Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.121
1.121 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.120 2004/12/07 15:15:52 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:
1.92 noro 1540: void do_diagonalize(int sugar,int m)
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);
1556: nd_nf(m,s,nd_ps,1,&dn,&nf);
1557: if ( !m ) {
1.113 noro 1558: NTOQ(NM(dn.z),SGN(dn.z),num);
1559: mulq(HCQ(head),num,&q); HCQ(head) = q;
1560: if ( DN(dn.z) ) {
1561: NTOQ(DN(dn.z),1,den);
1562: nd_mul_c_q(nf,den);
1563: }
1.92 noro 1564: }
1565: nf = nd_add(m,head,nf);
1566: ndv_free(nfv);
1567: nd_removecont(m,nf);
1568: nfv = ndtondv(m,nf);
1569: nd_free(nf);
1570: nd_bound[i] = ndv_compute_bound(nfv);
1571: if ( nd_demand ) {
1572: ndv_save(nfv,i);
1573: ndv_free(nfv);
1574: } else
1575: nd_ps[i] = nfv;
1576: }
1577: }
1578:
1.27 noro 1579: /* return value = 0 => input is not a GB */
1580:
1.77 noro 1581: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1582: {
1583: int i,nh,sugar,stat;
1.23 noro 1584: NODE r,g,t;
1.1 noro 1585: ND_pairs d;
1586: ND_pairs l;
1.117 noro 1587: ND h,nf,s,head,nf1;
1.63 noro 1588: NDV nfv;
1.113 noro 1589: Q q,num,den;
1590: union oNDC dn;
1.1 noro 1591:
1.23 noro 1592: g = 0; d = 0;
1593: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1594: d = update_pairs(d,g,i);
1595: g = update_base(g,i);
1596: }
1597: sugar = 0;
1598: while ( d ) {
1599: again:
1600: l = nd_minp(d,&d);
1.14 noro 1601: if ( SG(l) != sugar ) {
1.92 noro 1602: if ( ishomo ) do_diagonalize(sugar,m);
1603:
1.14 noro 1604: sugar = SG(l);
1.71 noro 1605: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1606: }
1.53 noro 1607: stat = nd_sp(m,0,l,&h);
1.1 noro 1608: if ( !stat ) {
1609: NEXT(l) = d; d = l;
1.103 noro 1610: d = nd_reconstruct(0,d);
1.1 noro 1611: goto again;
1612: }
1.41 noro 1613: #if USE_GEOBUCKET
1.69 noro 1614: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1615: #else
1.69 noro 1616: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1617: #endif
1.1 noro 1618: if ( !stat ) {
1619: NEXT(l) = d; d = l;
1.103 noro 1620: d = nd_reconstruct(0,d);
1.1 noro 1621: goto again;
1622: } else if ( nf ) {
1.27 noro 1623: if ( checkonly ) return 0;
1.71 noro 1624: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1625: nd_removecont(m,nf);
1.118 noro 1626: if ( !m && nd_nalg ) {
1627: nd_monic(0,&nf);
1628: nd_removecont(m,nf);
1.117 noro 1629: }
1.63 noro 1630: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1631: nh = ndv_newps(m,nfv,0);
1.1 noro 1632: d = update_pairs(d,g,nh);
1633: g = update_base(g,nh);
1634: FREENDP(l);
1635: } else {
1.71 noro 1636: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1637: FREENDP(l);
1638: }
1639: }
1.77 noro 1640: if ( nd_demand )
1641: for ( t = g; t; t = NEXT(t) )
1642: BDY(t) = (pointer)ndv_load((int)BDY(t));
1643: else
1644: for ( t = g; t; t = NEXT(t) )
1645: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1646: return g;
1647: }
1648:
1.92 noro 1649: void do_diagonalize_trace(int sugar,int m)
1.91 noro 1650: {
1651: int i,nh,stat;
1652: NODE r,g,t;
1653: ND h,nf,nfq,s,head;
1654: NDV nfv,nfqv;
1.113 noro 1655: Q q,den,num;
1656: union oNDC dn;
1.91 noro 1657:
1658: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1659: /* for nd_ps */
1660: s = ndvtond(m,nd_ps[i]);
1661: s = nd_separate_head(s,&head);
1662: nd_nf_pbucket(m,s,nd_ps,1,&nf);
1663: nf = nd_add(m,head,nf);
1664: ndv_free(nd_ps[i]);
1665: nd_ps[i] = ndtondv(m,nf);
1666: nd_free(nf);
1667:
1668: /* for nd_ps_trace */
1669: if ( nd_demand )
1670: nfv = ndv_load(i);
1671: else
1672: nfv = nd_ps_trace[i];
1673: s = ndvtond(0,nfv);
1674: s = nd_separate_head(s,&head);
1675: nd_nf(0,s,nd_ps_trace,1,&dn,&nf);
1.113 noro 1676: NTOQ(NM(dn.z),SGN(dn.z),num);
1677: mulq(HCQ(head),num,&q); HCQ(head) = q;
1678: if ( DN(dn.z) ) {
1679: NTOQ(DN(dn.z),1,den);
1680: nd_mul_c_q(nf,den);
1681: }
1.91 noro 1682: nf = nd_add(0,head,nf);
1683: ndv_free(nfv);
1684: nd_removecont(0,nf);
1685: nfv = ndtondv(0,nf);
1686: nd_free(nf);
1687: nd_bound[i] = ndv_compute_bound(nfv);
1688: if ( nd_demand ) {
1689: ndv_save(nfv,i);
1690: ndv_free(nfv);
1691: } else
1692: nd_ps_trace[i] = nfv;
1693: }
1694: }
1695:
1.118 noro 1696: static struct oEGT eg_invdalg;
1697: struct oEGT eg_le;
1698:
1.82 noro 1699: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 1700: {
1701: int i,nh,sugar,stat;
1.23 noro 1702: NODE r,g,t;
1.20 noro 1703: ND_pairs d;
1704: ND_pairs l;
1.82 noro 1705: ND h,nf,nfq,s,head;
1.63 noro 1706: NDV nfv,nfqv;
1.113 noro 1707: Q q,den,num;
1708: union oNDC dn;
1.118 noro 1709: struct oEGT eg_monic,egm0,egm1;
1.20 noro 1710:
1.118 noro 1711: init_eg(&eg_monic);
1712: init_eg(&eg_invdalg);
1713: init_eg(&eg_le);
1.23 noro 1714: g = 0; d = 0;
1715: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1716: d = update_pairs(d,g,i);
1717: g = update_base(g,i);
1718: }
1719: sugar = 0;
1720: while ( d ) {
1721: again:
1722: l = nd_minp(d,&d);
1723: if ( SG(l) != sugar ) {
1.93 noro 1724: if ( ishomo ) do_diagonalize_trace(sugar,m);
1.20 noro 1725: sugar = SG(l);
1.71 noro 1726: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1727: }
1.53 noro 1728: stat = nd_sp(m,0,l,&h);
1.20 noro 1729: if ( !stat ) {
1730: NEXT(l) = d; d = l;
1.103 noro 1731: d = nd_reconstruct(1,d);
1.20 noro 1732: goto again;
1733: }
1.41 noro 1734: #if USE_GEOBUCKET
1.53 noro 1735: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1736: #else
1.69 noro 1737: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1738: #endif
1.20 noro 1739: if ( !stat ) {
1740: NEXT(l) = d; d = l;
1.103 noro 1741: d = nd_reconstruct(1,d);
1.20 noro 1742: goto again;
1743: } else if ( nf ) {
1.77 noro 1744: if ( nd_demand ) {
1745: nfqv = ndv_load(nd_psn);
1746: nfq = ndvtond(0,nfqv);
1747: } else
1748: nfq = 0;
1749: if ( !nfq ) {
1750: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1751: NEXT(l) = d; d = l;
1.103 noro 1752: d = nd_reconstruct(1,d);
1.77 noro 1753: goto again;
1754: }
1.70 noro 1755: }
1.20 noro 1756: if ( nfq ) {
1.70 noro 1757: /* m|HC(nfq) => failure */
1.113 noro 1758: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1.63 noro 1759:
1.71 noro 1760: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.118 noro 1761: if ( nd_nalg ) {
1762: /* m|DN(HC(nf)^(-1)) => failure */
1763: get_eg(&egm0);
1764: if ( !nd_monic(m,&nfq) ) return 0;
1765: get_eg(&egm1); add_eg(&eg_monic,&egm0,&egm1);
1766: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1767: nfv = ndv_dup(0,nfqv); ndv_mod(m,nfv); nd_free(nf);
1768: } else {
1769: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1770: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1771: }
1.77 noro 1772: nh = ndv_newps(0,nfv,nfqv);
1.20 noro 1773: d = update_pairs(d,g,nh);
1774: g = update_base(g,nh);
1775: } else {
1.71 noro 1776: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1777: }
1778: } else {
1.71 noro 1779: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1780: }
1781: FREENDP(l);
1782: }
1.77 noro 1783: if ( nd_demand )
1784: for ( t = g; t; t = NEXT(t) )
1785: BDY(t) = (pointer)ndv_load((int)BDY(t));
1786: else
1787: for ( t = g; t; t = NEXT(t) )
1788: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.118 noro 1789: if ( nd_nalg ) {
1790: print_eg("monic",&eg_monic);
1791: print_eg("invdalg",&eg_invdalg);
1792: print_eg("le",&eg_le);
1793: }
1.20 noro 1794: return g;
1795: }
1796:
1.23 noro 1797: int ndv_compare(NDV *p1,NDV *p2)
1798: {
1.34 noro 1799: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1800: }
1801:
1802: int ndv_compare_rev(NDV *p1,NDV *p2)
1803: {
1.34 noro 1804: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1805: }
1806:
1.61 noro 1807: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1808: {
1.69 noro 1809: int i,n,stat;
1810: ND nf,g,head;
1.23 noro 1811: NODE t,a0,a;
1.113 noro 1812: union oNDC dn;
1.69 noro 1813: NDV *w;
1.113 noro 1814: Q q,num,den;
1.23 noro 1815:
1.69 noro 1816: n = length(f);
1817: #if 0
1818: w = (NDV *)ALLOCA(n*sizeof(NDV));
1819: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1820: qsort(w,n,sizeof(NDV),
1821: (int (*)(const void *,const void *))ndv_compare);
1822: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1823: #endif
1824: ndv_setup(m,0,f);
1825: for ( i = 0; i < n; ) {
1826: g = ndvtond(m,nd_ps[i]);
1827: g = nd_separate_head(g,&head);
1828: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1829: if ( !stat )
1.103 noro 1830: nd_reconstruct(0,0);
1.69 noro 1831: else {
1.71 noro 1832: if ( DP_Print ) { printf("."); fflush(stdout); }
1.85 noro 1833: if ( !m ) {
1.113 noro 1834: NTOQ(NM(dn.z),SGN(dn.z),num);
1835: mulq(HCQ(head),num,&q); HCQ(head) = q;
1836: if ( DN(dn.z) ) {
1837: NTOQ(DN(dn.z),1,den);
1838: nd_mul_c_q(nf,den);
1839: }
1.85 noro 1840: }
1.69 noro 1841: nf = nd_add(m,head,nf);
1842: ndv_free(nd_ps[i]);
1.24 noro 1843: nd_removecont(m,nf);
1.69 noro 1844: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1845: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1846: i++;
1.23 noro 1847: }
1848: }
1.71 noro 1849: if ( DP_Print ) { printf("\n"); }
1.23 noro 1850: for ( a0 = 0, i = 0; i < n; i++ ) {
1851: NEXTNODE(a0,a);
1.69 noro 1852: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1853: }
1854: NEXT(a) = 0;
1855: return a0;
1856: }
1857:
1.1 noro 1858: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1859: {
1860: ND_pairs d1,nd,cur,head,prev,remove;
1861:
1862: if ( !g ) return d;
1863: d = crit_B(d,t);
1864: d1 = nd_newpairs(g,t);
1865: d1 = crit_M(d1);
1866: d1 = crit_F(d1);
1.55 noro 1867: if ( do_weyl )
1868: head = d1;
1869: else {
1870: prev = 0; cur = head = d1;
1871: while ( cur ) {
1872: if ( crit_2( cur->i1,cur->i2 ) ) {
1873: remove = cur;
1874: if ( !prev ) head = cur = NEXT(cur);
1875: else cur = NEXT(prev) = NEXT(cur);
1876: FREENDP(remove);
1877: } else {
1878: prev = cur; cur = NEXT(cur);
1879: }
1.1 noro 1880: }
1881: }
1882: if ( !d )
1883: return head;
1884: else {
1885: nd = d;
1.34 noro 1886: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1887: NEXT(nd) = head;
1888: return d;
1889: }
1890: }
1891:
1892: ND_pairs nd_newpairs( NODE g, int t )
1893: {
1894: NODE h;
1.61 noro 1895: UINT *dl;
1.34 noro 1896: int ts,s;
1.1 noro 1897: ND_pairs r,r0;
1898:
1.20 noro 1899: dl = DL(nd_psh[t]);
1.34 noro 1900: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1901: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1902: NEXTND_pairs(r0,r);
1903: r->i1 = (int)BDY(h);
1904: r->i2 = t;
1.20 noro 1905: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1906: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1907: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1908: }
1909: NEXT(r) = 0;
1910: return r0;
1911: }
1912:
1913: ND_pairs crit_B( ND_pairs d, int s )
1914: {
1915: ND_pairs cur,head,prev,remove;
1.61 noro 1916: UINT *t,*tl,*lcm;
1.1 noro 1917: int td,tdl;
1918:
1919: if ( !d ) return 0;
1.20 noro 1920: t = DL(nd_psh[s]);
1.1 noro 1921: prev = 0;
1922: head = cur = d;
1.61 noro 1923: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1924: while ( cur ) {
1925: tl = cur->lcm;
1926: if ( ndl_reducible(tl,t)
1.20 noro 1927: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1928: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1929: remove = cur;
1930: if ( !prev ) {
1931: head = cur = NEXT(cur);
1932: } else {
1933: cur = NEXT(prev) = NEXT(cur);
1934: }
1935: FREENDP(remove);
1936: } else {
1.34 noro 1937: prev = cur; cur = NEXT(cur);
1.1 noro 1938: }
1939: }
1940: return head;
1941: }
1942:
1943: ND_pairs crit_M( ND_pairs d1 )
1944: {
1945: ND_pairs e,d2,d3,dd,p;
1.61 noro 1946: UINT *id,*jd;
1.1 noro 1947:
1948: for ( dd = 0, e = d1; e; e = d3 ) {
1949: if ( !(d2 = NEXT(e)) ) {
1950: NEXT(e) = dd;
1951: return e;
1952: }
1.34 noro 1953: id = LCM(e);
1.1 noro 1954: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 1955: p = NEXT(d2);
1956: jd = LCM(d2);
1957: if ( ndl_equal(jd,id) )
1958: ;
1959: else if ( TD(jd) > TD(id) )
1.1 noro 1960: if ( ndl_reducible(jd,id) ) continue;
1961: else ;
1.34 noro 1962: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 1963: NEXT(d2) = d3;
1964: d3 = d2;
1965: }
1966: NEXT(e) = dd;
1967: dd = e;
1968: continue;
1969: /**/
1970: delit: NEXT(d2) = d3;
1971: d3 = d2;
1972: for ( ; p; p = d2 ) {
1973: d2 = NEXT(p);
1974: NEXT(p) = d3;
1975: d3 = p;
1976: }
1977: FREENDP(e);
1978: }
1979: return dd;
1980: }
1981:
1982: ND_pairs crit_F( ND_pairs d1 )
1983: {
1984: ND_pairs rest, head,remove;
1985: ND_pairs last, p, r, w;
1986: int s;
1987:
1988: for ( head = last = 0, p = d1; NEXT(p); ) {
1989: r = w = equivalent_pairs(p,&rest);
1.14 noro 1990: s = SG(r);
1.1 noro 1991: w = NEXT(w);
1992: while ( w ) {
1993: if ( crit_2(w->i1,w->i2) ) {
1994: r = w;
1995: w = NEXT(w);
1996: while ( w ) {
1997: remove = w;
1998: w = NEXT(w);
1999: FREENDP(remove);
2000: }
2001: break;
1.14 noro 2002: } else if ( SG(w) < s ) {
1.1 noro 2003: FREENDP(r);
2004: r = w;
1.14 noro 2005: s = SG(r);
1.1 noro 2006: w = NEXT(w);
2007: } else {
2008: remove = w;
2009: w = NEXT(w);
2010: FREENDP(remove);
2011: }
2012: }
2013: if ( last ) NEXT(last) = r;
2014: else head = r;
2015: NEXT(last = r) = 0;
2016: p = rest;
2017: if ( !p ) return head;
2018: }
2019: if ( !last ) return p;
2020: NEXT(last) = p;
2021: return head;
2022: }
2023:
2024: int crit_2( int dp1, int dp2 )
2025: {
1.20 noro 2026: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2027: }
2028:
1.40 noro 2029: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2030: {
2031: ND_pairs w,p,r,s;
1.61 noro 2032: UINT *d;
1.1 noro 2033:
2034: w = d1;
1.34 noro 2035: d = LCM(w);
1.1 noro 2036: s = NEXT(w);
2037: NEXT(w) = 0;
2038: for ( r = 0; s; s = p ) {
2039: p = NEXT(s);
1.34 noro 2040: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2041: NEXT(s) = w; w = s;
1.1 noro 2042: } else {
1.39 noro 2043: NEXT(s) = r; r = s;
1.1 noro 2044: }
2045: }
2046: *prest = r;
2047: return w;
2048: }
2049:
2050: NODE update_base(NODE nd,int ndp)
2051: {
1.61 noro 2052: UINT *dl, *dln;
1.1 noro 2053: NODE last, p, head;
2054:
1.20 noro 2055: dl = DL(nd_psh[ndp]);
1.1 noro 2056: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2057: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2058: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2059: p = NEXT(p);
2060: if ( last ) NEXT(last) = p;
2061: } else {
2062: if ( !last ) head = p;
2063: p = NEXT(last = p);
2064: }
2065: }
2066: head = append_one(head,ndp);
2067: return head;
2068: }
2069:
2070: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2071: {
2072: ND_pairs m,ml,p,l;
1.61 noro 2073: UINT *lcm;
1.33 noro 2074: int s,td,len,tlen,c,c1;
1.1 noro 2075:
2076: if ( !(p = NEXT(m = d)) ) {
2077: *prest = p;
2078: NEXT(m) = 0;
2079: return m;
2080: }
1.14 noro 2081: s = SG(m);
1.77 noro 2082: if ( !NoSugar ) {
2083: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2084: if ( (SG(p) < s)
2085: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
2086: ml = l; m = p; s = SG(m);
2087: }
2088: } else {
2089: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2090: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
2091: ml = l; m = p; s = SG(m);
2092: }
2093: }
1.1 noro 2094: if ( !ml ) *prest = NEXT(m);
2095: else {
2096: NEXT(ml) = NEXT(m);
2097: *prest = d;
2098: }
2099: NEXT(m) = 0;
2100: return m;
2101: }
2102:
1.63 noro 2103: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2104: {
1.69 noro 2105: int msugar,i;
1.63 noro 2106: ND_pairs t,dm0,dm,dr0,dr;
2107:
2108: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2109: if ( SG(t) < msugar ) msugar = SG(t);
2110: dm0 = 0; dr0 = 0;
1.69 noro 2111: for ( i = 0, t = d; t; t = NEXT(t) )
2112: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2113: if ( dm0 ) NEXT(dm) = t;
2114: else dm0 = t;
2115: dm = t;
1.69 noro 2116: i++;
1.63 noro 2117: } else {
2118: if ( dr0 ) NEXT(dr) = t;
2119: else dr0 = t;
2120: dr = t;
2121: }
2122: NEXT(dm) = 0;
2123: if ( dr0 ) NEXT(dr) = 0;
2124: *prest = dr0;
2125: return dm0;
2126: }
2127:
1.77 noro 2128: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2129: {
1.3 noro 2130: int len;
1.13 noro 2131: RHist r;
1.20 noro 2132: NDV b;
1.3 noro 2133:
1.1 noro 2134: if ( nd_psn == nd_pslen ) {
2135: nd_pslen *= 2;
1.11 noro 2136: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2137: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2138: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2139: nd_bound = (UINT **)
2140: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2141: }
1.39 noro 2142: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2143: nd_ps[nd_psn] = a;
1.39 noro 2144: if ( aq ) {
1.63 noro 2145: nd_ps_trace[nd_psn] = aq;
2146: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2147: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2148: } else {
1.63 noro 2149: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2150: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2151: }
1.77 noro 2152: if ( nd_demand ) {
2153: if ( aq ) {
2154: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2155: nd_ps_trace[nd_psn] = 0;
2156: } else {
2157: ndv_save(nd_ps[nd_psn],nd_psn);
2158: nd_ps[nd_psn] = 0;
2159: }
2160: }
1.1 noro 2161: return nd_psn++;
2162: }
2163:
1.61 noro 2164: void ndv_setup(int mod,int trace,NODE f)
1.1 noro 2165: {
1.5 noro 2166: int i,j,td,len,max;
1.1 noro 2167: NODE s,s0,f0;
1.61 noro 2168: UINT *d;
1.13 noro 2169: RHist r;
1.69 noro 2170: NDV *w;
1.61 noro 2171: NDV a,am;
1.11 noro 2172:
2173: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2174:
1.69 noro 2175: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2176: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2177: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
2178: qsort(w,nd_psn,sizeof(NDV),
2179: (int (*)(const void *,const void *))ndv_compare);
2180: nd_pslen = 2*nd_psn;
1.11 noro 2181: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2182: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2183: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2184: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2185:
1.11 noro 2186: if ( !nd_red )
1.13 noro 2187: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2188: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2189: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2190: if ( trace ) {
1.69 noro 2191: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2192: ndv_removecont(0,a);
2193: am = nd_ps[i] = ndv_dup(mod,a);
2194: ndv_mod(mod,am);
2195: ndv_removecont(mod,am);
2196: } else {
1.69 noro 2197: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2198: ndv_removecont(mod,a);
1.39 noro 2199: }
1.61 noro 2200: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2201: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2202: nd_psh[i] = r;
1.77 noro 2203: if ( nd_demand ) {
2204: if ( trace ) {
2205: ndv_save(nd_ps_trace[i],i);
2206: nd_ps_trace[i] = 0;
2207: } else {
2208: ndv_save(nd_ps[i],i);
2209: nd_ps[i] = 0;
2210: }
2211: }
1.20 noro 2212: }
2213: }
2214:
1.119 noro 2215: struct order_spec *append_block(struct order_spec *spec,
2216: int nv,int nalg,int ord);
2217:
1.121 ! noro 2218: extern VECT current_dl_weight_vector_obj;
! 2219: static VECT prev_weight_vector_obj;
! 2220:
1.120 noro 2221: void preprocess_algcoef(VL vv,VL av,struct order_spec *ord,LIST f,
2222: struct order_spec **ord1p,LIST *f1p,NODE *alistp)
2223: {
1.121 ! noro 2224: NODE alist,t,s,r0,r,arg;
1.120 noro 2225: VL tv;
2226: P poly;
2227: DP d;
2228: Alg alpha,dp;
2229: DAlg inv,da,hc;
2230: MP m;
1.121 ! noro 2231: int i,nvar,nalg,n;
1.120 noro 2232: NumberField nf;
2233: LIST f1,f2;
2234: struct order_spec *current_spec;
1.121 ! noro 2235: VECT obj,obj0;
! 2236: Obj tmp;
1.120 noro 2237:
2238: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++);
2239: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++);
2240:
2241: for ( alist = 0, tv = av; tv; tv = NEXT(tv) ) {
2242: NEXTNODE(alist,t); MKV(tv->v,poly);
2243: MKAlg(poly,alpha); BDY(t) = (pointer)alpha;
2244: tv->v = tv->v->priv;
2245: }
2246: NEXT(t) = 0;
2247:
2248: /* simplification, makeing polynomials monic */
2249: setfield_dalg(alist);
2250: obj_algtodalg(f,&f1);
2251: for ( t = BDY(f); t; t = NEXT(t) ) {
2252: initd(ord); ptod(vv,vv,(P)BDY(t),&d);
2253: hc = (DAlg)BDY(d)->c;
2254: if ( NID(hc) == N_DA ) {
2255: invdalg(hc,&inv);
2256: for ( m = BDY(d); m; m = NEXT(m) ) {
2257: muldalg(inv,(DAlg)m->c,&da); m->c = (P)da;
2258: }
2259: }
2260: initd(ord); dtop(vv,vv,d,&poly); BDY(f) = (pointer)poly;
2261: }
2262: obj_dalgtoalg(f1,&f);
2263:
2264: /* append alg vars to the var list */
2265: for ( tv = vv; NEXT(tv); tv = NEXT(tv) );
2266: NEXT(tv) = av;
2267:
2268: /* append a block to ord */
2269: *ord1p = append_block(ord,nvar,nalg,2);
2270:
2271: /* create generator list */
2272: nf = get_numberfield();
2273: for ( i = nalg-1, t = BDY(f); i >= 0; i-- ) {
2274: MKAlg(nf->defpoly[i],dp);
2275: MKNODE(s,dp,t); t = s;
2276: }
2277: MKLIST(f1,t);
2278: *alistp = alist;
2279: algobjtorat(f1,f1p);
1.121 ! noro 2280:
! 2281: /* creating a new weight vector */
! 2282: prev_weight_vector_obj = obj0 = current_dl_weight_vector_obj;
! 2283: n = nvar+nalg+1;
! 2284: MKVECT(obj,n);
! 2285: if ( obj0 && obj0->len == nvar )
! 2286: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = BDY(obj0)[i];
! 2287: else
! 2288: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = (pointer)ONE;
! 2289: for ( i = 0; i < nalg; i++ ) BDY(obj)[i+nvar] = 0;
! 2290: BDY(obj)[n-1] = (pointer)ONE;
! 2291: arg = mknode(1,obj);
! 2292: Pdp_set_weight(arg,&tmp);
! 2293: }
! 2294:
! 2295: NODE postprocess_algcoef(VL av,NODE alist,NODE r)
! 2296: {
! 2297: NODE s,t,u0,u;
! 2298: P p;
! 2299: VL tv;
! 2300: Obj obj,tmp;
! 2301: NODE arg;
! 2302:
! 2303: u0 = 0;
! 2304: for ( t = r; t; t = NEXT(t) ) {
! 2305: p = (P)BDY(t);
! 2306: for ( tv = av, s = alist; tv; tv = NEXT(tv), s = NEXT(s) ) {
! 2307: substr(CO,0,(Obj)p,tv->v,(Obj)BDY(s),&obj); p = (P)obj;
! 2308: }
! 2309: if ( OID(p) == O_P || (OID(p) == O_N && NID((Num)p) != N_A) ) {
! 2310: NEXTNODE(u0,u);
! 2311: BDY(u) = (pointer)p;
! 2312: }
! 2313: }
! 2314: arg = mknode(1,prev_weight_vector_obj);
! 2315: Pdp_set_weight(arg,&tmp);
! 2316:
! 2317: return u0;
1.120 noro 2318: }
2319:
1.63 noro 2320: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2321: {
1.119 noro 2322: VL tv,fv,vv,vc,av;
2323: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2324: int e,max,nvar,i;
1.69 noro 2325: NDV b;
1.119 noro 2326: int ishomo,nalg;
2327: Alg alpha,dp;
2328: P p;
2329: LIST f1,f2;
2330: Obj obj;
2331: NumberField nf;
1.120 noro 2332: struct order_spec *ord1;
1.1 noro 2333:
1.78 noro 2334: if ( !m && Demand ) nd_demand = 1;
2335: else nd_demand = 0;
2336:
1.103 noro 2337: #if 0
1.74 noro 2338: ndv_alloc = 0;
1.103 noro 2339: #endif
1.1 noro 2340: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2341: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2342: switch ( ord->id ) {
2343: case 1:
2344: if ( ord->nv != nvar )
2345: error("nd_{gr,f4} : invalid order specification");
2346: break;
2347: default:
2348: break;
2349: }
1.119 noro 2350: nd_nalg = 0;
2351: if ( !m ) {
2352: get_algtree((Obj)f,&av);
2353: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 ! noro 2354: nd_ntrans = nvar;
1.119 noro 2355: nd_nalg = nalg;
2356: /* #i -> t#i */
2357: if ( nalg ) {
1.120 noro 2358: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2359: ord = ord1;
2360: f = f1;
1.119 noro 2361: }
1.120 noro 2362: nvar += nalg;
1.119 noro 2363: }
1.32 noro 2364: nd_init_ord(ord);
1.61 noro 2365: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2366: for ( tv = vv; tv; tv = NEXT(tv) ) {
2367: e = getdeg(tv->v,(P)BDY(t));
2368: max = MAX(e,max);
2369: }
2370: nd_setup_parameters(nvar,max);
1.77 noro 2371: ishomo = 1;
1.1 noro 2372: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2373: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.77 noro 2374: if ( ishomo )
2375: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2376: if ( m ) ndv_mod(m,b);
1.61 noro 2377: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2378: }
2379: if ( fd0 ) NEXT(fd) = 0;
1.61 noro 2380: ndv_setup(m,0,fd0);
1.77 noro 2381: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.82 noro 2382: nd_demand = 0;
1.61 noro 2383: x = ndv_reducebase(x);
2384: x = ndv_reduceall(m,x);
1.23 noro 2385: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2386: NEXTNODE(r0,r);
1.61 noro 2387: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2388: }
2389: if ( r0 ) NEXT(r) = 0;
1.121 ! noro 2390: if ( nalg )
! 2391: r0 = postprocess_algcoef(av,alist,r0);
1.20 noro 2392: MKLIST(*rp,r0);
1.103 noro 2393: #if 0
1.74 noro 2394: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 2395: #endif
1.20 noro 2396: }
2397:
1.52 noro 2398: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2399: {
1.119 noro 2400: VL tv,fv,vv,vc,av;
2401: NODE fd,fd0,in0,in,r,r0,t,s,cand,alist;
1.61 noro 2402: int m,nocheck,nvar,mindex,e,max;
2403: NDV c;
2404: NMV a;
1.27 noro 2405: P p;
1.61 noro 2406: EPOS oepos;
1.119 noro 2407: int obpe,oadv,wmax,i,len,cbpe,ishomo,nalg;
2408: Alg alpha,dp;
2409: P poly;
2410: LIST f1,f2;
2411: Obj obj;
2412: NumberField nf;
1.120 noro 2413: struct order_spec *ord1;
1.20 noro 2414:
2415: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2416: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2417: switch ( ord->id ) {
2418: case 1:
2419: if ( ord->nv != nvar )
2420: error("nd_gr_trace : invalid order specification");
2421: break;
2422: default:
2423: break;
2424: }
1.119 noro 2425:
2426: get_algtree((Obj)f,&av);
2427: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 ! noro 2428: nd_ntrans = nvar;
1.119 noro 2429: nd_nalg = nalg;
2430: /* #i -> t#i */
2431: if ( nalg ) {
1.120 noro 2432: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2433: ord = ord1;
2434: f = f1;
1.119 noro 2435: }
1.120 noro 2436: nvar += nalg;
1.119 noro 2437:
1.52 noro 2438: nocheck = 0;
2439: mindex = 0;
2440:
1.78 noro 2441: if ( Demand ) nd_demand = 1;
2442: else nd_demand = 0;
2443:
1.52 noro 2444: /* setup modulus */
2445: if ( trace < 0 ) {
2446: trace = -trace;
2447: nocheck = 1;
2448: }
2449: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2450: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2451: for ( tv = vv; tv; tv = NEXT(tv) ) {
2452: e = getdeg(tv->v,(P)BDY(t));
2453: max = MAX(e,max);
1.23 noro 2454: }
1.61 noro 2455: nd_init_ord(ord);
2456: nd_setup_parameters(nvar,max);
2457: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 noro 2458: ishomo = 1;
1.61 noro 2459: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2460: c = ptondv(CO,vv,(P)BDY(t));
1.82 noro 2461: if ( ishomo )
2462: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2463: if ( c ) {
2464: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2465: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2466: }
1.61 noro 2467: }
2468: if ( in0 ) NEXT(in) = 0;
2469: if ( fd0 ) NEXT(fd) = 0;
1.82 noro 2470: if ( !ishomo && homo ) {
1.61 noro 2471: for ( t = in0, wmax = 0; t; t = NEXT(t) ) {
2472: c = (NDV)BDY(t); len = LEN(c);
2473: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2474: wmax = MAX(TD(DL(a)),wmax);
2475: }
2476: homogenize_order(ord,nvar,&ord1);
1.89 noro 2477: nd_init_ord(ord1);
1.61 noro 2478: nd_setup_parameters(nvar+1,wmax);
2479: for ( t = fd0; t; t = NEXT(t) )
2480: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2481: }
1.52 noro 2482: while ( 1 ) {
1.77 noro 2483: if ( Demand )
2484: nd_demand = 1;
1.61 noro 2485: ndv_setup(m,1,fd0);
1.82 noro 2486: cand = nd_gb_trace(m,ishomo || homo);
1.52 noro 2487: if ( !cand ) {
2488: /* failure */
1.61 noro 2489: if ( trace > 1 ) { *rp = 0; return; }
2490: else m = get_lprime(++mindex);
1.52 noro 2491: continue;
2492: }
1.82 noro 2493: if ( !ishomo && homo ) {
1.27 noro 2494: /* dehomogenization */
1.61 noro 2495: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2496: nd_init_ord(ord);
1.61 noro 2497: nd_setup_parameters(nvar,0);
1.27 noro 2498: }
1.77 noro 2499: nd_demand = 0;
1.61 noro 2500: cand = ndv_reducebase(cand);
2501: cand = ndv_reduceall(0,cand);
1.78 noro 2502: cbpe = nd_bpe;
1.61 noro 2503: if ( nocheck )
2504: break;
2505: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2506: /* success */
2507: break;
2508: else if ( trace > 1 ) {
2509: /* failure */
2510: *rp = 0; return;
1.61 noro 2511: } else {
1.52 noro 2512: /* try the next modulus */
2513: m = get_lprime(++mindex);
1.61 noro 2514: /* reset the parameters */
1.82 noro 2515: if ( !ishomo && homo ) {
1.89 noro 2516: nd_init_ord(ord1);
1.61 noro 2517: nd_setup_parameters(nvar+1,wmax);
2518: } else {
2519: nd_init_ord(ord);
2520: nd_setup_parameters(nvar,max);
2521: }
2522: }
1.52 noro 2523: }
1.27 noro 2524: /* dp->p */
1.61 noro 2525: nd_bpe = cbpe;
1.73 noro 2526: nd_setup_parameters(nd_nvar,0);
1.121 ! noro 2527: for ( r = cand; r; r = NEXT(r) )
1.119 noro 2528: BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.121 ! noro 2529: if ( nalg )
! 2530: cand = postprocess_algcoef(av,alist,cand);
1.27 noro 2531: MKLIST(*rp,cand);
1.1 noro 2532: }
2533:
1.61 noro 2534: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2535: {
1.61 noro 2536: UINT *d;
1.57 noro 2537: int i,j,l,s,ord_l;
1.43 noro 2538: struct order_pair *op;
1.1 noro 2539:
2540: d = dl->d;
1.41 noro 2541: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2542: if ( nd_blockmask ) {
2543: l = nd_blockmask->n;
2544: op = nd_blockmask->order_pair;
2545: for ( j = 0, s = 0; j < l; j++ ) {
2546: ord_l = op[j].length;
1.57 noro 2547: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2548: }
2549: TD(r) = ndl_weight(r);
1.61 noro 2550: ndl_weight_mask(r);
1.43 noro 2551: } else {
1.56 noro 2552: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2553: TD(r) = ndl_weight(r);
2554: }
1.1 noro 2555: }
2556:
1.61 noro 2557: DL ndltodl(int n,UINT *ndl)
1.1 noro 2558: {
2559: DL dl;
2560: int *d;
1.57 noro 2561: int i,j,l,s,ord_l;
1.43 noro 2562: struct order_pair *op;
1.1 noro 2563:
2564: NEWDL(dl,n);
1.34 noro 2565: dl->td = TD(ndl);
1.1 noro 2566: d = dl->d;
1.43 noro 2567: if ( nd_blockmask ) {
2568: l = nd_blockmask->n;
2569: op = nd_blockmask->order_pair;
2570: for ( j = 0, s = 0; j < l; j++ ) {
2571: ord_l = op[j].length;
1.57 noro 2572: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2573: }
2574: } else {
1.56 noro 2575: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2576: }
1.1 noro 2577: return dl;
2578: }
2579:
1.61 noro 2580: void ndl_print(UINT *dl)
1.1 noro 2581: {
2582: int n;
1.57 noro 2583: int i,j,l,ord_l,s,s0;
1.43 noro 2584: struct order_pair *op;
1.1 noro 2585:
2586: n = nd_nvar;
2587: printf("<<");
1.43 noro 2588: if ( nd_blockmask ) {
2589: l = nd_blockmask->n;
2590: op = nd_blockmask->order_pair;
2591: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2592: ord_l = op[j].length;
1.57 noro 2593: for ( i = 0; i < ord_l; i++, s++ )
2594: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2595: }
2596: } else {
1.56 noro 2597: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2598: }
1.1 noro 2599: printf(">>");
2600: }
2601:
2602: void nd_print(ND p)
2603: {
2604: NM m;
2605:
2606: if ( !p )
2607: printf("0\n");
2608: else {
2609: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2610: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2611: else printf("+%d*",CM(m));
1.14 noro 2612: ndl_print(DL(m));
1.1 noro 2613: }
2614: printf("\n");
2615: }
2616: }
2617:
1.113 noro 2618: void nd_print_q(ND p)
1.16 noro 2619: {
2620: NM m;
2621:
2622: if ( !p )
2623: printf("0\n");
2624: else {
2625: for ( m = BDY(p); m; m = NEXT(m) ) {
2626: printf("+");
1.113 noro 2627: printexpr(CO,(Obj)CQ(m));
1.16 noro 2628: printf("*");
2629: ndl_print(DL(m));
2630: }
2631: printf("\n");
2632: }
2633: }
2634:
1.1 noro 2635: void ndp_print(ND_pairs d)
2636: {
2637: ND_pairs t;
2638:
1.34 noro 2639: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2640: printf("\n");
2641: }
2642:
1.20 noro 2643: void nd_removecont(int mod,ND p)
1.16 noro 2644: {
2645: int i,n;
1.113 noro 2646: Q *w;
2647: Q dvr,t;
1.16 noro 2648: NM m;
1.21 noro 2649: struct oVECT v;
1.113 noro 2650: N q,r;
1.16 noro 2651:
1.71 noro 2652: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2653: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2654: else {
2655: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
1.113 noro 2656: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2657: v.len = n;
2658: v.body = (pointer *)w;
1.113 noro 2659: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2660: removecont_array(w,n);
1.113 noro 2661: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2662: }
2663: }
2664:
1.21 noro 2665: void nd_removecont2(ND p1,ND p2)
2666: {
2667: int i,n1,n2,n;
1.113 noro 2668: Q *w;
2669: Q dvr,t;
1.21 noro 2670: NM m;
2671: struct oVECT v;
1.113 noro 2672: N q,r;
1.21 noro 2673:
2674: if ( !p1 ) {
2675: nd_removecont(0,p2); return;
2676: } else if ( !p2 ) {
2677: nd_removecont(0,p1); return;
2678: }
2679: n1 = nd_length(p1);
2680: n2 = nd_length(p2);
2681: n = n1+n2;
1.113 noro 2682: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2683: v.len = n;
2684: v.body = (pointer *)w;
1.113 noro 2685: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2686: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2687: removecont_array(w,n);
1.113 noro 2688: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2689: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.21 noro 2690: }
2691:
1.20 noro 2692: void ndv_removecont(int mod,NDV p)
1.16 noro 2693: {
2694: int i,len;
1.113 noro 2695: Q *w;
2696: Q dvr,t;
1.16 noro 2697: NMV m;
2698:
1.71 noro 2699: if ( mod == -1 )
2700: ndv_mul_c(mod,p,_invsf(HCM(p)));
2701: else if ( mod )
1.20 noro 2702: ndv_mul_c(mod,p,invm(HCM(p),mod));
2703: else {
2704: len = p->len;
1.113 noro 2705: w = (Q *)ALLOCA(len*sizeof(Q));
2706: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
2707: sortbynm(w,len);
2708: qltozl(w,len,&dvr);
2709: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2710: divq(CQ(m),dvr,&t); CQ(m) = t;
2711: }
1.16 noro 2712: }
1.21 noro 2713: }
2714:
1.61 noro 2715: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2716: {
2717: int len,i,max;
2718: NMV m,mr0,mr,t;
2719:
2720: len = p->len;
2721: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2722: max = MAX(max,TD(DL(m)));
2723: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2724: m = (NMV)((char *)mr0+(len-1)*oadv);
2725: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2726: t = (NMV)ALLOCA(nmv_adv);
2727: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2728: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
1.113 noro 2729: CQ(mr) = CQ(m);
1.61 noro 2730: ndl_copy(DL(t),DL(mr));
2731: }
2732: NV(p)++;
2733: BDY(p) = mr0;
2734: }
2735:
1.45 noro 2736: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2737: {
1.45 noro 2738: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.113 noro 2739: Q *w;
2740: Q dvr,t;
1.23 noro 2741: NMV m,r;
2742:
2743: len = p->len;
2744: newnvar = nd_nvar-1;
1.48 noro 2745: newexporigin = nd_get_exporigin(ord);
1.45 noro 2746: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2747: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2748: ndl_dehomogenize(DL(m));
1.23 noro 2749: if ( newwpd != nd_wpd ) {
1.90 noro 2750: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
1.61 noro 2751: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.113 noro 2752: CQ(r) = CQ(m);
1.45 noro 2753: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2754: adj = nd_exporigin-newexporigin;
2755: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2756: }
2757: }
2758: NV(p)--;
2759: }
2760:
1.113 noro 2761: void removecont_array(Q *c,int n)
1.21 noro 2762: {
2763: struct oVECT v;
1.113 noro 2764: Q d0,d1,a,u,u1,gcd;
1.110 noro 2765: int i,j;
1.113 noro 2766: N qn,rn,gn;
2767: Q *q,*r;
1.21 noro 2768:
1.113 noro 2769: q = (Q *)ALLOCA(n*sizeof(Q));
2770: r = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2771: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
1.113 noro 2772: igcdv_estimate(&v,&d0);
2773: for ( i = 0; i < n; i++ ) {
2774: divn(NM(c[i]),NM(d0),&qn,&rn);
2775: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2776: NTOQ(rn,SGN(c[i]),r[i]);
2777: }
1.34 noro 2778: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2779: if ( i < n ) {
2780: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
1.113 noro 2781: igcdv(&v,&d1);
2782: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2783: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
1.21 noro 2784: for ( i = 0; i < n; i++ ) {
1.113 noro 2785: mulq(a,q[i],&u);
2786: if ( r[i] ) {
2787: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2788: addq(u,u1,&q[i]);
2789: } else
2790: q[i] = u;
1.21 noro 2791: }
2792: }
1.34 noro 2793: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2794: }
2795:
1.19 noro 2796: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2797: {
2798: NM m;
2799: int c,c1;
2800:
1.34 noro 2801: if ( !p ) return;
1.115 noro 2802: if ( mul == 1 ) return;
1.71 noro 2803: if ( mod == -1 )
2804: for ( m = BDY(p); m; m = NEXT(m) )
2805: CM(m) = _mulsf(CM(m),mul);
2806: else
2807: for ( m = BDY(p); m; m = NEXT(m) ) {
2808: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2809: }
1.1 noro 2810: }
2811:
1.113 noro 2812: void nd_mul_c_q(ND p,Q mul)
1.16 noro 2813: {
2814: NM m;
1.113 noro 2815: Q c;
1.16 noro 2816:
1.113 noro 2817: if ( !p ) return;
1.115 noro 2818: if ( UNIQ(mul) ) return;
1.16 noro 2819: for ( m = BDY(p); m; m = NEXT(m) ) {
1.113 noro 2820: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 2821: }
2822: }
2823:
1.61 noro 2824: void nd_mul_c_p(VL vl,ND p,P mul)
2825: {
2826: NM m;
2827: P c;
2828:
2829: if ( !p ) return;
2830: for ( m = BDY(p); m; m = NEXT(m) ) {
2831: mulp(vl,CP(m),mul,&c); CP(m) = c;
2832: }
2833: }
2834:
1.1 noro 2835: void nd_free(ND p)
2836: {
2837: NM t,s;
2838:
1.34 noro 2839: if ( !p ) return;
1.1 noro 2840: t = BDY(p);
2841: while ( t ) {
2842: s = NEXT(t);
2843: FREENM(t);
2844: t = s;
2845: }
2846: FREEND(p);
2847: }
2848:
1.23 noro 2849: void ndv_free(NDV p)
2850: {
2851: GC_free(BDY(p));
2852: }
2853:
1.61 noro 2854: void nd_append_red(UINT *d,int i)
1.1 noro 2855: {
1.13 noro 2856: RHist m,m0;
1.1 noro 2857: int h;
2858:
1.13 noro 2859: NEWRHist(m);
1.34 noro 2860: h = ndl_hash_value(d);
1.13 noro 2861: m->index = i;
1.14 noro 2862: ndl_copy(d,DL(m));
1.1 noro 2863: NEXT(m) = nd_red[h];
2864: nd_red[h] = m;
2865: }
2866:
1.61 noro 2867: UINT *ndv_compute_bound(NDV p)
1.1 noro 2868: {
1.61 noro 2869: UINT *d1,*d2,*t;
2870: UINT u;
1.57 noro 2871: int i,j,k,l,len,ind;
1.45 noro 2872: NMV m;
1.1 noro 2873:
2874: if ( !p )
2875: return 0;
1.61 noro 2876: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2877: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2878: len = LEN(p);
2879: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2880: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2881: ndl_lcm(DL(m),d1,d2);
1.1 noro 2882: t = d1; d1 = d2; d2 = t;
2883: }
1.12 noro 2884: l = nd_nvar+31;
1.61 noro 2885: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 2886: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2887: u = d1[i];
2888: k = (nd_epw-1)*nd_bpe;
2889: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2890: t[ind] = (u>>k)&nd_mask0;
2891: }
2892: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 2893: return t;
2894: }
2895:
1.99 noro 2896: UINT *nd_compute_bound(ND p)
2897: {
2898: UINT *d1,*d2,*t;
2899: UINT u;
2900: int i,j,k,l,len,ind;
2901: NM m;
2902:
2903: if ( !p )
2904: return 0;
2905: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2906: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2907: len = LEN(p);
2908: m = BDY(p); ndl_copy(DL(m),d1); m = NEXT(m);
2909: for ( m = NEXT(m); m; m = NEXT(m) ) {
2910: ndl_lcm(DL(m),d1,d2);
2911: t = d1; d1 = d2; d2 = t;
2912: }
2913: l = nd_nvar+31;
2914: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
2915: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2916: u = d1[i];
2917: k = (nd_epw-1)*nd_bpe;
2918: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2919: t[ind] = (u>>k)&nd_mask0;
2920: }
2921: for ( ; ind < l; ind++ ) t[ind] = 0;
2922: return t;
2923: }
2924:
1.48 noro 2925: int nd_get_exporigin(struct order_spec *ord)
2926: {
1.51 noro 2927: switch ( ord->id ) {
1.96 noro 2928: case 0: case 2:
1.48 noro 2929: return 1;
1.41 noro 2930: case 1:
2931: /* block order */
1.43 noro 2932: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 2933: return ord->ord.block.length+1;
1.96 noro 2934: case 3:
2935: error("nd_get_exporigin : composite order is not supported yet.");
1.41 noro 2936: }
1.48 noro 2937: }
2938:
1.61 noro 2939: void nd_setup_parameters(int nvar,int max) {
1.79 noro 2940: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 2941: struct order_pair *op;
1.48 noro 2942:
1.73 noro 2943: nd_nvar = nvar;
1.79 noro 2944: if ( max ) {
1.83 noro 2945: /* XXX */
2946: if ( do_weyl ) nd_bpe = 32;
2947: else if ( max < 2 ) nd_bpe = 1;
1.79 noro 2948: else if ( max < 4 ) nd_bpe = 2;
2949: else if ( max < 8 ) nd_bpe = 3;
2950: else if ( max < 16 ) nd_bpe = 4;
2951: else if ( max < 32 ) nd_bpe = 5;
2952: else if ( max < 64 ) nd_bpe = 6;
2953: else if ( max < 256 ) nd_bpe = 8;
2954: else if ( max < 1024 ) nd_bpe = 10;
2955: else if ( max < 65536 ) nd_bpe = 16;
2956: else nd_bpe = 32;
2957: }
1.61 noro 2958: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 2959: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
2960: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 2961: wpd = nd_exporigin+elen;
2962: if ( wpd != nd_wpd ) {
2963: nd_free_private_storage();
2964: nd_wpd = wpd;
2965: }
1.1 noro 2966: if ( nd_bpe < 32 ) {
2967: nd_mask0 = (1<<nd_bpe)-1;
2968: } else {
2969: nd_mask0 = 0xffffffff;
2970: }
2971: bzero(nd_mask,sizeof(nd_mask));
2972: nd_mask1 = 0;
2973: for ( i = 0; i < nd_epw; i++ ) {
2974: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
2975: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
2976: }
1.90 noro 2977: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
1.57 noro 2978: nd_epos = nd_create_epos(nd_ord);
1.43 noro 2979: nd_blockmask = nd_create_blockmask(nd_ord);
1.96 noro 2980: nd_work_vector = (int *)REALLOC(nd_work_vector,nd_nvar*sizeof(int));
1.1 noro 2981: }
2982:
1.103 noro 2983: ND_pairs nd_reconstruct(int trace,ND_pairs d)
1.1 noro 2984: {
1.37 noro 2985: int i,obpe,oadv,h;
1.71 noro 2986: static NM prev_nm_free_list;
2987: static ND_pairs prev_ndp_free_list;
1.13 noro 2988: RHist mr0,mr;
2989: RHist r;
1.37 noro 2990: RHist *old_red;
1.71 noro 2991: ND_pairs s0,s,t;
1.43 noro 2992: EPOS oepos;
1.15 noro 2993:
1.1 noro 2994: obpe = nd_bpe;
1.11 noro 2995: oadv = nmv_adv;
1.43 noro 2996: oepos = nd_epos;
1.63 noro 2997: if ( obpe < 2 ) nd_bpe = 2;
2998: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 2999: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 3000: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 3001: else if ( obpe < 6 ) nd_bpe = 6;
3002: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 3003: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 3004: else if ( obpe < 16 ) nd_bpe = 16;
3005: else if ( obpe < 32 ) nd_bpe = 32;
3006: else error("nd_reconstruct : exponent too large");
1.5 noro 3007:
1.73 noro 3008: nd_setup_parameters(nd_nvar,0);
1.1 noro 3009: prev_nm_free_list = _nm_free_list;
3010: prev_ndp_free_list = _ndp_free_list;
3011: _nm_free_list = 0;
3012: _ndp_free_list = 0;
1.53 noro 3013: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
3014: if ( trace )
3015: for ( i = nd_psn-1; i >= 0; i-- )
3016: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 3017: s0 = 0;
3018: for ( t = d; t; t = NEXT(t) ) {
3019: NEXTND_pairs(s0,s);
3020: s->i1 = t->i1;
3021: s->i2 = t->i2;
1.14 noro 3022: SG(s) = SG(t);
1.61 noro 3023: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 3024: }
1.37 noro 3025:
3026: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 3027: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 3028: old_red[i] = nd_red[i];
3029: nd_red[i] = 0;
3030: }
3031: for ( i = 0; i < REDTAB_LEN; i++ )
3032: for ( r = old_red[i]; r; r = NEXT(r) ) {
3033: NEWRHist(mr);
1.13 noro 3034: mr->index = r->index;
1.20 noro 3035: SG(mr) = SG(r);
1.61 noro 3036: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 3037: h = ndl_hash_value(DL(mr));
3038: NEXT(mr) = nd_red[h];
3039: nd_red[h] = mr;
1.6 noro 3040: }
1.37 noro 3041: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
3042: old_red = 0;
1.11 noro 3043: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 3044: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 3045: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 3046: nd_psh[i] = r;
1.11 noro 3047: }
1.1 noro 3048: if ( s0 ) NEXT(s) = 0;
3049: prev_nm_free_list = 0;
3050: prev_ndp_free_list = 0;
1.71 noro 3051: #if 0
1.1 noro 3052: GC_gcollect();
1.71 noro 3053: #endif
1.1 noro 3054: return s0;
3055: }
3056:
1.61 noro 3057: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 3058: {
1.57 noro 3059: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 3060: struct order_pair *op;
1.1 noro 3061:
3062: n = nd_nvar;
1.61 noro 3063: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 3064: omask0 = (1<<obpe)-1;
1.34 noro 3065: TD(r) = TD(d);
1.41 noro 3066: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 3067: if ( nd_blockmask ) {
3068: l = nd_blockmask->n;
3069: op = nd_blockmask->order_pair;
3070: for ( i = 1; i < nd_exporigin; i++ )
3071: r[i] = d[i];
3072: for ( j = 0, s = 0; j < l; j++ ) {
3073: ord_l = op[j].length;
1.57 noro 3074: for ( i = 0; i < ord_l; i++, s++ ) {
3075: ei = GET_EXP_OLD(d,s);
3076: PUT_EXP(r,s,ei);
3077: }
1.1 noro 3078: }
1.43 noro 3079: } else {
1.56 noro 3080: for ( i = 0; i < n; i++ ) {
3081: ei = GET_EXP_OLD(d,i);
3082: PUT_EXP(r,i,ei);
3083: }
1.1 noro 3084: }
3085: }
1.3 noro 3086:
1.6 noro 3087: ND nd_copy(ND p)
3088: {
3089: NM m,mr,mr0;
1.41 noro 3090: int c,n;
1.6 noro 3091: ND r;
3092:
3093: if ( !p )
3094: return 0;
3095: else {
3096: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3097: NEXTNM(mr0,mr);
1.14 noro 3098: CM(mr) = CM(m);
3099: ndl_copy(DL(m),DL(mr));
1.6 noro 3100: }
3101: NEXT(mr) = 0;
1.31 noro 3102: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 3103: SG(r) = SG(p);
1.6 noro 3104: return r;
3105: }
3106: }
3107:
1.53 noro 3108: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3109: {
3110: NM m;
3111: NDV p1,p2;
3112: ND t1,t2;
1.61 noro 3113: UINT *lcm;
1.31 noro 3114: int td;
1.11 noro 3115:
1.77 noro 3116: if ( !mod && nd_demand ) {
3117: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 3118: } else {
1.77 noro 3119: if ( trace ) {
3120: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3121: } else {
3122: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3123: }
1.20 noro 3124: }
1.34 noro 3125: lcm = LCM(p);
1.11 noro 3126: NEWNM(m);
1.113 noro 3127: CQ(m) = HCQ(p2);
1.34 noro 3128: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 3129: if ( ndl_check_bound2(p->i1,DL(m)) )
3130: return 0;
1.55 noro 3131: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 3132: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
3133: else if ( mod ) CM(m) = mod-HCM(p1);
1.113 noro 3134: else chsgnq(HCQ(p1),&CQ(m));
1.34 noro 3135: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 3136: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 3137: nd_free(t1);
3138: return 0;
3139: }
1.55 noro 3140: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 3141: *rp = nd_add(mod,t1,t2);
1.11 noro 3142: FREENM(m);
3143: return 1;
3144: }
3145:
1.19 noro 3146: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3147: {
3148: NMV m;
3149: int c,c1,len,i;
3150:
1.34 noro 3151: if ( !p ) return;
1.14 noro 3152: len = LEN(p);
1.71 noro 3153: if ( mod == -1 )
3154: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3155: CM(m) = _mulsf(CM(m),mul);
3156: else
3157: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3158: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3159: }
1.11 noro 3160: }
3161:
1.113 noro 3162: void ndv_mul_c_q(NDV p,Q mul)
1.16 noro 3163: {
3164: NMV m;
1.113 noro 3165: Q c;
1.16 noro 3166: int len,i;
3167:
1.34 noro 3168: if ( !p ) return;
1.16 noro 3169: len = LEN(p);
3170: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 3171: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 3172: }
3173: }
3174:
1.55 noro 3175: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3176: int n2,i,j,l,n,tlen;
1.61 noro 3177: UINT *d0;
1.55 noro 3178: NM *tab,*psum;
3179: ND s,r;
3180: NM t;
3181: NMV m1;
3182:
3183: if ( !p ) return 0;
3184: n = NV(p); n2 = n>>1;
3185: d0 = DL(m0);
3186: l = LEN(p);
3187: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3188: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3189: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3190: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3191: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3192: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3193: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3194: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3195: for ( j = 0; j < tlen; j++ ) {
3196: if ( tab[j] ) {
3197: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3198: }
3199: }
3200: }
3201: for ( i = tlen-1, r = 0; i >= 0; i-- )
3202: if ( psum[i] ) {
3203: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3204: MKND(n,psum[i],j,s);
3205: r = nd_add(mod,r,s);
3206: }
1.56 noro 3207: if ( r ) SG(r) = SG(p)+TD(d0);
3208: return r;
1.55 noro 3209: }
3210:
1.56 noro 3211: /* product of monomials */
3212: /* XXX block order is not handled correctly */
3213:
1.55 noro 3214: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3215: {
1.56 noro 3216: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3217: UINT *d0,*d1,*d,*dt,*ctab;
1.113 noro 3218: Q *ctab_q;
3219: Q q,q1;
1.61 noro 3220: UINT c0,c1,c;
1.55 noro 3221: NM *p;
3222: NM m,t;
3223:
3224: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3225: if ( !m0 || !m1 ) return;
3226: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3227: NEWNM(m); d = DL(m);
1.56 noro 3228: if ( mod ) {
3229: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3230: } else
1.113 noro 3231: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3232: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3233: homo = n&1 ? 1 : 0;
3234: if ( homo ) {
3235: /* offset of h-degree */
3236: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3237: PUT_EXP(DL(m),n-1,h);
3238: TD(DL(m)) = h;
1.61 noro 3239: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3240: }
3241: tab[0] = m;
3242: NEWNM(m); d = DL(m);
1.57 noro 3243: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3244: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3245: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3246: /* xi^a*(Di^k*xi^l)*Di^b */
3247: a += l; b += k;
1.56 noro 3248: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3249: if ( !k || !l ) {
3250: for ( j = 0; j < curlen; j++ )
1.56 noro 3251: if ( t = tab[j] ) {
3252: dt = DL(t);
3253: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3254: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3255: }
3256: curlen *= k+1;
3257: continue;
3258: }
3259: min = MIN(k,l);
1.56 noro 3260: if ( mod ) {
1.61 noro 3261: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3262: mkwcm(k,l,mod,ctab);
3263: } else {
1.113 noro 3264: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3265: mkwc(k,l,ctab_q);
1.56 noro 3266: }
1.57 noro 3267: for ( j = min; j >= 0; j-- ) {
1.56 noro 3268: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3269: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3270: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3271: if ( homo ) {
3272: TD(d) = s;
1.56 noro 3273: PUT_EXP(d,n-1,s-h);
1.55 noro 3274: } else TD(d) = h;
1.61 noro 3275: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3276: if ( mod ) c = ctab[j];
1.113 noro 3277: else q = ctab_q[j];
1.57 noro 3278: p = tab+curlen*j;
3279: if ( j == 0 ) {
3280: for ( u = 0; u < curlen; u++, p++ ) {
3281: if ( tab[u] ) {
3282: ndl_addto(DL(tab[u]),d);
3283: if ( mod ) {
3284: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3285: } else {
1.113 noro 3286: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
1.57 noro 3287: }
3288: }
1.56 noro 3289: }
1.57 noro 3290: } else {
3291: for ( u = 0; u < curlen; u++, p++ ) {
3292: if ( tab[u] ) {
3293: NEWNM(t);
3294: ndl_add(DL(tab[u]),d,DL(t));
3295: if ( mod ) {
3296: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3297: } else
1.113 noro 3298: mulq(CQ(tab[u]),q,&CQ(t));
1.57 noro 3299: *p = t;
3300: }
1.55 noro 3301: }
3302: }
3303: }
3304: curlen *= k+1;
3305: }
3306: FREENM(m);
3307: }
3308:
1.63 noro 3309: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3310: {
3311: NM mr,mr0;
3312: NMV m;
3313: UINT *d,*dt,*dm;
3314: int c,n,td,i,c1,c2,len;
1.113 noro 3315: Q q;
1.63 noro 3316: ND r;
3317:
3318: if ( !p ) return 0;
3319: else {
3320: n = NV(p); m = BDY(p);
3321: d = DL(m0);
3322: len = LEN(p);
3323: mr0 = 0;
3324: td = TD(d);
3325: c = CM(m0);
3326: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3327: NEXTNM(mr0,mr);
3328: CM(mr) = 1;
3329: ndl_add(DL(m),d,DL(mr));
3330: }
3331: NEXT(mr) = 0;
3332: MKND(NV(p),mr0,len,r);
3333: SG(r) = SG(p) + TD(d);
3334: return r;
3335: }
3336: }
3337:
1.55 noro 3338: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3339: {
3340: NM mr,mr0;
3341: NMV m;
1.61 noro 3342: UINT *d,*dt,*dm;
1.9 noro 3343: int c,n,td,i,c1,c2,len;
1.113 noro 3344: Q q;
1.9 noro 3345: ND r;
3346:
1.34 noro 3347: if ( !p ) return 0;
1.55 noro 3348: else if ( do_weyl )
1.71 noro 3349: if ( mod == -1 )
3350: error("ndv_mul_nm : not implemented (weyl)");
3351: else
3352: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3353: else {
3354: n = NV(p); m = BDY(p);
1.34 noro 3355: d = DL(m0);
1.14 noro 3356: len = LEN(p);
1.9 noro 3357: mr0 = 0;
1.34 noro 3358: td = TD(d);
1.71 noro 3359: if ( mod == -1 ) {
3360: c = CM(m0);
3361: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3362: NEXTNM(mr0,mr);
3363: CM(mr) = _mulsf(CM(m),c);
3364: ndl_add(DL(m),d,DL(mr));
3365: }
3366: } else if ( mod ) {
1.16 noro 3367: c = CM(m0);
3368: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3369: NEXTNM(mr0,mr);
3370: c1 = CM(m);
1.19 noro 3371: DMAR(c1,c,0,mod,c2);
1.16 noro 3372: CM(mr) = c2;
3373: ndl_add(DL(m),d,DL(mr));
3374: }
3375: } else {
1.113 noro 3376: q = CQ(m0);
1.16 noro 3377: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3378: NEXTNM(mr0,mr);
1.113 noro 3379: mulq(CQ(m),q,&CQ(mr));
1.16 noro 3380: ndl_add(DL(m),d,DL(mr));
3381: }
1.4 noro 3382: }
1.9 noro 3383: NEXT(mr) = 0;
1.31 noro 3384: MKND(NV(p),mr0,len,r);
1.34 noro 3385: SG(r) = SG(p) + TD(d);
1.9 noro 3386: return r;
1.4 noro 3387: }
3388: }
3389:
1.104 noro 3390: ND nd_quo(int mod,PGeoBucket bucket,NDV d)
1.99 noro 3391: {
3392: NM mq0,mq;
1.102 noro 3393: NMV tm;
1.113 noro 3394: Q q;
1.104 noro 3395: int i,nv,sg,c,c1,c2,hindex;
3396: ND p,t,r;
3397: N tnm;
3398:
1.99 noro 3399: if ( !p ) return 0;
3400: else {
1.104 noro 3401: nv = NV(d);
1.99 noro 3402: mq0 = 0;
1.102 noro 3403: tm = (NMV)ALLOCA(nmv_adv);
1.104 noro 3404: while ( 1 ) {
1.113 noro 3405: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.104 noro 3406: if ( hindex < 0 ) break;
3407: p = bucket->body[hindex];
1.99 noro 3408: NEXTNM(mq0,mq);
3409: ndl_sub(HDL(p),HDL(d),DL(mq));
1.102 noro 3410: ndl_copy(DL(mq),DL(tm));
3411: if ( mod ) {
3412: c1 = invm(HCM(d),mod); c2 = HCM(p);
3413: DMAR(c1,c2,0,mod,c); CM(mq) = c;
3414: CM(tm) = mod-c;
3415: } else {
1.113 noro 3416: divsn(NM(HCQ(p)),NM(HCQ(d)),&tnm);
3417: NTOQ(tnm,SGN(HCQ(p))*SGN(HCQ(d)),CQ(mq));
3418: chsgnq(CQ(mq),&CQ(tm));
1.102 noro 3419: }
3420: t = ndv_mul_nmv_trunc(mod,tm,d,HDL(d));
1.104 noro 3421: bucket->body[hindex] = nd_remove_head(p);
3422: t = nd_remove_head(t);
3423: add_pbucket(mod,bucket,t);
3424: }
3425: if ( !mq0 )
3426: r = 0;
3427: else {
3428: NEXT(mq) = 0;
3429: for ( i = 0, mq = mq0; mq; mq = NEXT(mq), i++ );
3430: MKND(nv,mq0,i,r);
3431: /* XXX */
3432: SG(r) = HTD(r);
1.99 noro 3433: }
3434: return r;
3435: }
3436: }
3437:
1.43 noro 3438: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3439: {
1.13 noro 3440: NMV m,mr,mr0,t;
3441: int len,i,k;
1.11 noro 3442:
1.61 noro 3443: if ( !p ) return;
3444: m = BDY(p); len = LEN(p);
3445: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3446: m = (NMV)((char *)mr0+(len-1)*oadv);
3447: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3448: t = (NMV)ALLOCA(nmv_adv);
3449: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
1.113 noro 3450: CQ(t) = CQ(m);
1.61 noro 3451: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3452: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
1.113 noro 3453: CQ(mr) = CQ(t);
1.61 noro 3454: ndl_copy(DL(t),DL(mr));
3455: }
3456: BDY(p) = mr0;
3457: }
3458:
3459: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3460: {
3461: NMV m,mr,mr0;
3462: int len,i;
3463: NDV r;
1.11 noro 3464:
1.61 noro 3465: if ( !p ) return 0;
3466: m = BDY(p); len = LEN(p);
3467: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3468: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3469: ndl_zero(DL(mr));
3470: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
1.113 noro 3471: CQ(mr) = CQ(m);
1.11 noro 3472: }
1.61 noro 3473: MKNDV(NV(p),mr0,len,r);
3474: SG(r) = SG(p);
3475: return r;
1.11 noro 3476: }
3477:
1.61 noro 3478: /* duplicate p */
3479:
3480: NDV ndv_dup(int mod,NDV p)
1.3 noro 3481: {
3482: NDV d;
1.61 noro 3483: NMV t,m,m0;
1.3 noro 3484: int i,len;
3485:
1.34 noro 3486: if ( !p ) return 0;
1.31 noro 3487: len = LEN(p);
1.34 noro 3488: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3489: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3490: ndl_copy(DL(t),DL(m));
1.113 noro 3491: CQ(m) = CQ(t);
1.3 noro 3492: }
3493: MKNDV(NV(p),m0,len,d);
1.23 noro 3494: SG(d) = SG(p);
3495: return d;
3496: }
3497:
1.63 noro 3498: ND nd_dup(ND p)
3499: {
3500: ND d;
3501: NM t,m,m0;
3502:
3503: if ( !p ) return 0;
3504: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3505: NEXTNM(m0,m);
3506: ndl_copy(DL(t),DL(m));
1.113 noro 3507: CQ(m) = CQ(t);
1.63 noro 3508: }
3509: if ( m0 ) NEXT(m) = 0;
3510: MKND(NV(p),m0,LEN(p),d);
3511: SG(d) = SG(p);
3512: return d;
3513: }
3514:
1.61 noro 3515: /* XXX if p->len == 0 then it represents 0 */
3516:
3517: void ndv_mod(int mod,NDV p)
3518: {
3519: NMV t,d;
3520: int r;
3521: int i,len,dlen;
1.71 noro 3522: Obj gfs;
1.61 noro 3523:
3524: if ( !p ) return;
3525: len = LEN(p);
3526: dlen = 0;
1.71 noro 3527: if ( mod == -1 )
3528: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3529: simp_ff((Obj)CP(t),&gfs);
3530: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3531: CM(d) = r;
3532: ndl_copy(DL(t),DL(d));
3533: NMV_ADV(d);
3534: dlen++;
3535: }
1.71 noro 3536: else
3537: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
1.113 noro 3538: r = rem(NM(CQ(t)),mod);
1.71 noro 3539: if ( r ) {
1.113 noro 3540: if ( SGN(CQ(t)) < 0 )
3541: r = mod-r;
1.71 noro 3542: CM(d) = r;
3543: ndl_copy(DL(t),DL(d));
3544: NMV_ADV(d);
3545: dlen++;
3546: }
3547: }
1.61 noro 3548: LEN(p) = dlen;
3549: }
3550:
3551: NDV ptondv(VL vl,VL dvl,P p)
3552: {
3553: ND nd;
3554:
3555: nd = ptond(vl,dvl,p);
3556: return ndtondv(0,nd);
3557: }
3558:
3559: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3560: {
1.61 noro 3561: int n,i,j,k,e;
3562: VL tvl;
3563: V v;
3564: DCP dc;
3565: DCP *w;
3566: ND r,s,t,u;
3567: P x;
3568: int c;
3569: UINT *d;
1.23 noro 3570: NM m,m0;
1.61 noro 3571:
3572: if ( !p )
3573: return 0;
3574: else if ( NUM(p) ) {
3575: NEWNM(m);
3576: ndl_zero(DL(m));
1.113 noro 3577: CQ(m) = (Q)p;
1.61 noro 3578: NEXT(m) = 0;
3579: MKND(nd_nvar,m,1,r);
3580: SG(r) = 0;
3581: return r;
3582: } else {
3583: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3584: w = (DCP *)ALLOCA(k*sizeof(DCP));
3585: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3586: for ( i = 0, tvl = dvl, v = VR(p);
1.98 noro 3587: tvl && tvl->v != v; tvl = NEXT(tvl), i++ );
1.61 noro 3588: if ( !tvl ) {
3589: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3590: t = ptond(vl,dvl,COEF(w[j]));
3591: pwrp(vl,x,DEG(w[j]),&p);
3592: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3593: }
3594: return s;
3595: } else {
3596: NEWNM(m0); d = DL(m0);
3597: for ( j = k-1, s = 0; j >= 0; j-- ) {
3598: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3599: TD(d) = MUL_WEIGHT(e,i);
3600: if ( nd_blockmask) ndl_weight_mask(d);
3601: t = ptond(vl,dvl,COEF(w[j]));
3602: for ( m = BDY(t); m; m = NEXT(m) )
3603: ndl_addto(DL(m),d);
3604: SG(t) += TD(d);
3605: s = nd_add(0,s,t);
3606: }
3607: FREENM(m0);
3608: return s;
3609: }
3610: }
3611: }
3612:
3613: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3614: {
3615: VL tvl;
3616: int len,n,j,i,e;
3617: NMV m;
3618: Q q;
3619: P c;
3620: UINT *d;
3621: P s,r,u,t,w;
1.71 noro 3622: GFS gfs;
1.23 noro 3623:
1.34 noro 3624: if ( !p ) return 0;
1.61 noro 3625: else {
3626: len = LEN(p);
3627: n = NV(p);
3628: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3629: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3630: if ( mod == -1 ) {
3631: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3632: } else if ( mod ) {
1.61 noro 3633: STOQ(CM(m),q); c = (P)q;
3634: } else
1.113 noro 3635: c = CP(m);
1.61 noro 3636: d = DL(m);
3637: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3638: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3639: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3640: }
3641: addp(vl,s,t,&u); s = u;
3642: }
3643: return s;
1.23 noro 3644: }
1.3 noro 3645: }
3646:
1.61 noro 3647: NDV ndtondv(int mod,ND p)
1.11 noro 3648: {
3649: NDV d;
1.61 noro 3650: NMV m,m0;
3651: NM t;
3652: int i,len;
1.11 noro 3653:
1.34 noro 3654: if ( !p ) return 0;
1.61 noro 3655: len = LEN(p);
1.74 noro 3656: if ( mod )
3657: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3658: else
3659: m0 = m = MALLOC(len*nmv_adv);
1.103 noro 3660: #if 0
1.74 noro 3661: ndv_alloc += nmv_adv*len;
1.103 noro 3662: #endif
1.61 noro 3663: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3664: ndl_copy(DL(t),DL(m));
1.113 noro 3665: CQ(m) = CQ(t);
1.11 noro 3666: }
1.61 noro 3667: MKNDV(NV(p),m0,len,d);
1.14 noro 3668: SG(d) = SG(p);
1.11 noro 3669: return d;
3670: }
3671:
1.61 noro 3672: ND ndvtond(int mod,NDV p)
1.11 noro 3673: {
1.61 noro 3674: ND d;
3675: NM m,m0;
1.11 noro 3676: NMV t;
1.61 noro 3677: int i,len;
1.11 noro 3678:
1.34 noro 3679: if ( !p ) return 0;
1.11 noro 3680: m0 = 0;
1.61 noro 3681: len = p->len;
3682: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3683: NEXTNM(m0,m);
3684: ndl_copy(DL(t),DL(m));
1.113 noro 3685: CQ(m) = CQ(t);
1.11 noro 3686: }
3687: NEXT(m) = 0;
1.61 noro 3688: MKND(NV(p),m0,len,d);
1.14 noro 3689: SG(d) = SG(p);
1.11 noro 3690: return d;
3691: }
3692:
1.3 noro 3693: void ndv_print(NDV p)
3694: {
3695: NMV m;
3696: int i,len;
3697:
1.34 noro 3698: if ( !p ) printf("0\n");
1.3 noro 3699: else {
1.14 noro 3700: len = LEN(p);
1.3 noro 3701: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3702: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3703: else printf("+%d*",CM(m));
1.16 noro 3704: ndl_print(DL(m));
3705: }
3706: printf("\n");
3707: }
3708: }
3709:
1.113 noro 3710: void ndv_print_q(NDV p)
1.16 noro 3711: {
3712: NMV m;
3713: int i,len;
3714:
1.34 noro 3715: if ( !p ) printf("0\n");
1.16 noro 3716: else {
3717: len = LEN(p);
3718: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3719: printf("+");
1.113 noro 3720: printexpr(CO,(Obj)CQ(m));
1.16 noro 3721: printf("*");
1.14 noro 3722: ndl_print(DL(m));
1.3 noro 3723: }
3724: printf("\n");
3725: }
1.25 noro 3726: }
3727:
1.61 noro 3728: NODE ndv_reducebase(NODE x)
1.27 noro 3729: {
3730: int len,i,j;
3731: NDV *w;
3732: NODE t,t0;
3733:
3734: len = length(x);
3735: w = (NDV *)ALLOCA(len*sizeof(NDV));
3736: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3737: for ( i = 0; i < len; i++ ) {
3738: for ( j = 0; j < i; j++ ) {
3739: if ( w[i] && w[j] )
3740: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3741: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3742: }
3743: }
3744: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3745: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3746: }
3747: NEXT(t) = 0; x = t0;
3748: return x;
1.11 noro 3749: }
1.32 noro 3750:
1.43 noro 3751: /* XXX incomplete */
3752:
1.32 noro 3753: void nd_init_ord(struct order_spec *ord)
3754: {
1.43 noro 3755: switch ( ord->id ) {
1.32 noro 3756: case 0:
1.43 noro 3757: switch ( ord->ord.simple ) {
3758: case 0:
3759: nd_dcomp = 1;
3760: nd_isrlex = 1;
3761: break;
3762: case 1:
3763: nd_dcomp = 1;
3764: nd_isrlex = 0;
3765: break;
3766: case 2:
3767: nd_dcomp = 0;
3768: nd_isrlex = 0;
1.45 noro 3769: ndl_compare_function = ndl_lex_compare;
1.58 noro 3770: break;
3771: case 11:
3772: /* XXX */
3773: nd_dcomp = 0;
3774: nd_isrlex = 1;
3775: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3776: break;
3777: default:
3778: error("nd_gr : unsupported order");
3779: }
1.32 noro 3780: break;
3781: case 1:
1.96 noro 3782: /* block order */
1.43 noro 3783: /* XXX */
3784: nd_dcomp = -1;
1.32 noro 3785: nd_isrlex = 0;
1.45 noro 3786: ndl_compare_function = ndl_block_compare;
1.34 noro 3787: break;
1.43 noro 3788: case 2:
1.96 noro 3789: /* matrix order */
3790: /* XXX */
3791: nd_dcomp = -1;
3792: nd_isrlex = 0;
3793: nd_matrix_len = ord->ord.matrix.row;
3794: nd_matrix = ord->ord.matrix.matrix;
3795: ndl_compare_function = ndl_matrix_compare;
3796: break;
3797: case 3:
1.97 noro 3798: /* composite order */
3799: nd_dcomp = -1;
3800: nd_isrlex = 0;
3801: nd_worb_len = ord->ord.composite.length;
3802: nd_worb = ord->ord.composite.w_or_b;
3803: ndl_compare_function = ndl_composite_compare;
1.32 noro 3804: break;
3805: }
1.41 noro 3806: nd_ord = ord;
1.32 noro 3807: }
3808:
1.43 noro 3809: BlockMask nd_create_blockmask(struct order_spec *ord)
3810: {
3811: int n,i,j,s,l;
1.61 noro 3812: UINT *t;
1.43 noro 3813: BlockMask bm;
3814:
1.96 noro 3815: /* we only create mask table for block order */
3816: if ( ord->id != 1 )
1.43 noro 3817: return 0;
3818: n = ord->ord.block.length;
3819: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3820: bm->n = n;
3821: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3822: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3823: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3824: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3825: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3826: l = bm->order_pair[i].length;
3827: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3828: }
3829: return bm;
1.57 noro 3830: }
3831:
3832: EPOS nd_create_epos(struct order_spec *ord)
3833: {
3834: int i,j,l,s,ord_l,ord_o;
3835: EPOS epos;
3836: struct order_pair *op;
3837:
3838: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3839: switch ( ord->id ) {
3840: case 0:
3841: if ( nd_isrlex ) {
3842: for ( i = 0; i < nd_nvar; i++ ) {
3843: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3844: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3845: }
3846: } else {
3847: for ( i = 0; i < nd_nvar; i++ ) {
3848: epos[i].i = nd_exporigin + i/nd_epw;
3849: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3850: }
3851: }
3852: break;
3853: case 1:
3854: /* block order */
3855: l = ord->ord.block.length;
3856: op = ord->ord.block.order_pair;
3857: for ( j = 0, s = 0; j < l; j++ ) {
3858: ord_o = op[j].order;
3859: ord_l = op[j].length;
3860: if ( !ord_o )
3861: for ( i = 0; i < ord_l; i++ ) {
3862: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3863: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3864: }
3865: else
3866: for ( i = 0; i < ord_l; i++ ) {
3867: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3868: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3869: }
3870: s += ord_l;
3871: }
3872: break;
3873: case 2:
1.96 noro 3874: /* matrix order */
3875: case 3:
3876: /* composite order */
3877: for ( i = 0; i < nd_nvar; i++ ) {
3878: epos[i].i = nd_exporigin + i/nd_epw;
3879: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3880: }
3881: break;
1.57 noro 3882: }
3883: return epos;
1.43 noro 3884: }
1.59 noro 3885:
3886: /* external interface */
3887:
3888: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
3889: {
1.61 noro 3890: NODE t,in0,in;
1.59 noro 3891: ND nd,nf;
1.61 noro 3892: NDV ndv;
3893: VL vv,tv;
3894: int stat,nvar,max,e;
1.113 noro 3895: union oNDC dn;
1.59 noro 3896:
1.116 noro 3897: if ( !f ) {
3898: *rp = 0;
3899: return;
3900: }
1.59 noro 3901: pltovl(v,&vv);
1.74 noro 3902: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 3903:
3904: /* get the degree bound */
3905: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
3906: for ( tv = vv; tv; tv = NEXT(tv) ) {
3907: e = getdeg(tv->v,(P)BDY(t));
3908: max = MAX(e,max);
3909: }
3910: for ( tv = vv; tv; tv = NEXT(tv) ) {
3911: e = getdeg(tv->v,f);
3912: max = MAX(e,max);
3913: }
3914:
1.59 noro 3915: nd_init_ord(ord);
1.61 noro 3916: nd_setup_parameters(nvar,max);
3917:
3918: /* conversion to ndv */
3919: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
3920: NEXTNODE(in0,in);
3921: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 3922: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3923: }
3924: NEXTNODE(in0,in);
3925: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 3926: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3927: NEXT(in) = 0;
3928:
3929: ndv_setup(m,0,in0);
1.59 noro 3930: nd_psn--;
3931: nd_scale=2;
3932: while ( 1 ) {
3933: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 3934: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 3935: if ( !stat ) {
3936: nd_psn++;
1.103 noro 3937: nd_reconstruct(0,0);
1.59 noro 3938: nd_psn--;
3939: } else
3940: break;
3941: }
1.61 noro 3942: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 3943: }
3944:
3945: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
3946: {
3947: NM m;
3948: UINT *t,*s;
3949: int i;
3950:
3951: for ( i = 0; i < n; i++ ) r[i] = 0;
3952: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
3953: t = DL(m);
3954: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3955: r[i] = CM(m);
3956: }
3957: for ( i = 0; !r[i]; i++ );
3958: return i;
3959: }
3960:
1.113 noro 3961: int nd_to_vect_q(UINT *s0,int n,ND d,Q *r)
1.74 noro 3962: {
1.107 noro 3963: NM m;
1.74 noro 3964: UINT *t,*s;
1.107 noro 3965: int i;
1.74 noro 3966:
3967: for ( i = 0; i < n; i++ ) r[i] = 0;
1.107 noro 3968: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
1.74 noro 3969: t = DL(m);
3970: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.113 noro 3971: r[i] = CQ(m);
1.74 noro 3972: }
3973: for ( i = 0; !r[i]; i++ );
3974: return i;
3975: }
3976:
1.67 noro 3977: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 3978: {
3979: NM m;
3980: NMV mr;
3981: UINT *d,*t,*s;
3982: NDV p;
1.67 noro 3983: unsigned char *ivc;
3984: unsigned short *ivs;
1.81 noro 3985: UINT *v,*ivi,*s0v;
1.67 noro 3986: int i,j,len,prev,diff,cdiff;
3987: IndArray r;
1.64 noro 3988:
3989: m = pair->mul;
3990: d = DL(m);
3991: p = nd_ps[pair->index];
3992: len = LEN(p);
3993: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 3994: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 3995: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3996: ndl_add(d,DL(mr),t);
3997: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 3998: v[j] = i;
3999: }
1.81 noro 4000: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 4001: r->head = v[0];
4002: diff = 0;
4003: for ( i = 1; i < len; i++ ) {
4004: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
4005: }
4006: if ( diff < 256 ) {
4007: r->width = 1;
4008: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
4009: r->index.c = ivc;
4010: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
4011: } else if ( diff < 65536 ) {
4012: r->width = 2;
4013: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
4014: r->index.s = ivs;
4015: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
4016: } else {
4017: r->width = 4;
4018: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
4019: r->index.i = ivi;
4020: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 4021: }
1.67 noro 4022: return r;
1.64 noro 4023: }
4024:
1.65 noro 4025:
1.113 noro 4026: int ndv_reduce_vect_q(Q *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.107 noro 4027: {
4028: int i,j,k,len,pos,prev;
1.113 noro 4029: Q cs,mcs,c1,c2,cr,gcd,t;
1.107 noro 4030: IndArray ivect;
4031: unsigned char *ivc;
4032: unsigned short *ivs;
4033: unsigned int *ivi;
4034: NDV redv;
4035: NMV mr;
4036: NODE rp;
4037: int maxrs;
4038:
4039: maxrs = 0;
4040: for ( i = 0; i < nred; i++ ) {
4041: ivect = imat[i];
4042: k = ivect->head;
4043: if ( svect[k] ) {
4044: maxrs = MAX(maxrs,rp0[i]->sugar);
4045: redv = nd_ps[rp0[i]->index];
4046: len = LEN(redv); mr = BDY(redv);
1.113 noro 4047: igcd_cofactor(svect[k],CQ(mr),&gcd,&cs,&cr);
4048: chsgnq(cs,&mcs);
4049: if ( !UNIQ(cr) ) {
4050: for ( j = 0; j < col; j++ ) {
4051: mulq(svect[j],cr,&c1); svect[j] = c1;
4052: }
4053: }
1.107 noro 4054: svect[k] = 0; prev = k;
4055: switch ( ivect->width ) {
4056: case 1:
4057: ivc = ivect->index.c;
4058: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4059: pos = prev+ivc[j]; prev = pos;
1.113 noro 4060: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4061: }
4062: break;
4063: case 2:
4064: ivs = ivect->index.s;
4065: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4066: pos = prev+ivs[j]; prev = pos;
1.113 noro 4067: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4068: }
4069: break;
4070: case 4:
4071: ivi = ivect->index.i;
4072: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4073: pos = prev+ivi[j]; prev = pos;
1.113 noro 4074: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4075: }
4076: break;
4077: }
4078: }
4079: }
4080: return maxrs;
4081: }
4082:
1.76 noro 4083: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 4084: {
1.67 noro 4085: int i,j,k,len,pos,prev;
1.66 noro 4086: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 4087: IndArray ivect;
4088: unsigned char *ivc;
4089: unsigned short *ivs;
4090: unsigned int *ivi;
1.65 noro 4091: NDV redv;
1.67 noro 4092: NMV mr;
1.65 noro 4093: NODE rp;
1.76 noro 4094: int maxrs;
1.65 noro 4095:
1.76 noro 4096: maxrs = 0;
1.74 noro 4097: for ( i = 0; i < nred; i++ ) {
1.65 noro 4098: ivect = imat[i];
1.67 noro 4099: k = ivect->head; svect[k] %= m;
1.65 noro 4100: if ( c = svect[k] ) {
1.76 noro 4101: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4102: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 4103: len = LEN(redv); mr = BDY(redv);
4104: svect[k] = 0; prev = k;
4105: switch ( ivect->width ) {
4106: case 1:
4107: ivc = ivect->index.c;
4108: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4109: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
4110: prev = pos;
4111: DMA(c1,c,c2,up,lo);
4112: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4113: } else svect[pos] = lo;
4114: }
4115: break;
4116: case 2:
4117: ivs = ivect->index.s;
4118: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4119: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
4120: prev = pos;
4121: DMA(c1,c,c2,up,lo);
4122: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4123: } else svect[pos] = lo;
4124: }
4125: break;
4126: case 4:
4127: ivi = ivect->index.i;
4128: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4129: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
4130: prev = pos;
4131: DMA(c1,c,c2,up,lo);
4132: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4133: } else svect[pos] = lo;
4134: }
4135: break;
1.65 noro 4136: }
4137: }
4138: }
1.66 noro 4139: for ( i = 0; i < col; i++ )
4140: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 4141: return maxrs;
1.65 noro 4142: }
4143:
1.76 noro 4144: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 4145: {
4146: int i,j,k,len,pos,prev;
4147: UINT c,c1,c2,c3,up,lo,dmy;
4148: IndArray ivect;
4149: unsigned char *ivc;
4150: unsigned short *ivs;
4151: unsigned int *ivi;
4152: NDV redv;
4153: NMV mr;
4154: NODE rp;
1.76 noro 4155: int maxrs;
1.72 noro 4156:
1.76 noro 4157: maxrs = 0;
1.74 noro 4158: for ( i = 0; i < nred; i++ ) {
1.72 noro 4159: ivect = imat[i];
4160: k = ivect->head; svect[k] %= m;
4161: if ( c = svect[k] ) {
1.76 noro 4162: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4163: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 4164: len = LEN(redv); mr = BDY(redv);
4165: svect[k] = 0; prev = k;
4166: switch ( ivect->width ) {
4167: case 1:
4168: ivc = ivect->index.c;
4169: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4170: pos = prev+ivc[j]; prev = pos;
4171: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4172: }
4173: break;
4174: case 2:
4175: ivs = ivect->index.s;
4176: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4177: pos = prev+ivs[j]; prev = pos;
4178: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4179: }
4180: break;
4181: case 4:
4182: ivi = ivect->index.i;
4183: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4184: pos = prev+ivi[j]; prev = pos;
4185: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4186: }
4187: break;
4188: }
4189: }
4190: }
1.76 noro 4191: return maxrs;
1.72 noro 4192: }
4193:
1.65 noro 4194: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
4195: {
4196: int j,k,len;
4197: UINT *p;
4198: UINT c;
4199: NDV r;
4200: NMV mr0,mr;
4201:
4202: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4203: if ( !len ) return 0;
4204: else {
1.74 noro 4205: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4206: #if 0
1.74 noro 4207: ndv_alloc += nmv_adv*len;
1.103 noro 4208: #endif
1.65 noro 4209: mr = mr0;
4210: p = s0vect;
4211: for ( j = k = 0; j < col; j++, p += nd_wpd )
4212: if ( !rhead[j] ) {
4213: if ( c = vect[k++] ) {
4214: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
4215: }
4216: }
4217: MKNDV(nd_nvar,mr0,len,r);
4218: return r;
4219: }
4220: }
4221:
1.113 noro 4222: NDV vect_to_ndv_q(Q *vect,int spcol,int col,int *rhead,UINT *s0vect)
1.107 noro 4223: {
4224: int j,k,len;
4225: UINT *p;
1.113 noro 4226: Q c;
1.107 noro 4227: NDV r;
4228: NMV mr0,mr;
4229:
4230: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4231: if ( !len ) return 0;
4232: else {
1.109 noro 4233: mr0 = (NMV)GC_malloc(nmv_adv*len);
1.107 noro 4234: #if 0
4235: ndv_alloc += nmv_adv*len;
4236: #endif
4237: mr = mr0;
4238: p = s0vect;
4239: for ( j = k = 0; j < col; j++, p += nd_wpd )
4240: if ( !rhead[j] ) {
4241: if ( c = vect[k++] ) {
1.113 noro 4242: if ( DN(c) )
4243: error("afo");
4244: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
1.107 noro 4245: }
4246: }
4247: MKNDV(nd_nvar,mr0,len,r);
4248: return r;
4249: }
4250: }
4251:
1.74 noro 4252: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket)
1.65 noro 4253: {
4254: ND_pairs t;
4255: NODE sp0,sp;
4256: int stat;
4257: ND spol;
4258:
4259: for ( t = l; t; t = NEXT(t) ) {
4260: stat = nd_sp(m,0,t,&spol);
4261: if ( !stat ) return 0;
4262: if ( spol ) {
4263: add_pbucket_symbolic(bucket,spol);
4264: }
4265: }
1.68 noro 4266: return 1;
1.65 noro 4267: }
4268:
4269: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
4270: {
4271: NODE rp0,rp;
4272: NM mul,head,s0,s;
1.76 noro 4273: int index,col,i,sugar;
1.65 noro 4274: RHist h;
4275: UINT *s0v,*p;
4276: NM_ind_pair pair;
4277: ND red;
4278:
4279: s0 = 0; rp0 = 0; col = 0;
4280: while ( 1 ) {
4281: head = remove_head_pbucket_symbolic(bucket);
4282: if ( !head ) break;
4283: if ( !s0 ) s0 = head;
4284: else NEXT(s) = head;
4285: s = head;
4286: index = ndl_find_reducer(DL(head));
4287: if ( index >= 0 ) {
4288: h = nd_psh[index];
4289: NEWNM(mul);
4290: ndl_sub(DL(head),DL(h),DL(mul));
4291: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.76 noro 4292: sugar = TD(DL(mul))+SG(nd_ps[index]);
4293: MKNM_ind_pair(pair,mul,index,sugar);
1.65 noro 4294: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
4295: add_pbucket_symbolic(bucket,nd_remove_head(red));
4296: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4297: }
4298: col++;
4299: }
1.72 noro 4300: if ( rp0 ) NEXT(rp) = 0;
4301: NEXT(s) = 0;
1.65 noro 4302: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4303: for ( i = 0, p = s0v, s = s0; i < col;
4304: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4305: *s0vect = s0v;
4306: *r = rp0;
4307: return col;
4308: }
4309:
1.69 noro 4310: NODE nd_f4(int m)
4311: {
4312: int i,nh,stat,index;
4313: NODE r,g;
4314: ND_pairs d,l,t;
4315: ND spol,red;
4316: NDV nf,redv;
4317: NM s0,s;
1.84 noro 4318: NODE rp0,srp0,nflist;
1.69 noro 4319: int nsp,nred,col,rank,len,k,j,a;
4320: UINT c;
1.74 noro 4321: UINT **spmat;
1.69 noro 4322: UINT *s0vect,*svect,*p,*v;
4323: int *colstat;
4324: IndArray *imat;
4325: int *rhead;
4326: int spcol,sprow;
4327: int sugar;
4328: PGeoBucket bucket;
4329: struct oEGT eg0,eg1,eg_f4;
4330:
1.103 noro 4331: #if 0
1.74 noro 4332: ndv_alloc = 0;
1.103 noro 4333: #endif
1.69 noro 4334: g = 0; d = 0;
4335: for ( i = 0; i < nd_psn; i++ ) {
4336: d = update_pairs(d,g,i);
4337: g = update_base(g,i);
4338: }
4339: while ( d ) {
4340: get_eg(&eg0);
4341: l = nd_minsugarp(d,&d);
4342: sugar = SG(l);
4343: bucket = create_pbucket();
1.74 noro 4344: stat = nd_sp_f4(m,l,bucket);
1.69 noro 4345: if ( !stat ) {
4346: for ( t = l; NEXT(t); t = NEXT(t) );
4347: NEXT(t) = d; d = l;
1.103 noro 4348: d = nd_reconstruct(0,d);
1.69 noro 4349: continue;
4350: }
1.86 noro 4351: if ( bucket->m < 0 ) continue;
1.69 noro 4352: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
4353: if ( !col ) {
4354: for ( t = l; NEXT(t); t = NEXT(t) );
4355: NEXT(t) = d; d = l;
1.103 noro 4356: d = nd_reconstruct(0,d);
1.69 noro 4357: continue;
4358: }
4359: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4360: if ( DP_Print )
4361: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4362: sugar,eg_f4.exectime+eg_f4.gctime);
4363: if ( 1 )
4364: nflist = nd_f4_red(m,l,s0vect,col,rp0);
4365: else
4366: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0);
1.69 noro 4367: /* adding new bases */
1.74 noro 4368: for ( r = nflist; r; r = NEXT(r) ) {
4369: nf = (NDV)BDY(r);
1.69 noro 4370: ndv_removecont(m,nf);
1.77 noro 4371: nh = ndv_newps(m,nf,0);
1.69 noro 4372: d = update_pairs(d,g,nh);
4373: g = update_base(g,nh);
4374: }
4375: }
4376: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.103 noro 4377: #if 0
1.74 noro 4378: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 4379: #endif
1.69 noro 4380: return g;
4381: }
1.74 noro 4382:
4383: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
1.63 noro 4384: {
1.67 noro 4385: IndArray *imat;
1.106 noro 4386: int nsp,nred,i;
1.65 noro 4387: int *rhead;
1.106 noro 4388: NODE r0,rp;
1.74 noro 4389: ND_pairs sp;
4390: NM_ind_pair *rvect;
1.63 noro 4391:
1.74 noro 4392: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
1.106 noro 4393: nred = length(rp0);
1.74 noro 4394: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4395: rhead = (int *)ALLOCA(col*sizeof(int));
4396: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4397:
1.74 noro 4398: /* construction of index arrays */
4399: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4400: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4401: rvect[i] = (NM_ind_pair)BDY(rp);
4402: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4403: rhead[imat[i]->head] = 1;
4404: }
1.107 noro 4405: if ( m )
4406: r0 = nd_f4_red_main(m,sp0,nsp,s0vect,col,rvect,rhead,imat,nred);
4407: else
1.113 noro 4408: r0 = nd_f4_red_q_main(sp0,nsp,s0vect,col,rvect,rhead,imat,nred);
1.106 noro 4409: return r0;
4410: }
1.74 noro 4411:
1.106 noro 4412: NODE nd_f4_red_main(int m,ND_pairs sp0,int nsp,UINT *s0vect,int col,
4413: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4414: {
4415: int spcol,sprow,a;
4416: int i,j,k,l,rank;
4417: NODE r0,r;
4418: ND_pairs sp;
4419: ND spol;
4420: int **spmat;
4421: UINT *svect,*v;
4422: int *colstat;
4423: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4424: int maxrs;
4425: int *spsugar;
4426:
4427: spcol = col-nred;
4428: get_eg(&eg0);
1.74 noro 4429: /* elimination (1st step) */
4430: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4431: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4432: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4433: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4434: nd_sp(m,0,sp,&spol);
1.75 noro 4435: if ( !spol ) continue;
1.74 noro 4436: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4437: if ( m == -1 )
4438: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4439: else
4440: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4441: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4442: if ( i < col ) {
4443: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4444: for ( j = k = 0; j < col; j++ )
4445: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4446: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4447: sprow++;
4448: }
1.76 noro 4449: nd_free(spol);
1.74 noro 4450: }
1.80 noro 4451: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4452: if ( DP_Print ) {
4453: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4454: fflush(asir_out);
4455: }
1.74 noro 4456: /* free index arrays */
4457: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4458:
4459: /* elimination (2nd step) */
4460: colstat = (int *)ALLOCA(spcol*sizeof(int));
4461: if ( m == -1 )
1.76 noro 4462: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4463: else
1.76 noro 4464: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4465: r0 = 0;
4466: for ( i = 0; i < rank; i++ ) {
4467: NEXTNODE(r0,r); BDY(r) =
4468: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4469: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4470: GC_free(spmat[i]);
4471: }
1.109 noro 4472: if ( r0 ) NEXT(r) = 0;
1.74 noro 4473: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4474: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4475: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4476: if ( DP_Print ) {
1.80 noro 4477: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4478: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4479: nsp,nred,sprow,spcol,rank);
4480: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4481: }
1.74 noro 4482: return r0;
4483: }
4484:
1.113 noro 4485: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,UINT *s0vect,int col,
1.107 noro 4486: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4487: {
4488: int spcol,sprow,a;
4489: int i,j,k,l,rank;
4490: NODE r0,r;
4491: ND_pairs sp;
4492: ND spol;
1.113 noro 4493: Q **spmat;
4494: Q *svect,*v;
1.107 noro 4495: int *colstat;
4496: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4497: int maxrs;
4498: int *spsugar;
4499:
4500: spcol = col-nred;
4501: get_eg(&eg0);
4502: /* elimination (1st step) */
1.113 noro 4503: spmat = (Q **)ALLOCA(nsp*sizeof(Q *));
4504: svect = (Q *)ALLOCA(col*sizeof(Q));
4505: spsugar = (int *)ALLOCA(nsp*sizeof(Q));
1.107 noro 4506: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4507: nd_sp(0,0,sp,&spol);
4508: if ( !spol ) continue;
1.113 noro 4509: nd_to_vect_q(s0vect,col,spol,svect);
4510: maxrs = ndv_reduce_vect_q(svect,col,imat,rvect,nred);
1.107 noro 4511: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4512: if ( i < col ) {
1.113 noro 4513: spmat[sprow] = v = (Q *)MALLOC(spcol*sizeof(Q));
1.107 noro 4514: for ( j = k = 0; j < col; j++ )
4515: if ( !rhead[j] ) v[k++] = svect[j];
4516: spsugar[sprow] = MAX(maxrs,SG(spol));
4517: sprow++;
4518: }
1.109 noro 4519: /* nd_free(spol); */
1.107 noro 4520: }
4521: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4522: if ( DP_Print ) {
4523: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4524: fflush(asir_out);
4525: }
4526: /* free index arrays */
1.109 noro 4527: /* for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c); */
1.107 noro 4528:
4529: /* elimination (2nd step) */
4530: colstat = (int *)ALLOCA(spcol*sizeof(int));
1.113 noro 4531: rank = nd_gauss_elim_q(spmat,spsugar,sprow,spcol,colstat);
1.107 noro 4532: r0 = 0;
4533: for ( i = 0; i < rank; i++ ) {
4534: NEXTNODE(r0,r); BDY(r) =
1.113 noro 4535: (pointer)vect_to_ndv_q(spmat[i],spcol,col,rhead,s0vect);
1.107 noro 4536: SG((NDV)BDY(r)) = spsugar[i];
1.109 noro 4537: /* GC_free(spmat[i]); */
1.107 noro 4538: }
1.109 noro 4539: if ( r0 ) NEXT(r) = 0;
4540:
4541: /* for ( ; i < sprow; i++ ) GC_free(spmat[i]); */
1.107 noro 4542: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4543: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
4544: if ( DP_Print ) {
4545: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
4546: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4547: nsp,nred,sprow,spcol,rank);
4548: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
4549: }
4550: return r0;
4551: }
4552:
1.74 noro 4553: FILE *nd_write,*nd_read;
4554:
4555: void nd_send_int(int a) {
4556: write_int(nd_write,&a);
4557: }
4558:
4559: void nd_send_intarray(int *p,int len) {
4560: write_intarray(nd_write,p,len);
4561: }
4562:
4563: int nd_recv_int() {
4564: int a;
4565:
4566: read_int(nd_read,&a);
4567: return a;
4568: }
4569:
4570: void nd_recv_intarray(int *p,int len) {
4571: read_intarray(nd_read,p,len);
4572: }
4573:
4574: void nd_send_ndv(NDV p) {
4575: int len,i;
4576: NMV m;
4577:
4578: if ( !p ) nd_send_int(0);
4579: else {
4580: len = LEN(p);
4581: nd_send_int(len);
4582: m = BDY(p);
4583: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4584: nd_send_int(CM(m));
4585: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4586: }
1.74 noro 4587: }
4588: }
4589:
4590: void nd_send_nd(ND p) {
4591: int len,i;
4592: NM m;
4593:
4594: if ( !p ) nd_send_int(0);
4595: else {
4596: len = LEN(p);
4597: nd_send_int(len);
4598: m = BDY(p);
4599: for ( i = 0; i < len; i++, m = NEXT(m) ) {
4600: nd_send_int(CM(m));
4601: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4602: }
1.74 noro 4603: }
4604: }
1.65 noro 4605:
1.74 noro 4606: NDV nd_recv_ndv()
4607: {
4608: int len,i;
4609: NMV m,m0;
4610: NDV r;
1.65 noro 4611:
1.74 noro 4612: len = nd_recv_int();
4613: if ( !len ) return 0;
4614: else {
4615: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4616: #if 0
1.74 noro 4617: ndv_alloc += len*nmv_adv;
1.103 noro 4618: #endif
1.74 noro 4619: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4620: CM(m) = nd_recv_int();
4621: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4622: }
1.74 noro 4623: MKNDV(nd_nvar,m0,len,r);
4624: return r;
4625: }
4626: }
1.65 noro 4627:
1.74 noro 4628: int ox_exec_f4_red(Q proc)
4629: {
4630: Obj obj;
4631: STRING fname;
4632: NODE arg;
4633: int s;
4634: extern int ox_need_conv,ox_file_io;
4635:
4636: MKSTR(fname,"nd_exec_f4_red");
4637: arg = mknode(2,proc,fname);
4638: Pox_cmo_rpc(arg,&obj);
4639: s = get_ox_server_id(QTOS(proc));
4640: nd_write = iofp[s].out;
4641: nd_read = iofp[s].in;
4642: ox_need_conv = ox_file_io = 0;
4643: return s;
4644: }
4645:
4646: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
4647: {
4648: int nsp,nred;
4649: int i,rank,s;
4650: NODE rp,r0,r;
4651: ND_pairs sp;
4652: NM_ind_pair pair;
4653: NMV nmv;
4654: NM nm;
4655: NDV nf;
4656: Obj proc,dmy;
4657:
4658: ox_launch_main(0,0,&proc);
4659: s = ox_exec_f4_red((Q)proc);
4660:
4661: nd_send_int(m);
4662: nd_send_int(nd_nvar);
4663: nd_send_int(nd_bpe);
4664: nd_send_int(nd_wpd);
4665: nd_send_int(nmv_adv);
4666:
1.89 noro 4667: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
1.74 noro 4668:
4669: nd_send_int(nd_psn);
4670: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
4671:
4672: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4673: nd_send_int(nsp);
4674: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
4675: nd_send_int(sp->i1); nd_send_int(sp->i2);
4676: }
4677:
4678: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
4679:
4680: nred = length(rp0); nd_send_int(nred);
4681: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
4682: pair = (NM_ind_pair)BDY(rp);
4683: nd_send_int(pair->index);
4684: nd_send_intarray(pair->mul->dl,nd_wpd);
4685: }
4686: fflush(nd_write);
4687: rank = nd_recv_int();
4688: fprintf(asir_out,"rank=%d\n",rank);
4689: r0 = 0;
4690: for ( i = 0; i < rank; i++ ) {
4691: nf = nd_recv_ndv();
4692: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
4693: }
4694: Pox_shutdown(mknode(1,proc),&dmy);
4695: return r0;
4696: }
4697:
4698: /* server side */
4699:
4700: void nd_exec_f4_red_dist()
4701: {
4702: int m,i,nsp,col,s0size,nred,spcol,j,k;
4703: NM_ind_pair *rp0;
4704: NDV nf;
4705: UINT *s0vect;
4706: IndArray *imat;
4707: int *rhead;
4708: int **spmat;
4709: UINT *svect,*v;
4710: ND_pairs *sp0;
4711: int *colstat;
4712: int a,sprow,rank;
1.89 noro 4713: struct order_spec *ord;
1.74 noro 4714: Obj ordspec;
4715: ND spol;
1.76 noro 4716: int maxrs;
4717: int *spsugar;
1.74 noro 4718:
4719: nd_read = iofp[0].in;
4720: nd_write = iofp[0].out;
4721: m = nd_recv_int();
4722: nd_nvar = nd_recv_int();
4723: nd_bpe = nd_recv_int();
4724: nd_wpd = nd_recv_int();
4725: nmv_adv = nd_recv_int();
4726:
4727: loadobj(nd_read,&ordspec);
1.89 noro 4728: create_order_spec(0,ordspec,&ord);
4729: nd_init_ord(ord);
1.74 noro 4730: nd_setup_parameters(nd_nvar,0);
4731:
4732: nd_psn = nd_recv_int();
4733: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
4734: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
4735: for ( i = 0; i < nd_psn; i++ ) {
4736: nd_ps[i] = nd_recv_ndv();
4737: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
4738: }
4739:
4740: nsp = nd_recv_int();
4741: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
4742: for ( i = 0; i < nsp; i++ ) {
4743: NEWND_pairs(sp0[i]);
4744: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
4745: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
4746: }
4747:
4748: col = nd_recv_int();
4749: s0size = col*nd_wpd;
4750: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
4751: nd_recv_intarray(s0vect,s0size);
4752:
4753: nred = nd_recv_int();
4754: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
4755: for ( i = 0; i < nred; i++ ) {
4756: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
4757: rp0[i]->index = nd_recv_int();
4758: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
4759: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
4760: }
4761:
4762: spcol = col-nred;
4763: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
4764: rhead = (int *)MALLOC(col*sizeof(int));
4765: for ( i = 0; i < col; i++ ) rhead[i] = 0;
4766:
4767: /* construction of index arrays */
4768: for ( i = 0; i < nred; i++ ) {
4769: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
4770: rhead[imat[i]->head] = 1;
4771: }
4772:
4773: /* elimination (1st step) */
4774: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
4775: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 4776: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4777: for ( a = sprow = 0; a < nsp; a++ ) {
4778: nd_sp(m,0,sp0[a],&spol);
1.75 noro 4779: if ( !spol ) continue;
1.74 noro 4780: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4781: if ( m == -1 )
4782: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
4783: else
4784: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 4785: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4786: if ( i < col ) {
4787: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
4788: for ( j = k = 0; j < col; j++ )
4789: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4790: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4791: sprow++;
4792: }
1.76 noro 4793: nd_free(spol);
1.74 noro 4794: }
4795: /* elimination (2nd step) */
4796: colstat = (int *)ALLOCA(spcol*sizeof(int));
4797: if ( m == -1 )
1.76 noro 4798: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4799: else
1.76 noro 4800: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4801: nd_send_int(rank);
4802: for ( i = 0; i < rank; i++ ) {
4803: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
4804: nd_send_ndv(nf);
1.63 noro 4805: }
1.74 noro 4806: fflush(nd_write);
1.107 noro 4807: }
4808:
1.113 noro 4809: int nd_gauss_elim_q(Q **mat0,int *sugar,int row,int col,int *colstat)
1.107 noro 4810: {
1.109 noro 4811: int mod,i,j,t,c,rank,rank0,inv;
4812: int *ci,*ri;
1.113 noro 4813: Q dn;
1.109 noro 4814: MAT m,nm;
1.108 noro 4815: int **wmat;
4816:
4817: /* XXX */
4818: mod = 99999989;
4819: wmat = (int **)ALLOCA(row*sizeof(int *));
4820: for ( i = 0; i < row; i++ ) {
4821: wmat[i] = (int *)ALLOCA(col*sizeof(int));
4822: for ( j = 0; j < col; j++ ) {
1.113 noro 4823: if ( mat0[i][j] ) {
4824: t = rem(NM(mat0[i][j]),mod);
4825: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
4826: wmat[i][j] = t;
4827: } else
1.108 noro 4828: wmat[i][j] = 0;
4829: }
4830: }
1.109 noro 4831: rank0 = nd_gauss_elim_mod(wmat,sugar,row,col,mod,colstat);
4832: NEWMAT(m); m->row = row; m->col = col; m->body = (pointer **)mat0;
1.113 noro 4833: rank = generic_gauss_elim(m,&nm,&dn,&ri,&ci);
1.109 noro 4834: if ( rank != rank0 )
4835: error("afo");
4836: for ( i = 0; i < row; i++ )
4837: for ( j = 0; j < col; j++ )
4838: mat0[i][j] = 0;
4839: c = col-rank;
4840: for ( i = 0; i < rank; i++ ) {
1.113 noro 4841: mat0[i][ri[i]] = dn;
1.109 noro 4842: for ( j = 0; j < c; j++ )
1.113 noro 4843: mat0[i][ci[j]] = (Q)BDY(nm)[i][j];
1.109 noro 4844: }
1.113 noro 4845: inv = invm(rem(NM(dn),mod),mod);
4846: if ( SGN(dn) < 0 ) inv = mod-inv;
1.109 noro 4847: for ( i = 0; i < row; i++ )
4848: for ( j = 0; j < col; j++ ) {
1.113 noro 4849: if ( mat0[i][j] ) {
4850: t = rem(NM(mat0[i][j]),mod);
4851: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
4852: } else
1.109 noro 4853: t = 0;
4854: c = dmar(t,inv,0,mod);
4855: if ( wmat[i][j] != c )
4856: error("afo");
4857: }
4858: return rank;
1.76 noro 4859: }
4860:
4861: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4862: {
4863: int i,j,k,l,inv,a,rank,s;
4864: unsigned int *t,*pivot,*pk;
4865: unsigned int **mat;
4866:
4867: mat = (unsigned int **)mat0;
4868: for ( rank = 0, j = 0; j < col; j++ ) {
4869: for ( i = rank; i < row; i++ )
4870: mat[i][j] %= md;
4871: for ( i = rank; i < row; i++ )
4872: if ( mat[i][j] )
4873: break;
4874: if ( i == row ) {
4875: colstat[j] = 0;
4876: continue;
4877: } else
4878: colstat[j] = 1;
4879: if ( i != rank ) {
4880: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4881: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4882: }
4883: pivot = mat[rank];
4884: s = sugar[rank];
4885: inv = invm(pivot[j],md);
4886: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4887: if ( *pk ) {
4888: if ( *pk >= (unsigned int)md )
4889: *pk %= md;
4890: DMAR(*pk,inv,0,md,*pk)
4891: }
4892: for ( i = rank+1; i < row; i++ ) {
4893: t = mat[i];
4894: if ( a = t[j] ) {
4895: sugar[i] = MAX(sugar[i],s);
4896: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4897: }
4898: }
4899: rank++;
4900: }
4901: for ( j = col-1, l = rank-1; j >= 0; j-- )
4902: if ( colstat[j] ) {
4903: pivot = mat[l];
4904: s = sugar[l];
4905: for ( i = 0; i < l; i++ ) {
4906: t = mat[i];
4907: t[j] %= md;
4908: if ( a = t[j] ) {
4909: sugar[i] = MAX(sugar[i],s);
4910: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4911: }
4912: }
4913: l--;
4914: }
4915: for ( j = 0, l = 0; l < rank; j++ )
4916: if ( colstat[j] ) {
4917: t = mat[l];
4918: for ( k = j; k < col; k++ )
4919: if ( t[k] >= (unsigned int)md )
4920: t[k] %= md;
4921: l++;
4922: }
4923: return rank;
4924: }
4925:
4926: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4927: {
4928: int i,j,k,l,inv,a,rank,s;
4929: unsigned int *t,*pivot,*pk;
4930: unsigned int **mat;
4931:
4932: mat = (unsigned int **)mat0;
4933: for ( rank = 0, j = 0; j < col; j++ ) {
4934: for ( i = rank; i < row; i++ )
4935: if ( mat[i][j] )
4936: break;
4937: if ( i == row ) {
4938: colstat[j] = 0;
4939: continue;
4940: } else
4941: colstat[j] = 1;
4942: if ( i != rank ) {
4943: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4944: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4945: }
4946: pivot = mat[rank];
4947: s = sugar[rank];
4948: inv = _invsf(pivot[j]);
4949: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4950: if ( *pk )
4951: *pk = _mulsf(*pk,inv);
4952: for ( i = rank+1; i < row; i++ ) {
4953: t = mat[i];
4954: if ( a = t[j] ) {
4955: sugar[i] = MAX(sugar[i],s);
4956: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4957: }
4958: }
4959: rank++;
4960: }
4961: for ( j = col-1, l = rank-1; j >= 0; j-- )
4962: if ( colstat[j] ) {
4963: pivot = mat[l];
4964: s = sugar[l];
4965: for ( i = 0; i < l; i++ ) {
4966: t = mat[i];
4967: if ( a = t[j] ) {
4968: sugar[i] = MAX(sugar[i],s);
4969: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4970: }
4971: }
4972: l--;
4973: }
4974: return rank;
1.77 noro 4975: }
4976:
4977: int ndv_ishomo(NDV p)
4978: {
4979: NMV m;
4980: int len,h;
4981:
4982: if ( !p ) return 1;
4983: len = LEN(p);
4984: m = BDY(p);
4985: h = TD(DL(m));
4986: NMV_ADV(m);
4987: for ( len--; len; len--, NMV_ADV(m) )
4988: if ( TD(DL(m)) != h ) return 0;
4989: return 1;
4990: }
4991:
4992: void ndv_save(NDV p,int index)
4993: {
4994: FILE *s;
4995: char name[BUFSIZ];
4996: short id;
4997: int nv,sugar,len,n,i,td,e,j;
4998: NMV m;
4999: unsigned int *dl;
5000:
5001: sprintf(name,"%s/%d",Demand,index);
5002: s = fopen(name,"w");
5003: savevl(s,0);
5004: if ( !p ) {
5005: saveobj(s,0);
5006: return;
5007: }
5008: id = O_DP;
5009: nv = NV(p);
5010: sugar = SG(p);
5011: len = LEN(p);
5012: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
5013: write_int(s,&len);
5014:
5015: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5016: saveobj(s,(Obj)CQ(m));
1.77 noro 5017: dl = DL(m);
5018: td = TD(dl);
5019: write_int(s,&td);
5020: for ( j = 0; j < nv; j++ ) {
5021: e = GET_EXP(dl,j);
5022: write_int(s,&e);
5023: }
5024: }
5025: fclose(s);
5026: }
5027:
5028: NDV ndv_load(int index)
5029: {
5030: FILE *s;
5031: char name[BUFSIZ];
5032: short id;
5033: int nv,sugar,len,n,i,td,e,j;
5034: NDV d;
5035: NMV m0,m;
5036: unsigned int *dl;
5037: Obj obj;
5038:
5039: sprintf(name,"%s/%d",Demand,index);
5040: s = fopen(name,"r");
5041: if ( !s ) return 0;
5042:
5043: skipvl(s);
5044: read_short(s,&id);
5045: if ( !id ) return 0;
5046: read_int(s,&nv);
5047: read_int(s,&sugar);
5048: read_int(s,&len);
5049:
5050: m0 = m = MALLOC(len*nmv_adv);
5051: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5052: loadobj(s,&obj); CQ(m) = (Q)obj;
1.77 noro 5053: dl = DL(m);
5054: ndl_zero(dl);
5055: read_int(s,&td); TD(dl) = td;
5056: for ( j = 0; j < nv; j++ ) {
5057: read_int(s,&e);
5058: PUT_EXP(dl,j,e);
5059: }
5060: if ( nd_blockmask ) ndl_weight_mask(dl);
5061: }
5062: fclose(s);
5063: MKNDV(nv,m0,len,d);
5064: SG(d) = sugar;
5065: return d;
1.99 noro 5066: }
5067:
1.102 noro 5068: void nd_det(int mod,MAT f,P *rp)
1.99 noro 5069: {
5070: VL fv,tv;
1.102 noro 5071: int n,i,j,max,e,nvar,sgn,k0,l0,len0,len,k,l,a;
1.99 noro 5072: pointer **m;
1.113 noro 5073: Q mone;
1.102 noro 5074: NDV **dm;
5075: NDV *t,*mi,*mj;
5076: NDV d,s,mij,mjj;
5077: ND u;
5078: NMV nmv;
1.114 noro 5079: UINT *bound;
1.99 noro 5080: PGeoBucket bucket;
5081: struct order_spec *ord;
5082:
5083: create_order_spec(0,0,&ord);
5084: nd_init_ord(ord);
5085: get_vars((Obj)f,&fv);
5086: if ( f->row != f->col )
5087: error("nd_det : non-square matrix");
5088: n = f->row;
5089: for ( nvar = 0, tv = fv; tv; tv = NEXT(tv), nvar++ );
5090: m = f->body;
5091: for ( i = 0, max = 0; i < n; i++ )
5092: for ( j = 0; j < n; j++ )
5093: for ( tv = fv; tv; tv = NEXT(tv) ) {
5094: e = getdeg(tv->v,(P)m[i][j]);
5095: max = MAX(e,max);
5096: }
1.114 noro 5097: nd_setup_parameters(nvar,max);
1.102 noro 5098: dm = (NDV **)almat_pointer(n,n);
1.99 noro 5099: for ( i = 0, max = 0; i < n; i++ )
1.102 noro 5100: for ( j = 0; j < n; j++ ) {
5101: dm[i][j] = ptondv(CO,fv,m[i][j]);
5102: if ( mod ) ndv_mod(mod,dm[i][j]);
5103: if ( dm[i][j] && !LEN(dm[i][j]) ) dm[i][j] = 0;
5104: }
5105: d = ptondv(CO,fv,(P)ONE);
5106: if ( mod ) ndv_mod(mod,d);
1.113 noro 5107: chsgnq(ONE,&mone);
1.99 noro 5108: for ( j = 0, sgn = 1; j < n; j++ ) {
1.104 noro 5109: if ( DP_Print ) fprintf(stderr,"j=%d\n",j);
1.99 noro 5110: for ( i = j; i < n && !dm[i][j]; i++ );
5111: if ( i == n ) {
5112: *rp = 0;
5113: return;
5114: }
1.102 noro 5115: k0 = i; l0 = j; len0 = LEN(dm[k0][l0]);
1.99 noro 5116: for ( k = j; k < n; k++ )
5117: for ( l = j; l < n; l++ )
1.102 noro 5118: if ( dm[k][l] && LEN(dm[k][l]) < len0 ) {
5119: k0 = k; l0 = l; len0 = LEN(dm[k][l]);
1.99 noro 5120: }
5121: if ( k0 != j ) {
5122: t = dm[j]; dm[j] = dm[k0]; dm[k0] = t;
5123: sgn = -sgn;
5124: }
5125: if ( l0 != j ) {
5126: for ( k = j; k < n; k++ ) {
5127: s = dm[k][j]; dm[k][j] = dm[k][l0]; dm[k][l0] = s;
5128: }
5129: sgn = -sgn;
5130: }
1.114 noro 5131: bound = nd_det_compute_bound(dm,n,j);
5132: if ( ndl_check_bound(bound,bound) )
5133: nd_det_reconstruct(dm,n,j,d);
5134:
1.99 noro 5135: for ( i = j+1, mj = dm[j], mjj = mj[j]; i < n; i++ ) {
1.114 noro 5136: /* if ( DP_Print ) fprintf(stderr," i=%d\n ",i); */
1.99 noro 5137: mi = dm[i]; mij = mi[j];
1.102 noro 5138: if ( mod )
5139: ndv_mul_c(mod,mij,mod-1);
5140: else
1.113 noro 5141: ndv_mul_c_q(mij,mone);
1.99 noro 5142: for ( k = j+1; k < n; k++ ) {
1.114 noro 5143: /* if ( DP_Print ) fprintf(stderr,"k=%d ",k); */
1.99 noro 5144: bucket = create_pbucket();
1.104 noro 5145: if ( mi[k] ) {
1.102 noro 5146: nmv = BDY(mjj); len = LEN(mjj);
5147: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5148: u = ndv_mul_nmv_trunc(mod,nmv,mi[k],DL(BDY(d)));
5149: add_pbucket(mod,bucket,u);
1.99 noro 5150: }
1.104 noro 5151: }
1.99 noro 5152: if ( mj[k] && mij ) {
1.102 noro 5153: nmv = BDY(mij); len = LEN(mij);
5154: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5155: u = ndv_mul_nmv_trunc(mod,nmv,mj[k],DL(BDY(d)));
5156: add_pbucket(mod,bucket,u);
1.99 noro 5157: }
5158: }
1.104 noro 5159: u = nd_quo(mod,bucket,d);
1.102 noro 5160: mi[k] = ndtondv(mod,u);
1.99 noro 5161: }
1.114 noro 5162: /* if ( DP_Print ) fprintf(stderr,"\n",k); */
1.99 noro 5163: }
5164: d = mjj;
5165: }
1.100 noro 5166: if ( sgn < 0 )
1.102 noro 5167: if ( mod )
5168: ndv_mul_c(mod,d,mod-1);
5169: else
1.113 noro 5170: ndv_mul_c_q(d,mone);
1.102 noro 5171: *rp = ndvtop(mod,CO,fv,d);
1.99 noro 5172: }
5173:
1.102 noro 5174: ND ndv_mul_nmv_trunc(int mod,NMV m0,NDV p,UINT *d)
1.99 noro 5175: {
5176: NM mr,mr0;
1.102 noro 5177: NM tnm;
5178: NMV m;
1.99 noro 5179: UINT *d0,*dt,*dm;
5180: int c,n,td,i,c1,c2,len;
1.113 noro 5181: Q q;
1.99 noro 5182: ND r;
5183:
5184: if ( !p ) return 0;
5185: else {
1.102 noro 5186: n = NV(p); m = BDY(p); len = LEN(p);
1.99 noro 5187: d0 = DL(m0);
1.102 noro 5188: td = TD(d);
1.99 noro 5189: mr0 = 0;
5190: NEWNM(tnm);
1.102 noro 5191: if ( mod ) {
5192: c = CM(m0);
5193: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5194: ndl_add(DL(m),d0,DL(tnm));
5195: if ( ndl_reducible(DL(tnm),d) ) {
5196: NEXTNM(mr0,mr);
5197: c1 = CM(m); DMAR(c1,c,0,mod,c2); CM(mr) = c2;
1.104 noro 5198: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5199: }
5200: }
5201: } else {
1.113 noro 5202: q = CQ(m0);
1.102 noro 5203: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5204: ndl_add(DL(m),d0,DL(tnm));
5205: if ( ndl_reducible(DL(tnm),d) ) {
5206: NEXTNM(mr0,mr);
1.113 noro 5207: mulq(CQ(m),q,&CQ(mr));
1.104 noro 5208: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5209: }
1.99 noro 5210: }
5211: }
5212: if ( !mr0 )
5213: return 0;
5214: else {
5215: NEXT(mr) = 0;
1.105 noro 5216: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
1.99 noro 5217: MKND(NV(p),mr0,len,r);
5218: SG(r) = SG(p) + TD(d0);
5219: return r;
5220: }
5221: }
1.114 noro 5222: }
5223:
5224: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d)
5225: {
5226: int i,obpe,oadv,h,k,l;
5227: static NM prev_nm_free_list;
5228: EPOS oepos;
5229:
5230: obpe = nd_bpe;
5231: oadv = nmv_adv;
5232: oepos = nd_epos;
5233: if ( obpe < 2 ) nd_bpe = 2;
5234: else if ( obpe < 3 ) nd_bpe = 3;
5235: else if ( obpe < 4 ) nd_bpe = 4;
5236: else if ( obpe < 5 ) nd_bpe = 5;
5237: else if ( obpe < 6 ) nd_bpe = 6;
5238: else if ( obpe < 8 ) nd_bpe = 8;
5239: else if ( obpe < 10 ) nd_bpe = 10;
5240: else if ( obpe < 16 ) nd_bpe = 16;
5241: else if ( obpe < 32 ) nd_bpe = 32;
5242: else error("nd_det_reconstruct : exponent too large");
5243:
5244: nd_setup_parameters(nd_nvar,0);
5245: prev_nm_free_list = _nm_free_list;
5246: _nm_free_list = 0;
5247: for ( k = j; k < n; k++ )
5248: for (l = j; l < n; l++ )
5249: ndv_realloc(dm[k][l],obpe,oadv,oepos);
5250: ndv_realloc(d,obpe,oadv,oepos);
5251: prev_nm_free_list = 0;
5252: #if 0
5253: GC_gcollect();
5254: #endif
5255: }
5256:
5257: UINT *nd_det_compute_bound(NDV **dm,int n,int j)
5258: {
5259: UINT *d0,*d1,*d,*t,*r;
5260: int k,l;
5261:
5262: d0 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5263: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5264: for ( k = 0; k < nd_wpd; k++ ) d0[k] = 0;
5265: for ( k = j; k < n; k++ )
5266: for ( l = j; l < n; l++ )
5267: if ( dm[k][l] ) {
5268: d = ndv_compute_bound(dm[k][l]);
5269: ndl_lcm(d,d0,d1);
5270: t = d1; d1 = d0; d0 = t;
5271: }
5272: r = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5273: for ( k = 0; k < nd_wpd; k++ ) r[k] = d0[k];
5274: return r;
1.117 noro 5275: }
5276:
5277: DL nd_separate_d(UINT *d,UINT *trans)
5278: {
1.121 ! noro 5279: int n,td,i,e,j;
1.117 noro 5280: DL a;
5281:
5282: ndl_zero(trans);
5283: td = 0;
1.121 ! noro 5284: for ( i = 0; i < nd_ntrans; i++ ) {
! 5285: e = GET_EXP(d,i);
! 5286: PUT_EXP(trans,i,e);
! 5287: td += MUL_WEIGHT(e,i);
! 5288: }
! 5289: if ( nd_ntrans+nd_nalg < nd_nvar ) {
! 5290: /* homogenized */
! 5291: i = nd_nvar-1;
1.117 noro 5292: e = GET_EXP(d,i);
5293: PUT_EXP(trans,i,e);
5294: td += MUL_WEIGHT(e,i);
5295: }
5296: TD(trans) = td;
5297: if ( nd_blockmask) ndl_weight_mask(trans);
5298: NEWDL(a,nd_nalg);
5299: td = 0;
1.121 ! noro 5300: for ( i = 0; i < nd_nalg; i++ ) {
! 5301: j = nd_ntrans+i;
! 5302: e = GET_EXP(d,j);
! 5303: a->d[i] = e;
1.117 noro 5304: td += e;
5305: }
5306: a->td = td;
5307: return a;
5308: }
5309:
1.118 noro 5310: int nd_monic(int mod,ND *p)
1.117 noro 5311: {
5312: UINT *trans,*t;
5313: DL alg;
5314: MP mp0,mp;
1.118 noro 5315: NM m,m0,m1,ma0,ma,mb,mr0,mr;
5316: ND r;
1.117 noro 5317: DL dl;
5318: DP nm;
5319: NDV ndv;
1.118 noro 5320: DAlg inv,cd;
1.117 noro 5321: ND s,c;
1.118 noro 5322: Q l,mul;
5323: N ln;
5324: int n,ntrans,i,e,td,is_lc,len;
5325: NumberField nf;
5326: struct oEGT eg0,eg1;
5327:
5328: if ( !(nf = get_numberfield()) )
5329: error("nd_monic : current_numberfield is not set");
1.117 noro 5330:
1.118 noro 5331: /* Q coef -> DAlg coef */
5332: NEWNM(ma0); ma = ma0;
5333: m = BDY(*p);
5334: is_lc = 1;
5335: while ( 1 ) {
5336: NEWMP(mp0); mp = mp0;
5337: mp->c = (P)CQ(m);
5338: mp->dl = nd_separate_d(DL(m),DL(ma));
5339: NEWNM(mb);
5340: for ( m = NEXT(m); m; m = NEXT(m) ) {
5341: alg = nd_separate_d(DL(m),DL(mb));
5342: if ( !ndl_equal(DL(ma),DL(mb)) )
5343: break;
5344: NEXTMP(mp0,mp); mp->c = (P)CQ(m); mp->dl = alg;
5345: }
5346: NEXT(mp) = 0;
5347: MKDP(nd_nalg,mp0,nm);
5348: MKDAlg(nm,ONE,cd);
5349: if ( is_lc == 1 ) {
5350: /* if the lc is a rational number, we have nothing to do */
5351: if ( !mp0->dl->td )
5352: return 1;
5353:
5354: get_eg(&eg0);
5355: invdalg(cd,&inv);
5356: get_eg(&eg1); add_eg(&eg_invdalg,&eg0,&eg1);
5357: /* check the validity of inv */
5358: if ( mod && !rem(NM(inv->dn),mod) )
5359: return 0;
5360: CA(ma) = nf->one;
5361: is_lc = 0;
5362: ln = ONEN;
5363: } else {
5364: muldalg(cd,inv,&CA(ma));
5365: lcmn(ln,NM(CA(ma)->dn),&ln);
5366: }
5367: if ( m ) {
5368: NEXT(ma) = mb; ma = mb;
5369: } else {
5370: NEXT(ma) = 0;
5371: break;
5372: }
5373: }
5374: /* l = lcm(denoms) */
5375: NTOQ(ln,1,l);
5376: for ( mr0 = 0, m = ma0; m; m = NEXT(m) ) {
5377: divq(l,CA(m)->dn,&mul);
5378: for ( mp = BDY(CA(m)->nm); mp; mp = NEXT(mp) ) {
5379: NEXTNM(mr0,mr);
5380: mulq((Q)mp->c,mul,&CQ(mr));
5381: dl = mp->dl;
5382: td = TD(DL(m));
5383: ndl_copy(DL(m),DL(mr));
1.121 ! noro 5384: for ( i = 0; i < nd_nalg; i++ ) {
! 5385: e = dl->d[i];
! 5386: PUT_EXP(DL(mr),i+nd_ntrans,e);
! 5387: td += MUL_WEIGHT(e,i+nd_ntrans);
1.118 noro 5388: }
5389: TD(DL(mr)) = td;
5390: if ( nd_blockmask) ndl_weight_mask(DL(mr));
5391: }
1.117 noro 5392: }
1.118 noro 5393: NEXT(mr) = 0;
5394: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
5395: MKND(NV(*p),mr0,len,r);
5396: /* XXX */
5397: SG(r) = SG(*p);
5398: nd_free(*p);
5399: *p = r;
5400: return 1;
1.59 noro 5401: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>