Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.127
1.127 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.126 2005/02/09 14:30:47 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:
1.61 noro 1287: ndv_setup(0,0,cand);
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
1842: ndv_setup(m,0,f);
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:
1.61 noro 2182: void ndv_setup(int mod,int trace,NODE f)
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);
2196: qsort(w,nd_psn,sizeof(NDV),
2197: (int (*)(const void *,const void *))ndv_compare);
2198: nd_pslen = 2*nd_psn;
1.11 noro 2199: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2200: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2201: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2202: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2203:
1.11 noro 2204: if ( !nd_red )
1.13 noro 2205: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2206: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2207: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2208: if ( trace ) {
1.69 noro 2209: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2210: ndv_removecont(0,a);
2211: am = nd_ps[i] = ndv_dup(mod,a);
2212: ndv_mod(mod,am);
2213: ndv_removecont(mod,am);
2214: } else {
1.69 noro 2215: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2216: ndv_removecont(mod,a);
1.39 noro 2217: }
1.61 noro 2218: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2219: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2220: nd_psh[i] = r;
1.77 noro 2221: if ( nd_demand ) {
2222: if ( trace ) {
2223: ndv_save(nd_ps_trace[i],i);
2224: nd_ps_trace[i] = 0;
2225: } else {
2226: ndv_save(nd_ps[i],i);
2227: nd_ps[i] = 0;
2228: }
2229: }
1.20 noro 2230: }
2231: }
2232:
1.119 noro 2233: struct order_spec *append_block(struct order_spec *spec,
2234: int nv,int nalg,int ord);
2235:
1.121 noro 2236: extern VECT current_dl_weight_vector_obj;
2237: static VECT prev_weight_vector_obj;
2238:
1.120 noro 2239: void preprocess_algcoef(VL vv,VL av,struct order_spec *ord,LIST f,
2240: struct order_spec **ord1p,LIST *f1p,NODE *alistp)
2241: {
1.121 noro 2242: NODE alist,t,s,r0,r,arg;
1.120 noro 2243: VL tv;
2244: P poly;
2245: DP d;
2246: Alg alpha,dp;
2247: DAlg inv,da,hc;
2248: MP m;
1.121 noro 2249: int i,nvar,nalg,n;
1.120 noro 2250: NumberField nf;
2251: LIST f1,f2;
2252: struct order_spec *current_spec;
1.121 noro 2253: VECT obj,obj0;
2254: Obj tmp;
1.120 noro 2255:
2256: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++);
2257: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++);
2258:
2259: for ( alist = 0, tv = av; tv; tv = NEXT(tv) ) {
2260: NEXTNODE(alist,t); MKV(tv->v,poly);
2261: MKAlg(poly,alpha); BDY(t) = (pointer)alpha;
2262: tv->v = tv->v->priv;
2263: }
2264: NEXT(t) = 0;
2265:
2266: /* simplification, makeing polynomials monic */
2267: setfield_dalg(alist);
2268: obj_algtodalg(f,&f1);
2269: for ( t = BDY(f); t; t = NEXT(t) ) {
2270: initd(ord); ptod(vv,vv,(P)BDY(t),&d);
2271: hc = (DAlg)BDY(d)->c;
2272: if ( NID(hc) == N_DA ) {
2273: invdalg(hc,&inv);
2274: for ( m = BDY(d); m; m = NEXT(m) ) {
2275: muldalg(inv,(DAlg)m->c,&da); m->c = (P)da;
2276: }
2277: }
2278: initd(ord); dtop(vv,vv,d,&poly); BDY(f) = (pointer)poly;
2279: }
2280: obj_dalgtoalg(f1,&f);
2281:
2282: /* append alg vars to the var list */
2283: for ( tv = vv; NEXT(tv); tv = NEXT(tv) );
2284: NEXT(tv) = av;
2285:
2286: /* append a block to ord */
2287: *ord1p = append_block(ord,nvar,nalg,2);
2288:
2289: /* create generator list */
2290: nf = get_numberfield();
2291: for ( i = nalg-1, t = BDY(f); i >= 0; i-- ) {
2292: MKAlg(nf->defpoly[i],dp);
2293: MKNODE(s,dp,t); t = s;
2294: }
2295: MKLIST(f1,t);
2296: *alistp = alist;
2297: algobjtorat(f1,f1p);
1.121 noro 2298:
2299: /* creating a new weight vector */
2300: prev_weight_vector_obj = obj0 = current_dl_weight_vector_obj;
2301: n = nvar+nalg+1;
2302: MKVECT(obj,n);
2303: if ( obj0 && obj0->len == nvar )
2304: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = BDY(obj0)[i];
2305: else
2306: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = (pointer)ONE;
2307: for ( i = 0; i < nalg; i++ ) BDY(obj)[i+nvar] = 0;
2308: BDY(obj)[n-1] = (pointer)ONE;
2309: arg = mknode(1,obj);
2310: Pdp_set_weight(arg,&tmp);
2311: }
2312:
2313: NODE postprocess_algcoef(VL av,NODE alist,NODE r)
2314: {
2315: NODE s,t,u0,u;
2316: P p;
2317: VL tv;
2318: Obj obj,tmp;
2319: NODE arg;
2320:
2321: u0 = 0;
2322: for ( t = r; t; t = NEXT(t) ) {
2323: p = (P)BDY(t);
2324: for ( tv = av, s = alist; tv; tv = NEXT(tv), s = NEXT(s) ) {
2325: substr(CO,0,(Obj)p,tv->v,(Obj)BDY(s),&obj); p = (P)obj;
2326: }
2327: if ( OID(p) == O_P || (OID(p) == O_N && NID((Num)p) != N_A) ) {
2328: NEXTNODE(u0,u);
2329: BDY(u) = (pointer)p;
2330: }
2331: }
2332: arg = mknode(1,prev_weight_vector_obj);
2333: Pdp_set_weight(arg,&tmp);
2334:
2335: return u0;
1.120 noro 2336: }
2337:
1.63 noro 2338: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2339: {
1.119 noro 2340: VL tv,fv,vv,vc,av;
2341: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2342: int e,max,nvar,i;
1.69 noro 2343: NDV b;
1.119 noro 2344: int ishomo,nalg;
2345: Alg alpha,dp;
2346: P p;
2347: LIST f1,f2;
2348: Obj obj;
2349: NumberField nf;
1.120 noro 2350: struct order_spec *ord1;
1.1 noro 2351:
1.78 noro 2352: if ( !m && Demand ) nd_demand = 1;
2353: else nd_demand = 0;
2354:
1.103 noro 2355: #if 0
1.74 noro 2356: ndv_alloc = 0;
1.103 noro 2357: #endif
1.1 noro 2358: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2359: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2360: switch ( ord->id ) {
2361: case 1:
2362: if ( ord->nv != nvar )
2363: error("nd_{gr,f4} : invalid order specification");
2364: break;
2365: default:
2366: break;
2367: }
1.119 noro 2368: nd_nalg = 0;
1.123 noro 2369: av = 0;
1.119 noro 2370: if ( !m ) {
2371: get_algtree((Obj)f,&av);
2372: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2373: nd_ntrans = nvar;
1.119 noro 2374: nd_nalg = nalg;
2375: /* #i -> t#i */
2376: if ( nalg ) {
1.120 noro 2377: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2378: ord = ord1;
2379: f = f1;
1.119 noro 2380: }
1.120 noro 2381: nvar += nalg;
1.119 noro 2382: }
1.32 noro 2383: nd_init_ord(ord);
1.61 noro 2384: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2385: for ( tv = vv; tv; tv = NEXT(tv) ) {
2386: e = getdeg(tv->v,(P)BDY(t));
2387: max = MAX(e,max);
2388: }
2389: nd_setup_parameters(nvar,max);
1.77 noro 2390: ishomo = 1;
1.1 noro 2391: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2392: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.77 noro 2393: if ( ishomo )
2394: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2395: if ( m ) ndv_mod(m,b);
1.61 noro 2396: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2397: }
2398: if ( fd0 ) NEXT(fd) = 0;
1.61 noro 2399: ndv_setup(m,0,fd0);
1.77 noro 2400: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.82 noro 2401: nd_demand = 0;
1.61 noro 2402: x = ndv_reducebase(x);
2403: x = ndv_reduceall(m,x);
1.23 noro 2404: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2405: NEXTNODE(r0,r);
1.61 noro 2406: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2407: }
2408: if ( r0 ) NEXT(r) = 0;
1.121 noro 2409: if ( nalg )
2410: r0 = postprocess_algcoef(av,alist,r0);
1.20 noro 2411: MKLIST(*rp,r0);
1.103 noro 2412: #if 0
1.74 noro 2413: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 2414: #endif
1.127 ! noro 2415: }
! 2416:
! 2417: void nd_gr_postproc(LIST f,LIST v,int m,struct order_spec *ord,int do_check,LIST *rp)
! 2418: {
! 2419: VL tv,fv,vv,vc,av;
! 2420: NODE fd,fd0,r,r0,t,x,s,xx,alist;
! 2421: int e,max,nvar,i;
! 2422: NDV b;
! 2423: int ishomo,nalg;
! 2424: Alg alpha,dp;
! 2425: P p;
! 2426: LIST f1,f2;
! 2427: Obj obj;
! 2428: NumberField nf;
! 2429: struct order_spec *ord1;
! 2430:
! 2431: get_vars((Obj)f,&fv); pltovl(v,&vv);
! 2432: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
! 2433: switch ( ord->id ) {
! 2434: case 1:
! 2435: if ( ord->nv != nvar )
! 2436: error("nd_check : invalid order specification");
! 2437: break;
! 2438: default:
! 2439: break;
! 2440: }
! 2441: nd_nalg = 0;
! 2442: av = 0;
! 2443: if ( !m ) {
! 2444: get_algtree((Obj)f,&av);
! 2445: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
! 2446: nd_ntrans = nvar;
! 2447: nd_nalg = nalg;
! 2448: /* #i -> t#i */
! 2449: if ( nalg ) {
! 2450: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
! 2451: ord = ord1;
! 2452: f = f1;
! 2453: }
! 2454: nvar += nalg;
! 2455: }
! 2456: nd_init_ord(ord);
! 2457: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
! 2458: for ( tv = vv; tv; tv = NEXT(tv) ) {
! 2459: e = getdeg(tv->v,(P)BDY(t));
! 2460: max = MAX(e,max);
! 2461: }
! 2462: nd_setup_parameters(nvar,max);
! 2463: ishomo = 1;
! 2464: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
! 2465: b = (pointer)ptondv(CO,vv,(P)BDY(t));
! 2466: if ( ishomo )
! 2467: ishomo = ishomo && ndv_ishomo(b);
! 2468: if ( m ) ndv_mod(m,b);
! 2469: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
! 2470: }
! 2471: if ( fd0 ) NEXT(fd) = 0;
! 2472: ndv_setup(m,0,fd0);
! 2473: for ( x = 0, i = 0; i < nd_psn; i++ )
! 2474: x = update_base(x,i);
! 2475: if ( do_check ) {
! 2476: x = nd_gb(m,ishomo,1);
! 2477: if ( !x ) {
! 2478: *rp = 0;
! 2479: return;
! 2480: }
! 2481: } else {
! 2482: for ( t = x; t; t = NEXT(t) )
! 2483: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
! 2484: }
! 2485: x = ndv_reducebase(x);
! 2486: x = ndv_reduceall(m,x);
! 2487: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
! 2488: NEXTNODE(r0,r);
! 2489: BDY(r) = ndvtop(m,CO,vv,BDY(t));
! 2490: }
! 2491: if ( r0 ) NEXT(r) = 0;
! 2492: if ( nalg )
! 2493: r0 = postprocess_algcoef(av,alist,r0);
! 2494: MKLIST(*rp,r0);
1.20 noro 2495: }
2496:
1.52 noro 2497: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2498: {
1.119 noro 2499: VL tv,fv,vv,vc,av;
2500: NODE fd,fd0,in0,in,r,r0,t,s,cand,alist;
1.61 noro 2501: int m,nocheck,nvar,mindex,e,max;
2502: NDV c;
2503: NMV a;
1.27 noro 2504: P p;
1.61 noro 2505: EPOS oepos;
1.119 noro 2506: int obpe,oadv,wmax,i,len,cbpe,ishomo,nalg;
2507: Alg alpha,dp;
2508: P poly;
2509: LIST f1,f2;
2510: Obj obj;
2511: NumberField nf;
1.120 noro 2512: struct order_spec *ord1;
1.20 noro 2513:
2514: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2515: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2516: switch ( ord->id ) {
2517: case 1:
2518: if ( ord->nv != nvar )
2519: error("nd_gr_trace : invalid order specification");
2520: break;
2521: default:
2522: break;
2523: }
1.119 noro 2524:
2525: get_algtree((Obj)f,&av);
2526: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2527: nd_ntrans = nvar;
1.119 noro 2528: nd_nalg = nalg;
2529: /* #i -> t#i */
2530: if ( nalg ) {
1.120 noro 2531: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2532: ord = ord1;
2533: f = f1;
1.119 noro 2534: }
1.120 noro 2535: nvar += nalg;
1.119 noro 2536:
1.52 noro 2537: nocheck = 0;
2538: mindex = 0;
2539:
1.78 noro 2540: if ( Demand ) nd_demand = 1;
2541: else nd_demand = 0;
2542:
1.52 noro 2543: /* setup modulus */
2544: if ( trace < 0 ) {
2545: trace = -trace;
2546: nocheck = 1;
2547: }
2548: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2549: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2550: for ( tv = vv; tv; tv = NEXT(tv) ) {
2551: e = getdeg(tv->v,(P)BDY(t));
2552: max = MAX(e,max);
1.23 noro 2553: }
1.61 noro 2554: nd_init_ord(ord);
2555: nd_setup_parameters(nvar,max);
2556: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 noro 2557: ishomo = 1;
1.61 noro 2558: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2559: c = ptondv(CO,vv,(P)BDY(t));
1.82 noro 2560: if ( ishomo )
2561: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2562: if ( c ) {
2563: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2564: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2565: }
1.61 noro 2566: }
2567: if ( in0 ) NEXT(in) = 0;
2568: if ( fd0 ) NEXT(fd) = 0;
1.82 noro 2569: if ( !ishomo && homo ) {
1.122 noro 2570: for ( t = in0, wmax = max; t; t = NEXT(t) ) {
1.61 noro 2571: c = (NDV)BDY(t); len = LEN(c);
2572: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2573: wmax = MAX(TD(DL(a)),wmax);
2574: }
2575: homogenize_order(ord,nvar,&ord1);
1.89 noro 2576: nd_init_ord(ord1);
1.61 noro 2577: nd_setup_parameters(nvar+1,wmax);
2578: for ( t = fd0; t; t = NEXT(t) )
2579: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2580: }
1.52 noro 2581: while ( 1 ) {
1.77 noro 2582: if ( Demand )
2583: nd_demand = 1;
1.61 noro 2584: ndv_setup(m,1,fd0);
1.82 noro 2585: cand = nd_gb_trace(m,ishomo || homo);
1.52 noro 2586: if ( !cand ) {
2587: /* failure */
1.61 noro 2588: if ( trace > 1 ) { *rp = 0; return; }
2589: else m = get_lprime(++mindex);
1.52 noro 2590: continue;
2591: }
1.82 noro 2592: if ( !ishomo && homo ) {
1.27 noro 2593: /* dehomogenization */
1.61 noro 2594: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2595: nd_init_ord(ord);
1.61 noro 2596: nd_setup_parameters(nvar,0);
1.27 noro 2597: }
1.77 noro 2598: nd_demand = 0;
1.61 noro 2599: cand = ndv_reducebase(cand);
2600: cand = ndv_reduceall(0,cand);
1.78 noro 2601: cbpe = nd_bpe;
1.61 noro 2602: if ( nocheck )
2603: break;
2604: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2605: /* success */
2606: break;
2607: else if ( trace > 1 ) {
2608: /* failure */
2609: *rp = 0; return;
1.61 noro 2610: } else {
1.52 noro 2611: /* try the next modulus */
2612: m = get_lprime(++mindex);
1.61 noro 2613: /* reset the parameters */
1.82 noro 2614: if ( !ishomo && homo ) {
1.89 noro 2615: nd_init_ord(ord1);
1.61 noro 2616: nd_setup_parameters(nvar+1,wmax);
2617: } else {
2618: nd_init_ord(ord);
2619: nd_setup_parameters(nvar,max);
2620: }
2621: }
1.52 noro 2622: }
1.27 noro 2623: /* dp->p */
1.61 noro 2624: nd_bpe = cbpe;
1.73 noro 2625: nd_setup_parameters(nd_nvar,0);
1.121 noro 2626: for ( r = cand; r; r = NEXT(r) )
1.119 noro 2627: BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.121 noro 2628: if ( nalg )
2629: cand = postprocess_algcoef(av,alist,cand);
1.27 noro 2630: MKLIST(*rp,cand);
1.1 noro 2631: }
2632:
1.61 noro 2633: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2634: {
1.61 noro 2635: UINT *d;
1.57 noro 2636: int i,j,l,s,ord_l;
1.43 noro 2637: struct order_pair *op;
1.1 noro 2638:
2639: d = dl->d;
1.41 noro 2640: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2641: if ( nd_blockmask ) {
2642: l = nd_blockmask->n;
2643: op = nd_blockmask->order_pair;
2644: for ( j = 0, s = 0; j < l; j++ ) {
2645: ord_l = op[j].length;
1.57 noro 2646: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2647: }
2648: TD(r) = ndl_weight(r);
1.61 noro 2649: ndl_weight_mask(r);
1.43 noro 2650: } else {
1.56 noro 2651: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2652: TD(r) = ndl_weight(r);
2653: }
1.1 noro 2654: }
2655:
1.61 noro 2656: DL ndltodl(int n,UINT *ndl)
1.1 noro 2657: {
2658: DL dl;
2659: int *d;
1.57 noro 2660: int i,j,l,s,ord_l;
1.43 noro 2661: struct order_pair *op;
1.1 noro 2662:
2663: NEWDL(dl,n);
1.34 noro 2664: dl->td = TD(ndl);
1.1 noro 2665: d = dl->d;
1.43 noro 2666: if ( nd_blockmask ) {
2667: l = nd_blockmask->n;
2668: op = nd_blockmask->order_pair;
2669: for ( j = 0, s = 0; j < l; j++ ) {
2670: ord_l = op[j].length;
1.57 noro 2671: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2672: }
2673: } else {
1.56 noro 2674: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2675: }
1.1 noro 2676: return dl;
2677: }
2678:
1.61 noro 2679: void ndl_print(UINT *dl)
1.1 noro 2680: {
2681: int n;
1.57 noro 2682: int i,j,l,ord_l,s,s0;
1.43 noro 2683: struct order_pair *op;
1.1 noro 2684:
2685: n = nd_nvar;
2686: printf("<<");
1.43 noro 2687: if ( nd_blockmask ) {
2688: l = nd_blockmask->n;
2689: op = nd_blockmask->order_pair;
2690: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2691: ord_l = op[j].length;
1.57 noro 2692: for ( i = 0; i < ord_l; i++, s++ )
2693: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2694: }
2695: } else {
1.56 noro 2696: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2697: }
1.1 noro 2698: printf(">>");
2699: }
2700:
2701: void nd_print(ND p)
2702: {
2703: NM m;
2704:
2705: if ( !p )
2706: printf("0\n");
2707: else {
2708: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2709: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2710: else printf("+%d*",CM(m));
1.14 noro 2711: ndl_print(DL(m));
1.1 noro 2712: }
2713: printf("\n");
2714: }
2715: }
2716:
1.113 noro 2717: void nd_print_q(ND p)
1.16 noro 2718: {
2719: NM m;
2720:
2721: if ( !p )
2722: printf("0\n");
2723: else {
2724: for ( m = BDY(p); m; m = NEXT(m) ) {
2725: printf("+");
1.113 noro 2726: printexpr(CO,(Obj)CQ(m));
1.16 noro 2727: printf("*");
2728: ndl_print(DL(m));
2729: }
2730: printf("\n");
2731: }
2732: }
2733:
1.1 noro 2734: void ndp_print(ND_pairs d)
2735: {
2736: ND_pairs t;
2737:
1.34 noro 2738: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2739: printf("\n");
2740: }
2741:
1.20 noro 2742: void nd_removecont(int mod,ND p)
1.16 noro 2743: {
2744: int i,n;
1.113 noro 2745: Q *w;
2746: Q dvr,t;
1.16 noro 2747: NM m;
1.21 noro 2748: struct oVECT v;
1.113 noro 2749: N q,r;
1.16 noro 2750:
1.71 noro 2751: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2752: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2753: else {
2754: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
1.113 noro 2755: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2756: v.len = n;
2757: v.body = (pointer *)w;
1.113 noro 2758: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2759: removecont_array(w,n);
1.113 noro 2760: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2761: }
2762: }
2763:
1.21 noro 2764: void nd_removecont2(ND p1,ND p2)
2765: {
2766: int i,n1,n2,n;
1.113 noro 2767: Q *w;
2768: Q dvr,t;
1.21 noro 2769: NM m;
2770: struct oVECT v;
1.113 noro 2771: N q,r;
1.21 noro 2772:
2773: if ( !p1 ) {
2774: nd_removecont(0,p2); return;
2775: } else if ( !p2 ) {
2776: nd_removecont(0,p1); return;
2777: }
2778: n1 = nd_length(p1);
2779: n2 = nd_length(p2);
2780: n = n1+n2;
1.113 noro 2781: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2782: v.len = n;
2783: v.body = (pointer *)w;
1.113 noro 2784: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2785: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2786: removecont_array(w,n);
1.113 noro 2787: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2788: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.21 noro 2789: }
2790:
1.20 noro 2791: void ndv_removecont(int mod,NDV p)
1.16 noro 2792: {
2793: int i,len;
1.113 noro 2794: Q *w;
2795: Q dvr,t;
1.16 noro 2796: NMV m;
2797:
1.71 noro 2798: if ( mod == -1 )
2799: ndv_mul_c(mod,p,_invsf(HCM(p)));
2800: else if ( mod )
1.20 noro 2801: ndv_mul_c(mod,p,invm(HCM(p),mod));
2802: else {
2803: len = p->len;
1.113 noro 2804: w = (Q *)ALLOCA(len*sizeof(Q));
2805: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
2806: sortbynm(w,len);
2807: qltozl(w,len,&dvr);
2808: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2809: divq(CQ(m),dvr,&t); CQ(m) = t;
2810: }
1.16 noro 2811: }
1.21 noro 2812: }
2813:
1.61 noro 2814: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2815: {
2816: int len,i,max;
2817: NMV m,mr0,mr,t;
2818:
2819: len = p->len;
2820: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2821: max = MAX(max,TD(DL(m)));
2822: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2823: m = (NMV)((char *)mr0+(len-1)*oadv);
2824: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2825: t = (NMV)ALLOCA(nmv_adv);
2826: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2827: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
1.113 noro 2828: CQ(mr) = CQ(m);
1.61 noro 2829: ndl_copy(DL(t),DL(mr));
2830: }
2831: NV(p)++;
2832: BDY(p) = mr0;
2833: }
2834:
1.45 noro 2835: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2836: {
1.45 noro 2837: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.113 noro 2838: Q *w;
2839: Q dvr,t;
1.23 noro 2840: NMV m,r;
2841:
2842: len = p->len;
2843: newnvar = nd_nvar-1;
1.48 noro 2844: newexporigin = nd_get_exporigin(ord);
1.45 noro 2845: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2846: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2847: ndl_dehomogenize(DL(m));
1.23 noro 2848: if ( newwpd != nd_wpd ) {
1.90 noro 2849: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
1.61 noro 2850: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.113 noro 2851: CQ(r) = CQ(m);
1.45 noro 2852: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2853: adj = nd_exporigin-newexporigin;
2854: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2855: }
2856: }
2857: NV(p)--;
2858: }
2859:
1.113 noro 2860: void removecont_array(Q *c,int n)
1.21 noro 2861: {
2862: struct oVECT v;
1.113 noro 2863: Q d0,d1,a,u,u1,gcd;
1.110 noro 2864: int i,j;
1.113 noro 2865: N qn,rn,gn;
2866: Q *q,*r;
1.21 noro 2867:
1.113 noro 2868: q = (Q *)ALLOCA(n*sizeof(Q));
2869: r = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2870: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
1.113 noro 2871: igcdv_estimate(&v,&d0);
2872: for ( i = 0; i < n; i++ ) {
2873: divn(NM(c[i]),NM(d0),&qn,&rn);
2874: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2875: NTOQ(rn,SGN(c[i]),r[i]);
2876: }
1.34 noro 2877: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2878: if ( i < n ) {
2879: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
1.113 noro 2880: igcdv(&v,&d1);
2881: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2882: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
1.21 noro 2883: for ( i = 0; i < n; i++ ) {
1.113 noro 2884: mulq(a,q[i],&u);
2885: if ( r[i] ) {
2886: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2887: addq(u,u1,&q[i]);
2888: } else
2889: q[i] = u;
1.21 noro 2890: }
2891: }
1.34 noro 2892: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2893: }
2894:
1.19 noro 2895: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2896: {
2897: NM m;
2898: int c,c1;
2899:
1.34 noro 2900: if ( !p ) return;
1.115 noro 2901: if ( mul == 1 ) return;
1.71 noro 2902: if ( mod == -1 )
2903: for ( m = BDY(p); m; m = NEXT(m) )
2904: CM(m) = _mulsf(CM(m),mul);
2905: else
2906: for ( m = BDY(p); m; m = NEXT(m) ) {
2907: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2908: }
1.1 noro 2909: }
2910:
1.113 noro 2911: void nd_mul_c_q(ND p,Q mul)
1.16 noro 2912: {
2913: NM m;
1.113 noro 2914: Q c;
1.16 noro 2915:
1.113 noro 2916: if ( !p ) return;
1.115 noro 2917: if ( UNIQ(mul) ) return;
1.16 noro 2918: for ( m = BDY(p); m; m = NEXT(m) ) {
1.113 noro 2919: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 2920: }
2921: }
2922:
1.61 noro 2923: void nd_mul_c_p(VL vl,ND p,P mul)
2924: {
2925: NM m;
2926: P c;
2927:
2928: if ( !p ) return;
2929: for ( m = BDY(p); m; m = NEXT(m) ) {
2930: mulp(vl,CP(m),mul,&c); CP(m) = c;
2931: }
2932: }
2933:
1.1 noro 2934: void nd_free(ND p)
2935: {
2936: NM t,s;
2937:
1.34 noro 2938: if ( !p ) return;
1.1 noro 2939: t = BDY(p);
2940: while ( t ) {
2941: s = NEXT(t);
2942: FREENM(t);
2943: t = s;
2944: }
2945: FREEND(p);
2946: }
2947:
1.23 noro 2948: void ndv_free(NDV p)
2949: {
2950: GC_free(BDY(p));
2951: }
2952:
1.61 noro 2953: void nd_append_red(UINT *d,int i)
1.1 noro 2954: {
1.13 noro 2955: RHist m,m0;
1.1 noro 2956: int h;
2957:
1.13 noro 2958: NEWRHist(m);
1.34 noro 2959: h = ndl_hash_value(d);
1.13 noro 2960: m->index = i;
1.14 noro 2961: ndl_copy(d,DL(m));
1.1 noro 2962: NEXT(m) = nd_red[h];
2963: nd_red[h] = m;
2964: }
2965:
1.61 noro 2966: UINT *ndv_compute_bound(NDV p)
1.1 noro 2967: {
1.61 noro 2968: UINT *d1,*d2,*t;
2969: UINT u;
1.57 noro 2970: int i,j,k,l,len,ind;
1.45 noro 2971: NMV m;
1.1 noro 2972:
2973: if ( !p )
2974: return 0;
1.61 noro 2975: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2976: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2977: len = LEN(p);
2978: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2979: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2980: ndl_lcm(DL(m),d1,d2);
1.1 noro 2981: t = d1; d1 = d2; d2 = t;
2982: }
1.12 noro 2983: l = nd_nvar+31;
1.61 noro 2984: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 2985: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2986: u = d1[i];
2987: k = (nd_epw-1)*nd_bpe;
2988: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2989: t[ind] = (u>>k)&nd_mask0;
2990: }
2991: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 2992: return t;
2993: }
2994:
1.99 noro 2995: UINT *nd_compute_bound(ND p)
2996: {
2997: UINT *d1,*d2,*t;
2998: UINT u;
2999: int i,j,k,l,len,ind;
3000: NM m;
3001:
3002: if ( !p )
3003: return 0;
3004: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3005: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3006: len = LEN(p);
3007: m = BDY(p); ndl_copy(DL(m),d1); m = NEXT(m);
3008: for ( m = NEXT(m); m; m = NEXT(m) ) {
3009: ndl_lcm(DL(m),d1,d2);
3010: t = d1; d1 = d2; d2 = t;
3011: }
3012: l = nd_nvar+31;
3013: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
3014: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3015: u = d1[i];
3016: k = (nd_epw-1)*nd_bpe;
3017: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3018: t[ind] = (u>>k)&nd_mask0;
3019: }
3020: for ( ; ind < l; ind++ ) t[ind] = 0;
3021: return t;
3022: }
3023:
1.48 noro 3024: int nd_get_exporigin(struct order_spec *ord)
3025: {
1.51 noro 3026: switch ( ord->id ) {
1.96 noro 3027: case 0: case 2:
1.48 noro 3028: return 1;
1.41 noro 3029: case 1:
3030: /* block order */
1.43 noro 3031: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 3032: return ord->ord.block.length+1;
1.96 noro 3033: case 3:
3034: error("nd_get_exporigin : composite order is not supported yet.");
1.41 noro 3035: }
1.48 noro 3036: }
3037:
1.61 noro 3038: void nd_setup_parameters(int nvar,int max) {
1.79 noro 3039: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 3040: struct order_pair *op;
1.48 noro 3041:
1.73 noro 3042: nd_nvar = nvar;
1.79 noro 3043: if ( max ) {
1.83 noro 3044: /* XXX */
3045: if ( do_weyl ) nd_bpe = 32;
3046: else if ( max < 2 ) nd_bpe = 1;
1.79 noro 3047: else if ( max < 4 ) nd_bpe = 2;
3048: else if ( max < 8 ) nd_bpe = 3;
3049: else if ( max < 16 ) nd_bpe = 4;
3050: else if ( max < 32 ) nd_bpe = 5;
3051: else if ( max < 64 ) nd_bpe = 6;
3052: else if ( max < 256 ) nd_bpe = 8;
3053: else if ( max < 1024 ) nd_bpe = 10;
3054: else if ( max < 65536 ) nd_bpe = 16;
3055: else nd_bpe = 32;
3056: }
1.61 noro 3057: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 3058: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
3059: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 3060: wpd = nd_exporigin+elen;
3061: if ( wpd != nd_wpd ) {
3062: nd_free_private_storage();
3063: nd_wpd = wpd;
3064: }
1.1 noro 3065: if ( nd_bpe < 32 ) {
3066: nd_mask0 = (1<<nd_bpe)-1;
3067: } else {
3068: nd_mask0 = 0xffffffff;
3069: }
3070: bzero(nd_mask,sizeof(nd_mask));
3071: nd_mask1 = 0;
3072: for ( i = 0; i < nd_epw; i++ ) {
3073: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
3074: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
3075: }
1.90 noro 3076: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
1.57 noro 3077: nd_epos = nd_create_epos(nd_ord);
1.43 noro 3078: nd_blockmask = nd_create_blockmask(nd_ord);
1.96 noro 3079: nd_work_vector = (int *)REALLOC(nd_work_vector,nd_nvar*sizeof(int));
1.1 noro 3080: }
3081:
1.103 noro 3082: ND_pairs nd_reconstruct(int trace,ND_pairs d)
1.1 noro 3083: {
1.37 noro 3084: int i,obpe,oadv,h;
1.71 noro 3085: static NM prev_nm_free_list;
3086: static ND_pairs prev_ndp_free_list;
1.13 noro 3087: RHist mr0,mr;
3088: RHist r;
1.37 noro 3089: RHist *old_red;
1.71 noro 3090: ND_pairs s0,s,t;
1.43 noro 3091: EPOS oepos;
1.15 noro 3092:
1.1 noro 3093: obpe = nd_bpe;
1.11 noro 3094: oadv = nmv_adv;
1.43 noro 3095: oepos = nd_epos;
1.63 noro 3096: if ( obpe < 2 ) nd_bpe = 2;
3097: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 3098: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 3099: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 3100: else if ( obpe < 6 ) nd_bpe = 6;
3101: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 3102: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 3103: else if ( obpe < 16 ) nd_bpe = 16;
3104: else if ( obpe < 32 ) nd_bpe = 32;
3105: else error("nd_reconstruct : exponent too large");
1.5 noro 3106:
1.73 noro 3107: nd_setup_parameters(nd_nvar,0);
1.1 noro 3108: prev_nm_free_list = _nm_free_list;
3109: prev_ndp_free_list = _ndp_free_list;
3110: _nm_free_list = 0;
3111: _ndp_free_list = 0;
1.53 noro 3112: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
3113: if ( trace )
3114: for ( i = nd_psn-1; i >= 0; i-- )
3115: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 3116: s0 = 0;
3117: for ( t = d; t; t = NEXT(t) ) {
3118: NEXTND_pairs(s0,s);
3119: s->i1 = t->i1;
3120: s->i2 = t->i2;
1.14 noro 3121: SG(s) = SG(t);
1.61 noro 3122: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 3123: }
1.37 noro 3124:
3125: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 3126: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 3127: old_red[i] = nd_red[i];
3128: nd_red[i] = 0;
3129: }
3130: for ( i = 0; i < REDTAB_LEN; i++ )
3131: for ( r = old_red[i]; r; r = NEXT(r) ) {
3132: NEWRHist(mr);
1.13 noro 3133: mr->index = r->index;
1.20 noro 3134: SG(mr) = SG(r);
1.61 noro 3135: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 3136: h = ndl_hash_value(DL(mr));
3137: NEXT(mr) = nd_red[h];
3138: nd_red[h] = mr;
1.6 noro 3139: }
1.37 noro 3140: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
3141: old_red = 0;
1.11 noro 3142: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 3143: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 3144: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 3145: nd_psh[i] = r;
1.11 noro 3146: }
1.1 noro 3147: if ( s0 ) NEXT(s) = 0;
3148: prev_nm_free_list = 0;
3149: prev_ndp_free_list = 0;
1.71 noro 3150: #if 0
1.1 noro 3151: GC_gcollect();
1.71 noro 3152: #endif
1.1 noro 3153: return s0;
3154: }
3155:
1.61 noro 3156: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 3157: {
1.57 noro 3158: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 3159: struct order_pair *op;
1.1 noro 3160:
3161: n = nd_nvar;
1.61 noro 3162: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 3163: omask0 = (1<<obpe)-1;
1.34 noro 3164: TD(r) = TD(d);
1.41 noro 3165: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 3166: if ( nd_blockmask ) {
3167: l = nd_blockmask->n;
3168: op = nd_blockmask->order_pair;
3169: for ( i = 1; i < nd_exporigin; i++ )
3170: r[i] = d[i];
3171: for ( j = 0, s = 0; j < l; j++ ) {
3172: ord_l = op[j].length;
1.57 noro 3173: for ( i = 0; i < ord_l; i++, s++ ) {
3174: ei = GET_EXP_OLD(d,s);
3175: PUT_EXP(r,s,ei);
3176: }
1.1 noro 3177: }
1.43 noro 3178: } else {
1.56 noro 3179: for ( i = 0; i < n; i++ ) {
3180: ei = GET_EXP_OLD(d,i);
3181: PUT_EXP(r,i,ei);
3182: }
1.1 noro 3183: }
3184: }
1.3 noro 3185:
1.6 noro 3186: ND nd_copy(ND p)
3187: {
3188: NM m,mr,mr0;
1.41 noro 3189: int c,n;
1.6 noro 3190: ND r;
3191:
3192: if ( !p )
3193: return 0;
3194: else {
3195: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3196: NEXTNM(mr0,mr);
1.14 noro 3197: CM(mr) = CM(m);
3198: ndl_copy(DL(m),DL(mr));
1.6 noro 3199: }
3200: NEXT(mr) = 0;
1.31 noro 3201: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 3202: SG(r) = SG(p);
1.6 noro 3203: return r;
3204: }
3205: }
3206:
1.53 noro 3207: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3208: {
3209: NM m;
3210: NDV p1,p2;
3211: ND t1,t2;
1.61 noro 3212: UINT *lcm;
1.31 noro 3213: int td;
1.11 noro 3214:
1.77 noro 3215: if ( !mod && nd_demand ) {
3216: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 3217: } else {
1.77 noro 3218: if ( trace ) {
3219: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3220: } else {
3221: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3222: }
1.20 noro 3223: }
1.34 noro 3224: lcm = LCM(p);
1.11 noro 3225: NEWNM(m);
1.113 noro 3226: CQ(m) = HCQ(p2);
1.34 noro 3227: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 3228: if ( ndl_check_bound2(p->i1,DL(m)) )
3229: return 0;
1.55 noro 3230: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 3231: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
3232: else if ( mod ) CM(m) = mod-HCM(p1);
1.113 noro 3233: else chsgnq(HCQ(p1),&CQ(m));
1.34 noro 3234: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 3235: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 3236: nd_free(t1);
3237: return 0;
3238: }
1.55 noro 3239: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 3240: *rp = nd_add(mod,t1,t2);
1.11 noro 3241: FREENM(m);
3242: return 1;
3243: }
3244:
1.19 noro 3245: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3246: {
3247: NMV m;
3248: int c,c1,len,i;
3249:
1.34 noro 3250: if ( !p ) return;
1.14 noro 3251: len = LEN(p);
1.71 noro 3252: if ( mod == -1 )
3253: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3254: CM(m) = _mulsf(CM(m),mul);
3255: else
3256: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3257: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3258: }
1.11 noro 3259: }
3260:
1.113 noro 3261: void ndv_mul_c_q(NDV p,Q mul)
1.16 noro 3262: {
3263: NMV m;
1.113 noro 3264: Q c;
1.16 noro 3265: int len,i;
3266:
1.34 noro 3267: if ( !p ) return;
1.16 noro 3268: len = LEN(p);
3269: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 3270: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 3271: }
3272: }
3273:
1.55 noro 3274: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3275: int n2,i,j,l,n,tlen;
1.61 noro 3276: UINT *d0;
1.55 noro 3277: NM *tab,*psum;
3278: ND s,r;
3279: NM t;
3280: NMV m1;
3281:
3282: if ( !p ) return 0;
3283: n = NV(p); n2 = n>>1;
3284: d0 = DL(m0);
3285: l = LEN(p);
3286: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3287: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3288: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3289: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3290: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3291: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3292: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3293: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3294: for ( j = 0; j < tlen; j++ ) {
3295: if ( tab[j] ) {
3296: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3297: }
3298: }
3299: }
3300: for ( i = tlen-1, r = 0; i >= 0; i-- )
3301: if ( psum[i] ) {
3302: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3303: MKND(n,psum[i],j,s);
3304: r = nd_add(mod,r,s);
3305: }
1.56 noro 3306: if ( r ) SG(r) = SG(p)+TD(d0);
3307: return r;
1.55 noro 3308: }
3309:
1.56 noro 3310: /* product of monomials */
3311: /* XXX block order is not handled correctly */
3312:
1.55 noro 3313: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3314: {
1.56 noro 3315: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3316: UINT *d0,*d1,*d,*dt,*ctab;
1.113 noro 3317: Q *ctab_q;
3318: Q q,q1;
1.61 noro 3319: UINT c0,c1,c;
1.55 noro 3320: NM *p;
3321: NM m,t;
3322:
3323: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3324: if ( !m0 || !m1 ) return;
3325: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3326: NEWNM(m); d = DL(m);
1.56 noro 3327: if ( mod ) {
3328: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3329: } else
1.113 noro 3330: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3331: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3332: homo = n&1 ? 1 : 0;
3333: if ( homo ) {
3334: /* offset of h-degree */
3335: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3336: PUT_EXP(DL(m),n-1,h);
3337: TD(DL(m)) = h;
1.61 noro 3338: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3339: }
3340: tab[0] = m;
3341: NEWNM(m); d = DL(m);
1.57 noro 3342: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3343: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3344: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3345: /* xi^a*(Di^k*xi^l)*Di^b */
3346: a += l; b += k;
1.56 noro 3347: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3348: if ( !k || !l ) {
3349: for ( j = 0; j < curlen; j++ )
1.56 noro 3350: if ( t = tab[j] ) {
3351: dt = DL(t);
3352: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3353: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3354: }
3355: curlen *= k+1;
3356: continue;
3357: }
3358: min = MIN(k,l);
1.56 noro 3359: if ( mod ) {
1.61 noro 3360: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3361: mkwcm(k,l,mod,ctab);
3362: } else {
1.113 noro 3363: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3364: mkwc(k,l,ctab_q);
1.56 noro 3365: }
1.57 noro 3366: for ( j = min; j >= 0; j-- ) {
1.56 noro 3367: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3368: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3369: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3370: if ( homo ) {
3371: TD(d) = s;
1.56 noro 3372: PUT_EXP(d,n-1,s-h);
1.55 noro 3373: } else TD(d) = h;
1.61 noro 3374: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3375: if ( mod ) c = ctab[j];
1.113 noro 3376: else q = ctab_q[j];
1.57 noro 3377: p = tab+curlen*j;
3378: if ( j == 0 ) {
3379: for ( u = 0; u < curlen; u++, p++ ) {
3380: if ( tab[u] ) {
3381: ndl_addto(DL(tab[u]),d);
3382: if ( mod ) {
3383: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3384: } else {
1.113 noro 3385: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
1.57 noro 3386: }
3387: }
1.56 noro 3388: }
1.57 noro 3389: } else {
3390: for ( u = 0; u < curlen; u++, p++ ) {
3391: if ( tab[u] ) {
3392: NEWNM(t);
3393: ndl_add(DL(tab[u]),d,DL(t));
3394: if ( mod ) {
3395: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3396: } else
1.113 noro 3397: mulq(CQ(tab[u]),q,&CQ(t));
1.57 noro 3398: *p = t;
3399: }
1.55 noro 3400: }
3401: }
3402: }
3403: curlen *= k+1;
3404: }
3405: FREENM(m);
3406: }
3407:
1.63 noro 3408: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3409: {
3410: NM mr,mr0;
3411: NMV m;
3412: UINT *d,*dt,*dm;
3413: int c,n,td,i,c1,c2,len;
1.113 noro 3414: Q q;
1.63 noro 3415: ND r;
3416:
3417: if ( !p ) return 0;
3418: else {
3419: n = NV(p); m = BDY(p);
3420: d = DL(m0);
3421: len = LEN(p);
3422: mr0 = 0;
3423: td = TD(d);
3424: c = CM(m0);
3425: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3426: NEXTNM(mr0,mr);
3427: CM(mr) = 1;
3428: ndl_add(DL(m),d,DL(mr));
3429: }
3430: NEXT(mr) = 0;
3431: MKND(NV(p),mr0,len,r);
3432: SG(r) = SG(p) + TD(d);
3433: return r;
3434: }
3435: }
3436:
1.55 noro 3437: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3438: {
3439: NM mr,mr0;
3440: NMV m;
1.61 noro 3441: UINT *d,*dt,*dm;
1.9 noro 3442: int c,n,td,i,c1,c2,len;
1.113 noro 3443: Q q;
1.9 noro 3444: ND r;
3445:
1.34 noro 3446: if ( !p ) return 0;
1.55 noro 3447: else if ( do_weyl )
1.71 noro 3448: if ( mod == -1 )
3449: error("ndv_mul_nm : not implemented (weyl)");
3450: else
3451: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3452: else {
3453: n = NV(p); m = BDY(p);
1.34 noro 3454: d = DL(m0);
1.14 noro 3455: len = LEN(p);
1.9 noro 3456: mr0 = 0;
1.34 noro 3457: td = TD(d);
1.71 noro 3458: if ( mod == -1 ) {
3459: c = CM(m0);
3460: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3461: NEXTNM(mr0,mr);
3462: CM(mr) = _mulsf(CM(m),c);
3463: ndl_add(DL(m),d,DL(mr));
3464: }
3465: } else if ( mod ) {
1.16 noro 3466: c = CM(m0);
3467: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3468: NEXTNM(mr0,mr);
3469: c1 = CM(m);
1.19 noro 3470: DMAR(c1,c,0,mod,c2);
1.16 noro 3471: CM(mr) = c2;
3472: ndl_add(DL(m),d,DL(mr));
3473: }
3474: } else {
1.113 noro 3475: q = CQ(m0);
1.16 noro 3476: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3477: NEXTNM(mr0,mr);
1.113 noro 3478: mulq(CQ(m),q,&CQ(mr));
1.16 noro 3479: ndl_add(DL(m),d,DL(mr));
3480: }
1.4 noro 3481: }
1.9 noro 3482: NEXT(mr) = 0;
1.31 noro 3483: MKND(NV(p),mr0,len,r);
1.34 noro 3484: SG(r) = SG(p) + TD(d);
1.9 noro 3485: return r;
1.4 noro 3486: }
3487: }
3488:
1.104 noro 3489: ND nd_quo(int mod,PGeoBucket bucket,NDV d)
1.99 noro 3490: {
3491: NM mq0,mq;
1.102 noro 3492: NMV tm;
1.113 noro 3493: Q q;
1.104 noro 3494: int i,nv,sg,c,c1,c2,hindex;
3495: ND p,t,r;
3496: N tnm;
3497:
1.124 noro 3498: if ( bucket->m < 0 ) return 0;
1.99 noro 3499: else {
1.104 noro 3500: nv = NV(d);
1.99 noro 3501: mq0 = 0;
1.102 noro 3502: tm = (NMV)ALLOCA(nmv_adv);
1.104 noro 3503: while ( 1 ) {
1.113 noro 3504: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.104 noro 3505: if ( hindex < 0 ) break;
3506: p = bucket->body[hindex];
1.99 noro 3507: NEXTNM(mq0,mq);
3508: ndl_sub(HDL(p),HDL(d),DL(mq));
1.102 noro 3509: ndl_copy(DL(mq),DL(tm));
3510: if ( mod ) {
3511: c1 = invm(HCM(d),mod); c2 = HCM(p);
3512: DMAR(c1,c2,0,mod,c); CM(mq) = c;
3513: CM(tm) = mod-c;
3514: } else {
1.113 noro 3515: divsn(NM(HCQ(p)),NM(HCQ(d)),&tnm);
3516: NTOQ(tnm,SGN(HCQ(p))*SGN(HCQ(d)),CQ(mq));
3517: chsgnq(CQ(mq),&CQ(tm));
1.102 noro 3518: }
3519: t = ndv_mul_nmv_trunc(mod,tm,d,HDL(d));
1.104 noro 3520: bucket->body[hindex] = nd_remove_head(p);
3521: t = nd_remove_head(t);
3522: add_pbucket(mod,bucket,t);
3523: }
3524: if ( !mq0 )
3525: r = 0;
3526: else {
3527: NEXT(mq) = 0;
3528: for ( i = 0, mq = mq0; mq; mq = NEXT(mq), i++ );
3529: MKND(nv,mq0,i,r);
3530: /* XXX */
3531: SG(r) = HTD(r);
1.99 noro 3532: }
3533: return r;
3534: }
3535: }
3536:
1.43 noro 3537: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3538: {
1.13 noro 3539: NMV m,mr,mr0,t;
3540: int len,i,k;
1.11 noro 3541:
1.61 noro 3542: if ( !p ) return;
3543: m = BDY(p); len = LEN(p);
3544: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3545: m = (NMV)((char *)mr0+(len-1)*oadv);
3546: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3547: t = (NMV)ALLOCA(nmv_adv);
3548: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
1.113 noro 3549: CQ(t) = CQ(m);
1.61 noro 3550: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3551: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
1.113 noro 3552: CQ(mr) = CQ(t);
1.61 noro 3553: ndl_copy(DL(t),DL(mr));
3554: }
3555: BDY(p) = mr0;
3556: }
3557:
3558: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3559: {
3560: NMV m,mr,mr0;
3561: int len,i;
3562: NDV r;
1.11 noro 3563:
1.61 noro 3564: if ( !p ) return 0;
3565: m = BDY(p); len = LEN(p);
3566: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3567: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3568: ndl_zero(DL(mr));
3569: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
1.113 noro 3570: CQ(mr) = CQ(m);
1.11 noro 3571: }
1.61 noro 3572: MKNDV(NV(p),mr0,len,r);
3573: SG(r) = SG(p);
3574: return r;
1.11 noro 3575: }
3576:
1.61 noro 3577: /* duplicate p */
3578:
3579: NDV ndv_dup(int mod,NDV p)
1.3 noro 3580: {
3581: NDV d;
1.61 noro 3582: NMV t,m,m0;
1.3 noro 3583: int i,len;
3584:
1.34 noro 3585: if ( !p ) return 0;
1.31 noro 3586: len = LEN(p);
1.34 noro 3587: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3588: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3589: ndl_copy(DL(t),DL(m));
1.113 noro 3590: CQ(m) = CQ(t);
1.3 noro 3591: }
3592: MKNDV(NV(p),m0,len,d);
1.23 noro 3593: SG(d) = SG(p);
3594: return d;
3595: }
3596:
1.63 noro 3597: ND nd_dup(ND p)
3598: {
3599: ND d;
3600: NM t,m,m0;
3601:
3602: if ( !p ) return 0;
3603: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3604: NEXTNM(m0,m);
3605: ndl_copy(DL(t),DL(m));
1.113 noro 3606: CQ(m) = CQ(t);
1.63 noro 3607: }
3608: if ( m0 ) NEXT(m) = 0;
3609: MKND(NV(p),m0,LEN(p),d);
3610: SG(d) = SG(p);
3611: return d;
3612: }
3613:
1.61 noro 3614: /* XXX if p->len == 0 then it represents 0 */
3615:
3616: void ndv_mod(int mod,NDV p)
3617: {
3618: NMV t,d;
1.125 noro 3619: int r,s,u;
1.61 noro 3620: int i,len,dlen;
1.71 noro 3621: Obj gfs;
1.61 noro 3622:
3623: if ( !p ) return;
3624: len = LEN(p);
3625: dlen = 0;
1.71 noro 3626: if ( mod == -1 )
3627: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3628: simp_ff((Obj)CP(t),&gfs);
3629: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3630: CM(d) = r;
3631: ndl_copy(DL(t),DL(d));
3632: NMV_ADV(d);
3633: dlen++;
3634: }
1.71 noro 3635: else
3636: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
1.113 noro 3637: r = rem(NM(CQ(t)),mod);
1.71 noro 3638: if ( r ) {
1.113 noro 3639: if ( SGN(CQ(t)) < 0 )
3640: r = mod-r;
1.125 noro 3641: if ( DN(CQ(t)) ) {
3642: s = rem(DN(CQ(t)),mod);
3643: if ( !s )
3644: error("ndv_mod : division by 0");
3645: s = invm(s,mod);
3646: DMAR(r,s,0,mod,u); r = u;
3647: }
1.71 noro 3648: CM(d) = r;
3649: ndl_copy(DL(t),DL(d));
3650: NMV_ADV(d);
3651: dlen++;
3652: }
3653: }
1.61 noro 3654: LEN(p) = dlen;
3655: }
3656:
3657: NDV ptondv(VL vl,VL dvl,P p)
3658: {
3659: ND nd;
3660:
3661: nd = ptond(vl,dvl,p);
3662: return ndtondv(0,nd);
3663: }
3664:
3665: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3666: {
1.61 noro 3667: int n,i,j,k,e;
3668: VL tvl;
3669: V v;
3670: DCP dc;
3671: DCP *w;
3672: ND r,s,t,u;
3673: P x;
3674: int c;
3675: UINT *d;
1.23 noro 3676: NM m,m0;
1.61 noro 3677:
3678: if ( !p )
3679: return 0;
3680: else if ( NUM(p) ) {
3681: NEWNM(m);
3682: ndl_zero(DL(m));
1.113 noro 3683: CQ(m) = (Q)p;
1.61 noro 3684: NEXT(m) = 0;
3685: MKND(nd_nvar,m,1,r);
3686: SG(r) = 0;
3687: return r;
3688: } else {
3689: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3690: w = (DCP *)ALLOCA(k*sizeof(DCP));
3691: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3692: for ( i = 0, tvl = dvl, v = VR(p);
1.98 noro 3693: tvl && tvl->v != v; tvl = NEXT(tvl), i++ );
1.61 noro 3694: if ( !tvl ) {
3695: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3696: t = ptond(vl,dvl,COEF(w[j]));
3697: pwrp(vl,x,DEG(w[j]),&p);
3698: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3699: }
3700: return s;
3701: } else {
3702: NEWNM(m0); d = DL(m0);
3703: for ( j = k-1, s = 0; j >= 0; j-- ) {
3704: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3705: TD(d) = MUL_WEIGHT(e,i);
3706: if ( nd_blockmask) ndl_weight_mask(d);
3707: t = ptond(vl,dvl,COEF(w[j]));
3708: for ( m = BDY(t); m; m = NEXT(m) )
3709: ndl_addto(DL(m),d);
3710: SG(t) += TD(d);
3711: s = nd_add(0,s,t);
3712: }
3713: FREENM(m0);
3714: return s;
3715: }
3716: }
3717: }
3718:
3719: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3720: {
3721: VL tvl;
3722: int len,n,j,i,e;
3723: NMV m;
3724: Q q;
3725: P c;
3726: UINT *d;
3727: P s,r,u,t,w;
1.71 noro 3728: GFS gfs;
1.23 noro 3729:
1.34 noro 3730: if ( !p ) return 0;
1.61 noro 3731: else {
3732: len = LEN(p);
3733: n = NV(p);
3734: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3735: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3736: if ( mod == -1 ) {
3737: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3738: } else if ( mod ) {
1.61 noro 3739: STOQ(CM(m),q); c = (P)q;
3740: } else
1.113 noro 3741: c = CP(m);
1.61 noro 3742: d = DL(m);
3743: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3744: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3745: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3746: }
3747: addp(vl,s,t,&u); s = u;
3748: }
3749: return s;
1.23 noro 3750: }
1.3 noro 3751: }
3752:
1.61 noro 3753: NDV ndtondv(int mod,ND p)
1.11 noro 3754: {
3755: NDV d;
1.61 noro 3756: NMV m,m0;
3757: NM t;
3758: int i,len;
1.11 noro 3759:
1.34 noro 3760: if ( !p ) return 0;
1.61 noro 3761: len = LEN(p);
1.74 noro 3762: if ( mod )
3763: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3764: else
3765: m0 = m = MALLOC(len*nmv_adv);
1.103 noro 3766: #if 0
1.74 noro 3767: ndv_alloc += nmv_adv*len;
1.103 noro 3768: #endif
1.61 noro 3769: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3770: ndl_copy(DL(t),DL(m));
1.113 noro 3771: CQ(m) = CQ(t);
1.11 noro 3772: }
1.61 noro 3773: MKNDV(NV(p),m0,len,d);
1.14 noro 3774: SG(d) = SG(p);
1.11 noro 3775: return d;
3776: }
3777:
1.61 noro 3778: ND ndvtond(int mod,NDV p)
1.11 noro 3779: {
1.61 noro 3780: ND d;
3781: NM m,m0;
1.11 noro 3782: NMV t;
1.61 noro 3783: int i,len;
1.11 noro 3784:
1.34 noro 3785: if ( !p ) return 0;
1.11 noro 3786: m0 = 0;
1.61 noro 3787: len = p->len;
3788: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3789: NEXTNM(m0,m);
3790: ndl_copy(DL(t),DL(m));
1.113 noro 3791: CQ(m) = CQ(t);
1.11 noro 3792: }
3793: NEXT(m) = 0;
1.61 noro 3794: MKND(NV(p),m0,len,d);
1.14 noro 3795: SG(d) = SG(p);
1.11 noro 3796: return d;
3797: }
3798:
1.3 noro 3799: void ndv_print(NDV p)
3800: {
3801: NMV m;
3802: int i,len;
3803:
1.34 noro 3804: if ( !p ) printf("0\n");
1.3 noro 3805: else {
1.14 noro 3806: len = LEN(p);
1.3 noro 3807: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3808: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3809: else printf("+%d*",CM(m));
1.16 noro 3810: ndl_print(DL(m));
3811: }
3812: printf("\n");
3813: }
3814: }
3815:
1.113 noro 3816: void ndv_print_q(NDV p)
1.16 noro 3817: {
3818: NMV m;
3819: int i,len;
3820:
1.34 noro 3821: if ( !p ) printf("0\n");
1.16 noro 3822: else {
3823: len = LEN(p);
3824: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3825: printf("+");
1.113 noro 3826: printexpr(CO,(Obj)CQ(m));
1.16 noro 3827: printf("*");
1.14 noro 3828: ndl_print(DL(m));
1.3 noro 3829: }
3830: printf("\n");
3831: }
1.25 noro 3832: }
3833:
1.61 noro 3834: NODE ndv_reducebase(NODE x)
1.27 noro 3835: {
3836: int len,i,j;
3837: NDV *w;
3838: NODE t,t0;
3839:
3840: len = length(x);
3841: w = (NDV *)ALLOCA(len*sizeof(NDV));
3842: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3843: for ( i = 0; i < len; i++ ) {
3844: for ( j = 0; j < i; j++ ) {
3845: if ( w[i] && w[j] )
3846: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3847: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3848: }
3849: }
3850: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3851: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3852: }
3853: NEXT(t) = 0; x = t0;
3854: return x;
1.11 noro 3855: }
1.32 noro 3856:
1.43 noro 3857: /* XXX incomplete */
3858:
1.32 noro 3859: void nd_init_ord(struct order_spec *ord)
3860: {
1.43 noro 3861: switch ( ord->id ) {
1.32 noro 3862: case 0:
1.43 noro 3863: switch ( ord->ord.simple ) {
3864: case 0:
3865: nd_dcomp = 1;
3866: nd_isrlex = 1;
3867: break;
3868: case 1:
3869: nd_dcomp = 1;
3870: nd_isrlex = 0;
3871: break;
3872: case 2:
3873: nd_dcomp = 0;
3874: nd_isrlex = 0;
1.45 noro 3875: ndl_compare_function = ndl_lex_compare;
1.58 noro 3876: break;
3877: case 11:
3878: /* XXX */
3879: nd_dcomp = 0;
3880: nd_isrlex = 1;
3881: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3882: break;
3883: default:
3884: error("nd_gr : unsupported order");
3885: }
1.32 noro 3886: break;
3887: case 1:
1.96 noro 3888: /* block order */
1.43 noro 3889: /* XXX */
3890: nd_dcomp = -1;
1.32 noro 3891: nd_isrlex = 0;
1.45 noro 3892: ndl_compare_function = ndl_block_compare;
1.34 noro 3893: break;
1.43 noro 3894: case 2:
1.96 noro 3895: /* matrix order */
3896: /* XXX */
3897: nd_dcomp = -1;
3898: nd_isrlex = 0;
3899: nd_matrix_len = ord->ord.matrix.row;
3900: nd_matrix = ord->ord.matrix.matrix;
3901: ndl_compare_function = ndl_matrix_compare;
3902: break;
3903: case 3:
1.97 noro 3904: /* composite order */
3905: nd_dcomp = -1;
3906: nd_isrlex = 0;
3907: nd_worb_len = ord->ord.composite.length;
3908: nd_worb = ord->ord.composite.w_or_b;
3909: ndl_compare_function = ndl_composite_compare;
1.32 noro 3910: break;
3911: }
1.41 noro 3912: nd_ord = ord;
1.32 noro 3913: }
3914:
1.43 noro 3915: BlockMask nd_create_blockmask(struct order_spec *ord)
3916: {
3917: int n,i,j,s,l;
1.61 noro 3918: UINT *t;
1.43 noro 3919: BlockMask bm;
3920:
1.96 noro 3921: /* we only create mask table for block order */
3922: if ( ord->id != 1 )
1.43 noro 3923: return 0;
3924: n = ord->ord.block.length;
3925: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3926: bm->n = n;
3927: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3928: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3929: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3930: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3931: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3932: l = bm->order_pair[i].length;
3933: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3934: }
3935: return bm;
1.57 noro 3936: }
3937:
3938: EPOS nd_create_epos(struct order_spec *ord)
3939: {
3940: int i,j,l,s,ord_l,ord_o;
3941: EPOS epos;
3942: struct order_pair *op;
3943:
3944: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3945: switch ( ord->id ) {
3946: case 0:
3947: if ( nd_isrlex ) {
3948: for ( i = 0; i < nd_nvar; i++ ) {
3949: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3950: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3951: }
3952: } else {
3953: for ( i = 0; i < nd_nvar; i++ ) {
3954: epos[i].i = nd_exporigin + i/nd_epw;
3955: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3956: }
3957: }
3958: break;
3959: case 1:
3960: /* block order */
3961: l = ord->ord.block.length;
3962: op = ord->ord.block.order_pair;
3963: for ( j = 0, s = 0; j < l; j++ ) {
3964: ord_o = op[j].order;
3965: ord_l = op[j].length;
3966: if ( !ord_o )
3967: for ( i = 0; i < ord_l; i++ ) {
3968: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3969: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3970: }
3971: else
3972: for ( i = 0; i < ord_l; i++ ) {
3973: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3974: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3975: }
3976: s += ord_l;
3977: }
3978: break;
3979: case 2:
1.96 noro 3980: /* matrix order */
3981: case 3:
3982: /* composite order */
3983: for ( i = 0; i < nd_nvar; i++ ) {
3984: epos[i].i = nd_exporigin + i/nd_epw;
3985: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3986: }
3987: break;
1.57 noro 3988: }
3989: return epos;
1.43 noro 3990: }
1.59 noro 3991:
3992: /* external interface */
3993:
3994: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
3995: {
1.61 noro 3996: NODE t,in0,in;
1.59 noro 3997: ND nd,nf;
1.61 noro 3998: NDV ndv;
3999: VL vv,tv;
4000: int stat,nvar,max,e;
1.113 noro 4001: union oNDC dn;
1.59 noro 4002:
1.116 noro 4003: if ( !f ) {
4004: *rp = 0;
4005: return;
4006: }
1.59 noro 4007: pltovl(v,&vv);
1.74 noro 4008: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 4009:
4010: /* get the degree bound */
4011: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
4012: for ( tv = vv; tv; tv = NEXT(tv) ) {
4013: e = getdeg(tv->v,(P)BDY(t));
4014: max = MAX(e,max);
4015: }
4016: for ( tv = vv; tv; tv = NEXT(tv) ) {
4017: e = getdeg(tv->v,f);
4018: max = MAX(e,max);
4019: }
4020:
1.59 noro 4021: nd_init_ord(ord);
1.61 noro 4022: nd_setup_parameters(nvar,max);
4023:
4024: /* conversion to ndv */
4025: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
4026: NEXTNODE(in0,in);
4027: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 4028: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4029: }
4030: NEXTNODE(in0,in);
4031: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 4032: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4033: NEXT(in) = 0;
4034:
4035: ndv_setup(m,0,in0);
1.59 noro 4036: nd_psn--;
4037: nd_scale=2;
4038: while ( 1 ) {
4039: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 4040: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 4041: if ( !stat ) {
4042: nd_psn++;
1.103 noro 4043: nd_reconstruct(0,0);
1.59 noro 4044: nd_psn--;
4045: } else
4046: break;
4047: }
1.61 noro 4048: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 4049: }
4050:
4051: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
4052: {
4053: NM m;
4054: UINT *t,*s;
4055: int i;
4056:
4057: for ( i = 0; i < n; i++ ) r[i] = 0;
4058: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
4059: t = DL(m);
4060: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4061: r[i] = CM(m);
4062: }
4063: for ( i = 0; !r[i]; i++ );
4064: return i;
4065: }
4066:
1.113 noro 4067: int nd_to_vect_q(UINT *s0,int n,ND d,Q *r)
1.74 noro 4068: {
1.107 noro 4069: NM m;
1.74 noro 4070: UINT *t,*s;
1.107 noro 4071: int i;
1.74 noro 4072:
4073: for ( i = 0; i < n; i++ ) r[i] = 0;
1.107 noro 4074: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
1.74 noro 4075: t = DL(m);
4076: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.113 noro 4077: r[i] = CQ(m);
1.74 noro 4078: }
4079: for ( i = 0; !r[i]; i++ );
4080: return i;
4081: }
4082:
1.67 noro 4083: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 4084: {
4085: NM m;
4086: NMV mr;
4087: UINT *d,*t,*s;
4088: NDV p;
1.67 noro 4089: unsigned char *ivc;
4090: unsigned short *ivs;
1.81 noro 4091: UINT *v,*ivi,*s0v;
1.67 noro 4092: int i,j,len,prev,diff,cdiff;
4093: IndArray r;
1.64 noro 4094:
4095: m = pair->mul;
4096: d = DL(m);
4097: p = nd_ps[pair->index];
4098: len = LEN(p);
4099: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 4100: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 4101: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4102: ndl_add(d,DL(mr),t);
4103: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 4104: v[j] = i;
4105: }
1.81 noro 4106: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 4107: r->head = v[0];
4108: diff = 0;
4109: for ( i = 1; i < len; i++ ) {
4110: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
4111: }
4112: if ( diff < 256 ) {
4113: r->width = 1;
4114: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
4115: r->index.c = ivc;
4116: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
4117: } else if ( diff < 65536 ) {
4118: r->width = 2;
4119: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
4120: r->index.s = ivs;
4121: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
4122: } else {
4123: r->width = 4;
4124: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
4125: r->index.i = ivi;
4126: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 4127: }
1.67 noro 4128: return r;
1.64 noro 4129: }
4130:
1.65 noro 4131:
1.113 noro 4132: int ndv_reduce_vect_q(Q *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.107 noro 4133: {
4134: int i,j,k,len,pos,prev;
1.113 noro 4135: Q cs,mcs,c1,c2,cr,gcd,t;
1.107 noro 4136: IndArray ivect;
4137: unsigned char *ivc;
4138: unsigned short *ivs;
4139: unsigned int *ivi;
4140: NDV redv;
4141: NMV mr;
4142: NODE rp;
4143: int maxrs;
4144:
4145: maxrs = 0;
4146: for ( i = 0; i < nred; i++ ) {
4147: ivect = imat[i];
4148: k = ivect->head;
4149: if ( svect[k] ) {
4150: maxrs = MAX(maxrs,rp0[i]->sugar);
4151: redv = nd_ps[rp0[i]->index];
4152: len = LEN(redv); mr = BDY(redv);
1.113 noro 4153: igcd_cofactor(svect[k],CQ(mr),&gcd,&cs,&cr);
4154: chsgnq(cs,&mcs);
4155: if ( !UNIQ(cr) ) {
4156: for ( j = 0; j < col; j++ ) {
4157: mulq(svect[j],cr,&c1); svect[j] = c1;
4158: }
4159: }
1.107 noro 4160: svect[k] = 0; prev = k;
4161: switch ( ivect->width ) {
4162: case 1:
4163: ivc = ivect->index.c;
4164: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4165: pos = prev+ivc[j]; prev = pos;
1.113 noro 4166: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4167: }
4168: break;
4169: case 2:
4170: ivs = ivect->index.s;
4171: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4172: pos = prev+ivs[j]; prev = pos;
1.113 noro 4173: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4174: }
4175: break;
4176: case 4:
4177: ivi = ivect->index.i;
4178: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4179: pos = prev+ivi[j]; prev = pos;
1.113 noro 4180: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4181: }
4182: break;
4183: }
4184: }
4185: }
4186: return maxrs;
4187: }
4188:
1.76 noro 4189: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 4190: {
1.67 noro 4191: int i,j,k,len,pos,prev;
1.66 noro 4192: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 4193: IndArray ivect;
4194: unsigned char *ivc;
4195: unsigned short *ivs;
4196: unsigned int *ivi;
1.65 noro 4197: NDV redv;
1.67 noro 4198: NMV mr;
1.65 noro 4199: NODE rp;
1.76 noro 4200: int maxrs;
1.65 noro 4201:
1.76 noro 4202: maxrs = 0;
1.74 noro 4203: for ( i = 0; i < nred; i++ ) {
1.65 noro 4204: ivect = imat[i];
1.67 noro 4205: k = ivect->head; svect[k] %= m;
1.65 noro 4206: if ( c = svect[k] ) {
1.76 noro 4207: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4208: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 4209: len = LEN(redv); mr = BDY(redv);
4210: svect[k] = 0; prev = k;
4211: switch ( ivect->width ) {
4212: case 1:
4213: ivc = ivect->index.c;
4214: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4215: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
4216: prev = pos;
4217: DMA(c1,c,c2,up,lo);
4218: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4219: } else svect[pos] = lo;
4220: }
4221: break;
4222: case 2:
4223: ivs = ivect->index.s;
4224: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4225: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
4226: prev = pos;
4227: DMA(c1,c,c2,up,lo);
4228: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4229: } else svect[pos] = lo;
4230: }
4231: break;
4232: case 4:
4233: ivi = ivect->index.i;
4234: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4235: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
4236: prev = pos;
4237: DMA(c1,c,c2,up,lo);
4238: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4239: } else svect[pos] = lo;
4240: }
4241: break;
1.65 noro 4242: }
4243: }
4244: }
1.66 noro 4245: for ( i = 0; i < col; i++ )
4246: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 4247: return maxrs;
1.65 noro 4248: }
4249:
1.76 noro 4250: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 4251: {
4252: int i,j,k,len,pos,prev;
4253: UINT c,c1,c2,c3,up,lo,dmy;
4254: IndArray ivect;
4255: unsigned char *ivc;
4256: unsigned short *ivs;
4257: unsigned int *ivi;
4258: NDV redv;
4259: NMV mr;
4260: NODE rp;
1.76 noro 4261: int maxrs;
1.72 noro 4262:
1.76 noro 4263: maxrs = 0;
1.74 noro 4264: for ( i = 0; i < nred; i++ ) {
1.72 noro 4265: ivect = imat[i];
4266: k = ivect->head; svect[k] %= m;
4267: if ( c = svect[k] ) {
1.76 noro 4268: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4269: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 4270: len = LEN(redv); mr = BDY(redv);
4271: svect[k] = 0; prev = k;
4272: switch ( ivect->width ) {
4273: case 1:
4274: ivc = ivect->index.c;
4275: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4276: pos = prev+ivc[j]; prev = pos;
4277: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4278: }
4279: break;
4280: case 2:
4281: ivs = ivect->index.s;
4282: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4283: pos = prev+ivs[j]; prev = pos;
4284: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4285: }
4286: break;
4287: case 4:
4288: ivi = ivect->index.i;
4289: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4290: pos = prev+ivi[j]; prev = pos;
4291: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4292: }
4293: break;
4294: }
4295: }
4296: }
1.76 noro 4297: return maxrs;
1.72 noro 4298: }
4299:
1.65 noro 4300: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
4301: {
4302: int j,k,len;
4303: UINT *p;
4304: UINT c;
4305: NDV r;
4306: NMV mr0,mr;
4307:
4308: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4309: if ( !len ) return 0;
4310: else {
1.74 noro 4311: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4312: #if 0
1.74 noro 4313: ndv_alloc += nmv_adv*len;
1.103 noro 4314: #endif
1.65 noro 4315: mr = mr0;
4316: p = s0vect;
4317: for ( j = k = 0; j < col; j++, p += nd_wpd )
4318: if ( !rhead[j] ) {
4319: if ( c = vect[k++] ) {
4320: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
4321: }
4322: }
4323: MKNDV(nd_nvar,mr0,len,r);
4324: return r;
4325: }
4326: }
4327:
1.113 noro 4328: NDV vect_to_ndv_q(Q *vect,int spcol,int col,int *rhead,UINT *s0vect)
1.107 noro 4329: {
4330: int j,k,len;
4331: UINT *p;
1.113 noro 4332: Q c;
1.107 noro 4333: NDV r;
4334: NMV mr0,mr;
4335:
4336: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4337: if ( !len ) return 0;
4338: else {
1.109 noro 4339: mr0 = (NMV)GC_malloc(nmv_adv*len);
1.107 noro 4340: #if 0
4341: ndv_alloc += nmv_adv*len;
4342: #endif
4343: mr = mr0;
4344: p = s0vect;
4345: for ( j = k = 0; j < col; j++, p += nd_wpd )
4346: if ( !rhead[j] ) {
4347: if ( c = vect[k++] ) {
1.113 noro 4348: if ( DN(c) )
4349: error("afo");
4350: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
1.107 noro 4351: }
4352: }
4353: MKNDV(nd_nvar,mr0,len,r);
4354: return r;
4355: }
4356: }
4357:
1.74 noro 4358: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket)
1.65 noro 4359: {
4360: ND_pairs t;
4361: NODE sp0,sp;
4362: int stat;
4363: ND spol;
4364:
4365: for ( t = l; t; t = NEXT(t) ) {
4366: stat = nd_sp(m,0,t,&spol);
4367: if ( !stat ) return 0;
4368: if ( spol ) {
4369: add_pbucket_symbolic(bucket,spol);
4370: }
4371: }
1.68 noro 4372: return 1;
1.65 noro 4373: }
4374:
4375: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
4376: {
4377: NODE rp0,rp;
4378: NM mul,head,s0,s;
1.76 noro 4379: int index,col,i,sugar;
1.65 noro 4380: RHist h;
4381: UINT *s0v,*p;
4382: NM_ind_pair pair;
4383: ND red;
4384:
4385: s0 = 0; rp0 = 0; col = 0;
4386: while ( 1 ) {
4387: head = remove_head_pbucket_symbolic(bucket);
4388: if ( !head ) break;
4389: if ( !s0 ) s0 = head;
4390: else NEXT(s) = head;
4391: s = head;
4392: index = ndl_find_reducer(DL(head));
4393: if ( index >= 0 ) {
4394: h = nd_psh[index];
4395: NEWNM(mul);
4396: ndl_sub(DL(head),DL(h),DL(mul));
4397: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.76 noro 4398: sugar = TD(DL(mul))+SG(nd_ps[index]);
4399: MKNM_ind_pair(pair,mul,index,sugar);
1.65 noro 4400: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
4401: add_pbucket_symbolic(bucket,nd_remove_head(red));
4402: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4403: }
4404: col++;
4405: }
1.72 noro 4406: if ( rp0 ) NEXT(rp) = 0;
4407: NEXT(s) = 0;
1.65 noro 4408: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4409: for ( i = 0, p = s0v, s = s0; i < col;
4410: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4411: *s0vect = s0v;
4412: *r = rp0;
4413: return col;
4414: }
4415:
1.69 noro 4416: NODE nd_f4(int m)
4417: {
4418: int i,nh,stat,index;
4419: NODE r,g;
4420: ND_pairs d,l,t;
4421: ND spol,red;
4422: NDV nf,redv;
4423: NM s0,s;
1.84 noro 4424: NODE rp0,srp0,nflist;
1.69 noro 4425: int nsp,nred,col,rank,len,k,j,a;
4426: UINT c;
1.74 noro 4427: UINT **spmat;
1.69 noro 4428: UINT *s0vect,*svect,*p,*v;
4429: int *colstat;
4430: IndArray *imat;
4431: int *rhead;
4432: int spcol,sprow;
4433: int sugar;
4434: PGeoBucket bucket;
4435: struct oEGT eg0,eg1,eg_f4;
4436:
1.103 noro 4437: #if 0
1.74 noro 4438: ndv_alloc = 0;
1.103 noro 4439: #endif
1.69 noro 4440: g = 0; d = 0;
4441: for ( i = 0; i < nd_psn; i++ ) {
4442: d = update_pairs(d,g,i);
4443: g = update_base(g,i);
4444: }
4445: while ( d ) {
4446: get_eg(&eg0);
4447: l = nd_minsugarp(d,&d);
4448: sugar = SG(l);
4449: bucket = create_pbucket();
1.74 noro 4450: stat = nd_sp_f4(m,l,bucket);
1.69 noro 4451: if ( !stat ) {
4452: for ( t = l; NEXT(t); t = NEXT(t) );
4453: NEXT(t) = d; d = l;
1.103 noro 4454: d = nd_reconstruct(0,d);
1.69 noro 4455: continue;
4456: }
1.86 noro 4457: if ( bucket->m < 0 ) continue;
1.69 noro 4458: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
4459: if ( !col ) {
4460: for ( t = l; NEXT(t); t = NEXT(t) );
4461: NEXT(t) = d; d = l;
1.103 noro 4462: d = nd_reconstruct(0,d);
1.69 noro 4463: continue;
4464: }
4465: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4466: if ( DP_Print )
4467: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4468: sugar,eg_f4.exectime+eg_f4.gctime);
4469: if ( 1 )
4470: nflist = nd_f4_red(m,l,s0vect,col,rp0);
4471: else
4472: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0);
1.69 noro 4473: /* adding new bases */
1.74 noro 4474: for ( r = nflist; r; r = NEXT(r) ) {
4475: nf = (NDV)BDY(r);
1.69 noro 4476: ndv_removecont(m,nf);
1.77 noro 4477: nh = ndv_newps(m,nf,0);
1.69 noro 4478: d = update_pairs(d,g,nh);
4479: g = update_base(g,nh);
4480: }
4481: }
4482: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.103 noro 4483: #if 0
1.74 noro 4484: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 4485: #endif
1.69 noro 4486: return g;
4487: }
1.74 noro 4488:
4489: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
1.63 noro 4490: {
1.67 noro 4491: IndArray *imat;
1.106 noro 4492: int nsp,nred,i;
1.65 noro 4493: int *rhead;
1.106 noro 4494: NODE r0,rp;
1.74 noro 4495: ND_pairs sp;
4496: NM_ind_pair *rvect;
1.63 noro 4497:
1.74 noro 4498: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
1.106 noro 4499: nred = length(rp0);
1.74 noro 4500: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4501: rhead = (int *)ALLOCA(col*sizeof(int));
4502: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4503:
1.74 noro 4504: /* construction of index arrays */
4505: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4506: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4507: rvect[i] = (NM_ind_pair)BDY(rp);
4508: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4509: rhead[imat[i]->head] = 1;
4510: }
1.107 noro 4511: if ( m )
4512: r0 = nd_f4_red_main(m,sp0,nsp,s0vect,col,rvect,rhead,imat,nred);
4513: else
1.113 noro 4514: r0 = nd_f4_red_q_main(sp0,nsp,s0vect,col,rvect,rhead,imat,nred);
1.106 noro 4515: return r0;
4516: }
1.74 noro 4517:
1.106 noro 4518: NODE nd_f4_red_main(int m,ND_pairs sp0,int nsp,UINT *s0vect,int col,
4519: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4520: {
4521: int spcol,sprow,a;
4522: int i,j,k,l,rank;
4523: NODE r0,r;
4524: ND_pairs sp;
4525: ND spol;
4526: int **spmat;
4527: UINT *svect,*v;
4528: int *colstat;
4529: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4530: int maxrs;
4531: int *spsugar;
4532:
4533: spcol = col-nred;
4534: get_eg(&eg0);
1.74 noro 4535: /* elimination (1st step) */
4536: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4537: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4538: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4539: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4540: nd_sp(m,0,sp,&spol);
1.75 noro 4541: if ( !spol ) continue;
1.74 noro 4542: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4543: if ( m == -1 )
4544: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4545: else
4546: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4547: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4548: if ( i < col ) {
4549: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4550: for ( j = k = 0; j < col; j++ )
4551: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4552: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4553: sprow++;
4554: }
1.76 noro 4555: nd_free(spol);
1.74 noro 4556: }
1.80 noro 4557: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4558: if ( DP_Print ) {
4559: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4560: fflush(asir_out);
4561: }
1.74 noro 4562: /* free index arrays */
4563: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4564:
4565: /* elimination (2nd step) */
4566: colstat = (int *)ALLOCA(spcol*sizeof(int));
4567: if ( m == -1 )
1.76 noro 4568: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4569: else
1.76 noro 4570: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4571: r0 = 0;
4572: for ( i = 0; i < rank; i++ ) {
4573: NEXTNODE(r0,r); BDY(r) =
4574: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4575: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4576: GC_free(spmat[i]);
4577: }
1.109 noro 4578: if ( r0 ) NEXT(r) = 0;
1.74 noro 4579: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4580: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4581: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4582: if ( DP_Print ) {
1.80 noro 4583: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4584: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4585: nsp,nred,sprow,spcol,rank);
4586: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4587: }
1.74 noro 4588: return r0;
4589: }
4590:
1.113 noro 4591: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,UINT *s0vect,int col,
1.107 noro 4592: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4593: {
4594: int spcol,sprow,a;
4595: int i,j,k,l,rank;
4596: NODE r0,r;
4597: ND_pairs sp;
4598: ND spol;
1.113 noro 4599: Q **spmat;
4600: Q *svect,*v;
1.107 noro 4601: int *colstat;
4602: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4603: int maxrs;
4604: int *spsugar;
4605:
4606: spcol = col-nred;
4607: get_eg(&eg0);
4608: /* elimination (1st step) */
1.113 noro 4609: spmat = (Q **)ALLOCA(nsp*sizeof(Q *));
4610: svect = (Q *)ALLOCA(col*sizeof(Q));
4611: spsugar = (int *)ALLOCA(nsp*sizeof(Q));
1.107 noro 4612: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4613: nd_sp(0,0,sp,&spol);
4614: if ( !spol ) continue;
1.113 noro 4615: nd_to_vect_q(s0vect,col,spol,svect);
4616: maxrs = ndv_reduce_vect_q(svect,col,imat,rvect,nred);
1.107 noro 4617: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4618: if ( i < col ) {
1.113 noro 4619: spmat[sprow] = v = (Q *)MALLOC(spcol*sizeof(Q));
1.107 noro 4620: for ( j = k = 0; j < col; j++ )
4621: if ( !rhead[j] ) v[k++] = svect[j];
4622: spsugar[sprow] = MAX(maxrs,SG(spol));
4623: sprow++;
4624: }
1.109 noro 4625: /* nd_free(spol); */
1.107 noro 4626: }
4627: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4628: if ( DP_Print ) {
4629: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4630: fflush(asir_out);
4631: }
4632: /* free index arrays */
1.109 noro 4633: /* for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c); */
1.107 noro 4634:
4635: /* elimination (2nd step) */
4636: colstat = (int *)ALLOCA(spcol*sizeof(int));
1.113 noro 4637: rank = nd_gauss_elim_q(spmat,spsugar,sprow,spcol,colstat);
1.107 noro 4638: r0 = 0;
4639: for ( i = 0; i < rank; i++ ) {
4640: NEXTNODE(r0,r); BDY(r) =
1.113 noro 4641: (pointer)vect_to_ndv_q(spmat[i],spcol,col,rhead,s0vect);
1.107 noro 4642: SG((NDV)BDY(r)) = spsugar[i];
1.109 noro 4643: /* GC_free(spmat[i]); */
1.107 noro 4644: }
1.109 noro 4645: if ( r0 ) NEXT(r) = 0;
4646:
4647: /* for ( ; i < sprow; i++ ) GC_free(spmat[i]); */
1.107 noro 4648: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4649: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
4650: if ( DP_Print ) {
4651: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
4652: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4653: nsp,nred,sprow,spcol,rank);
4654: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
4655: }
4656: return r0;
4657: }
4658:
1.74 noro 4659: FILE *nd_write,*nd_read;
4660:
4661: void nd_send_int(int a) {
4662: write_int(nd_write,&a);
4663: }
4664:
4665: void nd_send_intarray(int *p,int len) {
4666: write_intarray(nd_write,p,len);
4667: }
4668:
4669: int nd_recv_int() {
4670: int a;
4671:
4672: read_int(nd_read,&a);
4673: return a;
4674: }
4675:
4676: void nd_recv_intarray(int *p,int len) {
4677: read_intarray(nd_read,p,len);
4678: }
4679:
4680: void nd_send_ndv(NDV p) {
4681: int len,i;
4682: NMV m;
4683:
4684: if ( !p ) nd_send_int(0);
4685: else {
4686: len = LEN(p);
4687: nd_send_int(len);
4688: m = BDY(p);
4689: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4690: nd_send_int(CM(m));
4691: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4692: }
1.74 noro 4693: }
4694: }
4695:
4696: void nd_send_nd(ND p) {
4697: int len,i;
4698: NM m;
4699:
4700: if ( !p ) nd_send_int(0);
4701: else {
4702: len = LEN(p);
4703: nd_send_int(len);
4704: m = BDY(p);
4705: for ( i = 0; i < len; i++, m = NEXT(m) ) {
4706: nd_send_int(CM(m));
4707: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4708: }
1.74 noro 4709: }
4710: }
1.65 noro 4711:
1.74 noro 4712: NDV nd_recv_ndv()
4713: {
4714: int len,i;
4715: NMV m,m0;
4716: NDV r;
1.65 noro 4717:
1.74 noro 4718: len = nd_recv_int();
4719: if ( !len ) return 0;
4720: else {
4721: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4722: #if 0
1.74 noro 4723: ndv_alloc += len*nmv_adv;
1.103 noro 4724: #endif
1.74 noro 4725: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4726: CM(m) = nd_recv_int();
4727: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4728: }
1.74 noro 4729: MKNDV(nd_nvar,m0,len,r);
4730: return r;
4731: }
4732: }
1.65 noro 4733:
1.74 noro 4734: int ox_exec_f4_red(Q proc)
4735: {
4736: Obj obj;
4737: STRING fname;
4738: NODE arg;
4739: int s;
4740: extern int ox_need_conv,ox_file_io;
4741:
4742: MKSTR(fname,"nd_exec_f4_red");
4743: arg = mknode(2,proc,fname);
4744: Pox_cmo_rpc(arg,&obj);
4745: s = get_ox_server_id(QTOS(proc));
4746: nd_write = iofp[s].out;
4747: nd_read = iofp[s].in;
4748: ox_need_conv = ox_file_io = 0;
4749: return s;
4750: }
4751:
4752: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
4753: {
4754: int nsp,nred;
4755: int i,rank,s;
4756: NODE rp,r0,r;
4757: ND_pairs sp;
4758: NM_ind_pair pair;
4759: NMV nmv;
4760: NM nm;
4761: NDV nf;
4762: Obj proc,dmy;
4763:
4764: ox_launch_main(0,0,&proc);
4765: s = ox_exec_f4_red((Q)proc);
4766:
4767: nd_send_int(m);
4768: nd_send_int(nd_nvar);
4769: nd_send_int(nd_bpe);
4770: nd_send_int(nd_wpd);
4771: nd_send_int(nmv_adv);
4772:
1.89 noro 4773: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
1.74 noro 4774:
4775: nd_send_int(nd_psn);
4776: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
4777:
4778: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4779: nd_send_int(nsp);
4780: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
4781: nd_send_int(sp->i1); nd_send_int(sp->i2);
4782: }
4783:
4784: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
4785:
4786: nred = length(rp0); nd_send_int(nred);
4787: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
4788: pair = (NM_ind_pair)BDY(rp);
4789: nd_send_int(pair->index);
4790: nd_send_intarray(pair->mul->dl,nd_wpd);
4791: }
4792: fflush(nd_write);
4793: rank = nd_recv_int();
4794: fprintf(asir_out,"rank=%d\n",rank);
4795: r0 = 0;
4796: for ( i = 0; i < rank; i++ ) {
4797: nf = nd_recv_ndv();
4798: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
4799: }
4800: Pox_shutdown(mknode(1,proc),&dmy);
4801: return r0;
4802: }
4803:
4804: /* server side */
4805:
4806: void nd_exec_f4_red_dist()
4807: {
4808: int m,i,nsp,col,s0size,nred,spcol,j,k;
4809: NM_ind_pair *rp0;
4810: NDV nf;
4811: UINT *s0vect;
4812: IndArray *imat;
4813: int *rhead;
4814: int **spmat;
4815: UINT *svect,*v;
4816: ND_pairs *sp0;
4817: int *colstat;
4818: int a,sprow,rank;
1.89 noro 4819: struct order_spec *ord;
1.74 noro 4820: Obj ordspec;
4821: ND spol;
1.76 noro 4822: int maxrs;
4823: int *spsugar;
1.74 noro 4824:
4825: nd_read = iofp[0].in;
4826: nd_write = iofp[0].out;
4827: m = nd_recv_int();
4828: nd_nvar = nd_recv_int();
4829: nd_bpe = nd_recv_int();
4830: nd_wpd = nd_recv_int();
4831: nmv_adv = nd_recv_int();
4832:
4833: loadobj(nd_read,&ordspec);
1.89 noro 4834: create_order_spec(0,ordspec,&ord);
4835: nd_init_ord(ord);
1.74 noro 4836: nd_setup_parameters(nd_nvar,0);
4837:
4838: nd_psn = nd_recv_int();
4839: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
4840: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
4841: for ( i = 0; i < nd_psn; i++ ) {
4842: nd_ps[i] = nd_recv_ndv();
4843: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
4844: }
4845:
4846: nsp = nd_recv_int();
4847: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
4848: for ( i = 0; i < nsp; i++ ) {
4849: NEWND_pairs(sp0[i]);
4850: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
4851: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
4852: }
4853:
4854: col = nd_recv_int();
4855: s0size = col*nd_wpd;
4856: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
4857: nd_recv_intarray(s0vect,s0size);
4858:
4859: nred = nd_recv_int();
4860: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
4861: for ( i = 0; i < nred; i++ ) {
4862: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
4863: rp0[i]->index = nd_recv_int();
4864: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
4865: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
4866: }
4867:
4868: spcol = col-nred;
4869: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
4870: rhead = (int *)MALLOC(col*sizeof(int));
4871: for ( i = 0; i < col; i++ ) rhead[i] = 0;
4872:
4873: /* construction of index arrays */
4874: for ( i = 0; i < nred; i++ ) {
4875: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
4876: rhead[imat[i]->head] = 1;
4877: }
4878:
4879: /* elimination (1st step) */
4880: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
4881: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 4882: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4883: for ( a = sprow = 0; a < nsp; a++ ) {
4884: nd_sp(m,0,sp0[a],&spol);
1.75 noro 4885: if ( !spol ) continue;
1.74 noro 4886: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4887: if ( m == -1 )
4888: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
4889: else
4890: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 4891: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4892: if ( i < col ) {
4893: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
4894: for ( j = k = 0; j < col; j++ )
4895: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4896: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4897: sprow++;
4898: }
1.76 noro 4899: nd_free(spol);
1.74 noro 4900: }
4901: /* elimination (2nd step) */
4902: colstat = (int *)ALLOCA(spcol*sizeof(int));
4903: if ( m == -1 )
1.76 noro 4904: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4905: else
1.76 noro 4906: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4907: nd_send_int(rank);
4908: for ( i = 0; i < rank; i++ ) {
4909: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
4910: nd_send_ndv(nf);
1.63 noro 4911: }
1.74 noro 4912: fflush(nd_write);
1.107 noro 4913: }
4914:
1.113 noro 4915: int nd_gauss_elim_q(Q **mat0,int *sugar,int row,int col,int *colstat)
1.107 noro 4916: {
1.109 noro 4917: int mod,i,j,t,c,rank,rank0,inv;
4918: int *ci,*ri;
1.113 noro 4919: Q dn;
1.109 noro 4920: MAT m,nm;
1.108 noro 4921: int **wmat;
4922:
4923: /* XXX */
4924: mod = 99999989;
4925: wmat = (int **)ALLOCA(row*sizeof(int *));
4926: for ( i = 0; i < row; i++ ) {
4927: wmat[i] = (int *)ALLOCA(col*sizeof(int));
4928: for ( j = 0; j < col; j++ ) {
1.113 noro 4929: if ( mat0[i][j] ) {
4930: t = rem(NM(mat0[i][j]),mod);
4931: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
4932: wmat[i][j] = t;
4933: } else
1.108 noro 4934: wmat[i][j] = 0;
4935: }
4936: }
1.109 noro 4937: rank0 = nd_gauss_elim_mod(wmat,sugar,row,col,mod,colstat);
4938: NEWMAT(m); m->row = row; m->col = col; m->body = (pointer **)mat0;
1.113 noro 4939: rank = generic_gauss_elim(m,&nm,&dn,&ri,&ci);
1.109 noro 4940: if ( rank != rank0 )
4941: error("afo");
4942: for ( i = 0; i < row; i++ )
4943: for ( j = 0; j < col; j++ )
4944: mat0[i][j] = 0;
4945: c = col-rank;
4946: for ( i = 0; i < rank; i++ ) {
1.113 noro 4947: mat0[i][ri[i]] = dn;
1.109 noro 4948: for ( j = 0; j < c; j++ )
1.113 noro 4949: mat0[i][ci[j]] = (Q)BDY(nm)[i][j];
1.109 noro 4950: }
1.113 noro 4951: inv = invm(rem(NM(dn),mod),mod);
4952: if ( SGN(dn) < 0 ) inv = mod-inv;
1.109 noro 4953: for ( i = 0; i < row; i++ )
4954: for ( j = 0; j < col; j++ ) {
1.113 noro 4955: if ( mat0[i][j] ) {
4956: t = rem(NM(mat0[i][j]),mod);
4957: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
4958: } else
1.109 noro 4959: t = 0;
4960: c = dmar(t,inv,0,mod);
4961: if ( wmat[i][j] != c )
4962: error("afo");
4963: }
4964: return rank;
1.76 noro 4965: }
4966:
4967: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4968: {
4969: int i,j,k,l,inv,a,rank,s;
4970: unsigned int *t,*pivot,*pk;
4971: unsigned int **mat;
4972:
4973: mat = (unsigned int **)mat0;
4974: for ( rank = 0, j = 0; j < col; j++ ) {
4975: for ( i = rank; i < row; i++ )
4976: mat[i][j] %= md;
4977: for ( i = rank; i < row; i++ )
4978: if ( mat[i][j] )
4979: break;
4980: if ( i == row ) {
4981: colstat[j] = 0;
4982: continue;
4983: } else
4984: colstat[j] = 1;
4985: if ( i != rank ) {
4986: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4987: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4988: }
4989: pivot = mat[rank];
4990: s = sugar[rank];
4991: inv = invm(pivot[j],md);
4992: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4993: if ( *pk ) {
4994: if ( *pk >= (unsigned int)md )
4995: *pk %= md;
4996: DMAR(*pk,inv,0,md,*pk)
4997: }
4998: for ( i = rank+1; i < row; i++ ) {
4999: t = mat[i];
5000: if ( a = t[j] ) {
5001: sugar[i] = MAX(sugar[i],s);
5002: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5003: }
5004: }
5005: rank++;
5006: }
5007: for ( j = col-1, l = rank-1; j >= 0; j-- )
5008: if ( colstat[j] ) {
5009: pivot = mat[l];
5010: s = sugar[l];
5011: for ( i = 0; i < l; i++ ) {
5012: t = mat[i];
5013: t[j] %= md;
5014: if ( a = t[j] ) {
5015: sugar[i] = MAX(sugar[i],s);
5016: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5017: }
5018: }
5019: l--;
5020: }
5021: for ( j = 0, l = 0; l < rank; j++ )
5022: if ( colstat[j] ) {
5023: t = mat[l];
5024: for ( k = j; k < col; k++ )
5025: if ( t[k] >= (unsigned int)md )
5026: t[k] %= md;
5027: l++;
5028: }
5029: return rank;
5030: }
5031:
5032: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
5033: {
5034: int i,j,k,l,inv,a,rank,s;
5035: unsigned int *t,*pivot,*pk;
5036: unsigned int **mat;
5037:
5038: mat = (unsigned int **)mat0;
5039: for ( rank = 0, j = 0; j < col; j++ ) {
5040: for ( i = rank; i < row; i++ )
5041: if ( mat[i][j] )
5042: break;
5043: if ( i == row ) {
5044: colstat[j] = 0;
5045: continue;
5046: } else
5047: colstat[j] = 1;
5048: if ( i != rank ) {
5049: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5050: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
5051: }
5052: pivot = mat[rank];
5053: s = sugar[rank];
5054: inv = _invsf(pivot[j]);
5055: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5056: if ( *pk )
5057: *pk = _mulsf(*pk,inv);
5058: for ( i = rank+1; i < row; i++ ) {
5059: t = mat[i];
5060: if ( a = t[j] ) {
5061: sugar[i] = MAX(sugar[i],s);
5062: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5063: }
5064: }
5065: rank++;
5066: }
5067: for ( j = col-1, l = rank-1; j >= 0; j-- )
5068: if ( colstat[j] ) {
5069: pivot = mat[l];
5070: s = sugar[l];
5071: for ( i = 0; i < l; i++ ) {
5072: t = mat[i];
5073: if ( a = t[j] ) {
5074: sugar[i] = MAX(sugar[i],s);
5075: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5076: }
5077: }
5078: l--;
5079: }
5080: return rank;
1.77 noro 5081: }
5082:
5083: int ndv_ishomo(NDV p)
5084: {
5085: NMV m;
5086: int len,h;
5087:
5088: if ( !p ) return 1;
5089: len = LEN(p);
5090: m = BDY(p);
5091: h = TD(DL(m));
5092: NMV_ADV(m);
5093: for ( len--; len; len--, NMV_ADV(m) )
5094: if ( TD(DL(m)) != h ) return 0;
5095: return 1;
5096: }
5097:
5098: void ndv_save(NDV p,int index)
5099: {
5100: FILE *s;
5101: char name[BUFSIZ];
5102: short id;
5103: int nv,sugar,len,n,i,td,e,j;
5104: NMV m;
5105: unsigned int *dl;
5106:
5107: sprintf(name,"%s/%d",Demand,index);
5108: s = fopen(name,"w");
5109: savevl(s,0);
5110: if ( !p ) {
5111: saveobj(s,0);
5112: return;
5113: }
5114: id = O_DP;
5115: nv = NV(p);
5116: sugar = SG(p);
5117: len = LEN(p);
5118: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
5119: write_int(s,&len);
5120:
5121: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5122: saveobj(s,(Obj)CQ(m));
1.77 noro 5123: dl = DL(m);
5124: td = TD(dl);
5125: write_int(s,&td);
5126: for ( j = 0; j < nv; j++ ) {
5127: e = GET_EXP(dl,j);
5128: write_int(s,&e);
5129: }
5130: }
5131: fclose(s);
5132: }
5133:
5134: NDV ndv_load(int index)
5135: {
5136: FILE *s;
5137: char name[BUFSIZ];
5138: short id;
5139: int nv,sugar,len,n,i,td,e,j;
5140: NDV d;
5141: NMV m0,m;
5142: unsigned int *dl;
5143: Obj obj;
5144:
5145: sprintf(name,"%s/%d",Demand,index);
5146: s = fopen(name,"r");
5147: if ( !s ) return 0;
5148:
5149: skipvl(s);
5150: read_short(s,&id);
5151: if ( !id ) return 0;
5152: read_int(s,&nv);
5153: read_int(s,&sugar);
5154: read_int(s,&len);
5155:
5156: m0 = m = MALLOC(len*nmv_adv);
5157: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5158: loadobj(s,&obj); CQ(m) = (Q)obj;
1.77 noro 5159: dl = DL(m);
5160: ndl_zero(dl);
5161: read_int(s,&td); TD(dl) = td;
5162: for ( j = 0; j < nv; j++ ) {
5163: read_int(s,&e);
5164: PUT_EXP(dl,j,e);
5165: }
5166: if ( nd_blockmask ) ndl_weight_mask(dl);
5167: }
5168: fclose(s);
5169: MKNDV(nv,m0,len,d);
5170: SG(d) = sugar;
5171: return d;
1.99 noro 5172: }
5173:
1.102 noro 5174: void nd_det(int mod,MAT f,P *rp)
1.99 noro 5175: {
5176: VL fv,tv;
1.102 noro 5177: int n,i,j,max,e,nvar,sgn,k0,l0,len0,len,k,l,a;
1.99 noro 5178: pointer **m;
1.113 noro 5179: Q mone;
1.125 noro 5180: P **w;
1.126 noro 5181: P mp,r;
1.102 noro 5182: NDV **dm;
5183: NDV *t,*mi,*mj;
5184: NDV d,s,mij,mjj;
5185: ND u;
5186: NMV nmv;
1.114 noro 5187: UINT *bound;
1.99 noro 5188: PGeoBucket bucket;
5189: struct order_spec *ord;
1.126 noro 5190: Q dq,dt,ds;
5191: N gn,qn,dn0,nm,dn;
1.99 noro 5192:
5193: create_order_spec(0,0,&ord);
5194: nd_init_ord(ord);
5195: get_vars((Obj)f,&fv);
5196: if ( f->row != f->col )
5197: error("nd_det : non-square matrix");
5198: n = f->row;
1.125 noro 5199: m = f->body;
1.99 noro 5200: for ( nvar = 0, tv = fv; tv; tv = NEXT(tv), nvar++ );
1.125 noro 5201:
5202: if ( !nvar ) {
5203: if ( !mod )
5204: detp(CO,(P **)m,n,rp);
5205: else {
5206: w = (P **)almat_pointer(n,n);
5207: for ( i = 0; i < n; i++ )
5208: for ( j = 0; j < n; j++ )
5209: ptomp(mod,(P)m[i][j],&w[i][j]);
5210: detmp(CO,mod,w,n,&mp);
5211: mptop(mp,rp);
5212: }
5213: return;
5214: }
1.126 noro 5215:
5216: if ( !mod ) {
5217: w = (P **)almat_pointer(n,n);
5218: dq = ONE;
5219: for ( i = 0; i < n; i++ ) {
5220: dn0 = ONEN;
5221: for ( j = 0; j < n; j++ ) {
5222: if ( !m[i][j] ) continue;
5223: lgp(m[i][j],&nm,&dn);
5224: gcdn(dn0,dn,&gn); divsn(dn0,gn,&qn); muln(qn,dn,&dn0);
5225: }
5226: if ( !UNIN(dn0) ) {
5227: NTOQ(dn0,1,ds);
5228: for ( j = 0; j < n; j++ )
5229: mulp(CO,(P)m[i][j],(P)ds,&w[i][j]);
5230: mulq(dq,ds,&dt); dq = dt;
5231: } else
5232: for ( j = 0; j < n; j++ )
5233: w[i][j] = (P)m[i][j];
5234: }
5235: m = (pointer **)w;
5236: }
1.125 noro 5237:
1.99 noro 5238: for ( i = 0, max = 0; i < n; i++ )
5239: for ( j = 0; j < n; j++ )
5240: for ( tv = fv; tv; tv = NEXT(tv) ) {
5241: e = getdeg(tv->v,(P)m[i][j]);
5242: max = MAX(e,max);
5243: }
1.114 noro 5244: nd_setup_parameters(nvar,max);
1.102 noro 5245: dm = (NDV **)almat_pointer(n,n);
1.99 noro 5246: for ( i = 0, max = 0; i < n; i++ )
1.102 noro 5247: for ( j = 0; j < n; j++ ) {
5248: dm[i][j] = ptondv(CO,fv,m[i][j]);
5249: if ( mod ) ndv_mod(mod,dm[i][j]);
5250: if ( dm[i][j] && !LEN(dm[i][j]) ) dm[i][j] = 0;
5251: }
5252: d = ptondv(CO,fv,(P)ONE);
5253: if ( mod ) ndv_mod(mod,d);
1.113 noro 5254: chsgnq(ONE,&mone);
1.99 noro 5255: for ( j = 0, sgn = 1; j < n; j++ ) {
1.125 noro 5256: if ( DP_Print ) fprintf(stderr,".",j);
1.99 noro 5257: for ( i = j; i < n && !dm[i][j]; i++ );
5258: if ( i == n ) {
5259: *rp = 0;
5260: return;
5261: }
1.102 noro 5262: k0 = i; l0 = j; len0 = LEN(dm[k0][l0]);
1.99 noro 5263: for ( k = j; k < n; k++ )
5264: for ( l = j; l < n; l++ )
1.102 noro 5265: if ( dm[k][l] && LEN(dm[k][l]) < len0 ) {
5266: k0 = k; l0 = l; len0 = LEN(dm[k][l]);
1.99 noro 5267: }
5268: if ( k0 != j ) {
5269: t = dm[j]; dm[j] = dm[k0]; dm[k0] = t;
5270: sgn = -sgn;
5271: }
5272: if ( l0 != j ) {
5273: for ( k = j; k < n; k++ ) {
5274: s = dm[k][j]; dm[k][j] = dm[k][l0]; dm[k][l0] = s;
5275: }
5276: sgn = -sgn;
5277: }
1.114 noro 5278: bound = nd_det_compute_bound(dm,n,j);
5279: if ( ndl_check_bound(bound,bound) )
5280: nd_det_reconstruct(dm,n,j,d);
5281:
1.99 noro 5282: for ( i = j+1, mj = dm[j], mjj = mj[j]; i < n; i++ ) {
1.114 noro 5283: /* if ( DP_Print ) fprintf(stderr," i=%d\n ",i); */
1.99 noro 5284: mi = dm[i]; mij = mi[j];
1.102 noro 5285: if ( mod )
5286: ndv_mul_c(mod,mij,mod-1);
5287: else
1.113 noro 5288: ndv_mul_c_q(mij,mone);
1.99 noro 5289: for ( k = j+1; k < n; k++ ) {
1.114 noro 5290: /* if ( DP_Print ) fprintf(stderr,"k=%d ",k); */
1.99 noro 5291: bucket = create_pbucket();
1.104 noro 5292: if ( mi[k] ) {
1.102 noro 5293: nmv = BDY(mjj); len = LEN(mjj);
5294: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5295: u = ndv_mul_nmv_trunc(mod,nmv,mi[k],DL(BDY(d)));
5296: add_pbucket(mod,bucket,u);
1.99 noro 5297: }
1.104 noro 5298: }
1.99 noro 5299: if ( mj[k] && mij ) {
1.102 noro 5300: nmv = BDY(mij); len = LEN(mij);
5301: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5302: u = ndv_mul_nmv_trunc(mod,nmv,mj[k],DL(BDY(d)));
5303: add_pbucket(mod,bucket,u);
1.99 noro 5304: }
5305: }
1.104 noro 5306: u = nd_quo(mod,bucket,d);
1.102 noro 5307: mi[k] = ndtondv(mod,u);
1.99 noro 5308: }
1.114 noro 5309: /* if ( DP_Print ) fprintf(stderr,"\n",k); */
1.99 noro 5310: }
5311: d = mjj;
5312: }
1.125 noro 5313: if ( DP_Print ) fprintf(stderr,"\n",k);
1.100 noro 5314: if ( sgn < 0 )
1.102 noro 5315: if ( mod )
5316: ndv_mul_c(mod,d,mod-1);
5317: else
1.113 noro 5318: ndv_mul_c_q(d,mone);
1.126 noro 5319: r = ndvtop(mod,CO,fv,d);
5320: if ( !mod && !UNIQ(dq) )
5321: divsp(CO,r,(P)dq,rp);
5322: else
5323: *rp = r;
1.99 noro 5324: }
5325:
1.102 noro 5326: ND ndv_mul_nmv_trunc(int mod,NMV m0,NDV p,UINT *d)
1.99 noro 5327: {
5328: NM mr,mr0;
1.102 noro 5329: NM tnm;
5330: NMV m;
1.99 noro 5331: UINT *d0,*dt,*dm;
5332: int c,n,td,i,c1,c2,len;
1.113 noro 5333: Q q;
1.99 noro 5334: ND r;
5335:
5336: if ( !p ) return 0;
5337: else {
1.102 noro 5338: n = NV(p); m = BDY(p); len = LEN(p);
1.99 noro 5339: d0 = DL(m0);
1.102 noro 5340: td = TD(d);
1.99 noro 5341: mr0 = 0;
5342: NEWNM(tnm);
1.102 noro 5343: if ( mod ) {
5344: c = CM(m0);
5345: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5346: ndl_add(DL(m),d0,DL(tnm));
5347: if ( ndl_reducible(DL(tnm),d) ) {
5348: NEXTNM(mr0,mr);
5349: c1 = CM(m); DMAR(c1,c,0,mod,c2); CM(mr) = c2;
1.104 noro 5350: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5351: }
5352: }
5353: } else {
1.113 noro 5354: q = CQ(m0);
1.102 noro 5355: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5356: ndl_add(DL(m),d0,DL(tnm));
5357: if ( ndl_reducible(DL(tnm),d) ) {
5358: NEXTNM(mr0,mr);
1.113 noro 5359: mulq(CQ(m),q,&CQ(mr));
1.104 noro 5360: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5361: }
1.99 noro 5362: }
5363: }
5364: if ( !mr0 )
5365: return 0;
5366: else {
5367: NEXT(mr) = 0;
1.105 noro 5368: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
1.99 noro 5369: MKND(NV(p),mr0,len,r);
5370: SG(r) = SG(p) + TD(d0);
5371: return r;
5372: }
5373: }
1.114 noro 5374: }
5375:
5376: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d)
5377: {
5378: int i,obpe,oadv,h,k,l;
5379: static NM prev_nm_free_list;
5380: EPOS oepos;
5381:
5382: obpe = nd_bpe;
5383: oadv = nmv_adv;
5384: oepos = nd_epos;
5385: if ( obpe < 2 ) nd_bpe = 2;
5386: else if ( obpe < 3 ) nd_bpe = 3;
5387: else if ( obpe < 4 ) nd_bpe = 4;
5388: else if ( obpe < 5 ) nd_bpe = 5;
5389: else if ( obpe < 6 ) nd_bpe = 6;
5390: else if ( obpe < 8 ) nd_bpe = 8;
5391: else if ( obpe < 10 ) nd_bpe = 10;
5392: else if ( obpe < 16 ) nd_bpe = 16;
5393: else if ( obpe < 32 ) nd_bpe = 32;
5394: else error("nd_det_reconstruct : exponent too large");
5395:
5396: nd_setup_parameters(nd_nvar,0);
5397: prev_nm_free_list = _nm_free_list;
5398: _nm_free_list = 0;
5399: for ( k = j; k < n; k++ )
5400: for (l = j; l < n; l++ )
5401: ndv_realloc(dm[k][l],obpe,oadv,oepos);
5402: ndv_realloc(d,obpe,oadv,oepos);
5403: prev_nm_free_list = 0;
5404: #if 0
5405: GC_gcollect();
5406: #endif
5407: }
5408:
5409: UINT *nd_det_compute_bound(NDV **dm,int n,int j)
5410: {
5411: UINT *d0,*d1,*d,*t,*r;
5412: int k,l;
5413:
5414: d0 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5415: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5416: for ( k = 0; k < nd_wpd; k++ ) d0[k] = 0;
5417: for ( k = j; k < n; k++ )
5418: for ( l = j; l < n; l++ )
5419: if ( dm[k][l] ) {
5420: d = ndv_compute_bound(dm[k][l]);
5421: ndl_lcm(d,d0,d1);
5422: t = d1; d1 = d0; d0 = t;
5423: }
5424: r = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5425: for ( k = 0; k < nd_wpd; k++ ) r[k] = d0[k];
5426: return r;
1.117 noro 5427: }
5428:
5429: DL nd_separate_d(UINT *d,UINT *trans)
5430: {
1.121 noro 5431: int n,td,i,e,j;
1.117 noro 5432: DL a;
5433:
5434: ndl_zero(trans);
5435: td = 0;
1.121 noro 5436: for ( i = 0; i < nd_ntrans; i++ ) {
5437: e = GET_EXP(d,i);
5438: PUT_EXP(trans,i,e);
5439: td += MUL_WEIGHT(e,i);
5440: }
5441: if ( nd_ntrans+nd_nalg < nd_nvar ) {
5442: /* homogenized */
5443: i = nd_nvar-1;
1.117 noro 5444: e = GET_EXP(d,i);
5445: PUT_EXP(trans,i,e);
5446: td += MUL_WEIGHT(e,i);
5447: }
5448: TD(trans) = td;
5449: if ( nd_blockmask) ndl_weight_mask(trans);
5450: NEWDL(a,nd_nalg);
5451: td = 0;
1.121 noro 5452: for ( i = 0; i < nd_nalg; i++ ) {
5453: j = nd_ntrans+i;
5454: e = GET_EXP(d,j);
5455: a->d[i] = e;
1.117 noro 5456: td += e;
5457: }
5458: a->td = td;
5459: return a;
5460: }
5461:
1.118 noro 5462: int nd_monic(int mod,ND *p)
1.117 noro 5463: {
5464: UINT *trans,*t;
5465: DL alg;
5466: MP mp0,mp;
1.118 noro 5467: NM m,m0,m1,ma0,ma,mb,mr0,mr;
5468: ND r;
1.117 noro 5469: DL dl;
5470: DP nm;
5471: NDV ndv;
1.118 noro 5472: DAlg inv,cd;
1.117 noro 5473: ND s,c;
1.118 noro 5474: Q l,mul;
5475: N ln;
5476: int n,ntrans,i,e,td,is_lc,len;
5477: NumberField nf;
5478: struct oEGT eg0,eg1;
5479:
5480: if ( !(nf = get_numberfield()) )
5481: error("nd_monic : current_numberfield is not set");
1.117 noro 5482:
1.118 noro 5483: /* Q coef -> DAlg coef */
5484: NEWNM(ma0); ma = ma0;
5485: m = BDY(*p);
5486: is_lc = 1;
5487: while ( 1 ) {
5488: NEWMP(mp0); mp = mp0;
5489: mp->c = (P)CQ(m);
5490: mp->dl = nd_separate_d(DL(m),DL(ma));
5491: NEWNM(mb);
5492: for ( m = NEXT(m); m; m = NEXT(m) ) {
5493: alg = nd_separate_d(DL(m),DL(mb));
5494: if ( !ndl_equal(DL(ma),DL(mb)) )
5495: break;
5496: NEXTMP(mp0,mp); mp->c = (P)CQ(m); mp->dl = alg;
5497: }
5498: NEXT(mp) = 0;
5499: MKDP(nd_nalg,mp0,nm);
5500: MKDAlg(nm,ONE,cd);
5501: if ( is_lc == 1 ) {
5502: /* if the lc is a rational number, we have nothing to do */
5503: if ( !mp0->dl->td )
5504: return 1;
5505:
5506: get_eg(&eg0);
5507: invdalg(cd,&inv);
5508: get_eg(&eg1); add_eg(&eg_invdalg,&eg0,&eg1);
5509: /* check the validity of inv */
5510: if ( mod && !rem(NM(inv->dn),mod) )
5511: return 0;
5512: CA(ma) = nf->one;
5513: is_lc = 0;
5514: ln = ONEN;
5515: } else {
5516: muldalg(cd,inv,&CA(ma));
5517: lcmn(ln,NM(CA(ma)->dn),&ln);
5518: }
5519: if ( m ) {
5520: NEXT(ma) = mb; ma = mb;
5521: } else {
5522: NEXT(ma) = 0;
5523: break;
5524: }
5525: }
5526: /* l = lcm(denoms) */
5527: NTOQ(ln,1,l);
5528: for ( mr0 = 0, m = ma0; m; m = NEXT(m) ) {
5529: divq(l,CA(m)->dn,&mul);
5530: for ( mp = BDY(CA(m)->nm); mp; mp = NEXT(mp) ) {
5531: NEXTNM(mr0,mr);
5532: mulq((Q)mp->c,mul,&CQ(mr));
5533: dl = mp->dl;
5534: td = TD(DL(m));
5535: ndl_copy(DL(m),DL(mr));
1.121 noro 5536: for ( i = 0; i < nd_nalg; i++ ) {
5537: e = dl->d[i];
5538: PUT_EXP(DL(mr),i+nd_ntrans,e);
5539: td += MUL_WEIGHT(e,i+nd_ntrans);
1.118 noro 5540: }
5541: TD(DL(mr)) = td;
5542: if ( nd_blockmask) ndl_weight_mask(DL(mr));
5543: }
1.117 noro 5544: }
1.118 noro 5545: NEXT(mr) = 0;
5546: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
5547: MKND(NV(*p),mr0,len,r);
5548: /* XXX */
5549: SG(r) = SG(*p);
5550: nd_free(*p);
5551: *p = r;
5552: return 1;
1.59 noro 5553: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>