Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.131
1.131 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.130 2006/05/30 07:35:31 noro Exp $ */
1.2 noro 2:
1.94 noro 3: #include "nd.h"
1.63 noro 4:
1.131 ! noro 5: int diag_period = 6;
1.61 noro 6: int (*ndl_compare_function)(UINT *a1,UINT *a2);
1.94 noro 7: int nd_dcomp;
8: NM _nm_free_list;
9: ND _nd_free_list;
10: ND_pairs _ndp_free_list;
1.32 noro 11:
1.121 noro 12: static int nd_ntrans;
1.117 noro 13: static int nd_nalg;
1.103 noro 14: #if 0
1.74 noro 15: static int ndv_alloc;
1.103 noro 16: #endif
1.87 noro 17: #if 1
1.69 noro 18: static int nd_f4_nsp=0x7fffffff;
1.87 noro 19: #else
20: static int nd_f4_nsp=50;
21: #endif
1.42 noro 22: static double nd_scale=2;
1.61 noro 23: static UINT **nd_bound;
1.42 noro 24: static struct order_spec *nd_ord;
25: static EPOS nd_epos;
1.43 noro 26: static BlockMask nd_blockmask;
1.42 noro 27: static int nd_nvar;
28: static int nd_isrlex;
29: static int nd_epw,nd_bpe,nd_wpd,nd_exporigin;
1.61 noro 30: static UINT nd_mask[32];
31: static UINT nd_mask0,nd_mask1;
1.42 noro 32:
1.20 noro 33: static NDV *nd_ps;
1.53 noro 34: static NDV *nd_ps_trace;
1.42 noro 35: static RHist *nd_psh;
36: static int nd_psn,nd_pslen;
37: static RHist *nd_red;
1.96 noro 38: static int *nd_work_vector;
39: static int **nd_matrix;
40: static int nd_matrix_len;
1.97 noro 41: static struct weight_or_block *nd_worb;
42: static int nd_worb_len;
1.42 noro 43: static int nd_found,nd_create,nd_notfirst;
44: static int nmv_adv;
1.77 noro 45: static int nd_demand;
1.1 noro 46:
1.119 noro 47: NumberField get_numberfield();
1.114 noro 48: UINT *nd_det_compute_bound(NDV **dm,int n,int j);
49: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d);
1.118 noro 50: int nd_monic(int m,ND *p);
1.129 noro 51: NDV plain_vect_to_ndv_q(Q *mat,int col,UINT *s0vect);
1.114 noro 52:
1.1 noro 53: void nd_free_private_storage()
54: {
55: _nm_free_list = 0;
1.5 noro 56: _ndp_free_list = 0;
1.71 noro 57: #if 0
1.1 noro 58: GC_gcollect();
1.71 noro 59: #endif
1.1 noro 60: }
61:
62: void _NM_alloc()
63: {
64: NM p;
65: int i;
66:
1.11 noro 67: for ( i = 0; i < 1024; i++ ) {
1.61 noro 68: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 69: p->next = _nm_free_list; _nm_free_list = p;
70: }
71: }
72:
73: void _ND_alloc()
74: {
75: ND p;
76: int i;
77:
78: for ( i = 0; i < 1024; i++ ) {
79: p = (ND)GC_malloc(sizeof(struct oND));
80: p->body = (NM)_nd_free_list; _nd_free_list = p;
81: }
82: }
83:
84: void _NDP_alloc()
85: {
86: ND_pairs p;
87: int i;
88:
1.11 noro 89: for ( i = 0; i < 1024; i++ ) {
1.1 noro 90: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
1.61 noro 91: +(nd_wpd-1)*sizeof(UINT));
1.1 noro 92: p->next = _ndp_free_list; _ndp_free_list = p;
93: }
94: }
95:
1.30 noro 96: INLINE int nd_length(ND p)
1.1 noro 97: {
98: NM m;
99: int i;
100:
101: if ( !p )
102: return 0;
103: else {
104: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
105: return i;
106: }
107: }
108:
1.61 noro 109: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 110: {
1.61 noro 111: UINT u1,u2;
1.1 noro 112: int i,j;
113:
1.34 noro 114: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 115: #if USE_UNROLL
1.1 noro 116: switch ( nd_bpe ) {
1.62 noro 117: case 3:
118: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
119: u1 = d1[i]; u2 = d2[i];
120: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
121: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
122: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
123: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
124: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
125: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
126: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
127: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
128: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
129: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
130: }
131: return 1;
132: break;
1.1 noro 133: case 4:
1.41 noro 134: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 135: u1 = d1[i]; u2 = d2[i];
136: if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0;
1.62 noro 137: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
138: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
139: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
140: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
141: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
142: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
143: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
1.1 noro 144: }
145: return 1;
146: break;
147: case 6:
1.41 noro 148: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 149: u1 = d1[i]; u2 = d2[i];
150: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
1.62 noro 151: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
152: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
153: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
154: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
1.1 noro 155: }
156: return 1;
157: break;
158: case 8:
1.41 noro 159: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 160: u1 = d1[i]; u2 = d2[i];
161: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
1.62 noro 162: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
163: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
164: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
1.1 noro 165: }
166: return 1;
167: break;
168: case 16:
1.41 noro 169: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 170: u1 = d1[i]; u2 = d2[i];
171: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
1.62 noro 172: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
1.1 noro 173: }
174: return 1;
175: break;
176: case 32:
1.41 noro 177: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 178: if ( d1[i] < d2[i] ) return 0;
179: return 1;
180: break;
181: default:
1.41 noro 182: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 183: u1 = d1[i]; u2 = d2[i];
184: for ( j = 0; j < nd_epw; j++ )
185: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
186: }
187: return 1;
188: }
1.65 noro 189: #else
190: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
191: u1 = d1[i]; u2 = d2[i];
192: for ( j = 0; j < nd_epw; j++ )
193: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
194: }
195: return 1;
196: #endif
1.1 noro 197: }
198:
1.61 noro 199: /*
200: * If the current order is a block order,
201: * then the last block is length 1 and contains
202: * the homo variable. Otherwise, the original
203: * order is either 0 or 2.
204: */
205:
206: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 207: {
1.61 noro 208: int w,i,e,n,omask0;
209:
210: omask0 = (1<<obpe)-1;
211: n = nd_nvar-1;
212: ndl_zero(r);
213: for ( i = 0; i < n; i++ ) {
214: e = GET_EXP_OLD(d,i);
215: PUT_EXP(r,i,e);
216: }
217: w = TD(d);
218: PUT_EXP(r,nd_nvar-1,weight-w);
219: TD(r) = weight;
220: if ( nd_blockmask ) ndl_weight_mask(r);
221: }
222:
223: void ndl_dehomogenize(UINT *d)
224: {
225: UINT mask;
226: UINT h;
1.31 noro 227: int i,bits;
1.23 noro 228:
1.44 noro 229: if ( nd_blockmask ) {
230: h = GET_EXP(d,nd_nvar-1);
1.45 noro 231: XOR_EXP(d,nd_nvar-1,h);
1.44 noro 232: TD(d) -= h;
233: d[nd_exporigin-1] -= h;
234: } else {
235: if ( nd_isrlex ) {
236: if ( nd_bpe == 32 ) {
237: h = d[nd_exporigin];
238: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
239: d[i-1] = d[i];
240: d[i-1] = 0;
241: TD(d) -= h;
242: } else {
243: bits = nd_epw*nd_bpe;
244: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
245: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
246: for ( i = nd_exporigin; i < nd_wpd; i++ )
247: d[i] = ((d[i]<<nd_bpe)&mask)
248: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
249: TD(d) -= h;
250: }
1.45 noro 251: } else {
252: h = GET_EXP(d,nd_nvar-1);
253: XOR_EXP(d,nd_nvar-1,h);
254: TD(d) -= h;
255: }
1.44 noro 256: }
1.23 noro 257: }
258:
1.61 noro 259: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 260: {
1.61 noro 261: UINT t1,t2,u,u1,u2;
1.43 noro 262: int i,j,l;
1.1 noro 263:
1.65 noro 264: #if USE_UNROLL
1.1 noro 265: switch ( nd_bpe ) {
1.62 noro 266: case 3:
267: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
268: u1 = d1[i]; u2 = d2[i];
269: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
270: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
271: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
272: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
273: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
274: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
275: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
276: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
277: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
278: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
279: d[i] = u;
280: }
281: break;
1.1 noro 282: case 4:
1.41 noro 283: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 284: u1 = d1[i]; u2 = d2[i];
285: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
1.62 noro 286: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
287: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
288: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
289: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
290: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
291: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
292: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
1.1 noro 293: d[i] = u;
294: }
295: break;
296: case 6:
1.41 noro 297: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 298: u1 = d1[i]; u2 = d2[i];
299: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
1.62 noro 300: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
301: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
302: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
303: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
1.1 noro 304: d[i] = u;
305: }
306: break;
307: case 8:
1.41 noro 308: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 309: u1 = d1[i]; u2 = d2[i];
310: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
1.62 noro 311: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
312: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
313: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
1.1 noro 314: d[i] = u;
315: }
316: break;
317: case 16:
1.41 noro 318: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 319: u1 = d1[i]; u2 = d2[i];
320: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
1.62 noro 321: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
1.1 noro 322: d[i] = u;
323: }
324: break;
325: case 32:
1.41 noro 326: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 327: u1 = d1[i]; u2 = d2[i];
328: d[i] = u1>u2?u1:u2;
329: }
330: break;
331: default:
1.41 noro 332: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 333: u1 = d1[i]; u2 = d2[i];
334: for ( j = 0, u = 0; j < nd_epw; j++ ) {
335: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
336: }
337: d[i] = u;
338: }
339: break;
340: }
1.65 noro 341: #else
342: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
343: u1 = d1[i]; u2 = d2[i];
344: for ( j = 0, u = 0; j < nd_epw; j++ ) {
345: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
346: }
347: d[i] = u;
348: }
349: #endif
1.39 noro 350: TD(d) = ndl_weight(d);
1.61 noro 351: if ( nd_blockmask ) ndl_weight_mask(d);
1.57 noro 352: }
353:
1.61 noro 354: int ndl_weight(UINT *d)
1.1 noro 355: {
1.61 noro 356: UINT t,u;
1.1 noro 357: int i,j;
358:
1.60 noro 359: if ( current_dl_weight_vector )
360: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
361: u = GET_EXP(d,i);
362: t += MUL_WEIGHT(u,i);
363: }
364: else
365: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
366: u = d[i];
367: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
368: t += (u&nd_mask0);
369: }
1.1 noro 370: return t;
371: }
372:
1.61 noro 373: void ndl_weight_mask(UINT *d)
1.43 noro 374: {
1.61 noro 375: UINT t,u;
376: UINT *mask;
377: int i,j,k,l;
1.43 noro 378:
1.61 noro 379: l = nd_blockmask->n;
380: for ( k = 0; k < l; k++ ) {
381: mask = nd_blockmask->mask[k];
382: if ( current_dl_weight_vector )
383: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
384: u = GET_EXP_MASK(d,i,mask);
385: t += MUL_WEIGHT(u,i);
386: }
387: else
388: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
389: u = d[i]&mask[i];
390: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
391: t += (u&nd_mask0);
392: }
393: d[k+1] = t;
394: }
1.43 noro 395: }
396:
1.61 noro 397: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 398: {
399: int i;
400:
1.41 noro 401: d1 += nd_exporigin;
402: d2 += nd_exporigin;
403: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
1.1 noro 404: if ( *d1 > *d2 )
1.32 noro 405: return nd_isrlex ? -1 : 1;
1.1 noro 406: else if ( *d1 < *d2 )
1.32 noro 407: return nd_isrlex ? 1 : -1;
1.1 noro 408: return 0;
409: }
410:
1.61 noro 411: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 412: {
413: int i,l,j,ord_o,ord_l;
414: struct order_pair *op;
1.61 noro 415: UINT t1,t2,m;
416: UINT *mask;
1.43 noro 417:
418: l = nd_blockmask->n;
419: op = nd_blockmask->order_pair;
420: for ( j = 0; j < l; j++ ) {
421: mask = nd_blockmask->mask[j];
422: ord_o = op[j].order;
423: if ( ord_o < 2 )
1.44 noro 424: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
425: else if ( t1 < t2 ) return -1;
1.43 noro 426: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.44 noro 427: m = mask[i];
428: t1 = d1[i]&m;
429: t2 = d2[i]&m;
1.43 noro 430: if ( t1 > t2 )
431: return !ord_o ? -1 : 1;
432: else if ( t1 < t2 )
433: return !ord_o ? 1 : -1;
434: }
435: }
436: return 0;
437: }
438:
1.96 noro 439: int ndl_matrix_compare(UINT *d1,UINT *d2)
440: {
441: int i,j,s;
442: int *v;
443:
444: for ( j = 0; j < nd_nvar; j++ )
445: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
446: for ( i = 0; i < nd_matrix_len; i++ ) {
447: v = nd_matrix[i];
448: for ( j = 0, s = 0; j < nd_nvar; j++ )
449: s += v[j]*nd_work_vector[j];
450: if ( s > 0 ) return 1;
451: else if ( s < 0 ) return -1;
452: }
453: return 0;
454: }
455:
1.97 noro 456: int ndl_composite_compare(UINT *d1,UINT *d2)
457: {
458: int i,j,s,start,end,len,o;
459: int *v;
460: struct sparse_weight *sw;
461:
462: for ( j = 0; j < nd_nvar; j++ )
463: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
464: for ( i = 0; i < nd_worb_len; i++ ) {
465: len = nd_worb[i].length;
466: switch ( nd_worb[i].type ) {
467: case IS_DENSE_WEIGHT:
468: v = nd_worb[i].body.dense_weight;
469: for ( j = 0, s = 0; j < len; j++ )
470: s += v[j]*nd_work_vector[j];
471: if ( s > 0 ) return 1;
472: else if ( s < 0 ) return -1;
473: break;
474: case IS_SPARSE_WEIGHT:
475: sw = nd_worb[i].body.sparse_weight;
476: for ( j = 0, s = 0; j < len; j++ )
477: s += sw[j].value*nd_work_vector[sw[j].pos];
478: if ( s > 0 ) return 1;
479: else if ( s < 0 ) return -1;
480: break;
481: case IS_BLOCK:
482: o = nd_worb[i].body.block.order;
483: start = nd_worb[i].body.block.start;
484: switch ( o ) {
485: case 0:
486: end = start+len;
487: for ( j = start, s = 0; j < end; j++ )
488: s += MUL_WEIGHT(nd_work_vector[j],j);
489: if ( s > 0 ) return 1;
490: else if ( s < 0 ) return -1;
491: for ( j = end-1; j >= start; j-- )
492: if ( nd_work_vector[j] < 0 ) return 1;
493: else if ( nd_work_vector[j] > 0 ) return -1;
494: break;
495: case 1:
496: end = start+len;
497: for ( j = start, s = 0; j < end; j++ )
498: s += MUL_WEIGHT(nd_work_vector[j],j);
499: if ( s > 0 ) return 1;
500: else if ( s < 0 ) return -1;
501: for ( j = start; j < end; j++ )
502: if ( nd_work_vector[j] > 0 ) return 1;
503: else if ( nd_work_vector[j] < 0 ) return -1;
504: break;
505: case 2:
506: for ( j = start; j < end; j++ )
507: if ( nd_work_vector[j] > 0 ) return 1;
508: else if ( nd_work_vector[j] < 0 ) return -1;
509: break;
510: }
511: break;
512: }
513: }
514: return 0;
515: }
516:
1.58 noro 517: /* TDH -> WW -> TD-> RL */
518:
1.61 noro 519: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 520: {
521: int i,m,e1,e2;
522:
523: if ( TD(d1) > TD(d2) ) return 1;
524: else if ( TD(d1) < TD(d2) ) return -1;
525: m = nd_nvar>>1;
526: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
527: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
528: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
529: }
530: if ( e1 > e2 ) return 1;
531: else if ( e1 < e2 ) return -1;
532: return ndl_lex_compare(d1,d2);
533: }
534:
1.61 noro 535: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 536: {
537: int i;
538:
1.81 noro 539: switch ( nd_wpd ) {
540: case 2:
541: if ( TD(d2) != TD(d1) ) return 0;
542: if ( d2[1] != d1[1] ) return 0;
543: return 1;
544: break;
545: case 3:
546: if ( TD(d2) != TD(d1) ) return 0;
547: if ( d2[1] != d1[1] ) return 0;
548: if ( d2[2] != d1[2] ) return 0;
549: return 1;
550: break;
551: default:
552: for ( i = 0; i < nd_wpd; i++ )
553: if ( *d1++ != *d2++ ) return 0;
554: return 1;
555: break;
556: }
1.1 noro 557: }
558:
1.61 noro 559: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 560: {
561: int i;
562:
563: switch ( nd_wpd ) {
1.41 noro 564: case 2:
1.34 noro 565: TD(d2) = TD(d1);
566: d2[1] = d1[1];
1.6 noro 567: break;
1.41 noro 568: case 3:
1.34 noro 569: TD(d2) = TD(d1);
1.6 noro 570: d2[1] = d1[1];
1.34 noro 571: d2[2] = d1[2];
1.6 noro 572: break;
573: default:
1.41 noro 574: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 575: d2[i] = d1[i];
576: break;
577: }
578: }
579:
1.61 noro 580: INLINE void ndl_zero(UINT *d)
581: {
582: int i;
583: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
584: }
585:
586: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 587: {
588: int i;
589:
1.43 noro 590: #if 1
1.6 noro 591: switch ( nd_wpd ) {
1.41 noro 592: case 2:
593: TD(d) = TD(d1)+TD(d2);
1.34 noro 594: d[1] = d1[1]+d2[1];
1.6 noro 595: break;
1.41 noro 596: case 3:
597: TD(d) = TD(d1)+TD(d2);
1.6 noro 598: d[1] = d1[1]+d2[1];
1.34 noro 599: d[2] = d1[2]+d2[2];
1.6 noro 600: break;
601: default:
1.43 noro 602: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 603: break;
604: }
1.43 noro 605: #else
606: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
607: #endif
1.6 noro 608: }
609:
1.55 noro 610: /* d1 += d2 */
1.61 noro 611: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 612: {
613: int i;
614:
615: #if 1
616: switch ( nd_wpd ) {
617: case 2:
618: TD(d1) += TD(d2);
619: d1[1] += d2[1];
620: break;
621: case 3:
622: TD(d1) += TD(d2);
623: d1[1] += d2[1];
624: d1[2] += d2[2];
625: break;
626: default:
627: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
628: break;
629: }
630: #else
631: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
632: #endif
633: }
634:
1.61 noro 635: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 636: {
637: int i;
638:
1.43 noro 639: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 640: }
641:
1.61 noro 642: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 643: {
1.61 noro 644: UINT t1,t2,u,u1,u2;
1.1 noro 645: int i,j;
646:
1.65 noro 647: #if USE_UNROLL
1.1 noro 648: switch ( nd_bpe ) {
1.62 noro 649: case 3:
650: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
651: u1 = d1[i]; u2 = d2[i];
652: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
653: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
654: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
655: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
656: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
657: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
658: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
659: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
660: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
661: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
662: }
663: return 1;
664: break;
1.1 noro 665: case 4:
1.41 noro 666: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 667: u1 = d1[i]; u2 = d2[i];
668: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 669: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
670: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
671: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
672: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
673: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
674: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
675: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 676: }
677: return 1;
678: break;
679: case 6:
1.41 noro 680: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 681: u1 = d1[i]; u2 = d2[i];
682: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 683: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
684: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
685: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
686: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 687: }
688: return 1;
689: break;
690: case 8:
1.41 noro 691: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 692: u1 = d1[i]; u2 = d2[i];
693: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 694: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
695: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
696: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 697: }
698: return 1;
699: break;
700: case 16:
1.41 noro 701: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 702: u1 = d1[i]; u2 = d2[i];
703: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 704: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 705: }
706: return 1;
707: break;
708: case 32:
1.41 noro 709: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 710: if ( d1[i] && d2[i] ) return 0;
711: return 1;
712: break;
713: default:
1.41 noro 714: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 715: u1 = d1[i]; u2 = d2[i];
716: for ( j = 0; j < nd_epw; j++ ) {
717: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
718: u1 >>= nd_bpe; u2 >>= nd_bpe;
719: }
720: }
721: return 1;
722: break;
723: }
1.65 noro 724: #else
725: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
726: u1 = d1[i]; u2 = d2[i];
727: for ( j = 0; j < nd_epw; j++ ) {
728: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
729: u1 >>= nd_bpe; u2 >>= nd_bpe;
730: }
731: }
732: return 1;
733: #endif
1.1 noro 734: }
735:
1.114 noro 736: int ndl_check_bound(UINT *d1,UINT *d2)
1.1 noro 737: {
1.61 noro 738: UINT u2;
1.5 noro 739: int i,j,ind,k;
1.1 noro 740:
1.5 noro 741: ind = 0;
1.65 noro 742: #if USE_UNROLL
1.5 noro 743: switch ( nd_bpe ) {
1.62 noro 744: case 3:
745: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
746: u2 = d2[i];
747: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
748: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
749: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
750: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
751: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
752: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
753: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
754: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
755: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
756: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
757: }
758: return 0;
759: break;
1.5 noro 760: case 4:
1.41 noro 761: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 762: u2 = d2[i];
763: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
764: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
765: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
766: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
767: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
768: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
769: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
770: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
771: }
772: return 0;
773: break;
774: case 6:
1.41 noro 775: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 776: u2 = d2[i];
777: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
778: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
779: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
780: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
781: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
782: }
783: return 0;
784: break;
785: case 8:
1.41 noro 786: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 787: u2 = d2[i];
788: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
789: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
790: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
791: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
792: }
793: return 0;
794: break;
795: case 16:
1.41 noro 796: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 797: u2 = d2[i];
798: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
799: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
800: }
801: return 0;
802: break;
803: case 32:
1.41 noro 804: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 805: if ( d1[i]+d2[i]<d1[i] ) return 1;
806: return 0;
807: break;
808: default:
1.41 noro 809: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 810: u2 = d2[i];
811: k = (nd_epw-1)*nd_bpe;
812: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
813: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
814: }
815: return 0;
816: break;
817: }
1.65 noro 818: #else
819: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
820: u2 = d2[i];
821: k = (nd_epw-1)*nd_bpe;
822: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
823: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
824: }
825: return 0;
826: #endif
1.1 noro 827: }
828:
1.114 noro 829: int ndl_check_bound2(int index,UINT *d2)
830: {
831: return ndl_check_bound(nd_bound[index],d2);
832: }
833:
1.61 noro 834: INLINE int ndl_hash_value(UINT *d)
1.1 noro 835: {
836: int i;
837: int r;
838:
1.34 noro 839: r = 0;
1.41 noro 840: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 841: r = ((r<<16)+d[i])%REDTAB_LEN;
842: return r;
843: }
844:
1.63 noro 845: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 846: {
1.13 noro 847: RHist r;
1.6 noro 848: int d,k,i;
1.1 noro 849:
1.63 noro 850: d = ndl_hash_value(dg);
1.13 noro 851: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 852: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 853: if ( k > 0 ) nd_notfirst++;
854: nd_found++;
1.13 noro 855: return r->index;
1.1 noro 856: }
857: }
1.13 noro 858: if ( Reverse )
859: for ( i = nd_psn-1; i >= 0; i-- ) {
860: r = nd_psh[i];
1.34 noro 861: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 862: nd_create++;
1.34 noro 863: nd_append_red(dg,i);
1.13 noro 864: return i;
865: }
866: }
867: else
868: for ( i = 0; i < nd_psn; i++ ) {
869: r = nd_psh[i];
1.34 noro 870: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 871: nd_create++;
1.34 noro 872: nd_append_red(dg,i);
1.13 noro 873: return i;
874: }
1.1 noro 875: }
1.6 noro 876: return -1;
1.1 noro 877: }
878:
1.63 noro 879: ND nd_merge(ND p1,ND p2)
880: {
881: int n,c;
882: int t,can,td1,td2;
883: ND r;
884: NM m1,m2,mr0,mr,s;
885:
886: if ( !p1 ) return p2;
887: else if ( !p2 ) return p1;
888: else {
889: can = 0;
890: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
891: c = DL_COMPARE(DL(m1),DL(m2));
892: switch ( c ) {
893: case 0:
894: s = m1; m1 = NEXT(m1);
895: can++; NEXTNM2(mr0,mr,s);
896: s = m2; m2 = NEXT(m2); FREENM(s);
897: break;
898: case 1:
899: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
900: break;
901: case -1:
902: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
903: break;
904: }
905: }
906: if ( !mr0 )
907: if ( m1 ) mr0 = m1;
908: else if ( m2 ) mr0 = m2;
909: else return 0;
910: else if ( m1 ) NEXT(mr) = m1;
911: else if ( m2 ) NEXT(mr) = m2;
912: else NEXT(mr) = 0;
913: BDY(p1) = mr0;
914: SG(p1) = MAX(SG(p1),SG(p2));
915: LEN(p1) = LEN(p1)+LEN(p2)-can;
916: FREEND(p2);
917: return p1;
918: }
919: }
920:
1.31 noro 921: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 922: {
923: int n,c;
1.34 noro 924: int t,can,td1,td2;
1.1 noro 925: ND r;
926: NM m1,m2,mr0,mr,s;
927:
1.34 noro 928: if ( !p1 ) return p2;
929: else if ( !p2 ) return p1;
1.71 noro 930: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1.113 noro 931: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 932: else {
1.30 noro 933: can = 0;
1.1 noro 934: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 935: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 936: switch ( c ) {
937: case 0:
1.19 noro 938: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 939: if ( t < 0 ) t += mod;
1.1 noro 940: s = m1; m1 = NEXT(m1);
941: if ( t ) {
1.34 noro 942: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.17 noro 943: } else {
1.34 noro 944: can += 2; FREENM(s);
1.17 noro 945: }
946: s = m2; m2 = NEXT(m2); FREENM(s);
1.95 noro 947: break;
948: case 1:
949: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
950: break;
951: case -1:
952: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
953: break;
954: }
955: }
956: if ( !mr0 )
957: if ( m1 ) mr0 = m1;
958: else if ( m2 ) mr0 = m2;
959: else return 0;
960: else if ( m1 ) NEXT(mr) = m1;
961: else if ( m2 ) NEXT(mr) = m2;
962: else NEXT(mr) = 0;
963: BDY(p1) = mr0;
964: SG(p1) = MAX(SG(p1),SG(p2));
965: LEN(p1) = LEN(p1)+LEN(p2)-can;
966: FREEND(p2);
967: return p1;
968: }
969: }
970:
971: /* XXX on opteron, the inlined manipulation of destructive additon of
972: * two NM seems to make gcc optimizer get confused, so the part is
973: * done in a function.
974: */
975:
1.113 noro 976: int nm_destructive_add_q(NM *m1,NM *m2,NM *mr0,NM *mr)
1.95 noro 977: {
978: NM s;
1.113 noro 979: Q t;
1.95 noro 980: int can;
981:
1.113 noro 982: addq(CQ(*m1),CQ(*m2),&t);
1.95 noro 983: s = *m1; *m1 = NEXT(*m1);
984: if ( t ) {
1.113 noro 985: can = 1; NEXTNM2(*mr0,*mr,s); CQ(*mr) = (t);
1.95 noro 986: } else {
987: can = 2; FREENM(s);
988: }
989: s = *m2; *m2 = NEXT(*m2); FREENM(s);
990: return can;
991: }
992:
1.113 noro 993: ND nd_add_q(ND p1,ND p2)
1.95 noro 994: {
995: int n,c,can;
996: ND r;
997: NM m1,m2,mr0,mr,s;
1.113 noro 998: Q t;
1.95 noro 999:
1000: if ( !p1 ) return p2;
1001: else if ( !p2 ) return p1;
1002: else {
1003: can = 0;
1004: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1005: c = DL_COMPARE(DL(m1),DL(m2));
1006: switch ( c ) {
1007: case 0:
1008: #if defined(__x86_64__)
1.113 noro 1009: can += nm_destructive_add_q(&m1,&m2,&mr0,&mr);
1.95 noro 1010: #else
1.113 noro 1011: addq(CQ(m1),CQ(m2),&t);
1.95 noro 1012: s = m1; m1 = NEXT(m1);
1013: if ( t ) {
1.113 noro 1014: can++; NEXTNM2(mr0,mr,s); CQ(mr) = (t);
1.95 noro 1015: } else {
1016: can += 2; FREENM(s);
1017: }
1018: s = m2; m2 = NEXT(m2); FREENM(s);
1019: #endif
1.17 noro 1020: break;
1021: case 1:
1022: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1023: break;
1024: case -1:
1025: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1026: break;
1027: }
1028: }
1029: if ( !mr0 )
1.34 noro 1030: if ( m1 ) mr0 = m1;
1031: else if ( m2 ) mr0 = m2;
1032: else return 0;
1033: else if ( m1 ) NEXT(mr) = m1;
1034: else if ( m2 ) NEXT(mr) = m2;
1035: else NEXT(mr) = 0;
1.17 noro 1036: BDY(p1) = mr0;
1037: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1038: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 1039: FREEND(p2);
1040: return p1;
1041: }
1042: }
1043:
1.71 noro 1044: ND nd_add_sf(ND p1,ND p2)
1045: {
1046: int n,c,can;
1047: ND r;
1048: NM m1,m2,mr0,mr,s;
1049: int t;
1050:
1051: if ( !p1 ) return p2;
1052: else if ( !p2 ) return p1;
1053: else {
1054: can = 0;
1055: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1056: c = DL_COMPARE(DL(m1),DL(m2));
1057: switch ( c ) {
1058: case 0:
1059: t = _addsf(CM(m1),CM(m2));
1060: s = m1; m1 = NEXT(m1);
1061: if ( t ) {
1062: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1063: } else {
1064: can += 2; FREENM(s);
1065: }
1066: s = m2; m2 = NEXT(m2); FREENM(s);
1067: break;
1068: case 1:
1069: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1070: break;
1071: case -1:
1072: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1073: break;
1074: }
1075: }
1076: if ( !mr0 )
1077: if ( m1 ) mr0 = m1;
1078: else if ( m2 ) mr0 = m2;
1079: else return 0;
1080: else if ( m1 ) NEXT(mr) = m1;
1081: else if ( m2 ) NEXT(mr) = m2;
1082: else NEXT(mr) = 0;
1083: BDY(p1) = mr0;
1084: SG(p1) = MAX(SG(p1),SG(p2));
1085: LEN(p1) = LEN(p1)+LEN(p2)-can;
1086: FREEND(p2);
1087: return p1;
1088: }
1089: }
1090:
1.1 noro 1091: /* ret=1 : success, ret=0 : overflow */
1.113 noro 1092: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 1093: {
1.11 noro 1094: ND d;
1.1 noro 1095: NM m,mrd,tail;
1.7 noro 1096: NM mul;
1.10 noro 1097: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1098: int c,c1,c2,dummy;
1.17 noro 1099: RHist h;
1.11 noro 1100: NDV p,red;
1.113 noro 1101: Q cg,cred,gcd,tq,qq;
1.21 noro 1102: double hmag;
1.1 noro 1103:
1.113 noro 1104: if ( dn ) {
1105: if ( mod )
1106: dn->m = 1;
1107: else
1108: dn->z = ONE;
1109: }
1.1 noro 1110: if ( !g ) {
1111: *rp = 0;
1112: return 1;
1113: }
1.113 noro 1114: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1115:
1.14 noro 1116: sugar0 = sugar = SG(g);
1.1 noro 1117: n = NV(g);
1.61 noro 1118: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 1119: for ( d = 0; g; ) {
1.63 noro 1120: index = ndl_find_reducer(HDL(g));
1.6 noro 1121: if ( index >= 0 ) {
1.17 noro 1122: h = nd_psh[index];
1123: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1124: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1125: nd_free(g); nd_free(d);
1126: return 0;
1127: }
1.77 noro 1128: if ( nd_demand )
1129: p = ndv_load(index);
1130: else
1131: p = ps[index];
1.71 noro 1132: if ( mod == -1 )
1133: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1134: else if ( mod ) {
1.19 noro 1135: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1136: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1.16 noro 1137: } else {
1.113 noro 1138: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1139: chsgnq(cg,&CQ(mul));
1140: nd_mul_c_q(d,cred); nd_mul_c_q(g,cred);
1.69 noro 1141: if ( dn ) {
1.113 noro 1142: mulq(dn->z,cred,&tq); dn->z = tq;
1.69 noro 1143: }
1.16 noro 1144: }
1.55 noro 1145: g = nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.34 noro 1146: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.113 noro 1147: if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) {
1148: tq = HCQ(g);
1.21 noro 1149: nd_removecont2(d,g);
1.69 noro 1150: if ( dn ) {
1.113 noro 1151: divq(tq,HCQ(g),&qq); divq(dn->z,qq,&tq); dn->z = tq;
1.69 noro 1152: }
1.113 noro 1153: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1154: }
1.1 noro 1155: } else if ( !full ) {
1156: *rp = g;
1157: return 1;
1158: } else {
1159: m = BDY(g);
1160: if ( NEXT(m) ) {
1.34 noro 1161: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1162: } else {
1163: FREEND(g); g = 0;
1164: }
1165: if ( d ) {
1.34 noro 1166: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1167: } else {
1.34 noro 1168: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1169: }
1170: }
1171: }
1.34 noro 1172: if ( d ) SG(d) = sugar;
1.1 noro 1173: *rp = d;
1174: return 1;
1175: }
1.28 noro 1176:
1.53 noro 1177: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1178: {
1179: int hindex,index;
1180: NDV p;
1181: ND u,d,red;
1182: NODE l;
1.31 noro 1183: NM mul,m,mrd,tail;
1.25 noro 1184: int sugar,psugar,n,h_reducible;
1185: PGeoBucket bucket;
1186: int c,c1,c2;
1.113 noro 1187: Q cg,cred,gcd,zzz;
1.25 noro 1188: RHist h;
1.28 noro 1189: double hmag,gmag;
1.77 noro 1190: int count = 0;
1191: int hcount = 0;
1.25 noro 1192:
1193: if ( !g ) {
1194: *rp = 0;
1195: return 1;
1196: }
1197: sugar = SG(g);
1198: n = NV(g);
1.113 noro 1199: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1200: bucket = create_pbucket();
1.31 noro 1201: add_pbucket(mod,bucket,g);
1.25 noro 1202: d = 0;
1.61 noro 1203: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1204: while ( 1 ) {
1.113 noro 1205: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1206: if ( hindex < 0 ) {
1.77 noro 1207: if ( DP_Print > 3 ) printf("(%d %d)",count,hcount);
1.34 noro 1208: if ( d ) SG(d) = sugar;
1.25 noro 1209: *rp = d;
1210: return 1;
1211: }
1212: g = bucket->body[hindex];
1.63 noro 1213: index = ndl_find_reducer(HDL(g));
1.25 noro 1214: if ( index >= 0 ) {
1.77 noro 1215: count++;
1216: if ( !d ) hcount++;
1.25 noro 1217: h = nd_psh[index];
1218: ndl_sub(HDL(g),DL(h),DL(mul));
1219: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1220: nd_free(d);
1.25 noro 1221: free_pbucket(bucket);
1222: *rp = 0;
1223: return 0;
1224: }
1.53 noro 1225: p = ps[index];
1.71 noro 1226: if ( mod == -1 )
1227: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1228: else if ( mod ) {
1.25 noro 1229: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1230: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1231: } else {
1.113 noro 1232: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1233: chsgnq(cg,&CQ(mul));
1234: nd_mul_c_q(d,cred);
1235: mulq_pbucket(bucket,cred);
1.26 noro 1236: g = bucket->body[hindex];
1.113 noro 1237: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1238: }
1.55 noro 1239: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1240: bucket->body[hindex] = nd_remove_head(g);
1241: red = nd_remove_head(red);
1.31 noro 1242: add_pbucket(mod,bucket,red);
1.34 noro 1243: psugar = SG(p)+TD(DL(mul));
1244: sugar = MAX(sugar,psugar);
1.28 noro 1245: if ( !mod && hmag && (gmag > hmag) ) {
1246: g = normalize_pbucket(mod,bucket);
1247: if ( !g ) {
1.34 noro 1248: if ( d ) SG(d) = sugar;
1.28 noro 1249: *rp = d;
1250: return 1;
1251: }
1252: nd_removecont2(d,g);
1.113 noro 1253: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1254: add_pbucket(mod,bucket,g);
1.28 noro 1255: }
1.25 noro 1256: } else if ( !full ) {
1257: g = normalize_pbucket(mod,bucket);
1.34 noro 1258: if ( g ) SG(g) = sugar;
1.25 noro 1259: *rp = g;
1260: return 1;
1261: } else {
1262: m = BDY(g);
1263: if ( NEXT(m) ) {
1.34 noro 1264: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1265: } else {
1266: FREEND(g); g = 0;
1267: }
1268: bucket->body[hindex] = g;
1269: NEXT(m) = 0;
1270: if ( d ) {
1.34 noro 1271: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1272: } else {
1.34 noro 1273: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1274: }
1275: }
1276: }
1277: }
1.27 noro 1278:
1.61 noro 1279: /* input : list of NDV, cand : list of NDV */
1.28 noro 1280:
1.61 noro 1281: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1282: {
1283: int n,i,stat;
1284: ND nf,d;
1.61 noro 1285: NDV r;
1.45 noro 1286: NODE t,s;
1.113 noro 1287: union oNDC dn;
1.45 noro 1288:
1.128 noro 1289: ndv_setup(0,0,cand,0);
1.31 noro 1290: n = length(cand);
1.28 noro 1291:
1292: /* membercheck : list is a subset of Id(cand) ? */
1293: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1294: again:
1.61 noro 1295: if ( nd_bpe > obpe )
1296: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1297: else
1298: r = (NDV)BDY(t);
1299: d = ndvtond(0,r);
1.69 noro 1300: stat = nd_nf(0,d,nd_ps,0,0,&nf);
1.45 noro 1301: if ( !stat ) {
1.103 noro 1302: nd_reconstruct(0,0);
1.45 noro 1303: goto again;
1304: } else if ( nf ) return 0;
1.71 noro 1305: if ( DP_Print ) { printf("."); fflush(stdout); }
1.28 noro 1306: }
1.71 noro 1307: if ( DP_Print ) { printf("\n"); }
1.28 noro 1308: /* gbcheck : cand is a GB of Id(cand) ? */
1.77 noro 1309: if ( !nd_gb(0,0,1) ) return 0;
1.28 noro 1310: /* XXX */
1.23 noro 1311: return 1;
1312: }
1.1 noro 1313:
1314: ND nd_remove_head(ND p)
1315: {
1316: NM m;
1317:
1318: m = BDY(p);
1319: if ( !NEXT(m) ) {
1.34 noro 1320: FREEND(p); p = 0;
1.31 noro 1321: } else {
1.34 noro 1322: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1323: }
1.1 noro 1324: FREENM(m);
1325: return p;
1326: }
1327:
1.69 noro 1328: ND nd_separate_head(ND p,ND *head)
1329: {
1330: NM m,m0;
1331: ND r;
1332:
1333: m = BDY(p);
1334: if ( !NEXT(m) ) {
1335: *head = p; p = 0;
1336: } else {
1337: m0 = m;
1338: BDY(p) = NEXT(m); LEN(p)--;
1339: NEXT(m0) = 0;
1340: MKND(NV(p),m0,1,r);
1341: *head = r;
1342: }
1343: return p;
1344: }
1345:
1.1 noro 1346: PGeoBucket create_pbucket()
1347: {
1348: PGeoBucket g;
1349:
1350: g = CALLOC(1,sizeof(struct oPGeoBucket));
1351: g->m = -1;
1352: return g;
1353: }
1354:
1.25 noro 1355: void free_pbucket(PGeoBucket b) {
1356: int i;
1357:
1.26 noro 1358: for ( i = 0; i <= b->m; i++ )
1.25 noro 1359: if ( b->body[i] ) {
1360: nd_free(b->body[i]);
1361: b->body[i] = 0;
1362: }
1363: GC_free(b);
1364: }
1365:
1.63 noro 1366: void add_pbucket_symbolic(PGeoBucket g,ND d)
1367: {
1368: int l,i,k,m;
1369:
1370: if ( !d )
1371: return;
1372: l = LEN(d);
1373: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1374: /* 2^(k-1) < l <= 2^k (=m) */
1375: d = nd_merge(g->body[k],d);
1376: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1377: g->body[k] = 0;
1378: d = nd_merge(g->body[k+1],d);
1379: }
1380: g->body[k] = d;
1381: g->m = MAX(g->m,k);
1382: }
1383:
1.31 noro 1384: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1385: {
1.31 noro 1386: int l,i,k,m;
1.1 noro 1387:
1.31 noro 1388: if ( !d )
1389: return;
1390: l = LEN(d);
1.29 noro 1391: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1392: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1393: d = nd_add(mod,g->body[k],d);
1394: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1395: g->body[k] = 0;
1.31 noro 1396: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1397: }
1398: g->body[k] = d;
1399: g->m = MAX(g->m,k);
1400: }
1401:
1.113 noro 1402: void mulq_pbucket(PGeoBucket g,Q c)
1.26 noro 1403: {
1404: int k;
1405:
1406: for ( k = 0; k <= g->m; k++ )
1.113 noro 1407: nd_mul_c_q(g->body[k],c);
1.26 noro 1408: }
1409:
1.63 noro 1410: NM remove_head_pbucket_symbolic(PGeoBucket g)
1411: {
1412: int j,i,k,c;
1413: NM head;
1414:
1415: k = g->m;
1416: j = -1;
1417: for ( i = 0; i <= k; i++ ) {
1418: if ( !g->body[i] ) continue;
1419: if ( j < 0 ) j = i;
1420: else {
1421: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1422: if ( c > 0 )
1423: j = i;
1424: else if ( c == 0 )
1425: g->body[i] = nd_remove_head(g->body[i]);
1426: }
1427: }
1428: if ( j < 0 ) return 0;
1429: else {
1430: head = BDY(g->body[j]);
1431: if ( !NEXT(head) ) {
1432: FREEND(g->body[j]);
1433: g->body[j] = 0;
1434: } else {
1435: BDY(g->body[j]) = NEXT(head);
1436: LEN(g->body[j])--;
1437: }
1438: return head;
1439: }
1440: }
1441:
1.19 noro 1442: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1443: {
1444: int j,i,c,k,nv,sum;
1.61 noro 1445: UINT *di,*dj;
1.1 noro 1446: ND gi,gj;
1447:
1448: k = g->m;
1449: while ( 1 ) {
1450: j = -1;
1451: for ( i = 0; i <= k; i++ ) {
1452: if ( !(gi = g->body[i]) )
1453: continue;
1454: if ( j < 0 ) {
1455: j = i;
1456: gj = g->body[j];
1457: dj = HDL(gj);
1.14 noro 1458: sum = HCM(gj);
1.1 noro 1459: } else {
1.34 noro 1460: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1461: if ( c > 0 ) {
1.34 noro 1462: if ( sum ) HCM(gj) = sum;
1463: else g->body[j] = nd_remove_head(gj);
1.1 noro 1464: j = i;
1465: gj = g->body[j];
1466: dj = HDL(gj);
1.14 noro 1467: sum = HCM(gj);
1.1 noro 1468: } else if ( c == 0 ) {
1.71 noro 1469: if ( mod == -1 )
1470: sum = _addsf(sum,HCM(gi));
1471: else {
1472: sum = sum+HCM(gi)-mod;
1473: if ( sum < 0 ) sum += mod;
1474: }
1.1 noro 1475: g->body[i] = nd_remove_head(gi);
1476: }
1477: }
1478: }
1.34 noro 1479: if ( j < 0 ) return -1;
1.1 noro 1480: else if ( sum ) {
1.14 noro 1481: HCM(gj) = sum;
1.26 noro 1482: return j;
1.31 noro 1483: } else
1.26 noro 1484: g->body[j] = nd_remove_head(gj);
1485: }
1486: }
1487:
1.113 noro 1488: int head_pbucket_q(PGeoBucket g)
1.26 noro 1489: {
1490: int j,i,c,k,nv;
1.113 noro 1491: Q sum,t;
1.26 noro 1492: ND gi,gj;
1493:
1494: k = g->m;
1495: while ( 1 ) {
1496: j = -1;
1497: for ( i = 0; i <= k; i++ ) {
1.34 noro 1498: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1499: if ( j < 0 ) {
1500: j = i;
1501: gj = g->body[j];
1.113 noro 1502: sum = HCQ(gj);
1.26 noro 1503: } else {
1504: nv = NV(gi);
1.34 noro 1505: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1506: if ( c > 0 ) {
1.113 noro 1507: if ( sum ) HCQ(gj) = sum;
1.34 noro 1508: else g->body[j] = nd_remove_head(gj);
1.26 noro 1509: j = i;
1510: gj = g->body[j];
1.113 noro 1511: sum = HCQ(gj);
1.26 noro 1512: } else if ( c == 0 ) {
1.113 noro 1513: addq(sum,HCQ(gi),&t);
1514: sum = t;
1.26 noro 1515: g->body[i] = nd_remove_head(gi);
1516: }
1517: }
1518: }
1.34 noro 1519: if ( j < 0 ) return -1;
1.26 noro 1520: else if ( sum ) {
1.113 noro 1521: HCQ(gj) = sum;
1.1 noro 1522: return j;
1.31 noro 1523: } else
1.1 noro 1524: g->body[j] = nd_remove_head(gj);
1525: }
1526: }
1527:
1.25 noro 1528: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1529: {
1.31 noro 1530: int i;
1.1 noro 1531: ND r,t;
1532:
1533: r = 0;
1.28 noro 1534: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1535: r = nd_add(mod,r,g->body[i]);
1.28 noro 1536: g->body[i] = 0;
1537: }
1538: g->m = -1;
1.1 noro 1539: return r;
1540: }
1541:
1.122 noro 1542: int do_diagonalize(int sugar,int m)
1.92 noro 1543: {
1544: int i,nh,stat;
1545: NODE r,g,t;
1546: ND h,nf,s,head;
1547: NDV nfv;
1.113 noro 1548: Q q,num,den;
1549: union oNDC dn;
1.92 noro 1550:
1551: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1552: if ( nd_demand )
1553: nfv = ndv_load(i);
1554: else
1555: nfv = nd_ps[i];
1556: s = ndvtond(m,nfv);
1557: s = nd_separate_head(s,&head);
1.122 noro 1558: stat = nd_nf(m,s,nd_ps,1,&dn,&nf);
1559: if ( !stat ) return 0;
1.92 noro 1560: if ( !m ) {
1.113 noro 1561: NTOQ(NM(dn.z),SGN(dn.z),num);
1562: mulq(HCQ(head),num,&q); HCQ(head) = q;
1563: if ( DN(dn.z) ) {
1564: NTOQ(DN(dn.z),1,den);
1565: nd_mul_c_q(nf,den);
1566: }
1.92 noro 1567: }
1568: nf = nd_add(m,head,nf);
1569: ndv_free(nfv);
1570: nd_removecont(m,nf);
1571: nfv = ndtondv(m,nf);
1572: nd_free(nf);
1573: nd_bound[i] = ndv_compute_bound(nfv);
1574: if ( nd_demand ) {
1575: ndv_save(nfv,i);
1576: ndv_free(nfv);
1577: } else
1578: nd_ps[i] = nfv;
1579: }
1.122 noro 1580: return 1;
1.92 noro 1581: }
1582:
1.27 noro 1583: /* return value = 0 => input is not a GB */
1584:
1.77 noro 1585: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1586: {
1587: int i,nh,sugar,stat;
1.23 noro 1588: NODE r,g,t;
1.1 noro 1589: ND_pairs d;
1590: ND_pairs l;
1.117 noro 1591: ND h,nf,s,head,nf1;
1.63 noro 1592: NDV nfv;
1.113 noro 1593: Q q,num,den;
1594: union oNDC dn;
1.131 ! noro 1595: int diag_count = 0;
1.1 noro 1596:
1.23 noro 1597: g = 0; d = 0;
1598: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1599: d = update_pairs(d,g,i);
1600: g = update_base(g,i);
1601: }
1602: sugar = 0;
1603: while ( d ) {
1604: again:
1605: l = nd_minp(d,&d);
1.14 noro 1606: if ( SG(l) != sugar ) {
1.122 noro 1607: if ( ishomo ) {
1608: stat = do_diagonalize(sugar,m);
1609: if ( !stat ) {
1610: NEXT(l) = d; d = l;
1611: d = nd_reconstruct(0,d);
1612: goto again;
1613: }
1614: }
1.14 noro 1615: sugar = SG(l);
1.71 noro 1616: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1617: }
1.53 noro 1618: stat = nd_sp(m,0,l,&h);
1.1 noro 1619: if ( !stat ) {
1620: NEXT(l) = d; d = l;
1.103 noro 1621: d = nd_reconstruct(0,d);
1.1 noro 1622: goto again;
1623: }
1.41 noro 1624: #if USE_GEOBUCKET
1.69 noro 1625: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1626: #else
1.69 noro 1627: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1628: #endif
1.1 noro 1629: if ( !stat ) {
1630: NEXT(l) = d; d = l;
1.103 noro 1631: d = nd_reconstruct(0,d);
1.1 noro 1632: goto again;
1633: } else if ( nf ) {
1.27 noro 1634: if ( checkonly ) return 0;
1.71 noro 1635: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1636: nd_removecont(m,nf);
1.118 noro 1637: if ( !m && nd_nalg ) {
1638: nd_monic(0,&nf);
1639: nd_removecont(m,nf);
1.117 noro 1640: }
1.63 noro 1641: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1642: nh = ndv_newps(m,nfv,0);
1.131 ! noro 1643: if ( ishomo && ++diag_count == diag_period ) {
! 1644: diag_count = 0;
! 1645: stat = do_diagonalize(sugar,m);
! 1646: if ( !stat ) {
! 1647: NEXT(l) = d; d = l;
! 1648: d = nd_reconstruct(0,d);
! 1649: goto again;
! 1650: }
! 1651: }
1.1 noro 1652: d = update_pairs(d,g,nh);
1653: g = update_base(g,nh);
1654: FREENDP(l);
1655: } else {
1.71 noro 1656: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1657: FREENDP(l);
1658: }
1659: }
1.77 noro 1660: if ( nd_demand )
1661: for ( t = g; t; t = NEXT(t) )
1662: BDY(t) = (pointer)ndv_load((int)BDY(t));
1663: else
1664: for ( t = g; t; t = NEXT(t) )
1665: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1666: return g;
1667: }
1668:
1.122 noro 1669: int do_diagonalize_trace(int sugar,int m)
1.91 noro 1670: {
1671: int i,nh,stat;
1672: NODE r,g,t;
1673: ND h,nf,nfq,s,head;
1674: NDV nfv,nfqv;
1.113 noro 1675: Q q,den,num;
1676: union oNDC dn;
1.91 noro 1677:
1678: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1679: /* for nd_ps */
1680: s = ndvtond(m,nd_ps[i]);
1681: s = nd_separate_head(s,&head);
1.122 noro 1682: stat = nd_nf_pbucket(m,s,nd_ps,1,&nf);
1683: if ( !stat ) return 0;
1.91 noro 1684: nf = nd_add(m,head,nf);
1685: ndv_free(nd_ps[i]);
1686: nd_ps[i] = ndtondv(m,nf);
1687: nd_free(nf);
1688:
1689: /* for nd_ps_trace */
1690: if ( nd_demand )
1691: nfv = ndv_load(i);
1692: else
1693: nfv = nd_ps_trace[i];
1694: s = ndvtond(0,nfv);
1695: s = nd_separate_head(s,&head);
1.122 noro 1696: stat = nd_nf(0,s,nd_ps_trace,1,&dn,&nf);
1697: if ( !stat ) return 0;
1.113 noro 1698: NTOQ(NM(dn.z),SGN(dn.z),num);
1699: mulq(HCQ(head),num,&q); HCQ(head) = q;
1700: if ( DN(dn.z) ) {
1701: NTOQ(DN(dn.z),1,den);
1702: nd_mul_c_q(nf,den);
1703: }
1.91 noro 1704: nf = nd_add(0,head,nf);
1705: ndv_free(nfv);
1706: nd_removecont(0,nf);
1707: nfv = ndtondv(0,nf);
1708: nd_free(nf);
1709: nd_bound[i] = ndv_compute_bound(nfv);
1710: if ( nd_demand ) {
1711: ndv_save(nfv,i);
1712: ndv_free(nfv);
1713: } else
1714: nd_ps_trace[i] = nfv;
1715: }
1.122 noro 1716: return 1;
1.91 noro 1717: }
1718:
1.118 noro 1719: static struct oEGT eg_invdalg;
1720: struct oEGT eg_le;
1721:
1.82 noro 1722: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 1723: {
1724: int i,nh,sugar,stat;
1.23 noro 1725: NODE r,g,t;
1.20 noro 1726: ND_pairs d;
1727: ND_pairs l;
1.82 noro 1728: ND h,nf,nfq,s,head;
1.63 noro 1729: NDV nfv,nfqv;
1.113 noro 1730: Q q,den,num;
1731: union oNDC dn;
1.118 noro 1732: struct oEGT eg_monic,egm0,egm1;
1.130 noro 1733: int diag_count = 0;
1.20 noro 1734:
1.118 noro 1735: init_eg(&eg_monic);
1736: init_eg(&eg_invdalg);
1737: init_eg(&eg_le);
1.23 noro 1738: g = 0; d = 0;
1739: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1740: d = update_pairs(d,g,i);
1741: g = update_base(g,i);
1742: }
1743: sugar = 0;
1744: while ( d ) {
1745: again:
1746: l = nd_minp(d,&d);
1747: if ( SG(l) != sugar ) {
1.130 noro 1748: #if 1
1.122 noro 1749: if ( ishomo ) {
1750: stat = do_diagonalize_trace(sugar,m);
1.130 noro 1751: diag_count = 0;
1.122 noro 1752: if ( !stat ) {
1753: NEXT(l) = d; d = l;
1754: d = nd_reconstruct(1,d);
1755: goto again;
1756: }
1757: }
1.130 noro 1758: #endif
1.20 noro 1759: sugar = SG(l);
1.71 noro 1760: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1761: }
1.53 noro 1762: stat = nd_sp(m,0,l,&h);
1.20 noro 1763: if ( !stat ) {
1764: NEXT(l) = d; d = l;
1.103 noro 1765: d = nd_reconstruct(1,d);
1.20 noro 1766: goto again;
1767: }
1.41 noro 1768: #if USE_GEOBUCKET
1.53 noro 1769: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1770: #else
1.69 noro 1771: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1772: #endif
1.20 noro 1773: if ( !stat ) {
1774: NEXT(l) = d; d = l;
1.103 noro 1775: d = nd_reconstruct(1,d);
1.20 noro 1776: goto again;
1777: } else if ( nf ) {
1.77 noro 1778: if ( nd_demand ) {
1779: nfqv = ndv_load(nd_psn);
1780: nfq = ndvtond(0,nfqv);
1781: } else
1782: nfq = 0;
1783: if ( !nfq ) {
1784: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1785: NEXT(l) = d; d = l;
1.103 noro 1786: d = nd_reconstruct(1,d);
1.77 noro 1787: goto again;
1788: }
1.70 noro 1789: }
1.20 noro 1790: if ( nfq ) {
1.70 noro 1791: /* m|HC(nfq) => failure */
1.113 noro 1792: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1.63 noro 1793:
1.71 noro 1794: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.118 noro 1795: if ( nd_nalg ) {
1796: /* m|DN(HC(nf)^(-1)) => failure */
1797: get_eg(&egm0);
1798: if ( !nd_monic(m,&nfq) ) return 0;
1799: get_eg(&egm1); add_eg(&eg_monic,&egm0,&egm1);
1800: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1801: nfv = ndv_dup(0,nfqv); ndv_mod(m,nfv); nd_free(nf);
1802: } else {
1803: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1804: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1805: }
1.77 noro 1806: nh = ndv_newps(0,nfv,nfqv);
1.130 noro 1807: if ( ishomo && ++diag_count == diag_period ) {
1808: diag_count = 0;
1809: stat = do_diagonalize_trace(sugar,m);
1810: if ( !stat ) {
1811: NEXT(l) = d; d = l;
1812: d = nd_reconstruct(1,d);
1813: goto again;
1814: }
1815: }
1.20 noro 1816: d = update_pairs(d,g,nh);
1817: g = update_base(g,nh);
1818: } else {
1.71 noro 1819: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1820: }
1821: } else {
1.71 noro 1822: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1823: }
1824: FREENDP(l);
1825: }
1.77 noro 1826: if ( nd_demand )
1827: for ( t = g; t; t = NEXT(t) )
1828: BDY(t) = (pointer)ndv_load((int)BDY(t));
1829: else
1830: for ( t = g; t; t = NEXT(t) )
1831: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.118 noro 1832: if ( nd_nalg ) {
1833: print_eg("monic",&eg_monic);
1834: print_eg("invdalg",&eg_invdalg);
1835: print_eg("le",&eg_le);
1836: }
1.20 noro 1837: return g;
1838: }
1839:
1.23 noro 1840: int ndv_compare(NDV *p1,NDV *p2)
1841: {
1.34 noro 1842: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1843: }
1844:
1845: int ndv_compare_rev(NDV *p1,NDV *p2)
1846: {
1.34 noro 1847: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1848: }
1849:
1.61 noro 1850: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1851: {
1.69 noro 1852: int i,n,stat;
1853: ND nf,g,head;
1.23 noro 1854: NODE t,a0,a;
1.113 noro 1855: union oNDC dn;
1.69 noro 1856: NDV *w;
1.113 noro 1857: Q q,num,den;
1.23 noro 1858:
1.69 noro 1859: n = length(f);
1860: #if 0
1861: w = (NDV *)ALLOCA(n*sizeof(NDV));
1862: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1863: qsort(w,n,sizeof(NDV),
1864: (int (*)(const void *,const void *))ndv_compare);
1865: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1866: #endif
1.128 noro 1867: ndv_setup(m,0,f,0);
1.69 noro 1868: for ( i = 0; i < n; ) {
1869: g = ndvtond(m,nd_ps[i]);
1870: g = nd_separate_head(g,&head);
1871: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1872: if ( !stat )
1.103 noro 1873: nd_reconstruct(0,0);
1.69 noro 1874: else {
1.71 noro 1875: if ( DP_Print ) { printf("."); fflush(stdout); }
1.85 noro 1876: if ( !m ) {
1.113 noro 1877: NTOQ(NM(dn.z),SGN(dn.z),num);
1878: mulq(HCQ(head),num,&q); HCQ(head) = q;
1879: if ( DN(dn.z) ) {
1880: NTOQ(DN(dn.z),1,den);
1881: nd_mul_c_q(nf,den);
1882: }
1.85 noro 1883: }
1.69 noro 1884: nf = nd_add(m,head,nf);
1885: ndv_free(nd_ps[i]);
1.24 noro 1886: nd_removecont(m,nf);
1.69 noro 1887: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1888: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1889: i++;
1.23 noro 1890: }
1891: }
1.71 noro 1892: if ( DP_Print ) { printf("\n"); }
1.23 noro 1893: for ( a0 = 0, i = 0; i < n; i++ ) {
1894: NEXTNODE(a0,a);
1.69 noro 1895: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1896: }
1897: NEXT(a) = 0;
1898: return a0;
1899: }
1900:
1.1 noro 1901: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1902: {
1903: ND_pairs d1,nd,cur,head,prev,remove;
1904:
1905: if ( !g ) return d;
1906: d = crit_B(d,t);
1907: d1 = nd_newpairs(g,t);
1908: d1 = crit_M(d1);
1909: d1 = crit_F(d1);
1.55 noro 1910: if ( do_weyl )
1911: head = d1;
1912: else {
1913: prev = 0; cur = head = d1;
1914: while ( cur ) {
1915: if ( crit_2( cur->i1,cur->i2 ) ) {
1916: remove = cur;
1917: if ( !prev ) head = cur = NEXT(cur);
1918: else cur = NEXT(prev) = NEXT(cur);
1919: FREENDP(remove);
1920: } else {
1921: prev = cur; cur = NEXT(cur);
1922: }
1.1 noro 1923: }
1924: }
1925: if ( !d )
1926: return head;
1927: else {
1928: nd = d;
1.34 noro 1929: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1930: NEXT(nd) = head;
1931: return d;
1932: }
1933: }
1934:
1935: ND_pairs nd_newpairs( NODE g, int t )
1936: {
1937: NODE h;
1.61 noro 1938: UINT *dl;
1.34 noro 1939: int ts,s;
1.1 noro 1940: ND_pairs r,r0;
1941:
1.20 noro 1942: dl = DL(nd_psh[t]);
1.34 noro 1943: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1944: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1945: NEXTND_pairs(r0,r);
1946: r->i1 = (int)BDY(h);
1947: r->i2 = t;
1.20 noro 1948: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1949: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1950: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1951: }
1952: NEXT(r) = 0;
1953: return r0;
1954: }
1955:
1956: ND_pairs crit_B( ND_pairs d, int s )
1957: {
1958: ND_pairs cur,head,prev,remove;
1.61 noro 1959: UINT *t,*tl,*lcm;
1.1 noro 1960: int td,tdl;
1961:
1962: if ( !d ) return 0;
1.20 noro 1963: t = DL(nd_psh[s]);
1.1 noro 1964: prev = 0;
1965: head = cur = d;
1.61 noro 1966: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1967: while ( cur ) {
1968: tl = cur->lcm;
1969: if ( ndl_reducible(tl,t)
1.20 noro 1970: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1971: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1972: remove = cur;
1973: if ( !prev ) {
1974: head = cur = NEXT(cur);
1975: } else {
1976: cur = NEXT(prev) = NEXT(cur);
1977: }
1978: FREENDP(remove);
1979: } else {
1.34 noro 1980: prev = cur; cur = NEXT(cur);
1.1 noro 1981: }
1982: }
1983: return head;
1984: }
1985:
1986: ND_pairs crit_M( ND_pairs d1 )
1987: {
1988: ND_pairs e,d2,d3,dd,p;
1.61 noro 1989: UINT *id,*jd;
1.1 noro 1990:
1991: for ( dd = 0, e = d1; e; e = d3 ) {
1992: if ( !(d2 = NEXT(e)) ) {
1993: NEXT(e) = dd;
1994: return e;
1995: }
1.34 noro 1996: id = LCM(e);
1.1 noro 1997: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 1998: p = NEXT(d2);
1999: jd = LCM(d2);
2000: if ( ndl_equal(jd,id) )
2001: ;
2002: else if ( TD(jd) > TD(id) )
1.1 noro 2003: if ( ndl_reducible(jd,id) ) continue;
2004: else ;
1.34 noro 2005: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 2006: NEXT(d2) = d3;
2007: d3 = d2;
2008: }
2009: NEXT(e) = dd;
2010: dd = e;
2011: continue;
2012: /**/
2013: delit: NEXT(d2) = d3;
2014: d3 = d2;
2015: for ( ; p; p = d2 ) {
2016: d2 = NEXT(p);
2017: NEXT(p) = d3;
2018: d3 = p;
2019: }
2020: FREENDP(e);
2021: }
2022: return dd;
2023: }
2024:
2025: ND_pairs crit_F( ND_pairs d1 )
2026: {
2027: ND_pairs rest, head,remove;
2028: ND_pairs last, p, r, w;
2029: int s;
2030:
2031: for ( head = last = 0, p = d1; NEXT(p); ) {
2032: r = w = equivalent_pairs(p,&rest);
1.14 noro 2033: s = SG(r);
1.1 noro 2034: w = NEXT(w);
2035: while ( w ) {
2036: if ( crit_2(w->i1,w->i2) ) {
2037: r = w;
2038: w = NEXT(w);
2039: while ( w ) {
2040: remove = w;
2041: w = NEXT(w);
2042: FREENDP(remove);
2043: }
2044: break;
1.14 noro 2045: } else if ( SG(w) < s ) {
1.1 noro 2046: FREENDP(r);
2047: r = w;
1.14 noro 2048: s = SG(r);
1.1 noro 2049: w = NEXT(w);
2050: } else {
2051: remove = w;
2052: w = NEXT(w);
2053: FREENDP(remove);
2054: }
2055: }
2056: if ( last ) NEXT(last) = r;
2057: else head = r;
2058: NEXT(last = r) = 0;
2059: p = rest;
2060: if ( !p ) return head;
2061: }
2062: if ( !last ) return p;
2063: NEXT(last) = p;
2064: return head;
2065: }
2066:
2067: int crit_2( int dp1, int dp2 )
2068: {
1.20 noro 2069: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2070: }
2071:
1.40 noro 2072: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2073: {
2074: ND_pairs w,p,r,s;
1.61 noro 2075: UINT *d;
1.1 noro 2076:
2077: w = d1;
1.34 noro 2078: d = LCM(w);
1.1 noro 2079: s = NEXT(w);
2080: NEXT(w) = 0;
2081: for ( r = 0; s; s = p ) {
2082: p = NEXT(s);
1.34 noro 2083: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2084: NEXT(s) = w; w = s;
1.1 noro 2085: } else {
1.39 noro 2086: NEXT(s) = r; r = s;
1.1 noro 2087: }
2088: }
2089: *prest = r;
2090: return w;
2091: }
2092:
2093: NODE update_base(NODE nd,int ndp)
2094: {
1.61 noro 2095: UINT *dl, *dln;
1.1 noro 2096: NODE last, p, head;
2097:
1.20 noro 2098: dl = DL(nd_psh[ndp]);
1.1 noro 2099: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2100: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2101: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2102: p = NEXT(p);
2103: if ( last ) NEXT(last) = p;
2104: } else {
2105: if ( !last ) head = p;
2106: p = NEXT(last = p);
2107: }
2108: }
2109: head = append_one(head,ndp);
2110: return head;
2111: }
2112:
2113: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2114: {
2115: ND_pairs m,ml,p,l;
1.61 noro 2116: UINT *lcm;
1.33 noro 2117: int s,td,len,tlen,c,c1;
1.1 noro 2118:
2119: if ( !(p = NEXT(m = d)) ) {
2120: *prest = p;
2121: NEXT(m) = 0;
2122: return m;
2123: }
1.14 noro 2124: s = SG(m);
1.77 noro 2125: if ( !NoSugar ) {
2126: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2127: if ( (SG(p) < s)
2128: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
2129: ml = l; m = p; s = SG(m);
2130: }
2131: } else {
2132: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2133: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
2134: ml = l; m = p; s = SG(m);
2135: }
2136: }
1.1 noro 2137: if ( !ml ) *prest = NEXT(m);
2138: else {
2139: NEXT(ml) = NEXT(m);
2140: *prest = d;
2141: }
2142: NEXT(m) = 0;
2143: return m;
2144: }
2145:
1.63 noro 2146: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2147: {
1.69 noro 2148: int msugar,i;
1.63 noro 2149: ND_pairs t,dm0,dm,dr0,dr;
2150:
2151: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2152: if ( SG(t) < msugar ) msugar = SG(t);
2153: dm0 = 0; dr0 = 0;
1.69 noro 2154: for ( i = 0, t = d; t; t = NEXT(t) )
2155: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2156: if ( dm0 ) NEXT(dm) = t;
2157: else dm0 = t;
2158: dm = t;
1.69 noro 2159: i++;
1.63 noro 2160: } else {
2161: if ( dr0 ) NEXT(dr) = t;
2162: else dr0 = t;
2163: dr = t;
2164: }
2165: NEXT(dm) = 0;
2166: if ( dr0 ) NEXT(dr) = 0;
2167: *prest = dr0;
2168: return dm0;
2169: }
2170:
1.77 noro 2171: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2172: {
1.3 noro 2173: int len;
1.13 noro 2174: RHist r;
1.20 noro 2175: NDV b;
1.3 noro 2176:
1.1 noro 2177: if ( nd_psn == nd_pslen ) {
2178: nd_pslen *= 2;
1.11 noro 2179: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2180: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2181: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2182: nd_bound = (UINT **)
2183: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2184: }
1.39 noro 2185: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2186: nd_ps[nd_psn] = a;
1.39 noro 2187: if ( aq ) {
1.63 noro 2188: nd_ps_trace[nd_psn] = aq;
2189: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2190: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2191: } else {
1.63 noro 2192: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2193: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2194: }
1.77 noro 2195: if ( nd_demand ) {
2196: if ( aq ) {
2197: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2198: nd_ps_trace[nd_psn] = 0;
2199: } else {
2200: ndv_save(nd_ps[nd_psn],nd_psn);
2201: nd_ps[nd_psn] = 0;
2202: }
2203: }
1.1 noro 2204: return nd_psn++;
2205: }
2206:
1.128 noro 2207: void ndv_setup(int mod,int trace,NODE f,int dont_sort)
1.1 noro 2208: {
1.5 noro 2209: int i,j,td,len,max;
1.1 noro 2210: NODE s,s0,f0;
1.61 noro 2211: UINT *d;
1.13 noro 2212: RHist r;
1.69 noro 2213: NDV *w;
1.61 noro 2214: NDV a,am;
1.11 noro 2215:
2216: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2217:
1.69 noro 2218: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2219: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2220: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
1.128 noro 2221: if ( !dont_sort )
2222: qsort(w,nd_psn,sizeof(NDV),
2223: (int (*)(const void *,const void *))ndv_compare);
1.69 noro 2224: nd_pslen = 2*nd_psn;
1.11 noro 2225: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2226: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2227: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2228: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2229:
1.11 noro 2230: if ( !nd_red )
1.13 noro 2231: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2232: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2233: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2234: if ( trace ) {
1.69 noro 2235: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2236: ndv_removecont(0,a);
2237: am = nd_ps[i] = ndv_dup(mod,a);
2238: ndv_mod(mod,am);
2239: ndv_removecont(mod,am);
2240: } else {
1.69 noro 2241: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2242: ndv_removecont(mod,a);
1.39 noro 2243: }
1.61 noro 2244: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2245: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2246: nd_psh[i] = r;
1.77 noro 2247: if ( nd_demand ) {
2248: if ( trace ) {
2249: ndv_save(nd_ps_trace[i],i);
2250: nd_ps_trace[i] = 0;
2251: } else {
2252: ndv_save(nd_ps[i],i);
2253: nd_ps[i] = 0;
2254: }
2255: }
1.20 noro 2256: }
2257: }
2258:
1.119 noro 2259: struct order_spec *append_block(struct order_spec *spec,
2260: int nv,int nalg,int ord);
2261:
1.121 noro 2262: extern VECT current_dl_weight_vector_obj;
2263: static VECT prev_weight_vector_obj;
2264:
1.120 noro 2265: void preprocess_algcoef(VL vv,VL av,struct order_spec *ord,LIST f,
2266: struct order_spec **ord1p,LIST *f1p,NODE *alistp)
2267: {
1.121 noro 2268: NODE alist,t,s,r0,r,arg;
1.120 noro 2269: VL tv;
2270: P poly;
2271: DP d;
2272: Alg alpha,dp;
2273: DAlg inv,da,hc;
2274: MP m;
1.121 noro 2275: int i,nvar,nalg,n;
1.120 noro 2276: NumberField nf;
2277: LIST f1,f2;
2278: struct order_spec *current_spec;
1.121 noro 2279: VECT obj,obj0;
2280: Obj tmp;
1.120 noro 2281:
2282: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++);
2283: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++);
2284:
2285: for ( alist = 0, tv = av; tv; tv = NEXT(tv) ) {
2286: NEXTNODE(alist,t); MKV(tv->v,poly);
2287: MKAlg(poly,alpha); BDY(t) = (pointer)alpha;
2288: tv->v = tv->v->priv;
2289: }
2290: NEXT(t) = 0;
2291:
2292: /* simplification, makeing polynomials monic */
2293: setfield_dalg(alist);
2294: obj_algtodalg(f,&f1);
2295: for ( t = BDY(f); t; t = NEXT(t) ) {
2296: initd(ord); ptod(vv,vv,(P)BDY(t),&d);
2297: hc = (DAlg)BDY(d)->c;
2298: if ( NID(hc) == N_DA ) {
2299: invdalg(hc,&inv);
2300: for ( m = BDY(d); m; m = NEXT(m) ) {
2301: muldalg(inv,(DAlg)m->c,&da); m->c = (P)da;
2302: }
2303: }
2304: initd(ord); dtop(vv,vv,d,&poly); BDY(f) = (pointer)poly;
2305: }
2306: obj_dalgtoalg(f1,&f);
2307:
2308: /* append alg vars to the var list */
2309: for ( tv = vv; NEXT(tv); tv = NEXT(tv) );
2310: NEXT(tv) = av;
2311:
2312: /* append a block to ord */
2313: *ord1p = append_block(ord,nvar,nalg,2);
2314:
2315: /* create generator list */
2316: nf = get_numberfield();
2317: for ( i = nalg-1, t = BDY(f); i >= 0; i-- ) {
2318: MKAlg(nf->defpoly[i],dp);
2319: MKNODE(s,dp,t); t = s;
2320: }
2321: MKLIST(f1,t);
2322: *alistp = alist;
2323: algobjtorat(f1,f1p);
1.121 noro 2324:
2325: /* creating a new weight vector */
2326: prev_weight_vector_obj = obj0 = current_dl_weight_vector_obj;
2327: n = nvar+nalg+1;
2328: MKVECT(obj,n);
2329: if ( obj0 && obj0->len == nvar )
2330: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = BDY(obj0)[i];
2331: else
2332: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = (pointer)ONE;
2333: for ( i = 0; i < nalg; i++ ) BDY(obj)[i+nvar] = 0;
2334: BDY(obj)[n-1] = (pointer)ONE;
2335: arg = mknode(1,obj);
2336: Pdp_set_weight(arg,&tmp);
2337: }
2338:
2339: NODE postprocess_algcoef(VL av,NODE alist,NODE r)
2340: {
2341: NODE s,t,u0,u;
2342: P p;
2343: VL tv;
2344: Obj obj,tmp;
2345: NODE arg;
2346:
2347: u0 = 0;
2348: for ( t = r; t; t = NEXT(t) ) {
2349: p = (P)BDY(t);
2350: for ( tv = av, s = alist; tv; tv = NEXT(tv), s = NEXT(s) ) {
2351: substr(CO,0,(Obj)p,tv->v,(Obj)BDY(s),&obj); p = (P)obj;
2352: }
2353: if ( OID(p) == O_P || (OID(p) == O_N && NID((Num)p) != N_A) ) {
2354: NEXTNODE(u0,u);
2355: BDY(u) = (pointer)p;
2356: }
2357: }
2358: arg = mknode(1,prev_weight_vector_obj);
2359: Pdp_set_weight(arg,&tmp);
2360:
2361: return u0;
1.120 noro 2362: }
2363:
1.63 noro 2364: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2365: {
1.119 noro 2366: VL tv,fv,vv,vc,av;
2367: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2368: int e,max,nvar,i;
1.69 noro 2369: NDV b;
1.119 noro 2370: int ishomo,nalg;
2371: Alg alpha,dp;
2372: P p;
2373: LIST f1,f2;
2374: Obj obj;
2375: NumberField nf;
1.120 noro 2376: struct order_spec *ord1;
1.1 noro 2377:
1.78 noro 2378: if ( !m && Demand ) nd_demand = 1;
2379: else nd_demand = 0;
2380:
1.103 noro 2381: #if 0
1.74 noro 2382: ndv_alloc = 0;
1.103 noro 2383: #endif
1.1 noro 2384: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2385: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2386: switch ( ord->id ) {
2387: case 1:
2388: if ( ord->nv != nvar )
2389: error("nd_{gr,f4} : invalid order specification");
2390: break;
2391: default:
2392: break;
2393: }
1.119 noro 2394: nd_nalg = 0;
1.123 noro 2395: av = 0;
1.119 noro 2396: if ( !m ) {
2397: get_algtree((Obj)f,&av);
2398: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2399: nd_ntrans = nvar;
1.119 noro 2400: nd_nalg = nalg;
2401: /* #i -> t#i */
2402: if ( nalg ) {
1.120 noro 2403: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2404: ord = ord1;
2405: f = f1;
1.119 noro 2406: }
1.120 noro 2407: nvar += nalg;
1.119 noro 2408: }
1.32 noro 2409: nd_init_ord(ord);
1.61 noro 2410: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2411: for ( tv = vv; tv; tv = NEXT(tv) ) {
2412: e = getdeg(tv->v,(P)BDY(t));
2413: max = MAX(e,max);
2414: }
2415: nd_setup_parameters(nvar,max);
1.77 noro 2416: ishomo = 1;
1.1 noro 2417: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2418: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.77 noro 2419: if ( ishomo )
2420: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2421: if ( m ) ndv_mod(m,b);
1.61 noro 2422: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2423: }
2424: if ( fd0 ) NEXT(fd) = 0;
1.128 noro 2425: ndv_setup(m,0,fd0,0);
1.77 noro 2426: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.82 noro 2427: nd_demand = 0;
1.61 noro 2428: x = ndv_reducebase(x);
2429: x = ndv_reduceall(m,x);
1.23 noro 2430: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2431: NEXTNODE(r0,r);
1.61 noro 2432: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2433: }
2434: if ( r0 ) NEXT(r) = 0;
1.121 noro 2435: if ( nalg )
2436: r0 = postprocess_algcoef(av,alist,r0);
1.20 noro 2437: MKLIST(*rp,r0);
1.103 noro 2438: #if 0
1.74 noro 2439: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 2440: #endif
1.127 noro 2441: }
2442:
2443: void nd_gr_postproc(LIST f,LIST v,int m,struct order_spec *ord,int do_check,LIST *rp)
2444: {
2445: VL tv,fv,vv,vc,av;
2446: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2447: int e,max,nvar,i;
2448: NDV b;
2449: int ishomo,nalg;
2450: Alg alpha,dp;
2451: P p;
2452: LIST f1,f2;
2453: Obj obj;
2454: NumberField nf;
2455: struct order_spec *ord1;
2456:
2457: get_vars((Obj)f,&fv); pltovl(v,&vv);
2458: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
2459: switch ( ord->id ) {
2460: case 1:
2461: if ( ord->nv != nvar )
2462: error("nd_check : invalid order specification");
2463: break;
2464: default:
2465: break;
2466: }
2467: nd_nalg = 0;
2468: av = 0;
2469: if ( !m ) {
2470: get_algtree((Obj)f,&av);
2471: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
2472: nd_ntrans = nvar;
2473: nd_nalg = nalg;
2474: /* #i -> t#i */
2475: if ( nalg ) {
2476: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2477: ord = ord1;
2478: f = f1;
2479: }
2480: nvar += nalg;
2481: }
2482: nd_init_ord(ord);
2483: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2484: for ( tv = vv; tv; tv = NEXT(tv) ) {
2485: e = getdeg(tv->v,(P)BDY(t));
2486: max = MAX(e,max);
2487: }
2488: nd_setup_parameters(nvar,max);
2489: ishomo = 1;
2490: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2491: b = (pointer)ptondv(CO,vv,(P)BDY(t));
2492: if ( ishomo )
2493: ishomo = ishomo && ndv_ishomo(b);
2494: if ( m ) ndv_mod(m,b);
2495: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
2496: }
2497: if ( fd0 ) NEXT(fd) = 0;
1.128 noro 2498: ndv_setup(m,0,fd0,0);
1.127 noro 2499: for ( x = 0, i = 0; i < nd_psn; i++ )
2500: x = update_base(x,i);
2501: if ( do_check ) {
2502: x = nd_gb(m,ishomo,1);
2503: if ( !x ) {
2504: *rp = 0;
2505: return;
2506: }
2507: } else {
2508: for ( t = x; t; t = NEXT(t) )
2509: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
2510: }
2511: x = ndv_reducebase(x);
2512: x = ndv_reduceall(m,x);
2513: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
2514: NEXTNODE(r0,r);
2515: BDY(r) = ndvtop(m,CO,vv,BDY(t));
2516: }
2517: if ( r0 ) NEXT(r) = 0;
2518: if ( nalg )
2519: r0 = postprocess_algcoef(av,alist,r0);
2520: MKLIST(*rp,r0);
1.20 noro 2521: }
2522:
1.52 noro 2523: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2524: {
1.119 noro 2525: VL tv,fv,vv,vc,av;
2526: NODE fd,fd0,in0,in,r,r0,t,s,cand,alist;
1.61 noro 2527: int m,nocheck,nvar,mindex,e,max;
2528: NDV c;
2529: NMV a;
1.27 noro 2530: P p;
1.61 noro 2531: EPOS oepos;
1.119 noro 2532: int obpe,oadv,wmax,i,len,cbpe,ishomo,nalg;
2533: Alg alpha,dp;
2534: P poly;
2535: LIST f1,f2;
2536: Obj obj;
2537: NumberField nf;
1.120 noro 2538: struct order_spec *ord1;
1.20 noro 2539:
2540: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2541: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2542: switch ( ord->id ) {
2543: case 1:
2544: if ( ord->nv != nvar )
2545: error("nd_gr_trace : invalid order specification");
2546: break;
2547: default:
2548: break;
2549: }
1.119 noro 2550:
2551: get_algtree((Obj)f,&av);
2552: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2553: nd_ntrans = nvar;
1.119 noro 2554: nd_nalg = nalg;
2555: /* #i -> t#i */
2556: if ( nalg ) {
1.120 noro 2557: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2558: ord = ord1;
2559: f = f1;
1.119 noro 2560: }
1.120 noro 2561: nvar += nalg;
1.119 noro 2562:
1.52 noro 2563: nocheck = 0;
2564: mindex = 0;
2565:
1.78 noro 2566: if ( Demand ) nd_demand = 1;
2567: else nd_demand = 0;
2568:
1.52 noro 2569: /* setup modulus */
2570: if ( trace < 0 ) {
2571: trace = -trace;
2572: nocheck = 1;
2573: }
2574: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2575: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2576: for ( tv = vv; tv; tv = NEXT(tv) ) {
2577: e = getdeg(tv->v,(P)BDY(t));
2578: max = MAX(e,max);
1.23 noro 2579: }
1.61 noro 2580: nd_init_ord(ord);
2581: nd_setup_parameters(nvar,max);
2582: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 noro 2583: ishomo = 1;
1.61 noro 2584: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2585: c = ptondv(CO,vv,(P)BDY(t));
1.82 noro 2586: if ( ishomo )
2587: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2588: if ( c ) {
2589: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2590: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2591: }
1.61 noro 2592: }
2593: if ( in0 ) NEXT(in) = 0;
2594: if ( fd0 ) NEXT(fd) = 0;
1.82 noro 2595: if ( !ishomo && homo ) {
1.122 noro 2596: for ( t = in0, wmax = max; t; t = NEXT(t) ) {
1.61 noro 2597: c = (NDV)BDY(t); len = LEN(c);
2598: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2599: wmax = MAX(TD(DL(a)),wmax);
2600: }
2601: homogenize_order(ord,nvar,&ord1);
1.89 noro 2602: nd_init_ord(ord1);
1.61 noro 2603: nd_setup_parameters(nvar+1,wmax);
2604: for ( t = fd0; t; t = NEXT(t) )
2605: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2606: }
1.52 noro 2607: while ( 1 ) {
1.77 noro 2608: if ( Demand )
2609: nd_demand = 1;
1.128 noro 2610: ndv_setup(m,1,fd0,0);
1.82 noro 2611: cand = nd_gb_trace(m,ishomo || homo);
1.52 noro 2612: if ( !cand ) {
2613: /* failure */
1.61 noro 2614: if ( trace > 1 ) { *rp = 0; return; }
2615: else m = get_lprime(++mindex);
1.52 noro 2616: continue;
2617: }
1.82 noro 2618: if ( !ishomo && homo ) {
1.27 noro 2619: /* dehomogenization */
1.61 noro 2620: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2621: nd_init_ord(ord);
1.61 noro 2622: nd_setup_parameters(nvar,0);
1.27 noro 2623: }
1.77 noro 2624: nd_demand = 0;
1.61 noro 2625: cand = ndv_reducebase(cand);
2626: cand = ndv_reduceall(0,cand);
1.78 noro 2627: cbpe = nd_bpe;
1.61 noro 2628: if ( nocheck )
2629: break;
2630: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2631: /* success */
2632: break;
2633: else if ( trace > 1 ) {
2634: /* failure */
2635: *rp = 0; return;
1.61 noro 2636: } else {
1.52 noro 2637: /* try the next modulus */
2638: m = get_lprime(++mindex);
1.61 noro 2639: /* reset the parameters */
1.82 noro 2640: if ( !ishomo && homo ) {
1.89 noro 2641: nd_init_ord(ord1);
1.61 noro 2642: nd_setup_parameters(nvar+1,wmax);
2643: } else {
2644: nd_init_ord(ord);
2645: nd_setup_parameters(nvar,max);
2646: }
2647: }
1.52 noro 2648: }
1.27 noro 2649: /* dp->p */
1.61 noro 2650: nd_bpe = cbpe;
1.73 noro 2651: nd_setup_parameters(nd_nvar,0);
1.121 noro 2652: for ( r = cand; r; r = NEXT(r) )
1.119 noro 2653: BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.121 noro 2654: if ( nalg )
2655: cand = postprocess_algcoef(av,alist,cand);
1.27 noro 2656: MKLIST(*rp,cand);
1.1 noro 2657: }
2658:
1.61 noro 2659: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2660: {
1.61 noro 2661: UINT *d;
1.57 noro 2662: int i,j,l,s,ord_l;
1.43 noro 2663: struct order_pair *op;
1.1 noro 2664:
2665: d = dl->d;
1.41 noro 2666: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2667: if ( nd_blockmask ) {
2668: l = nd_blockmask->n;
2669: op = nd_blockmask->order_pair;
2670: for ( j = 0, s = 0; j < l; j++ ) {
2671: ord_l = op[j].length;
1.57 noro 2672: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2673: }
2674: TD(r) = ndl_weight(r);
1.61 noro 2675: ndl_weight_mask(r);
1.43 noro 2676: } else {
1.56 noro 2677: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2678: TD(r) = ndl_weight(r);
2679: }
1.1 noro 2680: }
2681:
1.61 noro 2682: DL ndltodl(int n,UINT *ndl)
1.1 noro 2683: {
2684: DL dl;
2685: int *d;
1.57 noro 2686: int i,j,l,s,ord_l;
1.43 noro 2687: struct order_pair *op;
1.1 noro 2688:
2689: NEWDL(dl,n);
1.34 noro 2690: dl->td = TD(ndl);
1.1 noro 2691: d = dl->d;
1.43 noro 2692: if ( nd_blockmask ) {
2693: l = nd_blockmask->n;
2694: op = nd_blockmask->order_pair;
2695: for ( j = 0, s = 0; j < l; j++ ) {
2696: ord_l = op[j].length;
1.57 noro 2697: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2698: }
2699: } else {
1.56 noro 2700: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2701: }
1.1 noro 2702: return dl;
2703: }
2704:
1.61 noro 2705: void ndl_print(UINT *dl)
1.1 noro 2706: {
2707: int n;
1.57 noro 2708: int i,j,l,ord_l,s,s0;
1.43 noro 2709: struct order_pair *op;
1.1 noro 2710:
2711: n = nd_nvar;
2712: printf("<<");
1.43 noro 2713: if ( nd_blockmask ) {
2714: l = nd_blockmask->n;
2715: op = nd_blockmask->order_pair;
2716: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2717: ord_l = op[j].length;
1.57 noro 2718: for ( i = 0; i < ord_l; i++, s++ )
2719: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2720: }
2721: } else {
1.56 noro 2722: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2723: }
1.1 noro 2724: printf(">>");
2725: }
2726:
2727: void nd_print(ND p)
2728: {
2729: NM m;
2730:
2731: if ( !p )
2732: printf("0\n");
2733: else {
2734: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2735: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2736: else printf("+%d*",CM(m));
1.14 noro 2737: ndl_print(DL(m));
1.1 noro 2738: }
2739: printf("\n");
2740: }
2741: }
2742:
1.113 noro 2743: void nd_print_q(ND p)
1.16 noro 2744: {
2745: NM m;
2746:
2747: if ( !p )
2748: printf("0\n");
2749: else {
2750: for ( m = BDY(p); m; m = NEXT(m) ) {
2751: printf("+");
1.113 noro 2752: printexpr(CO,(Obj)CQ(m));
1.16 noro 2753: printf("*");
2754: ndl_print(DL(m));
2755: }
2756: printf("\n");
2757: }
2758: }
2759:
1.1 noro 2760: void ndp_print(ND_pairs d)
2761: {
2762: ND_pairs t;
2763:
1.34 noro 2764: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2765: printf("\n");
2766: }
2767:
1.20 noro 2768: void nd_removecont(int mod,ND p)
1.16 noro 2769: {
2770: int i,n;
1.113 noro 2771: Q *w;
2772: Q dvr,t;
1.16 noro 2773: NM m;
1.21 noro 2774: struct oVECT v;
1.113 noro 2775: N q,r;
1.16 noro 2776:
1.71 noro 2777: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2778: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2779: else {
2780: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
1.113 noro 2781: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2782: v.len = n;
2783: v.body = (pointer *)w;
1.113 noro 2784: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2785: removecont_array(w,n);
1.113 noro 2786: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2787: }
2788: }
2789:
1.21 noro 2790: void nd_removecont2(ND p1,ND p2)
2791: {
2792: int i,n1,n2,n;
1.113 noro 2793: Q *w;
2794: Q dvr,t;
1.21 noro 2795: NM m;
2796: struct oVECT v;
1.113 noro 2797: N q,r;
1.21 noro 2798:
2799: if ( !p1 ) {
2800: nd_removecont(0,p2); return;
2801: } else if ( !p2 ) {
2802: nd_removecont(0,p1); return;
2803: }
2804: n1 = nd_length(p1);
2805: n2 = nd_length(p2);
2806: n = n1+n2;
1.113 noro 2807: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2808: v.len = n;
2809: v.body = (pointer *)w;
1.113 noro 2810: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2811: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2812: removecont_array(w,n);
1.113 noro 2813: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2814: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.21 noro 2815: }
2816:
1.20 noro 2817: void ndv_removecont(int mod,NDV p)
1.16 noro 2818: {
2819: int i,len;
1.113 noro 2820: Q *w;
2821: Q dvr,t;
1.16 noro 2822: NMV m;
2823:
1.71 noro 2824: if ( mod == -1 )
2825: ndv_mul_c(mod,p,_invsf(HCM(p)));
2826: else if ( mod )
1.20 noro 2827: ndv_mul_c(mod,p,invm(HCM(p),mod));
2828: else {
2829: len = p->len;
1.113 noro 2830: w = (Q *)ALLOCA(len*sizeof(Q));
2831: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
2832: sortbynm(w,len);
2833: qltozl(w,len,&dvr);
2834: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2835: divq(CQ(m),dvr,&t); CQ(m) = t;
2836: }
1.16 noro 2837: }
1.21 noro 2838: }
2839:
1.61 noro 2840: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2841: {
2842: int len,i,max;
2843: NMV m,mr0,mr,t;
2844:
2845: len = p->len;
2846: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2847: max = MAX(max,TD(DL(m)));
2848: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2849: m = (NMV)((char *)mr0+(len-1)*oadv);
2850: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2851: t = (NMV)ALLOCA(nmv_adv);
2852: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2853: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
1.113 noro 2854: CQ(mr) = CQ(m);
1.61 noro 2855: ndl_copy(DL(t),DL(mr));
2856: }
2857: NV(p)++;
2858: BDY(p) = mr0;
2859: }
2860:
1.45 noro 2861: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2862: {
1.45 noro 2863: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.113 noro 2864: Q *w;
2865: Q dvr,t;
1.23 noro 2866: NMV m,r;
2867:
2868: len = p->len;
2869: newnvar = nd_nvar-1;
1.48 noro 2870: newexporigin = nd_get_exporigin(ord);
1.45 noro 2871: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2872: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2873: ndl_dehomogenize(DL(m));
1.23 noro 2874: if ( newwpd != nd_wpd ) {
1.90 noro 2875: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
1.61 noro 2876: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.113 noro 2877: CQ(r) = CQ(m);
1.45 noro 2878: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2879: adj = nd_exporigin-newexporigin;
2880: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2881: }
2882: }
2883: NV(p)--;
2884: }
2885:
1.113 noro 2886: void removecont_array(Q *c,int n)
1.21 noro 2887: {
2888: struct oVECT v;
1.113 noro 2889: Q d0,d1,a,u,u1,gcd;
1.110 noro 2890: int i,j;
1.113 noro 2891: N qn,rn,gn;
2892: Q *q,*r;
1.21 noro 2893:
1.113 noro 2894: q = (Q *)ALLOCA(n*sizeof(Q));
2895: r = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2896: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
1.113 noro 2897: igcdv_estimate(&v,&d0);
2898: for ( i = 0; i < n; i++ ) {
2899: divn(NM(c[i]),NM(d0),&qn,&rn);
2900: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2901: NTOQ(rn,SGN(c[i]),r[i]);
2902: }
1.34 noro 2903: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2904: if ( i < n ) {
2905: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
1.113 noro 2906: igcdv(&v,&d1);
2907: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2908: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
1.21 noro 2909: for ( i = 0; i < n; i++ ) {
1.113 noro 2910: mulq(a,q[i],&u);
2911: if ( r[i] ) {
2912: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2913: addq(u,u1,&q[i]);
2914: } else
2915: q[i] = u;
1.21 noro 2916: }
2917: }
1.34 noro 2918: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2919: }
2920:
1.19 noro 2921: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2922: {
2923: NM m;
2924: int c,c1;
2925:
1.34 noro 2926: if ( !p ) return;
1.115 noro 2927: if ( mul == 1 ) return;
1.71 noro 2928: if ( mod == -1 )
2929: for ( m = BDY(p); m; m = NEXT(m) )
2930: CM(m) = _mulsf(CM(m),mul);
2931: else
2932: for ( m = BDY(p); m; m = NEXT(m) ) {
2933: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2934: }
1.1 noro 2935: }
2936:
1.113 noro 2937: void nd_mul_c_q(ND p,Q mul)
1.16 noro 2938: {
2939: NM m;
1.113 noro 2940: Q c;
1.16 noro 2941:
1.113 noro 2942: if ( !p ) return;
1.115 noro 2943: if ( UNIQ(mul) ) return;
1.16 noro 2944: for ( m = BDY(p); m; m = NEXT(m) ) {
1.113 noro 2945: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 2946: }
2947: }
2948:
1.61 noro 2949: void nd_mul_c_p(VL vl,ND p,P mul)
2950: {
2951: NM m;
2952: P c;
2953:
2954: if ( !p ) return;
2955: for ( m = BDY(p); m; m = NEXT(m) ) {
2956: mulp(vl,CP(m),mul,&c); CP(m) = c;
2957: }
2958: }
2959:
1.1 noro 2960: void nd_free(ND p)
2961: {
2962: NM t,s;
2963:
1.34 noro 2964: if ( !p ) return;
1.1 noro 2965: t = BDY(p);
2966: while ( t ) {
2967: s = NEXT(t);
2968: FREENM(t);
2969: t = s;
2970: }
2971: FREEND(p);
2972: }
2973:
1.23 noro 2974: void ndv_free(NDV p)
2975: {
2976: GC_free(BDY(p));
2977: }
2978:
1.61 noro 2979: void nd_append_red(UINT *d,int i)
1.1 noro 2980: {
1.13 noro 2981: RHist m,m0;
1.1 noro 2982: int h;
2983:
1.13 noro 2984: NEWRHist(m);
1.34 noro 2985: h = ndl_hash_value(d);
1.13 noro 2986: m->index = i;
1.14 noro 2987: ndl_copy(d,DL(m));
1.1 noro 2988: NEXT(m) = nd_red[h];
2989: nd_red[h] = m;
2990: }
2991:
1.61 noro 2992: UINT *ndv_compute_bound(NDV p)
1.1 noro 2993: {
1.61 noro 2994: UINT *d1,*d2,*t;
2995: UINT u;
1.57 noro 2996: int i,j,k,l,len,ind;
1.45 noro 2997: NMV m;
1.1 noro 2998:
2999: if ( !p )
3000: return 0;
1.61 noro 3001: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3002: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 3003: len = LEN(p);
3004: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
3005: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 3006: ndl_lcm(DL(m),d1,d2);
1.1 noro 3007: t = d1; d1 = d2; d2 = t;
3008: }
1.12 noro 3009: l = nd_nvar+31;
1.61 noro 3010: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 3011: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3012: u = d1[i];
3013: k = (nd_epw-1)*nd_bpe;
3014: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3015: t[ind] = (u>>k)&nd_mask0;
3016: }
3017: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 3018: return t;
3019: }
3020:
1.99 noro 3021: UINT *nd_compute_bound(ND p)
3022: {
3023: UINT *d1,*d2,*t;
3024: UINT u;
3025: int i,j,k,l,len,ind;
3026: NM m;
3027:
3028: if ( !p )
3029: return 0;
3030: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3031: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3032: len = LEN(p);
3033: m = BDY(p); ndl_copy(DL(m),d1); m = NEXT(m);
3034: for ( m = NEXT(m); m; m = NEXT(m) ) {
3035: ndl_lcm(DL(m),d1,d2);
3036: t = d1; d1 = d2; d2 = t;
3037: }
3038: l = nd_nvar+31;
3039: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
3040: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3041: u = d1[i];
3042: k = (nd_epw-1)*nd_bpe;
3043: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3044: t[ind] = (u>>k)&nd_mask0;
3045: }
3046: for ( ; ind < l; ind++ ) t[ind] = 0;
3047: return t;
3048: }
3049:
1.48 noro 3050: int nd_get_exporigin(struct order_spec *ord)
3051: {
1.51 noro 3052: switch ( ord->id ) {
1.96 noro 3053: case 0: case 2:
1.48 noro 3054: return 1;
1.41 noro 3055: case 1:
3056: /* block order */
1.43 noro 3057: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 3058: return ord->ord.block.length+1;
1.96 noro 3059: case 3:
3060: error("nd_get_exporigin : composite order is not supported yet.");
1.41 noro 3061: }
1.48 noro 3062: }
3063:
1.61 noro 3064: void nd_setup_parameters(int nvar,int max) {
1.79 noro 3065: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 3066: struct order_pair *op;
1.48 noro 3067:
1.73 noro 3068: nd_nvar = nvar;
1.79 noro 3069: if ( max ) {
1.83 noro 3070: /* XXX */
3071: if ( do_weyl ) nd_bpe = 32;
3072: else if ( max < 2 ) nd_bpe = 1;
1.79 noro 3073: else if ( max < 4 ) nd_bpe = 2;
3074: else if ( max < 8 ) nd_bpe = 3;
3075: else if ( max < 16 ) nd_bpe = 4;
3076: else if ( max < 32 ) nd_bpe = 5;
3077: else if ( max < 64 ) nd_bpe = 6;
3078: else if ( max < 256 ) nd_bpe = 8;
3079: else if ( max < 1024 ) nd_bpe = 10;
3080: else if ( max < 65536 ) nd_bpe = 16;
3081: else nd_bpe = 32;
3082: }
1.61 noro 3083: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 3084: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
3085: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 3086: wpd = nd_exporigin+elen;
3087: if ( wpd != nd_wpd ) {
3088: nd_free_private_storage();
3089: nd_wpd = wpd;
3090: }
1.1 noro 3091: if ( nd_bpe < 32 ) {
3092: nd_mask0 = (1<<nd_bpe)-1;
3093: } else {
3094: nd_mask0 = 0xffffffff;
3095: }
3096: bzero(nd_mask,sizeof(nd_mask));
3097: nd_mask1 = 0;
3098: for ( i = 0; i < nd_epw; i++ ) {
3099: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
3100: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
3101: }
1.90 noro 3102: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
1.57 noro 3103: nd_epos = nd_create_epos(nd_ord);
1.43 noro 3104: nd_blockmask = nd_create_blockmask(nd_ord);
1.96 noro 3105: nd_work_vector = (int *)REALLOC(nd_work_vector,nd_nvar*sizeof(int));
1.1 noro 3106: }
3107:
1.103 noro 3108: ND_pairs nd_reconstruct(int trace,ND_pairs d)
1.1 noro 3109: {
1.37 noro 3110: int i,obpe,oadv,h;
1.71 noro 3111: static NM prev_nm_free_list;
3112: static ND_pairs prev_ndp_free_list;
1.13 noro 3113: RHist mr0,mr;
3114: RHist r;
1.37 noro 3115: RHist *old_red;
1.71 noro 3116: ND_pairs s0,s,t;
1.43 noro 3117: EPOS oepos;
1.15 noro 3118:
1.1 noro 3119: obpe = nd_bpe;
1.11 noro 3120: oadv = nmv_adv;
1.43 noro 3121: oepos = nd_epos;
1.63 noro 3122: if ( obpe < 2 ) nd_bpe = 2;
3123: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 3124: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 3125: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 3126: else if ( obpe < 6 ) nd_bpe = 6;
3127: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 3128: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 3129: else if ( obpe < 16 ) nd_bpe = 16;
3130: else if ( obpe < 32 ) nd_bpe = 32;
3131: else error("nd_reconstruct : exponent too large");
1.5 noro 3132:
1.73 noro 3133: nd_setup_parameters(nd_nvar,0);
1.1 noro 3134: prev_nm_free_list = _nm_free_list;
3135: prev_ndp_free_list = _ndp_free_list;
3136: _nm_free_list = 0;
3137: _ndp_free_list = 0;
1.53 noro 3138: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
3139: if ( trace )
3140: for ( i = nd_psn-1; i >= 0; i-- )
3141: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 3142: s0 = 0;
3143: for ( t = d; t; t = NEXT(t) ) {
3144: NEXTND_pairs(s0,s);
3145: s->i1 = t->i1;
3146: s->i2 = t->i2;
1.14 noro 3147: SG(s) = SG(t);
1.61 noro 3148: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 3149: }
1.37 noro 3150:
3151: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 3152: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 3153: old_red[i] = nd_red[i];
3154: nd_red[i] = 0;
3155: }
3156: for ( i = 0; i < REDTAB_LEN; i++ )
3157: for ( r = old_red[i]; r; r = NEXT(r) ) {
3158: NEWRHist(mr);
1.13 noro 3159: mr->index = r->index;
1.20 noro 3160: SG(mr) = SG(r);
1.61 noro 3161: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 3162: h = ndl_hash_value(DL(mr));
3163: NEXT(mr) = nd_red[h];
3164: nd_red[h] = mr;
1.6 noro 3165: }
1.37 noro 3166: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
3167: old_red = 0;
1.11 noro 3168: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 3169: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 3170: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 3171: nd_psh[i] = r;
1.11 noro 3172: }
1.1 noro 3173: if ( s0 ) NEXT(s) = 0;
3174: prev_nm_free_list = 0;
3175: prev_ndp_free_list = 0;
1.71 noro 3176: #if 0
1.1 noro 3177: GC_gcollect();
1.71 noro 3178: #endif
1.1 noro 3179: return s0;
3180: }
3181:
1.61 noro 3182: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 3183: {
1.57 noro 3184: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 3185: struct order_pair *op;
1.1 noro 3186:
3187: n = nd_nvar;
1.61 noro 3188: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 3189: omask0 = (1<<obpe)-1;
1.34 noro 3190: TD(r) = TD(d);
1.41 noro 3191: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 3192: if ( nd_blockmask ) {
3193: l = nd_blockmask->n;
3194: op = nd_blockmask->order_pair;
3195: for ( i = 1; i < nd_exporigin; i++ )
3196: r[i] = d[i];
3197: for ( j = 0, s = 0; j < l; j++ ) {
3198: ord_l = op[j].length;
1.57 noro 3199: for ( i = 0; i < ord_l; i++, s++ ) {
3200: ei = GET_EXP_OLD(d,s);
3201: PUT_EXP(r,s,ei);
3202: }
1.1 noro 3203: }
1.43 noro 3204: } else {
1.56 noro 3205: for ( i = 0; i < n; i++ ) {
3206: ei = GET_EXP_OLD(d,i);
3207: PUT_EXP(r,i,ei);
3208: }
1.1 noro 3209: }
3210: }
1.3 noro 3211:
1.6 noro 3212: ND nd_copy(ND p)
3213: {
3214: NM m,mr,mr0;
1.41 noro 3215: int c,n;
1.6 noro 3216: ND r;
3217:
3218: if ( !p )
3219: return 0;
3220: else {
3221: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3222: NEXTNM(mr0,mr);
1.14 noro 3223: CM(mr) = CM(m);
3224: ndl_copy(DL(m),DL(mr));
1.6 noro 3225: }
3226: NEXT(mr) = 0;
1.31 noro 3227: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 3228: SG(r) = SG(p);
1.6 noro 3229: return r;
3230: }
3231: }
3232:
1.53 noro 3233: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3234: {
3235: NM m;
3236: NDV p1,p2;
3237: ND t1,t2;
1.61 noro 3238: UINT *lcm;
1.31 noro 3239: int td;
1.11 noro 3240:
1.77 noro 3241: if ( !mod && nd_demand ) {
3242: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 3243: } else {
1.77 noro 3244: if ( trace ) {
3245: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3246: } else {
3247: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3248: }
1.20 noro 3249: }
1.34 noro 3250: lcm = LCM(p);
1.11 noro 3251: NEWNM(m);
1.113 noro 3252: CQ(m) = HCQ(p2);
1.34 noro 3253: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 3254: if ( ndl_check_bound2(p->i1,DL(m)) )
3255: return 0;
1.55 noro 3256: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 3257: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
3258: else if ( mod ) CM(m) = mod-HCM(p1);
1.113 noro 3259: else chsgnq(HCQ(p1),&CQ(m));
1.34 noro 3260: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 3261: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 3262: nd_free(t1);
3263: return 0;
3264: }
1.55 noro 3265: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 3266: *rp = nd_add(mod,t1,t2);
1.11 noro 3267: FREENM(m);
3268: return 1;
3269: }
3270:
1.19 noro 3271: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3272: {
3273: NMV m;
3274: int c,c1,len,i;
3275:
1.34 noro 3276: if ( !p ) return;
1.14 noro 3277: len = LEN(p);
1.71 noro 3278: if ( mod == -1 )
3279: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3280: CM(m) = _mulsf(CM(m),mul);
3281: else
3282: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3283: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3284: }
1.11 noro 3285: }
3286:
1.113 noro 3287: void ndv_mul_c_q(NDV p,Q mul)
1.16 noro 3288: {
3289: NMV m;
1.113 noro 3290: Q c;
1.16 noro 3291: int len,i;
3292:
1.34 noro 3293: if ( !p ) return;
1.16 noro 3294: len = LEN(p);
3295: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 3296: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 3297: }
3298: }
3299:
1.55 noro 3300: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3301: int n2,i,j,l,n,tlen;
1.61 noro 3302: UINT *d0;
1.55 noro 3303: NM *tab,*psum;
3304: ND s,r;
3305: NM t;
3306: NMV m1;
3307:
3308: if ( !p ) return 0;
3309: n = NV(p); n2 = n>>1;
3310: d0 = DL(m0);
3311: l = LEN(p);
3312: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3313: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3314: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3315: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3316: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3317: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3318: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3319: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3320: for ( j = 0; j < tlen; j++ ) {
3321: if ( tab[j] ) {
3322: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3323: }
3324: }
3325: }
3326: for ( i = tlen-1, r = 0; i >= 0; i-- )
3327: if ( psum[i] ) {
3328: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3329: MKND(n,psum[i],j,s);
3330: r = nd_add(mod,r,s);
3331: }
1.56 noro 3332: if ( r ) SG(r) = SG(p)+TD(d0);
3333: return r;
1.55 noro 3334: }
3335:
1.56 noro 3336: /* product of monomials */
3337: /* XXX block order is not handled correctly */
3338:
1.55 noro 3339: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3340: {
1.56 noro 3341: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3342: UINT *d0,*d1,*d,*dt,*ctab;
1.113 noro 3343: Q *ctab_q;
3344: Q q,q1;
1.61 noro 3345: UINT c0,c1,c;
1.55 noro 3346: NM *p;
3347: NM m,t;
3348:
3349: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3350: if ( !m0 || !m1 ) return;
3351: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3352: NEWNM(m); d = DL(m);
1.56 noro 3353: if ( mod ) {
3354: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3355: } else
1.113 noro 3356: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3357: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3358: homo = n&1 ? 1 : 0;
3359: if ( homo ) {
3360: /* offset of h-degree */
3361: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3362: PUT_EXP(DL(m),n-1,h);
3363: TD(DL(m)) = h;
1.61 noro 3364: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3365: }
3366: tab[0] = m;
3367: NEWNM(m); d = DL(m);
1.57 noro 3368: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3369: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3370: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3371: /* xi^a*(Di^k*xi^l)*Di^b */
3372: a += l; b += k;
1.56 noro 3373: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3374: if ( !k || !l ) {
3375: for ( j = 0; j < curlen; j++ )
1.56 noro 3376: if ( t = tab[j] ) {
3377: dt = DL(t);
3378: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3379: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3380: }
3381: curlen *= k+1;
3382: continue;
3383: }
3384: min = MIN(k,l);
1.56 noro 3385: if ( mod ) {
1.61 noro 3386: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3387: mkwcm(k,l,mod,ctab);
3388: } else {
1.113 noro 3389: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3390: mkwc(k,l,ctab_q);
1.56 noro 3391: }
1.57 noro 3392: for ( j = min; j >= 0; j-- ) {
1.56 noro 3393: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3394: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3395: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3396: if ( homo ) {
3397: TD(d) = s;
1.56 noro 3398: PUT_EXP(d,n-1,s-h);
1.55 noro 3399: } else TD(d) = h;
1.61 noro 3400: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3401: if ( mod ) c = ctab[j];
1.113 noro 3402: else q = ctab_q[j];
1.57 noro 3403: p = tab+curlen*j;
3404: if ( j == 0 ) {
3405: for ( u = 0; u < curlen; u++, p++ ) {
3406: if ( tab[u] ) {
3407: ndl_addto(DL(tab[u]),d);
3408: if ( mod ) {
3409: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3410: } else {
1.113 noro 3411: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
1.57 noro 3412: }
3413: }
1.56 noro 3414: }
1.57 noro 3415: } else {
3416: for ( u = 0; u < curlen; u++, p++ ) {
3417: if ( tab[u] ) {
3418: NEWNM(t);
3419: ndl_add(DL(tab[u]),d,DL(t));
3420: if ( mod ) {
3421: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3422: } else
1.113 noro 3423: mulq(CQ(tab[u]),q,&CQ(t));
1.57 noro 3424: *p = t;
3425: }
1.55 noro 3426: }
3427: }
3428: }
3429: curlen *= k+1;
3430: }
3431: FREENM(m);
3432: }
3433:
1.63 noro 3434: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3435: {
3436: NM mr,mr0;
3437: NMV m;
3438: UINT *d,*dt,*dm;
3439: int c,n,td,i,c1,c2,len;
1.113 noro 3440: Q q;
1.63 noro 3441: ND r;
3442:
3443: if ( !p ) return 0;
3444: else {
3445: n = NV(p); m = BDY(p);
3446: d = DL(m0);
3447: len = LEN(p);
3448: mr0 = 0;
3449: td = TD(d);
3450: c = CM(m0);
3451: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3452: NEXTNM(mr0,mr);
3453: CM(mr) = 1;
3454: ndl_add(DL(m),d,DL(mr));
3455: }
3456: NEXT(mr) = 0;
3457: MKND(NV(p),mr0,len,r);
3458: SG(r) = SG(p) + TD(d);
3459: return r;
3460: }
3461: }
3462:
1.55 noro 3463: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3464: {
3465: NM mr,mr0;
3466: NMV m;
1.61 noro 3467: UINT *d,*dt,*dm;
1.9 noro 3468: int c,n,td,i,c1,c2,len;
1.113 noro 3469: Q q;
1.9 noro 3470: ND r;
3471:
1.34 noro 3472: if ( !p ) return 0;
1.55 noro 3473: else if ( do_weyl )
1.71 noro 3474: if ( mod == -1 )
3475: error("ndv_mul_nm : not implemented (weyl)");
3476: else
3477: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3478: else {
3479: n = NV(p); m = BDY(p);
1.34 noro 3480: d = DL(m0);
1.14 noro 3481: len = LEN(p);
1.9 noro 3482: mr0 = 0;
1.34 noro 3483: td = TD(d);
1.71 noro 3484: if ( mod == -1 ) {
3485: c = CM(m0);
3486: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3487: NEXTNM(mr0,mr);
3488: CM(mr) = _mulsf(CM(m),c);
3489: ndl_add(DL(m),d,DL(mr));
3490: }
3491: } else if ( mod ) {
1.16 noro 3492: c = CM(m0);
3493: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3494: NEXTNM(mr0,mr);
3495: c1 = CM(m);
1.19 noro 3496: DMAR(c1,c,0,mod,c2);
1.16 noro 3497: CM(mr) = c2;
3498: ndl_add(DL(m),d,DL(mr));
3499: }
3500: } else {
1.113 noro 3501: q = CQ(m0);
1.16 noro 3502: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3503: NEXTNM(mr0,mr);
1.113 noro 3504: mulq(CQ(m),q,&CQ(mr));
1.16 noro 3505: ndl_add(DL(m),d,DL(mr));
3506: }
1.4 noro 3507: }
1.9 noro 3508: NEXT(mr) = 0;
1.31 noro 3509: MKND(NV(p),mr0,len,r);
1.34 noro 3510: SG(r) = SG(p) + TD(d);
1.9 noro 3511: return r;
1.4 noro 3512: }
3513: }
3514:
1.104 noro 3515: ND nd_quo(int mod,PGeoBucket bucket,NDV d)
1.99 noro 3516: {
3517: NM mq0,mq;
1.102 noro 3518: NMV tm;
1.113 noro 3519: Q q;
1.104 noro 3520: int i,nv,sg,c,c1,c2,hindex;
3521: ND p,t,r;
3522: N tnm;
3523:
1.124 noro 3524: if ( bucket->m < 0 ) return 0;
1.99 noro 3525: else {
1.104 noro 3526: nv = NV(d);
1.99 noro 3527: mq0 = 0;
1.102 noro 3528: tm = (NMV)ALLOCA(nmv_adv);
1.104 noro 3529: while ( 1 ) {
1.113 noro 3530: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.104 noro 3531: if ( hindex < 0 ) break;
3532: p = bucket->body[hindex];
1.99 noro 3533: NEXTNM(mq0,mq);
3534: ndl_sub(HDL(p),HDL(d),DL(mq));
1.102 noro 3535: ndl_copy(DL(mq),DL(tm));
3536: if ( mod ) {
3537: c1 = invm(HCM(d),mod); c2 = HCM(p);
3538: DMAR(c1,c2,0,mod,c); CM(mq) = c;
3539: CM(tm) = mod-c;
3540: } else {
1.113 noro 3541: divsn(NM(HCQ(p)),NM(HCQ(d)),&tnm);
3542: NTOQ(tnm,SGN(HCQ(p))*SGN(HCQ(d)),CQ(mq));
3543: chsgnq(CQ(mq),&CQ(tm));
1.102 noro 3544: }
3545: t = ndv_mul_nmv_trunc(mod,tm,d,HDL(d));
1.104 noro 3546: bucket->body[hindex] = nd_remove_head(p);
3547: t = nd_remove_head(t);
3548: add_pbucket(mod,bucket,t);
3549: }
3550: if ( !mq0 )
3551: r = 0;
3552: else {
3553: NEXT(mq) = 0;
3554: for ( i = 0, mq = mq0; mq; mq = NEXT(mq), i++ );
3555: MKND(nv,mq0,i,r);
3556: /* XXX */
3557: SG(r) = HTD(r);
1.99 noro 3558: }
3559: return r;
3560: }
3561: }
3562:
1.43 noro 3563: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3564: {
1.13 noro 3565: NMV m,mr,mr0,t;
3566: int len,i,k;
1.11 noro 3567:
1.61 noro 3568: if ( !p ) return;
3569: m = BDY(p); len = LEN(p);
3570: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3571: m = (NMV)((char *)mr0+(len-1)*oadv);
3572: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3573: t = (NMV)ALLOCA(nmv_adv);
3574: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
1.113 noro 3575: CQ(t) = CQ(m);
1.61 noro 3576: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3577: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
1.113 noro 3578: CQ(mr) = CQ(t);
1.61 noro 3579: ndl_copy(DL(t),DL(mr));
3580: }
3581: BDY(p) = mr0;
3582: }
3583:
3584: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3585: {
3586: NMV m,mr,mr0;
3587: int len,i;
3588: NDV r;
1.11 noro 3589:
1.61 noro 3590: if ( !p ) return 0;
3591: m = BDY(p); len = LEN(p);
3592: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3593: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3594: ndl_zero(DL(mr));
3595: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
1.113 noro 3596: CQ(mr) = CQ(m);
1.11 noro 3597: }
1.61 noro 3598: MKNDV(NV(p),mr0,len,r);
3599: SG(r) = SG(p);
3600: return r;
1.11 noro 3601: }
3602:
1.61 noro 3603: /* duplicate p */
3604:
3605: NDV ndv_dup(int mod,NDV p)
1.3 noro 3606: {
3607: NDV d;
1.61 noro 3608: NMV t,m,m0;
1.3 noro 3609: int i,len;
3610:
1.34 noro 3611: if ( !p ) return 0;
1.31 noro 3612: len = LEN(p);
1.34 noro 3613: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3614: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3615: ndl_copy(DL(t),DL(m));
1.113 noro 3616: CQ(m) = CQ(t);
1.3 noro 3617: }
3618: MKNDV(NV(p),m0,len,d);
1.23 noro 3619: SG(d) = SG(p);
3620: return d;
3621: }
3622:
1.63 noro 3623: ND nd_dup(ND p)
3624: {
3625: ND d;
3626: NM t,m,m0;
3627:
3628: if ( !p ) return 0;
3629: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3630: NEXTNM(m0,m);
3631: ndl_copy(DL(t),DL(m));
1.113 noro 3632: CQ(m) = CQ(t);
1.63 noro 3633: }
3634: if ( m0 ) NEXT(m) = 0;
3635: MKND(NV(p),m0,LEN(p),d);
3636: SG(d) = SG(p);
3637: return d;
3638: }
3639:
1.61 noro 3640: /* XXX if p->len == 0 then it represents 0 */
3641:
3642: void ndv_mod(int mod,NDV p)
3643: {
3644: NMV t,d;
1.125 noro 3645: int r,s,u;
1.61 noro 3646: int i,len,dlen;
1.71 noro 3647: Obj gfs;
1.61 noro 3648:
3649: if ( !p ) return;
3650: len = LEN(p);
3651: dlen = 0;
1.71 noro 3652: if ( mod == -1 )
3653: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3654: simp_ff((Obj)CP(t),&gfs);
3655: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3656: CM(d) = r;
3657: ndl_copy(DL(t),DL(d));
3658: NMV_ADV(d);
3659: dlen++;
3660: }
1.71 noro 3661: else
3662: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
1.113 noro 3663: r = rem(NM(CQ(t)),mod);
1.71 noro 3664: if ( r ) {
1.113 noro 3665: if ( SGN(CQ(t)) < 0 )
3666: r = mod-r;
1.125 noro 3667: if ( DN(CQ(t)) ) {
3668: s = rem(DN(CQ(t)),mod);
3669: if ( !s )
3670: error("ndv_mod : division by 0");
3671: s = invm(s,mod);
3672: DMAR(r,s,0,mod,u); r = u;
3673: }
1.71 noro 3674: CM(d) = r;
3675: ndl_copy(DL(t),DL(d));
3676: NMV_ADV(d);
3677: dlen++;
3678: }
3679: }
1.61 noro 3680: LEN(p) = dlen;
3681: }
3682:
3683: NDV ptondv(VL vl,VL dvl,P p)
3684: {
3685: ND nd;
3686:
3687: nd = ptond(vl,dvl,p);
3688: return ndtondv(0,nd);
3689: }
3690:
3691: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3692: {
1.61 noro 3693: int n,i,j,k,e;
3694: VL tvl;
3695: V v;
3696: DCP dc;
3697: DCP *w;
3698: ND r,s,t,u;
3699: P x;
3700: int c;
3701: UINT *d;
1.23 noro 3702: NM m,m0;
1.61 noro 3703:
3704: if ( !p )
3705: return 0;
3706: else if ( NUM(p) ) {
3707: NEWNM(m);
3708: ndl_zero(DL(m));
1.113 noro 3709: CQ(m) = (Q)p;
1.61 noro 3710: NEXT(m) = 0;
3711: MKND(nd_nvar,m,1,r);
3712: SG(r) = 0;
3713: return r;
3714: } else {
3715: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3716: w = (DCP *)ALLOCA(k*sizeof(DCP));
3717: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3718: for ( i = 0, tvl = dvl, v = VR(p);
1.98 noro 3719: tvl && tvl->v != v; tvl = NEXT(tvl), i++ );
1.61 noro 3720: if ( !tvl ) {
3721: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3722: t = ptond(vl,dvl,COEF(w[j]));
3723: pwrp(vl,x,DEG(w[j]),&p);
3724: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3725: }
3726: return s;
3727: } else {
3728: NEWNM(m0); d = DL(m0);
3729: for ( j = k-1, s = 0; j >= 0; j-- ) {
3730: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3731: TD(d) = MUL_WEIGHT(e,i);
3732: if ( nd_blockmask) ndl_weight_mask(d);
3733: t = ptond(vl,dvl,COEF(w[j]));
3734: for ( m = BDY(t); m; m = NEXT(m) )
3735: ndl_addto(DL(m),d);
3736: SG(t) += TD(d);
3737: s = nd_add(0,s,t);
3738: }
3739: FREENM(m0);
3740: return s;
3741: }
3742: }
3743: }
3744:
3745: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3746: {
3747: VL tvl;
3748: int len,n,j,i,e;
3749: NMV m;
3750: Q q;
3751: P c;
3752: UINT *d;
3753: P s,r,u,t,w;
1.71 noro 3754: GFS gfs;
1.23 noro 3755:
1.34 noro 3756: if ( !p ) return 0;
1.61 noro 3757: else {
3758: len = LEN(p);
3759: n = NV(p);
3760: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3761: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3762: if ( mod == -1 ) {
3763: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3764: } else if ( mod ) {
1.61 noro 3765: STOQ(CM(m),q); c = (P)q;
3766: } else
1.113 noro 3767: c = CP(m);
1.61 noro 3768: d = DL(m);
3769: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3770: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3771: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3772: }
3773: addp(vl,s,t,&u); s = u;
3774: }
3775: return s;
1.23 noro 3776: }
1.3 noro 3777: }
3778:
1.61 noro 3779: NDV ndtondv(int mod,ND p)
1.11 noro 3780: {
3781: NDV d;
1.61 noro 3782: NMV m,m0;
3783: NM t;
3784: int i,len;
1.11 noro 3785:
1.34 noro 3786: if ( !p ) return 0;
1.61 noro 3787: len = LEN(p);
1.74 noro 3788: if ( mod )
3789: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3790: else
3791: m0 = m = MALLOC(len*nmv_adv);
1.103 noro 3792: #if 0
1.74 noro 3793: ndv_alloc += nmv_adv*len;
1.103 noro 3794: #endif
1.61 noro 3795: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3796: ndl_copy(DL(t),DL(m));
1.113 noro 3797: CQ(m) = CQ(t);
1.11 noro 3798: }
1.61 noro 3799: MKNDV(NV(p),m0,len,d);
1.14 noro 3800: SG(d) = SG(p);
1.11 noro 3801: return d;
3802: }
3803:
1.61 noro 3804: ND ndvtond(int mod,NDV p)
1.11 noro 3805: {
1.61 noro 3806: ND d;
3807: NM m,m0;
1.11 noro 3808: NMV t;
1.61 noro 3809: int i,len;
1.11 noro 3810:
1.34 noro 3811: if ( !p ) return 0;
1.11 noro 3812: m0 = 0;
1.61 noro 3813: len = p->len;
3814: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3815: NEXTNM(m0,m);
3816: ndl_copy(DL(t),DL(m));
1.113 noro 3817: CQ(m) = CQ(t);
1.11 noro 3818: }
3819: NEXT(m) = 0;
1.61 noro 3820: MKND(NV(p),m0,len,d);
1.14 noro 3821: SG(d) = SG(p);
1.11 noro 3822: return d;
3823: }
3824:
1.3 noro 3825: void ndv_print(NDV p)
3826: {
3827: NMV m;
3828: int i,len;
3829:
1.34 noro 3830: if ( !p ) printf("0\n");
1.3 noro 3831: else {
1.14 noro 3832: len = LEN(p);
1.3 noro 3833: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3834: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3835: else printf("+%d*",CM(m));
1.16 noro 3836: ndl_print(DL(m));
3837: }
3838: printf("\n");
3839: }
3840: }
3841:
1.113 noro 3842: void ndv_print_q(NDV p)
1.16 noro 3843: {
3844: NMV m;
3845: int i,len;
3846:
1.34 noro 3847: if ( !p ) printf("0\n");
1.16 noro 3848: else {
3849: len = LEN(p);
3850: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3851: printf("+");
1.113 noro 3852: printexpr(CO,(Obj)CQ(m));
1.16 noro 3853: printf("*");
1.14 noro 3854: ndl_print(DL(m));
1.3 noro 3855: }
3856: printf("\n");
3857: }
1.25 noro 3858: }
3859:
1.61 noro 3860: NODE ndv_reducebase(NODE x)
1.27 noro 3861: {
3862: int len,i,j;
3863: NDV *w;
3864: NODE t,t0;
3865:
3866: len = length(x);
3867: w = (NDV *)ALLOCA(len*sizeof(NDV));
3868: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3869: for ( i = 0; i < len; i++ ) {
3870: for ( j = 0; j < i; j++ ) {
3871: if ( w[i] && w[j] )
3872: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3873: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3874: }
3875: }
3876: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3877: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3878: }
3879: NEXT(t) = 0; x = t0;
3880: return x;
1.11 noro 3881: }
1.32 noro 3882:
1.43 noro 3883: /* XXX incomplete */
3884:
1.32 noro 3885: void nd_init_ord(struct order_spec *ord)
3886: {
1.43 noro 3887: switch ( ord->id ) {
1.32 noro 3888: case 0:
1.43 noro 3889: switch ( ord->ord.simple ) {
3890: case 0:
3891: nd_dcomp = 1;
3892: nd_isrlex = 1;
3893: break;
3894: case 1:
3895: nd_dcomp = 1;
3896: nd_isrlex = 0;
3897: break;
3898: case 2:
3899: nd_dcomp = 0;
3900: nd_isrlex = 0;
1.45 noro 3901: ndl_compare_function = ndl_lex_compare;
1.58 noro 3902: break;
3903: case 11:
3904: /* XXX */
3905: nd_dcomp = 0;
3906: nd_isrlex = 1;
3907: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3908: break;
3909: default:
3910: error("nd_gr : unsupported order");
3911: }
1.32 noro 3912: break;
3913: case 1:
1.96 noro 3914: /* block order */
1.43 noro 3915: /* XXX */
3916: nd_dcomp = -1;
1.32 noro 3917: nd_isrlex = 0;
1.45 noro 3918: ndl_compare_function = ndl_block_compare;
1.34 noro 3919: break;
1.43 noro 3920: case 2:
1.96 noro 3921: /* matrix order */
3922: /* XXX */
3923: nd_dcomp = -1;
3924: nd_isrlex = 0;
3925: nd_matrix_len = ord->ord.matrix.row;
3926: nd_matrix = ord->ord.matrix.matrix;
3927: ndl_compare_function = ndl_matrix_compare;
3928: break;
3929: case 3:
1.97 noro 3930: /* composite order */
3931: nd_dcomp = -1;
3932: nd_isrlex = 0;
3933: nd_worb_len = ord->ord.composite.length;
3934: nd_worb = ord->ord.composite.w_or_b;
3935: ndl_compare_function = ndl_composite_compare;
1.32 noro 3936: break;
3937: }
1.41 noro 3938: nd_ord = ord;
1.32 noro 3939: }
3940:
1.43 noro 3941: BlockMask nd_create_blockmask(struct order_spec *ord)
3942: {
3943: int n,i,j,s,l;
1.61 noro 3944: UINT *t;
1.43 noro 3945: BlockMask bm;
3946:
1.96 noro 3947: /* we only create mask table for block order */
3948: if ( ord->id != 1 )
1.43 noro 3949: return 0;
3950: n = ord->ord.block.length;
3951: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3952: bm->n = n;
3953: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3954: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3955: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3956: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3957: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3958: l = bm->order_pair[i].length;
3959: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3960: }
3961: return bm;
1.57 noro 3962: }
3963:
3964: EPOS nd_create_epos(struct order_spec *ord)
3965: {
3966: int i,j,l,s,ord_l,ord_o;
3967: EPOS epos;
3968: struct order_pair *op;
3969:
3970: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3971: switch ( ord->id ) {
3972: case 0:
3973: if ( nd_isrlex ) {
3974: for ( i = 0; i < nd_nvar; i++ ) {
3975: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3976: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3977: }
3978: } else {
3979: for ( i = 0; i < nd_nvar; i++ ) {
3980: epos[i].i = nd_exporigin + i/nd_epw;
3981: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3982: }
3983: }
3984: break;
3985: case 1:
3986: /* block order */
3987: l = ord->ord.block.length;
3988: op = ord->ord.block.order_pair;
3989: for ( j = 0, s = 0; j < l; j++ ) {
3990: ord_o = op[j].order;
3991: ord_l = op[j].length;
3992: if ( !ord_o )
3993: for ( i = 0; i < ord_l; i++ ) {
3994: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3995: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3996: }
3997: else
3998: for ( i = 0; i < ord_l; i++ ) {
3999: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
4000: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
4001: }
4002: s += ord_l;
4003: }
4004: break;
4005: case 2:
1.96 noro 4006: /* matrix order */
4007: case 3:
4008: /* composite order */
4009: for ( i = 0; i < nd_nvar; i++ ) {
4010: epos[i].i = nd_exporigin + i/nd_epw;
4011: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
4012: }
4013: break;
1.57 noro 4014: }
4015: return epos;
1.43 noro 4016: }
1.59 noro 4017:
4018: /* external interface */
4019:
4020: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
4021: {
1.61 noro 4022: NODE t,in0,in;
1.59 noro 4023: ND nd,nf;
1.61 noro 4024: NDV ndv;
4025: VL vv,tv;
4026: int stat,nvar,max,e;
1.113 noro 4027: union oNDC dn;
1.59 noro 4028:
1.116 noro 4029: if ( !f ) {
4030: *rp = 0;
4031: return;
4032: }
1.59 noro 4033: pltovl(v,&vv);
1.74 noro 4034: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 4035:
4036: /* get the degree bound */
4037: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
4038: for ( tv = vv; tv; tv = NEXT(tv) ) {
4039: e = getdeg(tv->v,(P)BDY(t));
4040: max = MAX(e,max);
4041: }
4042: for ( tv = vv; tv; tv = NEXT(tv) ) {
4043: e = getdeg(tv->v,f);
4044: max = MAX(e,max);
4045: }
4046:
1.59 noro 4047: nd_init_ord(ord);
1.61 noro 4048: nd_setup_parameters(nvar,max);
4049:
4050: /* conversion to ndv */
4051: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
4052: NEXTNODE(in0,in);
4053: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 4054: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4055: }
4056: NEXTNODE(in0,in);
4057: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 4058: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4059: NEXT(in) = 0;
4060:
1.128 noro 4061: /* dont sort */
4062: ndv_setup(m,0,in0,1);
1.59 noro 4063: nd_psn--;
4064: nd_scale=2;
4065: while ( 1 ) {
4066: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 4067: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 4068: if ( !stat ) {
4069: nd_psn++;
1.103 noro 4070: nd_reconstruct(0,0);
1.59 noro 4071: nd_psn--;
4072: } else
4073: break;
4074: }
1.61 noro 4075: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 4076: }
4077:
4078: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
4079: {
4080: NM m;
4081: UINT *t,*s;
4082: int i;
4083:
4084: for ( i = 0; i < n; i++ ) r[i] = 0;
4085: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
4086: t = DL(m);
4087: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4088: r[i] = CM(m);
4089: }
4090: for ( i = 0; !r[i]; i++ );
4091: return i;
4092: }
4093:
1.113 noro 4094: int nd_to_vect_q(UINT *s0,int n,ND d,Q *r)
1.74 noro 4095: {
1.107 noro 4096: NM m;
1.74 noro 4097: UINT *t,*s;
1.107 noro 4098: int i;
1.74 noro 4099:
4100: for ( i = 0; i < n; i++ ) r[i] = 0;
1.107 noro 4101: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
1.74 noro 4102: t = DL(m);
4103: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.113 noro 4104: r[i] = CQ(m);
1.74 noro 4105: }
4106: for ( i = 0; !r[i]; i++ );
4107: return i;
4108: }
4109:
1.129 noro 4110: Q *nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair)
4111: {
4112: NM m;
4113: NMV mr;
4114: UINT *d,*t,*s;
4115: NDV p;
4116: int i,j,len;
4117: Q *r;
4118:
4119: m = pair->mul;
4120: d = DL(m);
4121: p = nd_ps[pair->index];
4122: len = LEN(p);
4123: r = (Q *)CALLOC(n,sizeof(Q));
4124: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
4125: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4126: ndl_add(d,DL(mr),t);
4127: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4128: r[i] = CQ(mr);
4129: }
4130: return r;
4131: }
4132:
1.67 noro 4133: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 4134: {
4135: NM m;
4136: NMV mr;
4137: UINT *d,*t,*s;
4138: NDV p;
1.67 noro 4139: unsigned char *ivc;
4140: unsigned short *ivs;
1.81 noro 4141: UINT *v,*ivi,*s0v;
1.67 noro 4142: int i,j,len,prev,diff,cdiff;
4143: IndArray r;
1.64 noro 4144:
4145: m = pair->mul;
4146: d = DL(m);
4147: p = nd_ps[pair->index];
4148: len = LEN(p);
4149: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 4150: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 4151: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4152: ndl_add(d,DL(mr),t);
4153: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 4154: v[j] = i;
4155: }
1.81 noro 4156: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 4157: r->head = v[0];
4158: diff = 0;
4159: for ( i = 1; i < len; i++ ) {
4160: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
4161: }
4162: if ( diff < 256 ) {
4163: r->width = 1;
4164: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
4165: r->index.c = ivc;
4166: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
4167: } else if ( diff < 65536 ) {
4168: r->width = 2;
4169: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
4170: r->index.s = ivs;
4171: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
4172: } else {
4173: r->width = 4;
4174: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
4175: r->index.i = ivi;
4176: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 4177: }
1.67 noro 4178: return r;
1.64 noro 4179: }
4180:
1.65 noro 4181:
1.113 noro 4182: int ndv_reduce_vect_q(Q *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.107 noro 4183: {
4184: int i,j,k,len,pos,prev;
1.113 noro 4185: Q cs,mcs,c1,c2,cr,gcd,t;
1.107 noro 4186: IndArray ivect;
4187: unsigned char *ivc;
4188: unsigned short *ivs;
4189: unsigned int *ivi;
4190: NDV redv;
4191: NMV mr;
4192: NODE rp;
4193: int maxrs;
4194:
4195: maxrs = 0;
4196: for ( i = 0; i < nred; i++ ) {
4197: ivect = imat[i];
4198: k = ivect->head;
4199: if ( svect[k] ) {
4200: maxrs = MAX(maxrs,rp0[i]->sugar);
4201: redv = nd_ps[rp0[i]->index];
4202: len = LEN(redv); mr = BDY(redv);
1.113 noro 4203: igcd_cofactor(svect[k],CQ(mr),&gcd,&cs,&cr);
4204: chsgnq(cs,&mcs);
4205: if ( !UNIQ(cr) ) {
4206: for ( j = 0; j < col; j++ ) {
4207: mulq(svect[j],cr,&c1); svect[j] = c1;
4208: }
4209: }
1.107 noro 4210: svect[k] = 0; prev = k;
4211: switch ( ivect->width ) {
4212: case 1:
4213: ivc = ivect->index.c;
4214: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4215: pos = prev+ivc[j]; prev = pos;
1.113 noro 4216: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4217: }
4218: break;
4219: case 2:
4220: ivs = ivect->index.s;
4221: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4222: pos = prev+ivs[j]; prev = pos;
1.113 noro 4223: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4224: }
4225: break;
4226: case 4:
4227: ivi = ivect->index.i;
4228: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4229: pos = prev+ivi[j]; prev = pos;
1.113 noro 4230: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4231: }
4232: break;
4233: }
4234: }
4235: }
4236: return maxrs;
4237: }
4238:
1.76 noro 4239: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 4240: {
1.67 noro 4241: int i,j,k,len,pos,prev;
1.66 noro 4242: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 4243: IndArray ivect;
4244: unsigned char *ivc;
4245: unsigned short *ivs;
4246: unsigned int *ivi;
1.65 noro 4247: NDV redv;
1.67 noro 4248: NMV mr;
1.65 noro 4249: NODE rp;
1.76 noro 4250: int maxrs;
1.65 noro 4251:
1.76 noro 4252: maxrs = 0;
1.74 noro 4253: for ( i = 0; i < nred; i++ ) {
1.65 noro 4254: ivect = imat[i];
1.67 noro 4255: k = ivect->head; svect[k] %= m;
1.65 noro 4256: if ( c = svect[k] ) {
1.76 noro 4257: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4258: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 4259: len = LEN(redv); mr = BDY(redv);
4260: svect[k] = 0; prev = k;
4261: switch ( ivect->width ) {
4262: case 1:
4263: ivc = ivect->index.c;
4264: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4265: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
4266: prev = pos;
4267: DMA(c1,c,c2,up,lo);
4268: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4269: } else svect[pos] = lo;
4270: }
4271: break;
4272: case 2:
4273: ivs = ivect->index.s;
4274: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4275: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
4276: prev = pos;
4277: DMA(c1,c,c2,up,lo);
4278: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4279: } else svect[pos] = lo;
4280: }
4281: break;
4282: case 4:
4283: ivi = ivect->index.i;
4284: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4285: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
4286: prev = pos;
4287: DMA(c1,c,c2,up,lo);
4288: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4289: } else svect[pos] = lo;
4290: }
4291: break;
1.65 noro 4292: }
4293: }
4294: }
1.66 noro 4295: for ( i = 0; i < col; i++ )
4296: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 4297: return maxrs;
1.65 noro 4298: }
4299:
1.76 noro 4300: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 4301: {
4302: int i,j,k,len,pos,prev;
4303: UINT c,c1,c2,c3,up,lo,dmy;
4304: IndArray ivect;
4305: unsigned char *ivc;
4306: unsigned short *ivs;
4307: unsigned int *ivi;
4308: NDV redv;
4309: NMV mr;
4310: NODE rp;
1.76 noro 4311: int maxrs;
1.72 noro 4312:
1.76 noro 4313: maxrs = 0;
1.74 noro 4314: for ( i = 0; i < nred; i++ ) {
1.72 noro 4315: ivect = imat[i];
4316: k = ivect->head; svect[k] %= m;
4317: if ( c = svect[k] ) {
1.76 noro 4318: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4319: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 4320: len = LEN(redv); mr = BDY(redv);
4321: svect[k] = 0; prev = k;
4322: switch ( ivect->width ) {
4323: case 1:
4324: ivc = ivect->index.c;
4325: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4326: pos = prev+ivc[j]; prev = pos;
4327: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4328: }
4329: break;
4330: case 2:
4331: ivs = ivect->index.s;
4332: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4333: pos = prev+ivs[j]; prev = pos;
4334: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4335: }
4336: break;
4337: case 4:
4338: ivi = ivect->index.i;
4339: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4340: pos = prev+ivi[j]; prev = pos;
4341: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4342: }
4343: break;
4344: }
4345: }
4346: }
1.76 noro 4347: return maxrs;
1.72 noro 4348: }
4349:
1.65 noro 4350: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
4351: {
4352: int j,k,len;
4353: UINT *p;
4354: UINT c;
4355: NDV r;
4356: NMV mr0,mr;
4357:
4358: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4359: if ( !len ) return 0;
4360: else {
1.74 noro 4361: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4362: #if 0
1.74 noro 4363: ndv_alloc += nmv_adv*len;
1.103 noro 4364: #endif
1.65 noro 4365: mr = mr0;
4366: p = s0vect;
4367: for ( j = k = 0; j < col; j++, p += nd_wpd )
4368: if ( !rhead[j] ) {
4369: if ( c = vect[k++] ) {
4370: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
4371: }
4372: }
4373: MKNDV(nd_nvar,mr0,len,r);
4374: return r;
4375: }
4376: }
4377:
1.129 noro 4378: /* for preprocessed vector */
4379:
1.113 noro 4380: NDV vect_to_ndv_q(Q *vect,int spcol,int col,int *rhead,UINT *s0vect)
1.107 noro 4381: {
4382: int j,k,len;
4383: UINT *p;
1.113 noro 4384: Q c;
1.107 noro 4385: NDV r;
4386: NMV mr0,mr;
4387:
4388: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4389: if ( !len ) return 0;
4390: else {
1.109 noro 4391: mr0 = (NMV)GC_malloc(nmv_adv*len);
1.107 noro 4392: #if 0
4393: ndv_alloc += nmv_adv*len;
4394: #endif
4395: mr = mr0;
4396: p = s0vect;
4397: for ( j = k = 0; j < col; j++, p += nd_wpd )
4398: if ( !rhead[j] ) {
4399: if ( c = vect[k++] ) {
1.113 noro 4400: if ( DN(c) )
4401: error("afo");
4402: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
1.107 noro 4403: }
4404: }
4405: MKNDV(nd_nvar,mr0,len,r);
4406: return r;
4407: }
4408: }
4409:
1.129 noro 4410: /* for plain vector */
4411:
4412: NDV plain_vect_to_ndv_q(Q *vect,int col,UINT *s0vect)
4413: {
4414: int j,k,len;
4415: UINT *p;
4416: Q c;
4417: NDV r;
4418: NMV mr0,mr;
4419:
4420: for ( j = 0, len = 0; j < col; j++ ) if ( vect[j] ) len++;
4421: if ( !len ) return 0;
4422: else {
4423: mr0 = (NMV)GC_malloc(nmv_adv*len);
4424: #if 0
4425: ndv_alloc += nmv_adv*len;
4426: #endif
4427: mr = mr0;
4428: p = s0vect;
4429: for ( j = k = 0; j < col; j++, p += nd_wpd, k++ )
4430: if ( c = vect[k] ) {
4431: if ( DN(c) )
4432: error("afo");
4433: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
4434: }
4435: MKNDV(nd_nvar,mr0,len,r);
4436: return r;
4437: }
4438: }
4439:
1.74 noro 4440: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket)
1.65 noro 4441: {
4442: ND_pairs t;
4443: NODE sp0,sp;
4444: int stat;
4445: ND spol;
4446:
4447: for ( t = l; t; t = NEXT(t) ) {
4448: stat = nd_sp(m,0,t,&spol);
4449: if ( !stat ) return 0;
4450: if ( spol ) {
4451: add_pbucket_symbolic(bucket,spol);
4452: }
4453: }
1.68 noro 4454: return 1;
1.65 noro 4455: }
4456:
4457: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
4458: {
4459: NODE rp0,rp;
4460: NM mul,head,s0,s;
1.76 noro 4461: int index,col,i,sugar;
1.65 noro 4462: RHist h;
4463: UINT *s0v,*p;
4464: NM_ind_pair pair;
4465: ND red;
4466:
4467: s0 = 0; rp0 = 0; col = 0;
4468: while ( 1 ) {
4469: head = remove_head_pbucket_symbolic(bucket);
4470: if ( !head ) break;
4471: if ( !s0 ) s0 = head;
4472: else NEXT(s) = head;
4473: s = head;
4474: index = ndl_find_reducer(DL(head));
4475: if ( index >= 0 ) {
4476: h = nd_psh[index];
4477: NEWNM(mul);
4478: ndl_sub(DL(head),DL(h),DL(mul));
4479: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.76 noro 4480: sugar = TD(DL(mul))+SG(nd_ps[index]);
4481: MKNM_ind_pair(pair,mul,index,sugar);
1.65 noro 4482: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
4483: add_pbucket_symbolic(bucket,nd_remove_head(red));
4484: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4485: }
4486: col++;
4487: }
1.72 noro 4488: if ( rp0 ) NEXT(rp) = 0;
4489: NEXT(s) = 0;
1.65 noro 4490: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4491: for ( i = 0, p = s0v, s = s0; i < col;
4492: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4493: *s0vect = s0v;
4494: *r = rp0;
4495: return col;
4496: }
4497:
1.69 noro 4498: NODE nd_f4(int m)
4499: {
4500: int i,nh,stat,index;
4501: NODE r,g;
4502: ND_pairs d,l,t;
4503: ND spol,red;
4504: NDV nf,redv;
4505: NM s0,s;
1.84 noro 4506: NODE rp0,srp0,nflist;
1.69 noro 4507: int nsp,nred,col,rank,len,k,j,a;
4508: UINT c;
1.74 noro 4509: UINT **spmat;
1.69 noro 4510: UINT *s0vect,*svect,*p,*v;
4511: int *colstat;
4512: IndArray *imat;
4513: int *rhead;
4514: int spcol,sprow;
4515: int sugar;
4516: PGeoBucket bucket;
4517: struct oEGT eg0,eg1,eg_f4;
4518:
1.103 noro 4519: #if 0
1.74 noro 4520: ndv_alloc = 0;
1.103 noro 4521: #endif
1.69 noro 4522: g = 0; d = 0;
4523: for ( i = 0; i < nd_psn; i++ ) {
4524: d = update_pairs(d,g,i);
4525: g = update_base(g,i);
4526: }
4527: while ( d ) {
4528: get_eg(&eg0);
4529: l = nd_minsugarp(d,&d);
4530: sugar = SG(l);
4531: bucket = create_pbucket();
1.74 noro 4532: stat = nd_sp_f4(m,l,bucket);
1.69 noro 4533: if ( !stat ) {
4534: for ( t = l; NEXT(t); t = NEXT(t) );
4535: NEXT(t) = d; d = l;
1.103 noro 4536: d = nd_reconstruct(0,d);
1.69 noro 4537: continue;
4538: }
1.86 noro 4539: if ( bucket->m < 0 ) continue;
1.69 noro 4540: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
4541: if ( !col ) {
4542: for ( t = l; NEXT(t); t = NEXT(t) );
4543: NEXT(t) = d; d = l;
1.103 noro 4544: d = nd_reconstruct(0,d);
1.69 noro 4545: continue;
4546: }
4547: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4548: if ( DP_Print )
4549: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4550: sugar,eg_f4.exectime+eg_f4.gctime);
4551: if ( 1 )
4552: nflist = nd_f4_red(m,l,s0vect,col,rp0);
4553: else
4554: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0);
1.69 noro 4555: /* adding new bases */
1.74 noro 4556: for ( r = nflist; r; r = NEXT(r) ) {
4557: nf = (NDV)BDY(r);
1.69 noro 4558: ndv_removecont(m,nf);
1.129 noro 4559: if ( !m && nd_nalg ) {
4560: ND nf1;
4561:
4562: nf1 = ndvtond(m,nf);
4563: nd_monic(0,&nf1);
4564: nd_removecont(m,nf1);
4565: nf = ndtondv(m,nf1);
4566: }
1.77 noro 4567: nh = ndv_newps(m,nf,0);
1.69 noro 4568: d = update_pairs(d,g,nh);
4569: g = update_base(g,nh);
4570: }
4571: }
4572: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.103 noro 4573: #if 0
1.74 noro 4574: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 4575: #endif
1.69 noro 4576: return g;
4577: }
1.74 noro 4578:
4579: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
1.63 noro 4580: {
1.67 noro 4581: IndArray *imat;
1.106 noro 4582: int nsp,nred,i;
1.65 noro 4583: int *rhead;
1.106 noro 4584: NODE r0,rp;
1.74 noro 4585: ND_pairs sp;
4586: NM_ind_pair *rvect;
1.63 noro 4587:
1.74 noro 4588: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
1.106 noro 4589: nred = length(rp0);
1.74 noro 4590: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4591: rhead = (int *)ALLOCA(col*sizeof(int));
4592: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4593:
1.74 noro 4594: /* construction of index arrays */
4595: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4596: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4597: rvect[i] = (NM_ind_pair)BDY(rp);
4598: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4599: rhead[imat[i]->head] = 1;
4600: }
1.107 noro 4601: if ( m )
4602: r0 = nd_f4_red_main(m,sp0,nsp,s0vect,col,rvect,rhead,imat,nred);
4603: else
1.113 noro 4604: r0 = nd_f4_red_q_main(sp0,nsp,s0vect,col,rvect,rhead,imat,nred);
1.106 noro 4605: return r0;
4606: }
1.74 noro 4607:
1.106 noro 4608: NODE nd_f4_red_main(int m,ND_pairs sp0,int nsp,UINT *s0vect,int col,
4609: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4610: {
4611: int spcol,sprow,a;
4612: int i,j,k,l,rank;
4613: NODE r0,r;
4614: ND_pairs sp;
4615: ND spol;
4616: int **spmat;
4617: UINT *svect,*v;
4618: int *colstat;
4619: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4620: int maxrs;
4621: int *spsugar;
4622:
4623: spcol = col-nred;
4624: get_eg(&eg0);
1.74 noro 4625: /* elimination (1st step) */
4626: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4627: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4628: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4629: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4630: nd_sp(m,0,sp,&spol);
1.75 noro 4631: if ( !spol ) continue;
1.74 noro 4632: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4633: if ( m == -1 )
4634: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4635: else
4636: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4637: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4638: if ( i < col ) {
4639: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4640: for ( j = k = 0; j < col; j++ )
4641: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4642: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4643: sprow++;
4644: }
1.76 noro 4645: nd_free(spol);
1.74 noro 4646: }
1.80 noro 4647: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4648: if ( DP_Print ) {
4649: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4650: fflush(asir_out);
4651: }
1.74 noro 4652: /* free index arrays */
4653: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4654:
4655: /* elimination (2nd step) */
4656: colstat = (int *)ALLOCA(spcol*sizeof(int));
4657: if ( m == -1 )
1.76 noro 4658: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4659: else
1.76 noro 4660: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4661: r0 = 0;
4662: for ( i = 0; i < rank; i++ ) {
4663: NEXTNODE(r0,r); BDY(r) =
4664: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4665: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4666: GC_free(spmat[i]);
4667: }
1.109 noro 4668: if ( r0 ) NEXT(r) = 0;
1.74 noro 4669: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4670: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4671: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4672: if ( DP_Print ) {
1.80 noro 4673: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4674: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4675: nsp,nred,sprow,spcol,rank);
4676: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4677: }
1.74 noro 4678: return r0;
4679: }
4680:
1.129 noro 4681: #if 0
1.113 noro 4682: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,UINT *s0vect,int col,
1.107 noro 4683: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4684: {
4685: int spcol,sprow,a;
4686: int i,j,k,l,rank;
4687: NODE r0,r;
4688: ND_pairs sp;
4689: ND spol;
1.113 noro 4690: Q **spmat;
4691: Q *svect,*v;
1.107 noro 4692: int *colstat;
4693: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4694: int maxrs;
4695: int *spsugar;
4696:
4697: spcol = col-nred;
4698: get_eg(&eg0);
4699: /* elimination (1st step) */
1.113 noro 4700: spmat = (Q **)ALLOCA(nsp*sizeof(Q *));
4701: svect = (Q *)ALLOCA(col*sizeof(Q));
4702: spsugar = (int *)ALLOCA(nsp*sizeof(Q));
1.107 noro 4703: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4704: nd_sp(0,0,sp,&spol);
4705: if ( !spol ) continue;
1.113 noro 4706: nd_to_vect_q(s0vect,col,spol,svect);
4707: maxrs = ndv_reduce_vect_q(svect,col,imat,rvect,nred);
1.107 noro 4708: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4709: if ( i < col ) {
1.113 noro 4710: spmat[sprow] = v = (Q *)MALLOC(spcol*sizeof(Q));
1.107 noro 4711: for ( j = k = 0; j < col; j++ )
4712: if ( !rhead[j] ) v[k++] = svect[j];
4713: spsugar[sprow] = MAX(maxrs,SG(spol));
4714: sprow++;
4715: }
1.109 noro 4716: /* nd_free(spol); */
1.107 noro 4717: }
4718: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4719: if ( DP_Print ) {
4720: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4721: fflush(asir_out);
4722: }
4723: /* free index arrays */
1.109 noro 4724: /* for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c); */
1.107 noro 4725:
4726: /* elimination (2nd step) */
4727: colstat = (int *)ALLOCA(spcol*sizeof(int));
1.113 noro 4728: rank = nd_gauss_elim_q(spmat,spsugar,sprow,spcol,colstat);
1.107 noro 4729: r0 = 0;
4730: for ( i = 0; i < rank; i++ ) {
4731: NEXTNODE(r0,r); BDY(r) =
1.113 noro 4732: (pointer)vect_to_ndv_q(spmat[i],spcol,col,rhead,s0vect);
1.107 noro 4733: SG((NDV)BDY(r)) = spsugar[i];
1.109 noro 4734: /* GC_free(spmat[i]); */
1.107 noro 4735: }
1.109 noro 4736: if ( r0 ) NEXT(r) = 0;
4737:
4738: /* for ( ; i < sprow; i++ ) GC_free(spmat[i]); */
1.107 noro 4739: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4740: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
4741: if ( DP_Print ) {
4742: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
4743: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4744: nsp,nred,sprow,spcol,rank);
4745: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
4746: }
4747: return r0;
4748: }
1.129 noro 4749: #else
4750: void printm(Q **mat,int row,int col)
4751: {
4752: int i,j;
4753: printf("[");
4754: for ( i = 0; i < row; i++ ) {
4755: for ( j = 0; j < col; j++ ) {
4756: printexpr(CO,mat[i][j]); printf(" ");
4757: }
4758: printf("]\n");
4759: }
4760: }
4761:
4762: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,UINT *s0vect,int col,
4763: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4764: {
4765: int row,a;
4766: int i,j,rank;
4767: NODE r0,r;
4768: ND_pairs sp;
4769: ND spol;
4770: Q **mat;
4771: int *colstat;
4772: int *sugar;
4773:
4774: row = nsp+nred;
4775: /* make the matrix */
4776: mat = (Q **)ALLOCA(row*sizeof(Q *));
4777: sugar = (int *)ALLOCA(row*sizeof(int));
4778: for ( row = a = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4779: nd_sp(0,0,sp,&spol);
4780: if ( !spol ) continue;
4781: mat[row] = (Q *)MALLOC(col*sizeof(Q));
4782: nd_to_vect_q(s0vect,col,spol,mat[row]);
4783: sugar[row] = SG(spol);
4784: row++;
4785: }
4786: for ( i = 0; i < nred; i++, row++ ) {
4787: mat[row] = nm_ind_pair_to_vect(0,s0vect,col,rvect[i]);
4788: sugar[row] = rvect[i]->sugar;
4789: }
4790: /* elimination */
4791: colstat = (int *)ALLOCA(col*sizeof(int));
4792: rank = nd_gauss_elim_q(mat,sugar,row,col,colstat);
4793: r0 = 0;
4794: for ( i = 0; i < rank; i++ ) {
4795: for ( j = 0; j < col; j++ ) if ( mat[i][j] ) break;
4796: if ( j == col ) error("nd_f4_red_q_main : cannot happen");
4797: if ( rhead[j] ) continue;
4798: NEXTNODE(r0,r); BDY(r) =
4799: (pointer)plain_vect_to_ndv_q(mat[i],col,s0vect);
4800: SG((NDV)BDY(r)) = sugar[i];
4801: }
4802: if ( r0 ) NEXT(r) = 0;
4803: printf("\n");
4804: return r0;
4805: }
4806: #endif
1.107 noro 4807:
1.74 noro 4808: FILE *nd_write,*nd_read;
4809:
4810: void nd_send_int(int a) {
4811: write_int(nd_write,&a);
4812: }
4813:
4814: void nd_send_intarray(int *p,int len) {
4815: write_intarray(nd_write,p,len);
4816: }
4817:
4818: int nd_recv_int() {
4819: int a;
4820:
4821: read_int(nd_read,&a);
4822: return a;
4823: }
4824:
4825: void nd_recv_intarray(int *p,int len) {
4826: read_intarray(nd_read,p,len);
4827: }
4828:
4829: void nd_send_ndv(NDV p) {
4830: int len,i;
4831: NMV m;
4832:
4833: if ( !p ) nd_send_int(0);
4834: else {
4835: len = LEN(p);
4836: nd_send_int(len);
4837: m = BDY(p);
4838: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4839: nd_send_int(CM(m));
4840: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4841: }
1.74 noro 4842: }
4843: }
4844:
4845: void nd_send_nd(ND p) {
4846: int len,i;
4847: NM m;
4848:
4849: if ( !p ) nd_send_int(0);
4850: else {
4851: len = LEN(p);
4852: nd_send_int(len);
4853: m = BDY(p);
4854: for ( i = 0; i < len; i++, m = NEXT(m) ) {
4855: nd_send_int(CM(m));
4856: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4857: }
1.74 noro 4858: }
4859: }
1.65 noro 4860:
1.74 noro 4861: NDV nd_recv_ndv()
4862: {
4863: int len,i;
4864: NMV m,m0;
4865: NDV r;
1.65 noro 4866:
1.74 noro 4867: len = nd_recv_int();
4868: if ( !len ) return 0;
4869: else {
4870: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4871: #if 0
1.74 noro 4872: ndv_alloc += len*nmv_adv;
1.103 noro 4873: #endif
1.74 noro 4874: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4875: CM(m) = nd_recv_int();
4876: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4877: }
1.74 noro 4878: MKNDV(nd_nvar,m0,len,r);
4879: return r;
4880: }
4881: }
1.65 noro 4882:
1.74 noro 4883: int ox_exec_f4_red(Q proc)
4884: {
4885: Obj obj;
4886: STRING fname;
4887: NODE arg;
4888: int s;
4889: extern int ox_need_conv,ox_file_io;
4890:
4891: MKSTR(fname,"nd_exec_f4_red");
4892: arg = mknode(2,proc,fname);
4893: Pox_cmo_rpc(arg,&obj);
4894: s = get_ox_server_id(QTOS(proc));
4895: nd_write = iofp[s].out;
4896: nd_read = iofp[s].in;
4897: ox_need_conv = ox_file_io = 0;
4898: return s;
4899: }
4900:
4901: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
4902: {
4903: int nsp,nred;
4904: int i,rank,s;
4905: NODE rp,r0,r;
4906: ND_pairs sp;
4907: NM_ind_pair pair;
4908: NMV nmv;
4909: NM nm;
4910: NDV nf;
4911: Obj proc,dmy;
4912:
4913: ox_launch_main(0,0,&proc);
4914: s = ox_exec_f4_red((Q)proc);
4915:
4916: nd_send_int(m);
4917: nd_send_int(nd_nvar);
4918: nd_send_int(nd_bpe);
4919: nd_send_int(nd_wpd);
4920: nd_send_int(nmv_adv);
4921:
1.89 noro 4922: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
1.74 noro 4923:
4924: nd_send_int(nd_psn);
4925: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
4926:
4927: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4928: nd_send_int(nsp);
4929: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
4930: nd_send_int(sp->i1); nd_send_int(sp->i2);
4931: }
4932:
4933: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
4934:
4935: nred = length(rp0); nd_send_int(nred);
4936: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
4937: pair = (NM_ind_pair)BDY(rp);
4938: nd_send_int(pair->index);
4939: nd_send_intarray(pair->mul->dl,nd_wpd);
4940: }
4941: fflush(nd_write);
4942: rank = nd_recv_int();
4943: fprintf(asir_out,"rank=%d\n",rank);
4944: r0 = 0;
4945: for ( i = 0; i < rank; i++ ) {
4946: nf = nd_recv_ndv();
4947: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
4948: }
4949: Pox_shutdown(mknode(1,proc),&dmy);
4950: return r0;
4951: }
4952:
4953: /* server side */
4954:
4955: void nd_exec_f4_red_dist()
4956: {
4957: int m,i,nsp,col,s0size,nred,spcol,j,k;
4958: NM_ind_pair *rp0;
4959: NDV nf;
4960: UINT *s0vect;
4961: IndArray *imat;
4962: int *rhead;
4963: int **spmat;
4964: UINT *svect,*v;
4965: ND_pairs *sp0;
4966: int *colstat;
4967: int a,sprow,rank;
1.89 noro 4968: struct order_spec *ord;
1.74 noro 4969: Obj ordspec;
4970: ND spol;
1.76 noro 4971: int maxrs;
4972: int *spsugar;
1.74 noro 4973:
4974: nd_read = iofp[0].in;
4975: nd_write = iofp[0].out;
4976: m = nd_recv_int();
4977: nd_nvar = nd_recv_int();
4978: nd_bpe = nd_recv_int();
4979: nd_wpd = nd_recv_int();
4980: nmv_adv = nd_recv_int();
4981:
4982: loadobj(nd_read,&ordspec);
1.89 noro 4983: create_order_spec(0,ordspec,&ord);
4984: nd_init_ord(ord);
1.74 noro 4985: nd_setup_parameters(nd_nvar,0);
4986:
4987: nd_psn = nd_recv_int();
4988: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
4989: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
4990: for ( i = 0; i < nd_psn; i++ ) {
4991: nd_ps[i] = nd_recv_ndv();
4992: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
4993: }
4994:
4995: nsp = nd_recv_int();
4996: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
4997: for ( i = 0; i < nsp; i++ ) {
4998: NEWND_pairs(sp0[i]);
4999: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
5000: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
5001: }
5002:
5003: col = nd_recv_int();
5004: s0size = col*nd_wpd;
5005: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
5006: nd_recv_intarray(s0vect,s0size);
5007:
5008: nred = nd_recv_int();
5009: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
5010: for ( i = 0; i < nred; i++ ) {
5011: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
5012: rp0[i]->index = nd_recv_int();
5013: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
5014: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
5015: }
5016:
5017: spcol = col-nred;
5018: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
5019: rhead = (int *)MALLOC(col*sizeof(int));
5020: for ( i = 0; i < col; i++ ) rhead[i] = 0;
5021:
5022: /* construction of index arrays */
5023: for ( i = 0; i < nred; i++ ) {
5024: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
5025: rhead[imat[i]->head] = 1;
5026: }
5027:
5028: /* elimination (1st step) */
5029: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
5030: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 5031: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 5032: for ( a = sprow = 0; a < nsp; a++ ) {
5033: nd_sp(m,0,sp0[a],&spol);
1.75 noro 5034: if ( !spol ) continue;
1.74 noro 5035: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 5036: if ( m == -1 )
5037: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
5038: else
5039: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 5040: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
5041: if ( i < col ) {
5042: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
5043: for ( j = k = 0; j < col; j++ )
5044: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 5045: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 5046: sprow++;
5047: }
1.76 noro 5048: nd_free(spol);
1.74 noro 5049: }
5050: /* elimination (2nd step) */
5051: colstat = (int *)ALLOCA(spcol*sizeof(int));
5052: if ( m == -1 )
1.76 noro 5053: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 5054: else
1.76 noro 5055: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 5056: nd_send_int(rank);
5057: for ( i = 0; i < rank; i++ ) {
5058: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
5059: nd_send_ndv(nf);
1.63 noro 5060: }
1.74 noro 5061: fflush(nd_write);
1.107 noro 5062: }
5063:
1.113 noro 5064: int nd_gauss_elim_q(Q **mat0,int *sugar,int row,int col,int *colstat)
1.107 noro 5065: {
1.109 noro 5066: int mod,i,j,t,c,rank,rank0,inv;
5067: int *ci,*ri;
1.113 noro 5068: Q dn;
1.109 noro 5069: MAT m,nm;
1.108 noro 5070: int **wmat;
5071:
5072: /* XXX */
5073: mod = 99999989;
5074: wmat = (int **)ALLOCA(row*sizeof(int *));
5075: for ( i = 0; i < row; i++ ) {
5076: wmat[i] = (int *)ALLOCA(col*sizeof(int));
5077: for ( j = 0; j < col; j++ ) {
1.113 noro 5078: if ( mat0[i][j] ) {
5079: t = rem(NM(mat0[i][j]),mod);
5080: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5081: wmat[i][j] = t;
5082: } else
1.108 noro 5083: wmat[i][j] = 0;
5084: }
5085: }
1.109 noro 5086: rank0 = nd_gauss_elim_mod(wmat,sugar,row,col,mod,colstat);
5087: NEWMAT(m); m->row = row; m->col = col; m->body = (pointer **)mat0;
1.113 noro 5088: rank = generic_gauss_elim(m,&nm,&dn,&ri,&ci);
1.109 noro 5089: if ( rank != rank0 )
5090: error("afo");
5091: for ( i = 0; i < row; i++ )
5092: for ( j = 0; j < col; j++ )
5093: mat0[i][j] = 0;
5094: c = col-rank;
5095: for ( i = 0; i < rank; i++ ) {
1.113 noro 5096: mat0[i][ri[i]] = dn;
1.109 noro 5097: for ( j = 0; j < c; j++ )
1.113 noro 5098: mat0[i][ci[j]] = (Q)BDY(nm)[i][j];
1.109 noro 5099: }
1.113 noro 5100: inv = invm(rem(NM(dn),mod),mod);
5101: if ( SGN(dn) < 0 ) inv = mod-inv;
1.109 noro 5102: for ( i = 0; i < row; i++ )
5103: for ( j = 0; j < col; j++ ) {
1.113 noro 5104: if ( mat0[i][j] ) {
5105: t = rem(NM(mat0[i][j]),mod);
5106: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5107: } else
1.109 noro 5108: t = 0;
5109: c = dmar(t,inv,0,mod);
5110: if ( wmat[i][j] != c )
5111: error("afo");
5112: }
5113: return rank;
1.76 noro 5114: }
5115:
5116: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat)
5117: {
5118: int i,j,k,l,inv,a,rank,s;
5119: unsigned int *t,*pivot,*pk;
5120: unsigned int **mat;
5121:
5122: mat = (unsigned int **)mat0;
5123: for ( rank = 0, j = 0; j < col; j++ ) {
5124: for ( i = rank; i < row; i++ )
5125: mat[i][j] %= md;
5126: for ( i = rank; i < row; i++ )
5127: if ( mat[i][j] )
5128: break;
5129: if ( i == row ) {
5130: colstat[j] = 0;
5131: continue;
5132: } else
5133: colstat[j] = 1;
5134: if ( i != rank ) {
5135: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5136: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
5137: }
5138: pivot = mat[rank];
5139: s = sugar[rank];
5140: inv = invm(pivot[j],md);
5141: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5142: if ( *pk ) {
5143: if ( *pk >= (unsigned int)md )
5144: *pk %= md;
5145: DMAR(*pk,inv,0,md,*pk)
5146: }
5147: for ( i = rank+1; i < row; i++ ) {
5148: t = mat[i];
5149: if ( a = t[j] ) {
5150: sugar[i] = MAX(sugar[i],s);
5151: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5152: }
5153: }
5154: rank++;
5155: }
5156: for ( j = col-1, l = rank-1; j >= 0; j-- )
5157: if ( colstat[j] ) {
5158: pivot = mat[l];
5159: s = sugar[l];
5160: for ( i = 0; i < l; i++ ) {
5161: t = mat[i];
5162: t[j] %= md;
5163: if ( a = t[j] ) {
5164: sugar[i] = MAX(sugar[i],s);
5165: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5166: }
5167: }
5168: l--;
5169: }
5170: for ( j = 0, l = 0; l < rank; j++ )
5171: if ( colstat[j] ) {
5172: t = mat[l];
5173: for ( k = j; k < col; k++ )
5174: if ( t[k] >= (unsigned int)md )
5175: t[k] %= md;
5176: l++;
5177: }
5178: return rank;
5179: }
5180:
5181: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
5182: {
5183: int i,j,k,l,inv,a,rank,s;
5184: unsigned int *t,*pivot,*pk;
5185: unsigned int **mat;
5186:
5187: mat = (unsigned int **)mat0;
5188: for ( rank = 0, j = 0; j < col; j++ ) {
5189: for ( i = rank; i < row; i++ )
5190: if ( mat[i][j] )
5191: break;
5192: if ( i == row ) {
5193: colstat[j] = 0;
5194: continue;
5195: } else
5196: colstat[j] = 1;
5197: if ( i != rank ) {
5198: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5199: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
5200: }
5201: pivot = mat[rank];
5202: s = sugar[rank];
5203: inv = _invsf(pivot[j]);
5204: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5205: if ( *pk )
5206: *pk = _mulsf(*pk,inv);
5207: for ( i = rank+1; i < row; i++ ) {
5208: t = mat[i];
5209: if ( a = t[j] ) {
5210: sugar[i] = MAX(sugar[i],s);
5211: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5212: }
5213: }
5214: rank++;
5215: }
5216: for ( j = col-1, l = rank-1; j >= 0; j-- )
5217: if ( colstat[j] ) {
5218: pivot = mat[l];
5219: s = sugar[l];
5220: for ( i = 0; i < l; i++ ) {
5221: t = mat[i];
5222: if ( a = t[j] ) {
5223: sugar[i] = MAX(sugar[i],s);
5224: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5225: }
5226: }
5227: l--;
5228: }
5229: return rank;
1.77 noro 5230: }
5231:
5232: int ndv_ishomo(NDV p)
5233: {
5234: NMV m;
5235: int len,h;
5236:
5237: if ( !p ) return 1;
5238: len = LEN(p);
5239: m = BDY(p);
5240: h = TD(DL(m));
5241: NMV_ADV(m);
5242: for ( len--; len; len--, NMV_ADV(m) )
5243: if ( TD(DL(m)) != h ) return 0;
5244: return 1;
5245: }
5246:
5247: void ndv_save(NDV p,int index)
5248: {
5249: FILE *s;
5250: char name[BUFSIZ];
5251: short id;
5252: int nv,sugar,len,n,i,td,e,j;
5253: NMV m;
5254: unsigned int *dl;
5255:
5256: sprintf(name,"%s/%d",Demand,index);
5257: s = fopen(name,"w");
5258: savevl(s,0);
5259: if ( !p ) {
5260: saveobj(s,0);
5261: return;
5262: }
5263: id = O_DP;
5264: nv = NV(p);
5265: sugar = SG(p);
5266: len = LEN(p);
5267: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
5268: write_int(s,&len);
5269:
5270: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5271: saveobj(s,(Obj)CQ(m));
1.77 noro 5272: dl = DL(m);
5273: td = TD(dl);
5274: write_int(s,&td);
5275: for ( j = 0; j < nv; j++ ) {
5276: e = GET_EXP(dl,j);
5277: write_int(s,&e);
5278: }
5279: }
5280: fclose(s);
5281: }
5282:
5283: NDV ndv_load(int index)
5284: {
5285: FILE *s;
5286: char name[BUFSIZ];
5287: short id;
5288: int nv,sugar,len,n,i,td,e,j;
5289: NDV d;
5290: NMV m0,m;
5291: unsigned int *dl;
5292: Obj obj;
5293:
5294: sprintf(name,"%s/%d",Demand,index);
5295: s = fopen(name,"r");
5296: if ( !s ) return 0;
5297:
5298: skipvl(s);
5299: read_short(s,&id);
5300: if ( !id ) return 0;
5301: read_int(s,&nv);
5302: read_int(s,&sugar);
5303: read_int(s,&len);
5304:
5305: m0 = m = MALLOC(len*nmv_adv);
5306: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5307: loadobj(s,&obj); CQ(m) = (Q)obj;
1.77 noro 5308: dl = DL(m);
5309: ndl_zero(dl);
5310: read_int(s,&td); TD(dl) = td;
5311: for ( j = 0; j < nv; j++ ) {
5312: read_int(s,&e);
5313: PUT_EXP(dl,j,e);
5314: }
5315: if ( nd_blockmask ) ndl_weight_mask(dl);
5316: }
5317: fclose(s);
5318: MKNDV(nv,m0,len,d);
5319: SG(d) = sugar;
5320: return d;
1.99 noro 5321: }
5322:
1.102 noro 5323: void nd_det(int mod,MAT f,P *rp)
1.99 noro 5324: {
5325: VL fv,tv;
1.102 noro 5326: int n,i,j,max,e,nvar,sgn,k0,l0,len0,len,k,l,a;
1.99 noro 5327: pointer **m;
1.113 noro 5328: Q mone;
1.125 noro 5329: P **w;
1.126 noro 5330: P mp,r;
1.102 noro 5331: NDV **dm;
5332: NDV *t,*mi,*mj;
5333: NDV d,s,mij,mjj;
5334: ND u;
5335: NMV nmv;
1.114 noro 5336: UINT *bound;
1.99 noro 5337: PGeoBucket bucket;
5338: struct order_spec *ord;
1.126 noro 5339: Q dq,dt,ds;
5340: N gn,qn,dn0,nm,dn;
1.99 noro 5341:
5342: create_order_spec(0,0,&ord);
5343: nd_init_ord(ord);
5344: get_vars((Obj)f,&fv);
5345: if ( f->row != f->col )
5346: error("nd_det : non-square matrix");
5347: n = f->row;
1.125 noro 5348: m = f->body;
1.99 noro 5349: for ( nvar = 0, tv = fv; tv; tv = NEXT(tv), nvar++ );
1.125 noro 5350:
5351: if ( !nvar ) {
5352: if ( !mod )
5353: detp(CO,(P **)m,n,rp);
5354: else {
5355: w = (P **)almat_pointer(n,n);
5356: for ( i = 0; i < n; i++ )
5357: for ( j = 0; j < n; j++ )
5358: ptomp(mod,(P)m[i][j],&w[i][j]);
5359: detmp(CO,mod,w,n,&mp);
5360: mptop(mp,rp);
5361: }
5362: return;
5363: }
1.126 noro 5364:
5365: if ( !mod ) {
5366: w = (P **)almat_pointer(n,n);
5367: dq = ONE;
5368: for ( i = 0; i < n; i++ ) {
5369: dn0 = ONEN;
5370: for ( j = 0; j < n; j++ ) {
5371: if ( !m[i][j] ) continue;
5372: lgp(m[i][j],&nm,&dn);
5373: gcdn(dn0,dn,&gn); divsn(dn0,gn,&qn); muln(qn,dn,&dn0);
5374: }
5375: if ( !UNIN(dn0) ) {
5376: NTOQ(dn0,1,ds);
5377: for ( j = 0; j < n; j++ )
5378: mulp(CO,(P)m[i][j],(P)ds,&w[i][j]);
5379: mulq(dq,ds,&dt); dq = dt;
5380: } else
5381: for ( j = 0; j < n; j++ )
5382: w[i][j] = (P)m[i][j];
5383: }
5384: m = (pointer **)w;
5385: }
1.125 noro 5386:
1.99 noro 5387: for ( i = 0, max = 0; i < n; i++ )
5388: for ( j = 0; j < n; j++ )
5389: for ( tv = fv; tv; tv = NEXT(tv) ) {
5390: e = getdeg(tv->v,(P)m[i][j]);
5391: max = MAX(e,max);
5392: }
1.114 noro 5393: nd_setup_parameters(nvar,max);
1.102 noro 5394: dm = (NDV **)almat_pointer(n,n);
1.99 noro 5395: for ( i = 0, max = 0; i < n; i++ )
1.102 noro 5396: for ( j = 0; j < n; j++ ) {
5397: dm[i][j] = ptondv(CO,fv,m[i][j]);
5398: if ( mod ) ndv_mod(mod,dm[i][j]);
5399: if ( dm[i][j] && !LEN(dm[i][j]) ) dm[i][j] = 0;
5400: }
5401: d = ptondv(CO,fv,(P)ONE);
5402: if ( mod ) ndv_mod(mod,d);
1.113 noro 5403: chsgnq(ONE,&mone);
1.99 noro 5404: for ( j = 0, sgn = 1; j < n; j++ ) {
1.125 noro 5405: if ( DP_Print ) fprintf(stderr,".",j);
1.99 noro 5406: for ( i = j; i < n && !dm[i][j]; i++ );
5407: if ( i == n ) {
5408: *rp = 0;
5409: return;
5410: }
1.102 noro 5411: k0 = i; l0 = j; len0 = LEN(dm[k0][l0]);
1.99 noro 5412: for ( k = j; k < n; k++ )
5413: for ( l = j; l < n; l++ )
1.102 noro 5414: if ( dm[k][l] && LEN(dm[k][l]) < len0 ) {
5415: k0 = k; l0 = l; len0 = LEN(dm[k][l]);
1.99 noro 5416: }
5417: if ( k0 != j ) {
5418: t = dm[j]; dm[j] = dm[k0]; dm[k0] = t;
5419: sgn = -sgn;
5420: }
5421: if ( l0 != j ) {
5422: for ( k = j; k < n; k++ ) {
5423: s = dm[k][j]; dm[k][j] = dm[k][l0]; dm[k][l0] = s;
5424: }
5425: sgn = -sgn;
5426: }
1.114 noro 5427: bound = nd_det_compute_bound(dm,n,j);
5428: if ( ndl_check_bound(bound,bound) )
5429: nd_det_reconstruct(dm,n,j,d);
5430:
1.99 noro 5431: for ( i = j+1, mj = dm[j], mjj = mj[j]; i < n; i++ ) {
1.114 noro 5432: /* if ( DP_Print ) fprintf(stderr," i=%d\n ",i); */
1.99 noro 5433: mi = dm[i]; mij = mi[j];
1.102 noro 5434: if ( mod )
5435: ndv_mul_c(mod,mij,mod-1);
5436: else
1.113 noro 5437: ndv_mul_c_q(mij,mone);
1.99 noro 5438: for ( k = j+1; k < n; k++ ) {
1.114 noro 5439: /* if ( DP_Print ) fprintf(stderr,"k=%d ",k); */
1.99 noro 5440: bucket = create_pbucket();
1.104 noro 5441: if ( mi[k] ) {
1.102 noro 5442: nmv = BDY(mjj); len = LEN(mjj);
5443: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5444: u = ndv_mul_nmv_trunc(mod,nmv,mi[k],DL(BDY(d)));
5445: add_pbucket(mod,bucket,u);
1.99 noro 5446: }
1.104 noro 5447: }
1.99 noro 5448: if ( mj[k] && mij ) {
1.102 noro 5449: nmv = BDY(mij); len = LEN(mij);
5450: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5451: u = ndv_mul_nmv_trunc(mod,nmv,mj[k],DL(BDY(d)));
5452: add_pbucket(mod,bucket,u);
1.99 noro 5453: }
5454: }
1.104 noro 5455: u = nd_quo(mod,bucket,d);
1.102 noro 5456: mi[k] = ndtondv(mod,u);
1.99 noro 5457: }
1.114 noro 5458: /* if ( DP_Print ) fprintf(stderr,"\n",k); */
1.99 noro 5459: }
5460: d = mjj;
5461: }
1.125 noro 5462: if ( DP_Print ) fprintf(stderr,"\n",k);
1.100 noro 5463: if ( sgn < 0 )
1.102 noro 5464: if ( mod )
5465: ndv_mul_c(mod,d,mod-1);
5466: else
1.113 noro 5467: ndv_mul_c_q(d,mone);
1.126 noro 5468: r = ndvtop(mod,CO,fv,d);
5469: if ( !mod && !UNIQ(dq) )
5470: divsp(CO,r,(P)dq,rp);
5471: else
5472: *rp = r;
1.99 noro 5473: }
5474:
1.102 noro 5475: ND ndv_mul_nmv_trunc(int mod,NMV m0,NDV p,UINT *d)
1.99 noro 5476: {
5477: NM mr,mr0;
1.102 noro 5478: NM tnm;
5479: NMV m;
1.99 noro 5480: UINT *d0,*dt,*dm;
5481: int c,n,td,i,c1,c2,len;
1.113 noro 5482: Q q;
1.99 noro 5483: ND r;
5484:
5485: if ( !p ) return 0;
5486: else {
1.102 noro 5487: n = NV(p); m = BDY(p); len = LEN(p);
1.99 noro 5488: d0 = DL(m0);
1.102 noro 5489: td = TD(d);
1.99 noro 5490: mr0 = 0;
5491: NEWNM(tnm);
1.102 noro 5492: if ( mod ) {
5493: c = CM(m0);
5494: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5495: ndl_add(DL(m),d0,DL(tnm));
5496: if ( ndl_reducible(DL(tnm),d) ) {
5497: NEXTNM(mr0,mr);
5498: c1 = CM(m); DMAR(c1,c,0,mod,c2); CM(mr) = c2;
1.104 noro 5499: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5500: }
5501: }
5502: } else {
1.113 noro 5503: q = CQ(m0);
1.102 noro 5504: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5505: ndl_add(DL(m),d0,DL(tnm));
5506: if ( ndl_reducible(DL(tnm),d) ) {
5507: NEXTNM(mr0,mr);
1.113 noro 5508: mulq(CQ(m),q,&CQ(mr));
1.104 noro 5509: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5510: }
1.99 noro 5511: }
5512: }
5513: if ( !mr0 )
5514: return 0;
5515: else {
5516: NEXT(mr) = 0;
1.105 noro 5517: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
1.99 noro 5518: MKND(NV(p),mr0,len,r);
5519: SG(r) = SG(p) + TD(d0);
5520: return r;
5521: }
5522: }
1.114 noro 5523: }
5524:
5525: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d)
5526: {
5527: int i,obpe,oadv,h,k,l;
5528: static NM prev_nm_free_list;
5529: EPOS oepos;
5530:
5531: obpe = nd_bpe;
5532: oadv = nmv_adv;
5533: oepos = nd_epos;
5534: if ( obpe < 2 ) nd_bpe = 2;
5535: else if ( obpe < 3 ) nd_bpe = 3;
5536: else if ( obpe < 4 ) nd_bpe = 4;
5537: else if ( obpe < 5 ) nd_bpe = 5;
5538: else if ( obpe < 6 ) nd_bpe = 6;
5539: else if ( obpe < 8 ) nd_bpe = 8;
5540: else if ( obpe < 10 ) nd_bpe = 10;
5541: else if ( obpe < 16 ) nd_bpe = 16;
5542: else if ( obpe < 32 ) nd_bpe = 32;
5543: else error("nd_det_reconstruct : exponent too large");
5544:
5545: nd_setup_parameters(nd_nvar,0);
5546: prev_nm_free_list = _nm_free_list;
5547: _nm_free_list = 0;
5548: for ( k = j; k < n; k++ )
5549: for (l = j; l < n; l++ )
5550: ndv_realloc(dm[k][l],obpe,oadv,oepos);
5551: ndv_realloc(d,obpe,oadv,oepos);
5552: prev_nm_free_list = 0;
5553: #if 0
5554: GC_gcollect();
5555: #endif
5556: }
5557:
5558: UINT *nd_det_compute_bound(NDV **dm,int n,int j)
5559: {
5560: UINT *d0,*d1,*d,*t,*r;
5561: int k,l;
5562:
5563: d0 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5564: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5565: for ( k = 0; k < nd_wpd; k++ ) d0[k] = 0;
5566: for ( k = j; k < n; k++ )
5567: for ( l = j; l < n; l++ )
5568: if ( dm[k][l] ) {
5569: d = ndv_compute_bound(dm[k][l]);
5570: ndl_lcm(d,d0,d1);
5571: t = d1; d1 = d0; d0 = t;
5572: }
5573: r = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5574: for ( k = 0; k < nd_wpd; k++ ) r[k] = d0[k];
5575: return r;
1.117 noro 5576: }
5577:
5578: DL nd_separate_d(UINT *d,UINT *trans)
5579: {
1.121 noro 5580: int n,td,i,e,j;
1.117 noro 5581: DL a;
5582:
5583: ndl_zero(trans);
5584: td = 0;
1.121 noro 5585: for ( i = 0; i < nd_ntrans; i++ ) {
5586: e = GET_EXP(d,i);
5587: PUT_EXP(trans,i,e);
5588: td += MUL_WEIGHT(e,i);
5589: }
5590: if ( nd_ntrans+nd_nalg < nd_nvar ) {
5591: /* homogenized */
5592: i = nd_nvar-1;
1.117 noro 5593: e = GET_EXP(d,i);
5594: PUT_EXP(trans,i,e);
5595: td += MUL_WEIGHT(e,i);
5596: }
5597: TD(trans) = td;
5598: if ( nd_blockmask) ndl_weight_mask(trans);
5599: NEWDL(a,nd_nalg);
5600: td = 0;
1.121 noro 5601: for ( i = 0; i < nd_nalg; i++ ) {
5602: j = nd_ntrans+i;
5603: e = GET_EXP(d,j);
5604: a->d[i] = e;
1.117 noro 5605: td += e;
5606: }
5607: a->td = td;
5608: return a;
5609: }
5610:
1.118 noro 5611: int nd_monic(int mod,ND *p)
1.117 noro 5612: {
5613: UINT *trans,*t;
5614: DL alg;
5615: MP mp0,mp;
1.118 noro 5616: NM m,m0,m1,ma0,ma,mb,mr0,mr;
5617: ND r;
1.117 noro 5618: DL dl;
5619: DP nm;
5620: NDV ndv;
1.118 noro 5621: DAlg inv,cd;
1.117 noro 5622: ND s,c;
1.118 noro 5623: Q l,mul;
5624: N ln;
5625: int n,ntrans,i,e,td,is_lc,len;
5626: NumberField nf;
5627: struct oEGT eg0,eg1;
5628:
5629: if ( !(nf = get_numberfield()) )
5630: error("nd_monic : current_numberfield is not set");
1.117 noro 5631:
1.118 noro 5632: /* Q coef -> DAlg coef */
5633: NEWNM(ma0); ma = ma0;
5634: m = BDY(*p);
5635: is_lc = 1;
5636: while ( 1 ) {
5637: NEWMP(mp0); mp = mp0;
5638: mp->c = (P)CQ(m);
5639: mp->dl = nd_separate_d(DL(m),DL(ma));
5640: NEWNM(mb);
5641: for ( m = NEXT(m); m; m = NEXT(m) ) {
5642: alg = nd_separate_d(DL(m),DL(mb));
5643: if ( !ndl_equal(DL(ma),DL(mb)) )
5644: break;
5645: NEXTMP(mp0,mp); mp->c = (P)CQ(m); mp->dl = alg;
5646: }
5647: NEXT(mp) = 0;
5648: MKDP(nd_nalg,mp0,nm);
5649: MKDAlg(nm,ONE,cd);
5650: if ( is_lc == 1 ) {
5651: /* if the lc is a rational number, we have nothing to do */
5652: if ( !mp0->dl->td )
5653: return 1;
5654:
5655: get_eg(&eg0);
5656: invdalg(cd,&inv);
5657: get_eg(&eg1); add_eg(&eg_invdalg,&eg0,&eg1);
5658: /* check the validity of inv */
5659: if ( mod && !rem(NM(inv->dn),mod) )
5660: return 0;
5661: CA(ma) = nf->one;
5662: is_lc = 0;
5663: ln = ONEN;
5664: } else {
5665: muldalg(cd,inv,&CA(ma));
5666: lcmn(ln,NM(CA(ma)->dn),&ln);
5667: }
5668: if ( m ) {
5669: NEXT(ma) = mb; ma = mb;
5670: } else {
5671: NEXT(ma) = 0;
5672: break;
5673: }
5674: }
5675: /* l = lcm(denoms) */
5676: NTOQ(ln,1,l);
5677: for ( mr0 = 0, m = ma0; m; m = NEXT(m) ) {
5678: divq(l,CA(m)->dn,&mul);
5679: for ( mp = BDY(CA(m)->nm); mp; mp = NEXT(mp) ) {
5680: NEXTNM(mr0,mr);
5681: mulq((Q)mp->c,mul,&CQ(mr));
5682: dl = mp->dl;
5683: td = TD(DL(m));
5684: ndl_copy(DL(m),DL(mr));
1.121 noro 5685: for ( i = 0; i < nd_nalg; i++ ) {
5686: e = dl->d[i];
5687: PUT_EXP(DL(mr),i+nd_ntrans,e);
5688: td += MUL_WEIGHT(e,i+nd_ntrans);
1.118 noro 5689: }
5690: TD(DL(mr)) = td;
5691: if ( nd_blockmask) ndl_weight_mask(DL(mr));
5692: }
1.117 noro 5693: }
1.118 noro 5694: NEXT(mr) = 0;
5695: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
5696: MKND(NV(*p),mr0,len,r);
5697: /* XXX */
5698: SG(r) = SG(*p);
5699: nd_free(*p);
5700: *p = r;
5701: return 1;
1.59 noro 5702: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>