Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.128
1.128 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.127 2005/08/03 05:01:01 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:
1.121 noro 11: static int nd_ntrans;
1.117 noro 12: static int nd_nalg;
1.103 noro 13: #if 0
1.74 noro 14: static int ndv_alloc;
1.103 noro 15: #endif
1.87 noro 16: #if 1
1.69 noro 17: static int nd_f4_nsp=0x7fffffff;
1.87 noro 18: #else
19: static int nd_f4_nsp=50;
20: #endif
1.42 noro 21: static double nd_scale=2;
1.61 noro 22: static UINT **nd_bound;
1.42 noro 23: static struct order_spec *nd_ord;
24: static EPOS nd_epos;
1.43 noro 25: static BlockMask nd_blockmask;
1.42 noro 26: static int nd_nvar;
27: static int nd_isrlex;
28: static int nd_epw,nd_bpe,nd_wpd,nd_exporigin;
1.61 noro 29: static UINT nd_mask[32];
30: static UINT nd_mask0,nd_mask1;
1.42 noro 31:
1.20 noro 32: static NDV *nd_ps;
1.53 noro 33: static NDV *nd_ps_trace;
1.42 noro 34: static RHist *nd_psh;
35: static int nd_psn,nd_pslen;
36: static RHist *nd_red;
1.96 noro 37: static int *nd_work_vector;
38: static int **nd_matrix;
39: static int nd_matrix_len;
1.97 noro 40: static struct weight_or_block *nd_worb;
41: static int nd_worb_len;
1.42 noro 42: static int nd_found,nd_create,nd_notfirst;
43: static int nmv_adv;
1.77 noro 44: static int nd_demand;
1.1 noro 45:
1.119 noro 46: NumberField get_numberfield();
1.114 noro 47: UINT *nd_det_compute_bound(NDV **dm,int n,int j);
48: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d);
1.118 noro 49: int nd_monic(int m,ND *p);
1.114 noro 50:
1.1 noro 51: void nd_free_private_storage()
52: {
53: _nm_free_list = 0;
1.5 noro 54: _ndp_free_list = 0;
1.71 noro 55: #if 0
1.1 noro 56: GC_gcollect();
1.71 noro 57: #endif
1.1 noro 58: }
59:
60: void _NM_alloc()
61: {
62: NM p;
63: int i;
64:
1.11 noro 65: for ( i = 0; i < 1024; i++ ) {
1.61 noro 66: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 67: p->next = _nm_free_list; _nm_free_list = p;
68: }
69: }
70:
71: void _ND_alloc()
72: {
73: ND p;
74: int i;
75:
76: for ( i = 0; i < 1024; i++ ) {
77: p = (ND)GC_malloc(sizeof(struct oND));
78: p->body = (NM)_nd_free_list; _nd_free_list = p;
79: }
80: }
81:
82: void _NDP_alloc()
83: {
84: ND_pairs p;
85: int i;
86:
1.11 noro 87: for ( i = 0; i < 1024; i++ ) {
1.1 noro 88: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
1.61 noro 89: +(nd_wpd-1)*sizeof(UINT));
1.1 noro 90: p->next = _ndp_free_list; _ndp_free_list = p;
91: }
92: }
93:
1.30 noro 94: INLINE int nd_length(ND p)
1.1 noro 95: {
96: NM m;
97: int i;
98:
99: if ( !p )
100: return 0;
101: else {
102: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
103: return i;
104: }
105: }
106:
1.61 noro 107: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 108: {
1.61 noro 109: UINT u1,u2;
1.1 noro 110: int i,j;
111:
1.34 noro 112: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 113: #if USE_UNROLL
1.1 noro 114: switch ( nd_bpe ) {
1.62 noro 115: case 3:
116: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
117: u1 = d1[i]; u2 = d2[i];
118: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
119: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
120: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
121: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
122: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
123: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
124: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
125: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
126: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
127: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
128: }
129: return 1;
130: break;
1.1 noro 131: case 4:
1.41 noro 132: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 133: u1 = d1[i]; u2 = d2[i];
134: if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0;
1.62 noro 135: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
136: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
137: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
138: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
139: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
140: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
141: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
1.1 noro 142: }
143: return 1;
144: break;
145: case 6:
1.41 noro 146: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 147: u1 = d1[i]; u2 = d2[i];
148: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
1.62 noro 149: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
150: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
151: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
152: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
1.1 noro 153: }
154: return 1;
155: break;
156: case 8:
1.41 noro 157: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 158: u1 = d1[i]; u2 = d2[i];
159: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
1.62 noro 160: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
161: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
162: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
1.1 noro 163: }
164: return 1;
165: break;
166: case 16:
1.41 noro 167: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 168: u1 = d1[i]; u2 = d2[i];
169: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
1.62 noro 170: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
1.1 noro 171: }
172: return 1;
173: break;
174: case 32:
1.41 noro 175: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 176: if ( d1[i] < d2[i] ) return 0;
177: return 1;
178: break;
179: default:
1.41 noro 180: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 181: u1 = d1[i]; u2 = d2[i];
182: for ( j = 0; j < nd_epw; j++ )
183: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
184: }
185: return 1;
186: }
1.65 noro 187: #else
188: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
189: u1 = d1[i]; u2 = d2[i];
190: for ( j = 0; j < nd_epw; j++ )
191: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
192: }
193: return 1;
194: #endif
1.1 noro 195: }
196:
1.61 noro 197: /*
198: * If the current order is a block order,
199: * then the last block is length 1 and contains
200: * the homo variable. Otherwise, the original
201: * order is either 0 or 2.
202: */
203:
204: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 205: {
1.61 noro 206: int w,i,e,n,omask0;
207:
208: omask0 = (1<<obpe)-1;
209: n = nd_nvar-1;
210: ndl_zero(r);
211: for ( i = 0; i < n; i++ ) {
212: e = GET_EXP_OLD(d,i);
213: PUT_EXP(r,i,e);
214: }
215: w = TD(d);
216: PUT_EXP(r,nd_nvar-1,weight-w);
217: TD(r) = weight;
218: if ( nd_blockmask ) ndl_weight_mask(r);
219: }
220:
221: void ndl_dehomogenize(UINT *d)
222: {
223: UINT mask;
224: UINT h;
1.31 noro 225: int i,bits;
1.23 noro 226:
1.44 noro 227: if ( nd_blockmask ) {
228: h = GET_EXP(d,nd_nvar-1);
1.45 noro 229: XOR_EXP(d,nd_nvar-1,h);
1.44 noro 230: TD(d) -= h;
231: d[nd_exporigin-1] -= h;
232: } else {
233: if ( nd_isrlex ) {
234: if ( nd_bpe == 32 ) {
235: h = d[nd_exporigin];
236: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
237: d[i-1] = d[i];
238: d[i-1] = 0;
239: TD(d) -= h;
240: } else {
241: bits = nd_epw*nd_bpe;
242: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
243: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
244: for ( i = nd_exporigin; i < nd_wpd; i++ )
245: d[i] = ((d[i]<<nd_bpe)&mask)
246: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
247: TD(d) -= h;
248: }
1.45 noro 249: } else {
250: h = GET_EXP(d,nd_nvar-1);
251: XOR_EXP(d,nd_nvar-1,h);
252: TD(d) -= h;
253: }
1.44 noro 254: }
1.23 noro 255: }
256:
1.61 noro 257: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 258: {
1.61 noro 259: UINT t1,t2,u,u1,u2;
1.43 noro 260: int i,j,l;
1.1 noro 261:
1.65 noro 262: #if USE_UNROLL
1.1 noro 263: switch ( nd_bpe ) {
1.62 noro 264: case 3:
265: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
266: u1 = d1[i]; u2 = d2[i];
267: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
268: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
269: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
270: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
271: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
272: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
273: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
274: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
275: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
276: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
277: d[i] = u;
278: }
279: break;
1.1 noro 280: case 4:
1.41 noro 281: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 282: u1 = d1[i]; u2 = d2[i];
283: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
1.62 noro 284: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
285: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
286: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
287: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
288: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
289: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
290: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
1.1 noro 291: d[i] = u;
292: }
293: break;
294: case 6:
1.41 noro 295: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 296: u1 = d1[i]; u2 = d2[i];
297: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
1.62 noro 298: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
299: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
300: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
301: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
1.1 noro 302: d[i] = u;
303: }
304: break;
305: case 8:
1.41 noro 306: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 307: u1 = d1[i]; u2 = d2[i];
308: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
1.62 noro 309: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
310: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
311: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
1.1 noro 312: d[i] = u;
313: }
314: break;
315: case 16:
1.41 noro 316: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 317: u1 = d1[i]; u2 = d2[i];
318: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
1.62 noro 319: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
1.1 noro 320: d[i] = u;
321: }
322: break;
323: case 32:
1.41 noro 324: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 325: u1 = d1[i]; u2 = d2[i];
326: d[i] = u1>u2?u1:u2;
327: }
328: break;
329: default:
1.41 noro 330: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 331: u1 = d1[i]; u2 = d2[i];
332: for ( j = 0, u = 0; j < nd_epw; j++ ) {
333: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
334: }
335: d[i] = u;
336: }
337: break;
338: }
1.65 noro 339: #else
340: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
341: u1 = d1[i]; u2 = d2[i];
342: for ( j = 0, u = 0; j < nd_epw; j++ ) {
343: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
344: }
345: d[i] = u;
346: }
347: #endif
1.39 noro 348: TD(d) = ndl_weight(d);
1.61 noro 349: if ( nd_blockmask ) ndl_weight_mask(d);
1.57 noro 350: }
351:
1.61 noro 352: int ndl_weight(UINT *d)
1.1 noro 353: {
1.61 noro 354: UINT t,u;
1.1 noro 355: int i,j;
356:
1.60 noro 357: if ( current_dl_weight_vector )
358: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
359: u = GET_EXP(d,i);
360: t += MUL_WEIGHT(u,i);
361: }
362: else
363: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
364: u = d[i];
365: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
366: t += (u&nd_mask0);
367: }
1.1 noro 368: return t;
369: }
370:
1.61 noro 371: void ndl_weight_mask(UINT *d)
1.43 noro 372: {
1.61 noro 373: UINT t,u;
374: UINT *mask;
375: int i,j,k,l;
1.43 noro 376:
1.61 noro 377: l = nd_blockmask->n;
378: for ( k = 0; k < l; k++ ) {
379: mask = nd_blockmask->mask[k];
380: if ( current_dl_weight_vector )
381: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
382: u = GET_EXP_MASK(d,i,mask);
383: t += MUL_WEIGHT(u,i);
384: }
385: else
386: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
387: u = d[i]&mask[i];
388: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
389: t += (u&nd_mask0);
390: }
391: d[k+1] = t;
392: }
1.43 noro 393: }
394:
1.61 noro 395: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 396: {
397: int i;
398:
1.41 noro 399: d1 += nd_exporigin;
400: d2 += nd_exporigin;
401: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
1.1 noro 402: if ( *d1 > *d2 )
1.32 noro 403: return nd_isrlex ? -1 : 1;
1.1 noro 404: else if ( *d1 < *d2 )
1.32 noro 405: return nd_isrlex ? 1 : -1;
1.1 noro 406: return 0;
407: }
408:
1.61 noro 409: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 410: {
411: int i,l,j,ord_o,ord_l;
412: struct order_pair *op;
1.61 noro 413: UINT t1,t2,m;
414: UINT *mask;
1.43 noro 415:
416: l = nd_blockmask->n;
417: op = nd_blockmask->order_pair;
418: for ( j = 0; j < l; j++ ) {
419: mask = nd_blockmask->mask[j];
420: ord_o = op[j].order;
421: if ( ord_o < 2 )
1.44 noro 422: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
423: else if ( t1 < t2 ) return -1;
1.43 noro 424: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.44 noro 425: m = mask[i];
426: t1 = d1[i]&m;
427: t2 = d2[i]&m;
1.43 noro 428: if ( t1 > t2 )
429: return !ord_o ? -1 : 1;
430: else if ( t1 < t2 )
431: return !ord_o ? 1 : -1;
432: }
433: }
434: return 0;
435: }
436:
1.96 noro 437: int ndl_matrix_compare(UINT *d1,UINT *d2)
438: {
439: int i,j,s;
440: int *v;
441:
442: for ( j = 0; j < nd_nvar; j++ )
443: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
444: for ( i = 0; i < nd_matrix_len; i++ ) {
445: v = nd_matrix[i];
446: for ( j = 0, s = 0; j < nd_nvar; j++ )
447: s += v[j]*nd_work_vector[j];
448: if ( s > 0 ) return 1;
449: else if ( s < 0 ) return -1;
450: }
451: return 0;
452: }
453:
1.97 noro 454: int ndl_composite_compare(UINT *d1,UINT *d2)
455: {
456: int i,j,s,start,end,len,o;
457: int *v;
458: struct sparse_weight *sw;
459:
460: for ( j = 0; j < nd_nvar; j++ )
461: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
462: for ( i = 0; i < nd_worb_len; i++ ) {
463: len = nd_worb[i].length;
464: switch ( nd_worb[i].type ) {
465: case IS_DENSE_WEIGHT:
466: v = nd_worb[i].body.dense_weight;
467: for ( j = 0, s = 0; j < len; j++ )
468: s += v[j]*nd_work_vector[j];
469: if ( s > 0 ) return 1;
470: else if ( s < 0 ) return -1;
471: break;
472: case IS_SPARSE_WEIGHT:
473: sw = nd_worb[i].body.sparse_weight;
474: for ( j = 0, s = 0; j < len; j++ )
475: s += sw[j].value*nd_work_vector[sw[j].pos];
476: if ( s > 0 ) return 1;
477: else if ( s < 0 ) return -1;
478: break;
479: case IS_BLOCK:
480: o = nd_worb[i].body.block.order;
481: start = nd_worb[i].body.block.start;
482: switch ( o ) {
483: case 0:
484: end = start+len;
485: for ( j = start, s = 0; j < end; j++ )
486: s += MUL_WEIGHT(nd_work_vector[j],j);
487: if ( s > 0 ) return 1;
488: else if ( s < 0 ) return -1;
489: for ( j = end-1; j >= start; j-- )
490: if ( nd_work_vector[j] < 0 ) return 1;
491: else if ( nd_work_vector[j] > 0 ) return -1;
492: break;
493: case 1:
494: end = start+len;
495: for ( j = start, s = 0; j < end; j++ )
496: s += MUL_WEIGHT(nd_work_vector[j],j);
497: if ( s > 0 ) return 1;
498: else if ( s < 0 ) return -1;
499: for ( j = start; j < end; j++ )
500: if ( nd_work_vector[j] > 0 ) return 1;
501: else if ( nd_work_vector[j] < 0 ) return -1;
502: break;
503: case 2:
504: for ( j = start; j < end; j++ )
505: if ( nd_work_vector[j] > 0 ) return 1;
506: else if ( nd_work_vector[j] < 0 ) return -1;
507: break;
508: }
509: break;
510: }
511: }
512: return 0;
513: }
514:
1.58 noro 515: /* TDH -> WW -> TD-> RL */
516:
1.61 noro 517: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 518: {
519: int i,m,e1,e2;
520:
521: if ( TD(d1) > TD(d2) ) return 1;
522: else if ( TD(d1) < TD(d2) ) return -1;
523: m = nd_nvar>>1;
524: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
525: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
526: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
527: }
528: if ( e1 > e2 ) return 1;
529: else if ( e1 < e2 ) return -1;
530: return ndl_lex_compare(d1,d2);
531: }
532:
1.61 noro 533: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 534: {
535: int i;
536:
1.81 noro 537: switch ( nd_wpd ) {
538: case 2:
539: if ( TD(d2) != TD(d1) ) return 0;
540: if ( d2[1] != d1[1] ) return 0;
541: return 1;
542: break;
543: case 3:
544: if ( TD(d2) != TD(d1) ) return 0;
545: if ( d2[1] != d1[1] ) return 0;
546: if ( d2[2] != d1[2] ) return 0;
547: return 1;
548: break;
549: default:
550: for ( i = 0; i < nd_wpd; i++ )
551: if ( *d1++ != *d2++ ) return 0;
552: return 1;
553: break;
554: }
1.1 noro 555: }
556:
1.61 noro 557: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 558: {
559: int i;
560:
561: switch ( nd_wpd ) {
1.41 noro 562: case 2:
1.34 noro 563: TD(d2) = TD(d1);
564: d2[1] = d1[1];
1.6 noro 565: break;
1.41 noro 566: case 3:
1.34 noro 567: TD(d2) = TD(d1);
1.6 noro 568: d2[1] = d1[1];
1.34 noro 569: d2[2] = d1[2];
1.6 noro 570: break;
571: default:
1.41 noro 572: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 573: d2[i] = d1[i];
574: break;
575: }
576: }
577:
1.61 noro 578: INLINE void ndl_zero(UINT *d)
579: {
580: int i;
581: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
582: }
583:
584: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 585: {
586: int i;
587:
1.43 noro 588: #if 1
1.6 noro 589: switch ( nd_wpd ) {
1.41 noro 590: case 2:
591: TD(d) = TD(d1)+TD(d2);
1.34 noro 592: d[1] = d1[1]+d2[1];
1.6 noro 593: break;
1.41 noro 594: case 3:
595: TD(d) = TD(d1)+TD(d2);
1.6 noro 596: d[1] = d1[1]+d2[1];
1.34 noro 597: d[2] = d1[2]+d2[2];
1.6 noro 598: break;
599: default:
1.43 noro 600: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 601: break;
602: }
1.43 noro 603: #else
604: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
605: #endif
1.6 noro 606: }
607:
1.55 noro 608: /* d1 += d2 */
1.61 noro 609: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 610: {
611: int i;
612:
613: #if 1
614: switch ( nd_wpd ) {
615: case 2:
616: TD(d1) += TD(d2);
617: d1[1] += d2[1];
618: break;
619: case 3:
620: TD(d1) += TD(d2);
621: d1[1] += d2[1];
622: d1[2] += d2[2];
623: break;
624: default:
625: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
626: break;
627: }
628: #else
629: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
630: #endif
631: }
632:
1.61 noro 633: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 634: {
635: int i;
636:
1.43 noro 637: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 638: }
639:
1.61 noro 640: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 641: {
1.61 noro 642: UINT t1,t2,u,u1,u2;
1.1 noro 643: int i,j;
644:
1.65 noro 645: #if USE_UNROLL
1.1 noro 646: switch ( nd_bpe ) {
1.62 noro 647: case 3:
648: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
649: u1 = d1[i]; u2 = d2[i];
650: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
651: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
652: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
653: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
654: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
655: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
656: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
657: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
658: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
659: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
660: }
661: return 1;
662: break;
1.1 noro 663: case 4:
1.41 noro 664: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 665: u1 = d1[i]; u2 = d2[i];
666: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 667: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
668: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
669: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
670: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
671: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
672: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
673: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 674: }
675: return 1;
676: break;
677: case 6:
1.41 noro 678: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 679: u1 = d1[i]; u2 = d2[i];
680: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 681: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
682: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
683: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
684: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 685: }
686: return 1;
687: break;
688: case 8:
1.41 noro 689: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 690: u1 = d1[i]; u2 = d2[i];
691: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 692: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
693: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
694: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 695: }
696: return 1;
697: break;
698: case 16:
1.41 noro 699: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 700: u1 = d1[i]; u2 = d2[i];
701: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 702: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 703: }
704: return 1;
705: break;
706: case 32:
1.41 noro 707: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 708: if ( d1[i] && d2[i] ) return 0;
709: return 1;
710: break;
711: default:
1.41 noro 712: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 713: u1 = d1[i]; u2 = d2[i];
714: for ( j = 0; j < nd_epw; j++ ) {
715: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
716: u1 >>= nd_bpe; u2 >>= nd_bpe;
717: }
718: }
719: return 1;
720: break;
721: }
1.65 noro 722: #else
723: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
724: u1 = d1[i]; u2 = d2[i];
725: for ( j = 0; j < nd_epw; j++ ) {
726: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
727: u1 >>= nd_bpe; u2 >>= nd_bpe;
728: }
729: }
730: return 1;
731: #endif
1.1 noro 732: }
733:
1.114 noro 734: int ndl_check_bound(UINT *d1,UINT *d2)
1.1 noro 735: {
1.61 noro 736: UINT u2;
1.5 noro 737: int i,j,ind,k;
1.1 noro 738:
1.5 noro 739: ind = 0;
1.65 noro 740: #if USE_UNROLL
1.5 noro 741: switch ( nd_bpe ) {
1.62 noro 742: case 3:
743: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
744: u2 = d2[i];
745: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
746: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
747: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
748: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
749: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
750: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
751: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
752: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
753: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
754: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
755: }
756: return 0;
757: break;
1.5 noro 758: case 4:
1.41 noro 759: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 760: u2 = d2[i];
761: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
762: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
763: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
764: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
765: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
766: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
767: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
768: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
769: }
770: return 0;
771: break;
772: case 6:
1.41 noro 773: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 774: u2 = d2[i];
775: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
776: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
777: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
778: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
779: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
780: }
781: return 0;
782: break;
783: case 8:
1.41 noro 784: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 785: u2 = d2[i];
786: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
787: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
788: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
789: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
790: }
791: return 0;
792: break;
793: case 16:
1.41 noro 794: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 795: u2 = d2[i];
796: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
797: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
798: }
799: return 0;
800: break;
801: case 32:
1.41 noro 802: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 803: if ( d1[i]+d2[i]<d1[i] ) return 1;
804: return 0;
805: break;
806: default:
1.41 noro 807: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 808: u2 = d2[i];
809: k = (nd_epw-1)*nd_bpe;
810: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
811: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
812: }
813: return 0;
814: break;
815: }
1.65 noro 816: #else
817: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
818: u2 = d2[i];
819: k = (nd_epw-1)*nd_bpe;
820: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
821: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
822: }
823: return 0;
824: #endif
1.1 noro 825: }
826:
1.114 noro 827: int ndl_check_bound2(int index,UINT *d2)
828: {
829: return ndl_check_bound(nd_bound[index],d2);
830: }
831:
1.61 noro 832: INLINE int ndl_hash_value(UINT *d)
1.1 noro 833: {
834: int i;
835: int r;
836:
1.34 noro 837: r = 0;
1.41 noro 838: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 839: r = ((r<<16)+d[i])%REDTAB_LEN;
840: return r;
841: }
842:
1.63 noro 843: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 844: {
1.13 noro 845: RHist r;
1.6 noro 846: int d,k,i;
1.1 noro 847:
1.63 noro 848: d = ndl_hash_value(dg);
1.13 noro 849: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 850: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 851: if ( k > 0 ) nd_notfirst++;
852: nd_found++;
1.13 noro 853: return r->index;
1.1 noro 854: }
855: }
1.13 noro 856: if ( Reverse )
857: for ( i = nd_psn-1; i >= 0; i-- ) {
858: r = nd_psh[i];
1.34 noro 859: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 860: nd_create++;
1.34 noro 861: nd_append_red(dg,i);
1.13 noro 862: return i;
863: }
864: }
865: else
866: for ( i = 0; i < nd_psn; i++ ) {
867: r = nd_psh[i];
1.34 noro 868: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 869: nd_create++;
1.34 noro 870: nd_append_red(dg,i);
1.13 noro 871: return i;
872: }
1.1 noro 873: }
1.6 noro 874: return -1;
1.1 noro 875: }
876:
1.63 noro 877: ND nd_merge(ND p1,ND p2)
878: {
879: int n,c;
880: int t,can,td1,td2;
881: ND r;
882: NM m1,m2,mr0,mr,s;
883:
884: if ( !p1 ) return p2;
885: else if ( !p2 ) return p1;
886: else {
887: can = 0;
888: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
889: c = DL_COMPARE(DL(m1),DL(m2));
890: switch ( c ) {
891: case 0:
892: s = m1; m1 = NEXT(m1);
893: can++; NEXTNM2(mr0,mr,s);
894: s = m2; m2 = NEXT(m2); FREENM(s);
895: break;
896: case 1:
897: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
898: break;
899: case -1:
900: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
901: break;
902: }
903: }
904: if ( !mr0 )
905: if ( m1 ) mr0 = m1;
906: else if ( m2 ) mr0 = m2;
907: else return 0;
908: else if ( m1 ) NEXT(mr) = m1;
909: else if ( m2 ) NEXT(mr) = m2;
910: else NEXT(mr) = 0;
911: BDY(p1) = mr0;
912: SG(p1) = MAX(SG(p1),SG(p2));
913: LEN(p1) = LEN(p1)+LEN(p2)-can;
914: FREEND(p2);
915: return p1;
916: }
917: }
918:
1.31 noro 919: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 920: {
921: int n,c;
1.34 noro 922: int t,can,td1,td2;
1.1 noro 923: ND r;
924: NM m1,m2,mr0,mr,s;
925:
1.34 noro 926: if ( !p1 ) return p2;
927: else if ( !p2 ) return p1;
1.71 noro 928: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1.113 noro 929: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 930: else {
1.30 noro 931: can = 0;
1.1 noro 932: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 933: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 934: switch ( c ) {
935: case 0:
1.19 noro 936: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 937: if ( t < 0 ) t += mod;
1.1 noro 938: s = m1; m1 = NEXT(m1);
939: if ( t ) {
1.34 noro 940: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.17 noro 941: } else {
1.34 noro 942: can += 2; FREENM(s);
1.17 noro 943: }
944: s = m2; m2 = NEXT(m2); FREENM(s);
1.95 noro 945: break;
946: case 1:
947: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
948: break;
949: case -1:
950: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
951: break;
952: }
953: }
954: if ( !mr0 )
955: if ( m1 ) mr0 = m1;
956: else if ( m2 ) mr0 = m2;
957: else return 0;
958: else if ( m1 ) NEXT(mr) = m1;
959: else if ( m2 ) NEXT(mr) = m2;
960: else NEXT(mr) = 0;
961: BDY(p1) = mr0;
962: SG(p1) = MAX(SG(p1),SG(p2));
963: LEN(p1) = LEN(p1)+LEN(p2)-can;
964: FREEND(p2);
965: return p1;
966: }
967: }
968:
969: /* XXX on opteron, the inlined manipulation of destructive additon of
970: * two NM seems to make gcc optimizer get confused, so the part is
971: * done in a function.
972: */
973:
1.113 noro 974: int nm_destructive_add_q(NM *m1,NM *m2,NM *mr0,NM *mr)
1.95 noro 975: {
976: NM s;
1.113 noro 977: Q t;
1.95 noro 978: int can;
979:
1.113 noro 980: addq(CQ(*m1),CQ(*m2),&t);
1.95 noro 981: s = *m1; *m1 = NEXT(*m1);
982: if ( t ) {
1.113 noro 983: can = 1; NEXTNM2(*mr0,*mr,s); CQ(*mr) = (t);
1.95 noro 984: } else {
985: can = 2; FREENM(s);
986: }
987: s = *m2; *m2 = NEXT(*m2); FREENM(s);
988: return can;
989: }
990:
1.113 noro 991: ND nd_add_q(ND p1,ND p2)
1.95 noro 992: {
993: int n,c,can;
994: ND r;
995: NM m1,m2,mr0,mr,s;
1.113 noro 996: Q t;
1.95 noro 997:
998: if ( !p1 ) return p2;
999: else if ( !p2 ) return p1;
1000: else {
1001: can = 0;
1002: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1003: c = DL_COMPARE(DL(m1),DL(m2));
1004: switch ( c ) {
1005: case 0:
1006: #if defined(__x86_64__)
1.113 noro 1007: can += nm_destructive_add_q(&m1,&m2,&mr0,&mr);
1.95 noro 1008: #else
1.113 noro 1009: addq(CQ(m1),CQ(m2),&t);
1.95 noro 1010: s = m1; m1 = NEXT(m1);
1011: if ( t ) {
1.113 noro 1012: can++; NEXTNM2(mr0,mr,s); CQ(mr) = (t);
1.95 noro 1013: } else {
1014: can += 2; FREENM(s);
1015: }
1016: s = m2; m2 = NEXT(m2); FREENM(s);
1017: #endif
1.17 noro 1018: break;
1019: case 1:
1020: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1021: break;
1022: case -1:
1023: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1024: break;
1025: }
1026: }
1027: if ( !mr0 )
1.34 noro 1028: if ( m1 ) mr0 = m1;
1029: else if ( m2 ) mr0 = m2;
1030: else return 0;
1031: else if ( m1 ) NEXT(mr) = m1;
1032: else if ( m2 ) NEXT(mr) = m2;
1033: else NEXT(mr) = 0;
1.17 noro 1034: BDY(p1) = mr0;
1035: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1036: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 1037: FREEND(p2);
1038: return p1;
1039: }
1040: }
1041:
1.71 noro 1042: ND nd_add_sf(ND p1,ND p2)
1043: {
1044: int n,c,can;
1045: ND r;
1046: NM m1,m2,mr0,mr,s;
1047: int t;
1048:
1049: if ( !p1 ) return p2;
1050: else if ( !p2 ) return p1;
1051: else {
1052: can = 0;
1053: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1054: c = DL_COMPARE(DL(m1),DL(m2));
1055: switch ( c ) {
1056: case 0:
1057: t = _addsf(CM(m1),CM(m2));
1058: s = m1; m1 = NEXT(m1);
1059: if ( t ) {
1060: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1061: } else {
1062: can += 2; FREENM(s);
1063: }
1064: s = m2; m2 = NEXT(m2); FREENM(s);
1065: break;
1066: case 1:
1067: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1068: break;
1069: case -1:
1070: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1071: break;
1072: }
1073: }
1074: if ( !mr0 )
1075: if ( m1 ) mr0 = m1;
1076: else if ( m2 ) mr0 = m2;
1077: else return 0;
1078: else if ( m1 ) NEXT(mr) = m1;
1079: else if ( m2 ) NEXT(mr) = m2;
1080: else NEXT(mr) = 0;
1081: BDY(p1) = mr0;
1082: SG(p1) = MAX(SG(p1),SG(p2));
1083: LEN(p1) = LEN(p1)+LEN(p2)-can;
1084: FREEND(p2);
1085: return p1;
1086: }
1087: }
1088:
1.1 noro 1089: /* ret=1 : success, ret=0 : overflow */
1.113 noro 1090: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 1091: {
1.11 noro 1092: ND d;
1.1 noro 1093: NM m,mrd,tail;
1.7 noro 1094: NM mul;
1.10 noro 1095: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1096: int c,c1,c2,dummy;
1.17 noro 1097: RHist h;
1.11 noro 1098: NDV p,red;
1.113 noro 1099: Q cg,cred,gcd,tq,qq;
1.21 noro 1100: double hmag;
1.1 noro 1101:
1.113 noro 1102: if ( dn ) {
1103: if ( mod )
1104: dn->m = 1;
1105: else
1106: dn->z = ONE;
1107: }
1.1 noro 1108: if ( !g ) {
1109: *rp = 0;
1110: return 1;
1111: }
1.113 noro 1112: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1113:
1.14 noro 1114: sugar0 = sugar = SG(g);
1.1 noro 1115: n = NV(g);
1.61 noro 1116: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 1117: for ( d = 0; g; ) {
1.63 noro 1118: index = ndl_find_reducer(HDL(g));
1.6 noro 1119: if ( index >= 0 ) {
1.17 noro 1120: h = nd_psh[index];
1121: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1122: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1123: nd_free(g); nd_free(d);
1124: return 0;
1125: }
1.77 noro 1126: if ( nd_demand )
1127: p = ndv_load(index);
1128: else
1129: p = ps[index];
1.71 noro 1130: if ( mod == -1 )
1131: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1132: else if ( mod ) {
1.19 noro 1133: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1134: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1.16 noro 1135: } else {
1.113 noro 1136: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1137: chsgnq(cg,&CQ(mul));
1138: nd_mul_c_q(d,cred); nd_mul_c_q(g,cred);
1.69 noro 1139: if ( dn ) {
1.113 noro 1140: mulq(dn->z,cred,&tq); dn->z = tq;
1.69 noro 1141: }
1.16 noro 1142: }
1.55 noro 1143: g = nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.34 noro 1144: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.113 noro 1145: if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) {
1146: tq = HCQ(g);
1.21 noro 1147: nd_removecont2(d,g);
1.69 noro 1148: if ( dn ) {
1.113 noro 1149: divq(tq,HCQ(g),&qq); divq(dn->z,qq,&tq); dn->z = tq;
1.69 noro 1150: }
1.113 noro 1151: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1152: }
1.1 noro 1153: } else if ( !full ) {
1154: *rp = g;
1155: return 1;
1156: } else {
1157: m = BDY(g);
1158: if ( NEXT(m) ) {
1.34 noro 1159: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1160: } else {
1161: FREEND(g); g = 0;
1162: }
1163: if ( d ) {
1.34 noro 1164: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1165: } else {
1.34 noro 1166: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1167: }
1168: }
1169: }
1.34 noro 1170: if ( d ) SG(d) = sugar;
1.1 noro 1171: *rp = d;
1172: return 1;
1173: }
1.28 noro 1174:
1.53 noro 1175: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1176: {
1177: int hindex,index;
1178: NDV p;
1179: ND u,d,red;
1180: NODE l;
1.31 noro 1181: NM mul,m,mrd,tail;
1.25 noro 1182: int sugar,psugar,n,h_reducible;
1183: PGeoBucket bucket;
1184: int c,c1,c2;
1.113 noro 1185: Q cg,cred,gcd,zzz;
1.25 noro 1186: RHist h;
1.28 noro 1187: double hmag,gmag;
1.77 noro 1188: int count = 0;
1189: int hcount = 0;
1.25 noro 1190:
1191: if ( !g ) {
1192: *rp = 0;
1193: return 1;
1194: }
1195: sugar = SG(g);
1196: n = NV(g);
1.113 noro 1197: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1198: bucket = create_pbucket();
1.31 noro 1199: add_pbucket(mod,bucket,g);
1.25 noro 1200: d = 0;
1.61 noro 1201: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1202: while ( 1 ) {
1.113 noro 1203: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1204: if ( hindex < 0 ) {
1.77 noro 1205: if ( DP_Print > 3 ) printf("(%d %d)",count,hcount);
1.34 noro 1206: if ( d ) SG(d) = sugar;
1.25 noro 1207: *rp = d;
1208: return 1;
1209: }
1210: g = bucket->body[hindex];
1.63 noro 1211: index = ndl_find_reducer(HDL(g));
1.25 noro 1212: if ( index >= 0 ) {
1.77 noro 1213: count++;
1214: if ( !d ) hcount++;
1.25 noro 1215: h = nd_psh[index];
1216: ndl_sub(HDL(g),DL(h),DL(mul));
1217: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1218: nd_free(d);
1.25 noro 1219: free_pbucket(bucket);
1220: *rp = 0;
1221: return 0;
1222: }
1.53 noro 1223: p = ps[index];
1.71 noro 1224: if ( mod == -1 )
1225: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1226: else if ( mod ) {
1.25 noro 1227: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1228: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1229: } else {
1.113 noro 1230: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1231: chsgnq(cg,&CQ(mul));
1232: nd_mul_c_q(d,cred);
1233: mulq_pbucket(bucket,cred);
1.26 noro 1234: g = bucket->body[hindex];
1.113 noro 1235: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1236: }
1.55 noro 1237: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1238: bucket->body[hindex] = nd_remove_head(g);
1239: red = nd_remove_head(red);
1.31 noro 1240: add_pbucket(mod,bucket,red);
1.34 noro 1241: psugar = SG(p)+TD(DL(mul));
1242: sugar = MAX(sugar,psugar);
1.28 noro 1243: if ( !mod && hmag && (gmag > hmag) ) {
1244: g = normalize_pbucket(mod,bucket);
1245: if ( !g ) {
1.34 noro 1246: if ( d ) SG(d) = sugar;
1.28 noro 1247: *rp = d;
1248: return 1;
1249: }
1250: nd_removecont2(d,g);
1.113 noro 1251: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1252: add_pbucket(mod,bucket,g);
1.28 noro 1253: }
1.25 noro 1254: } else if ( !full ) {
1255: g = normalize_pbucket(mod,bucket);
1.34 noro 1256: if ( g ) SG(g) = sugar;
1.25 noro 1257: *rp = g;
1258: return 1;
1259: } else {
1260: m = BDY(g);
1261: if ( NEXT(m) ) {
1.34 noro 1262: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1263: } else {
1264: FREEND(g); g = 0;
1265: }
1266: bucket->body[hindex] = g;
1267: NEXT(m) = 0;
1268: if ( d ) {
1.34 noro 1269: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1270: } else {
1.34 noro 1271: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1272: }
1273: }
1274: }
1275: }
1.27 noro 1276:
1.61 noro 1277: /* input : list of NDV, cand : list of NDV */
1.28 noro 1278:
1.61 noro 1279: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1280: {
1281: int n,i,stat;
1282: ND nf,d;
1.61 noro 1283: NDV r;
1.45 noro 1284: NODE t,s;
1.113 noro 1285: union oNDC dn;
1.45 noro 1286:
1.128 ! noro 1287: ndv_setup(0,0,cand,0);
1.31 noro 1288: n = length(cand);
1.28 noro 1289:
1290: /* membercheck : list is a subset of Id(cand) ? */
1291: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1292: again:
1.61 noro 1293: if ( nd_bpe > obpe )
1294: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1295: else
1296: r = (NDV)BDY(t);
1297: d = ndvtond(0,r);
1.69 noro 1298: stat = nd_nf(0,d,nd_ps,0,0,&nf);
1.45 noro 1299: if ( !stat ) {
1.103 noro 1300: nd_reconstruct(0,0);
1.45 noro 1301: goto again;
1302: } else if ( nf ) return 0;
1.71 noro 1303: if ( DP_Print ) { printf("."); fflush(stdout); }
1.28 noro 1304: }
1.71 noro 1305: if ( DP_Print ) { printf("\n"); }
1.28 noro 1306: /* gbcheck : cand is a GB of Id(cand) ? */
1.77 noro 1307: if ( !nd_gb(0,0,1) ) return 0;
1.28 noro 1308: /* XXX */
1.23 noro 1309: return 1;
1310: }
1.1 noro 1311:
1312: ND nd_remove_head(ND p)
1313: {
1314: NM m;
1315:
1316: m = BDY(p);
1317: if ( !NEXT(m) ) {
1.34 noro 1318: FREEND(p); p = 0;
1.31 noro 1319: } else {
1.34 noro 1320: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1321: }
1.1 noro 1322: FREENM(m);
1323: return p;
1324: }
1325:
1.69 noro 1326: ND nd_separate_head(ND p,ND *head)
1327: {
1328: NM m,m0;
1329: ND r;
1330:
1331: m = BDY(p);
1332: if ( !NEXT(m) ) {
1333: *head = p; p = 0;
1334: } else {
1335: m0 = m;
1336: BDY(p) = NEXT(m); LEN(p)--;
1337: NEXT(m0) = 0;
1338: MKND(NV(p),m0,1,r);
1339: *head = r;
1340: }
1341: return p;
1342: }
1343:
1.1 noro 1344: PGeoBucket create_pbucket()
1345: {
1346: PGeoBucket g;
1347:
1348: g = CALLOC(1,sizeof(struct oPGeoBucket));
1349: g->m = -1;
1350: return g;
1351: }
1352:
1.25 noro 1353: void free_pbucket(PGeoBucket b) {
1354: int i;
1355:
1.26 noro 1356: for ( i = 0; i <= b->m; i++ )
1.25 noro 1357: if ( b->body[i] ) {
1358: nd_free(b->body[i]);
1359: b->body[i] = 0;
1360: }
1361: GC_free(b);
1362: }
1363:
1.63 noro 1364: void add_pbucket_symbolic(PGeoBucket g,ND d)
1365: {
1366: int l,i,k,m;
1367:
1368: if ( !d )
1369: return;
1370: l = LEN(d);
1371: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1372: /* 2^(k-1) < l <= 2^k (=m) */
1373: d = nd_merge(g->body[k],d);
1374: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1375: g->body[k] = 0;
1376: d = nd_merge(g->body[k+1],d);
1377: }
1378: g->body[k] = d;
1379: g->m = MAX(g->m,k);
1380: }
1381:
1.31 noro 1382: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1383: {
1.31 noro 1384: int l,i,k,m;
1.1 noro 1385:
1.31 noro 1386: if ( !d )
1387: return;
1388: l = LEN(d);
1.29 noro 1389: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1390: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1391: d = nd_add(mod,g->body[k],d);
1392: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1393: g->body[k] = 0;
1.31 noro 1394: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1395: }
1396: g->body[k] = d;
1397: g->m = MAX(g->m,k);
1398: }
1399:
1.113 noro 1400: void mulq_pbucket(PGeoBucket g,Q c)
1.26 noro 1401: {
1402: int k;
1403:
1404: for ( k = 0; k <= g->m; k++ )
1.113 noro 1405: nd_mul_c_q(g->body[k],c);
1.26 noro 1406: }
1407:
1.63 noro 1408: NM remove_head_pbucket_symbolic(PGeoBucket g)
1409: {
1410: int j,i,k,c;
1411: NM head;
1412:
1413: k = g->m;
1414: j = -1;
1415: for ( i = 0; i <= k; i++ ) {
1416: if ( !g->body[i] ) continue;
1417: if ( j < 0 ) j = i;
1418: else {
1419: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1420: if ( c > 0 )
1421: j = i;
1422: else if ( c == 0 )
1423: g->body[i] = nd_remove_head(g->body[i]);
1424: }
1425: }
1426: if ( j < 0 ) return 0;
1427: else {
1428: head = BDY(g->body[j]);
1429: if ( !NEXT(head) ) {
1430: FREEND(g->body[j]);
1431: g->body[j] = 0;
1432: } else {
1433: BDY(g->body[j]) = NEXT(head);
1434: LEN(g->body[j])--;
1435: }
1436: return head;
1437: }
1438: }
1439:
1.19 noro 1440: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1441: {
1442: int j,i,c,k,nv,sum;
1.61 noro 1443: UINT *di,*dj;
1.1 noro 1444: ND gi,gj;
1445:
1446: k = g->m;
1447: while ( 1 ) {
1448: j = -1;
1449: for ( i = 0; i <= k; i++ ) {
1450: if ( !(gi = g->body[i]) )
1451: continue;
1452: if ( j < 0 ) {
1453: j = i;
1454: gj = g->body[j];
1455: dj = HDL(gj);
1.14 noro 1456: sum = HCM(gj);
1.1 noro 1457: } else {
1.34 noro 1458: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1459: if ( c > 0 ) {
1.34 noro 1460: if ( sum ) HCM(gj) = sum;
1461: else g->body[j] = nd_remove_head(gj);
1.1 noro 1462: j = i;
1463: gj = g->body[j];
1464: dj = HDL(gj);
1.14 noro 1465: sum = HCM(gj);
1.1 noro 1466: } else if ( c == 0 ) {
1.71 noro 1467: if ( mod == -1 )
1468: sum = _addsf(sum,HCM(gi));
1469: else {
1470: sum = sum+HCM(gi)-mod;
1471: if ( sum < 0 ) sum += mod;
1472: }
1.1 noro 1473: g->body[i] = nd_remove_head(gi);
1474: }
1475: }
1476: }
1.34 noro 1477: if ( j < 0 ) return -1;
1.1 noro 1478: else if ( sum ) {
1.14 noro 1479: HCM(gj) = sum;
1.26 noro 1480: return j;
1.31 noro 1481: } else
1.26 noro 1482: g->body[j] = nd_remove_head(gj);
1483: }
1484: }
1485:
1.113 noro 1486: int head_pbucket_q(PGeoBucket g)
1.26 noro 1487: {
1488: int j,i,c,k,nv;
1.113 noro 1489: Q sum,t;
1.26 noro 1490: ND gi,gj;
1491:
1492: k = g->m;
1493: while ( 1 ) {
1494: j = -1;
1495: for ( i = 0; i <= k; i++ ) {
1.34 noro 1496: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1497: if ( j < 0 ) {
1498: j = i;
1499: gj = g->body[j];
1.113 noro 1500: sum = HCQ(gj);
1.26 noro 1501: } else {
1502: nv = NV(gi);
1.34 noro 1503: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1504: if ( c > 0 ) {
1.113 noro 1505: if ( sum ) HCQ(gj) = sum;
1.34 noro 1506: else g->body[j] = nd_remove_head(gj);
1.26 noro 1507: j = i;
1508: gj = g->body[j];
1.113 noro 1509: sum = HCQ(gj);
1.26 noro 1510: } else if ( c == 0 ) {
1.113 noro 1511: addq(sum,HCQ(gi),&t);
1512: sum = t;
1.26 noro 1513: g->body[i] = nd_remove_head(gi);
1514: }
1515: }
1516: }
1.34 noro 1517: if ( j < 0 ) return -1;
1.26 noro 1518: else if ( sum ) {
1.113 noro 1519: HCQ(gj) = sum;
1.1 noro 1520: return j;
1.31 noro 1521: } else
1.1 noro 1522: g->body[j] = nd_remove_head(gj);
1523: }
1524: }
1525:
1.25 noro 1526: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1527: {
1.31 noro 1528: int i;
1.1 noro 1529: ND r,t;
1530:
1531: r = 0;
1.28 noro 1532: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1533: r = nd_add(mod,r,g->body[i]);
1.28 noro 1534: g->body[i] = 0;
1535: }
1536: g->m = -1;
1.1 noro 1537: return r;
1538: }
1539:
1.122 noro 1540: int do_diagonalize(int sugar,int m)
1.92 noro 1541: {
1542: int i,nh,stat;
1543: NODE r,g,t;
1544: ND h,nf,s,head;
1545: NDV nfv;
1.113 noro 1546: Q q,num,den;
1547: union oNDC dn;
1.92 noro 1548:
1549: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1550: if ( nd_demand )
1551: nfv = ndv_load(i);
1552: else
1553: nfv = nd_ps[i];
1554: s = ndvtond(m,nfv);
1555: s = nd_separate_head(s,&head);
1.122 noro 1556: stat = nd_nf(m,s,nd_ps,1,&dn,&nf);
1557: if ( !stat ) return 0;
1.92 noro 1558: if ( !m ) {
1.113 noro 1559: NTOQ(NM(dn.z),SGN(dn.z),num);
1560: mulq(HCQ(head),num,&q); HCQ(head) = q;
1561: if ( DN(dn.z) ) {
1562: NTOQ(DN(dn.z),1,den);
1563: nd_mul_c_q(nf,den);
1564: }
1.92 noro 1565: }
1566: nf = nd_add(m,head,nf);
1567: ndv_free(nfv);
1568: nd_removecont(m,nf);
1569: nfv = ndtondv(m,nf);
1570: nd_free(nf);
1571: nd_bound[i] = ndv_compute_bound(nfv);
1572: if ( nd_demand ) {
1573: ndv_save(nfv,i);
1574: ndv_free(nfv);
1575: } else
1576: nd_ps[i] = nfv;
1577: }
1.122 noro 1578: return 1;
1.92 noro 1579: }
1580:
1.27 noro 1581: /* return value = 0 => input is not a GB */
1582:
1.77 noro 1583: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1584: {
1585: int i,nh,sugar,stat;
1.23 noro 1586: NODE r,g,t;
1.1 noro 1587: ND_pairs d;
1588: ND_pairs l;
1.117 noro 1589: ND h,nf,s,head,nf1;
1.63 noro 1590: NDV nfv;
1.113 noro 1591: Q q,num,den;
1592: union oNDC dn;
1.1 noro 1593:
1.23 noro 1594: g = 0; d = 0;
1595: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1596: d = update_pairs(d,g,i);
1597: g = update_base(g,i);
1598: }
1599: sugar = 0;
1600: while ( d ) {
1601: again:
1602: l = nd_minp(d,&d);
1.14 noro 1603: if ( SG(l) != sugar ) {
1.122 noro 1604: if ( ishomo ) {
1605: stat = do_diagonalize(sugar,m);
1606: if ( !stat ) {
1607: NEXT(l) = d; d = l;
1608: d = nd_reconstruct(0,d);
1609: goto again;
1610: }
1611: }
1.14 noro 1612: sugar = SG(l);
1.71 noro 1613: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1614: }
1.53 noro 1615: stat = nd_sp(m,0,l,&h);
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: }
1.41 noro 1621: #if USE_GEOBUCKET
1.69 noro 1622: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1623: #else
1.69 noro 1624: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1625: #endif
1.1 noro 1626: if ( !stat ) {
1627: NEXT(l) = d; d = l;
1.103 noro 1628: d = nd_reconstruct(0,d);
1.1 noro 1629: goto again;
1630: } else if ( nf ) {
1.27 noro 1631: if ( checkonly ) return 0;
1.71 noro 1632: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1633: nd_removecont(m,nf);
1.118 noro 1634: if ( !m && nd_nalg ) {
1635: nd_monic(0,&nf);
1636: nd_removecont(m,nf);
1.117 noro 1637: }
1.63 noro 1638: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1639: nh = ndv_newps(m,nfv,0);
1.1 noro 1640: d = update_pairs(d,g,nh);
1641: g = update_base(g,nh);
1642: FREENDP(l);
1643: } else {
1.71 noro 1644: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1645: FREENDP(l);
1646: }
1647: }
1.77 noro 1648: if ( nd_demand )
1649: for ( t = g; t; t = NEXT(t) )
1650: BDY(t) = (pointer)ndv_load((int)BDY(t));
1651: else
1652: for ( t = g; t; t = NEXT(t) )
1653: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1654: return g;
1655: }
1656:
1.122 noro 1657: int do_diagonalize_trace(int sugar,int m)
1.91 noro 1658: {
1659: int i,nh,stat;
1660: NODE r,g,t;
1661: ND h,nf,nfq,s,head;
1662: NDV nfv,nfqv;
1.113 noro 1663: Q q,den,num;
1664: union oNDC dn;
1.91 noro 1665:
1666: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1667: /* for nd_ps */
1668: s = ndvtond(m,nd_ps[i]);
1669: s = nd_separate_head(s,&head);
1.122 noro 1670: stat = nd_nf_pbucket(m,s,nd_ps,1,&nf);
1671: if ( !stat ) return 0;
1.91 noro 1672: nf = nd_add(m,head,nf);
1673: ndv_free(nd_ps[i]);
1674: nd_ps[i] = ndtondv(m,nf);
1675: nd_free(nf);
1676:
1677: /* for nd_ps_trace */
1678: if ( nd_demand )
1679: nfv = ndv_load(i);
1680: else
1681: nfv = nd_ps_trace[i];
1682: s = ndvtond(0,nfv);
1683: s = nd_separate_head(s,&head);
1.122 noro 1684: stat = nd_nf(0,s,nd_ps_trace,1,&dn,&nf);
1685: if ( !stat ) return 0;
1.113 noro 1686: NTOQ(NM(dn.z),SGN(dn.z),num);
1687: mulq(HCQ(head),num,&q); HCQ(head) = q;
1688: if ( DN(dn.z) ) {
1689: NTOQ(DN(dn.z),1,den);
1690: nd_mul_c_q(nf,den);
1691: }
1.91 noro 1692: nf = nd_add(0,head,nf);
1693: ndv_free(nfv);
1694: nd_removecont(0,nf);
1695: nfv = ndtondv(0,nf);
1696: nd_free(nf);
1697: nd_bound[i] = ndv_compute_bound(nfv);
1698: if ( nd_demand ) {
1699: ndv_save(nfv,i);
1700: ndv_free(nfv);
1701: } else
1702: nd_ps_trace[i] = nfv;
1703: }
1.122 noro 1704: return 1;
1.91 noro 1705: }
1706:
1.118 noro 1707: static struct oEGT eg_invdalg;
1708: struct oEGT eg_le;
1709:
1.82 noro 1710: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 1711: {
1712: int i,nh,sugar,stat;
1.23 noro 1713: NODE r,g,t;
1.20 noro 1714: ND_pairs d;
1715: ND_pairs l;
1.82 noro 1716: ND h,nf,nfq,s,head;
1.63 noro 1717: NDV nfv,nfqv;
1.113 noro 1718: Q q,den,num;
1719: union oNDC dn;
1.118 noro 1720: struct oEGT eg_monic,egm0,egm1;
1.20 noro 1721:
1.118 noro 1722: init_eg(&eg_monic);
1723: init_eg(&eg_invdalg);
1724: init_eg(&eg_le);
1.23 noro 1725: g = 0; d = 0;
1726: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1727: d = update_pairs(d,g,i);
1728: g = update_base(g,i);
1729: }
1730: sugar = 0;
1731: while ( d ) {
1732: again:
1733: l = nd_minp(d,&d);
1734: if ( SG(l) != sugar ) {
1.122 noro 1735: if ( ishomo ) {
1736: stat = do_diagonalize_trace(sugar,m);
1737: if ( !stat ) {
1738: NEXT(l) = d; d = l;
1739: d = nd_reconstruct(1,d);
1740: goto again;
1741: }
1742: }
1.20 noro 1743: sugar = SG(l);
1.71 noro 1744: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1745: }
1.53 noro 1746: stat = nd_sp(m,0,l,&h);
1.20 noro 1747: if ( !stat ) {
1748: NEXT(l) = d; d = l;
1.103 noro 1749: d = nd_reconstruct(1,d);
1.20 noro 1750: goto again;
1751: }
1.41 noro 1752: #if USE_GEOBUCKET
1.53 noro 1753: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1754: #else
1.69 noro 1755: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1756: #endif
1.20 noro 1757: if ( !stat ) {
1758: NEXT(l) = d; d = l;
1.103 noro 1759: d = nd_reconstruct(1,d);
1.20 noro 1760: goto again;
1761: } else if ( nf ) {
1.77 noro 1762: if ( nd_demand ) {
1763: nfqv = ndv_load(nd_psn);
1764: nfq = ndvtond(0,nfqv);
1765: } else
1766: nfq = 0;
1767: if ( !nfq ) {
1768: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1769: NEXT(l) = d; d = l;
1.103 noro 1770: d = nd_reconstruct(1,d);
1.77 noro 1771: goto again;
1772: }
1.70 noro 1773: }
1.20 noro 1774: if ( nfq ) {
1.70 noro 1775: /* m|HC(nfq) => failure */
1.113 noro 1776: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1.63 noro 1777:
1.71 noro 1778: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.118 noro 1779: if ( nd_nalg ) {
1780: /* m|DN(HC(nf)^(-1)) => failure */
1781: get_eg(&egm0);
1782: if ( !nd_monic(m,&nfq) ) return 0;
1783: get_eg(&egm1); add_eg(&eg_monic,&egm0,&egm1);
1784: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1785: nfv = ndv_dup(0,nfqv); ndv_mod(m,nfv); nd_free(nf);
1786: } else {
1787: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1788: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1789: }
1.77 noro 1790: nh = ndv_newps(0,nfv,nfqv);
1.20 noro 1791: d = update_pairs(d,g,nh);
1792: g = update_base(g,nh);
1793: } else {
1.71 noro 1794: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1795: }
1796: } else {
1.71 noro 1797: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1798: }
1799: FREENDP(l);
1800: }
1.77 noro 1801: if ( nd_demand )
1802: for ( t = g; t; t = NEXT(t) )
1803: BDY(t) = (pointer)ndv_load((int)BDY(t));
1804: else
1805: for ( t = g; t; t = NEXT(t) )
1806: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.118 noro 1807: if ( nd_nalg ) {
1808: print_eg("monic",&eg_monic);
1809: print_eg("invdalg",&eg_invdalg);
1810: print_eg("le",&eg_le);
1811: }
1.20 noro 1812: return g;
1813: }
1814:
1.23 noro 1815: int ndv_compare(NDV *p1,NDV *p2)
1816: {
1.34 noro 1817: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1818: }
1819:
1820: int ndv_compare_rev(NDV *p1,NDV *p2)
1821: {
1.34 noro 1822: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1823: }
1824:
1.61 noro 1825: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1826: {
1.69 noro 1827: int i,n,stat;
1828: ND nf,g,head;
1.23 noro 1829: NODE t,a0,a;
1.113 noro 1830: union oNDC dn;
1.69 noro 1831: NDV *w;
1.113 noro 1832: Q q,num,den;
1.23 noro 1833:
1.69 noro 1834: n = length(f);
1835: #if 0
1836: w = (NDV *)ALLOCA(n*sizeof(NDV));
1837: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1838: qsort(w,n,sizeof(NDV),
1839: (int (*)(const void *,const void *))ndv_compare);
1840: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1841: #endif
1.128 ! noro 1842: ndv_setup(m,0,f,0);
1.69 noro 1843: for ( i = 0; i < n; ) {
1844: g = ndvtond(m,nd_ps[i]);
1845: g = nd_separate_head(g,&head);
1846: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1847: if ( !stat )
1.103 noro 1848: nd_reconstruct(0,0);
1.69 noro 1849: else {
1.71 noro 1850: if ( DP_Print ) { printf("."); fflush(stdout); }
1.85 noro 1851: if ( !m ) {
1.113 noro 1852: NTOQ(NM(dn.z),SGN(dn.z),num);
1853: mulq(HCQ(head),num,&q); HCQ(head) = q;
1854: if ( DN(dn.z) ) {
1855: NTOQ(DN(dn.z),1,den);
1856: nd_mul_c_q(nf,den);
1857: }
1.85 noro 1858: }
1.69 noro 1859: nf = nd_add(m,head,nf);
1860: ndv_free(nd_ps[i]);
1.24 noro 1861: nd_removecont(m,nf);
1.69 noro 1862: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1863: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1864: i++;
1.23 noro 1865: }
1866: }
1.71 noro 1867: if ( DP_Print ) { printf("\n"); }
1.23 noro 1868: for ( a0 = 0, i = 0; i < n; i++ ) {
1869: NEXTNODE(a0,a);
1.69 noro 1870: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1871: }
1872: NEXT(a) = 0;
1873: return a0;
1874: }
1875:
1.1 noro 1876: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1877: {
1878: ND_pairs d1,nd,cur,head,prev,remove;
1879:
1880: if ( !g ) return d;
1881: d = crit_B(d,t);
1882: d1 = nd_newpairs(g,t);
1883: d1 = crit_M(d1);
1884: d1 = crit_F(d1);
1.55 noro 1885: if ( do_weyl )
1886: head = d1;
1887: else {
1888: prev = 0; cur = head = d1;
1889: while ( cur ) {
1890: if ( crit_2( cur->i1,cur->i2 ) ) {
1891: remove = cur;
1892: if ( !prev ) head = cur = NEXT(cur);
1893: else cur = NEXT(prev) = NEXT(cur);
1894: FREENDP(remove);
1895: } else {
1896: prev = cur; cur = NEXT(cur);
1897: }
1.1 noro 1898: }
1899: }
1900: if ( !d )
1901: return head;
1902: else {
1903: nd = d;
1.34 noro 1904: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1905: NEXT(nd) = head;
1906: return d;
1907: }
1908: }
1909:
1910: ND_pairs nd_newpairs( NODE g, int t )
1911: {
1912: NODE h;
1.61 noro 1913: UINT *dl;
1.34 noro 1914: int ts,s;
1.1 noro 1915: ND_pairs r,r0;
1916:
1.20 noro 1917: dl = DL(nd_psh[t]);
1.34 noro 1918: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1919: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1920: NEXTND_pairs(r0,r);
1921: r->i1 = (int)BDY(h);
1922: r->i2 = t;
1.20 noro 1923: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1924: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1925: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1926: }
1927: NEXT(r) = 0;
1928: return r0;
1929: }
1930:
1931: ND_pairs crit_B( ND_pairs d, int s )
1932: {
1933: ND_pairs cur,head,prev,remove;
1.61 noro 1934: UINT *t,*tl,*lcm;
1.1 noro 1935: int td,tdl;
1936:
1937: if ( !d ) return 0;
1.20 noro 1938: t = DL(nd_psh[s]);
1.1 noro 1939: prev = 0;
1940: head = cur = d;
1.61 noro 1941: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1942: while ( cur ) {
1943: tl = cur->lcm;
1944: if ( ndl_reducible(tl,t)
1.20 noro 1945: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1946: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1947: remove = cur;
1948: if ( !prev ) {
1949: head = cur = NEXT(cur);
1950: } else {
1951: cur = NEXT(prev) = NEXT(cur);
1952: }
1953: FREENDP(remove);
1954: } else {
1.34 noro 1955: prev = cur; cur = NEXT(cur);
1.1 noro 1956: }
1957: }
1958: return head;
1959: }
1960:
1961: ND_pairs crit_M( ND_pairs d1 )
1962: {
1963: ND_pairs e,d2,d3,dd,p;
1.61 noro 1964: UINT *id,*jd;
1.1 noro 1965:
1966: for ( dd = 0, e = d1; e; e = d3 ) {
1967: if ( !(d2 = NEXT(e)) ) {
1968: NEXT(e) = dd;
1969: return e;
1970: }
1.34 noro 1971: id = LCM(e);
1.1 noro 1972: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 1973: p = NEXT(d2);
1974: jd = LCM(d2);
1975: if ( ndl_equal(jd,id) )
1976: ;
1977: else if ( TD(jd) > TD(id) )
1.1 noro 1978: if ( ndl_reducible(jd,id) ) continue;
1979: else ;
1.34 noro 1980: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 1981: NEXT(d2) = d3;
1982: d3 = d2;
1983: }
1984: NEXT(e) = dd;
1985: dd = e;
1986: continue;
1987: /**/
1988: delit: NEXT(d2) = d3;
1989: d3 = d2;
1990: for ( ; p; p = d2 ) {
1991: d2 = NEXT(p);
1992: NEXT(p) = d3;
1993: d3 = p;
1994: }
1995: FREENDP(e);
1996: }
1997: return dd;
1998: }
1999:
2000: ND_pairs crit_F( ND_pairs d1 )
2001: {
2002: ND_pairs rest, head,remove;
2003: ND_pairs last, p, r, w;
2004: int s;
2005:
2006: for ( head = last = 0, p = d1; NEXT(p); ) {
2007: r = w = equivalent_pairs(p,&rest);
1.14 noro 2008: s = SG(r);
1.1 noro 2009: w = NEXT(w);
2010: while ( w ) {
2011: if ( crit_2(w->i1,w->i2) ) {
2012: r = w;
2013: w = NEXT(w);
2014: while ( w ) {
2015: remove = w;
2016: w = NEXT(w);
2017: FREENDP(remove);
2018: }
2019: break;
1.14 noro 2020: } else if ( SG(w) < s ) {
1.1 noro 2021: FREENDP(r);
2022: r = w;
1.14 noro 2023: s = SG(r);
1.1 noro 2024: w = NEXT(w);
2025: } else {
2026: remove = w;
2027: w = NEXT(w);
2028: FREENDP(remove);
2029: }
2030: }
2031: if ( last ) NEXT(last) = r;
2032: else head = r;
2033: NEXT(last = r) = 0;
2034: p = rest;
2035: if ( !p ) return head;
2036: }
2037: if ( !last ) return p;
2038: NEXT(last) = p;
2039: return head;
2040: }
2041:
2042: int crit_2( int dp1, int dp2 )
2043: {
1.20 noro 2044: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2045: }
2046:
1.40 noro 2047: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2048: {
2049: ND_pairs w,p,r,s;
1.61 noro 2050: UINT *d;
1.1 noro 2051:
2052: w = d1;
1.34 noro 2053: d = LCM(w);
1.1 noro 2054: s = NEXT(w);
2055: NEXT(w) = 0;
2056: for ( r = 0; s; s = p ) {
2057: p = NEXT(s);
1.34 noro 2058: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2059: NEXT(s) = w; w = s;
1.1 noro 2060: } else {
1.39 noro 2061: NEXT(s) = r; r = s;
1.1 noro 2062: }
2063: }
2064: *prest = r;
2065: return w;
2066: }
2067:
2068: NODE update_base(NODE nd,int ndp)
2069: {
1.61 noro 2070: UINT *dl, *dln;
1.1 noro 2071: NODE last, p, head;
2072:
1.20 noro 2073: dl = DL(nd_psh[ndp]);
1.1 noro 2074: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2075: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2076: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2077: p = NEXT(p);
2078: if ( last ) NEXT(last) = p;
2079: } else {
2080: if ( !last ) head = p;
2081: p = NEXT(last = p);
2082: }
2083: }
2084: head = append_one(head,ndp);
2085: return head;
2086: }
2087:
2088: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2089: {
2090: ND_pairs m,ml,p,l;
1.61 noro 2091: UINT *lcm;
1.33 noro 2092: int s,td,len,tlen,c,c1;
1.1 noro 2093:
2094: if ( !(p = NEXT(m = d)) ) {
2095: *prest = p;
2096: NEXT(m) = 0;
2097: return m;
2098: }
1.14 noro 2099: s = SG(m);
1.77 noro 2100: if ( !NoSugar ) {
2101: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2102: if ( (SG(p) < s)
2103: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
2104: ml = l; m = p; s = SG(m);
2105: }
2106: } else {
2107: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2108: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
2109: ml = l; m = p; s = SG(m);
2110: }
2111: }
1.1 noro 2112: if ( !ml ) *prest = NEXT(m);
2113: else {
2114: NEXT(ml) = NEXT(m);
2115: *prest = d;
2116: }
2117: NEXT(m) = 0;
2118: return m;
2119: }
2120:
1.63 noro 2121: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2122: {
1.69 noro 2123: int msugar,i;
1.63 noro 2124: ND_pairs t,dm0,dm,dr0,dr;
2125:
2126: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2127: if ( SG(t) < msugar ) msugar = SG(t);
2128: dm0 = 0; dr0 = 0;
1.69 noro 2129: for ( i = 0, t = d; t; t = NEXT(t) )
2130: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2131: if ( dm0 ) NEXT(dm) = t;
2132: else dm0 = t;
2133: dm = t;
1.69 noro 2134: i++;
1.63 noro 2135: } else {
2136: if ( dr0 ) NEXT(dr) = t;
2137: else dr0 = t;
2138: dr = t;
2139: }
2140: NEXT(dm) = 0;
2141: if ( dr0 ) NEXT(dr) = 0;
2142: *prest = dr0;
2143: return dm0;
2144: }
2145:
1.77 noro 2146: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2147: {
1.3 noro 2148: int len;
1.13 noro 2149: RHist r;
1.20 noro 2150: NDV b;
1.3 noro 2151:
1.1 noro 2152: if ( nd_psn == nd_pslen ) {
2153: nd_pslen *= 2;
1.11 noro 2154: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2155: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2156: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2157: nd_bound = (UINT **)
2158: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2159: }
1.39 noro 2160: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2161: nd_ps[nd_psn] = a;
1.39 noro 2162: if ( aq ) {
1.63 noro 2163: nd_ps_trace[nd_psn] = aq;
2164: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2165: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2166: } else {
1.63 noro 2167: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2168: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2169: }
1.77 noro 2170: if ( nd_demand ) {
2171: if ( aq ) {
2172: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2173: nd_ps_trace[nd_psn] = 0;
2174: } else {
2175: ndv_save(nd_ps[nd_psn],nd_psn);
2176: nd_ps[nd_psn] = 0;
2177: }
2178: }
1.1 noro 2179: return nd_psn++;
2180: }
2181:
1.128 ! noro 2182: void ndv_setup(int mod,int trace,NODE f,int dont_sort)
1.1 noro 2183: {
1.5 noro 2184: int i,j,td,len,max;
1.1 noro 2185: NODE s,s0,f0;
1.61 noro 2186: UINT *d;
1.13 noro 2187: RHist r;
1.69 noro 2188: NDV *w;
1.61 noro 2189: NDV a,am;
1.11 noro 2190:
2191: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2192:
1.69 noro 2193: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2194: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2195: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
1.128 ! noro 2196: if ( !dont_sort )
! 2197: qsort(w,nd_psn,sizeof(NDV),
! 2198: (int (*)(const void *,const void *))ndv_compare);
1.69 noro 2199: nd_pslen = 2*nd_psn;
1.11 noro 2200: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2201: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2202: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2203: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2204:
1.11 noro 2205: if ( !nd_red )
1.13 noro 2206: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2207: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2208: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2209: if ( trace ) {
1.69 noro 2210: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2211: ndv_removecont(0,a);
2212: am = nd_ps[i] = ndv_dup(mod,a);
2213: ndv_mod(mod,am);
2214: ndv_removecont(mod,am);
2215: } else {
1.69 noro 2216: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2217: ndv_removecont(mod,a);
1.39 noro 2218: }
1.61 noro 2219: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2220: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2221: nd_psh[i] = r;
1.77 noro 2222: if ( nd_demand ) {
2223: if ( trace ) {
2224: ndv_save(nd_ps_trace[i],i);
2225: nd_ps_trace[i] = 0;
2226: } else {
2227: ndv_save(nd_ps[i],i);
2228: nd_ps[i] = 0;
2229: }
2230: }
1.20 noro 2231: }
2232: }
2233:
1.119 noro 2234: struct order_spec *append_block(struct order_spec *spec,
2235: int nv,int nalg,int ord);
2236:
1.121 noro 2237: extern VECT current_dl_weight_vector_obj;
2238: static VECT prev_weight_vector_obj;
2239:
1.120 noro 2240: void preprocess_algcoef(VL vv,VL av,struct order_spec *ord,LIST f,
2241: struct order_spec **ord1p,LIST *f1p,NODE *alistp)
2242: {
1.121 noro 2243: NODE alist,t,s,r0,r,arg;
1.120 noro 2244: VL tv;
2245: P poly;
2246: DP d;
2247: Alg alpha,dp;
2248: DAlg inv,da,hc;
2249: MP m;
1.121 noro 2250: int i,nvar,nalg,n;
1.120 noro 2251: NumberField nf;
2252: LIST f1,f2;
2253: struct order_spec *current_spec;
1.121 noro 2254: VECT obj,obj0;
2255: Obj tmp;
1.120 noro 2256:
2257: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++);
2258: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++);
2259:
2260: for ( alist = 0, tv = av; tv; tv = NEXT(tv) ) {
2261: NEXTNODE(alist,t); MKV(tv->v,poly);
2262: MKAlg(poly,alpha); BDY(t) = (pointer)alpha;
2263: tv->v = tv->v->priv;
2264: }
2265: NEXT(t) = 0;
2266:
2267: /* simplification, makeing polynomials monic */
2268: setfield_dalg(alist);
2269: obj_algtodalg(f,&f1);
2270: for ( t = BDY(f); t; t = NEXT(t) ) {
2271: initd(ord); ptod(vv,vv,(P)BDY(t),&d);
2272: hc = (DAlg)BDY(d)->c;
2273: if ( NID(hc) == N_DA ) {
2274: invdalg(hc,&inv);
2275: for ( m = BDY(d); m; m = NEXT(m) ) {
2276: muldalg(inv,(DAlg)m->c,&da); m->c = (P)da;
2277: }
2278: }
2279: initd(ord); dtop(vv,vv,d,&poly); BDY(f) = (pointer)poly;
2280: }
2281: obj_dalgtoalg(f1,&f);
2282:
2283: /* append alg vars to the var list */
2284: for ( tv = vv; NEXT(tv); tv = NEXT(tv) );
2285: NEXT(tv) = av;
2286:
2287: /* append a block to ord */
2288: *ord1p = append_block(ord,nvar,nalg,2);
2289:
2290: /* create generator list */
2291: nf = get_numberfield();
2292: for ( i = nalg-1, t = BDY(f); i >= 0; i-- ) {
2293: MKAlg(nf->defpoly[i],dp);
2294: MKNODE(s,dp,t); t = s;
2295: }
2296: MKLIST(f1,t);
2297: *alistp = alist;
2298: algobjtorat(f1,f1p);
1.121 noro 2299:
2300: /* creating a new weight vector */
2301: prev_weight_vector_obj = obj0 = current_dl_weight_vector_obj;
2302: n = nvar+nalg+1;
2303: MKVECT(obj,n);
2304: if ( obj0 && obj0->len == nvar )
2305: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = BDY(obj0)[i];
2306: else
2307: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = (pointer)ONE;
2308: for ( i = 0; i < nalg; i++ ) BDY(obj)[i+nvar] = 0;
2309: BDY(obj)[n-1] = (pointer)ONE;
2310: arg = mknode(1,obj);
2311: Pdp_set_weight(arg,&tmp);
2312: }
2313:
2314: NODE postprocess_algcoef(VL av,NODE alist,NODE r)
2315: {
2316: NODE s,t,u0,u;
2317: P p;
2318: VL tv;
2319: Obj obj,tmp;
2320: NODE arg;
2321:
2322: u0 = 0;
2323: for ( t = r; t; t = NEXT(t) ) {
2324: p = (P)BDY(t);
2325: for ( tv = av, s = alist; tv; tv = NEXT(tv), s = NEXT(s) ) {
2326: substr(CO,0,(Obj)p,tv->v,(Obj)BDY(s),&obj); p = (P)obj;
2327: }
2328: if ( OID(p) == O_P || (OID(p) == O_N && NID((Num)p) != N_A) ) {
2329: NEXTNODE(u0,u);
2330: BDY(u) = (pointer)p;
2331: }
2332: }
2333: arg = mknode(1,prev_weight_vector_obj);
2334: Pdp_set_weight(arg,&tmp);
2335:
2336: return u0;
1.120 noro 2337: }
2338:
1.63 noro 2339: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2340: {
1.119 noro 2341: VL tv,fv,vv,vc,av;
2342: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2343: int e,max,nvar,i;
1.69 noro 2344: NDV b;
1.119 noro 2345: int ishomo,nalg;
2346: Alg alpha,dp;
2347: P p;
2348: LIST f1,f2;
2349: Obj obj;
2350: NumberField nf;
1.120 noro 2351: struct order_spec *ord1;
1.1 noro 2352:
1.78 noro 2353: if ( !m && Demand ) nd_demand = 1;
2354: else nd_demand = 0;
2355:
1.103 noro 2356: #if 0
1.74 noro 2357: ndv_alloc = 0;
1.103 noro 2358: #endif
1.1 noro 2359: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2360: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2361: switch ( ord->id ) {
2362: case 1:
2363: if ( ord->nv != nvar )
2364: error("nd_{gr,f4} : invalid order specification");
2365: break;
2366: default:
2367: break;
2368: }
1.119 noro 2369: nd_nalg = 0;
1.123 noro 2370: av = 0;
1.119 noro 2371: if ( !m ) {
2372: get_algtree((Obj)f,&av);
2373: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2374: nd_ntrans = nvar;
1.119 noro 2375: nd_nalg = nalg;
2376: /* #i -> t#i */
2377: if ( nalg ) {
1.120 noro 2378: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2379: ord = ord1;
2380: f = f1;
1.119 noro 2381: }
1.120 noro 2382: nvar += nalg;
1.119 noro 2383: }
1.32 noro 2384: nd_init_ord(ord);
1.61 noro 2385: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2386: for ( tv = vv; tv; tv = NEXT(tv) ) {
2387: e = getdeg(tv->v,(P)BDY(t));
2388: max = MAX(e,max);
2389: }
2390: nd_setup_parameters(nvar,max);
1.77 noro 2391: ishomo = 1;
1.1 noro 2392: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2393: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.77 noro 2394: if ( ishomo )
2395: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2396: if ( m ) ndv_mod(m,b);
1.61 noro 2397: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2398: }
2399: if ( fd0 ) NEXT(fd) = 0;
1.128 ! noro 2400: ndv_setup(m,0,fd0,0);
1.77 noro 2401: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.82 noro 2402: nd_demand = 0;
1.61 noro 2403: x = ndv_reducebase(x);
2404: x = ndv_reduceall(m,x);
1.23 noro 2405: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2406: NEXTNODE(r0,r);
1.61 noro 2407: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2408: }
2409: if ( r0 ) NEXT(r) = 0;
1.121 noro 2410: if ( nalg )
2411: r0 = postprocess_algcoef(av,alist,r0);
1.20 noro 2412: MKLIST(*rp,r0);
1.103 noro 2413: #if 0
1.74 noro 2414: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 2415: #endif
1.127 noro 2416: }
2417:
2418: void nd_gr_postproc(LIST f,LIST v,int m,struct order_spec *ord,int do_check,LIST *rp)
2419: {
2420: VL tv,fv,vv,vc,av;
2421: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2422: int e,max,nvar,i;
2423: NDV b;
2424: int ishomo,nalg;
2425: Alg alpha,dp;
2426: P p;
2427: LIST f1,f2;
2428: Obj obj;
2429: NumberField nf;
2430: struct order_spec *ord1;
2431:
2432: get_vars((Obj)f,&fv); pltovl(v,&vv);
2433: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
2434: switch ( ord->id ) {
2435: case 1:
2436: if ( ord->nv != nvar )
2437: error("nd_check : invalid order specification");
2438: break;
2439: default:
2440: break;
2441: }
2442: nd_nalg = 0;
2443: av = 0;
2444: if ( !m ) {
2445: get_algtree((Obj)f,&av);
2446: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
2447: nd_ntrans = nvar;
2448: nd_nalg = nalg;
2449: /* #i -> t#i */
2450: if ( nalg ) {
2451: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2452: ord = ord1;
2453: f = f1;
2454: }
2455: nvar += nalg;
2456: }
2457: nd_init_ord(ord);
2458: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2459: for ( tv = vv; tv; tv = NEXT(tv) ) {
2460: e = getdeg(tv->v,(P)BDY(t));
2461: max = MAX(e,max);
2462: }
2463: nd_setup_parameters(nvar,max);
2464: ishomo = 1;
2465: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2466: b = (pointer)ptondv(CO,vv,(P)BDY(t));
2467: if ( ishomo )
2468: ishomo = ishomo && ndv_ishomo(b);
2469: if ( m ) ndv_mod(m,b);
2470: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
2471: }
2472: if ( fd0 ) NEXT(fd) = 0;
1.128 ! noro 2473: ndv_setup(m,0,fd0,0);
1.127 noro 2474: for ( x = 0, i = 0; i < nd_psn; i++ )
2475: x = update_base(x,i);
2476: if ( do_check ) {
2477: x = nd_gb(m,ishomo,1);
2478: if ( !x ) {
2479: *rp = 0;
2480: return;
2481: }
2482: } else {
2483: for ( t = x; t; t = NEXT(t) )
2484: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
2485: }
2486: x = ndv_reducebase(x);
2487: x = ndv_reduceall(m,x);
2488: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
2489: NEXTNODE(r0,r);
2490: BDY(r) = ndvtop(m,CO,vv,BDY(t));
2491: }
2492: if ( r0 ) NEXT(r) = 0;
2493: if ( nalg )
2494: r0 = postprocess_algcoef(av,alist,r0);
2495: MKLIST(*rp,r0);
1.20 noro 2496: }
2497:
1.52 noro 2498: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2499: {
1.119 noro 2500: VL tv,fv,vv,vc,av;
2501: NODE fd,fd0,in0,in,r,r0,t,s,cand,alist;
1.61 noro 2502: int m,nocheck,nvar,mindex,e,max;
2503: NDV c;
2504: NMV a;
1.27 noro 2505: P p;
1.61 noro 2506: EPOS oepos;
1.119 noro 2507: int obpe,oadv,wmax,i,len,cbpe,ishomo,nalg;
2508: Alg alpha,dp;
2509: P poly;
2510: LIST f1,f2;
2511: Obj obj;
2512: NumberField nf;
1.120 noro 2513: struct order_spec *ord1;
1.20 noro 2514:
2515: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2516: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2517: switch ( ord->id ) {
2518: case 1:
2519: if ( ord->nv != nvar )
2520: error("nd_gr_trace : invalid order specification");
2521: break;
2522: default:
2523: break;
2524: }
1.119 noro 2525:
2526: get_algtree((Obj)f,&av);
2527: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2528: nd_ntrans = nvar;
1.119 noro 2529: nd_nalg = nalg;
2530: /* #i -> t#i */
2531: if ( nalg ) {
1.120 noro 2532: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2533: ord = ord1;
2534: f = f1;
1.119 noro 2535: }
1.120 noro 2536: nvar += nalg;
1.119 noro 2537:
1.52 noro 2538: nocheck = 0;
2539: mindex = 0;
2540:
1.78 noro 2541: if ( Demand ) nd_demand = 1;
2542: else nd_demand = 0;
2543:
1.52 noro 2544: /* setup modulus */
2545: if ( trace < 0 ) {
2546: trace = -trace;
2547: nocheck = 1;
2548: }
2549: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2550: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2551: for ( tv = vv; tv; tv = NEXT(tv) ) {
2552: e = getdeg(tv->v,(P)BDY(t));
2553: max = MAX(e,max);
1.23 noro 2554: }
1.61 noro 2555: nd_init_ord(ord);
2556: nd_setup_parameters(nvar,max);
2557: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 noro 2558: ishomo = 1;
1.61 noro 2559: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2560: c = ptondv(CO,vv,(P)BDY(t));
1.82 noro 2561: if ( ishomo )
2562: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2563: if ( c ) {
2564: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2565: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2566: }
1.61 noro 2567: }
2568: if ( in0 ) NEXT(in) = 0;
2569: if ( fd0 ) NEXT(fd) = 0;
1.82 noro 2570: if ( !ishomo && homo ) {
1.122 noro 2571: for ( t = in0, wmax = max; t; t = NEXT(t) ) {
1.61 noro 2572: c = (NDV)BDY(t); len = LEN(c);
2573: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2574: wmax = MAX(TD(DL(a)),wmax);
2575: }
2576: homogenize_order(ord,nvar,&ord1);
1.89 noro 2577: nd_init_ord(ord1);
1.61 noro 2578: nd_setup_parameters(nvar+1,wmax);
2579: for ( t = fd0; t; t = NEXT(t) )
2580: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2581: }
1.52 noro 2582: while ( 1 ) {
1.77 noro 2583: if ( Demand )
2584: nd_demand = 1;
1.128 ! noro 2585: ndv_setup(m,1,fd0,0);
1.82 noro 2586: cand = nd_gb_trace(m,ishomo || homo);
1.52 noro 2587: if ( !cand ) {
2588: /* failure */
1.61 noro 2589: if ( trace > 1 ) { *rp = 0; return; }
2590: else m = get_lprime(++mindex);
1.52 noro 2591: continue;
2592: }
1.82 noro 2593: if ( !ishomo && homo ) {
1.27 noro 2594: /* dehomogenization */
1.61 noro 2595: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2596: nd_init_ord(ord);
1.61 noro 2597: nd_setup_parameters(nvar,0);
1.27 noro 2598: }
1.77 noro 2599: nd_demand = 0;
1.61 noro 2600: cand = ndv_reducebase(cand);
2601: cand = ndv_reduceall(0,cand);
1.78 noro 2602: cbpe = nd_bpe;
1.61 noro 2603: if ( nocheck )
2604: break;
2605: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2606: /* success */
2607: break;
2608: else if ( trace > 1 ) {
2609: /* failure */
2610: *rp = 0; return;
1.61 noro 2611: } else {
1.52 noro 2612: /* try the next modulus */
2613: m = get_lprime(++mindex);
1.61 noro 2614: /* reset the parameters */
1.82 noro 2615: if ( !ishomo && homo ) {
1.89 noro 2616: nd_init_ord(ord1);
1.61 noro 2617: nd_setup_parameters(nvar+1,wmax);
2618: } else {
2619: nd_init_ord(ord);
2620: nd_setup_parameters(nvar,max);
2621: }
2622: }
1.52 noro 2623: }
1.27 noro 2624: /* dp->p */
1.61 noro 2625: nd_bpe = cbpe;
1.73 noro 2626: nd_setup_parameters(nd_nvar,0);
1.121 noro 2627: for ( r = cand; r; r = NEXT(r) )
1.119 noro 2628: BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.121 noro 2629: if ( nalg )
2630: cand = postprocess_algcoef(av,alist,cand);
1.27 noro 2631: MKLIST(*rp,cand);
1.1 noro 2632: }
2633:
1.61 noro 2634: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2635: {
1.61 noro 2636: UINT *d;
1.57 noro 2637: int i,j,l,s,ord_l;
1.43 noro 2638: struct order_pair *op;
1.1 noro 2639:
2640: d = dl->d;
1.41 noro 2641: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2642: if ( nd_blockmask ) {
2643: l = nd_blockmask->n;
2644: op = nd_blockmask->order_pair;
2645: for ( j = 0, s = 0; j < l; j++ ) {
2646: ord_l = op[j].length;
1.57 noro 2647: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2648: }
2649: TD(r) = ndl_weight(r);
1.61 noro 2650: ndl_weight_mask(r);
1.43 noro 2651: } else {
1.56 noro 2652: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2653: TD(r) = ndl_weight(r);
2654: }
1.1 noro 2655: }
2656:
1.61 noro 2657: DL ndltodl(int n,UINT *ndl)
1.1 noro 2658: {
2659: DL dl;
2660: int *d;
1.57 noro 2661: int i,j,l,s,ord_l;
1.43 noro 2662: struct order_pair *op;
1.1 noro 2663:
2664: NEWDL(dl,n);
1.34 noro 2665: dl->td = TD(ndl);
1.1 noro 2666: d = dl->d;
1.43 noro 2667: if ( nd_blockmask ) {
2668: l = nd_blockmask->n;
2669: op = nd_blockmask->order_pair;
2670: for ( j = 0, s = 0; j < l; j++ ) {
2671: ord_l = op[j].length;
1.57 noro 2672: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2673: }
2674: } else {
1.56 noro 2675: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2676: }
1.1 noro 2677: return dl;
2678: }
2679:
1.61 noro 2680: void ndl_print(UINT *dl)
1.1 noro 2681: {
2682: int n;
1.57 noro 2683: int i,j,l,ord_l,s,s0;
1.43 noro 2684: struct order_pair *op;
1.1 noro 2685:
2686: n = nd_nvar;
2687: printf("<<");
1.43 noro 2688: if ( nd_blockmask ) {
2689: l = nd_blockmask->n;
2690: op = nd_blockmask->order_pair;
2691: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2692: ord_l = op[j].length;
1.57 noro 2693: for ( i = 0; i < ord_l; i++, s++ )
2694: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2695: }
2696: } else {
1.56 noro 2697: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2698: }
1.1 noro 2699: printf(">>");
2700: }
2701:
2702: void nd_print(ND p)
2703: {
2704: NM m;
2705:
2706: if ( !p )
2707: printf("0\n");
2708: else {
2709: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2710: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2711: else printf("+%d*",CM(m));
1.14 noro 2712: ndl_print(DL(m));
1.1 noro 2713: }
2714: printf("\n");
2715: }
2716: }
2717:
1.113 noro 2718: void nd_print_q(ND p)
1.16 noro 2719: {
2720: NM m;
2721:
2722: if ( !p )
2723: printf("0\n");
2724: else {
2725: for ( m = BDY(p); m; m = NEXT(m) ) {
2726: printf("+");
1.113 noro 2727: printexpr(CO,(Obj)CQ(m));
1.16 noro 2728: printf("*");
2729: ndl_print(DL(m));
2730: }
2731: printf("\n");
2732: }
2733: }
2734:
1.1 noro 2735: void ndp_print(ND_pairs d)
2736: {
2737: ND_pairs t;
2738:
1.34 noro 2739: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2740: printf("\n");
2741: }
2742:
1.20 noro 2743: void nd_removecont(int mod,ND p)
1.16 noro 2744: {
2745: int i,n;
1.113 noro 2746: Q *w;
2747: Q dvr,t;
1.16 noro 2748: NM m;
1.21 noro 2749: struct oVECT v;
1.113 noro 2750: N q,r;
1.16 noro 2751:
1.71 noro 2752: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2753: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2754: else {
2755: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
1.113 noro 2756: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2757: v.len = n;
2758: v.body = (pointer *)w;
1.113 noro 2759: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2760: removecont_array(w,n);
1.113 noro 2761: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2762: }
2763: }
2764:
1.21 noro 2765: void nd_removecont2(ND p1,ND p2)
2766: {
2767: int i,n1,n2,n;
1.113 noro 2768: Q *w;
2769: Q dvr,t;
1.21 noro 2770: NM m;
2771: struct oVECT v;
1.113 noro 2772: N q,r;
1.21 noro 2773:
2774: if ( !p1 ) {
2775: nd_removecont(0,p2); return;
2776: } else if ( !p2 ) {
2777: nd_removecont(0,p1); return;
2778: }
2779: n1 = nd_length(p1);
2780: n2 = nd_length(p2);
2781: n = n1+n2;
1.113 noro 2782: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2783: v.len = n;
2784: v.body = (pointer *)w;
1.113 noro 2785: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2786: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2787: removecont_array(w,n);
1.113 noro 2788: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2789: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.21 noro 2790: }
2791:
1.20 noro 2792: void ndv_removecont(int mod,NDV p)
1.16 noro 2793: {
2794: int i,len;
1.113 noro 2795: Q *w;
2796: Q dvr,t;
1.16 noro 2797: NMV m;
2798:
1.71 noro 2799: if ( mod == -1 )
2800: ndv_mul_c(mod,p,_invsf(HCM(p)));
2801: else if ( mod )
1.20 noro 2802: ndv_mul_c(mod,p,invm(HCM(p),mod));
2803: else {
2804: len = p->len;
1.113 noro 2805: w = (Q *)ALLOCA(len*sizeof(Q));
2806: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
2807: sortbynm(w,len);
2808: qltozl(w,len,&dvr);
2809: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2810: divq(CQ(m),dvr,&t); CQ(m) = t;
2811: }
1.16 noro 2812: }
1.21 noro 2813: }
2814:
1.61 noro 2815: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2816: {
2817: int len,i,max;
2818: NMV m,mr0,mr,t;
2819:
2820: len = p->len;
2821: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2822: max = MAX(max,TD(DL(m)));
2823: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2824: m = (NMV)((char *)mr0+(len-1)*oadv);
2825: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2826: t = (NMV)ALLOCA(nmv_adv);
2827: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2828: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
1.113 noro 2829: CQ(mr) = CQ(m);
1.61 noro 2830: ndl_copy(DL(t),DL(mr));
2831: }
2832: NV(p)++;
2833: BDY(p) = mr0;
2834: }
2835:
1.45 noro 2836: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2837: {
1.45 noro 2838: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.113 noro 2839: Q *w;
2840: Q dvr,t;
1.23 noro 2841: NMV m,r;
2842:
2843: len = p->len;
2844: newnvar = nd_nvar-1;
1.48 noro 2845: newexporigin = nd_get_exporigin(ord);
1.45 noro 2846: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2847: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2848: ndl_dehomogenize(DL(m));
1.23 noro 2849: if ( newwpd != nd_wpd ) {
1.90 noro 2850: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
1.61 noro 2851: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.113 noro 2852: CQ(r) = CQ(m);
1.45 noro 2853: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2854: adj = nd_exporigin-newexporigin;
2855: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2856: }
2857: }
2858: NV(p)--;
2859: }
2860:
1.113 noro 2861: void removecont_array(Q *c,int n)
1.21 noro 2862: {
2863: struct oVECT v;
1.113 noro 2864: Q d0,d1,a,u,u1,gcd;
1.110 noro 2865: int i,j;
1.113 noro 2866: N qn,rn,gn;
2867: Q *q,*r;
1.21 noro 2868:
1.113 noro 2869: q = (Q *)ALLOCA(n*sizeof(Q));
2870: r = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2871: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
1.113 noro 2872: igcdv_estimate(&v,&d0);
2873: for ( i = 0; i < n; i++ ) {
2874: divn(NM(c[i]),NM(d0),&qn,&rn);
2875: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2876: NTOQ(rn,SGN(c[i]),r[i]);
2877: }
1.34 noro 2878: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2879: if ( i < n ) {
2880: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
1.113 noro 2881: igcdv(&v,&d1);
2882: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2883: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
1.21 noro 2884: for ( i = 0; i < n; i++ ) {
1.113 noro 2885: mulq(a,q[i],&u);
2886: if ( r[i] ) {
2887: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2888: addq(u,u1,&q[i]);
2889: } else
2890: q[i] = u;
1.21 noro 2891: }
2892: }
1.34 noro 2893: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2894: }
2895:
1.19 noro 2896: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2897: {
2898: NM m;
2899: int c,c1;
2900:
1.34 noro 2901: if ( !p ) return;
1.115 noro 2902: if ( mul == 1 ) return;
1.71 noro 2903: if ( mod == -1 )
2904: for ( m = BDY(p); m; m = NEXT(m) )
2905: CM(m) = _mulsf(CM(m),mul);
2906: else
2907: for ( m = BDY(p); m; m = NEXT(m) ) {
2908: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2909: }
1.1 noro 2910: }
2911:
1.113 noro 2912: void nd_mul_c_q(ND p,Q mul)
1.16 noro 2913: {
2914: NM m;
1.113 noro 2915: Q c;
1.16 noro 2916:
1.113 noro 2917: if ( !p ) return;
1.115 noro 2918: if ( UNIQ(mul) ) return;
1.16 noro 2919: for ( m = BDY(p); m; m = NEXT(m) ) {
1.113 noro 2920: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 2921: }
2922: }
2923:
1.61 noro 2924: void nd_mul_c_p(VL vl,ND p,P mul)
2925: {
2926: NM m;
2927: P c;
2928:
2929: if ( !p ) return;
2930: for ( m = BDY(p); m; m = NEXT(m) ) {
2931: mulp(vl,CP(m),mul,&c); CP(m) = c;
2932: }
2933: }
2934:
1.1 noro 2935: void nd_free(ND p)
2936: {
2937: NM t,s;
2938:
1.34 noro 2939: if ( !p ) return;
1.1 noro 2940: t = BDY(p);
2941: while ( t ) {
2942: s = NEXT(t);
2943: FREENM(t);
2944: t = s;
2945: }
2946: FREEND(p);
2947: }
2948:
1.23 noro 2949: void ndv_free(NDV p)
2950: {
2951: GC_free(BDY(p));
2952: }
2953:
1.61 noro 2954: void nd_append_red(UINT *d,int i)
1.1 noro 2955: {
1.13 noro 2956: RHist m,m0;
1.1 noro 2957: int h;
2958:
1.13 noro 2959: NEWRHist(m);
1.34 noro 2960: h = ndl_hash_value(d);
1.13 noro 2961: m->index = i;
1.14 noro 2962: ndl_copy(d,DL(m));
1.1 noro 2963: NEXT(m) = nd_red[h];
2964: nd_red[h] = m;
2965: }
2966:
1.61 noro 2967: UINT *ndv_compute_bound(NDV p)
1.1 noro 2968: {
1.61 noro 2969: UINT *d1,*d2,*t;
2970: UINT u;
1.57 noro 2971: int i,j,k,l,len,ind;
1.45 noro 2972: NMV m;
1.1 noro 2973:
2974: if ( !p )
2975: return 0;
1.61 noro 2976: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2977: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2978: len = LEN(p);
2979: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2980: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2981: ndl_lcm(DL(m),d1,d2);
1.1 noro 2982: t = d1; d1 = d2; d2 = t;
2983: }
1.12 noro 2984: l = nd_nvar+31;
1.61 noro 2985: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 2986: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2987: u = d1[i];
2988: k = (nd_epw-1)*nd_bpe;
2989: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2990: t[ind] = (u>>k)&nd_mask0;
2991: }
2992: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 2993: return t;
2994: }
2995:
1.99 noro 2996: UINT *nd_compute_bound(ND p)
2997: {
2998: UINT *d1,*d2,*t;
2999: UINT u;
3000: int i,j,k,l,len,ind;
3001: NM m;
3002:
3003: if ( !p )
3004: return 0;
3005: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3006: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3007: len = LEN(p);
3008: m = BDY(p); ndl_copy(DL(m),d1); m = NEXT(m);
3009: for ( m = NEXT(m); m; m = NEXT(m) ) {
3010: ndl_lcm(DL(m),d1,d2);
3011: t = d1; d1 = d2; d2 = t;
3012: }
3013: l = nd_nvar+31;
3014: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
3015: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3016: u = d1[i];
3017: k = (nd_epw-1)*nd_bpe;
3018: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3019: t[ind] = (u>>k)&nd_mask0;
3020: }
3021: for ( ; ind < l; ind++ ) t[ind] = 0;
3022: return t;
3023: }
3024:
1.48 noro 3025: int nd_get_exporigin(struct order_spec *ord)
3026: {
1.51 noro 3027: switch ( ord->id ) {
1.96 noro 3028: case 0: case 2:
1.48 noro 3029: return 1;
1.41 noro 3030: case 1:
3031: /* block order */
1.43 noro 3032: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 3033: return ord->ord.block.length+1;
1.96 noro 3034: case 3:
3035: error("nd_get_exporigin : composite order is not supported yet.");
1.41 noro 3036: }
1.48 noro 3037: }
3038:
1.61 noro 3039: void nd_setup_parameters(int nvar,int max) {
1.79 noro 3040: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 3041: struct order_pair *op;
1.48 noro 3042:
1.73 noro 3043: nd_nvar = nvar;
1.79 noro 3044: if ( max ) {
1.83 noro 3045: /* XXX */
3046: if ( do_weyl ) nd_bpe = 32;
3047: else if ( max < 2 ) nd_bpe = 1;
1.79 noro 3048: else if ( max < 4 ) nd_bpe = 2;
3049: else if ( max < 8 ) nd_bpe = 3;
3050: else if ( max < 16 ) nd_bpe = 4;
3051: else if ( max < 32 ) nd_bpe = 5;
3052: else if ( max < 64 ) nd_bpe = 6;
3053: else if ( max < 256 ) nd_bpe = 8;
3054: else if ( max < 1024 ) nd_bpe = 10;
3055: else if ( max < 65536 ) nd_bpe = 16;
3056: else nd_bpe = 32;
3057: }
1.61 noro 3058: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 3059: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
3060: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 3061: wpd = nd_exporigin+elen;
3062: if ( wpd != nd_wpd ) {
3063: nd_free_private_storage();
3064: nd_wpd = wpd;
3065: }
1.1 noro 3066: if ( nd_bpe < 32 ) {
3067: nd_mask0 = (1<<nd_bpe)-1;
3068: } else {
3069: nd_mask0 = 0xffffffff;
3070: }
3071: bzero(nd_mask,sizeof(nd_mask));
3072: nd_mask1 = 0;
3073: for ( i = 0; i < nd_epw; i++ ) {
3074: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
3075: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
3076: }
1.90 noro 3077: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
1.57 noro 3078: nd_epos = nd_create_epos(nd_ord);
1.43 noro 3079: nd_blockmask = nd_create_blockmask(nd_ord);
1.96 noro 3080: nd_work_vector = (int *)REALLOC(nd_work_vector,nd_nvar*sizeof(int));
1.1 noro 3081: }
3082:
1.103 noro 3083: ND_pairs nd_reconstruct(int trace,ND_pairs d)
1.1 noro 3084: {
1.37 noro 3085: int i,obpe,oadv,h;
1.71 noro 3086: static NM prev_nm_free_list;
3087: static ND_pairs prev_ndp_free_list;
1.13 noro 3088: RHist mr0,mr;
3089: RHist r;
1.37 noro 3090: RHist *old_red;
1.71 noro 3091: ND_pairs s0,s,t;
1.43 noro 3092: EPOS oepos;
1.15 noro 3093:
1.1 noro 3094: obpe = nd_bpe;
1.11 noro 3095: oadv = nmv_adv;
1.43 noro 3096: oepos = nd_epos;
1.63 noro 3097: if ( obpe < 2 ) nd_bpe = 2;
3098: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 3099: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 3100: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 3101: else if ( obpe < 6 ) nd_bpe = 6;
3102: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 3103: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 3104: else if ( obpe < 16 ) nd_bpe = 16;
3105: else if ( obpe < 32 ) nd_bpe = 32;
3106: else error("nd_reconstruct : exponent too large");
1.5 noro 3107:
1.73 noro 3108: nd_setup_parameters(nd_nvar,0);
1.1 noro 3109: prev_nm_free_list = _nm_free_list;
3110: prev_ndp_free_list = _ndp_free_list;
3111: _nm_free_list = 0;
3112: _ndp_free_list = 0;
1.53 noro 3113: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
3114: if ( trace )
3115: for ( i = nd_psn-1; i >= 0; i-- )
3116: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 3117: s0 = 0;
3118: for ( t = d; t; t = NEXT(t) ) {
3119: NEXTND_pairs(s0,s);
3120: s->i1 = t->i1;
3121: s->i2 = t->i2;
1.14 noro 3122: SG(s) = SG(t);
1.61 noro 3123: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 3124: }
1.37 noro 3125:
3126: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 3127: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 3128: old_red[i] = nd_red[i];
3129: nd_red[i] = 0;
3130: }
3131: for ( i = 0; i < REDTAB_LEN; i++ )
3132: for ( r = old_red[i]; r; r = NEXT(r) ) {
3133: NEWRHist(mr);
1.13 noro 3134: mr->index = r->index;
1.20 noro 3135: SG(mr) = SG(r);
1.61 noro 3136: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 3137: h = ndl_hash_value(DL(mr));
3138: NEXT(mr) = nd_red[h];
3139: nd_red[h] = mr;
1.6 noro 3140: }
1.37 noro 3141: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
3142: old_red = 0;
1.11 noro 3143: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 3144: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 3145: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 3146: nd_psh[i] = r;
1.11 noro 3147: }
1.1 noro 3148: if ( s0 ) NEXT(s) = 0;
3149: prev_nm_free_list = 0;
3150: prev_ndp_free_list = 0;
1.71 noro 3151: #if 0
1.1 noro 3152: GC_gcollect();
1.71 noro 3153: #endif
1.1 noro 3154: return s0;
3155: }
3156:
1.61 noro 3157: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 3158: {
1.57 noro 3159: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 3160: struct order_pair *op;
1.1 noro 3161:
3162: n = nd_nvar;
1.61 noro 3163: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 3164: omask0 = (1<<obpe)-1;
1.34 noro 3165: TD(r) = TD(d);
1.41 noro 3166: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 3167: if ( nd_blockmask ) {
3168: l = nd_blockmask->n;
3169: op = nd_blockmask->order_pair;
3170: for ( i = 1; i < nd_exporigin; i++ )
3171: r[i] = d[i];
3172: for ( j = 0, s = 0; j < l; j++ ) {
3173: ord_l = op[j].length;
1.57 noro 3174: for ( i = 0; i < ord_l; i++, s++ ) {
3175: ei = GET_EXP_OLD(d,s);
3176: PUT_EXP(r,s,ei);
3177: }
1.1 noro 3178: }
1.43 noro 3179: } else {
1.56 noro 3180: for ( i = 0; i < n; i++ ) {
3181: ei = GET_EXP_OLD(d,i);
3182: PUT_EXP(r,i,ei);
3183: }
1.1 noro 3184: }
3185: }
1.3 noro 3186:
1.6 noro 3187: ND nd_copy(ND p)
3188: {
3189: NM m,mr,mr0;
1.41 noro 3190: int c,n;
1.6 noro 3191: ND r;
3192:
3193: if ( !p )
3194: return 0;
3195: else {
3196: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3197: NEXTNM(mr0,mr);
1.14 noro 3198: CM(mr) = CM(m);
3199: ndl_copy(DL(m),DL(mr));
1.6 noro 3200: }
3201: NEXT(mr) = 0;
1.31 noro 3202: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 3203: SG(r) = SG(p);
1.6 noro 3204: return r;
3205: }
3206: }
3207:
1.53 noro 3208: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3209: {
3210: NM m;
3211: NDV p1,p2;
3212: ND t1,t2;
1.61 noro 3213: UINT *lcm;
1.31 noro 3214: int td;
1.11 noro 3215:
1.77 noro 3216: if ( !mod && nd_demand ) {
3217: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 3218: } else {
1.77 noro 3219: if ( trace ) {
3220: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3221: } else {
3222: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3223: }
1.20 noro 3224: }
1.34 noro 3225: lcm = LCM(p);
1.11 noro 3226: NEWNM(m);
1.113 noro 3227: CQ(m) = HCQ(p2);
1.34 noro 3228: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 3229: if ( ndl_check_bound2(p->i1,DL(m)) )
3230: return 0;
1.55 noro 3231: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 3232: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
3233: else if ( mod ) CM(m) = mod-HCM(p1);
1.113 noro 3234: else chsgnq(HCQ(p1),&CQ(m));
1.34 noro 3235: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 3236: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 3237: nd_free(t1);
3238: return 0;
3239: }
1.55 noro 3240: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 3241: *rp = nd_add(mod,t1,t2);
1.11 noro 3242: FREENM(m);
3243: return 1;
3244: }
3245:
1.19 noro 3246: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3247: {
3248: NMV m;
3249: int c,c1,len,i;
3250:
1.34 noro 3251: if ( !p ) return;
1.14 noro 3252: len = LEN(p);
1.71 noro 3253: if ( mod == -1 )
3254: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3255: CM(m) = _mulsf(CM(m),mul);
3256: else
3257: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3258: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3259: }
1.11 noro 3260: }
3261:
1.113 noro 3262: void ndv_mul_c_q(NDV p,Q mul)
1.16 noro 3263: {
3264: NMV m;
1.113 noro 3265: Q c;
1.16 noro 3266: int len,i;
3267:
1.34 noro 3268: if ( !p ) return;
1.16 noro 3269: len = LEN(p);
3270: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 3271: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 3272: }
3273: }
3274:
1.55 noro 3275: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3276: int n2,i,j,l,n,tlen;
1.61 noro 3277: UINT *d0;
1.55 noro 3278: NM *tab,*psum;
3279: ND s,r;
3280: NM t;
3281: NMV m1;
3282:
3283: if ( !p ) return 0;
3284: n = NV(p); n2 = n>>1;
3285: d0 = DL(m0);
3286: l = LEN(p);
3287: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3288: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3289: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3290: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3291: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3292: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3293: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3294: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3295: for ( j = 0; j < tlen; j++ ) {
3296: if ( tab[j] ) {
3297: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3298: }
3299: }
3300: }
3301: for ( i = tlen-1, r = 0; i >= 0; i-- )
3302: if ( psum[i] ) {
3303: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3304: MKND(n,psum[i],j,s);
3305: r = nd_add(mod,r,s);
3306: }
1.56 noro 3307: if ( r ) SG(r) = SG(p)+TD(d0);
3308: return r;
1.55 noro 3309: }
3310:
1.56 noro 3311: /* product of monomials */
3312: /* XXX block order is not handled correctly */
3313:
1.55 noro 3314: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3315: {
1.56 noro 3316: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3317: UINT *d0,*d1,*d,*dt,*ctab;
1.113 noro 3318: Q *ctab_q;
3319: Q q,q1;
1.61 noro 3320: UINT c0,c1,c;
1.55 noro 3321: NM *p;
3322: NM m,t;
3323:
3324: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3325: if ( !m0 || !m1 ) return;
3326: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3327: NEWNM(m); d = DL(m);
1.56 noro 3328: if ( mod ) {
3329: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3330: } else
1.113 noro 3331: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3332: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3333: homo = n&1 ? 1 : 0;
3334: if ( homo ) {
3335: /* offset of h-degree */
3336: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3337: PUT_EXP(DL(m),n-1,h);
3338: TD(DL(m)) = h;
1.61 noro 3339: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3340: }
3341: tab[0] = m;
3342: NEWNM(m); d = DL(m);
1.57 noro 3343: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3344: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3345: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3346: /* xi^a*(Di^k*xi^l)*Di^b */
3347: a += l; b += k;
1.56 noro 3348: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3349: if ( !k || !l ) {
3350: for ( j = 0; j < curlen; j++ )
1.56 noro 3351: if ( t = tab[j] ) {
3352: dt = DL(t);
3353: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3354: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3355: }
3356: curlen *= k+1;
3357: continue;
3358: }
3359: min = MIN(k,l);
1.56 noro 3360: if ( mod ) {
1.61 noro 3361: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3362: mkwcm(k,l,mod,ctab);
3363: } else {
1.113 noro 3364: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3365: mkwc(k,l,ctab_q);
1.56 noro 3366: }
1.57 noro 3367: for ( j = min; j >= 0; j-- ) {
1.56 noro 3368: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3369: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3370: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3371: if ( homo ) {
3372: TD(d) = s;
1.56 noro 3373: PUT_EXP(d,n-1,s-h);
1.55 noro 3374: } else TD(d) = h;
1.61 noro 3375: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3376: if ( mod ) c = ctab[j];
1.113 noro 3377: else q = ctab_q[j];
1.57 noro 3378: p = tab+curlen*j;
3379: if ( j == 0 ) {
3380: for ( u = 0; u < curlen; u++, p++ ) {
3381: if ( tab[u] ) {
3382: ndl_addto(DL(tab[u]),d);
3383: if ( mod ) {
3384: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3385: } else {
1.113 noro 3386: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
1.57 noro 3387: }
3388: }
1.56 noro 3389: }
1.57 noro 3390: } else {
3391: for ( u = 0; u < curlen; u++, p++ ) {
3392: if ( tab[u] ) {
3393: NEWNM(t);
3394: ndl_add(DL(tab[u]),d,DL(t));
3395: if ( mod ) {
3396: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3397: } else
1.113 noro 3398: mulq(CQ(tab[u]),q,&CQ(t));
1.57 noro 3399: *p = t;
3400: }
1.55 noro 3401: }
3402: }
3403: }
3404: curlen *= k+1;
3405: }
3406: FREENM(m);
3407: }
3408:
1.63 noro 3409: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3410: {
3411: NM mr,mr0;
3412: NMV m;
3413: UINT *d,*dt,*dm;
3414: int c,n,td,i,c1,c2,len;
1.113 noro 3415: Q q;
1.63 noro 3416: ND r;
3417:
3418: if ( !p ) return 0;
3419: else {
3420: n = NV(p); m = BDY(p);
3421: d = DL(m0);
3422: len = LEN(p);
3423: mr0 = 0;
3424: td = TD(d);
3425: c = CM(m0);
3426: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3427: NEXTNM(mr0,mr);
3428: CM(mr) = 1;
3429: ndl_add(DL(m),d,DL(mr));
3430: }
3431: NEXT(mr) = 0;
3432: MKND(NV(p),mr0,len,r);
3433: SG(r) = SG(p) + TD(d);
3434: return r;
3435: }
3436: }
3437:
1.55 noro 3438: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3439: {
3440: NM mr,mr0;
3441: NMV m;
1.61 noro 3442: UINT *d,*dt,*dm;
1.9 noro 3443: int c,n,td,i,c1,c2,len;
1.113 noro 3444: Q q;
1.9 noro 3445: ND r;
3446:
1.34 noro 3447: if ( !p ) return 0;
1.55 noro 3448: else if ( do_weyl )
1.71 noro 3449: if ( mod == -1 )
3450: error("ndv_mul_nm : not implemented (weyl)");
3451: else
3452: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3453: else {
3454: n = NV(p); m = BDY(p);
1.34 noro 3455: d = DL(m0);
1.14 noro 3456: len = LEN(p);
1.9 noro 3457: mr0 = 0;
1.34 noro 3458: td = TD(d);
1.71 noro 3459: if ( mod == -1 ) {
3460: c = CM(m0);
3461: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3462: NEXTNM(mr0,mr);
3463: CM(mr) = _mulsf(CM(m),c);
3464: ndl_add(DL(m),d,DL(mr));
3465: }
3466: } else if ( mod ) {
1.16 noro 3467: c = CM(m0);
3468: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3469: NEXTNM(mr0,mr);
3470: c1 = CM(m);
1.19 noro 3471: DMAR(c1,c,0,mod,c2);
1.16 noro 3472: CM(mr) = c2;
3473: ndl_add(DL(m),d,DL(mr));
3474: }
3475: } else {
1.113 noro 3476: q = CQ(m0);
1.16 noro 3477: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3478: NEXTNM(mr0,mr);
1.113 noro 3479: mulq(CQ(m),q,&CQ(mr));
1.16 noro 3480: ndl_add(DL(m),d,DL(mr));
3481: }
1.4 noro 3482: }
1.9 noro 3483: NEXT(mr) = 0;
1.31 noro 3484: MKND(NV(p),mr0,len,r);
1.34 noro 3485: SG(r) = SG(p) + TD(d);
1.9 noro 3486: return r;
1.4 noro 3487: }
3488: }
3489:
1.104 noro 3490: ND nd_quo(int mod,PGeoBucket bucket,NDV d)
1.99 noro 3491: {
3492: NM mq0,mq;
1.102 noro 3493: NMV tm;
1.113 noro 3494: Q q;
1.104 noro 3495: int i,nv,sg,c,c1,c2,hindex;
3496: ND p,t,r;
3497: N tnm;
3498:
1.124 noro 3499: if ( bucket->m < 0 ) return 0;
1.99 noro 3500: else {
1.104 noro 3501: nv = NV(d);
1.99 noro 3502: mq0 = 0;
1.102 noro 3503: tm = (NMV)ALLOCA(nmv_adv);
1.104 noro 3504: while ( 1 ) {
1.113 noro 3505: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.104 noro 3506: if ( hindex < 0 ) break;
3507: p = bucket->body[hindex];
1.99 noro 3508: NEXTNM(mq0,mq);
3509: ndl_sub(HDL(p),HDL(d),DL(mq));
1.102 noro 3510: ndl_copy(DL(mq),DL(tm));
3511: if ( mod ) {
3512: c1 = invm(HCM(d),mod); c2 = HCM(p);
3513: DMAR(c1,c2,0,mod,c); CM(mq) = c;
3514: CM(tm) = mod-c;
3515: } else {
1.113 noro 3516: divsn(NM(HCQ(p)),NM(HCQ(d)),&tnm);
3517: NTOQ(tnm,SGN(HCQ(p))*SGN(HCQ(d)),CQ(mq));
3518: chsgnq(CQ(mq),&CQ(tm));
1.102 noro 3519: }
3520: t = ndv_mul_nmv_trunc(mod,tm,d,HDL(d));
1.104 noro 3521: bucket->body[hindex] = nd_remove_head(p);
3522: t = nd_remove_head(t);
3523: add_pbucket(mod,bucket,t);
3524: }
3525: if ( !mq0 )
3526: r = 0;
3527: else {
3528: NEXT(mq) = 0;
3529: for ( i = 0, mq = mq0; mq; mq = NEXT(mq), i++ );
3530: MKND(nv,mq0,i,r);
3531: /* XXX */
3532: SG(r) = HTD(r);
1.99 noro 3533: }
3534: return r;
3535: }
3536: }
3537:
1.43 noro 3538: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3539: {
1.13 noro 3540: NMV m,mr,mr0,t;
3541: int len,i,k;
1.11 noro 3542:
1.61 noro 3543: if ( !p ) return;
3544: m = BDY(p); len = LEN(p);
3545: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3546: m = (NMV)((char *)mr0+(len-1)*oadv);
3547: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3548: t = (NMV)ALLOCA(nmv_adv);
3549: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
1.113 noro 3550: CQ(t) = CQ(m);
1.61 noro 3551: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3552: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
1.113 noro 3553: CQ(mr) = CQ(t);
1.61 noro 3554: ndl_copy(DL(t),DL(mr));
3555: }
3556: BDY(p) = mr0;
3557: }
3558:
3559: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3560: {
3561: NMV m,mr,mr0;
3562: int len,i;
3563: NDV r;
1.11 noro 3564:
1.61 noro 3565: if ( !p ) return 0;
3566: m = BDY(p); len = LEN(p);
3567: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3568: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3569: ndl_zero(DL(mr));
3570: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
1.113 noro 3571: CQ(mr) = CQ(m);
1.11 noro 3572: }
1.61 noro 3573: MKNDV(NV(p),mr0,len,r);
3574: SG(r) = SG(p);
3575: return r;
1.11 noro 3576: }
3577:
1.61 noro 3578: /* duplicate p */
3579:
3580: NDV ndv_dup(int mod,NDV p)
1.3 noro 3581: {
3582: NDV d;
1.61 noro 3583: NMV t,m,m0;
1.3 noro 3584: int i,len;
3585:
1.34 noro 3586: if ( !p ) return 0;
1.31 noro 3587: len = LEN(p);
1.34 noro 3588: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3589: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3590: ndl_copy(DL(t),DL(m));
1.113 noro 3591: CQ(m) = CQ(t);
1.3 noro 3592: }
3593: MKNDV(NV(p),m0,len,d);
1.23 noro 3594: SG(d) = SG(p);
3595: return d;
3596: }
3597:
1.63 noro 3598: ND nd_dup(ND p)
3599: {
3600: ND d;
3601: NM t,m,m0;
3602:
3603: if ( !p ) return 0;
3604: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3605: NEXTNM(m0,m);
3606: ndl_copy(DL(t),DL(m));
1.113 noro 3607: CQ(m) = CQ(t);
1.63 noro 3608: }
3609: if ( m0 ) NEXT(m) = 0;
3610: MKND(NV(p),m0,LEN(p),d);
3611: SG(d) = SG(p);
3612: return d;
3613: }
3614:
1.61 noro 3615: /* XXX if p->len == 0 then it represents 0 */
3616:
3617: void ndv_mod(int mod,NDV p)
3618: {
3619: NMV t,d;
1.125 noro 3620: int r,s,u;
1.61 noro 3621: int i,len,dlen;
1.71 noro 3622: Obj gfs;
1.61 noro 3623:
3624: if ( !p ) return;
3625: len = LEN(p);
3626: dlen = 0;
1.71 noro 3627: if ( mod == -1 )
3628: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3629: simp_ff((Obj)CP(t),&gfs);
3630: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3631: CM(d) = r;
3632: ndl_copy(DL(t),DL(d));
3633: NMV_ADV(d);
3634: dlen++;
3635: }
1.71 noro 3636: else
3637: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
1.113 noro 3638: r = rem(NM(CQ(t)),mod);
1.71 noro 3639: if ( r ) {
1.113 noro 3640: if ( SGN(CQ(t)) < 0 )
3641: r = mod-r;
1.125 noro 3642: if ( DN(CQ(t)) ) {
3643: s = rem(DN(CQ(t)),mod);
3644: if ( !s )
3645: error("ndv_mod : division by 0");
3646: s = invm(s,mod);
3647: DMAR(r,s,0,mod,u); r = u;
3648: }
1.71 noro 3649: CM(d) = r;
3650: ndl_copy(DL(t),DL(d));
3651: NMV_ADV(d);
3652: dlen++;
3653: }
3654: }
1.61 noro 3655: LEN(p) = dlen;
3656: }
3657:
3658: NDV ptondv(VL vl,VL dvl,P p)
3659: {
3660: ND nd;
3661:
3662: nd = ptond(vl,dvl,p);
3663: return ndtondv(0,nd);
3664: }
3665:
3666: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3667: {
1.61 noro 3668: int n,i,j,k,e;
3669: VL tvl;
3670: V v;
3671: DCP dc;
3672: DCP *w;
3673: ND r,s,t,u;
3674: P x;
3675: int c;
3676: UINT *d;
1.23 noro 3677: NM m,m0;
1.61 noro 3678:
3679: if ( !p )
3680: return 0;
3681: else if ( NUM(p) ) {
3682: NEWNM(m);
3683: ndl_zero(DL(m));
1.113 noro 3684: CQ(m) = (Q)p;
1.61 noro 3685: NEXT(m) = 0;
3686: MKND(nd_nvar,m,1,r);
3687: SG(r) = 0;
3688: return r;
3689: } else {
3690: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3691: w = (DCP *)ALLOCA(k*sizeof(DCP));
3692: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3693: for ( i = 0, tvl = dvl, v = VR(p);
1.98 noro 3694: tvl && tvl->v != v; tvl = NEXT(tvl), i++ );
1.61 noro 3695: if ( !tvl ) {
3696: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3697: t = ptond(vl,dvl,COEF(w[j]));
3698: pwrp(vl,x,DEG(w[j]),&p);
3699: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3700: }
3701: return s;
3702: } else {
3703: NEWNM(m0); d = DL(m0);
3704: for ( j = k-1, s = 0; j >= 0; j-- ) {
3705: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3706: TD(d) = MUL_WEIGHT(e,i);
3707: if ( nd_blockmask) ndl_weight_mask(d);
3708: t = ptond(vl,dvl,COEF(w[j]));
3709: for ( m = BDY(t); m; m = NEXT(m) )
3710: ndl_addto(DL(m),d);
3711: SG(t) += TD(d);
3712: s = nd_add(0,s,t);
3713: }
3714: FREENM(m0);
3715: return s;
3716: }
3717: }
3718: }
3719:
3720: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3721: {
3722: VL tvl;
3723: int len,n,j,i,e;
3724: NMV m;
3725: Q q;
3726: P c;
3727: UINT *d;
3728: P s,r,u,t,w;
1.71 noro 3729: GFS gfs;
1.23 noro 3730:
1.34 noro 3731: if ( !p ) return 0;
1.61 noro 3732: else {
3733: len = LEN(p);
3734: n = NV(p);
3735: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3736: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3737: if ( mod == -1 ) {
3738: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3739: } else if ( mod ) {
1.61 noro 3740: STOQ(CM(m),q); c = (P)q;
3741: } else
1.113 noro 3742: c = CP(m);
1.61 noro 3743: d = DL(m);
3744: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3745: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3746: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3747: }
3748: addp(vl,s,t,&u); s = u;
3749: }
3750: return s;
1.23 noro 3751: }
1.3 noro 3752: }
3753:
1.61 noro 3754: NDV ndtondv(int mod,ND p)
1.11 noro 3755: {
3756: NDV d;
1.61 noro 3757: NMV m,m0;
3758: NM t;
3759: int i,len;
1.11 noro 3760:
1.34 noro 3761: if ( !p ) return 0;
1.61 noro 3762: len = LEN(p);
1.74 noro 3763: if ( mod )
3764: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3765: else
3766: m0 = m = MALLOC(len*nmv_adv);
1.103 noro 3767: #if 0
1.74 noro 3768: ndv_alloc += nmv_adv*len;
1.103 noro 3769: #endif
1.61 noro 3770: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3771: ndl_copy(DL(t),DL(m));
1.113 noro 3772: CQ(m) = CQ(t);
1.11 noro 3773: }
1.61 noro 3774: MKNDV(NV(p),m0,len,d);
1.14 noro 3775: SG(d) = SG(p);
1.11 noro 3776: return d;
3777: }
3778:
1.61 noro 3779: ND ndvtond(int mod,NDV p)
1.11 noro 3780: {
1.61 noro 3781: ND d;
3782: NM m,m0;
1.11 noro 3783: NMV t;
1.61 noro 3784: int i,len;
1.11 noro 3785:
1.34 noro 3786: if ( !p ) return 0;
1.11 noro 3787: m0 = 0;
1.61 noro 3788: len = p->len;
3789: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3790: NEXTNM(m0,m);
3791: ndl_copy(DL(t),DL(m));
1.113 noro 3792: CQ(m) = CQ(t);
1.11 noro 3793: }
3794: NEXT(m) = 0;
1.61 noro 3795: MKND(NV(p),m0,len,d);
1.14 noro 3796: SG(d) = SG(p);
1.11 noro 3797: return d;
3798: }
3799:
1.3 noro 3800: void ndv_print(NDV p)
3801: {
3802: NMV m;
3803: int i,len;
3804:
1.34 noro 3805: if ( !p ) printf("0\n");
1.3 noro 3806: else {
1.14 noro 3807: len = LEN(p);
1.3 noro 3808: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3809: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3810: else printf("+%d*",CM(m));
1.16 noro 3811: ndl_print(DL(m));
3812: }
3813: printf("\n");
3814: }
3815: }
3816:
1.113 noro 3817: void ndv_print_q(NDV p)
1.16 noro 3818: {
3819: NMV m;
3820: int i,len;
3821:
1.34 noro 3822: if ( !p ) printf("0\n");
1.16 noro 3823: else {
3824: len = LEN(p);
3825: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3826: printf("+");
1.113 noro 3827: printexpr(CO,(Obj)CQ(m));
1.16 noro 3828: printf("*");
1.14 noro 3829: ndl_print(DL(m));
1.3 noro 3830: }
3831: printf("\n");
3832: }
1.25 noro 3833: }
3834:
1.61 noro 3835: NODE ndv_reducebase(NODE x)
1.27 noro 3836: {
3837: int len,i,j;
3838: NDV *w;
3839: NODE t,t0;
3840:
3841: len = length(x);
3842: w = (NDV *)ALLOCA(len*sizeof(NDV));
3843: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3844: for ( i = 0; i < len; i++ ) {
3845: for ( j = 0; j < i; j++ ) {
3846: if ( w[i] && w[j] )
3847: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3848: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3849: }
3850: }
3851: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3852: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3853: }
3854: NEXT(t) = 0; x = t0;
3855: return x;
1.11 noro 3856: }
1.32 noro 3857:
1.43 noro 3858: /* XXX incomplete */
3859:
1.32 noro 3860: void nd_init_ord(struct order_spec *ord)
3861: {
1.43 noro 3862: switch ( ord->id ) {
1.32 noro 3863: case 0:
1.43 noro 3864: switch ( ord->ord.simple ) {
3865: case 0:
3866: nd_dcomp = 1;
3867: nd_isrlex = 1;
3868: break;
3869: case 1:
3870: nd_dcomp = 1;
3871: nd_isrlex = 0;
3872: break;
3873: case 2:
3874: nd_dcomp = 0;
3875: nd_isrlex = 0;
1.45 noro 3876: ndl_compare_function = ndl_lex_compare;
1.58 noro 3877: break;
3878: case 11:
3879: /* XXX */
3880: nd_dcomp = 0;
3881: nd_isrlex = 1;
3882: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3883: break;
3884: default:
3885: error("nd_gr : unsupported order");
3886: }
1.32 noro 3887: break;
3888: case 1:
1.96 noro 3889: /* block order */
1.43 noro 3890: /* XXX */
3891: nd_dcomp = -1;
1.32 noro 3892: nd_isrlex = 0;
1.45 noro 3893: ndl_compare_function = ndl_block_compare;
1.34 noro 3894: break;
1.43 noro 3895: case 2:
1.96 noro 3896: /* matrix order */
3897: /* XXX */
3898: nd_dcomp = -1;
3899: nd_isrlex = 0;
3900: nd_matrix_len = ord->ord.matrix.row;
3901: nd_matrix = ord->ord.matrix.matrix;
3902: ndl_compare_function = ndl_matrix_compare;
3903: break;
3904: case 3:
1.97 noro 3905: /* composite order */
3906: nd_dcomp = -1;
3907: nd_isrlex = 0;
3908: nd_worb_len = ord->ord.composite.length;
3909: nd_worb = ord->ord.composite.w_or_b;
3910: ndl_compare_function = ndl_composite_compare;
1.32 noro 3911: break;
3912: }
1.41 noro 3913: nd_ord = ord;
1.32 noro 3914: }
3915:
1.43 noro 3916: BlockMask nd_create_blockmask(struct order_spec *ord)
3917: {
3918: int n,i,j,s,l;
1.61 noro 3919: UINT *t;
1.43 noro 3920: BlockMask bm;
3921:
1.96 noro 3922: /* we only create mask table for block order */
3923: if ( ord->id != 1 )
1.43 noro 3924: return 0;
3925: n = ord->ord.block.length;
3926: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3927: bm->n = n;
3928: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3929: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3930: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3931: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3932: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3933: l = bm->order_pair[i].length;
3934: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3935: }
3936: return bm;
1.57 noro 3937: }
3938:
3939: EPOS nd_create_epos(struct order_spec *ord)
3940: {
3941: int i,j,l,s,ord_l,ord_o;
3942: EPOS epos;
3943: struct order_pair *op;
3944:
3945: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3946: switch ( ord->id ) {
3947: case 0:
3948: if ( nd_isrlex ) {
3949: for ( i = 0; i < nd_nvar; i++ ) {
3950: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3951: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3952: }
3953: } else {
3954: for ( i = 0; i < nd_nvar; i++ ) {
3955: epos[i].i = nd_exporigin + i/nd_epw;
3956: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3957: }
3958: }
3959: break;
3960: case 1:
3961: /* block order */
3962: l = ord->ord.block.length;
3963: op = ord->ord.block.order_pair;
3964: for ( j = 0, s = 0; j < l; j++ ) {
3965: ord_o = op[j].order;
3966: ord_l = op[j].length;
3967: if ( !ord_o )
3968: for ( i = 0; i < ord_l; i++ ) {
3969: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3970: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3971: }
3972: else
3973: for ( i = 0; i < ord_l; i++ ) {
3974: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3975: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3976: }
3977: s += ord_l;
3978: }
3979: break;
3980: case 2:
1.96 noro 3981: /* matrix order */
3982: case 3:
3983: /* composite order */
3984: for ( i = 0; i < nd_nvar; i++ ) {
3985: epos[i].i = nd_exporigin + i/nd_epw;
3986: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3987: }
3988: break;
1.57 noro 3989: }
3990: return epos;
1.43 noro 3991: }
1.59 noro 3992:
3993: /* external interface */
3994:
3995: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
3996: {
1.61 noro 3997: NODE t,in0,in;
1.59 noro 3998: ND nd,nf;
1.61 noro 3999: NDV ndv;
4000: VL vv,tv;
4001: int stat,nvar,max,e;
1.113 noro 4002: union oNDC dn;
1.59 noro 4003:
1.116 noro 4004: if ( !f ) {
4005: *rp = 0;
4006: return;
4007: }
1.59 noro 4008: pltovl(v,&vv);
1.74 noro 4009: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 4010:
4011: /* get the degree bound */
4012: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
4013: for ( tv = vv; tv; tv = NEXT(tv) ) {
4014: e = getdeg(tv->v,(P)BDY(t));
4015: max = MAX(e,max);
4016: }
4017: for ( tv = vv; tv; tv = NEXT(tv) ) {
4018: e = getdeg(tv->v,f);
4019: max = MAX(e,max);
4020: }
4021:
1.59 noro 4022: nd_init_ord(ord);
1.61 noro 4023: nd_setup_parameters(nvar,max);
4024:
4025: /* conversion to ndv */
4026: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
4027: NEXTNODE(in0,in);
4028: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 4029: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4030: }
4031: NEXTNODE(in0,in);
4032: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 4033: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4034: NEXT(in) = 0;
4035:
1.128 ! noro 4036: /* dont sort */
! 4037: ndv_setup(m,0,in0,1);
1.59 noro 4038: nd_psn--;
4039: nd_scale=2;
4040: while ( 1 ) {
4041: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 4042: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 4043: if ( !stat ) {
4044: nd_psn++;
1.103 noro 4045: nd_reconstruct(0,0);
1.59 noro 4046: nd_psn--;
4047: } else
4048: break;
4049: }
1.61 noro 4050: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 4051: }
4052:
4053: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
4054: {
4055: NM m;
4056: UINT *t,*s;
4057: int i;
4058:
4059: for ( i = 0; i < n; i++ ) r[i] = 0;
4060: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
4061: t = DL(m);
4062: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4063: r[i] = CM(m);
4064: }
4065: for ( i = 0; !r[i]; i++ );
4066: return i;
4067: }
4068:
1.113 noro 4069: int nd_to_vect_q(UINT *s0,int n,ND d,Q *r)
1.74 noro 4070: {
1.107 noro 4071: NM m;
1.74 noro 4072: UINT *t,*s;
1.107 noro 4073: int i;
1.74 noro 4074:
4075: for ( i = 0; i < n; i++ ) r[i] = 0;
1.107 noro 4076: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
1.74 noro 4077: t = DL(m);
4078: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.113 noro 4079: r[i] = CQ(m);
1.74 noro 4080: }
4081: for ( i = 0; !r[i]; i++ );
4082: return i;
4083: }
4084:
1.67 noro 4085: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 4086: {
4087: NM m;
4088: NMV mr;
4089: UINT *d,*t,*s;
4090: NDV p;
1.67 noro 4091: unsigned char *ivc;
4092: unsigned short *ivs;
1.81 noro 4093: UINT *v,*ivi,*s0v;
1.67 noro 4094: int i,j,len,prev,diff,cdiff;
4095: IndArray r;
1.64 noro 4096:
4097: m = pair->mul;
4098: d = DL(m);
4099: p = nd_ps[pair->index];
4100: len = LEN(p);
4101: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 4102: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 4103: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4104: ndl_add(d,DL(mr),t);
4105: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 4106: v[j] = i;
4107: }
1.81 noro 4108: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 4109: r->head = v[0];
4110: diff = 0;
4111: for ( i = 1; i < len; i++ ) {
4112: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
4113: }
4114: if ( diff < 256 ) {
4115: r->width = 1;
4116: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
4117: r->index.c = ivc;
4118: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
4119: } else if ( diff < 65536 ) {
4120: r->width = 2;
4121: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
4122: r->index.s = ivs;
4123: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
4124: } else {
4125: r->width = 4;
4126: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
4127: r->index.i = ivi;
4128: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 4129: }
1.67 noro 4130: return r;
1.64 noro 4131: }
4132:
1.65 noro 4133:
1.113 noro 4134: int ndv_reduce_vect_q(Q *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.107 noro 4135: {
4136: int i,j,k,len,pos,prev;
1.113 noro 4137: Q cs,mcs,c1,c2,cr,gcd,t;
1.107 noro 4138: IndArray ivect;
4139: unsigned char *ivc;
4140: unsigned short *ivs;
4141: unsigned int *ivi;
4142: NDV redv;
4143: NMV mr;
4144: NODE rp;
4145: int maxrs;
4146:
4147: maxrs = 0;
4148: for ( i = 0; i < nred; i++ ) {
4149: ivect = imat[i];
4150: k = ivect->head;
4151: if ( svect[k] ) {
4152: maxrs = MAX(maxrs,rp0[i]->sugar);
4153: redv = nd_ps[rp0[i]->index];
4154: len = LEN(redv); mr = BDY(redv);
1.113 noro 4155: igcd_cofactor(svect[k],CQ(mr),&gcd,&cs,&cr);
4156: chsgnq(cs,&mcs);
4157: if ( !UNIQ(cr) ) {
4158: for ( j = 0; j < col; j++ ) {
4159: mulq(svect[j],cr,&c1); svect[j] = c1;
4160: }
4161: }
1.107 noro 4162: svect[k] = 0; prev = k;
4163: switch ( ivect->width ) {
4164: case 1:
4165: ivc = ivect->index.c;
4166: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4167: pos = prev+ivc[j]; prev = pos;
1.113 noro 4168: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4169: }
4170: break;
4171: case 2:
4172: ivs = ivect->index.s;
4173: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4174: pos = prev+ivs[j]; prev = pos;
1.113 noro 4175: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4176: }
4177: break;
4178: case 4:
4179: ivi = ivect->index.i;
4180: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4181: pos = prev+ivi[j]; prev = pos;
1.113 noro 4182: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4183: }
4184: break;
4185: }
4186: }
4187: }
4188: return maxrs;
4189: }
4190:
1.76 noro 4191: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 4192: {
1.67 noro 4193: int i,j,k,len,pos,prev;
1.66 noro 4194: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 4195: IndArray ivect;
4196: unsigned char *ivc;
4197: unsigned short *ivs;
4198: unsigned int *ivi;
1.65 noro 4199: NDV redv;
1.67 noro 4200: NMV mr;
1.65 noro 4201: NODE rp;
1.76 noro 4202: int maxrs;
1.65 noro 4203:
1.76 noro 4204: maxrs = 0;
1.74 noro 4205: for ( i = 0; i < nred; i++ ) {
1.65 noro 4206: ivect = imat[i];
1.67 noro 4207: k = ivect->head; svect[k] %= m;
1.65 noro 4208: if ( c = svect[k] ) {
1.76 noro 4209: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4210: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 4211: len = LEN(redv); mr = BDY(redv);
4212: svect[k] = 0; prev = k;
4213: switch ( ivect->width ) {
4214: case 1:
4215: ivc = ivect->index.c;
4216: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4217: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
4218: prev = pos;
4219: DMA(c1,c,c2,up,lo);
4220: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4221: } else svect[pos] = lo;
4222: }
4223: break;
4224: case 2:
4225: ivs = ivect->index.s;
4226: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4227: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
4228: prev = pos;
4229: DMA(c1,c,c2,up,lo);
4230: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4231: } else svect[pos] = lo;
4232: }
4233: break;
4234: case 4:
4235: ivi = ivect->index.i;
4236: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4237: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
4238: prev = pos;
4239: DMA(c1,c,c2,up,lo);
4240: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4241: } else svect[pos] = lo;
4242: }
4243: break;
1.65 noro 4244: }
4245: }
4246: }
1.66 noro 4247: for ( i = 0; i < col; i++ )
4248: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 4249: return maxrs;
1.65 noro 4250: }
4251:
1.76 noro 4252: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 4253: {
4254: int i,j,k,len,pos,prev;
4255: UINT c,c1,c2,c3,up,lo,dmy;
4256: IndArray ivect;
4257: unsigned char *ivc;
4258: unsigned short *ivs;
4259: unsigned int *ivi;
4260: NDV redv;
4261: NMV mr;
4262: NODE rp;
1.76 noro 4263: int maxrs;
1.72 noro 4264:
1.76 noro 4265: maxrs = 0;
1.74 noro 4266: for ( i = 0; i < nred; i++ ) {
1.72 noro 4267: ivect = imat[i];
4268: k = ivect->head; svect[k] %= m;
4269: if ( c = svect[k] ) {
1.76 noro 4270: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4271: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 4272: len = LEN(redv); mr = BDY(redv);
4273: svect[k] = 0; prev = k;
4274: switch ( ivect->width ) {
4275: case 1:
4276: ivc = ivect->index.c;
4277: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4278: pos = prev+ivc[j]; prev = pos;
4279: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4280: }
4281: break;
4282: case 2:
4283: ivs = ivect->index.s;
4284: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4285: pos = prev+ivs[j]; prev = pos;
4286: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4287: }
4288: break;
4289: case 4:
4290: ivi = ivect->index.i;
4291: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4292: pos = prev+ivi[j]; prev = pos;
4293: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4294: }
4295: break;
4296: }
4297: }
4298: }
1.76 noro 4299: return maxrs;
1.72 noro 4300: }
4301:
1.65 noro 4302: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
4303: {
4304: int j,k,len;
4305: UINT *p;
4306: UINT c;
4307: NDV r;
4308: NMV mr0,mr;
4309:
4310: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4311: if ( !len ) return 0;
4312: else {
1.74 noro 4313: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4314: #if 0
1.74 noro 4315: ndv_alloc += nmv_adv*len;
1.103 noro 4316: #endif
1.65 noro 4317: mr = mr0;
4318: p = s0vect;
4319: for ( j = k = 0; j < col; j++, p += nd_wpd )
4320: if ( !rhead[j] ) {
4321: if ( c = vect[k++] ) {
4322: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
4323: }
4324: }
4325: MKNDV(nd_nvar,mr0,len,r);
4326: return r;
4327: }
4328: }
4329:
1.113 noro 4330: NDV vect_to_ndv_q(Q *vect,int spcol,int col,int *rhead,UINT *s0vect)
1.107 noro 4331: {
4332: int j,k,len;
4333: UINT *p;
1.113 noro 4334: Q c;
1.107 noro 4335: NDV r;
4336: NMV mr0,mr;
4337:
4338: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4339: if ( !len ) return 0;
4340: else {
1.109 noro 4341: mr0 = (NMV)GC_malloc(nmv_adv*len);
1.107 noro 4342: #if 0
4343: ndv_alloc += nmv_adv*len;
4344: #endif
4345: mr = mr0;
4346: p = s0vect;
4347: for ( j = k = 0; j < col; j++, p += nd_wpd )
4348: if ( !rhead[j] ) {
4349: if ( c = vect[k++] ) {
1.113 noro 4350: if ( DN(c) )
4351: error("afo");
4352: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
1.107 noro 4353: }
4354: }
4355: MKNDV(nd_nvar,mr0,len,r);
4356: return r;
4357: }
4358: }
4359:
1.74 noro 4360: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket)
1.65 noro 4361: {
4362: ND_pairs t;
4363: NODE sp0,sp;
4364: int stat;
4365: ND spol;
4366:
4367: for ( t = l; t; t = NEXT(t) ) {
4368: stat = nd_sp(m,0,t,&spol);
4369: if ( !stat ) return 0;
4370: if ( spol ) {
4371: add_pbucket_symbolic(bucket,spol);
4372: }
4373: }
1.68 noro 4374: return 1;
1.65 noro 4375: }
4376:
4377: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
4378: {
4379: NODE rp0,rp;
4380: NM mul,head,s0,s;
1.76 noro 4381: int index,col,i,sugar;
1.65 noro 4382: RHist h;
4383: UINT *s0v,*p;
4384: NM_ind_pair pair;
4385: ND red;
4386:
4387: s0 = 0; rp0 = 0; col = 0;
4388: while ( 1 ) {
4389: head = remove_head_pbucket_symbolic(bucket);
4390: if ( !head ) break;
4391: if ( !s0 ) s0 = head;
4392: else NEXT(s) = head;
4393: s = head;
4394: index = ndl_find_reducer(DL(head));
4395: if ( index >= 0 ) {
4396: h = nd_psh[index];
4397: NEWNM(mul);
4398: ndl_sub(DL(head),DL(h),DL(mul));
4399: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.76 noro 4400: sugar = TD(DL(mul))+SG(nd_ps[index]);
4401: MKNM_ind_pair(pair,mul,index,sugar);
1.65 noro 4402: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
4403: add_pbucket_symbolic(bucket,nd_remove_head(red));
4404: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4405: }
4406: col++;
4407: }
1.72 noro 4408: if ( rp0 ) NEXT(rp) = 0;
4409: NEXT(s) = 0;
1.65 noro 4410: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4411: for ( i = 0, p = s0v, s = s0; i < col;
4412: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4413: *s0vect = s0v;
4414: *r = rp0;
4415: return col;
4416: }
4417:
1.69 noro 4418: NODE nd_f4(int m)
4419: {
4420: int i,nh,stat,index;
4421: NODE r,g;
4422: ND_pairs d,l,t;
4423: ND spol,red;
4424: NDV nf,redv;
4425: NM s0,s;
1.84 noro 4426: NODE rp0,srp0,nflist;
1.69 noro 4427: int nsp,nred,col,rank,len,k,j,a;
4428: UINT c;
1.74 noro 4429: UINT **spmat;
1.69 noro 4430: UINT *s0vect,*svect,*p,*v;
4431: int *colstat;
4432: IndArray *imat;
4433: int *rhead;
4434: int spcol,sprow;
4435: int sugar;
4436: PGeoBucket bucket;
4437: struct oEGT eg0,eg1,eg_f4;
4438:
1.103 noro 4439: #if 0
1.74 noro 4440: ndv_alloc = 0;
1.103 noro 4441: #endif
1.69 noro 4442: g = 0; d = 0;
4443: for ( i = 0; i < nd_psn; i++ ) {
4444: d = update_pairs(d,g,i);
4445: g = update_base(g,i);
4446: }
4447: while ( d ) {
4448: get_eg(&eg0);
4449: l = nd_minsugarp(d,&d);
4450: sugar = SG(l);
4451: bucket = create_pbucket();
1.74 noro 4452: stat = nd_sp_f4(m,l,bucket);
1.69 noro 4453: if ( !stat ) {
4454: for ( t = l; NEXT(t); t = NEXT(t) );
4455: NEXT(t) = d; d = l;
1.103 noro 4456: d = nd_reconstruct(0,d);
1.69 noro 4457: continue;
4458: }
1.86 noro 4459: if ( bucket->m < 0 ) continue;
1.69 noro 4460: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
4461: if ( !col ) {
4462: for ( t = l; NEXT(t); t = NEXT(t) );
4463: NEXT(t) = d; d = l;
1.103 noro 4464: d = nd_reconstruct(0,d);
1.69 noro 4465: continue;
4466: }
4467: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4468: if ( DP_Print )
4469: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4470: sugar,eg_f4.exectime+eg_f4.gctime);
4471: if ( 1 )
4472: nflist = nd_f4_red(m,l,s0vect,col,rp0);
4473: else
4474: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0);
1.69 noro 4475: /* adding new bases */
1.74 noro 4476: for ( r = nflist; r; r = NEXT(r) ) {
4477: nf = (NDV)BDY(r);
1.69 noro 4478: ndv_removecont(m,nf);
1.77 noro 4479: nh = ndv_newps(m,nf,0);
1.69 noro 4480: d = update_pairs(d,g,nh);
4481: g = update_base(g,nh);
4482: }
4483: }
4484: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.103 noro 4485: #if 0
1.74 noro 4486: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 4487: #endif
1.69 noro 4488: return g;
4489: }
1.74 noro 4490:
4491: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
1.63 noro 4492: {
1.67 noro 4493: IndArray *imat;
1.106 noro 4494: int nsp,nred,i;
1.65 noro 4495: int *rhead;
1.106 noro 4496: NODE r0,rp;
1.74 noro 4497: ND_pairs sp;
4498: NM_ind_pair *rvect;
1.63 noro 4499:
1.74 noro 4500: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
1.106 noro 4501: nred = length(rp0);
1.74 noro 4502: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4503: rhead = (int *)ALLOCA(col*sizeof(int));
4504: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4505:
1.74 noro 4506: /* construction of index arrays */
4507: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4508: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4509: rvect[i] = (NM_ind_pair)BDY(rp);
4510: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4511: rhead[imat[i]->head] = 1;
4512: }
1.107 noro 4513: if ( m )
4514: r0 = nd_f4_red_main(m,sp0,nsp,s0vect,col,rvect,rhead,imat,nred);
4515: else
1.113 noro 4516: r0 = nd_f4_red_q_main(sp0,nsp,s0vect,col,rvect,rhead,imat,nred);
1.106 noro 4517: return r0;
4518: }
1.74 noro 4519:
1.106 noro 4520: NODE nd_f4_red_main(int m,ND_pairs sp0,int nsp,UINT *s0vect,int col,
4521: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4522: {
4523: int spcol,sprow,a;
4524: int i,j,k,l,rank;
4525: NODE r0,r;
4526: ND_pairs sp;
4527: ND spol;
4528: int **spmat;
4529: UINT *svect,*v;
4530: int *colstat;
4531: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4532: int maxrs;
4533: int *spsugar;
4534:
4535: spcol = col-nred;
4536: get_eg(&eg0);
1.74 noro 4537: /* elimination (1st step) */
4538: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4539: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4540: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4541: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4542: nd_sp(m,0,sp,&spol);
1.75 noro 4543: if ( !spol ) continue;
1.74 noro 4544: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4545: if ( m == -1 )
4546: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4547: else
4548: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4549: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4550: if ( i < col ) {
4551: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4552: for ( j = k = 0; j < col; j++ )
4553: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4554: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4555: sprow++;
4556: }
1.76 noro 4557: nd_free(spol);
1.74 noro 4558: }
1.80 noro 4559: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4560: if ( DP_Print ) {
4561: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4562: fflush(asir_out);
4563: }
1.74 noro 4564: /* free index arrays */
4565: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4566:
4567: /* elimination (2nd step) */
4568: colstat = (int *)ALLOCA(spcol*sizeof(int));
4569: if ( m == -1 )
1.76 noro 4570: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4571: else
1.76 noro 4572: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4573: r0 = 0;
4574: for ( i = 0; i < rank; i++ ) {
4575: NEXTNODE(r0,r); BDY(r) =
4576: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4577: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4578: GC_free(spmat[i]);
4579: }
1.109 noro 4580: if ( r0 ) NEXT(r) = 0;
1.74 noro 4581: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4582: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4583: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4584: if ( DP_Print ) {
1.80 noro 4585: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4586: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4587: nsp,nred,sprow,spcol,rank);
4588: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4589: }
1.74 noro 4590: return r0;
4591: }
4592:
1.113 noro 4593: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,UINT *s0vect,int col,
1.107 noro 4594: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4595: {
4596: int spcol,sprow,a;
4597: int i,j,k,l,rank;
4598: NODE r0,r;
4599: ND_pairs sp;
4600: ND spol;
1.113 noro 4601: Q **spmat;
4602: Q *svect,*v;
1.107 noro 4603: int *colstat;
4604: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4605: int maxrs;
4606: int *spsugar;
4607:
4608: spcol = col-nred;
4609: get_eg(&eg0);
4610: /* elimination (1st step) */
1.113 noro 4611: spmat = (Q **)ALLOCA(nsp*sizeof(Q *));
4612: svect = (Q *)ALLOCA(col*sizeof(Q));
4613: spsugar = (int *)ALLOCA(nsp*sizeof(Q));
1.107 noro 4614: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4615: nd_sp(0,0,sp,&spol);
4616: if ( !spol ) continue;
1.113 noro 4617: nd_to_vect_q(s0vect,col,spol,svect);
4618: maxrs = ndv_reduce_vect_q(svect,col,imat,rvect,nred);
1.107 noro 4619: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4620: if ( i < col ) {
1.113 noro 4621: spmat[sprow] = v = (Q *)MALLOC(spcol*sizeof(Q));
1.107 noro 4622: for ( j = k = 0; j < col; j++ )
4623: if ( !rhead[j] ) v[k++] = svect[j];
4624: spsugar[sprow] = MAX(maxrs,SG(spol));
4625: sprow++;
4626: }
1.109 noro 4627: /* nd_free(spol); */
1.107 noro 4628: }
4629: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4630: if ( DP_Print ) {
4631: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4632: fflush(asir_out);
4633: }
4634: /* free index arrays */
1.109 noro 4635: /* for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c); */
1.107 noro 4636:
4637: /* elimination (2nd step) */
4638: colstat = (int *)ALLOCA(spcol*sizeof(int));
1.113 noro 4639: rank = nd_gauss_elim_q(spmat,spsugar,sprow,spcol,colstat);
1.107 noro 4640: r0 = 0;
4641: for ( i = 0; i < rank; i++ ) {
4642: NEXTNODE(r0,r); BDY(r) =
1.113 noro 4643: (pointer)vect_to_ndv_q(spmat[i],spcol,col,rhead,s0vect);
1.107 noro 4644: SG((NDV)BDY(r)) = spsugar[i];
1.109 noro 4645: /* GC_free(spmat[i]); */
1.107 noro 4646: }
1.109 noro 4647: if ( r0 ) NEXT(r) = 0;
4648:
4649: /* for ( ; i < sprow; i++ ) GC_free(spmat[i]); */
1.107 noro 4650: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4651: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
4652: if ( DP_Print ) {
4653: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
4654: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4655: nsp,nred,sprow,spcol,rank);
4656: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
4657: }
4658: return r0;
4659: }
4660:
1.74 noro 4661: FILE *nd_write,*nd_read;
4662:
4663: void nd_send_int(int a) {
4664: write_int(nd_write,&a);
4665: }
4666:
4667: void nd_send_intarray(int *p,int len) {
4668: write_intarray(nd_write,p,len);
4669: }
4670:
4671: int nd_recv_int() {
4672: int a;
4673:
4674: read_int(nd_read,&a);
4675: return a;
4676: }
4677:
4678: void nd_recv_intarray(int *p,int len) {
4679: read_intarray(nd_read,p,len);
4680: }
4681:
4682: void nd_send_ndv(NDV p) {
4683: int len,i;
4684: NMV m;
4685:
4686: if ( !p ) nd_send_int(0);
4687: else {
4688: len = LEN(p);
4689: nd_send_int(len);
4690: m = BDY(p);
4691: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4692: nd_send_int(CM(m));
4693: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4694: }
1.74 noro 4695: }
4696: }
4697:
4698: void nd_send_nd(ND p) {
4699: int len,i;
4700: NM m;
4701:
4702: if ( !p ) nd_send_int(0);
4703: else {
4704: len = LEN(p);
4705: nd_send_int(len);
4706: m = BDY(p);
4707: for ( i = 0; i < len; i++, m = NEXT(m) ) {
4708: nd_send_int(CM(m));
4709: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4710: }
1.74 noro 4711: }
4712: }
1.65 noro 4713:
1.74 noro 4714: NDV nd_recv_ndv()
4715: {
4716: int len,i;
4717: NMV m,m0;
4718: NDV r;
1.65 noro 4719:
1.74 noro 4720: len = nd_recv_int();
4721: if ( !len ) return 0;
4722: else {
4723: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4724: #if 0
1.74 noro 4725: ndv_alloc += len*nmv_adv;
1.103 noro 4726: #endif
1.74 noro 4727: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4728: CM(m) = nd_recv_int();
4729: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4730: }
1.74 noro 4731: MKNDV(nd_nvar,m0,len,r);
4732: return r;
4733: }
4734: }
1.65 noro 4735:
1.74 noro 4736: int ox_exec_f4_red(Q proc)
4737: {
4738: Obj obj;
4739: STRING fname;
4740: NODE arg;
4741: int s;
4742: extern int ox_need_conv,ox_file_io;
4743:
4744: MKSTR(fname,"nd_exec_f4_red");
4745: arg = mknode(2,proc,fname);
4746: Pox_cmo_rpc(arg,&obj);
4747: s = get_ox_server_id(QTOS(proc));
4748: nd_write = iofp[s].out;
4749: nd_read = iofp[s].in;
4750: ox_need_conv = ox_file_io = 0;
4751: return s;
4752: }
4753:
4754: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
4755: {
4756: int nsp,nred;
4757: int i,rank,s;
4758: NODE rp,r0,r;
4759: ND_pairs sp;
4760: NM_ind_pair pair;
4761: NMV nmv;
4762: NM nm;
4763: NDV nf;
4764: Obj proc,dmy;
4765:
4766: ox_launch_main(0,0,&proc);
4767: s = ox_exec_f4_red((Q)proc);
4768:
4769: nd_send_int(m);
4770: nd_send_int(nd_nvar);
4771: nd_send_int(nd_bpe);
4772: nd_send_int(nd_wpd);
4773: nd_send_int(nmv_adv);
4774:
1.89 noro 4775: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
1.74 noro 4776:
4777: nd_send_int(nd_psn);
4778: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
4779:
4780: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4781: nd_send_int(nsp);
4782: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
4783: nd_send_int(sp->i1); nd_send_int(sp->i2);
4784: }
4785:
4786: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
4787:
4788: nred = length(rp0); nd_send_int(nred);
4789: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
4790: pair = (NM_ind_pair)BDY(rp);
4791: nd_send_int(pair->index);
4792: nd_send_intarray(pair->mul->dl,nd_wpd);
4793: }
4794: fflush(nd_write);
4795: rank = nd_recv_int();
4796: fprintf(asir_out,"rank=%d\n",rank);
4797: r0 = 0;
4798: for ( i = 0; i < rank; i++ ) {
4799: nf = nd_recv_ndv();
4800: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
4801: }
4802: Pox_shutdown(mknode(1,proc),&dmy);
4803: return r0;
4804: }
4805:
4806: /* server side */
4807:
4808: void nd_exec_f4_red_dist()
4809: {
4810: int m,i,nsp,col,s0size,nred,spcol,j,k;
4811: NM_ind_pair *rp0;
4812: NDV nf;
4813: UINT *s0vect;
4814: IndArray *imat;
4815: int *rhead;
4816: int **spmat;
4817: UINT *svect,*v;
4818: ND_pairs *sp0;
4819: int *colstat;
4820: int a,sprow,rank;
1.89 noro 4821: struct order_spec *ord;
1.74 noro 4822: Obj ordspec;
4823: ND spol;
1.76 noro 4824: int maxrs;
4825: int *spsugar;
1.74 noro 4826:
4827: nd_read = iofp[0].in;
4828: nd_write = iofp[0].out;
4829: m = nd_recv_int();
4830: nd_nvar = nd_recv_int();
4831: nd_bpe = nd_recv_int();
4832: nd_wpd = nd_recv_int();
4833: nmv_adv = nd_recv_int();
4834:
4835: loadobj(nd_read,&ordspec);
1.89 noro 4836: create_order_spec(0,ordspec,&ord);
4837: nd_init_ord(ord);
1.74 noro 4838: nd_setup_parameters(nd_nvar,0);
4839:
4840: nd_psn = nd_recv_int();
4841: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
4842: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
4843: for ( i = 0; i < nd_psn; i++ ) {
4844: nd_ps[i] = nd_recv_ndv();
4845: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
4846: }
4847:
4848: nsp = nd_recv_int();
4849: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
4850: for ( i = 0; i < nsp; i++ ) {
4851: NEWND_pairs(sp0[i]);
4852: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
4853: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
4854: }
4855:
4856: col = nd_recv_int();
4857: s0size = col*nd_wpd;
4858: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
4859: nd_recv_intarray(s0vect,s0size);
4860:
4861: nred = nd_recv_int();
4862: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
4863: for ( i = 0; i < nred; i++ ) {
4864: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
4865: rp0[i]->index = nd_recv_int();
4866: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
4867: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
4868: }
4869:
4870: spcol = col-nred;
4871: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
4872: rhead = (int *)MALLOC(col*sizeof(int));
4873: for ( i = 0; i < col; i++ ) rhead[i] = 0;
4874:
4875: /* construction of index arrays */
4876: for ( i = 0; i < nred; i++ ) {
4877: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
4878: rhead[imat[i]->head] = 1;
4879: }
4880:
4881: /* elimination (1st step) */
4882: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
4883: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 4884: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4885: for ( a = sprow = 0; a < nsp; a++ ) {
4886: nd_sp(m,0,sp0[a],&spol);
1.75 noro 4887: if ( !spol ) continue;
1.74 noro 4888: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4889: if ( m == -1 )
4890: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
4891: else
4892: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 4893: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4894: if ( i < col ) {
4895: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
4896: for ( j = k = 0; j < col; j++ )
4897: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4898: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4899: sprow++;
4900: }
1.76 noro 4901: nd_free(spol);
1.74 noro 4902: }
4903: /* elimination (2nd step) */
4904: colstat = (int *)ALLOCA(spcol*sizeof(int));
4905: if ( m == -1 )
1.76 noro 4906: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4907: else
1.76 noro 4908: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4909: nd_send_int(rank);
4910: for ( i = 0; i < rank; i++ ) {
4911: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
4912: nd_send_ndv(nf);
1.63 noro 4913: }
1.74 noro 4914: fflush(nd_write);
1.107 noro 4915: }
4916:
1.113 noro 4917: int nd_gauss_elim_q(Q **mat0,int *sugar,int row,int col,int *colstat)
1.107 noro 4918: {
1.109 noro 4919: int mod,i,j,t,c,rank,rank0,inv;
4920: int *ci,*ri;
1.113 noro 4921: Q dn;
1.109 noro 4922: MAT m,nm;
1.108 noro 4923: int **wmat;
4924:
4925: /* XXX */
4926: mod = 99999989;
4927: wmat = (int **)ALLOCA(row*sizeof(int *));
4928: for ( i = 0; i < row; i++ ) {
4929: wmat[i] = (int *)ALLOCA(col*sizeof(int));
4930: for ( j = 0; j < col; j++ ) {
1.113 noro 4931: if ( mat0[i][j] ) {
4932: t = rem(NM(mat0[i][j]),mod);
4933: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
4934: wmat[i][j] = t;
4935: } else
1.108 noro 4936: wmat[i][j] = 0;
4937: }
4938: }
1.109 noro 4939: rank0 = nd_gauss_elim_mod(wmat,sugar,row,col,mod,colstat);
4940: NEWMAT(m); m->row = row; m->col = col; m->body = (pointer **)mat0;
1.113 noro 4941: rank = generic_gauss_elim(m,&nm,&dn,&ri,&ci);
1.109 noro 4942: if ( rank != rank0 )
4943: error("afo");
4944: for ( i = 0; i < row; i++ )
4945: for ( j = 0; j < col; j++ )
4946: mat0[i][j] = 0;
4947: c = col-rank;
4948: for ( i = 0; i < rank; i++ ) {
1.113 noro 4949: mat0[i][ri[i]] = dn;
1.109 noro 4950: for ( j = 0; j < c; j++ )
1.113 noro 4951: mat0[i][ci[j]] = (Q)BDY(nm)[i][j];
1.109 noro 4952: }
1.113 noro 4953: inv = invm(rem(NM(dn),mod),mod);
4954: if ( SGN(dn) < 0 ) inv = mod-inv;
1.109 noro 4955: for ( i = 0; i < row; i++ )
4956: for ( j = 0; j < col; j++ ) {
1.113 noro 4957: if ( mat0[i][j] ) {
4958: t = rem(NM(mat0[i][j]),mod);
4959: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
4960: } else
1.109 noro 4961: t = 0;
4962: c = dmar(t,inv,0,mod);
4963: if ( wmat[i][j] != c )
4964: error("afo");
4965: }
4966: return rank;
1.76 noro 4967: }
4968:
4969: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4970: {
4971: int i,j,k,l,inv,a,rank,s;
4972: unsigned int *t,*pivot,*pk;
4973: unsigned int **mat;
4974:
4975: mat = (unsigned int **)mat0;
4976: for ( rank = 0, j = 0; j < col; j++ ) {
4977: for ( i = rank; i < row; i++ )
4978: mat[i][j] %= md;
4979: for ( i = rank; i < row; i++ )
4980: if ( mat[i][j] )
4981: break;
4982: if ( i == row ) {
4983: colstat[j] = 0;
4984: continue;
4985: } else
4986: colstat[j] = 1;
4987: if ( i != rank ) {
4988: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4989: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4990: }
4991: pivot = mat[rank];
4992: s = sugar[rank];
4993: inv = invm(pivot[j],md);
4994: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4995: if ( *pk ) {
4996: if ( *pk >= (unsigned int)md )
4997: *pk %= md;
4998: DMAR(*pk,inv,0,md,*pk)
4999: }
5000: for ( i = rank+1; i < row; i++ ) {
5001: t = mat[i];
5002: if ( a = t[j] ) {
5003: sugar[i] = MAX(sugar[i],s);
5004: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5005: }
5006: }
5007: rank++;
5008: }
5009: for ( j = col-1, l = rank-1; j >= 0; j-- )
5010: if ( colstat[j] ) {
5011: pivot = mat[l];
5012: s = sugar[l];
5013: for ( i = 0; i < l; i++ ) {
5014: t = mat[i];
5015: t[j] %= md;
5016: if ( a = t[j] ) {
5017: sugar[i] = MAX(sugar[i],s);
5018: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5019: }
5020: }
5021: l--;
5022: }
5023: for ( j = 0, l = 0; l < rank; j++ )
5024: if ( colstat[j] ) {
5025: t = mat[l];
5026: for ( k = j; k < col; k++ )
5027: if ( t[k] >= (unsigned int)md )
5028: t[k] %= md;
5029: l++;
5030: }
5031: return rank;
5032: }
5033:
5034: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
5035: {
5036: int i,j,k,l,inv,a,rank,s;
5037: unsigned int *t,*pivot,*pk;
5038: unsigned int **mat;
5039:
5040: mat = (unsigned int **)mat0;
5041: for ( rank = 0, j = 0; j < col; j++ ) {
5042: for ( i = rank; i < row; i++ )
5043: if ( mat[i][j] )
5044: break;
5045: if ( i == row ) {
5046: colstat[j] = 0;
5047: continue;
5048: } else
5049: colstat[j] = 1;
5050: if ( i != rank ) {
5051: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5052: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
5053: }
5054: pivot = mat[rank];
5055: s = sugar[rank];
5056: inv = _invsf(pivot[j]);
5057: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5058: if ( *pk )
5059: *pk = _mulsf(*pk,inv);
5060: for ( i = rank+1; i < row; i++ ) {
5061: t = mat[i];
5062: if ( a = t[j] ) {
5063: sugar[i] = MAX(sugar[i],s);
5064: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5065: }
5066: }
5067: rank++;
5068: }
5069: for ( j = col-1, l = rank-1; j >= 0; j-- )
5070: if ( colstat[j] ) {
5071: pivot = mat[l];
5072: s = sugar[l];
5073: for ( i = 0; i < l; i++ ) {
5074: t = mat[i];
5075: if ( a = t[j] ) {
5076: sugar[i] = MAX(sugar[i],s);
5077: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5078: }
5079: }
5080: l--;
5081: }
5082: return rank;
1.77 noro 5083: }
5084:
5085: int ndv_ishomo(NDV p)
5086: {
5087: NMV m;
5088: int len,h;
5089:
5090: if ( !p ) return 1;
5091: len = LEN(p);
5092: m = BDY(p);
5093: h = TD(DL(m));
5094: NMV_ADV(m);
5095: for ( len--; len; len--, NMV_ADV(m) )
5096: if ( TD(DL(m)) != h ) return 0;
5097: return 1;
5098: }
5099:
5100: void ndv_save(NDV p,int index)
5101: {
5102: FILE *s;
5103: char name[BUFSIZ];
5104: short id;
5105: int nv,sugar,len,n,i,td,e,j;
5106: NMV m;
5107: unsigned int *dl;
5108:
5109: sprintf(name,"%s/%d",Demand,index);
5110: s = fopen(name,"w");
5111: savevl(s,0);
5112: if ( !p ) {
5113: saveobj(s,0);
5114: return;
5115: }
5116: id = O_DP;
5117: nv = NV(p);
5118: sugar = SG(p);
5119: len = LEN(p);
5120: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
5121: write_int(s,&len);
5122:
5123: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5124: saveobj(s,(Obj)CQ(m));
1.77 noro 5125: dl = DL(m);
5126: td = TD(dl);
5127: write_int(s,&td);
5128: for ( j = 0; j < nv; j++ ) {
5129: e = GET_EXP(dl,j);
5130: write_int(s,&e);
5131: }
5132: }
5133: fclose(s);
5134: }
5135:
5136: NDV ndv_load(int index)
5137: {
5138: FILE *s;
5139: char name[BUFSIZ];
5140: short id;
5141: int nv,sugar,len,n,i,td,e,j;
5142: NDV d;
5143: NMV m0,m;
5144: unsigned int *dl;
5145: Obj obj;
5146:
5147: sprintf(name,"%s/%d",Demand,index);
5148: s = fopen(name,"r");
5149: if ( !s ) return 0;
5150:
5151: skipvl(s);
5152: read_short(s,&id);
5153: if ( !id ) return 0;
5154: read_int(s,&nv);
5155: read_int(s,&sugar);
5156: read_int(s,&len);
5157:
5158: m0 = m = MALLOC(len*nmv_adv);
5159: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5160: loadobj(s,&obj); CQ(m) = (Q)obj;
1.77 noro 5161: dl = DL(m);
5162: ndl_zero(dl);
5163: read_int(s,&td); TD(dl) = td;
5164: for ( j = 0; j < nv; j++ ) {
5165: read_int(s,&e);
5166: PUT_EXP(dl,j,e);
5167: }
5168: if ( nd_blockmask ) ndl_weight_mask(dl);
5169: }
5170: fclose(s);
5171: MKNDV(nv,m0,len,d);
5172: SG(d) = sugar;
5173: return d;
1.99 noro 5174: }
5175:
1.102 noro 5176: void nd_det(int mod,MAT f,P *rp)
1.99 noro 5177: {
5178: VL fv,tv;
1.102 noro 5179: int n,i,j,max,e,nvar,sgn,k0,l0,len0,len,k,l,a;
1.99 noro 5180: pointer **m;
1.113 noro 5181: Q mone;
1.125 noro 5182: P **w;
1.126 noro 5183: P mp,r;
1.102 noro 5184: NDV **dm;
5185: NDV *t,*mi,*mj;
5186: NDV d,s,mij,mjj;
5187: ND u;
5188: NMV nmv;
1.114 noro 5189: UINT *bound;
1.99 noro 5190: PGeoBucket bucket;
5191: struct order_spec *ord;
1.126 noro 5192: Q dq,dt,ds;
5193: N gn,qn,dn0,nm,dn;
1.99 noro 5194:
5195: create_order_spec(0,0,&ord);
5196: nd_init_ord(ord);
5197: get_vars((Obj)f,&fv);
5198: if ( f->row != f->col )
5199: error("nd_det : non-square matrix");
5200: n = f->row;
1.125 noro 5201: m = f->body;
1.99 noro 5202: for ( nvar = 0, tv = fv; tv; tv = NEXT(tv), nvar++ );
1.125 noro 5203:
5204: if ( !nvar ) {
5205: if ( !mod )
5206: detp(CO,(P **)m,n,rp);
5207: else {
5208: w = (P **)almat_pointer(n,n);
5209: for ( i = 0; i < n; i++ )
5210: for ( j = 0; j < n; j++ )
5211: ptomp(mod,(P)m[i][j],&w[i][j]);
5212: detmp(CO,mod,w,n,&mp);
5213: mptop(mp,rp);
5214: }
5215: return;
5216: }
1.126 noro 5217:
5218: if ( !mod ) {
5219: w = (P **)almat_pointer(n,n);
5220: dq = ONE;
5221: for ( i = 0; i < n; i++ ) {
5222: dn0 = ONEN;
5223: for ( j = 0; j < n; j++ ) {
5224: if ( !m[i][j] ) continue;
5225: lgp(m[i][j],&nm,&dn);
5226: gcdn(dn0,dn,&gn); divsn(dn0,gn,&qn); muln(qn,dn,&dn0);
5227: }
5228: if ( !UNIN(dn0) ) {
5229: NTOQ(dn0,1,ds);
5230: for ( j = 0; j < n; j++ )
5231: mulp(CO,(P)m[i][j],(P)ds,&w[i][j]);
5232: mulq(dq,ds,&dt); dq = dt;
5233: } else
5234: for ( j = 0; j < n; j++ )
5235: w[i][j] = (P)m[i][j];
5236: }
5237: m = (pointer **)w;
5238: }
1.125 noro 5239:
1.99 noro 5240: for ( i = 0, max = 0; i < n; i++ )
5241: for ( j = 0; j < n; j++ )
5242: for ( tv = fv; tv; tv = NEXT(tv) ) {
5243: e = getdeg(tv->v,(P)m[i][j]);
5244: max = MAX(e,max);
5245: }
1.114 noro 5246: nd_setup_parameters(nvar,max);
1.102 noro 5247: dm = (NDV **)almat_pointer(n,n);
1.99 noro 5248: for ( i = 0, max = 0; i < n; i++ )
1.102 noro 5249: for ( j = 0; j < n; j++ ) {
5250: dm[i][j] = ptondv(CO,fv,m[i][j]);
5251: if ( mod ) ndv_mod(mod,dm[i][j]);
5252: if ( dm[i][j] && !LEN(dm[i][j]) ) dm[i][j] = 0;
5253: }
5254: d = ptondv(CO,fv,(P)ONE);
5255: if ( mod ) ndv_mod(mod,d);
1.113 noro 5256: chsgnq(ONE,&mone);
1.99 noro 5257: for ( j = 0, sgn = 1; j < n; j++ ) {
1.125 noro 5258: if ( DP_Print ) fprintf(stderr,".",j);
1.99 noro 5259: for ( i = j; i < n && !dm[i][j]; i++ );
5260: if ( i == n ) {
5261: *rp = 0;
5262: return;
5263: }
1.102 noro 5264: k0 = i; l0 = j; len0 = LEN(dm[k0][l0]);
1.99 noro 5265: for ( k = j; k < n; k++ )
5266: for ( l = j; l < n; l++ )
1.102 noro 5267: if ( dm[k][l] && LEN(dm[k][l]) < len0 ) {
5268: k0 = k; l0 = l; len0 = LEN(dm[k][l]);
1.99 noro 5269: }
5270: if ( k0 != j ) {
5271: t = dm[j]; dm[j] = dm[k0]; dm[k0] = t;
5272: sgn = -sgn;
5273: }
5274: if ( l0 != j ) {
5275: for ( k = j; k < n; k++ ) {
5276: s = dm[k][j]; dm[k][j] = dm[k][l0]; dm[k][l0] = s;
5277: }
5278: sgn = -sgn;
5279: }
1.114 noro 5280: bound = nd_det_compute_bound(dm,n,j);
5281: if ( ndl_check_bound(bound,bound) )
5282: nd_det_reconstruct(dm,n,j,d);
5283:
1.99 noro 5284: for ( i = j+1, mj = dm[j], mjj = mj[j]; i < n; i++ ) {
1.114 noro 5285: /* if ( DP_Print ) fprintf(stderr," i=%d\n ",i); */
1.99 noro 5286: mi = dm[i]; mij = mi[j];
1.102 noro 5287: if ( mod )
5288: ndv_mul_c(mod,mij,mod-1);
5289: else
1.113 noro 5290: ndv_mul_c_q(mij,mone);
1.99 noro 5291: for ( k = j+1; k < n; k++ ) {
1.114 noro 5292: /* if ( DP_Print ) fprintf(stderr,"k=%d ",k); */
1.99 noro 5293: bucket = create_pbucket();
1.104 noro 5294: if ( mi[k] ) {
1.102 noro 5295: nmv = BDY(mjj); len = LEN(mjj);
5296: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5297: u = ndv_mul_nmv_trunc(mod,nmv,mi[k],DL(BDY(d)));
5298: add_pbucket(mod,bucket,u);
1.99 noro 5299: }
1.104 noro 5300: }
1.99 noro 5301: if ( mj[k] && mij ) {
1.102 noro 5302: nmv = BDY(mij); len = LEN(mij);
5303: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5304: u = ndv_mul_nmv_trunc(mod,nmv,mj[k],DL(BDY(d)));
5305: add_pbucket(mod,bucket,u);
1.99 noro 5306: }
5307: }
1.104 noro 5308: u = nd_quo(mod,bucket,d);
1.102 noro 5309: mi[k] = ndtondv(mod,u);
1.99 noro 5310: }
1.114 noro 5311: /* if ( DP_Print ) fprintf(stderr,"\n",k); */
1.99 noro 5312: }
5313: d = mjj;
5314: }
1.125 noro 5315: if ( DP_Print ) fprintf(stderr,"\n",k);
1.100 noro 5316: if ( sgn < 0 )
1.102 noro 5317: if ( mod )
5318: ndv_mul_c(mod,d,mod-1);
5319: else
1.113 noro 5320: ndv_mul_c_q(d,mone);
1.126 noro 5321: r = ndvtop(mod,CO,fv,d);
5322: if ( !mod && !UNIQ(dq) )
5323: divsp(CO,r,(P)dq,rp);
5324: else
5325: *rp = r;
1.99 noro 5326: }
5327:
1.102 noro 5328: ND ndv_mul_nmv_trunc(int mod,NMV m0,NDV p,UINT *d)
1.99 noro 5329: {
5330: NM mr,mr0;
1.102 noro 5331: NM tnm;
5332: NMV m;
1.99 noro 5333: UINT *d0,*dt,*dm;
5334: int c,n,td,i,c1,c2,len;
1.113 noro 5335: Q q;
1.99 noro 5336: ND r;
5337:
5338: if ( !p ) return 0;
5339: else {
1.102 noro 5340: n = NV(p); m = BDY(p); len = LEN(p);
1.99 noro 5341: d0 = DL(m0);
1.102 noro 5342: td = TD(d);
1.99 noro 5343: mr0 = 0;
5344: NEWNM(tnm);
1.102 noro 5345: if ( mod ) {
5346: c = CM(m0);
5347: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5348: ndl_add(DL(m),d0,DL(tnm));
5349: if ( ndl_reducible(DL(tnm),d) ) {
5350: NEXTNM(mr0,mr);
5351: c1 = CM(m); DMAR(c1,c,0,mod,c2); CM(mr) = c2;
1.104 noro 5352: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5353: }
5354: }
5355: } else {
1.113 noro 5356: q = CQ(m0);
1.102 noro 5357: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5358: ndl_add(DL(m),d0,DL(tnm));
5359: if ( ndl_reducible(DL(tnm),d) ) {
5360: NEXTNM(mr0,mr);
1.113 noro 5361: mulq(CQ(m),q,&CQ(mr));
1.104 noro 5362: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5363: }
1.99 noro 5364: }
5365: }
5366: if ( !mr0 )
5367: return 0;
5368: else {
5369: NEXT(mr) = 0;
1.105 noro 5370: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
1.99 noro 5371: MKND(NV(p),mr0,len,r);
5372: SG(r) = SG(p) + TD(d0);
5373: return r;
5374: }
5375: }
1.114 noro 5376: }
5377:
5378: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d)
5379: {
5380: int i,obpe,oadv,h,k,l;
5381: static NM prev_nm_free_list;
5382: EPOS oepos;
5383:
5384: obpe = nd_bpe;
5385: oadv = nmv_adv;
5386: oepos = nd_epos;
5387: if ( obpe < 2 ) nd_bpe = 2;
5388: else if ( obpe < 3 ) nd_bpe = 3;
5389: else if ( obpe < 4 ) nd_bpe = 4;
5390: else if ( obpe < 5 ) nd_bpe = 5;
5391: else if ( obpe < 6 ) nd_bpe = 6;
5392: else if ( obpe < 8 ) nd_bpe = 8;
5393: else if ( obpe < 10 ) nd_bpe = 10;
5394: else if ( obpe < 16 ) nd_bpe = 16;
5395: else if ( obpe < 32 ) nd_bpe = 32;
5396: else error("nd_det_reconstruct : exponent too large");
5397:
5398: nd_setup_parameters(nd_nvar,0);
5399: prev_nm_free_list = _nm_free_list;
5400: _nm_free_list = 0;
5401: for ( k = j; k < n; k++ )
5402: for (l = j; l < n; l++ )
5403: ndv_realloc(dm[k][l],obpe,oadv,oepos);
5404: ndv_realloc(d,obpe,oadv,oepos);
5405: prev_nm_free_list = 0;
5406: #if 0
5407: GC_gcollect();
5408: #endif
5409: }
5410:
5411: UINT *nd_det_compute_bound(NDV **dm,int n,int j)
5412: {
5413: UINT *d0,*d1,*d,*t,*r;
5414: int k,l;
5415:
5416: d0 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5417: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5418: for ( k = 0; k < nd_wpd; k++ ) d0[k] = 0;
5419: for ( k = j; k < n; k++ )
5420: for ( l = j; l < n; l++ )
5421: if ( dm[k][l] ) {
5422: d = ndv_compute_bound(dm[k][l]);
5423: ndl_lcm(d,d0,d1);
5424: t = d1; d1 = d0; d0 = t;
5425: }
5426: r = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5427: for ( k = 0; k < nd_wpd; k++ ) r[k] = d0[k];
5428: return r;
1.117 noro 5429: }
5430:
5431: DL nd_separate_d(UINT *d,UINT *trans)
5432: {
1.121 noro 5433: int n,td,i,e,j;
1.117 noro 5434: DL a;
5435:
5436: ndl_zero(trans);
5437: td = 0;
1.121 noro 5438: for ( i = 0; i < nd_ntrans; i++ ) {
5439: e = GET_EXP(d,i);
5440: PUT_EXP(trans,i,e);
5441: td += MUL_WEIGHT(e,i);
5442: }
5443: if ( nd_ntrans+nd_nalg < nd_nvar ) {
5444: /* homogenized */
5445: i = nd_nvar-1;
1.117 noro 5446: e = GET_EXP(d,i);
5447: PUT_EXP(trans,i,e);
5448: td += MUL_WEIGHT(e,i);
5449: }
5450: TD(trans) = td;
5451: if ( nd_blockmask) ndl_weight_mask(trans);
5452: NEWDL(a,nd_nalg);
5453: td = 0;
1.121 noro 5454: for ( i = 0; i < nd_nalg; i++ ) {
5455: j = nd_ntrans+i;
5456: e = GET_EXP(d,j);
5457: a->d[i] = e;
1.117 noro 5458: td += e;
5459: }
5460: a->td = td;
5461: return a;
5462: }
5463:
1.118 noro 5464: int nd_monic(int mod,ND *p)
1.117 noro 5465: {
5466: UINT *trans,*t;
5467: DL alg;
5468: MP mp0,mp;
1.118 noro 5469: NM m,m0,m1,ma0,ma,mb,mr0,mr;
5470: ND r;
1.117 noro 5471: DL dl;
5472: DP nm;
5473: NDV ndv;
1.118 noro 5474: DAlg inv,cd;
1.117 noro 5475: ND s,c;
1.118 noro 5476: Q l,mul;
5477: N ln;
5478: int n,ntrans,i,e,td,is_lc,len;
5479: NumberField nf;
5480: struct oEGT eg0,eg1;
5481:
5482: if ( !(nf = get_numberfield()) )
5483: error("nd_monic : current_numberfield is not set");
1.117 noro 5484:
1.118 noro 5485: /* Q coef -> DAlg coef */
5486: NEWNM(ma0); ma = ma0;
5487: m = BDY(*p);
5488: is_lc = 1;
5489: while ( 1 ) {
5490: NEWMP(mp0); mp = mp0;
5491: mp->c = (P)CQ(m);
5492: mp->dl = nd_separate_d(DL(m),DL(ma));
5493: NEWNM(mb);
5494: for ( m = NEXT(m); m; m = NEXT(m) ) {
5495: alg = nd_separate_d(DL(m),DL(mb));
5496: if ( !ndl_equal(DL(ma),DL(mb)) )
5497: break;
5498: NEXTMP(mp0,mp); mp->c = (P)CQ(m); mp->dl = alg;
5499: }
5500: NEXT(mp) = 0;
5501: MKDP(nd_nalg,mp0,nm);
5502: MKDAlg(nm,ONE,cd);
5503: if ( is_lc == 1 ) {
5504: /* if the lc is a rational number, we have nothing to do */
5505: if ( !mp0->dl->td )
5506: return 1;
5507:
5508: get_eg(&eg0);
5509: invdalg(cd,&inv);
5510: get_eg(&eg1); add_eg(&eg_invdalg,&eg0,&eg1);
5511: /* check the validity of inv */
5512: if ( mod && !rem(NM(inv->dn),mod) )
5513: return 0;
5514: CA(ma) = nf->one;
5515: is_lc = 0;
5516: ln = ONEN;
5517: } else {
5518: muldalg(cd,inv,&CA(ma));
5519: lcmn(ln,NM(CA(ma)->dn),&ln);
5520: }
5521: if ( m ) {
5522: NEXT(ma) = mb; ma = mb;
5523: } else {
5524: NEXT(ma) = 0;
5525: break;
5526: }
5527: }
5528: /* l = lcm(denoms) */
5529: NTOQ(ln,1,l);
5530: for ( mr0 = 0, m = ma0; m; m = NEXT(m) ) {
5531: divq(l,CA(m)->dn,&mul);
5532: for ( mp = BDY(CA(m)->nm); mp; mp = NEXT(mp) ) {
5533: NEXTNM(mr0,mr);
5534: mulq((Q)mp->c,mul,&CQ(mr));
5535: dl = mp->dl;
5536: td = TD(DL(m));
5537: ndl_copy(DL(m),DL(mr));
1.121 noro 5538: for ( i = 0; i < nd_nalg; i++ ) {
5539: e = dl->d[i];
5540: PUT_EXP(DL(mr),i+nd_ntrans,e);
5541: td += MUL_WEIGHT(e,i+nd_ntrans);
1.118 noro 5542: }
5543: TD(DL(mr)) = td;
5544: if ( nd_blockmask) ndl_weight_mask(DL(mr));
5545: }
1.117 noro 5546: }
1.118 noro 5547: NEXT(mr) = 0;
5548: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
5549: MKND(NV(*p),mr0,len,r);
5550: /* XXX */
5551: SG(r) = SG(*p);
5552: nd_free(*p);
5553: *p = r;
5554: return 1;
1.59 noro 5555: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>