 noro 3132: mulq(CQ(tab[u]),q,&CQ(t));
1.57 noro 3133: *p = t;
3134: }
1.55 noro 3135: }
3136: }
3137: }
3138: curlen *= k+1;
3139: }
3140: FREENM(m);
3141: }
3142:
1.63 noro 3143: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3144: {
3145: NM mr,mr0;
3146: NMV m;
3147: UINT *d,*dt,*dm;
3148: int c,n,td,i,c1,c2,len;
1.113 noro 3149: Q q;
1.63 noro 3150: ND r;
3151:
3152: if ( !p ) return 0;
3153: else {
3154: n = NV(p); m = BDY(p);
3155: d = DL(m0);
3156: len = LEN(p);
3157: mr0 = 0;
3158: td = TD(d);
3159: c = CM(m0);
3160: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3161: NEXTNM(mr0,mr);
3162: CM(mr) = 1;
3163: ndl_add(DL(m),d,DL(mr));
3164: }
3165: NEXT(mr) = 0;
3166: MKND(NV(p),mr0,len,r);
3167: SG(r) = SG(p) + TD(d);
3168: return r;
3169: }
3170: }
3171:
1.55 noro 3172: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3173: {
3174: NM mr,mr0;
3175: NMV m;
1.61 noro 3176: UINT *d,*dt,*dm;
1.9 noro 3177: int c,n,td,i,c1,c2,len;
1.113 noro 3178: Q q;
1.9 noro 3179: ND r;
3180:
1.34 noro 3181: if ( !p ) return 0;
1.55 noro 3182: else if ( do_weyl )
1.71 noro 3183: if ( mod == -1 )
3184: error("ndv_mul_nm : not implemented (weyl)");
3185: else
3186: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3187: else {
3188: n = NV(p); m = BDY(p);
1.34 noro 3189: d = DL(m0);
1.14 noro 3190: len = LEN(p);
1.9 noro 3191: mr0 = 0;
1.34 noro 3192: td = TD(d);
1.71 noro 3193: if ( mod == -1 ) {
3194: c = CM(m0);
3195: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3196: NEXTNM(mr0,mr);
3197: CM(mr) = _mulsf(CM(m),c);
3198: ndl_add(DL(m),d,DL(mr));
3199: }
3200: } else if ( mod ) {
1.16 noro 3201: c = CM(m0);
3202: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3203: NEXTNM(mr0,mr);
3204: c1 = CM(m);
1.19 noro 3205: DMAR(c1,c,0,mod,c2);
1.16 noro 3206: CM(mr) = c2;
3207: ndl_add(DL(m),d,DL(mr));
3208: }
3209: } else {
1.113 noro 3210: q = CQ(m0);
1.16 noro 3211: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3212: NEXTNM(mr0,mr);
1.113 noro 3213: mulq(CQ(m),q,&CQ(mr));
1.16 noro 3214: ndl_add(DL(m),d,DL(mr));
3215: }
1.4 noro 3216: }
1.9 noro 3217: NEXT(mr) = 0;
1.31 noro 3218: MKND(NV(p),mr0,len,r);
1.34 noro 3219: SG(r) = SG(p) + TD(d);
1.9 noro 3220: return r;
1.4 noro 3221: }
3222: }
3223:
1.104 noro 3224: ND nd_quo(int mod,PGeoBucket bucket,NDV d)
1.99 noro 3225: {
3226: NM mq0,mq;
1.102 noro 3227: NMV tm;
1.113 noro 3228: Q q;
1.104 noro 3229: int i,nv,sg,c,c1,c2,hindex;
3230: ND p,t,r;
3231: N tnm;
3232:
1.99 noro 3233: if ( !p ) return 0;
3234: else {
1.104 noro 3235: nv = NV(d);
1.99 noro 3236: mq0 = 0;
1.102 noro 3237: tm = (NMV)ALLOCA(nmv_adv);
1.104 noro 3238: while ( 1 ) {
1.113 noro 3239: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.104 noro 3240: if ( hindex < 0 ) break;
3241: p = bucket->body[hindex];
1.99 noro 3242: NEXTNM(mq0,mq);
3243: ndl_sub(HDL(p),HDL(d),DL(mq));
1.102 noro 3244: ndl_copy(DL(mq),DL(tm));
3245: if ( mod ) {
3246: c1 = invm(HCM(d),mod); c2 = HCM(p);
3247: DMAR(c1,c2,0,mod,c); CM(mq) = c;
3248: CM(tm) = mod-c;
3249: } else {
1.113 noro 3250: divsn(NM(HCQ(p)),NM(HCQ(d)),&tnm);
3251: NTOQ(tnm,SGN(HCQ(p))*SGN(HCQ(d)),CQ(mq));
3252: chsgnq(CQ(mq),&CQ(tm));
1.102 noro 3253: }
3254: t = ndv_mul_nmv_trunc(mod,tm,d,HDL(d));
1.104 noro 3255: bucket->body[hindex] = nd_remove_head(p);
3256: t = nd_remove_head(t);
3257: add_pbucket(mod,bucket,t);
3258: }
3259: if ( !mq0 )
3260: r = 0;
3261: else {
3262: NEXT(mq) = 0;
3263: for ( i = 0, mq = mq0; mq; mq = NEXT(mq), i++ );
3264: MKND(nv,mq0,i,r);
3265: /* XXX */
3266: SG(r) = HTD(r);
1.99 noro 3267: }
3268: return r;
3269: }
3270: }
3271:
1.43 noro 3272: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3273: {
1.13 noro 3274: NMV m,mr,mr0,t;
3275: int len,i,k;
1.11 noro 3276:
1.61 noro 3277: if ( !p ) return;
3278: m = BDY(p); len = LEN(p);
3279: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3280: m = (NMV)((char *)mr0+(len-1)*oadv);
3281: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3282: t = (NMV)ALLOCA(nmv_adv);
3283: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
1.113 noro 3284: CQ(t) = CQ(m);
1.61 noro 3285: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3286: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
1.113 noro 3287: CQ(mr) = CQ(t);
1.61 noro 3288: ndl_copy(DL(t),DL(mr));
3289: }
3290: BDY(p) = mr0;
3291: }
3292:
3293: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3294: {
3295: NMV m,mr,mr0;
3296: int len,i;
3297: NDV r;
1.11 noro 3298:
1.61 noro 3299: if ( !p ) return 0;
3300: m = BDY(p); len = LEN(p);
3301: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3302: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3303: ndl_zero(DL(mr));
3304: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
1.113 noro 3305: CQ(mr) = CQ(m);
1.11 noro 3306: }
1.61 noro 3307: MKNDV(NV(p),mr0,len,r);
3308: SG(r) = SG(p);
3309: return r;
1.11 noro 3310: }
3311:
1.61 noro 3312: /* duplicate p */
3313:
3314: NDV ndv_dup(int mod,NDV p)
1.3 noro 3315: {
3316: NDV d;
1.61 noro 3317: NMV t,m,m0;
1.3 noro 3318: int i,len;
3319:
1.34 noro 3320: if ( !p ) return 0;
1.31 noro 3321: len = LEN(p);
1.34 noro 3322: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3323: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3324: ndl_copy(DL(t),DL(m));
1.113 noro 3325: CQ(m) = CQ(t);
1.3 noro 3326: }
3327: MKNDV(NV(p),m0,len,d);
1.23 noro 3328: SG(d) = SG(p);
3329: return d;
3330: }
3331:
1.63 noro 3332: ND nd_dup(ND p)
3333: {
3334: ND d;
3335: NM t,m,m0;
3336:
3337: if ( !p ) return 0;
3338: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3339: NEXTNM(m0,m);
3340: ndl_copy(DL(t),DL(m));
1.113 noro 3341: CQ(m) = CQ(t);
1.63 noro 3342: }
3343: if ( m0 ) NEXT(m) = 0;
3344: MKND(NV(p),m0,LEN(p),d);
3345: SG(d) = SG(p);
3346: return d;
3347: }
3348:
1.61 noro 3349: /* XXX if p->len == 0 then it represents 0 */
3350:
3351: void ndv_mod(int mod,NDV p)
3352: {
3353: NMV t,d;
3354: int r;
3355: int i,len,dlen;
1.71 noro 3356: Obj gfs;
1.61 noro 3357:
3358: if ( !p ) return;
3359: len = LEN(p);
3360: dlen = 0;
1.71 noro 3361: if ( mod == -1 )
3362: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3363: simp_ff((Obj)CP(t),&gfs);
3364: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3365: CM(d) = r;
3366: ndl_copy(DL(t),DL(d));
3367: NMV_ADV(d);
3368: dlen++;
3369: }
1.71 noro 3370: else
3371: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
1.113 noro 3372: r = rem(NM(CQ(t)),mod);
1.71 noro 3373: if ( r ) {
1.113 noro 3374: if ( SGN(CQ(t)) < 0 )
3375: r = mod-r;
1.71 noro 3376: CM(d) = r;
3377: ndl_copy(DL(t),DL(d));
3378: NMV_ADV(d);
3379: dlen++;
3380: }
3381: }
1.61 noro 3382: LEN(p) = dlen;
3383: }
3384:
3385: NDV ptondv(VL vl,VL dvl,P p)
3386: {
3387: ND nd;
3388:
3389: nd = ptond(vl,dvl,p);
3390: return ndtondv(0,nd);
3391: }
3392:
3393: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3394: {
1.61 noro 3395: int n,i,j,k,e;
3396: VL tvl;
3397: V v;
3398: DCP dc;
3399: DCP *w;
3400: ND r,s,t,u;
3401: P x;
3402: int c;
3403: UINT *d;
1.23 noro 3404: NM m,m0;
1.61 noro 3405:
3406: if ( !p )
3407: return 0;
3408: else if ( NUM(p) ) {
3409: NEWNM(m);
3410: ndl_zero(DL(m));
1.113 noro 3411: CQ(m) = (Q)p;
1.61 noro 3412: NEXT(m) = 0;
3413: MKND(nd_nvar,m,1,r);
3414: SG(r) = 0;
3415: return r;
3416: } else {
3417: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3418: w = (DCP *)ALLOCA(k*sizeof(DCP));
3419: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3420: for ( i = 0, tvl = dvl, v = VR(p);
1.98 noro 3421: tvl && tvl->v != v; tvl = NEXT(tvl), i++ );
1.61 noro 3422: if ( !tvl ) {
3423: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3424: t = ptond(vl,dvl,COEF(w[j]));
3425: pwrp(vl,x,DEG(w[j]),&p);
3426: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3427: }
3428: return s;
3429: } else {
3430: NEWNM(m0); d = DL(m0);
3431: for ( j = k-1, s = 0; j >= 0; j-- ) {
3432: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3433: TD(d) = MUL_WEIGHT(e,i);
3434: if ( nd_blockmask) ndl_weight_mask(d);
3435: t = ptond(vl,dvl,COEF(w[j]));
3436: for ( m = BDY(t); m; m = NEXT(m) )
3437: ndl_addto(DL(m),d);
3438: SG(t) += TD(d);
3439: s = nd_add(0,s,t);
3440: }
3441: FREENM(m0);
3442: return s;
3443: }
3444: }
3445: }
3446:
3447: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3448: {
3449: VL tvl;
3450: int len,n,j,i,e;
3451: NMV m;
3452: Q q;
3453: P c;
3454: UINT *d;
3455: P s,r,u,t,w;
1.71 noro 3456: GFS gfs;
1.23 noro 3457:
1.34 noro 3458: if ( !p ) return 0;
1.61 noro 3459: else {
3460: len = LEN(p);
3461: n = NV(p);
3462: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3463: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3464: if ( mod == -1 ) {
3465: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3466: } else if ( mod ) {
1.61 noro 3467: STOQ(CM(m),q); c = (P)q;
3468: } else
1.113 noro 3469: c = CP(m);
1.61 noro 3470: d = DL(m);
3471: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3472: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3473: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3474: }
3475: addp(vl,s,t,&u); s = u;
3476: }
3477: return s;
1.23 noro 3478: }
1.3 noro 3479: }
3480:
1.61 noro 3481: NDV ndtondv(int mod,ND p)
1.11 noro 3482: {
3483: NDV d;
1.61 noro 3484: NMV m,m0;
3485: NM t;
3486: int i,len;
1.11 noro 3487:
1.34 noro 3488: if ( !p ) return 0;
1.61 noro 3489: len = LEN(p);
1.74 noro 3490: if ( mod )
3491: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3492: else
3493: m0 = m = MALLOC(len*nmv_adv);
1.103 noro 3494: #if 0
1.74 noro 3495: ndv_alloc += nmv_adv*len;
1.103 noro 3496: #endif
1.61 noro 3497: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3498: ndl_copy(DL(t),DL(m));
1.113 noro 3499: CQ(m) = CQ(t);
1.11 noro 3500: }
1.61 noro 3501: MKNDV(NV(p),m0,len,d);
1.14 noro 3502: SG(d) = SG(p);
1.11 noro 3503: return d;
3504: }
3505:
1.61 noro 3506: ND ndvtond(int mod,NDV p)
1.11 noro 3507: {
1.61 noro 3508: ND d;
3509: NM m,m0;
1.11 noro 3510: NMV t;
1.61 noro 3511: int i,len;
1.11 noro 3512:
1.34 noro 3513: if ( !p ) return 0;
1.11 noro 3514: m0 = 0;
1.61 noro 3515: len = p->len;
3516: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3517: NEXTNM(m0,m);
3518: ndl_copy(DL(t),DL(m));
1.113 noro 3519: CQ(m) = CQ(t);
1.11 noro 3520: }
3521: NEXT(m) = 0;
1.61 noro 3522: MKND(NV(p),m0,len,d);
1.14 noro 3523: SG(d) = SG(p);
1.11 noro 3524: return d;
3525: }
3526:
1.3 noro 3527: void ndv_print(NDV p)
3528: {
3529: NMV m;
3530: int i,len;
3531:
1.34 noro 3532: if ( !p ) printf("0\n");
1.3 noro 3533: else {
1.14 noro 3534: len = LEN(p);
1.3 noro 3535: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3536: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3537: else printf("+%d*",CM(m));
1.16 noro 3538: ndl_print(DL(m));
3539: }
3540: printf("\n");
3541: }
3542: }
3543:
1.113 noro 3544: void ndv_print_q(NDV p)
1.16 noro 3545: {
3546: NMV m;
3547: int i,len;
3548:
1.34 noro 3549: if ( !p ) printf("0\n");
1.16 noro 3550: else {
3551: len = LEN(p);
3552: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3553: printf("+");
1.113 noro 3554: printexpr(CO,(Obj)CQ(m));
1.16 noro 3555: printf("*");
1.14 noro 3556: ndl_print(DL(m));
1.3 noro 3557: }
3558: printf("\n");
3559: }
1.25 noro 3560: }
3561:
1.61 noro 3562: NODE ndv_reducebase(NODE x)
1.27 noro 3563: {
3564: int len,i,j;
3565: NDV *w;
3566: NODE t,t0;
3567:
3568: len = length(x);
3569: w = (NDV *)ALLOCA(len*sizeof(NDV));
3570: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3571: for ( i = 0; i < len; i++ ) {
3572: for ( j = 0; j < i; j++ ) {
3573: if ( w[i] && w[j] )
3574: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3575: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3576: }
3577: }
3578: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3579: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3580: }
3581: NEXT(t) = 0; x = t0;
3582: return x;
1.11 noro 3583: }
1.32 noro 3584:
1.43 noro 3585: /* XXX incomplete */
3586:
1.32 noro 3587: void nd_init_ord(struct order_spec *ord)
3588: {
1.43 noro 3589: switch ( ord->id ) {
1.32 noro 3590: case 0:
1.43 noro 3591: switch ( ord->ord.simple ) {
3592: case 0:
3593: nd_dcomp = 1;
3594: nd_isrlex = 1;
3595: break;
3596: case 1:
3597: nd_dcomp = 1;
3598: nd_isrlex = 0;
3599: break;
3600: case 2:
3601: nd_dcomp = 0;
3602: nd_isrlex = 0;
1.45 noro 3603: ndl_compare_function = ndl_lex_compare;
1.58 noro 3604: break;
3605: case 11:
3606: /* XXX */
3607: nd_dcomp = 0;
3608: nd_isrlex = 1;
3609: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3610: break;
3611: default:
3612: error("nd_gr : unsupported order");
3613: }
1.32 noro 3614: break;
3615: case 1:
1.96 noro 3616: /* block order */
1.43 noro 3617: /* XXX */
3618: nd_dcomp = -1;
1.32 noro 3619: nd_isrlex = 0;
1.45 noro 3620: ndl_compare_function = ndl_block_compare;
1.34 noro 3621: break;
1.43 noro 3622: case 2:
1.96 noro 3623: /* matrix order */
3624: /* XXX */
3625: nd_dcomp = -1;
3626: nd_isrlex = 0;
3627: nd_matrix_len = ord->ord.matrix.row;
3628: nd_matrix = ord->ord.matrix.matrix;
3629: ndl_compare_function = ndl_matrix_compare;
3630: break;
3631: case 3:
1.97 noro 3632: /* composite order */
3633: nd_dcomp = -1;
3634: nd_isrlex = 0;
3635: nd_worb_len = ord->ord.composite.length;
3636: nd_worb = ord->ord.composite.w_or_b;
3637: ndl_compare_function = ndl_composite_compare;
1.32 noro 3638: break;
3639: }
1.41 noro 3640: nd_ord = ord;
1.32 noro 3641: }
3642:
1.43 noro 3643: BlockMask nd_create_blockmask(struct order_spec *ord)
3644: {
3645: int n,i,j,s,l;
1.61 noro 3646: UINT *t;
1.43 noro 3647: BlockMask bm;
3648:
1.96 noro 3649: /* we only create mask table for block order */
3650: if ( ord->id != 1 )
1.43 noro 3651: return 0;
3652: n = ord->ord.block.length;
3653: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3654: bm->n = n;
3655: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3656: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3657: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3658: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3659: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3660: l = bm->order_pair[i].length;
3661: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3662: }
3663: return bm;
1.57 noro 3664: }
3665:
3666: EPOS nd_create_epos(struct order_spec *ord)
3667: {
3668: int i,j,l,s,ord_l,ord_o;
3669: EPOS epos;
3670: struct order_pair *op;
3671:
3672: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3673: switch ( ord->id ) {
3674: case 0:
3675: if ( nd_isrlex ) {
3676: for ( i = 0; i < nd_nvar; i++ ) {
3677: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3678: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3679: }
3680: } else {
3681: for ( i = 0; i < nd_nvar; i++ ) {
3682: epos[i].i = nd_exporigin + i/nd_epw;
3683: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3684: }
3685: }
3686: break;
3687: case 1:
3688: /* block order */
3689: l = ord->ord.block.length;
3690: op = ord->ord.block.order_pair;
3691: for ( j = 0, s = 0; j < l; j++ ) {
3692: ord_o = op[j].order;
3693: ord_l = op[j].length;
3694: if ( !ord_o )
3695: for ( i = 0; i < ord_l; i++ ) {
3696: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3697: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3698: }
3699: else
3700: for ( i = 0; i < ord_l; i++ ) {
3701: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3702: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3703: }
3704: s += ord_l;
3705: }
3706: break;
3707: case 2:
1.96 noro 3708: /* matrix order */
3709: case 3:
3710: /* composite order */
3711: for ( i = 0; i < nd_nvar; i++ ) {
3712: epos[i].i = nd_exporigin + i/nd_epw;
3713: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3714: }
3715: break;
1.57 noro 3716: }
3717: return epos;
1.43 noro 3718: }
1.59 noro 3719:
3720: /* external interface */
3721:
3722: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
3723: {
1.61 noro 3724: NODE t,in0,in;
1.59 noro 3725: ND nd,nf;
1.61 noro 3726: NDV ndv;
3727: VL vv,tv;
3728: int stat,nvar,max,e;
1.113 noro 3729: union oNDC dn;
1.59 noro 3730:
1.116 noro 3731: if ( !f ) {
3732: *rp = 0;
3733: return;
3734: }
1.59 noro 3735: pltovl(v,&vv);
1.74 noro 3736: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 3737:
3738: /* get the degree bound */
3739: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
3740: for ( tv = vv; tv; tv = NEXT(tv) ) {
3741: e = getdeg(tv->v,(P)BDY(t));
3742: max = MAX(e,max);
3743: }
3744: for ( tv = vv; tv; tv = NEXT(tv) ) {
3745: e = getdeg(tv->v,f);
3746: max = MAX(e,max);
3747: }
3748:
1.59 noro 3749: nd_init_ord(ord);
1.61 noro 3750: nd_setup_parameters(nvar,max);
3751:
3752: /* conversion to ndv */
3753: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
3754: NEXTNODE(in0,in);
3755: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 3756: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3757: }
3758: NEXTNODE(in0,in);
3759: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 3760: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3761: NEXT(in) = 0;
3762:
3763: ndv_setup(m,0,in0);
1.59 noro 3764: nd_psn--;
3765: nd_scale=2;
3766: while ( 1 ) {
3767: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 3768: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 3769: if ( !stat ) {
3770: nd_psn++;
1.103 noro 3771: nd_reconstruct(0,0);
1.59 noro 3772: nd_psn--;
3773: } else
3774: break;
3775: }
1.61 noro 3776: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 3777: }
3778:
3779: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
3780: {
3781: NM m;
3782: UINT *t,*s;
3783: int i;
3784:
3785: for ( i = 0; i < n; i++ ) r[i] = 0;
3786: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
3787: t = DL(m);
3788: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3789: r[i] = CM(m);
3790: }
3791: for ( i = 0; !r[i]; i++ );
3792: return i;
3793: }
3794:
1.113 noro 3795: int nd_to_vect_q(UINT *s0,int n,ND d,Q *r)
1.74 noro 3796: {
1.107 noro 3797: NM m;
1.74 noro 3798: UINT *t,*s;
1.107 noro 3799: int i;
1.74 noro 3800:
3801: for ( i = 0; i < n; i++ ) r[i] = 0;
1.107 noro 3802: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
1.74 noro 3803: t = DL(m);
3804: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.113 noro 3805: r[i] = CQ(m);
1.74 noro 3806: }
3807: for ( i = 0; !r[i]; i++ );
3808: return i;
3809: }
3810:
1.67 noro 3811: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 3812: {
3813: NM m;
3814: NMV mr;
3815: UINT *d,*t,*s;
3816: NDV p;
1.67 noro 3817: unsigned char *ivc;
3818: unsigned short *ivs;
1.81 noro 3819: UINT *v,*ivi,*s0v;
1.67 noro 3820: int i,j,len,prev,diff,cdiff;
3821: IndArray r;
1.64 noro 3822:
3823: m = pair->mul;
3824: d = DL(m);
3825: p = nd_ps[pair->index];
3826: len = LEN(p);
3827: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 3828: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 3829: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3830: ndl_add(d,DL(mr),t);
3831: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 3832: v[j] = i;
3833: }
1.81 noro 3834: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 3835: r->head = v[0];
3836: diff = 0;
3837: for ( i = 1; i < len; i++ ) {
3838: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
3839: }
3840: if ( diff < 256 ) {
3841: r->width = 1;
3842: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
3843: r->index.c = ivc;
3844: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
3845: } else if ( diff < 65536 ) {
3846: r->width = 2;
3847: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
3848: r->index.s = ivs;
3849: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
3850: } else {
3851: r->width = 4;
3852: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
3853: r->index.i = ivi;
3854: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 3855: }
1.67 noro 3856: return r;
1.64 noro 3857: }
3858:
1.65 noro 3859:
1.113 noro 3860: int ndv_reduce_vect_q(Q *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.107 noro 3861: {
3862: int i,j,k,len,pos,prev;
1.113 noro 3863: Q cs,mcs,c1,c2,cr,gcd,t;
1.107 noro 3864: IndArray ivect;
3865: unsigned char *ivc;
3866: unsigned short *ivs;
3867: unsigned int *ivi;
3868: NDV redv;
3869: NMV mr;
3870: NODE rp;
3871: int maxrs;
3872:
3873: maxrs = 0;
3874: for ( i = 0; i < nred; i++ ) {
3875: ivect = imat[i];
3876: k = ivect->head;
3877: if ( svect[k] ) {
3878: maxrs = MAX(maxrs,rp0[i]->sugar);
3879: redv = nd_ps[rp0[i]->index];
3880: len = LEN(redv); mr = BDY(redv);
1.113 noro 3881: igcd_cofactor(svect[k],CQ(mr),&gcd,&cs,&cr);
3882: chsgnq(cs,&mcs);
3883: if ( !UNIQ(cr) ) {
3884: for ( j = 0; j < col; j++ ) {
3885: mulq(svect[j],cr,&c1); svect[j] = c1;
3886: }
3887: }
1.107 noro 3888: svect[k] = 0; prev = k;
3889: switch ( ivect->width ) {
3890: case 1:
3891: ivc = ivect->index.c;
3892: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3893: pos = prev+ivc[j]; prev = pos;
1.113 noro 3894: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 3895: }
3896: break;
3897: case 2:
3898: ivs = ivect->index.s;
3899: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3900: pos = prev+ivs[j]; prev = pos;
1.113 noro 3901: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 3902: }
3903: break;
3904: case 4:
3905: ivi = ivect->index.i;
3906: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3907: pos = prev+ivi[j]; prev = pos;
1.113 noro 3908: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 3909: }
3910: break;
3911: }
3912: }
3913: }
3914: return maxrs;
3915: }
3916:
1.76 noro 3917: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 3918: {
1.67 noro 3919: int i,j,k,len,pos,prev;
1.66 noro 3920: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 3921: IndArray ivect;
3922: unsigned char *ivc;
3923: unsigned short *ivs;
3924: unsigned int *ivi;
1.65 noro 3925: NDV redv;
1.67 noro 3926: NMV mr;
1.65 noro 3927: NODE rp;
1.76 noro 3928: int maxrs;
1.65 noro 3929:
1.76 noro 3930: maxrs = 0;
1.74 noro 3931: for ( i = 0; i < nred; i++ ) {
1.65 noro 3932: ivect = imat[i];
1.67 noro 3933: k = ivect->head; svect[k] %= m;
1.65 noro 3934: if ( c = svect[k] ) {
1.76 noro 3935: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3936: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 3937: len = LEN(redv); mr = BDY(redv);
3938: svect[k] = 0; prev = k;
3939: switch ( ivect->width ) {
3940: case 1:
3941: ivc = ivect->index.c;
3942: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3943: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
3944: prev = pos;
3945: DMA(c1,c,c2,up,lo);
3946: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3947: } else svect[pos] = lo;
3948: }
3949: break;
3950: case 2:
3951: ivs = ivect->index.s;
3952: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3953: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
3954: prev = pos;
3955: DMA(c1,c,c2,up,lo);
3956: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3957: } else svect[pos] = lo;
3958: }
3959: break;
3960: case 4:
3961: ivi = ivect->index.i;
3962: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3963: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
3964: prev = pos;
3965: DMA(c1,c,c2,up,lo);
3966: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3967: } else svect[pos] = lo;
3968: }
3969: break;
1.65 noro 3970: }
3971: }
3972: }
1.66 noro 3973: for ( i = 0; i < col; i++ )
3974: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 3975: return maxrs;
1.65 noro 3976: }
3977:
1.76 noro 3978: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 3979: {
3980: int i,j,k,len,pos,prev;
3981: UINT c,c1,c2,c3,up,lo,dmy;
3982: IndArray ivect;
3983: unsigned char *ivc;
3984: unsigned short *ivs;
3985: unsigned int *ivi;
3986: NDV redv;
3987: NMV mr;
3988: NODE rp;
1.76 noro 3989: int maxrs;
1.72 noro 3990:
1.76 noro 3991: maxrs = 0;
1.74 noro 3992: for ( i = 0; i < nred; i++ ) {
1.72 noro 3993: ivect = imat[i];
3994: k = ivect->head; svect[k] %= m;
3995: if ( c = svect[k] ) {
1.76 noro 3996: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3997: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 3998: len = LEN(redv); mr = BDY(redv);
3999: svect[k] = 0; prev = k;
4000: switch ( ivect->width ) {
4001: case 1:
4002: ivc = ivect->index.c;
4003: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4004: pos = prev+ivc[j]; prev = pos;
4005: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4006: }
4007: break;
4008: case 2:
4009: ivs = ivect->index.s;
4010: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4011: pos = prev+ivs[j]; prev = pos;
4012: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4013: }
4014: break;
4015: case 4:
4016: ivi = ivect->index.i;
4017: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4018: pos = prev+ivi[j]; prev = pos;
4019: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4020: }
4021: break;
4022: }
4023: }
4024: }
1.76 noro 4025: return maxrs;
1.72 noro 4026: }
4027:
1.65 noro 4028: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
4029: {
4030: int j,k,len;
4031: UINT *p;
4032: UINT c;
4033: NDV r;
4034: NMV mr0,mr;
4035:
4036: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4037: if ( !len ) return 0;
4038: else {
1.74 noro 4039: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4040: #if 0
1.74 noro 4041: ndv_alloc += nmv_adv*len;
1.103 noro 4042: #endif
1.65 noro 4043: mr = mr0;
4044: p = s0vect;
4045: for ( j = k = 0; j < col; j++, p += nd_wpd )
4046: if ( !rhead[j] ) {
4047: if ( c = vect[k++] ) {
4048: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
4049: }
4050: }
4051: MKNDV(nd_nvar,mr0,len,r);
4052: return r;
4053: }
4054: }
4055:
1.113 noro 4056: NDV vect_to_ndv_q(Q *vect,int spcol,int col,int *rhead,UINT *s0vect)
1.107 noro 4057: {
4058: int j,k,len;
4059: UINT *p;
1.113 noro 4060: Q c;
1.107 noro 4061: NDV r;
4062: NMV mr0,mr;
4063:
4064: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4065: if ( !len ) return 0;
4066: else {
1.109 noro 4067: mr0 = (NMV)GC_malloc(nmv_adv*len);
1.107 noro 4068: #if 0
4069: ndv_alloc += nmv_adv*len;
4070: #endif
4071: mr = mr0;
4072: p = s0vect;
4073: for ( j = k = 0; j < col; j++, p += nd_wpd )
4074: if ( !rhead[j] ) {
4075: if ( c = vect[k++] ) {
1.113 noro 4076: if ( DN(c) )
4077: error("afo");
4078: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
1.107 noro 4079: }
4080: }
4081: MKNDV(nd_nvar,mr0,len,r);
4082: return r;
4083: }
4084: }
4085:
1.74 noro 4086: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket)
1.65 noro 4087: {
4088: ND_pairs t;
4089: NODE sp0,sp;
4090: int stat;
4091: ND spol;
4092:
4093: for ( t = l; t; t = NEXT(t) ) {
4094: stat = nd_sp(m,0,t,&spol);
4095: if ( !stat ) return 0;
4096: if ( spol ) {
4097: add_pbucket_symbolic(bucket,spol);
4098: }
4099: }
1.68 noro 4100: return 1;
1.65 noro 4101: }
4102:
4103: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
4104: {
4105: NODE rp0,rp;
4106: NM mul,head,s0,s;
1.76 noro 4107: int index,col,i,sugar;
1.65 noro 4108: RHist h;
4109: UINT *s0v,*p;
4110: NM_ind_pair pair;
4111: ND red;
4112:
4113: s0 = 0; rp0 = 0; col = 0;
4114: while ( 1 ) {
4115: head = remove_head_pbucket_symbolic(bucket);
4116: if ( !head ) break;
4117: if ( !s0 ) s0 = head;
4118: else NEXT(s) = head;
4119: s = head;
4120: index = ndl_find_reducer(DL(head));
4121: if ( index >= 0 ) {
4122: h = nd_psh[index];
4123: NEWNM(mul);
4124: ndl_sub(DL(head),DL(h),DL(mul));
4125: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.76 noro 4126: sugar = TD(DL(mul))+SG(nd_ps[index]);
4127: MKNM_ind_pair(pair,mul,index,sugar);
1.65 noro 4128: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
4129: add_pbucket_symbolic(bucket,nd_remove_head(red));
4130: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4131: }
4132: col++;
4133: }
1.72 noro 4134: if ( rp0 ) NEXT(rp) = 0;
4135: NEXT(s) = 0;
1.65 noro 4136: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4137: for ( i = 0, p = s0v, s = s0; i < col;
4138: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4139: *s0vect = s0v;
4140: *r = rp0;
4141: return col;
4142: }
4143:
1.69 noro 4144: NODE nd_f4(int m)
4145: {
4146: int i,nh,stat,index;
4147: NODE r,g;
4148: ND_pairs d,l,t;
4149: ND spol,red;
4150: NDV nf,redv;
4151: NM s0,s;
1.84 noro 4152: NODE rp0,srp0,nflist;
1.69 noro 4153: int nsp,nred,col,rank,len,k,j,a;
4154: UINT c;
1.74 noro 4155: UINT **spmat;
1.69 noro 4156: UINT *s0vect,*svect,*p,*v;
4157: int *colstat;
4158: IndArray *imat;
4159: int *rhead;
4160: int spcol,sprow;
4161: int sugar;
4162: PGeoBucket bucket;
4163: struct oEGT eg0,eg1,eg_f4;
4164:
1.103 noro 4165: #if 0
1.74 noro 4166: ndv_alloc = 0;
1.103 noro 4167: #endif
1.69 noro 4168: g = 0; d = 0;
4169: for ( i = 0; i < nd_psn; i++ ) {
4170: d = update_pairs(d,g,i);
4171: g = update_base(g,i);
4172: }
4173: while ( d ) {
4174: get_eg(&eg0);
4175: l = nd_minsugarp(d,&d);
4176: sugar = SG(l);
4177: bucket = create_pbucket();
1.74 noro 4178: stat = nd_sp_f4(m,l,bucket);
1.69 noro 4179: if ( !stat ) {
4180: for ( t = l; NEXT(t); t = NEXT(t) );
4181: NEXT(t) = d; d = l;
1.103 noro 4182: d = nd_reconstruct(0,d);
1.69 noro 4183: continue;
4184: }
1.86 noro 4185: if ( bucket->m < 0 ) continue;
1.69 noro 4186: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
4187: if ( !col ) {
4188: for ( t = l; NEXT(t); t = NEXT(t) );
4189: NEXT(t) = d; d = l;
1.103 noro 4190: d = nd_reconstruct(0,d);
1.69 noro 4191: continue;
4192: }
4193: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4194: if ( DP_Print )
4195: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4196: sugar,eg_f4.exectime+eg_f4.gctime);
4197: if ( 1 )
4198: nflist = nd_f4_red(m,l,s0vect,col,rp0);
4199: else
4200: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0);
1.69 noro 4201: /* adding new bases */
1.74 noro 4202: for ( r = nflist; r; r = NEXT(r) ) {
4203: nf = (NDV)BDY(r);
1.69 noro 4204: ndv_removecont(m,nf);
1.77 noro 4205: nh = ndv_newps(m,nf,0);
1.69 noro 4206: d = update_pairs(d,g,nh);
4207: g = update_base(g,nh);
4208: }
4209: }
4210: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.103 noro 4211: #if 0
1.74 noro 4212: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 4213: #endif
1.69 noro 4214: return g;
4215: }
1.74 noro 4216:
4217: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
1.63 noro 4218: {
1.67 noro 4219: IndArray *imat;
1.106 noro 4220: int nsp,nred,i;
1.65 noro 4221: int *rhead;
1.106 noro 4222: NODE r0,rp;
1.74 noro 4223: ND_pairs sp;
4224: NM_ind_pair *rvect;
1.63 noro 4225:
1.74 noro 4226: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
1.106 noro 4227: nred = length(rp0);
1.74 noro 4228: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4229: rhead = (int *)ALLOCA(col*sizeof(int));
4230: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4231:
1.74 noro 4232: /* construction of index arrays */
4233: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4234: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4235: rvect[i] = (NM_ind_pair)BDY(rp);
4236: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4237: rhead[imat[i]->head] = 1;
4238: }
1.107 noro 4239: if ( m )
4240: r0 = nd_f4_red_main(m,sp0,nsp,s0vect,col,rvect,rhead,imat,nred);
4241: else
1.113 noro 4242: r0 = nd_f4_red_q_main(sp0,nsp,s0vect,col,rvect,rhead,imat,nred);
1.106 noro 4243: return r0;
4244: }
1.74 noro 4245:
1.106 noro 4246: NODE nd_f4_red_main(int m,ND_pairs sp0,int nsp,UINT *s0vect,int col,
4247: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4248: {
4249: int spcol,sprow,a;
4250: int i,j,k,l,rank;
4251: NODE r0,r;
4252: ND_pairs sp;
4253: ND spol;
4254: int **spmat;
4255: UINT *svect,*v;
4256: int *colstat;
4257: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4258: int maxrs;
4259: int *spsugar;
4260:
4261: spcol = col-nred;
4262: get_eg(&eg0);
1.74 noro 4263: /* elimination (1st step) */
4264: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4265: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4266: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4267: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4268: nd_sp(m,0,sp,&spol);
1.75 noro 4269: if ( !spol ) continue;
1.74 noro 4270: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4271: if ( m == -1 )
4272: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4273: else
4274: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4275: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4276: if ( i < col ) {
4277: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4278: for ( j = k = 0; j < col; j++ )
4279: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4280: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4281: sprow++;
4282: }
1.76 noro 4283: nd_free(spol);
1.74 noro 4284: }
1.80 noro 4285: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4286: if ( DP_Print ) {
4287: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4288: fflush(asir_out);
4289: }
1.74 noro 4290: /* free index arrays */
4291: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4292:
4293: /* elimination (2nd step) */
4294: colstat = (int *)ALLOCA(spcol*sizeof(int));
4295: if ( m == -1 )
1.76 noro 4296: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4297: else
1.76 noro 4298: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4299: r0 = 0;
4300: for ( i = 0; i < rank; i++ ) {
4301: NEXTNODE(r0,r); BDY(r) =
4302: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4303: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4304: GC_free(spmat[i]);
4305: }
1.109 noro 4306: if ( r0 ) NEXT(r) = 0;
1.74 noro 4307: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4308: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4309: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4310: if ( DP_Print ) {
1.80 noro 4311: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4312: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4313: nsp,nred,sprow,spcol,rank);
4314: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4315: }
1.74 noro 4316: return r0;
4317: }
4318:
1.113 noro 4319: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,UINT *s0vect,int col,
1.107 noro 4320: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4321: {
4322: int spcol,sprow,a;
4323: int i,j,k,l,rank;
4324: NODE r0,r;
4325: ND_pairs sp;
4326: ND spol;
1.113 noro 4327: Q **spmat;
4328: Q *svect,*v;
1.107 noro 4329: int *colstat;
4330: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4331: int maxrs;
4332: int *spsugar;
4333:
4334: spcol = col-nred;
4335: get_eg(&eg0);
4336: /* elimination (1st step) */
1.113 noro 4337: spmat = (Q **)ALLOCA(nsp*sizeof(Q *));
4338: svect = (Q *)ALLOCA(col*sizeof(Q));
4339: spsugar = (int *)ALLOCA(nsp*sizeof(Q));
1.107 noro 4340: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4341: nd_sp(0,0,sp,&spol);
4342: if ( !spol ) continue;
1.113 noro 4343: nd_to_vect_q(s0vect,col,spol,svect);
4344: maxrs = ndv_reduce_vect_q(svect,col,imat,rvect,nred);
1.107 noro 4345: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4346: if ( i < col ) {
1.113 noro 4347: spmat[sprow] = v = (Q *)MALLOC(spcol*sizeof(Q));
1.107 noro 4348: for ( j = k = 0; j < col; j++ )
4349: if ( !rhead[j] ) v[k++] = svect[j];
4350: spsugar[sprow] = MAX(maxrs,SG(spol));
4351: sprow++;
4352: }
1.109 noro 4353: /* nd_free(spol); */
1.107 noro 4354: }
4355: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4356: if ( DP_Print ) {
4357: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4358: fflush(asir_out);
4359: }
4360: /* free index arrays */
1.109 noro 4361: /* for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c); */
1.107 noro 4362:
4363: /* elimination (2nd step) */
4364: colstat = (int *)ALLOCA(spcol*sizeof(int));
1.113 noro 4365: rank = nd_gauss_elim_q(spmat,spsugar,sprow,spcol,colstat);
1.107 noro 4366: r0 = 0;
4367: for ( i = 0; i < rank; i++ ) {
4368: NEXTNODE(r0,r); BDY(r) =
1.113 noro 4369: (pointer)vect_to_ndv_q(spmat[i],spcol,col,rhead,s0vect);
1.107 noro 4370: SG((NDV)BDY(r)) = spsugar[i];
1.109 noro 4371: /* GC_free(spmat[i]); */
1.107 noro 4372: }
1.109 noro 4373: if ( r0 ) NEXT(r) = 0;
4374:
4375: /* for ( ; i < sprow; i++ ) GC_free(spmat[i]); */
1.107 noro 4376: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4377: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
4378: if ( DP_Print ) {
4379: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
4380: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4381: nsp,nred,sprow,spcol,rank);
4382: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
4383: }
4384: return r0;
4385: }
4386:
1.74 noro 4387: FILE *nd_write,*nd_read;
4388:
4389: void nd_send_int(int a) {
4390: write_int(nd_write,&a);
4391: }
4392:
4393: void nd_send_intarray(int *p,int len) {
4394: write_intarray(nd_write,p,len);
4395: }
4396:
4397: int nd_recv_int() {
4398: int a;
4399:
4400: read_int(nd_read,&a);
4401: return a;
4402: }
4403:
4404: void nd_recv_intarray(int *p,int len) {
4405: read_intarray(nd_read,p,len);
4406: }
4407:
4408: void nd_send_ndv(NDV p) {
4409: int len,i;
4410: NMV m;
4411:
4412: if ( !p ) nd_send_int(0);
4413: else {
4414: len = LEN(p);
4415: nd_send_int(len);
4416: m = BDY(p);
4417: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4418: nd_send_int(CM(m));
4419: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4420: }
1.74 noro 4421: }
4422: }
4423:
4424: void nd_send_nd(ND p) {
4425: int len,i;
4426: NM m;
4427:
4428: if ( !p ) nd_send_int(0);
4429: else {
4430: len = LEN(p);
4431: nd_send_int(len);
4432: m = BDY(p);
4433: for ( i = 0; i < len; i++, m = NEXT(m) ) {
4434: nd_send_int(CM(m));
4435: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4436: }
1.74 noro 4437: }
4438: }
1.65 noro 4439:
1.74 noro 4440: NDV nd_recv_ndv()
4441: {
4442: int len,i;
4443: NMV m,m0;
4444: NDV r;
1.65 noro 4445:
1.74 noro 4446: len = nd_recv_int();
4447: if ( !len ) return 0;
4448: else {
4449: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4450: #if 0
1.74 noro 4451: ndv_alloc += len*nmv_adv;
1.103 noro 4452: #endif
1.74 noro 4453: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4454: CM(m) = nd_recv_int();
4455: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4456: }
1.74 noro 4457: MKNDV(nd_nvar,m0,len,r);
4458: return r;
4459: }
4460: }
1.65 noro 4461:
1.74 noro 4462: int ox_exec_f4_red(Q proc)
4463: {
4464: Obj obj;
4465: STRING fname;
4466: NODE arg;
4467: int s;
4468: extern int ox_need_conv,ox_file_io;
4469:
4470: MKSTR(fname,"nd_exec_f4_red");
4471: arg = mknode(2,proc,fname);
4472: Pox_cmo_rpc(arg,&obj);
4473: s = get_ox_server_id(QTOS(proc));
4474: nd_write = iofp[s].out;
4475: nd_read = iofp[s].in;
4476: ox_need_conv = ox_file_io = 0;
4477: return s;
4478: }
4479:
4480: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
4481: {
4482: int nsp,nred;
4483: int i,rank,s;
4484: NODE rp,r0,r;
4485: ND_pairs sp;
4486: NM_ind_pair pair;
4487: NMV nmv;
4488: NM nm;
4489: NDV nf;
4490: Obj proc,dmy;
4491:
4492: ox_launch_main(0,0,&proc);
4493: s = ox_exec_f4_red((Q)proc);
4494:
4495: nd_send_int(m);
4496: nd_send_int(nd_nvar);
4497: nd_send_int(nd_bpe);
4498: nd_send_int(nd_wpd);
4499: nd_send_int(nmv_adv);
4500:
1.89 noro 4501: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
1.74 noro 4502:
4503: nd_send_int(nd_psn);
4504: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
4505:
4506: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4507: nd_send_int(nsp);
4508: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
4509: nd_send_int(sp->i1); nd_send_int(sp->i2);
4510: }
4511:
4512: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
4513:
4514: nred = length(rp0); nd_send_int(nred);
4515: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
4516: pair = (NM_ind_pair)BDY(rp);
4517: nd_send_int(pair->index);
4518: nd_send_intarray(pair->mul->dl,nd_wpd);
4519: }
4520: fflush(nd_write);
4521: rank = nd_recv_int();
4522: fprintf(asir_out,"rank=%d\n",rank);
4523: r0 = 0;
4524: for ( i = 0; i < rank; i++ ) {
4525: nf = nd_recv_ndv();
4526: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
4527: }
4528: Pox_shutdown(mknode(1,proc),&dmy);
4529: return r0;
4530: }
4531:
4532: /* server side */
4533:
4534: void nd_exec_f4_red_dist()
4535: {
4536: int m,i,nsp,col,s0size,nred,spcol,j,k;
4537: NM_ind_pair *rp0;
4538: NDV nf;
4539: UINT *s0vect;
4540: IndArray *imat;
4541: int *rhead;
4542: int **spmat;
4543: UINT *svect,*v;
4544: ND_pairs *sp0;
4545: int *colstat;
4546: int a,sprow,rank;
1.89 noro 4547: struct order_spec *ord;
1.74 noro 4548: Obj ordspec;
4549: ND spol;
1.76 noro 4550: int maxrs;
4551: int *spsugar;
1.74 noro 4552:
4553: nd_read = iofp[0].in;
4554: nd_write = iofp[0].out;
4555: m = nd_recv_int();
4556: nd_nvar = nd_recv_int();
4557: nd_bpe = nd_recv_int();
4558: nd_wpd = nd_recv_int();
4559: nmv_adv = nd_recv_int();
4560:
4561: loadobj(nd_read,&ordspec);
1.89 noro 4562: create_order_spec(0,ordspec,&ord);
4563: nd_init_ord(ord);
1.74 noro 4564: nd_setup_parameters(nd_nvar,0);
4565:
4566: nd_psn = nd_recv_int();
4567: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
4568: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
4569: for ( i = 0; i < nd_psn; i++ ) {
4570: nd_ps[i] = nd_recv_ndv();
4571: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
4572: }
4573:
4574: nsp = nd_recv_int();
4575: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
4576: for ( i = 0; i < nsp; i++ ) {
4577: NEWND_pairs(sp0[i]);
4578: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
4579: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
4580: }
4581:
4582: col = nd_recv_int();
4583: s0size = col*nd_wpd;
4584: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
4585: nd_recv_intarray(s0vect,s0size);
4586:
4587: nred = nd_recv_int();
4588: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
4589: for ( i = 0; i < nred; i++ ) {
4590: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
4591: rp0[i]->index = nd_recv_int();
4592: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
4593: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
4594: }
4595:
4596: spcol = col-nred;
4597: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
4598: rhead = (int *)MALLOC(col*sizeof(int));
4599: for ( i = 0; i < col; i++ ) rhead[i] = 0;
4600:
4601: /* construction of index arrays */
4602: for ( i = 0; i < nred; i++ ) {
4603: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
4604: rhead[imat[i]->head] = 1;
4605: }
4606:
4607: /* elimination (1st step) */
4608: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
4609: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 4610: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4611: for ( a = sprow = 0; a < nsp; a++ ) {
4612: nd_sp(m,0,sp0[a],&spol);
1.75 noro 4613: if ( !spol ) continue;
1.74 noro 4614: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4615: if ( m == -1 )
4616: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
4617: else
4618: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 4619: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4620: if ( i < col ) {
4621: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
4622: for ( j = k = 0; j < col; j++ )
4623: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4624: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4625: sprow++;
4626: }
1.76 noro 4627: nd_free(spol);
1.74 noro 4628: }
4629: /* elimination (2nd step) */
4630: colstat = (int *)ALLOCA(spcol*sizeof(int));
4631: if ( m == -1 )
1.76 noro 4632: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4633: else
1.76 noro 4634: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4635: nd_send_int(rank);
4636: for ( i = 0; i < rank; i++ ) {
4637: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
4638: nd_send_ndv(nf);
1.63 noro 4639: }
1.74 noro 4640: fflush(nd_write);
1.107 noro 4641: }
4642:
1.113 noro 4643: int nd_gauss_elim_q(Q **mat0,int *sugar,int row,int col,int *colstat)
1.107 noro 4644: {
1.109 noro 4645: int mod,i,j,t,c,rank,rank0,inv;
4646: int *ci,*ri;
1.113 noro 4647: Q dn;
1.109 noro 4648: MAT m,nm;
1.108 noro 4649: int **wmat;
4650:
4651: /* XXX */
4652: mod = 99999989;
4653: wmat = (int **)ALLOCA(row*sizeof(int *));
4654: for ( i = 0; i < row; i++ ) {
4655: wmat[i] = (int *)ALLOCA(col*sizeof(int));
4656: for ( j = 0; j < col; j++ ) {
1.113 noro 4657: if ( mat0[i][j] ) {
4658: t = rem(NM(mat0[i][j]),mod);
4659: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
4660: wmat[i][j] = t;
4661: } else
1.108 noro 4662: wmat[i][j] = 0;
4663: }
4664: }
1.109 noro 4665: rank0 = nd_gauss_elim_mod(wmat,sugar,row,col,mod,colstat);
4666: NEWMAT(m); m->row = row; m->col = col; m->body = (pointer **)mat0;
1.113 noro 4667: rank = generic_gauss_elim(m,&nm,&dn,&ri,&ci);
1.109 noro 4668: if ( rank != rank0 )
4669: error("afo");
4670: for ( i = 0; i < row; i++ )
4671: for ( j = 0; j < col; j++ )
4672: mat0[i][j] = 0;
4673: c = col-rank;
4674: for ( i = 0; i < rank; i++ ) {
1.113 noro 4675: mat0[i][ri[i]] = dn;
1.109 noro 4676: for ( j = 0; j < c; j++ )
1.113 noro 4677: mat0[i][ci[j]] = (Q)BDY(nm)[i][j];
1.109 noro 4678: }
1.113 noro 4679: inv = invm(rem(NM(dn),mod),mod);
4680: if ( SGN(dn) < 0 ) inv = mod-inv;
1.109 noro 4681: for ( i = 0; i < row; i++ )
4682: for ( j = 0; j < col; j++ ) {
1.113 noro 4683: if ( mat0[i][j] ) {
4684: t = rem(NM(mat0[i][j]),mod);
4685: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
4686: } else
1.109 noro 4687: t = 0;
4688: c = dmar(t,inv,0,mod);
4689: if ( wmat[i][j] != c )
4690: error("afo");
4691: }
4692: return rank;
1.76 noro 4693: }
4694:
4695: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4696: {
4697: int i,j,k,l,inv,a,rank,s;
4698: unsigned int *t,*pivot,*pk;
4699: unsigned int **mat;
4700:
4701: mat = (unsigned int **)mat0;
4702: for ( rank = 0, j = 0; j < col; j++ ) {
4703: for ( i = rank; i < row; i++ )
4704: mat[i][j] %= md;
4705: for ( i = rank; i < row; i++ )
4706: if ( mat[i][j] )
4707: break;
4708: if ( i == row ) {
4709: colstat[j] = 0;
4710: continue;
4711: } else
4712: colstat[j] = 1;
4713: if ( i != rank ) {
4714: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4715: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4716: }
4717: pivot = mat[rank];
4718: s = sugar[rank];
4719: inv = invm(pivot[j],md);
4720: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4721: if ( *pk ) {
4722: if ( *pk >= (unsigned int)md )
4723: *pk %= md;
4724: DMAR(*pk,inv,0,md,*pk)
4725: }
4726: for ( i = rank+1; i < row; i++ ) {
4727: t = mat[i];
4728: if ( a = t[j] ) {
4729: sugar[i] = MAX(sugar[i],s);
4730: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4731: }
4732: }
4733: rank++;
4734: }
4735: for ( j = col-1, l = rank-1; j >= 0; j-- )
4736: if ( colstat[j] ) {
4737: pivot = mat[l];
4738: s = sugar[l];
4739: for ( i = 0; i < l; i++ ) {
4740: t = mat[i];
4741: t[j] %= md;
4742: if ( a = t[j] ) {
4743: sugar[i] = MAX(sugar[i],s);
4744: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4745: }
4746: }
4747: l--;
4748: }
4749: for ( j = 0, l = 0; l < rank; j++ )
4750: if ( colstat[j] ) {
4751: t = mat[l];
4752: for ( k = j; k < col; k++ )
4753: if ( t[k] >= (unsigned int)md )
4754: t[k] %= md;
4755: l++;
4756: }
4757: return rank;
4758: }
4759:
4760: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4761: {
4762: int i,j,k,l,inv,a,rank,s;
4763: unsigned int *t,*pivot,*pk;
4764: unsigned int **mat;
4765:
4766: mat = (unsigned int **)mat0;
4767: for ( rank = 0, j = 0; j < col; j++ ) {
4768: for ( i = rank; i < row; i++ )
4769: if ( mat[i][j] )
4770: break;
4771: if ( i == row ) {
4772: colstat[j] = 0;
4773: continue;
4774: } else
4775: colstat[j] = 1;
4776: if ( i != rank ) {
4777: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4778: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4779: }
4780: pivot = mat[rank];
4781: s = sugar[rank];
4782: inv = _invsf(pivot[j]);
4783: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4784: if ( *pk )
4785: *pk = _mulsf(*pk,inv);
4786: for ( i = rank+1; i < row; i++ ) {
4787: t = mat[i];
4788: if ( a = t[j] ) {
4789: sugar[i] = MAX(sugar[i],s);
4790: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4791: }
4792: }
4793: rank++;
4794: }
4795: for ( j = col-1, l = rank-1; j >= 0; j-- )
4796: if ( colstat[j] ) {
4797: pivot = mat[l];
4798: s = sugar[l];
4799: for ( i = 0; i < l; i++ ) {
4800: t = mat[i];
4801: if ( a = t[j] ) {
4802: sugar[i] = MAX(sugar[i],s);
4803: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4804: }
4805: }
4806: l--;
4807: }
4808: return rank;
1.77 noro 4809: }
4810:
4811: int ndv_ishomo(NDV p)
4812: {
4813: NMV m;
4814: int len,h;
4815:
4816: if ( !p ) return 1;
4817: len = LEN(p);
4818: m = BDY(p);
4819: h = TD(DL(m));
4820: NMV_ADV(m);
4821: for ( len--; len; len--, NMV_ADV(m) )
4822: if ( TD(DL(m)) != h ) return 0;
4823: return 1;
4824: }
4825:
4826: void ndv_save(NDV p,int index)
4827: {
4828: FILE *s;
4829: char name[BUFSIZ];
4830: short id;
4831: int nv,sugar,len,n,i,td,e,j;
4832: NMV m;
4833: unsigned int *dl;
4834:
4835: sprintf(name,"%s/%d",Demand,index);
4836: s = fopen(name,"w");
4837: savevl(s,0);
4838: if ( !p ) {
4839: saveobj(s,0);
4840: return;
4841: }
4842: id = O_DP;
4843: nv = NV(p);
4844: sugar = SG(p);
4845: len = LEN(p);
4846: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
4847: write_int(s,&len);
4848:
4849: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 4850: saveobj(s,(Obj)CQ(m));
1.77 noro 4851: dl = DL(m);
4852: td = TD(dl);
4853: write_int(s,&td);
4854: for ( j = 0; j < nv; j++ ) {
4855: e = GET_EXP(dl,j);
4856: write_int(s,&e);
4857: }
4858: }
4859: fclose(s);
4860: }
4861:
4862: NDV ndv_load(int index)
4863: {
4864: FILE *s;
4865: char name[BUFSIZ];
4866: short id;
4867: int nv,sugar,len,n,i,td,e,j;
4868: NDV d;
4869: NMV m0,m;
4870: unsigned int *dl;
4871: Obj obj;
4872:
4873: sprintf(name,"%s/%d",Demand,index);
4874: s = fopen(name,"r");
4875: if ( !s ) return 0;
4876:
4877: skipvl(s);
4878: read_short(s,&id);
4879: if ( !id ) return 0;
4880: read_int(s,&nv);
4881: read_int(s,&sugar);
4882: read_int(s,&len);
4883:
4884: m0 = m = MALLOC(len*nmv_adv);
4885: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 4886: loadobj(s,&obj); CQ(m) = (Q)obj;
1.77 noro 4887: dl = DL(m);
4888: ndl_zero(dl);
4889: read_int(s,&td); TD(dl) = td;
4890: for ( j = 0; j < nv; j++ ) {
4891: read_int(s,&e);
4892: PUT_EXP(dl,j,e);
4893: }
4894: if ( nd_blockmask ) ndl_weight_mask(dl);
4895: }
4896: fclose(s);
4897: MKNDV(nv,m0,len,d);
4898: SG(d) = sugar;
4899: return d;
1.99 noro 4900: }
4901:
1.102 noro 4902: void nd_det(int mod,MAT f,P *rp)
1.99 noro 4903: {
4904: VL fv,tv;
1.102 noro 4905: int n,i,j,max,e,nvar,sgn,k0,l0,len0,len,k,l,a;
1.99 noro 4906: pointer **m;
1.113 noro 4907: Q mone;
1.102 noro 4908: NDV **dm;
4909: NDV *t,*mi,*mj;
4910: NDV d,s,mij,mjj;
4911: ND u;
4912: NMV nmv;
1.114 noro 4913: UINT *bound;
1.99 noro 4914: PGeoBucket bucket;
4915: struct order_spec *ord;
4916:
4917: create_order_spec(0,0,&ord);
4918: nd_init_ord(ord);
4919: get_vars((Obj)f,&fv);
4920: if ( f->row != f->col )
4921: error("nd_det : non-square matrix");
4922: n = f->row;
4923: for ( nvar = 0, tv = fv; tv; tv = NEXT(tv), nvar++ );
4924: m = f->body;
4925: for ( i = 0, max = 0; i < n; i++ )
4926: for ( j = 0; j < n; j++ )
4927: for ( tv = fv; tv; tv = NEXT(tv) ) {
4928: e = getdeg(tv->v,(P)m[i][j]);
4929: max = MAX(e,max);
4930: }
1.114 noro 4931: nd_setup_parameters(nvar,max);
1.102 noro 4932: dm = (NDV **)almat_pointer(n,n);
1.99 noro 4933: for ( i = 0, max = 0; i < n; i++ )
1.102 noro 4934: for ( j = 0; j < n; j++ ) {
4935: dm[i][j] = ptondv(CO,fv,m[i][j]);
4936: if ( mod ) ndv_mod(mod,dm[i][j]);
4937: if ( dm[i][j] && !LEN(dm[i][j]) ) dm[i][j] = 0;
4938: }
4939: d = ptondv(CO,fv,(P)ONE);
4940: if ( mod ) ndv_mod(mod,d);
1.113 noro 4941: chsgnq(ONE,&mone);
1.99 noro 4942: for ( j = 0, sgn = 1; j < n; j++ ) {
1.104 noro 4943: if ( DP_Print ) fprintf(stderr,"j=%d\n",j);
1.99 noro 4944: for ( i = j; i < n && !dm[i][j]; i++ );
4945: if ( i == n ) {
4946: *rp = 0;
4947: return;
4948: }
1.102 noro 4949: k0 = i; l0 = j; len0 = LEN(dm[k0][l0]);
1.99 noro 4950: for ( k = j; k < n; k++ )
4951: for ( l = j; l < n; l++ )
1.102 noro 4952: if ( dm[k][l] && LEN(dm[k][l]) < len0 ) {
4953: k0 = k; l0 = l; len0 = LEN(dm[k][l]);
1.99 noro 4954: }
4955: if ( k0 != j ) {
4956: t = dm[j]; dm[j] = dm[k0]; dm[k0] = t;
4957: sgn = -sgn;
4958: }
4959: if ( l0 != j ) {
4960: for ( k = j; k < n; k++ ) {
4961: s = dm[k][j]; dm[k][j] = dm[k][l0]; dm[k][l0] = s;
4962: }
4963: sgn = -sgn;
4964: }
1.114 noro 4965: bound = nd_det_compute_bound(dm,n,j);
4966: if ( ndl_check_bound(bound,bound) )
4967: nd_det_reconstruct(dm,n,j,d);
4968:
1.99 noro 4969: for ( i = j+1, mj = dm[j], mjj = mj[j]; i < n; i++ ) {
1.114 noro 4970: /* if ( DP_Print ) fprintf(stderr," i=%d\n ",i); */
1.99 noro 4971: mi = dm[i]; mij = mi[j];
1.102 noro 4972: if ( mod )
4973: ndv_mul_c(mod,mij,mod-1);
4974: else
1.113 noro 4975: ndv_mul_c_q(mij,mone);
1.99 noro 4976: for ( k = j+1; k < n; k++ ) {
1.114 noro 4977: /* if ( DP_Print ) fprintf(stderr,"k=%d ",k); */
1.99 noro 4978: bucket = create_pbucket();
1.104 noro 4979: if ( mi[k] ) {
1.102 noro 4980: nmv = BDY(mjj); len = LEN(mjj);
4981: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
4982: u = ndv_mul_nmv_trunc(mod,nmv,mi[k],DL(BDY(d)));
4983: add_pbucket(mod,bucket,u);
1.99 noro 4984: }
1.104 noro 4985: }
1.99 noro 4986: if ( mj[k] && mij ) {
1.102 noro 4987: nmv = BDY(mij); len = LEN(mij);
4988: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
4989: u = ndv_mul_nmv_trunc(mod,nmv,mj[k],DL(BDY(d)));
4990: add_pbucket(mod,bucket,u);
1.99 noro 4991: }
4992: }
1.104 noro 4993: u = nd_quo(mod,bucket,d);
1.102 noro 4994: mi[k] = ndtondv(mod,u);
1.99 noro 4995: }
1.114 noro 4996: /* if ( DP_Print ) fprintf(stderr,"\n",k); */
1.99 noro 4997: }
4998: d = mjj;
4999: }
1.100 noro 5000: if ( sgn < 0 )
1.102 noro 5001: if ( mod )
5002: ndv_mul_c(mod,d,mod-1);
5003: else
1.113 noro 5004: ndv_mul_c_q(d,mone);
1.102 noro 5005: *rp = ndvtop(mod,CO,fv,d);
1.99 noro 5006: }
5007:
1.102 noro 5008: ND ndv_mul_nmv_trunc(int mod,NMV m0,NDV p,UINT *d)
1.99 noro 5009: {
5010: NM mr,mr0;
1.102 noro 5011: NM tnm;
5012: NMV m;
1.99 noro 5013: UINT *d0,*dt,*dm;
5014: int c,n,td,i,c1,c2,len;
1.113 noro 5015: Q q;
1.99 noro 5016: ND r;
5017:
5018: if ( !p ) return 0;
5019: else {
1.102 noro 5020: n = NV(p); m = BDY(p); len = LEN(p);
1.99 noro 5021: d0 = DL(m0);
1.102 noro 5022: td = TD(d);
1.99 noro 5023: mr0 = 0;
5024: NEWNM(tnm);
1.102 noro 5025: if ( mod ) {
5026: c = CM(m0);
5027: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5028: ndl_add(DL(m),d0,DL(tnm));
5029: if ( ndl_reducible(DL(tnm),d) ) {
5030: NEXTNM(mr0,mr);
5031: c1 = CM(m); DMAR(c1,c,0,mod,c2); CM(mr) = c2;
1.104 noro 5032: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5033: }
5034: }
5035: } else {
1.113 noro 5036: q = CQ(m0);
1.102 noro 5037: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5038: ndl_add(DL(m),d0,DL(tnm));
5039: if ( ndl_reducible(DL(tnm),d) ) {
5040: NEXTNM(mr0,mr);
1.113 noro 5041: mulq(CQ(m),q,&CQ(mr));
1.104 noro 5042: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5043: }
1.99 noro 5044: }
5045: }
5046: if ( !mr0 )
5047: return 0;
5048: else {
5049: NEXT(mr) = 0;
1.105 noro 5050: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
1.99 noro 5051: MKND(NV(p),mr0,len,r);
5052: SG(r) = SG(p) + TD(d0);
5053: return r;
5054: }
5055: }
1.114 noro 5056: }
5057:
5058: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d)
5059: {
5060: int i,obpe,oadv,h,k,l;
5061: static NM prev_nm_free_list;
5062: EPOS oepos;
5063:
5064: obpe = nd_bpe;
5065: oadv = nmv_adv;
5066: oepos = nd_epos;
5067: if ( obpe < 2 ) nd_bpe = 2;
5068: else if ( obpe < 3 ) nd_bpe = 3;
5069: else if ( obpe < 4 ) nd_bpe = 4;
5070: else if ( obpe < 5 ) nd_bpe = 5;
5071: else if ( obpe < 6 ) nd_bpe = 6;
5072: else if ( obpe < 8 ) nd_bpe = 8;
5073: else if ( obpe < 10 ) nd_bpe = 10;
5074: else if ( obpe < 16 ) nd_bpe = 16;
5075: else if ( obpe < 32 ) nd_bpe = 32;
5076: else error("nd_det_reconstruct : exponent too large");
5077:
5078: nd_setup_parameters(nd_nvar,0);
5079: prev_nm_free_list = _nm_free_list;
5080: _nm_free_list = 0;
5081: for ( k = j; k < n; k++ )
5082: for (l = j; l < n; l++ )
5083: ndv_realloc(dm[k][l],obpe,oadv,oepos);
5084: ndv_realloc(d,obpe,oadv,oepos);
5085: prev_nm_free_list = 0;
5086: #if 0
5087: GC_gcollect();
5088: #endif
5089: }
5090:
5091: UINT *nd_det_compute_bound(NDV **dm,int n,int j)
5092: {
5093: UINT *d0,*d1,*d,*t,*r;
5094: int k,l;
5095:
5096: d0 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5097: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5098: for ( k = 0; k < nd_wpd; k++ ) d0[k] = 0;
5099: for ( k = j; k < n; k++ )
5100: for ( l = j; l < n; l++ )
5101: if ( dm[k][l] ) {
5102: d = ndv_compute_bound(dm[k][l]);
5103: ndl_lcm(d,d0,d1);
5104: t = d1; d1 = d0; d0 = t;
5105: }
5106: r = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5107: for ( k = 0; k < nd_wpd; k++ ) r[k] = d0[k];
5108: return r;
1.117 ! noro 5109: }
! 5110:
! 5111: DL nd_separate_d(UINT *d,UINT *trans)
! 5112: {
! 5113: int n,ntrans,td,i,e;
! 5114: DL a;
! 5115:
! 5116: n = nd_nvar; ntrans = n-nd_nalg;
! 5117: ndl_zero(trans);
! 5118: td = 0;
! 5119: for ( i = 0; i < ntrans; i++ ) {
! 5120: e = GET_EXP(d,i);
! 5121: PUT_EXP(trans,i,e);
! 5122: td += MUL_WEIGHT(e,i);
! 5123: }
! 5124: TD(trans) = td;
! 5125: if ( nd_blockmask) ndl_weight_mask(trans);
! 5126: NEWDL(a,nd_nalg);
! 5127: td = 0;
! 5128: for ( ; i < n; i++ ) {
! 5129: e = GET_EXP(d,i);
! 5130: a->d[i-ntrans] = e;
! 5131: td += e;
! 5132: }
! 5133: a->td = td;
! 5134: return a;
! 5135: }
! 5136:
! 5137: ND nd_pseudo_monic(int mod,ND p)
! 5138: {
! 5139: UINT *trans,*t;
! 5140: DL alg;
! 5141: MP mp0,mp;
! 5142: NM m,m0,m1;
! 5143: DL dl;
! 5144: DP nm;
! 5145: NDV ndv;
! 5146: DAlg lc,inv;
! 5147: ND s,c;
! 5148: int n,ntrans,i,e,td;
! 5149:
! 5150: n = nd_nvar; ntrans = n-nd_nalg;
! 5151: NEWNM(m0);
! 5152: NEWNM(m1);
! 5153: alg = nd_separate_d(HDL(p),DL(m0));
! 5154: mp0 = 0; NEXTMP(mp0,mp); mp->c = (P)HCQ(p); mp->dl = alg;
! 5155: if ( !mp->dl->td )
! 5156: return p;
! 5157: for ( m = NEXT(BDY(p)); m; m = NEXT(m) ) {
! 5158: alg = nd_separate_d(DL(m),DL(m1));
! 5159: if ( !ndl_equal(DL(m0),DL(m1)) )
! 5160: break;
! 5161: NEXTMP(mp0,mp); mp->c = (P)CQ(m); mp->dl = alg;
! 5162: }
! 5163: NEXT(mp) = 0;
! 5164: MKDP(nd_nalg,mp0,nm);
! 5165: MKDAlg(nm,ONE,lc);
! 5166: invdalg(lc,&inv);
! 5167: ndv = ndtondv(0,p);
! 5168: for ( s = 0, mp = BDY(inv->nm); mp; mp = NEXT(mp) ) {
! 5169: CQ(m0) = (Q)mp->c;
! 5170: dl = mp->dl;
! 5171: for ( td = 0, i = ntrans; i < n; i++ ) {
! 5172: e = dl->d[i-ntrans];
! 5173: ndl_zero(DL(m0));
! 5174: PUT_EXP(DL(m0),i,e);
! 5175: td += MUL_WEIGHT(e,i);
! 5176: }
! 5177: TD(DL(m0)) = td;
! 5178: if ( nd_blockmask) ndl_weight_mask(trans);
! 5179: s = nd_add(0,s,ndv_mul_nm(0,m0,ndv));
! 5180: }
! 5181: ndv_free(ndv);
! 5182: return s;
1.59 noro 5183: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>